Linear Technology Magazine Circuit Collection, Volume III

Application Note 67
September 1996
Linear Technology Magazine Circuit Collection, Volume III
Data Conversion, Interface and Signal Processing
Richard Markell, Editor
INTRODUCTION
Application Note 67 is a collection of circuits from the first
five years of Linear Technology, targeting data conversion,
interface and signal processing applications. This
Application Note includes circuits such as fast video
multiplexers for high speed video, an ultraselective
bandpass filter circuit with adjustable gain and a fully
differential, 8-channel, 12-bit A/D system. The categories
included herein are data conversion, interface, filters,
instrumentation, video/op amps and miscellaneous
circuits. Application Note 66, which covers power products
and circuits from Linear Technology ’s first five years, is
also available from LTC.
ARTICLE INDEX
Data Conversion ........................................................................................................... 3
Fully Differential, 8-Channel, 12-Bit A/D System Using the LTC®1390 and LTC1410 ......................................... 3
12-Bit DAC Applications ..................................................................................................................................... 5
LTC1329 Micropower, 8-Bit, Current Output DAC Used for Power Supply Adjustment,
Trimmer Pot Replacement ................................................................................................................................. 7
12-Bit Cold Junction Compensated, Temperature Control System with Shutdown ............................................ 8
A 12-Bit Micropower Battery Current Monitor .................................................................................................... 9
Interface ................................................................................................................... 10
V.35 Transceivers Allow 3-Chip V.35 Port Solution ......................................................................................... 10
Switching, Active GTL Terminator .................................................................................................................... 12
RS232 Transceivers for DTE/DCE Switching .................................................................................................... 14
Active Negation Bus Terminators ..................................................................................................................... 16
RS485 Repeater Extends System Capability..................................................................................................... 18
An LT ®1087-Based 1.2V GTL Terminator ......................................................................................................... 19
LTC1145/LTC1146 Achieve Low Profile Isolation with Capacitive Lead Frame................................................. 19
LTC485 Line Termination ................................................................................................................................. 21
Filters ...................................................................................................................... 22
Sallen and Key Filters Use 5% Values .............................................................................................................. 22
Low Power Signal Detection in a Noisy Environment ....................................................................................... 25
Bandpass Filter Has Adjustable Q ..................................................................................................................... 28
An Ultraselective Bandpass Filter with Adjustable Gain .................................................................................... 30
LT1367 Builds Rail-to-Rail Butterworth Filter ................................................................................................... 32
DC Accurate, Clock Tunable Lowpass Filter with Input Antialiasing Filter......................................................... 33
The LTC1066-1 DC Accurate Elliptic Lowpass Filter ......................................................................................... 35
Clock Tunable Bandpass Filter Operates to 160kHz in Single Supply Systems................................................. 37
A Linear Phase Bandpass Filter for Digital Communications ............................................................................ 39
, LTC and LT are registered trademarks of Linear Technology Corporation.
AN67-1
Application Note 67
Instrumentation ........................................................................................................... 41
Wideband RMS Noise Meter ............................................................................................................................ 41
LTC1392 Micropower Temperature and Voltage Measurement Sensor ........................................................... 45
Humidity Sensor to Data Acquisition System Interface .................................................................................... 46
A Single Cell Barometer.................................................................................................................................... 48
Noise Generators for Multiple Uses—A Broadband Random Noise Generator................................................. 49
Noise Generators for Multiple Uses—A Diode Noise Generator for “Eye Diagram” Testing ............................. 52
Video/Op Amps............................................................................................................ 53
LT1251 Circuit Smoothly Fades Video to Black ................................................................................................ 53
Luma Keying with the LT1203 Video Multiplexer ............................................................................................. 54
LT1251/LT1256 Video Fader and DC Gain Controlled Amplifier ....................................................................... 56
Extending Op Amp Supplies to Get More Output Voltage ................................................................................. 58
Using Super Op Amps to Push Technological Frontiers: An Ultrapure Oscillator ............................................. 62
Fast Video MUX Uses LT1203/LT1205 ............................................................................................................. 66
Using a Fast Analog Multiplexer to Switch Video Signals for NTSC “Picture-in-Picture” Displays ................... 67
Applications for the LT1113 Dual JFET Op Amp ............................................................................................... 69
LT1206 and LT1115 Make Low Noise Audio Line Driver .................................................................................. 70
Driving Multiple Video Cables with the LT1206 ................................................................................................ 71
Optimizing a Video Gain Control Stage Using the LT1228 ................................................................................ 72
Differential Gain and Phase .............................................................................................................................. 74
LT1190 Family Ultrahigh Speed Op Amp Circuits............................................................................................. 75
An LT1112 Dual Output Buffered Reference ..................................................................................................... 78
Three Op Amp Instrumentation Amp Using the LT1112/LT1114...................................................................... 79
Ultralow Noise Three Op Amp Instrumentation Amplifier ................................................................................. 80
A Temperature Compensated, Voltage Controlled Gain Amplifier Using the LT1228 ........................................ 80
The LTC1100, LT1101 and LT1102: A Trio of Effective Instrumentation Amplifiers ......................................... 83
Miscellaneous Circuits .................................................................................................. 86
Driving a High Level Diode Ring Mixer with an Operational Amplifier .............................................................. 86
AN67-2
Application Note 67
Data Conversion
FULLY DIFFERENTIAL, 8-CHANNEL, 12-BIT A/D
SYSTEM USING THE LTC1390 AND LTC1410
by Kevin R. Hoskins
The LTC1410’s fast 1.25Msps conversion rate and differential ±2.5V input range make it ideal for applications that
require multichannel acquisition of fast, wide bandwidth
signals. These applications include multitransducer vibration analysis, race vehicle telemetry data acquisition and
multichannel telecommunications. The LTC1410 can be
+5V
combined with the LTC1390 8-channel serial interfaced
analog multiplexer to create a differential ADC system with
conversion throughput rates up to 625ksps. This rate
applies to situations where the selected channel changes
with each conversion. The conversion rate increases to
1.25Msps if the same channel is used for consecutive
conversions.
Figure 1 shows the complete differential, 8-channel A/D
circuit. Two LTC1390s, U1 and U2, are used as noninvert-
–5V
0.1µF
S0
+CH1
S1
V+
U1
LTC1390
D
+CH2
S2
V–
+CH3
S3
DATA2
+CH4
S4
DATA1
+CH5
S5
CS
0.1µF
+
10µF
0.1µF
+AIN
AVDD
–AIN
DVDD
VREF
VSS
REFCOMP
+CH6
S6
CLK
+CH7
S7
GND
D10
SERIAL DATA
0.1µF
0.1µF
+
BUSY
AGND
D11
10µF
+
+CH0
10µF
CS
CONVST
U3
LTC1410
RD
D9
SHDN
D8
NAP/SLP
D7
OGND
D6
D0
D5
D1
D4
D2
DGND
D3
µP CONTROL
LINES
CHIP SELECT MUX
SERIAL CLK
0.1µF
V+
–CH0
S0
–CH1
S1
–CH2
S2
V–
–CH3
S3
DATA2
–CH4
S4
DATA1
–CH5
S5
CS
–CH6
S6
CLK
–CH7
S7
GND
U2
LTC1390
D
0.1µF
12-BIT
PARALLEL
BUS
AN67 F01
Figure 1. Fully Differential 8-Channel Data Acquisition System Achieves 625ksps Throughput
AN67-3
Application Note 67
ing and inverting input multiplexers. The outputs of the
noninverting and inverting multiplexers are applied to the
LTC1410’s +AIN and – AIN inputs, respectively. The
LTC1390 share the Chip Select MUX, Serial Data and Serial
Clock control signals. This arrangement simultaneously
selects the same channel on each multiplexer: S0 for both
+CH0 and – CH0, S1 for both +CH1 and – CH1, and so on.
As shown in the timing diagram (Figure 2), MUX channel
selection and A/D conversion are pipelined to maximize
the converter’s throughput. The conversion process begins with selecting the desired multiplexer channel pair.
With a logic high applied to the LTC1390’s CS input, the
channel pair data is clocked into each Data 1 input on the
rising edge of the 5MHz clock signal. Chip Select MUX is
then pulled low, latching the channel pair selection data.
The signals on the selected MUX inputs are then applied to
the LTC1410’s differential inputs. Chip Select MUX is
pulled low 700ns before the LTC1410’s conversion start
input, CONVST, is pulled low. This corresponds to the
maximum time needed by the LTC1390’s MUX switches to
fully turn on. This ensures that the input signals are fully
settled before the LTC1410’s S/H captures its sample.
The LTC1410’s S/H acquires the input signal and begins
conversion on CONVST’s falling edge. During the conversion, the LTC1390’s CS input is pulled high and the data for
the next channel pair is clocked into Data 1. This pipelined
operation continues until a conversion sequence is completed. When a new channel pair is selected for each
conversion, the sampling rate of each channel is 78ksps,
allowing an input signal bandwidth of 39kHz for each
channel of the LTC1390/LTC1410 system.
To maximize the throughput rate, the LTC1410’s CS input
is pulled low at the beginning of a series of conversions.
The LTC1410’s data output drivers are controlled by the
signal applied to RD. The conversion’s results are available 20ns before the rising edge of Busy. The rising edge
of the Busy output signal can be used to notify a processor
that a conversion has ended and data is ready to read.
This circuit takes advantage of the LTC1410’s very high
1.25Msps conversion rate and differential inputs and the
LTC1390’s ease of programming to create an A/D system
that maintains wide input signal bandwidth while sampling multiple input signals.
SERIAL CLOCK LTC1390
5MHz, 200ns
CHIP SELECT MUX LTC1390
DATA LTC1390
EN
B2
B1
B0
EN
B2
B1
B0
EN
B2
B1
B0
EN
B2
B1
B0
CONVST LTC1410
BUSY LTC1410
700ns
CONVERSION 0
CONVERSION 1
CONVERSION 2
RD LTC1410
OUTPUT DATA LTC1410
DATA 0
DATA 1
DATA 2
AN67 F02
A LOGIC LOW IS APPLIED TO THE LTC1410S CS INPUT DURING A CONVERSION SEQUENCE.
Figure 2. The Figure 1 Circuit Timing Diagram
AN67-4
Application Note 67
12-BIT DAC APPLICATIONS
by Kevin R. Hoskins
System Autoranging
System autoranging, adjusting an ADC’s full-scale range,
is an application area for which the LTC1257 is appropriate. Autoranging is especially useful when using an ADC
with multiplexed inputs. Without autoranging only two
reference values are used: one to set the full-scale magnitude and another to set the zero scale magnitude. Since it
is common to have input signals with different zero scale
and full-scale magnitude requirements, fixed reference
voltages present a problem. Although the ranges selected
for some of the inputs may take advantage of the full range
of ADC output codes, inputs that do not span the same
range will not generate all codes, reducing the ADC’s
effective resolution. One possible solution is to match the
reference voltage span to each multiplexer input.
The circuit shown in Figure 3 uses two LTC1257s to set
the full-scale and zero operating points of the LTC1296
+22µF
VCC
•
•
•
8 ANALOG
INPUT CHANNELS
U1
LTC1296
CH7
COM
REF + REF –
CS
DOUT
CLK
µP
DIN
SSO
50k
50k
5V
74HC04
0.1µF
VCC
100Ω
VOUT
0.1µF
VCC
100Ω
VOUT
0.1µF
DIN
CLK
U2
LTC1257 LOAD
GND VREF
DOUT
DIN
CLK
U3
LTC1257 LOAD
GND VREF
DOUT
During the conversion process, U2 and U3 receive the full
and zero scale codes, respectively, that correspond to a
selected multiplexer channel. For example, let channel 2’s
span begin at 2V and end at 4.5V. When a host processor
wants a conversion of channel 2’s input signal, it first
sends code that sets the output of U2 to 2V and U3 to 4.5V,
fixing the span to 2.5V. The processor then sends data to
the LTC1296 selecting channel 2. The processor next
clocks the LTC1296 and reads the data generated during
the conversion of the 3.5VP-P signal applied to channel 2.
As other multiplexer channels are selected the DAC outputs are changed to match their spans.
Computer-Controlled 4 – 20mA Current Loop
A common and useful circuit is the 4 – 20mA current loop.
It is used to transmit information over long distances
using varying current levels. The advantage of using
current over voltage is the absence of IR losses and the
transmission errors and signal losses they can create.
5V
CH0
12-bit, 8-channel ADC. The ADC shares its serial interface with the DACs. To further simplify bus connections,
the DACs’ data is daisy-chained. Two chip selects are
used, one to select the LTC1296 when programming its
multiplexer and the other to select the DACs when setting
their output voltages.
AN67 F03
Figure 3. Using Two LTC1257 12-Bit Voltage Output DACs to Set
the Input Span of the 12-Bit 8-Channel LTC1296
The circuit in Figure 4 is a computer-controlled 4 – 20mA
current loop. It is designed to operate on a single supply
over a range of 3.3V to 30V. The circuit’s zero output
reference signal, 4mA, is set by R1 and calibrated using
R2, and its full-scale output current is set by R3 and
calibrated using R4. The zero and full-scale output currents are set as follows: with a zero input code applied to
the LTC1453, the output current, IOUT, is set to 4mA by
adjusting R2; next, with a full-scale code applied to the
DAC the full-scale output current is set to 20mA by
adjusting R4.
The circuit is self-regulating, forcing the output current to
remain stable for a fixed DAC output voltage. This selfregulation works as follows: starting at t = 0, the LTC1453’s
fixed output (in this example, 2.5V) is applied to the left
side of R3; instantaneously, the voltage applied to the
LT1077’s input is 1.25V; this turns on Q1 and the voltage
across RS starts increasing beginning from 1.25V; as the
AN67-5
Application Note 67
VLOOP
3.3V TO 30V
LT1121-3.3
IN
OUT
R1
90k
R2
5k
R3
45k
R4
5k
1µF
CLK
FROM
OPTOISOLATED
INPUTS
OPTOISOLATORS
DIN
500Ω
VREF
LTC1453
10k
4N28
+
VOUT
CS/LD
3.3V
CLK
DIN
CS/LD
VCC
R5
3k
1k
LT1077
–
Q1
2N3440
RS
10Ω
CLK
DIN
CS/LD
IOUT
AN67 F04
Figure 4. The LTC1453 Forms the Heart of This Isolated 4 – 20mA Current Loop
voltage across RS increases it lifts the LTC1453’s GND pin
above 0V; the voltage across RS continues to increase until
it equals the DAC’s output voltage.
Once the circuit reaches this stable condition, the constant
DAC output voltage sets a constant current through R3 +
R4 and R5. This constant current fixes a constant voltage
across R5 that is also applied to the LT1077’s noninverting
input. Feedback from the top of RS is applied to the
inverting input. As the op amp forces its inputs to the same
voltage, it will fix the voltage at the top of RS. This in turn
fixes the output current to a constant value.
AN67-6
Optoisolated Serial Interface
The serial interface of the LTC1451 family and the LTC1257
make optoisolated interfaces very easy and cost effective.
Only three optoisolators are needed for serial data communications. Since the inputs of the LTC1451, LTC1452
and LTC1453 have generous hysteresis, the switching
speed of the optoisolators is not critical. Further, because
each of these DACs can be daisy-chained to others, only
three optoisolators are required.
Application Note 67
LTC1329 MICROPOWER, 8-BIT, CURRENT OUTPUT
DAC USED FOR POWER SUPPLY ADJUSTMENT,
TRIMMER POT REPLACEMENT
by K.S. Yap
By simply clocking the LTC1329, the DAC current output
is decreased or increased (decreased if DIN = 0, increased
if DIN = 1), causing VOUT to change accordingly.
Power Supply Voltage Adjustment
Figure 6 is a schematic of a digitally controlled offset
voltage adjustment circuit using a 1-wire interface. By
clocking the LTC1329, the DAC current output is increased, causing VR2 to increase accordingly. When the
DAC current output reaches full scale it will roll over to
zero, causing VR2 to change from the maximum offset trim
voltage to the minimum offset trim voltage.
Trimmer Pot Replacement
Figure 5 is a schematic of a digitally controlled power
supply voltage adjustment circuit using a 2-wire interface.
The LT1107 is configured as a step-up DC/DC converter,
with the output voltage (VOUT) determined by the values of
the feedback resistors. The LTC1329’s DAC current output
is connected to the feedback node of these resistors, and
an 8051 microprocessor is used to interface to the LTC1329.
L1*
100µH
VIN
2V TO 3V
1N5817
VOUT
4V TO 6V
20mA
47Ω
ILIM
+
VIN
100µF
39k
LTC1329-50
SW1
1
LT1107
VCC = 3.3V
FB
SW2
GND
2
+
3
10k
47µF
4
IOUT
DOUT
VCC
DIN
SHDN
CLK
GND
CS
8
7
MPU
(e.g. 8051)
6
P1.1
5
P1.0
AN67 F05
*L1 = COILTRONICS CTX100-4
Figure 5. LTC1329 Digitally Controls the Output Voltage of a Power Supply
RF
V–
(0.5) (IFULL SCALE)
TRIM RANGE = ±(0.5) (IFULL SCALE) (R2)
R1 =
RI
–
VIN
VOUT
LT1006
+
1
VR2
VCC = 3.3V
R2
100Ω
R1
600k
2
3
V–
4
IOUT
DOUT
VCC
DIN
LTC1329-50
SHDN
CLK
GND
CS
8
7
6
MPU
(e.g. 8051)
5
P1.0
AN67 F06
Figure 6. LTC1329 Used to Null Op Amp’s Offset Voltage
AN67-7
Application Note 67
12-BIT COLD JUNCTION COMPENSATED,
TEMPERATURE CONTROL SYSTEM WITH SHUTDOWN
by Robert Reay
the chopping noise before the signal is sent to the A/D
converter. The LTC1297 A/D converter uses the reference of the LTC1257 after it has been filtered to set full
scale. After the A/D measurement is taken CS is pulled
high and everything except the LTC1257 is powered
down, reducing the system supply current to about
350µA. A word can then be written to the LTC1257 and its
output can be used as a temperature control signal for the
system being monitored.
The circuit in Figure 7 is a 12-bit, single 5V supply
temperature control system with shutdown. An external
temperature is monitored by a J-type thermocouple. The
LT1025A provides the cold junction compensation for
the thermocouple and the LTC1050 chopper op amp
provides signal gain. The 47kΩ, 1µF RC network filters
5V
100k
10µF
10k
µP
2N3906
VCC
–IN
CS
DOUT
0.1µF
–
+IN
VIN
J
LT1025A
+
GND
VREF
47k
+
1µF
DATA
DAC LOAD
LTC1297
COMM
1µF
CLK
CS/
POWER DOWN
CLK
GND
1µF
100k
LTC1050
–
74k
VREF
VCC
DIN
1k
CONTROL
OUTPUT
VOUT
LTC1257
CLK
LOAD
DOUT
GND
AN67 F07
Figure 7. 12-Bit Single 5V Control System with Shutdown
AN67-8
Application Note 67
The Battery Current Monitor
Introduction
The LTC1297 forms the core of the micropower battery
current monitor shown in Figure 8. This 12-bit data
acquisition system features an automatic power shutdown that is activated after each conversion. In shutdown
the supply current is reduced to 6µA, typically. As shown
in Figure 9, the average power supply current of the
LTC1297 varies from milliamperes to a few microamperes
as the sampling frequency is reduced. This circuit draws
only 190µA from a 6V to 12V battery when the sampling
frequency is less than 10 samples per second. Wake-up
time is limited by that required by the LTC1297 (5.5µs).
For long periods of inactivity, the circuit’s supply current
can be further reduced to 20µA by using the shutdown
feature on the LT1121. More wake-up time is required
when using this mode of shutdown. It is usually determined by the amount of capacitance in the circuit and the
available charging current from the regulator.
IN
GND
10000
1000
100
10
1
0.001
0.01
0.1
1
10
fSAMPLE (kHz)
100
AN67 F09
Figure 9. Power Supply Current vs Sampling Frequency
for the LTC1297
5V
OUT
LT1121
510k
The battery voltage of 6V to 12V is regulated down to 5V
by the LT1121 micropower regulator. A sense resistor of
0.05Ω is placed in series with the battery to convert the
battery current to a voltage. Full scale is designed for 2A,
giving a resolution of 0.5mA with the 12-bit ADC. The
LTC1047 amplifies the voltage across the sense resistor
AVERAGE ICC (µA)
A 12-BIT MICROPOWER
BATTERY CURRENT MONITOR
by Sammy Lum
0.33µF
SHDN
6V
TO 12V
100k
1N4148
0.1µF
750k
LOAD
30.1k
0.05Ω
2A FULL
SCALE
+
1/2
LTC1047
LTC1297
10Ω
1µF
470k
VCC
CS
–
+IN
CLK
–IN
DOUT
GND
VREF
+
TO AND
FROM µP
22µF
TANT
LOW BATTERY
0.1µF
–
LT1004-2.5
1/2
+LTC1047
20M
AN67 F08
Figure 8. A Micropower Battery Current Monitor Using the LTC1297 12-Bit Data Acquisition System
AN67-9
Application Note 67
by 25 V/V. This goes through an RC lowpass filter before
being fed into the input of the LTC1297. The RC filter
serves two functions. First it helps band limit the input
noise to the ADC. Second the capacitor helps the LTC1047
recover from transients due to the switching input capacitor of the LTC1297. The LT1004 provides the full-scale
reference for the ADC. A low-battery detection circuit has
been created by using the other half of the LTC1047 as a
comparator. Its trip point has been set to 5V plus the
dropout voltage of the LT1121. Because data is transmitted serially to and from the microprocessor or microcontroller, this current monitor circuit can be located close to
the battery.
Interface
V.35 TRANSCEIVERS ALLOW 3-CHIP
V.35 PORT SOLUTION
by Y.K. Sim
Two new LTC interface devices, the LTC1345 and the
LTC1346, provide the differential drivers and receivers
needed to implement a V.35 interface. When used in
conjunction with an RS232 transceiver like the LT1134A,
they allow a complete V.35 interface to be implemented
with just two transceiver chips and one resistor termination chip. The LTC1345 and LTC1346 provide the three
differential drivers and receivers necessary to implement
the high speed path and the LT1134A provides the four
RS232 drivers and receivers required for the handshaking
interface. Both the LTC1345 and the LT1134A provide
onboard charge pump power supplies allowing a complete V.35 interface to be powered from a single 5V supply.
For systems where ±5V supplies are present the LTC1346
is offered without charge pumps, representing a 30%
power savings.
The differential transceivers are capable of operating at
data rates above 10MBd in nonreturn-to-zero (NRZ)
format.
AN67-10
The RS232 handshaking lines can be implemented with
standard RS232 transceivers. The LT1134A provides four
RS232 drivers and four receivers, enough to implement
the extended 8-line handshaking protocol specified in
V.35. The LT1134A also includes an onboard charge pump
to generate the higher voltages required by RS232 from a
single 5V supply, making it an ideal companion to the
LTC1345. These two chips, together with the BI Technologies termination resistor network chip provide a complete
surface mountable 5V-only V.35 data port. Systems that
have multiple power supply voltages available and use
only the simpler 5-signal V.35 handshaking protocol can
use the LTC1346 with the LT1135A or LT1039 RS232
transceivers; this combination provides a complete port
while saving board space and complexity. Figure 10 shows
a typical LTC1345/LT1134A V.35 implementation with five
differential signals and five basic handshaking signals
with an option for three additional handshaking signals.
Application Note 67
1µF
50Ω
1µF
+ +
VCC1
5V
4
T
1
2
27
LTC1345
6
+
1µF
DX
7
DX
11
RX
12
RX
13
RX
10
9
1µF
26
1
25
2
24
3
23
4
20
14
19
13
18
12
17
11
16
10
15
9
5
7
4
P
TXD (103)
S
U
T
S
SCTE (113)
W
AA
T
TXC (114)
T
V
T
T
R
T
X
V
RXD (104)
T
AA
Y
RXC (115)
T
U
W
Y
X
P
BI
627T500/
1250
(SOIC)
T
T
R
B
GND (102)
B
A
CABLE SHIELD
A
T
1
3
22
2
12
16
11
15
10
14
9
13
1
24
2
23
3
22
4
21
5
20
6
19
7
3
10
11
RX
4
DX
5
DX
6
DX
7
8
12
VCC2
0.2µF
23
4
LT1134A
0.2µF
3
22
1
0.1µF
0.1µF
19
20
18
OPTIONAL SIGNALS
16
14
17
15
23
24
LT1134A
2
21
0.1µF
RX
8
24
+
LTC1346
0.1µF +
0.2µF
1
0.1µF
T
8
14
VCC1
0.2µF
BI
627T500/
1250
(SOIC)
VCC2
5V
VEE2
50Ω
28
3
+
125Ω
=
0.1µF
2
DX
DX
RX
RX
RX
RX
DX
DX
13
5
H
DTR (108)
7
C
RTS (105)
6
E
DSR (107)
8
D
CTS (106)
10
F
DCD (109)
12
NN
TM (142)
9
N
RDL (140)
11
L
LLB (141)
H
6
C
8
E
5
D
7
F
9
NN
11
N
10
L
12
ISO 2593
ISO 2593
34-PIN DTE/DCE
34-PIN DTE/DCE
INTERFACE CONNECTOR INTERFACE CONNECTOR
20
RX
18
RX
DX
DX
DX
DX
21
19
17
15
16
RX
14
RX
13
AN67 F10
Figure 10. Typical V.35 Implementation Using LTC1345 and LTC1346
AN67-11
Application Note 67
SWITCHING, ACTIVE GTL TERMINATOR
by Dale Eagar
Introduction
New high speed microprocessors, especially those used
in multiprocessor workstations and video graphics terminals, require high speed backplanes that support peak
data rates of up to 1Gbps. The backplane is a passive
component, whereas all drivers and receivers are implemented in low voltage swing CMOS (also referred to as
GTL logic). These applications require bidirectional terminators, terminators that will either source or sink current
(in this case, at 1.55V). The current requirements of the
terminator depended on the number of terminations on
the backplane. Present applications may require up to
10A. This specification may, of course, be reduced if
required.
Circuit Operation
The complete schematic of the terminator is shown in
Figure 11. The circuit is based on the LT1158 half-bridge,
N-channel, power MOSFET driver. The LT1158 is configured to provide bidirectional synchronous switching to
MOSFETs Q1 through Q6. VR1, an LT1004-1.2, R1 and C1
generate a 1.25V reference voltage that programs the
terminator’s output voltage. U1A, an LT1215, is a moderate speed (23MHz GBW) precision operational amplifier
that subtracts the error voltage at its inverting input from
the 1.25V reference. U1A is also used to amplify this error
signal. Components R3 and C2 tailor the phase and gain of
this section and are selected when evaluating the system’s
load step response.
U1B and part of U2 provide the gain and the phase
inversion necessary to form an oscillator. C3 and C4
provide positive feedback at high frequencies, which is
necessary for the system to oscillate in a controlled
AN67-12
manner while keeping the voltage excursions within the
common mode range of U1B. R8, U2 and C6 provide
phase inversion and negative feedback at the middle
frequencies, causing U1B to oscillate at a frequency much
higher than the feedback loop’s response. The DC path for
the oscillator is closed through the power MOSFETs Q1 to
Q6, the output choke L1, the output capacitor C11 and
through the feedback path with the error amplifier. R4 and
R7 set the center of the common mode voltage of U1B and
are selected to limit the maximum duty factor the oscillator
can achieve.
R9, R10, R12 and C9 provide output current sense to U2,
allowing it to shut down the oscillator via the Fault pin (Pin
5) to prevent catastrophic or even cataclysmic events from
occurring. D2, C8 and the circuitry behind the Boost pin
(Pin 16) of U2 work together to provide more than sufficient gate drive for the N-channel FETs Q1-3. D3, R11 and
C7 allow the oscillator to start up regardless of the state of
the oscillator on powerup.
Performance
The circuit provides excellent transient response, efficiencies in the source mode of better than 80% and efficiencies
in the sink mode of better than 90%. Figure 12 shows the
step response of the terminator.
AN67 F12
Figure 12. Step Response of LT1158-Based Terminator
C1
0.1µF
VR1
LT1004
1.2V
R1
100k
+
–
R2
120k
3
2
8
4
U1A
LT1215
R3
1M
1
C2
680pF
6
5
R8
10k
U1B
LT1215
C4
0.001µF
–
+
C6
0.1µF
R6
30k
6
7
8
7
5
C3
220pF
4
3
2
U2
LT1158
 R6 
V OUT = 1.25  1 +

 R2 
D3
1N4148
9
C7
0.1µF
Figure 11. GTL 1.55V Terminator Provides 10A Max Current
Kool M µ is a registered trademark of Magnetics, Inc.
11
12
13
14
15
16
10
V+
SENSE–
SENSE+
T SOURCE
T GATE FB
T GATE DR
BOOST
B GATE DR B GATE DR
GND
INPUT
FAULT
ENABLE
BIAS
V+
BOOST DR
Q1 TO Q6: SILICONIX Si9410.
L1: Kool M µ CORE #77 548-A7 10 TURNS OF #14AWG.
C10 AND C11: NICHICON HFQ 6.3V.
R12: LR2512-R010. (MFG. IRC)
R7
36k
R5
30k
R4
15k
C5
0.01µF
1
D2
BAT85
R11
5.1k
R10
100Ω
C9
R9
0.01µF 100Ω
C8
0.1µF
Q4
R12
0.01Ω
Q1
Q5
L1
7.5µH
Q2
Q6
Q3
VIN
AN67 F11
C11
12000µF
6.3V
VOUT
C10
4000µF
6.3V
Application Note 67
AN67-13
Application Note 67
RS232 TRANSCEIVERS FOR DTE/DCE SWITCHING
by Gary Maulding
Switched DTE/DCE Port
There are situations where a data port is required to act
alternately as either a DTE or a DCE. Examples include test
equipment and data multiplexers. Figure 13 shows a
circuit that can switch from a 9-pin DTE to a 9-pin DCE
configuration while maintaining full compliance with the
RS232 standards.
The circuit uses an LT1137A DTE transceiver and an
LT1138A DCE transceiver. A DTE/DCE select logic signal
alternately activates or shuts down one of the two transceivers. In addition to drawing no power, the OFF
transceiver’s drivers achieve a high impedance state,
removing themselves from the data line. The receiver
inputs will continue to load the line, but this presents no
operational problem and does not violate the RS232
standard. The drivers on the activated transceiver can
easily drive the extra load of the companion transceiver’s
inputs along with the termination at the opposite end of
the cable. The scope photograph (Figure 14) shows the
signal outputs of the DTE/DCE switched circuit driving
3k || 1000pF at 120kBd.
To the transceiver at the opposite end of the data line the
data port always appears to be a normal fixed port. All
signals into the port are properly terminated in 5k.
AN67-14
The schematic in Figure 13 shows the essential features
needed to implement DTE/DCE switching but other features can be easily included. Shutdown of both transceivers could be implemented by adding an additional logic
control signal. Multiplexing of the logic level signals is also
possible since receiver outputs remain in a high impedance state when the transceivers are shut down. Two
capacitors can be saved by sharing the V + and V – filter
capacitors between the two transceivers, but the charge
pump capacitors must not be shared.
The circuits used in the demonstration circuit are bipolar,
but Linear Technology’s CMOS transceivers, such as the
LTC1327 and 1328 could be substituted where the absolute minimum power dissipation is required.
TX IN
RX IN
RX OUT
AN67 F14
Figure 14. Oscillograph Showing Signal Outputs of the DTE/DCE
Circuit of Figure 13 Driving 3k || 1000pF at 120kBd
Application Note 67
0.1µF
VCC
0.1µF
×2
1
LT1137A
2
27
3
26
4
25
5
24
6
23
7
22
8
21
9
20
10
19
11
18
12
17
13
1
LT1138A
28
27
0.1µF
×2
3
26
4
25
RI
5
24
CTS
6
23
DTR
7
22
RD
8
21
TD
9
20
DSR
10
19
RTS
11
18
DCD
12
17
13
0.1µF
×2
15
2
VCC
0.1µF
16
ON/OFF
14
0.1µF
28
ON/OFF
14
0.1µF
0.1µF
×2
16
15
DTE/DCE
AN67 F13
Figure 13. Switchable, 9-Pin DTE/DCE Data Port Circuitry
AN67-15
Application Note 67
ACTIVE NEGATION BUS TERMINATORS
by Dale Eagar
High speed data buses require transmission line techniques, including termination, to preserve signal integrity.
Lost data on a bus can be attributed to reflections of the
signals from the discontinuities of the bus. The solution to
this problem is proper termination of the bus.
The circuit will source or sink current. Current is sourced
from the 5V supply through Q1, an NPN Darlington, to the
output. The sink current flows through CR1 into the
collector (Pin 1) of the LT1431, and to ground. The LT1431
regulates a scaled version of the output voltage against the
VCC
VCC
Early designs of bus terminators were passive (see Figure
15). Passive termination works great but wastes lots of
precious power, especially when the bus is not being used.
BUS
LINE
The ideal solution is a voltage source capable of both
sourcing and sinking current. Such a voltage source, with
termination resistors, is shown in Figure 16. This is called
active negation. Active negation uses minimal quiescent
current, essentially providing only the power needed to
properly terminate the bus.
BUS
LINE
AN67 F15
Figure 15. Passive Termination Technique
BUS
LINE
BUS
LINE
Active Negation Bus Terminator
Using Linear Voltage Regulation
E 2.85V
The active negation circuit shown in Figure 17 provides the
power to the output at an efficiency of about 50%; the rest
of the power is dissipated in either Q1 or U1 depending on
the polarity of the output current.
AN67 F16
Figure 16. Active Negation Termination Technique
VCC = 5V
0.1µF
1k
Q1
TIP121
1nF
1
COLL
3
V+
CR1
1N4001
U1
LT1431
8
+
+ 100µF
6.3V
–
7
5k
2.5V
SGND
5
R1
680Ω
FGND
6
AN67 F17
Figure 17. Linear Active Negation Voltage Source
AN67-16
BUS
LINES
2.85V
•
•
•
Application Note 67
internal 2.5V bandgap reference, driving the base of Q1 or
drawing current through CR1 to regulate the output voltage. R1 and the internal 5k resistor of the LT1431 scale the
output voltage.
duty factor oscillator. The duty factor is controlled by the
output of the regulator, U3, and is maintained at the ratio
of 2.85V/VIN. VIN is the 5V supply that powers U1, U2 and
U3. The output voltage is the average voltage of the square
wave (VIN)(duty factor) from the outputs of U1B–U1F and
U2A–U2F. L1 and C2 filter the AC component of the 0V to
5V signal yielding a DC output voltage of 2.85V.
Switching Power Supply, Active Negation Network
The switching active negation terminator shown in Figure
18 is a synchronous switcher. This solution further reduces dissipation and therefore achieves higher efficiency.
This type of switcher can both source and sink current.
CR1 is added to prevent latchup of U1 and U2 during
adverse conditions.
A logic gate could easily be added to the oscillator to add
a disable function to this terminator, further lowering the
quiescent power when termination is not needed.
The switching power supply operates as follows. The
74AC04 hex inverters (U1 and U2) form a 1MHz variable
C1
20pF
U1B
39k
•
•
•
U1F
U1A
CR1
1N5818
U2A
+
•
•
•
U2F
5V
4.7µF
+
14
74AC04
U1
7
C2
100µF
6.3V
L1
40µH
BUS
LINES
VIN = 5V
20k
3.9k
0.1µF
•
•
•
5V
5V
4.7µF
+
3
10k
1
14
74AC04
U2
7
8
U3
LT1431
4
5
6
7
24.9k
0.01µF
5.1k
AN67 F18
Figure 18. Switching Active Negation Termination
AN67-17
Application Note 67
RS485 REPEATER EXTENDS SYSTEM CAPABILITY
by Mitchell Lee
RS485 data communications are specified for distances of
up to 4000 feet. This limit is the consequence of losses in
the twisted pair used to carry the data signals. Beyond
4000 feet, skin effect and dielectric losses take their toll,
attenuating the signal beyond use.
If greater distances must be covered some means of
repeating the data is necessary. One method is to terminate a long run of cable with a microprocessor-based node
capable of relaying data to yet another length of cable.
A more simple solution* is shown in Figure 19. Two
RS485 transceivers are connected back-to-back so as to
relay incoming data from either side to the other. A pair of
cross coupled one-shots furnish a means of “flow control”
so that one and only one transmitter is turned on at any
given time. Incoming data is sensed by detecting a 1-0
transition at the output of either idling receiver. The first
receiver to spot such a transition triggers its associated
one-shot, which, in turn, activates the opposite transmitter and ensures smooth data flow from one side to the
other. At the same time the one-shot locks out the other
receiver/transmitter/one-shot combination so that only
one data path is open.
The one-shot is retriggered by successive 1-0 transitions
and start bits, holding the data path in this configuration.
The one-shot time constant is set slightly greater than the
interval between any two start bits. When the received data
stops, the line idles high, producing a 1 at the receiver’s
output. The one-shot resets, returning the opposite transceiver to the receive mode—ready for any subsequent
data flow.
In order to allow adequate time for the one-shot to reset,
the software protocol must wait one word length after the
end of any data transmission before responding to a call
or initiating a new conversation. As shown, the repeater is
set up for 100kBd data rates and an 8-bit word length (plus
start and stop bits).
*Honeywell Inc. patent 4, 670, 886 may apply.
5V
5V
LTC485
750Ω
LTC485
10nF
10nF
RX
750Ω
RX
130Ω
130Ω
TX
TX
750Ω
750Ω
5V
5V
5V
10k
5V
B
10k
10k
10nF
Q
Q
1/2 74HC123
A
CLR
10nF
B
10k
1/2 74HC123
Q
Q
CLR
A
AN67 F19
Figure 19. RS485 Repeater Schematic Diagram
AN67-18
Application Note 67
AN LT1087-BASED 1.2V GTL TERMINATOR
by Mitchell Lee
A recent development in high speed digital design has
resulted in a new family of logic chips called Gunning
Transition Logic (GTL). Because of the speeds involved,
careful attention must be paid to the transmission line
characteristics of the interconnections between these
chips; active termination is required.
The termination voltage is 1.20V and currents of several
amperes are common in a complete system. One method
of generating 1.2V is to use a linear regulator operating
from 3.3V or 5V. Unfortunately, this method suffers from
two major drawbacks. First, the minimum adjust voltage,
without the aid of a negative supply, is 1.25V for most
adjustable linear regulators. Second, most low voltage
linear regulators do not feature low dropout characteristics, rendering them unusable on a 3.3V input. The LT1087
solves both of these problems with an output that can be
adjusted to less than the reference voltage and a low
dropout architecture.
LTC1145/LTC1146 ACHIEVE LOW PROFILE ISOLATION
WITH CAPACITIVE LEAD FRAME
by James Herr
The LTC1145 and LTC1146 are a new generation of signal
isolators. Previously, signal isolation was accomplished
by means of optoisolators. Light from an LED was detected across a physical isolation barrier by either a photo
diode or transistor and converted to an electrical signal.
Isolation levels up to thousands of volts were easily
achieved.
Attempts have been made to provide signal isolation on a
single silicon die. Problems arose due to reliability constraints of damage from ESD or overvoltage. A new
technique, using a capacitive lead frame, overcomes the
problems associated with single package signal isolation.
Further, this technique is suitable for use in thin surface
mount packages — a solution not available with
optoisolators. The data rates are 200kbps for the LTC1145
and 20kbps for the LTC1146. Both parts can sustain over
1000V across their isolation barriers.
Figure 20 shows the complete circuit. The LT1087 features feedback sense, which, in its original application,
was used for remote Kelvin sensing. In the GTL terminator
circuit the Sense pins are used to adjust the internal 1.25V
reference downward. The result is a 1.20V, 5A regulator
with 2% output tolerance over all conditions of line, load
and temperature. To minimize power dissipation a 3.3V
input source is recommended.
1.20V
5A
2%
VIN = 3.3V
VIN
+
VOUT
SENSE –
4.42Ω
1%
LT1087CT
+
10µF
22µF
TANT
SENSE +
ADJ
1.21k
1%
AN67 F20
Figure 20. 1.2V GTL Termination Voltage Schematic Diagram
Applications
The LTC1145/LTC1146 can be used in a wide range of
applications where voltage transients, differential ground
potentials or high noise may be encountered, such as
isolated serial data interfaces, isolated analog-to-digital
converters for process control, isolated FET drivers and
low power optoisolator replacement. One possible application is an isolated RS232 receiver. The DIN pin of the
LTC1145 is driven by an RS232 signal through a 5.1k
resistor (Figure 21). The DOUT pin of the LTC1145 presents
RS232
IN
5.1k
1 DIN
GND1 18
ISOLATION
BARRIER
5V
7 NC
VCC 12
8 OSCIN
OS 11
9 GND2
DOUT 10
TTL OUT
AN67 F21
Figure 21. Isolated Low Power RS232 Receiver
AN67-19
Application Note 67
isolated TTL-compatible output signals. The GND2 pin of
the LTC1145 is connected to the same ground potential as
the receiving end of the link. The isolator can accommodate differences of up to 1kV between GND1 and GND2.
sponse to a 0°C to 100°C temperature excursion (see LTC
Application Note 45 for the details). The pulses from I3
drive the DIN pin of LTC1146. The GND1 pin is connected
to the same ground potential as I3. The DOUT pin of
LTC1146 presents isolated, TTL-compatible output signals. The circuit consumes only 460µA maximum, allowing it to operate from a 9V battery.
Another application is an isolated, thermocouple-sensed
temperature-to-frequency converter (see Figure 22). The
output of I3 produces a 0kHz to1kHz pulse train in re-
CONTROL
AMPLIFIER
6V
TYPE K
THERMOCOUPLE
V+
LT1025
GND
–
+
10k
–
1M
A1
LTC1049
"A"
Q2
2N3906
+
R–
6.81k*
Q1
2N3904
0.02µF
100k
k
NC
6V
C1
100pF
C3
0.47µF
I1
V➞ F
100k
I2
C4
300pF
240k
+
1.5k
100°C
TRIM
OUTPUT
0°C TO100°C
= 0kHz TO 1kHz
I3
6V
6.8µF
DIN
LT1004-1.2
DOUT
LTC1146
16
9
14
15
C2
390pF† 11
GND1
10
S1
S4
S3
S2
2
3
1
LTC201
CHARGE
PUMP
6
7
8
GND2
*IRC/TRW-MTR-5/+120PPM
= 74C14
† = POLYSTYRENE
FOR GENERAL PURPOSE 4mV FULL SCALE V ➞ F DELETE
THERMOCOUPLE/LT1025 PAIR AND DRIVE POINT "A."
AN67 F22
Figure 22. Isolated Temperature-to-Frequency Converter
AN67-20
Application Note 67
LTC485 LINE TERMINATION
by Bob Reay
The termination of the data line connecting LTC485 transceivers is very important because an improperly terminated line can cause data errors. The data line is usually a
120Ω shielded twisted pair of wires that is terminated at
each end with a 120Ω resistor (Figure 23). For some
applications a problem occurs when the output of the
drivers is forced into a high impedance state because the
termination resistors short the inputs to the receivers.
Since the receivers are differential comparators with builtin hysteresis, their output will remain in the last logic state.
For the applications that must force the outputs of the
receivers to a known state, but still maintain low power
consumption, the cable can be terminated as in Figure 24.
A capacitor (typically 0.1µF) has been connected in series
DI
120Ω
D
RO
with the 120Ω termination resistor R2 and two bias
resistors (R1 and R3) have been added. When data is
being transmitted the capacitor looks basically like a short
circuit and a differential signal is developed across the
termination resistor. When the drivers are forced into a
high impedance state, the bias resistors force the receiver
into a logic 1 state. The receiver inputs can be reversed
when the output must be a logic 0.
Because the capacitor is in series with the bias string, no
DC current flows when data is not being transmitted. Care
must be taken to transmit data at a high enough rate to
prevent the bias resistors from charging the capacitor to
the wrong state before the next data bit arrives. Also note
that differences in the V + supplies or grounds will cause
DC current to flow in the cable, but this can be kept to a
minimum by using high value bias resistors.
120Ω
D
R
R
DI
RO
AN67 F23
Figure 23. DC Coupled Termination
5V
DI
D
5V
R1
25k
R1
25k
R2
120Ω
R2
120Ω
0.1µ F
RO
R
R3
25k
D
DI
0.1µ F
R3
25k
R
RO
AN67 F24
Figure 24. AC Coupled Termination
AN67-21
Application Note 67
Filters
How to Design a Filter from the Tables:
SALLEN AND KEY FILTERS USE 5% VALUES
by Dale Eagar
Pick a cutoff frequency in Hertz as if it were a standard 5%
resistor value in Ohms. (that is, if you want a cutoff
frequency of 1.7kHz you must choose between 1.6k and
1.8k)
Taking advantage of the wider range of resistor values is
not altogether trivial; the mathematics can be quite cumbersome and time consuming.
This Design Idea includes tables of resistor and capacitor
values for third-order Sallen and Key lowpass filters. The
resistor values are selected from the standard 5% value
pool, and the capacitor values are selected from the
standard 10% value pool. Frequencies are selected from
the standard 5% value pool used for resistors. Frequencies are in Hertz, capacitance in Farads and resistance in
Ohms.
1. Select a diagonal that represents the frequency multiplier (think of the third color band on a 5% resistor).
2. Choose a particular diagonal box by either choosing a
capacitor multiplier from the rows of the table that give
you a desired capacitor value or by choosing a resistor
multiplier from the columns of the table that gives you
a desired resistance value.
Multiply the resistors and capacitors by the scale factors
for the rows and columns that intersect at the chosen
frequency multiplier box. (for example, 0.68 • 1µF =
0.68µF, 0.47 • 1kΩ = 470Ω).
PSpice is a trademark of MicroSim Corporation.
20
0
1.6kHz LPF
VOUT
Figure 26 details the PSpiceTM simulation of a 1.6kHz
Butterworth filter designed from these tables.
Select a scale factor for the resistors and capacitors from
Table 3 by the following method:
(dB)
When the three resistors are the same value, the pole
placement, and thus the filter characteristics, are set by
the capacitor values (C1, C2 and C3). This procedure,
although great for the mathematician, can lead to problems. The problem is that, in the real world, the resistors,
not the capacitors, are available in a large selection of
values.
Select the component values from Table 1 or Table 2 as
listed for the frequency (think of the first two color bands
on a resistor).
C2
–20
0.47µF
VIN
Lowpass filters designed after Sallen and Key usually take
the form shown in Figure 25. In the classic Sallen and Key
circuit, resistors R1, R2 and R3 are set to the same value
to simplify the design equations.
–
–40
270Ω 430Ω 820Ω
–
R1
0Ω
VIN
R2
C1
R3
+
C3
VOUT
–60
0.47µF
0.047µF
AV
–80
0.01
0.1
1
10
FREQUENCY (kHz)
100
AN67 F26
VOUT
Figure 26. PSpice Simulation of 1.6kHz Butterworth Filter
AN67 F25
Figure 25. Sallen and Key Lowpass Filter
AN67-22
LT1007
+
VIN
Application Note 67
Table 1. Bessel Lowpass Filter
FREQ
Table 2. Butterworth Lowpass Filter
R1
R2
R3
C1
C2
C3
R1
R2
R3
C1
C2
C3
1.0
1.1
1.2
0.39
0.36
0.33
0.43
0.39
0.36
8.20
7.50
6.80
0.47
0.47
0.47
0.22
0.22
0.22
0.01
0.01
0.01
FREQ
1.0
1.1
1.2
0.36
0.47
0.36
3.3
0.47
0.62
3.3
6.2
1.0
0.47
0.47
0.47
0.10
0.47
0.47
0.022
0.010
0.047
1.3
1.5
1.6
0.36
0.33
0.30
2.40
4.70
0.10
0.033
0.012
0.240
0.22
0.22
0.47
2.20
4.70
2.20
0.047
0.022
0.047
1.3
1.5
1.6
0.27
0.24
0.27
2.00
1.60
0.43
0.33
0.3
0.82
0.47
0.47
0.47
0.47
0.47
0.47
0.047
0.047
0.047
1.8
2.0
2.2
0.30
0.27
0.24
3.30
0.51
2.70
5.10
0.027
0.43
0.22
0.22
0.22
0.022
2.20
0.10
0.010
0.100
0.022
1.8
2.0
2.2
0.43
0.36
0.24
1.20
7.50
0.24
0.13
0.18
3.00
0.22
0.22
0.47
1.00
0.47
0.47
0.047
0.010
0.010
2.4
2.7
3.0
0.22
0.27
0.18
2.70
0.43
0.82
3.60
1.30
0.16
0.22
0.22
0.22
0.022
0.10
0.22
0.010
0.022
0.047
2.4
2.7
3.0
0.33
0.27
0.24
0.91
5.60
5.10
0.043
0.062
0.056
0.22
0.22
0.22
2.20
1.00
1.00
0.047
0.010
0.010
3.3
3.6
3.9
0.15
0.18
0.15
0.056
0.16
1.50
1.00
0.022
2.20
0.47
0.22
0.22
1.00
2.20
0.022
0.010
0.100
0.010
3.3
3.6
3.9
0.22
0.22
0.24
1.60
0.56
0.39
0.30
0.068
0.68
0.22
0.22
0.22
0.22
1.00
0.22
0.022
0.047
0.022
4.3
4.7
5.1
0.13
0.20
0.18
0.22
0.12
0.068
0.013
1.20
0.039
0.22
0.22
0.22
2.20
0.22
2.20
0.100
0.010
0.047
4.3
4.7
5.1
0.18
0.16
0.16
0.51
1.30
0.36
0.024
0.039
0.051
0.22
0.22
0.22
2.20
1.00
1.00
0.047
0.022
0.047
5.6
6.2
6.8
0.20
0.15
0.16
1.10
0.091
0.91
0.036
0.91
0.03
0.10
0.22
0.10
0.47
0.22
0.47
0.022
0.010
0.022
5.6
6.2
6.8
0.13
0.13
0.24
1.10
0.36
1.60
0.033
0.016
0.33
0.22
0.22
0.10
1.00
2.20
0.10
0.022
0.047
0.010
7.5
8.2
9.1
0.15
0.10
0.13
1.80
0.12
0.56
0.27
1.00
0.12
0.10
0.22
0.10
0.047
0.10
0.10
0.010
0.010
0.022
7.5
8.2
9.1
0.12
0.12
0.18
0.30
0.11
1.50
1.20
0.024
0.091
0.22
0.22
0.10
0.10
2.20
0.22
0.010
0.047
0.010
AN67-23
Application Note 67
Table 3. Frequency Multipliers
0.1Ω
1Ω
10Ω
100Ω
1k
10k
100k
1M
10M
100M
1F
10
1
0.1
0.001
—
—
—
—
—
—
0.1F
100
10
1
0.1
0.01
0.001
—
—
—
—
10,000µF
1k
100
10
1
0.1
0.01
0.001
—
—
—
1,000µF
10k
1k
100
10
1
0.1
0.01
0.001
—
—
100µF
100k
10k
1k
100
10
1
0.1
0.01
0.001
—
10µF
1M
100k
10k
1k
100
10
1
0.1
0.01
0.001
1µF
10M
1M
100k
10k
1k
100
10
1
0.1
0.01
0.1µF
100M
10M
1M
100k
10k
1k
100
10
1
0.1
0.01µF
1G
100M
10M
1M
100k
10k
1k
100
10
1
1,000pF
—
1G
100M
10M
1M
100k
10k
1k
100
10
100pF
—
—
1G
100M
10M
1M
1OOk
10k
1k
100
AN67-24
Application Note 67
passband response and the area of passband gain variation. Outside the filter’s passband, signal attenuation
increases to more than 50dB for frequencies between
0.96fCENTER and 1.04fCENTER. Quiescent current is typically 2.3mA with a single 5V power supply.
LOW POWER SIGNAL DETECTION
IN A NOISY ENVIRONMENT
by Philip Karantzalis and Jimmylee Lawson
Introduction
In signal detection applications where a small narrowband
signal is to be detected in the presence of wideband noise,
one can design an asynchronous (nonphase sensitive)
tone detector using an ultraselective bandpass filter, such
as the LTC1164-8. The ultranarrow passband of the
LTC1164-8 filter band limits any random noise and increases the detector’s signal sensitivity.
An Ultraselective Bandpass Filter and a Dual
Comparator Build a High Performance Tone Detector
The LTC1164-8 has excellent selectivity, which limits the
noise that passes from the input to the output of the filter.
As a result, one can build a tone detector that can extract
small signals from the “mud.” Figure 28 shows the block
diagram of such a tone detector. The detector’s input is an
LTC1164-8 bandpass filter whose output is AC coupled to
a dual comparator circuit. The first comparator converts
the filter’s output to a variable pulsewidth signal. The
pulsewidth varies depending on the signal amplitude. The
average DC value of the pulse signal is extracted by a
lowpass RC filter and applied to the second comparator.
The identification of a tone is indicated by a logic high at
the output of the second comparator.
The LTC1164-8 is an eighth-order, elliptic bandpass filter
with the following features: the filter’s fCENTER (the center
frequency of the filter’s passband) is clock tunable and is
equal to the clock frequency divided by 100; the filter’s
passband is from 0.995fCENTER to 1.005fCENTER (±0.5%
from fCENTER). Figure 27 shows a typical LTC1164-8
3
0
One of the key benefits of using a high selectivity bandpass
filter for tone detection is that when wideband noise (white
noise) appears at the input of the filter, only a small
amount of input noise will reach the filter’s output. This
results in a dramatically improved signal-to-noise ratio at
the output of the filter compared to the signal-to-noise at
the input of the filter. If the output noise of the LTC1164-8
is neglected, the signal-to-noise ratio at the output of the
filter divided by the signal-to-noise ratio at the input of the
filter is:
GAIN (dB)
–3
–6
AREA OF PASSBAND
GAIN VARIATION
–9
–12
–15
LTC1164-8 PASSBAND
(fCENTER = fCLK/100)
RF = 61.9k
RIN = 340k
–18
–1.0 –0.75 –0.5 –0.25
0.25 0.50 0.75 1.0
fCENTER
PERCENT DEVIATION FROM fCENTER
AN67 F27
(S/N)OUT
= 20 Log
(S/N)IN
Figure 27. Detail of LTC1164-8 Passband
VIN
f
fIN = CLK
100
ULTRANARROW
BANDPASS FILTER
WITH GAIN
REF 1
(BW)IN
(BW)f
+
AC BUFFER
COMP 1
VARIABLE PULSE WIDTH OUTPUT
–
LTC1164-8
PULSE AVERAGE
+
COMP 2
REF 2
LOGIC HI WHEN SIGNAL PRESENT
LOGIC LO WHEN NO SIGNAL PRESENT
–
fCLK
AN67 F28
Figure 28. Tone Detector Block Diagram
AN67-25
Application Note 67
where: (BW)IN = the noise bandwidth at the input of the
filter and (BW)f = (0.01)(fCENTER) is the filter’s noise
equivalent bandwidth.
For example, a small 1kHz signal is sent through a cable
that is also conducting random noise with a 3.4kHz
bandwidth. An LTC1164-8 is used to detect the 1kHz
signal. The signal-to-noise ratio at the output of the filter
is 25.3db larger than the signal-to-noise ratio at the input
of the filter:
(BW)IN
= 20 Log
(BW)f
frequency at 1kHz (fCENTER = fCLK/100). A low frequency
op amp (LT1013) and resistors RIN and RF set the filter’s
gain. In order to minimize the filter’s output noise and
maintain optimum dynamic range, the output feedback
resistor RF should be 61.9k. Capacitor CF across resistor
RF is added to reduce the clock feedthrough at the filter’s
output.
To set the gain for the LTC1164-8, RIN should be calculated by the equation:
RIN = 340k/Gain
3.4kHz = 25.3dB
(0.01)(1kHz)
Figure 29 shows the complete circuit for a 1kHz tone
detector operating with a single 5V supply. An LTC1164-8
with a clock input set at 100kHz sets the tone detector’s
In Figure 29, the filter’s gain is 10 (RIN = 34k). Capacitor
C1 and a unity-gain op amp (LT1013) AC couple the signal
at the filter’s output to an LTC1040 dual low power
comparator. AC coupling is required to eliminate any DC
offset caused by the LTC1164-8.
5V
0.1µF
5V
fCLK
R2
10k
4
RIN
34k
2
VIN
1
CF
200pF
7
0.1µF
LTC1164-8
1.0µF
6
RF
61.9k
AGND
5
2
–
3
+
6
8
10 12 13
R1
10k
–
8
1/2 LT1013
5
REF. 1
(1.9V)
7
+
5
6
7
8
C1
0.22µF
1/2 LT1013
1
18
5V
14
3
0.1µF
11
STROBE
+
–
+ COMP 1
–
1
4
LTC1040
14
4
R3
10k
REF. 2 13
C2
0.47µF (1V) 12
11
+
–
+ COMP 2
–
15
5V
30.1k
AGND (2V)
0.1µF
RIN = 340k/GAIN, fCENTER = fCLK/100
(1/(2 π RF CF) ≥ 10 • fCENTER)
(1/(2 π R1 C1) ≤ fCENTER/10)
(1/(2 π R3 C2) ≤ fCENTER/32)
1k
8.87k
REF. 2 (1V)
0.1µF
10k
Figure 29. 1kHz Tone Detector with Gain of 10
AN67-26
10
AN67 F29
REF. 1 (1.9V)
0.1µF
9
VOUT
Application Note 67
A resistive divider generates a 2V bias for the LTC1164-8
“ground” (Pins 3 and 5) and the positive input of the
LT1013 dual op amps. For single 5V operation the output
swing of the LTC1164-8 is from 0.5V to 3.5V, centered at
2V. The divider also provides the reference voltages for the
LTC1040 dual comparators (Ref. 1 = 1.9V and Ref. 2 = 1V).
Power supply variations do not affect the performance of
this circuit because all DC reference voltages are derived
from the same resistor divider and will track any changes
in the 5V power supply.
of the maximum peaks of wideband noise with uniform
spectral density). Therefore, the maximum allowable noise
at the filter’s output is 32mVRMS (160mVPEAK/5). The
noise at the filter’s output depends on the filter’s gain and
noise equivalent bandwidth and the spectral density of the
noise at the filter’s input. Therefore, the maximum input
noise spectral density for Figure 3’s circuit is:
Theory of Operation
where: Gain is the filter’s gain at its center frequency and
(BW)f is the filter’s noise equivalent bandwidth.
The tone detector works by looking at the negative peaks
at the output of the filter. Signals below 1.9V at the output
of the filter trip the first comparator. The second comparator has a 1V reference and detects the average value of the
output of the first comparator. The R3/C2 time constant is
set to allow detection only if the duty cycle of the first
comparator’s output exceeds 25%. Waveforms with duty
cycles below 25% are arbitrarily assumed to carry false
information
The circuitry is designed so that two or more negative
signal peaks of 160mV at the filter’s output produce a 25%
duty cycle pulse waveform at the output of the first
detector (the 1.9V and 1V references for comparators 1
and 2 respectively, set the 160mVPEAK and the 25% duty
cycle). The 25% duty cycle requirement establishes an
operating point or “minimum detectable signal” for the
detector circuit. Thus, the circuitry outputs a “tone present”
condition only when the duty cycle is greater than or equal
to 25%. The 25% duty cycle requirement sets two conditions for optimum tone detection at the detector’s input.
The first input condition is the maximum input noise
spectral density that will not trigger the detector’s output
to indicate the presence of a tone. When only noise is
present at the filter’s input, the maximum input noise
spectral density is conservatively defined as the amount
required to produce noise peaks at the filter’s output of
160mV or lower amplitude. The 160mV maximum noise
peak specification at the filter’s output can be converted to
output noise in mVRMS by using a crest factor of 5 (the
crest factor of a signal is the ratio of its peak value to its
RMS value—a theoretical crest factor of 5 predicts 99.3%
V
eIN ≤ 32mVRMS /(Gain • √(BW)f) RMS
√Hz
Note: Compared to 32mVRMS the 270mVRMS output noise
of the LTC1164-8 is negligible. The output noise of the
LTC1164-8 is independent of the chosen filter signal gain.
The second input condition is the minimum input signal
required so that a tone can be detected when it is buried by
the maximum noise, as defined by the first input condition.
When a tone plus noise is present at the filter’s input, the
output of the filter will be a tone whose amplitude is
modulated by the bandlimited noise at the filter’s output.
If a maximum noise peak of 160mV modulates the tone’s
amplitude, a 320mV tone peak at the filter’s output can be
detected because the product of the noise and the tone
crosses the (negative) 160mVPEAK detection threshold
and the 25% duty cycle requirement is exceeded. Therefore, a conservative value for the minimum signal at the
filter’s output can be set to 320mVPEAK or 226mVRMS, but
a value of 200mVRMS was established experimentally.
Therefore, the minimum input signal for reliable tone
detection in the presence of the maximum input noise
spectral density is:
VIN(MIN) = 200mVRMS /Gain
For optimum tone detection, the signal’s frequency should
be in the filter’s passband, within ±0.1% of fCENTER.
Conclusion
A very selective bandpass filter, the LTC1164-8, can be
configured as a nonphase-sensitive tone detector. This
allows signals to be detected in the presence of comparatively large amounts of noise or signal-to-noise ratios that
are less than unity.
AN67-27
Application Note 67
gives the complete expression for the forward gain as a
function of frequency:
BANDPASS FILTER HAS ADJUSTABLE Q
by Frank Cox
The bandpass filter circuit shown in Figure 30 features an
electronically controlled Q. Q for a bandpass filter is
defined as the ratio of the 3dB pass bandwidth to the stop
bandwidth at some specified attenuation. The center
frequency of the bandpass filter in this example is 3MHz,
but this can be adjusted with appropriate LC tank components. The upper limit of the usable frequency range is
about 10MHz. The width of the passband is adjusted by
the current into Pin 5 (set current or ISET) of the transconductance amplifier segment of IC1, an LT1228. Figure 31
is a network analyzer plot of frequency response versus
set current. This plot shows the variation in Q while the
center frequency and the passband gain remain relatively
constant.
A(s) = 10 ISET ACFA
R7
R6 + R7
R4 + R5
and ACFA =
R4
1
Setting B(s) =
R
ACFA RATIO
B(s) =
and substituting these expressions into the first equation
gives:
H(s) =
A(s)
1 + A(s) B(s)
where A(s) is the forward gain and B(s) is the reverse gain.
The forward gain is the product of the transconductance
stage gain (gm) and the gain of the CFA (ACFA). For this
circuit, gm is ten times the product of ISET and the impedance of the tank circuit as a function of frequency. This
R7
75Ω
RRATIO
)
1 + 10 ISET
)
H(s) =
S
1
RRATIO
S2 + S
)
)
3
R3
75Ω
+
–
CFA
(LT1228)
1
1
6
1 10 ISET √LC + 1
C
LC
√LC
50Ω
+
5
5.3µH
536pF
ISET
AN67 F30
Figure 30. LT1228 Bandpass Filter Circuit Diagram
AN67-28
)
)
1 10 ISET √LC
C
√LC
R5
750Ω
–
gm
(LT1228)
)
sL
1 + s2 LC
R6
750Ω
2
)
sL
1 + s2 LC
The last equation can be rewritten as:
8
R1
50Ω
10 ISET
1
R4
75Ω
R2
1k
)
sL
1 + s2 LC
The reverse gain is simply:
The circuit’s operation is best understood by analyzing the
closed-loop transfer function. This can be written in the
form of the classic negative feedback equation:
H(s) =
)
Application Note 67
The transfer function of a second order bandpass filter can
be expressed in the form1:
H(s) = HBP
S (ωO/Q)
S2 + S (ωO/Q) + ωO2
Comparing the last two equations note that
10 ISET √LC
ωO = 1 and 1 =
C
Q
√LC
And therefore Q =
C
10 ISET √LC
It can be seen from the last equation that the Q is inversely
proportional to the set current.
Many variations of the circuit are possible. The center
frequency of the filter can be tuned over a small range by
the addition of a varactor diode. To increase the maximum realizable Q, add a series LC network tuned to the
same frequency as the LC tank on Pin 1 of IC1. To lower
the minimum obtainable Q, add a resistor in parallel with
the tank circuit. To create a variable Q notch filter,
connect the inductor and capacitor at Pin 1 in series
rather than in parallel.
A variable Q bandpass filter can be used to make a variable
bandwidth IF or RF stage. Another application for this
circuit is as a variable-loop filter in a phase locked loop
phase demodulator. The variable Q bandpass filter is set
for a wide bandwidth while the loop acquires the signal and
is then adjusted to a narrow bandwidth for best noise
performance after lock is achieved.
1Thanks to Doug La Porte for this equation hack.
0
ISET =
200µA
RELATIVE AMPLITUDE (dB)
–10
250µA
150µA
–20
100µA
50µA
–30
25µA
–40
300kHz
3MHz
6MHz
AN67 F31
Figure 31. Network Analyzer Plot of Frequency Response vs “Set” Current
AN67-29
Application Note 67
AN ULTRASELECTIVE BANDPASS FILTER
WITH ADJUSTABLE GAIN
by Philip Karantzalis
VIN
5V
Introduction
The LTC1164-8 is a monolithic, ultraselective, eighth
order elliptic bandpass filter. The passband of the LTC11648 is tuned with an external clock; the clock-to-centerfrequency ratio is 100:1. The stopband attenuation of the
LTC1164-8 is greater than 50dB for input frequencies
outside a narrow band defined as ±4% of the center
frequency of the filter (see Figure 32).
One Op Amp and Two Resistors
Build an Ultraselective Filter
10
0
–10
–20
–30
–40
–50
–60
–70
–80
–90
0.90
0.95
1.00
1.05
FREQUENCY (kHz)
1.10
AN67 F32
Figure 32. LTC1164-8 Gain vs Frequency Response
AN67-30
1
14
2
13
3
12
4
LTC1164-8 11
5
10
6
9
7
8
CF
160pF
RF
90.9k
– 5V
100kHz
5V
2
–
3
+
7
LT1006
6
VOUT
4
–5V
AN67 F33
Figure 33. LTC1164-8 Ultranarrow, 1kHz Bandpass Filter
with Gain (Gain = 340k/RIN, 1/2πRFCF = 10 fCENTER)
Signal Detection in a Hostile Environment
The LTC1164-8 requires an external op amp and two
external resistors. The filter’s gain at its center frequency
is equal to 3.4RF/RIN. For optimum dynamic range with a
gain equal to one, the external resistor RF should be 90.9k
and the external resistor RIN should be 340k. For gains
other than 1, RIN = 340k/gain. Gains of up to 1000 are
possible. The complete configuration is shown in Figure
33. Note that programming the filter’s gain with input
resistor RIN is equivalent to providing the LTC1164-8 with
noiseless preamplification, since the filter’s internal noise
is not amplified. The wideband noise of the LTC1164-8
measures 400µVRMS at ±5V and is independent of the
filter’s gain and center frequency. A capacitor, CF, across
resistor RF reduces clock feedthrough and provides a
smooth sine wave output.
GAIN (dB)
RIN
340k
An outstanding feature of the LTC1164-8 is its
ultraselectivity. A bandpass filter with ultraselectivity is
ideal for signal detection applications. One signal detection application occurs when two signals are very closely
spaced in the frequency spectrum and only one of the
signals has useful information. The LTC1164-8 can extract
the signal of interest and suppress its unwanted neighbor.
For example, a small 1kHz, 10mVRMS signal is combined
with an unwanted 950Hz, 40mVRMS signal. The two signals differ in frequency by only 5% and the 950Hz signal
is four times larger than the 1kHz signal. To detect the 1kHz
signal, the LTC1164-8 is set to a gain of 100 and the clock
frequency is set to 100kHz. At the filtered output of the
LTC1164-8 the following signals will be present: an extracted 1kHz, 1VRMS signal and a rejected 950Hz, 2.7mVRMS
signal, as shown in Figure 34. In a narrowband signal
separation and extraction application, as described previously, the LTC1164-8 provides a simple and reliable detection circuit solution.
A second signal detection application occurs when a small
signal is to be detected in the presence of noise. For
example, a 1kHz, 10mVRMS signal is mixed with a wideband
noise signal that measures 5mVRMS in a 400Hz frequency
band. The signal-to-noise ratio is just 6dB. With the
LTC1164-8 set for a center frequency of 1kHz (fCLK is equal
to 100kHz) and a gain of 100, the 1kHz, 10mVRMS signal
will be detected and amplified. The wideband noise will be
band limited by the very narrow band gain response of the
LTC1164-8. At the output of a LTC1164-8, the 1kHz signal
will be 1VRMS as shown in Figure 35. The total band limited
Application Note 67
noise will be 70mVRMS with a signal-to-noise ratio of more
than 20dB, as shown in Figure 36. In applications of signal
detection in the presence of noise, the LTC1164-8 provides asynchronous detection. Signal detection circuits
such as synchronous demodulators and lock-in amplifiers
require the presence of a reference or carrier signal to
provide phase and frequency information of the signal to
be detected. With an LTC1164-8, signal detection is accomplished by selecting a very narrow signal detection
band around the frequency of the desired signal, which is
defined as fCLK divided by 100 (fCLK is the clock frequency
of the LTC1164-8), and by selecting the filter gain by
choosing the value of a resistor.
Figure 34. Narrow Band Signal Extraction Showing Input to and
Output from the LTC1164-8 Filter. Filter fCENTER set to 1kHz with
Gain = 100
Figure 36. Wideband Noise Input to LTC1164-8 Filter. Plots
Show Input to and Output from the Filter. Filter fCENTER set to
1kHz with Gain = 100
Figure 35. Signal Detection in the Presence of Noise Example
Showing Input to and Output from the LTC1164-8 Filter. Filter
fCENTER set to 1kHz with Gain = 100
AN67-31
Application Note 67
For cutoff frequencies other than the 1kHz example shown,
use the following formula for each section:
LT1367 BUILDS RAIL-TO-RAIL
BUTTERWORTH FILTER
by William Jett and Sean Gold
ω02 = 1/(R1 C1 R2 C2),
where R1 = 1/(ω0 Q C1) and R2 = Q/(ω0C2)
Single Supply 1kHz, 4th Order Butterworth Filter
The circuit shown in Figure 37 takes advantage of all four
op amps in the LT1367 to form a 4th order Butterworth
filter. The filter is a simplified state-variable architecture
consisting of two cascaded second order sections. Each
section uses the 360 degree phase shift around the two op
amp loop to create a negative summing junction at A1’s
positive input.1 The circuit has two-thirds the power
dissipation and component count as the classic three op
amp biquad,2 yet it has the same low component sensitivities for center frequency, ω0 and Q.
The DC bias applied to A2 and A4 for single supply
operation is not needed when split supplies are available.
The circuit’s output can swing rail-to-rail and displays the
maximally flat amplitude response with a 1kHz cutoff
frequency with 80dB/decade rolloff (Figure 38).
1 Hahn, James. 1982. State Variable Filter Trims Predecessor’s Component Count. Electronics,
April 21, 1982.
2 Thomas, L.C. 1971. The Biquad: Part I—Some Practical Design Considerations.
IEEE Transactions on Circuit Theory, 3:350-357, May 1971.
SECTION 1
SECTION 2
C1
10,000pF
VIN
R1
29.5k*
–
A1
1/4 LT1367
C2
10,000pF
R2
8.6k*
10,000pF
–
A2
1/4 LT1367
+
+
10,000pF
11.8k*
–
A3
1/4 LT1367
21.5k*
–
A4
1/4 LT1367
+
29.5k*
VOUT
+
3.3V
20k
13k
1µF
11.8k*
AN67 F37
* = 1% RESISTORS
Figure 37. 1kHz 4th Order Butterworth Filter
Figure 38. Frequency Response of 4th Order Butterworth Filter
AN67-32
Application Note 67
DC ACCURATE, CLOCK TUNABLE LOWPASS FILTER
WITH INPUT ANTIALIASING FILTER
by Philip Karantzalis
In a sampled data system, the sampling theorem says that
if an input signal has any frequency components greater
than one half the sampling frequency, aliasing errors will
appear at the output. In practice aliasing is not always a
serious problem. High order switched capacitor lowpass
filters are band limited and significant aliasing occurs only
for input signals centered around the clock frequency and
its multiples.
Figure 39 shows the LTC1066-1 aliasing response when
operated with a clock-to-fC ratio of 50:1. With a 50:1 ratio,
the LTC1066-1 samples its input twice during one clock
period and the effective sampling frequency is twice the
clock frequency. Figure 39 shows that the maximum
aliased output is generated for inputs in the range of
2( fCLK ±fC). (fC is the cutoff frequency of the LTC1066-1.)
For instance, if the LTC1066-1 is programmed to produce
a cutoff frequency of 20kHz with a 1MHz clock, maximum
aliasing will occur only for input signals in the narrow
range of 2MHz ±20kHz and its multiples.
The simplest antialiasing filter is a passive 1st order
lowpass RC filter. The – 3dB frequency of the RC filter
should be chosen so that the passband of the RC filter does
not influence the passband of the LTC1066-1. When the
LTC1066-1 clock frequency is 500kHz, an RC filter with the
– 3dB frequency set at 50kHz attenuates by 26dB any
ALIASED OUTPUT (dB)
0
–60
–80
fCLK – fC fCLK + fC
fCLK
2fCLK – 2.3fC
2fCLK
2fCLK + 2.3fC
2fCLK – fC 2fCLK + fC
INPUT FREQUENCY
AN67 F39
Figure 39. Aliasing vs Frequency fCLK/fC = 50:1 (Pin 8 to V +);
Clock is a 50% Duty Cycle Square Wave
possible aliasing inputs in the range 1MHz ±10kHz. The
passband shape of the 50kHz RC filter does not degrade
the flat passband of the LTC1066-1 at 10kHz (the passband attenuation of the 50kHz RC filter for frequencies less
than 10kHz is less than 0.2dB). If the LTC1066-1 is clock
tuned to a cutoff frequency of 5kHz (with a clock frequency
of 250kHz), the 50kHz RC filter will provide 20dB attenuation for aliasing inputs in the range of 500kHz ±5kHz.
Therefore, a 1st order lowpass RC filter will attenuate all
aliasing signals to the LTC1066-1 by a minimum of 20dB
for a clock tunable range of one octave.
For added antialiasing bandwidth, a 1st order lowpass RC
filter can be tuned by the clock signal of LTC1066-1 to
follow the cutoff frequency of the higher order filter. The
circuit is shown in Figure 40. The circuit operation is as
follows. The six comparators inside the LTC1045 detect
the clock frequency. The clock signal of the LTC1066-1 is
converted to a pulse output whose duty cycle changes
with clock frequency. The average voltage of the pulse
signal is delivered to a 4-window comparator whose
outputs drive the four analog switches of the LTC202.
When the LTC1066-1 clock frequency increases or decreases by more than one octave (2x or x/2), a capacitor
is switched in or out of the 1st order lowpass filter formed
by resistor R1 (1k) and capacitor C1. The – 3dB frequency
of the lowpass RC filter is therefore doubled or halved if the
cutoff frequency of the LTC1066-1 is doubled or halved.
Resistor R1 and capacitors C1 through C5 allow the
lowpass RC filter to be tuned over a range of five octaves,
providing at least 20dB attenuation to any LTC1066-1
input signals in the range 2(fCLK ±fC) (the RC filter also
attenuates all aliasing signals near any multiples of the
clock frequency).
The circuit in Figure 40 can be used for any clock tunable,
5-octave range for cutoff frequencies from 10Hz to
80kHz (with ±5V supplies for LTC1066-1) or for cutoff
frequencies as high as 100kHz (with ±8V supplies for the
LTC1066-1). For cutoff frequencies greater than 50kHz,
a 15pF capacitor in series with a 30k resistor should be
connected between Pins 11 and 13 of the LTC1066-1 to
minimize passband gain peaking.
AN67-33
Application Note 67
5V
1µF
0.1µF
+
1
12.1k
20
LTC1045
2
+
19
0.1µF
–
2k
0.1µF
3
+
18
–
FIRST ORDER RC
LOWPASS ANTIALIASING FILTER
–5V
5V
0.1µF
0.1µF
1k
4
4
+
17
0.1µF
–
500Ω
1
VIN
13
2
CLOCK TUNABLE,
8TH ORDER LOWPASS FILTER
R1
1k
LTC202
3
C2
14
C3
11
C4
6
C5
C1
20Ω
16
15
RF
9
0.1µF
5
10
+
16
9
1
8
2
7
–
CF
RIN
4
5
500Ω
12
10
3
11
12
5V
CIN
0.1µF
13
5
6
0.1µF
7
8
200Ω
CLOCK INPUT
(TTL OR CMOS)
LTC1045
6
9
LTC1066-1
18
OUT A
V+
17
–IN A
OUT B
16
+IN A
+IN B
15
–
V
GND
14
+
V
FIN
13
CON 1
COMP 2
12
F OUT
CON 2
11
50/100
COMP 1
10
–
CLK
V
–5V
0.1µF
0.1µF
AN67 F40
–
CP
50pF
7
20Ω
+
15
8
VOUT
0.1µF
PULSE
AVERAGE
+
14
RA
RP
CA
0.047µF
–
PULSE
OUTPUT
CLOCK FREQUENCY DETECTOR
Figure 40. DC-Accurate, Clock-Tunable Lowpass Filter with Input Antialiasing
AN67-34
Application Note 67
Use the following design guide for choosing the component values of RA, RP, RF, RIN, CF, C1 through C5, CP and
CA.
CP = 50pF, RP =
Definitions
CA = 0.047µF, RA =
1. The cutoff frequency of the LTC1066-1 is abbreviated
as fC.
2. fC(LOW) is the lowest cutoff frequency of interest
3. A range of five octaves is from fC(LOW) to 32 • fC(LOW)
Component Calculations
f
1
= C(LOW)
250
2πRFCF
C1 =
1
fC(LOW)
RIN = RF (If RF can be
chosen as 20k, RIN and CIN
are not needed)
(fC(LOW) in Hz); R1 = 1k
105 k
50 fC(LOW)
5(105)
k
50 fC(LOW)
Example
For a five octave range from 1kHz to 32kHz:
fC(LOW) = 1kHz
Let CF = 1µF ±20%, then RF = 40.2k ±1%. RIN = RF =
40.2k ±1%, CIN = 0.1µF
C1 = 0.001µF ±5%, C2 = 0.001µF ±5%, C3 = 0.0022µF
±5%
C4 = 0.0039µF ±55, C5 = 0.0082µF ±5%
CP = 50pF, RP = 2k, CA = 0.047µF, RA = 10k
C2 = C1 ±5%, C3 = 2(C1) ±5%, C4 = 4(C1) ±5%,
C5 = 8(C1) ±5%
THE LTC1066-1 DC ACCURATE
ELLIPTIC LOWPASS FILTER
by Nello Sevastopoulos
Figure 41 shows an application allowing clock tunability
from 10Hz to 100kHz. The RCCC frequency compensating
components (needed only for cutoff frequencies above
60kHz) maintain a flat passband for cutoff frequencies
between 50kHz and 100kHz. The input resistor, RI, reduces the output DC offset caused by the op amp bias
current through the 100k feedback resistor, RF. The measured DC offset and the gain nonlinearity are 4mV and
±0.0063% (84dB), respectively. The 0.1µF bypass capacitor, CB, helps keep the total harmonic distortion of the filter
from being degraded by the 100k input resistor.
Clock Tunability
An external clock tunes the cutoff frequency of the internal
switched capacitor network. The device has been optimized for a clock-to-cutoff-frequency ratio of 50:1. The
internal double sampling greatly reduces the risk of aliasing.
The maximum obtainable cutoff frequency, fCUTOFF(MAX),
depends on power supply, clock duty cycle and tempera-
RF
100k
CB
0.1µF
33µF
1
RI
100k
VIN
–7.5V
7.5V
0.1µF
0.1µF
2
3
fCLK
–IN A
OUT B
+IN A
+IN B
V–
LTC1066-1 GND
5
V+
FILTERIN
6
8
200Ω
V+
4
7
7.5V
OUT A
9
CONNECT 1
COMP 2
FILTEROUT CONNECT 2
50/100
CLK
COMP 1
V–
18
17
7.5V
VOUT
16
0.1µF
15
14
13
RC
30k
CC
15pF
12
11
10
0.1µF
–7.5V
MAXIMUM OUTPUT VOLTAGE OFFSET = 4mV, DC LINEARITY = ±0.0063%, TA = 25°C.
THE PIN 6 TO 12 CONNECTION SHOULD BE UNDER THE IC AND SHIELDED BY AN
ANALOG SYSTEM GROUND PLANE.
RC COMPENSATION BETWEEN PINS 11 AND 13 REQUIRED ONLY FOR fCUTOFF > 50kHz.
THE 33µF CAPACITOR IS A NONPOLARIZED, ALUMINUM ELECTROLYTIC, ±20%, 16V
(NICHICON UUPIC 330MCRIGS OR NIC NACEN 33M16V 6.3 × 5.5 OR EQUIVALENT).
AN67 F41
Figure 41. DC Accurate, 10MHz to 100kHz 8th Order Elliptic
Lowpass Filter, fCLK/fC = 50:1
ture; fCUTOFF(MAX) does not depend on the value of the
external resistor/capacitor combination RFCF. The RCCC
compensation is shown in Figure 41. The data detailed in
AN67-35
Application Note 67
Figure 42 reveals the important fact that for a cutoff
frequency of 100kHz, the stopband attenuation still remains greater than 70dB for input frequencies up to 1MHz.
–40
)
20 log THD + NOISE (dB)
VIN
–45
(
The minimum obtainable cutoff frequency depends on the
RFCF time constant of the servo loop. For a given RFCF time
constant, the minimum obtainable cutoff frequency of the
LTC1066-1 is:
–50
fIN = 1kHz
fCLK = 1MHz
fCLK /fC = 50:1
–55
VS = ±5V
–60
–65
–70
–75
VS = ±7.5V
–80
fCUTOFF(MIN) = 250(1/2πRFCF).
–85
fCUTOFF(MAX) = 100kHz
–90
0.1
1
INPUT VOLTAGE (VRMS)
For instance, if RF = 20k, CF = 1µF, fCUTOFF(MIN) = 2kHz, and
fCLOCK(MIN) = 100kHz.
Under these conditions, a clock frequency below 100kHz
will “warp” the passband gain by more than 0.1dB. Please
see the LTC1066-1 data sheet for more details.
10
0
fCLK= 5MHz
–10
–20
fCLK= 500kHz
GAIN (dB)
–30
–40
–50
–60
–70
VS = ±7.5V
fCLK/fC = 50:1
COMPENSATION
= 30k, 15pF
–80
–90
–100
fCLK= 2.5Mz
–110
1k
10k
100k
FREQUENCY (Hz)
1M
AN67 F42
Figure 42. LTC1066-1 Amplitude vs Frequency
5
AN67 F43
Figure 43. LTC1066-1 Dynamic Range
Aliasing and Antialiasing
All sampled data systems will alias if their input signals
exceed half the sampling rate, but aliasing for high order,
band limited, switched capacitor filters need not be a
serious problem. The LTC1066-1, when operating with a
50:1 clock-to-cutoff-frequency ratio, will have significant
aliasing only for input signals centered around twice the
clock frequency and its even multiples. Figure 44 shows
the input frequencies that will generate aliasing at the filter
output. For instance, if the filter is tuned to a 50kHz cutoff
frequency using a 2.5MHz clock, significant aliasing will
occur only for input frequencies of 5MHz ±50kHz. The
filter user should be aware of the spectrum at the input to
the filter. Next, an assessment should be made as to
whether a simple, continuous-time antialiasing filter in
front of the LTC1066-1 is required. The antialiasing filter
should do precisely what it is meant to do, that is, provide
Dynamic Range
AN67-36
ALIASED OUTPUT (dB)
0
The LTC1066-1 wideband noise is 100µVRMS. Figure 43
shows the noise plus distortion versus RMS input voltage
at 1kHz. With a ±5V supply, the filter can swing ±2.5V (5V
full scale) with better than 0.01% distortion plus noise.
The maximum signal-to-noise ratio, in excess of 90dB, is
achieved with ±7.5V supplies. Unlike previous monolithic
filters the data shown in Figure 43 is taken without using
any input or output op amp buffers. The output buffer of
the LTC1066-1 can drive a 200Ω load without dynamic
range degradation.
–60
–80
fCLK – fC fCLK + fC
fCLK
2fCLK – 2.3fC
2fCLK
2fCLK + 2.3fC
2fCLK – fC 2fCLK + fC
INPUT FREQUENCY
AN67 F44
Figure 44. Aliasing vs Frequency fCLK/fC = 50:1 (Pin 8 to V +).
Clock is a 50% Duty Cycle Square Wave
Application Note 67
band limiting. The antialiasing filter should not degrade
the DC or AC performance of the LTC1066-1.
20k
1µF
C
10nF
For fixed cutoff frequency applications, the antialiasing
function is quite trivial. Figure 45 shows the internal
precision input op amp configuration used to perform
both the DC accurate function of the LTC1066-1 and the
input antialiasing configuration. The cutoff frequency of
the RC antialiasing filter is set three times higher than the
cutoff frequency of the LTC1066-1. For the example
shown in Figure 45 the input antialiasing filter provides a
62dB attenuation at twice the clock frequency of the
switched capacitor filter.
R1
2.37k
1
2
R2
9k
3
VIN
C1
3.3nF
–7.5V
7.5V
V+
–IN A
OUT B
+IN A
+IN B
4
V–
LTC1066-1 GND
5
+
FILTERIN
6
7
8
9
fCLK = 100kHz
OUT A
V
CONNECT 1
COMP 2
FILTEROUT CONNECT 2
COMP 1
50/100
V–
CLK
18
17
7.5V
VOUT
16
15
14
13
12
11
10
–7.5V
AN67 F45
Figure 45. Adding a 2-Pole Butterworth Input Antialiasing Filter.
Set C1 = 0.33C, R2 = 3.8(R1); f– 3dB (Input Antialiasing) =
0.8993/(2πR1C)
CLOCK TUNABLE BANDPASS FILTER OPERATES TO
160kHz IN SINGLE SUPPLY SYSTEMS
by Philip Karantzalis
When the only available power supply in a system is 5V or
12V and a precision bandpass filter is needed at cutoff
frequencies greater than 20kHz, the LTC1264 switched
capacitor active filter building block can be configured to
realize an 8th order bandpass filter accurate to ±1% or
better over temperature (– 40°C to 85°C). Figure 46 is a
schematic diagram of an 8th order bandpass filter tunable
with a TTL clock signal to any center frequency up to 70kHz
with a 5V supply or to 100kHz with a 12V supply. The clock
frequency-to-center frequency ratio is 20:1. The gain
response for a 50kHz bandpass filter is shown in Figure 47
and the input dynamic range with a 5V supply is shown in
Figure 48.
R1C, 39.2k
R1B, 39.2k
VIN
1
R2B, 10k
2
1µF
+
3
R3B, 39.2k
4
5
0.1µF
6
15k
7
10k
0.1µF
8
9
VS
(5V OR 12V) R3A, 10k
10
11
R2A, 39.2k 12
INV B
INV C
HP B
HP C
BP B
LP B
BP C
LP C
24
23
R2C, 10k
22
21
20
SC
LTC1264
19
V–
GND
18
fCLK
V+
17
SA
SHDN
16
LP A
LP D
15
BP A
BP D
14
HP A
HP D
13
INV A
INV D
R3C, 39.2k
(USE ONLY FOR
12V SUPPLY)
12V
SB
22.1k 0.1µF
fCLK
(TTL)
50k
R3D, 10k
R2D, 39.2k
VOUT
R1A, 39.2k
R1D, 39.2k
VS
fCENTER
R1
(EACH SECTION)
5V
5V
12V
12V
12V
80kHz
90kHz
120kHz
140kHz
160kHz
42.2k
47.5k
40.2k
42.2k
47.5k
AN67 F46
Figure 46. Single Supply Bandpass Filter
AN67-37
Application Note 67
The passband frequency range (the frequency range where
the filter’s attenuation is 3dB or less) is equal to the center
frequency divided by ten. The stopband attenuation reaches
60dB at twice the center frequency and at one-half the
center frequency. The typical gain variation at the center
frequency is ±0.5dB at 25°C and ±1.5dB over temperature. (Note that an additional ±0.4dB should be added to
account for the gain variation due to the 1% resistors). If
the operating temperature range is 25°C (±20°C) and the
power supply voltage can be controlled to ±2%, the center
frequency can be extended to 90kHz for a 5V supply or
160kHz for a 12V supply. Note that the gain error for center
frequencies greater than 70kHz with a 5V supply and
greater than 100kHz with a 12V supply increases from 1dB
to 7dB. Therefore, the value of resistor R1 for each
LTC1264 section should be increased to reduce the error
to ±1dB (see the table in Figure 46).
If the power supply for this filter is a switching regulator,
the regulator’s output noise can appear at the filter’s
output if the center frequency of the filter is tuned to the
noise frequency of the regulator. This is due to the filter’s
low power supply rejection near its center frequency. The
LTC1264 is not a low power device. The typical quiescent
current is 11mA with a 5V supply or 18mA with a 12V
supply.
–40
10
VS = 5V
fCLK = 1MHz
VS = 5V
fCLK = 1MHz
0
–10
THD + NOISE (dB)
–50
GAIN (dB)
– 20
– 30
– 40
– 50
–60
–70
– 60
–70
– 80
100
10
200
FREQUENCY (kHz)
–80
0.01
0.1
INPUT (VRMS)
1
AN67 F48
AN67 F47
Figure 47. LTC1264 Single 5V Supply,
50kHz Bandpass Response
AN67-38
Figure 48. Dynamic Range vs Input Signal.
LTC1264 Single 5V Supply, 50kHz Bandpass Filter
Application Note 67
Bandpass filters with linear passband phase are useful
for a variety of data-communications tasks, the most
noteworthy of which may be in modulation-demodulation (modem) circuitry. Modems generate signals that
must be processed without phase distortion to allow
error free transmission and reception of information (or
the closest approach to that ideal we can achieve).
An interesting feature of linear-phase bandpass filters is
that their response to a step input produces a short
transient sine wave burst with a symmetrical envelope.
Figure 52 shows a comparison of the transient responses
to a step input for the linear-phase bandpass filter of
10
0
– 10
– 20
GAIN (dB)
A LINEAR-PHASE BANDPASS FILTER
FOR DIGITAL COMMUNICATIONS
By Philip Karantzalis
Figure 49 shows a linear-phase bandpass filter using the
LTC1264 high frequency, universal switched capacitor
filter building block. This filter is an 8th order narrow
bandpass filter, centered at 50kHz for a 1MHz clock input,
with flat group delay in its passband. The fCLK-to-fCENTER
frequency ratio is 20:1. Figure 50 shows the filter’s
narrowband gain response and Figure 51 shows the
passband group delay.
– 30
– 40
– 50
– 60
– 70
– 80
– 90
10
100
200
FREQUENCY (kHz)
AN67 F50
Figure 50. Filter Gain vs Frequency
RlB, 30.1k
RhB, 107k
R1B, 60.4k
1
VIN
R2B, 10.7k
R3B, 29.4k
R4B, 13.3k
2
3
4
5
6
7
7.5V
0.1µF
8
R4A, 11.5k
9
R3A, 12.4k
R2A, 10.7k
10
11
12
INV B
INV C
HP B
HP C
BP B
BP C
LP B
LP C
SB
LTC1264
V–
GND
V+
fCLK
SA
SHDN
LP A
LP D
BP A
BP D
HP A
HP A
INV A
INV D
RhD, 27.4k
R1A, 16.2k
SC
24
23
22
21
R2C, 12.4k
R3C, 37.4k
R4C, 10.7k
CC
20
19
0.1µF
18
–7.5V
1MHz
17
16
15
14
R4D, 10.7k
CD
R3D, 29.4k
R2D, 10k
VOUT
13
RlD, 100k
AN67 F49
Figure 49. LTC1264 Linear Phase, 8th Order Bandpass Filter
AN67-39
Application Note 67
Figure 49 and a bandpass filter with a similar passband
and nonlinear phase response. The response of a bandpass
filter to a step input is a simple qualitative test for determining the linearity of its phase response, although in data
transmission systems the measurements are usually made
with eye diagrams and constellation displays.
100
95
90
DELAY (µs)
85
80
75
70
65
60
55
50
46
47
48
49 50 51 52
FREQUENCY (kHz)
53
54
AN67 F51
Figure 51. Filter Group Delay vs Frequency
NONLINEAR PHASE
FILTER
Capacitor C, across R4 in sections C and D, minimizes
gain and phase variations when the filter is used with
clock frequencies greater than 1.4MHz. For ±5V supplies
the maximum clock frequency is 1.6MHz. Use the Table
1 as a guide for the selection of capacitor C.
Table 1. Capacitor Selection Guide
LINEAR PHASE
FILTER
AN67 F52
Figure 52. Step Response
AN67-40
The maximum clock frequency for the filter is 2MHz with
±7.5V supplies. This allows bandpass filters with center
frequencies up to 100kHz to be realized without significant phase distortion in the passband.
VS
fCLK
CC = CD
±7.5V
1.8MHz
2.0MHz
3pF
5pF
±5V
1.6MHz
1.4MHz
5pF
3pF
Application Note 67
Instrumentation
40MHz bandwidth, this didn’t sound like it would have a
good chance of working first time — built by hand and
without benefit of a custom casting.
WIDEBAND RMS NOISE METER
by Mitchell Lee
Rather than build a circuit with 40MHz bandwidth and a
gain of 100dB, I decided to use just enough gain to put my
desired noise performance around twice minimum scale.
Aside from gain, this amplifier would also need less than
5nV/√Hz input noise, and the output stage would have to
drive the 50Ω load presented by the LT1088.
Recently, I needed to measure and optimize the wideband
RMS noise of a power supply over about a 40MHz bandwidth. A quick calculation showed that the 12nV to 15nV/
√Hz noise floor of my spectrum analyzer would come up
short—my circuit was predicted to exhibit a spot noise of
perhaps 8nV to 10nV/√Hz. In fact, I didn’t have a single
instrument in my lab that would measure 50µVRMS to
60µVRMS.
It wasn’t hard to find an appropriate output stage. The
LT1206 (see Figure 53) can easily drive the required
120mA peak current into the LT1088 converter and
there’s plenty left over for handling noise spikes. To
preserve 40MHz bandwidth, the LT1206 was set to run at
a gain of 2.
For the 40MHz bandwidth, the HP3403C RMS voltmeter is
a good choice but its most sensitive range is 100mV, about
66dB shy of my requirement. This obsolete instrument
today carries a hefty price on the used market. The fact that
here in the Silicon Valley HP3403Cs are a common sight
at flea markets is of little consolation to most customers
wishing to reproduce my measurements. We have several
of these meters in the LTC design lab but they are in
constant use and closely guarded by “The Keepers of the
Secret RMS Knowledge.” I resolved to build my own meter
using an LT1088 thermal RMS converter.
The front end was harder to solve. I needed a low noise,
high speed amplifier that could give me plenty of gain.
Here I selected the LT1226. This is a 1GHz GBW op amp
with only 2.6nV/√Hz input noise. It has a minimum stable
gain of 25 but in this circuit high gain is an advantage.
Cascading two LT1226s on the front end gives a gain of
625, a little shy of the 5,000 to 10,000 required. Another
gain of 5, plus the gain of 2 in the LT1206 adds up to a gain
of 6,250 — just about right.
Full scale on the LT1088 is 4.25VRMS. To measure 50µV
full scale, I’d need an amplifier with a gain of 100,000. At
5V
10nF
10nF
5V
3
SIGNAL
INPUT
+
100nF
–
10nF
10nF
7
6
LT1226
2
5V
10nF
10nF
3
+
100nF
4
7
5
6
LT1226
2
10nF
–
3
100nF
4
7
–
4
10nF
1200Ω
–5V
TO
LT1206
6
LT1192
2
10nF
1200Ω
+
1200Ω
–5V
–5V
1000Ω
100Ω
100Ω
1000Ω
100Ω
100Ω
1000Ω
300Ω
AN67 F53
5V
–5V
100nF
100nF
Figure 53 Noise Meter Gain Stage
AN67-41
Application Note 67
10nF
There are several ways to get 40MHz bandwidth at a gain
of 5, including the LT1223 and LT1227 current feedback
amplifiers, but I settled on the LT1192 voltage amplifier
because it is the lowest cost solution. This brings the gain
up to 6250, for a minimum scale sensitivity of 34µVRMS
and a full-scale sensitivity of 680µVRMS.
15V
10nF
100nF
24k
SHUTDOWN
CONTROL
100nF
FROM
LT1192
2
4
+
3
7
LT1206
1000Ω
1
My advice-filled coworkers assured me that there was no
way I could build a wideband amplifier with a gain of 6250
and make it stable. Nevertheless, I built my amplifier on a
1.5" × 6" copperclad board, taking care to maintain a linear
layout. The finished circuit was stable provided that a
coaxial connection was made to the input. The amplifier
was flat with 3dB points at 4kHz and 43MHz and some
peaking at high frequencies.
TO LT1088
5 10nF
–
6
10nF
100nF
620Ω
–15V
620Ω
AN67 F54
Figure 54. LT1206 Buffer/Driver Section
3.3nF
15V
500Ω
27.4k
1%
2
12
6
1k
1%
10nF
10nF
FROM
LT1206
1k
1%
27.4k
1%
5
9
5
3
1
14
SUB
6
13
2
7
8
–15V
7
1k
2N2219
+
+
U4
1/4
LT1014
1
–
5V
33k
4
2N3904
U1
1/4
LT1014
15V
SHUTDOWN CONTROL
30k
–
15V
9.09M
1%
LT1088
SUB
9.09M
1%
+
8
12
10
10k
U3
1/4
LT1014 9
22nF
–
12k
13
+
U2
1/4
LT1014
–
14
OUTPUT
10k
11
10k
–5V
5V
100nF
100nF
AN67 F55
Figure 55. LT1088 RMS Detector Section
AN67-42
Application Note 67
Coaxial Measurements
When measuring low level signals it is difficult to get a
clean, accurate result. Scope probes have two problems. 10 × probes attenuate the already small signal,
and both 1× and 10 × probes suffer from circuitous
grounds. Coaxial adapters are a partial solution but
these are expensive. They make for a lot of wear and tear
on the probes and, without a little forethought, they can
be a bear to attach to the circuit under test. My favorite
way to get clean measurements of small signals is to
directly attach a short length of coaxial cable as shown
in Figure 56.
I use the good part of a damaged BNC cable, cutting away
the shorter portion to leave at least 18" of RG-58/U and
one good connector. At the cut, or as I call it, “real world”
end of the cable, I unbraid, twist and tin a very small
amount of outer conductor to form a stub 1/4" to 3/8"
long. Next, I cut away the dielectric, exposing a similar
length of center conductor, which I also tin. Now the
probe is ready for use. It can be soldered directly to a
circuit or breadboard, eliminating any lead length that
might otherwise pick up stray noise, or worse, act as an
antenna in a sensitive high gain circuit.
Small signals aren’t the only beneficiaries of this technique. This works great for looking at ripple on the
outputs of switching supplies. Ripple measurements are
simplified because the large voltage swings associated
with the switch node are completely isolated and no loop
is formed where di/dt could inject magnetically coupled
noise.
In some instances, I’ve found it important to retain the
50Ω termination impedance on the cable, but it is rarely
possible to place a terminator at the “BNC” end of the
cable, since this creates a DC path directly across the
circuit under test. There is, however, another way as
shown in Figure 57. Here, a technique known as back
termination is used. No termination is used at the far end
of the cable, but a 51Ω resistor is connected in series
with the measurement end. Signals sent down the cable
reach the BNC connector without attenuation, and fast
edges that bounce off the unterminated end are absorbed
back into the 51Ω source resistor. I’ve found this especially useful for measuring fast switch signals, or when
measuring the RMS value of small signals, for ensuring
that the amplifier input sees a properly terminated source.
The resistor trick does not work if the node under test is
high impedance; a FET probe is a better choice for high
impedance measurements.
While I’m at it, I might as well give away my only other
secret. We’ve all encountered ground loop problems,
giving rise to 60Hz (50Hz for my friends overseas)
injection into sensitive circuits. Every lab is replete with
isolation transformers and “controlled substance” line
cords with missing ground prongs for battling ground
loops. There are similar problems at high frequencies,
but the victim is wave fidelity, not AC pick up. To
determine whether or not high frequency grounding,
ground loops or common mode rejection is a problem for
your oscilloscope, simply clamp a small ferrite E core
around the probe lead while observing any effects on the
waveform (see Figure 58). Sometimes the news is bad;
the waveform really is messed up and there is some work
to be done on the circuit. But occasionally the circuit is
exonerated, the unexplainable aberrations disappear,
proving that high frequency gremlins are at work. If
necessary, several passes of the probe cable can be
made through the E-Core and it can be taped together for
as long as needed.
51Ω
AN67 F56
Figure 56. BNC Cable Used as a “Probe”
AN67 F57
Figure 57. BNC Cable Probe Back Terminated
AN67-43
Application Note 67
AN67 F58
Figure 58. An E-Core Serves to Attenuate High Frequency Common Mode Currents from a Scope Probe
The performance of the amplifier and thermal converter
can be optimized by adjusting the value of the feedback
and gain setting resistors around the LT1206. Slightly
more bandwidth can be achieved at the expense of higher
peaking by reducing the resistor values 10%. Reducing
the resistor values will decrease peaking effects at the
expense of bandwidth. A good compromise value is 680Ω.
I’ve shown the LT1226 amplifiers operating from ±5
supplies, which puts their bandwidth on the edge at
40MHz. Their bandwidth can be improved by operating at
±15V.
AN67-44
Because the LT1206 operates on 15V rails, it is possible to
overdrive the LT1088 and possibly cause permanent damage. One section of the LT1014 (U3) is used to sense an
overdrive condition on the LT1088 and shut down the
LT1206. Sensing the feedback heater instead of the input
heater allows the LT1088 to accommodate high crest
factor waveforms, shutting down only when the average
input exceeds maximum ratings.
By the way, my power supply noise measured 200µV;
filtering brought it down to less than 60µV.
Application Note 67
LTC1392 MICROPOWER TEMPERATURE AND
VOLTAGE MEASUREMENT SENSOR
by Ricky Chow and Dave Dwelley
The LTC1392 is a micropower data acquisition system
designed to measure temperature, on-chip supply voltage
and differential rail-to-rail common mode voltage. The
device incorporates a temperature sensor, a 10-bit A/D
converter, a high accuracy bandgap reference and a 3-wire
half-duplex serial interface.
Figure 59 shows a typical LTC1392 application. A single
point “star” ground is used along with a ground plane to
minimize errors in the voltage measurements. The power
supply is bypassed directly to the ground plane with a 1µF
tantalum capacitor in parallel with an 0.1µF ceramic capacitor.
The conversion time is set by the frequency of the signal
applied to the CLK pin. The conversion starts when the CS
pin goes low. The falling edge of CS signals the LTC1392
to wake up from micropower shutdown mode. After the
LTC1392 recognizes the wake-up signal, it requires an
additional 80µs delay for a temperature measurement, or
a 10µs delay for a voltage measurement, followed by a
4-bit configuration word shifted into DIN pin. This word
configures the LTC1392 for the selected measurement
P1.4
DIN
VCC
DOUT
–VIN
and initiates the A/D conversion cycle. The DIN pin is then
disabled and the DOUT pin switches from three-state mode
to an active output. A null bit is then shifted out of the DOUT
pin on the falling edge of the CLK, followed by the result of
the selected conversion. The output data can be formatted
as an MSB-first sequence or as an MSB-first followed by
an LSB-first sequence, providing easy interface to either
LSB-first or MSB-first serial ports. The minimum conversion time for the LTC1392 is 142µs in temperature mode
or 72µs in the voltage conversion modes, both at the
maximum clock frequency of 250kHz.
Conclusion
The LTC1392 provides a versatile data acquisition and
environmental monitoring system with an easy-to-use
interface. Its low supply current, coupled with space
saving SO-8 or PDIP packaging, makes the LTC1392 ideal
for systems that require temperature, voltage and current
measurement while minimizing space, power consumption and external component count. The combination of
temperature and voltage measurement capability on one
chip makes the LTC1392 unique in the market, providing
the smallest, lowest power multifunction data acquisition
system available.
RSENSE
MPU
(e.g. 68HC11)
LTC1392
1µF
P1.3
CLK
+VIN
P1.2
CS
GND
0.1µF
5V
ILOAD
AN67 F59
Figure 59. Typical LTC1392 Application
AN67-45
Application Note 67
HUMIDITY SENSOR TO
DATA ACQUISITION SYSTEM INTERFACE
by Richard Markell
Introduction
It can be difficult to interface humidity sensors to data
acquisition systems because of the sensors’ drive requirements and their wide dynamic range. By carefully selecting
the devices that comprise the analog front end, users can
customize the circuit to meet their humidity sensing requirements and achieve reasonable accuracy throughout
the chosen range. This article details the analog front end
interface between a Phys-Chem Scientific Corp.1 model
EMD-2000 humidity sensor and a user selected (probably
microprocessor-based) data acquisition system.
Design Considerations
The Phys-Chem humidity sensor is a small, low cost,
accurate resistance-type relative humidity (RH) sensor.
This sensor has a well-defined stable response curve and
can be replaced in circuit without system recalibration.
The design criteria call for a low cost, high precision
analog front end that requires few calibration “tweaks” and
operates on a single 5V supply. The sensor requires a
square wave or sine wave excitation with no DC component. The sensor reactance varies over an extremely wide
range (approximately 700Ω to 20MΩ). The wide dynamic
range (approximately 90dB) required to obtain the full RH
range of the sensor results in some challenges for the
designer.
The circuit shown in the schematic features zero drift
operational amplifiers (LTC1250 and LTC1050) and a
precision instrumentation switched capacitor block
(LTC1043). This design will maintain excellent DC accuracy down to microvolt levels. This method was chosen
over the use of a true RMS-to-DC or log converter because
of the expense and temperature sensitivity of these parts.
Circuit Description
Figure 60 is a schematic diagram of the circuit. Only a
single 5V power supply is required. Integrated circuit U1,
an LTC1046, converts the 5V supply to – 5V to supply
power to U2, U3 and U4. U2A, part of an LTC1043
switched capacitor building block, provides the excitation
for the sensor, switching between 5V and – 5V at a rate of
approximately 2.2kHz. This rate can be varied, but we
recommended that it be kept below approximately 2.4kHz,
which is one-half the auto zero rate of U3. We believe the
5V
1
2
+
C1
10µF
BOOST
V+
8
7
OSC
U1
LTC1046
3
6
GND
LV
4
C3
0.1µF
C+
C–
VO
5
C8
0.01µF
+
R4
10k
R3
10k
C2
10µF
C5
62pF
11
C4
1µF
4
2
HUMIDITY
SENSOR
2
R1
3
CALIBRATION
–
U3
LTC1250
+
8
7
6
13
7
U2
LTC1043
17
C9
2000pF*
3
C8
1µF
12
4
R2
1k
14
16
7
U4
LTC1050
6
VOUT
+
4
C7
1µF
NOTES: UNLESS OTHERWISE SPECIFIED
1. ALL RESISTANCES ARE IN OHMS, 1/4 W 5%
*C9 ADJUSTS OSC. FREQUENCY 2000pF YIELDS ~ 2.2kHz
Figure 60. Schematic Diagram of Humidity Sensor Circuit
AN67-46
–
AN67 F60
Application Note 67
deviation from the Phys-Chem response curves taken at
5kHz is insignificant.
converter that changes ranges somewhere in the humidity range.
Variable resistor R2 sets the full-scale output. Since the
sensor resistance is 700Ω at approximately 90% humidity, setting R2 at 700Ω will provide a 2:1 voltage divider
that, when combined with the gain of U4 (×2), results in an
overall gain of one. U3 must be included in order for the
circuit to function properly; otherwise C4 and C7 form a
voltage divider that is dependent on the resistance of the
RH sensor. U3 is a precision auto zero operational amplifier with an auto zero frequency of approximately 4.75kHz.
U2B (the “lower” switch) samples the output of U3 and
provides this sample to the input of U4. U4 is set to provide
a gain of 2.
All of the above solutions measure output voltage from a
voltage divider consisting of the RH sensor and a fixed
“calibration” resistor. The resistance of the sensor at a
fixed output voltage can be calculated from the formula
R (Ω) =
R2 VFULL SCALE
– R2
VOUT/2
In this case, if R2 is set to 700Ω and VFULL SCALE = 5.00V,
then
R (Ω) =
It is easy to digitize the output of U4. Figure 61 is the
schematic of a 12-bit converter that can be used for this
purpose. The range of humidity that can be sensed
depends on the resolution of the converter. The full-scale
output (which is equivalent to approximately 90% humidity) is essentially independent of the number of bits
in the A/D converter, but the dry (low RH) end of the scale
is dependent on the A/D resolution. As an example, the
above referenced 12-bit converter will process humidity
signals that translate to approximately 20% RH, since the
voltage output at this humidity is approximately 2.3mV,
while 0.5LSB is 1.2mV. Digitization down to 10% RH
requires the conversion of 350µV signals or a 16-bit
converter. From a cost standpoint this seems unwieldy.
It is much more economical to use a 2-channel 12-bit
3500
– 700
VOUT/2
Once R is calculated (probably by the microprocessor),
the humidity can be calculated from the quadratic approximation in the Phys-Chem literature:
RH = LnR – 13.95 – √(13.95 – LnR)2 + 24.288
–0.184
If a suitable humidity chamber is not available, the sensor
can be removed and fixed resistors substituted. The circuit
should then be calibrated from the EMD-2000 “typical
response curve.” This should provide approximately 2%
accuracy.
1Phys-Chem Scientific Corporation, 26 West 20th Street, New York, NY 10011.
(212) 924-2070 Phone, (212) 243-7352 FAX
22µF
TANTALUM
5V
+
1
FROM VOUT
LTC1050
2
2-CHANNEL
MUX*
CS
VCC (VREF)
CH0
CLK
8
7
DO
0.1µF
SCK
LTC1291
3
4
CH1
DOUT
GND
DIN
6
5
MC68HC11
MISO
MOSI
AN67 F61
*FOR OVERVOLTAGE PROTECTION LIMIT THE INPUT CURRENT TO 15mA
PER PIN OR CLAMP THE INPUTS TO VCC AND GND WITH 1N4148 DIODES.
Figure 61. LTC1291 12-Bit A/D Converter Interfaced to MC68HC11
AN67-47
Application Note 67
L1 Pins 1 and 2 source a boosted, fully floating voltage,
which is rectified and filtered. This potential powers A2.
Because A2 floats with respect to T1, it can look differentially across T1’s outputs, Pins 10 and 4. In practice, Pin
10 becomes “ground” and A2 measures Pin 4’s output
with respect to this point. A2’s gain scaled output is the
circuit’s output, conveniently scaled at 3.000V = 30.00"Hg.
A SINGLE CELL BAROMETER
by Jim Williams and Steve Pietkiewicz
Figure 62, a complete barometric pressure signal conditioner, operates from a single 1.5V battery. Until recently,
high accuracy and stability have been obtainable only
with bonded strain gauge and capacitively-based transducers, which are quite expensive. This design, using a
recently introduced semiconductor transducer, achieves
0.01"Hg (inches of mercury) uncertainty over time and
temperature. The 1.5V powered operation permits portable application.
To calibrate the circuit, adjust R1 for 150mV across the
100Ω resistor in T1’s return path. This sets T1’s current to
the manufacturer’s specified calibration point. Next, adjust R2 at a scale factor of 3.000V = 30.00"Hg. If R2 cannot
capture the calibration, reselect the 200k resistor in series
with it. If a pressure standard is not available, the transducer is supplied with individual calibration data permitting circuit calibration.
The 6kΩ transducer (T1) requires precisely 1.5mA of
excitation, necessitating a relatively high voltage drive.
A1’s positive input senses T1’s current by monitoring the
voltage drop across the resistor string in T1’s return path.
A1’s negative input is fixed by the 1.2V LT1004 reference.
A1’s output biases the 1.5V powered LT1110 switching
regulator. The LT1110’s switching produces two outputs
from L1. Pin 4’s rectified and filtered output powers A1 and
T1. A1’s output, in turn, closes a feedback loop at the
regulator. This loop generates whatever voltage step-up is
required to force precisely 1.5mA through T1. This arrangement provides the required high voltage drive while
minimizing power consumption. This occurs because the
switching regulator produces only enough voltage to
satisfy T1’s current requirements.
68k
5
T1
10
4
+
OUTPUT
A2
LT1077
–
0.1µF
1µF
NONPOLAR
NOVASENSOR
NPH-8-100AH†
This circuit, compared to a high order pressure standard,
maintained 0.01"Hg accuracy over months with widely
varying ambient pressure shifts. Changes in pressure,
particularly rapid ones, correlated quite nicely to changing
weather conditions. Additionally, because 0.01"Hg corresponds to about 10 feet of altitude at sea level, driving over
hills and freeway overpasses becomes quite interesting.
The circuit pulls 14mA from the battery, allowing about
250 hours operation from one D cell.
200k 10k
1%* 1%
R2
1k
6
100Ω
+
0.1µF
100k
2
+
A1
LT1077
–
698Ω
1%
100k
R1
50Ω**
100k
100Ω
1%
1N4148
2
1µF
8
6
AA CELL
LT1004-1.2
* NOMINAL VALUE EACH SENSOR REQUIRES
** TRIM FOR 150mV ACROSS A1-A2
LT1110
AO
GND
5
SW1
1N5818
+
SET
SW2
4
3
430k
7
† LUCAS NOVASENSOR
FREMONT, CA (510) 490-9100
COILTRONICS (407) 241-7876
Figure 62. Single Cell Barometer Schematic Diagram
AN67-48
100Ω
LL
VIN
FB
1N4148
1
3
4
L1
COILTRONICS
1 CTX50-1
150Ω
+
100µF
39k
AN67 F62
390µF
16V
NICHICON
PL
Application Note 67
NOISE GENERATORS FOR MULTIPLE USES
A Broadband Random Noise Generator
by Jim Williams
Filter, audio and RF communications testing often require
a random noise source. Figure 63’s circuit provides an
RMS amplitude regulated noise source with selectable
bandwidth. RMS output is 300mV with a 1kHz to 5MHz
bandwidth, selectable in decade ranges.
rent feedback amplifier. A4’s output, which is also the
circuit’s output, is sampled by the A5-based gain control
configuration. This closes a gain control loop to A3. A3’s
ISET current controls gain, allowing overall output level
control.
Noise source D1 is AC coupled to A2, which provides a
broadband gain of 100. A2’s output feeds a gain control
stage via a simple, selectable lowpass filter. The filter’s
output is applied to A3, an LT1228 operational transconductance amplifier. A1’s output feeds LT1228 A4, a cur-
Figure 64 plots noise at a 1MHz bandpass, whereas Figure
65 shows RMS noise versus frequency in the same
bandpass. Figure 66 plots similar information at full bandwidth (5MHz). RMS output is essentially flat to 1.5MHz,
with about ±2dB control to 5MHz before sagging badly.
0.1(1kHz)
1µF
16k
+
15V
D1
NC201
1k
0.01(10kHz)
1.6k
A2
LT1226
–
0.001(100kHz)
1k
1k
100pF(1MHz)
10Ω
NC
(5MHz)
15V
+
A3
LT1228
gm
+
–
A4
LT1228
CFA
910Ω
3k
–
0.1
510Ω
1µF
NONPOLAR
–15V
15V
10Ω
–
+
A5
LT1006
+
–15V
22µF
22µF
10k
– – +
0.5µF
10k
1M
4.7k
–15V
NC 201 = NOISE COM CORP.
NOISE COM = (201) 261-8797
1N4148
THERMALLY
COUPLED
LT1004
1.2V
AN67 F63
Figure 63. Broadband Random Noise Generator Schematic
AN67-49
Application Note 67
1V/DIV
AN67 F64
10µs/DIV
Figure 64. Figure 63’s Output In the 1MHz Filter Position
12
AMPLITUDE VARIANCE (dB)
9
6
3
0
–3
–6
–9
–12
–15
–18
0
0.1
0.2
0.3
0.4
0.5
0.6
FREQUENCY (MHz)
0.7
0.8
0.9
1.0
AN67 F65
Figure 65. RMS Noise vs Frequency at 1MHz Bandpass
9
6
AMPLITUDE VARIANCE (dB)
3
0
–3
–6
–9
–12
–15
–18
–21
0
1
2
3
4
5
6
FREQUENCY (MHz)
7
8
Figure 66. RMS Noise vs Frequency at 5MHz Bandpass
AN67-50
9
10
AN67 F66
Application Note 67
SYMMETRICAL WHITE GAUSSIAN NOISE
by Bent Hessen-Schmidt, NOISE COM, INC.
high level of symmetrical white Gaussian noise. The level
is often specified in terms of excess noise ratio (ENR).
White noise provides instantaneous coverage of all frequencies within a band of interest with a very flat output
spectrum. This makes it useful both as a broadband
stimulus and as a power level reference.
Symmetrical white Gaussian noise is naturally generated
in resistors. The noise in resistors is due to vibrations of
the conducting electrons and holes as described by
Johnson and Nyquist.1, 2 The distribution of the noise
voltage is symmetrically Gaussian, and the average
noise voltage is:
Vn = 2√kTƒ R(f) p(f) df
(1)
where:
k = 1.38E-23 J/K (Boltzmann’s constant)
T = temperature of the resistor in Kelvin
f = frequency in Hz
h = 6.62E-34 Js (Planck’s constant)
R(f) = resistance in ohms as a function of frequency
p(f) =
hf
kT[exp(hf/kT) – 1]
(2)
p(f) is close to unity for frequencies below 40GHz when
T is equal to 290°K. The resistance is often assumed to
be independent of frequency, and ƒdf is equal to the
noise bandwidth (B). The available noise power is obtained when the load is a conjugate match to the resistor,
and it is:
V 2
N = n = kTB
4R
(3)
where the “4” results from the fact that only half of the
noise voltage and hence only 1/4 of the noise power is
delivered to a matched load.
Equation 3 shows that the available noise power is
proportional to the temperature of the resistor; thus it is
often called thermal noise power. Equation 3 also shows
that white noise power is proportional to the bandwidth.
An important source of symmetrical white Gaussian
noise is the noise diode. A good noise diode generates a
ENR (in dB) = 10Log
(Te – 290)
290
(4)
Te is the physical temperature that a load (with the same
impedance as the noise diode) must be at to generate the
same amount of noise.
The ENR expresses how many times the effective noise
power delivered to a nonemitting, nonreflecting load
exceeds the noise power available from a load held at the
reference temperature of 290°K (16.8°C or 62.3°F).
The importance of a high ENR becomes obvious when
the noise is amplified, because the noise contributions
of the amplifier may be disregarded when the ENR is
17dB larger than the noise figure of the amplifier (the
difference in total noise power is then less than 0.1dB).
The ENR can easily be converted to noise spectral
density in dBm/Hz or µV/√Hz by use of the white noise
conversion formulas in Table 1.
Table 1. Useful White Noise Conversion
dBm = dBm/Hz + 10log(BW)
dBm = 20log(Vn) – 10log(R) + 30dB
dBm = 20log(Vn) + 13dB, for R = 50Ω
dBm/Hz = 20log(µVn√Hz – 10log(R) – 90dB
dBm/Hz = –174dBm/Hz + ENR, for ENR > 17dB
When amplifying noise it is important to remember that
the noise voltage has a Gaussian distribution. The peak
voltages of noise are therefore much larger than the
average or RMS voltage. The ratio of peak voltage to
RMS voltage is called crest factor, and a good crest
factor for Gaussian noise is between 5:1 and 10:1 (14dB
to 20dB). An amplifier’s 1dB gain compression point
should therefore be typically 20dB larger than the desired average noise output power to avoid clipping of the
noise.
For more information about noise diodes, please contact
NOISE COM, INC. at (201) 261-8797.
1 Johnson, J.B. “Thermal Agitation of Electricity in Conductors,”
Physical Review,
July 1928, pp. 97-109.
2 Nyquist, H. “Thermal Agitation of Electric Charge in Conductors,” Physical Review,
July 1928, pp. 110-113.
AN67-51
Application Note 67
NOISE GENERATORS FOR MULITPLE USES
A Diode Noise Generator for “Eye Diagram” Testing
by Richard Markell
sensing comparator. The 2kΩ pot at the inverting input of
the LT1116 sets the threshold to the comparator so that a
quasiequal number of 1’s and 0’s are output. U3 latches
the output from U2 so that the output from the comparator
remains latched throughout one clock period. The twolevel output is taken from U3’s Q0 output.
The circuit that Jim Williams describes evolved from my
desire to build a circuit for testing communications channels by means of “eye diagrams.” (See Linear Technology,
Volume I, Number 2 for a short explanation of the eye
diagram.) I wanted to replace my pseudorandom code
generator circuit, which used a PROM, with a more “analog” design—one that more people could build without
specialized components. What evolved was a noise source
sampled by a very fast comparator (see Figure 67). The
comparator outputs a random pattern of 1’s and 0’s.
The additional circuitry shown in the schematic diagram
allows the circuit to output four-level data for PAM (pulse
amplitude modulation) testing. The random data from the
two-level output is input to a shift register, which is reset
on every fourth clock pulse. The output from the shift
register is weighted by the three 5k resistors and summed
into the LT1220 operational amplifier from which the
output is taken. The filter network between the 74HC74
output and the 74HC4094 strobe input is necessary to
ensure that the output data is correct.
The noise diode (an NC201) is filtered and amplified by the
LT1190 high speed operational amplifier (U1). The output
feeds the LT1116 (U2), a 12ns single supply, ground10k
10pF
5V
100k
5V
1µF
3
NC201
NOISE DIODE
1M 2
–5V
+
U1
LT1190
–
5V
7
6
1k
3
4
2
100k
–5V
1k
1µF
TANT.
+
+
8
3
U2
LT1116
–
4
6
20
VCC
2
DO
5 7
LE
U3
74HC373
11
5V
LE
–5V
2 LEVEL
OUTPUT
Q0
10 1
100k
+
5V
–5V
0E
U5
74HC04
5
6
2k
10µF
CLOCK
1
4
2 1RD
1SD 5
1D
1Q
3
6
1CLK
1Q
1.7k
5V
2 D
3 CP
8-STAGE SHIFT
REGISTER
8V
QS1 9
2
1 STR
8-BIT STORAGE
REGISTER
15 OE
3-STATE OUTPUTS
–
7
LT1220
3
5V
13
10
12 2RD
2SD 9
QS2 10
74HC4094
VCC = PIN 16
GND = PIN 8
+
8
2D
2Q
11
8
2CLK
4
4 LEVEL
OUTPUT
– 8V
100pF
1k
2Q
74HC74
VCC = PIN 14
GND = PIN 7
QP0 QP1 QP2 QP3 QP4 QP5 QP6 QP7
4
5
6
7
14 13 12 11
5k × 3
NC 201 = NOISE COM DIODE (201) 261-8797
Figure 67. Pseudorandom Code Generator Schematic Diagram
AN67-52
AN67 F67
Application Note 67
Video/Op Amps
op amp, and its associated components form an elementary sync separator. C1, R1 and D1 clamp the composite
video. D2 biases the input of U1 to compensate for the
drop across D1. When D1 conducts, the most negative
portion of the waveform containing the sync information
is amplified by U1. The clamp circuit in the feedback
network of U1 (D4 to D8) prevents the amplifier from
saturating. D3 and the CMOS inverter U4 complete the
shaping of the sync waveform. This sync separator works
with most video signals but, because of its simplicity, will
not work with very noisy or distorted video. The remainder
of the circuit is an LT1251 video fader (U2) configured to
fade between the original video and the sync stripped from
that video. Thus, the video fades to black.
LT1251 CIRCUIT SMOOTHLY FADES VIDEO TO BLACK
by Frank Cox
When a video signal is attenuated, there is a point were the
sync amplitude is too small for a monitor to process
properly. Instead of making a smooth transition to black,
the picture rolls and tears. One solution to this problem is
to run a separate sync signal into the monitor. This may
not be a viable solution in a system where cost and
complexity are the prime concerns. What is needed is a
simple video “volume control.”
The circuit in Figure 68 can perform a smooth fade to
black, while maintaining good video fidelity. U1, an LT1360
15V
5k
D4
D7
D8
D5
D6
5k
C1
10µF
VIDEO IN
D1
1N5711
R1
10k
75Ω
D8 = 1N5226
D4 TO D7 = 1N4148
–15V
10k
U4A
74HC14
D3
1N4148
–
U1
LT1360
30k
+
15V
1500Ω
1500Ω
2
1
U2
LT1251
–
+
8
10k
COMPOSITE
SYNC
14
+
VFS
200Ω
13
5k
–
VC
RC
75Ω
VIDEO OUT
10k
12
15V
3
10k
RFS
5
SYNC LEVEL
COMPOSITE SYNC
500Ω
D2
1N5711
U3
LT1004-2.5
10
FADE CONTROL
1500Ω
AN67 F68
Figure 68. LT1251 Video Fader Schematic Diagram
AN67-53
Application Note 67
The control voltage for the fader is generated by a voltage
reference and a 10k variable resistor. If this control potentiometer is mounted an appreciable distance from the
circuit or if the control generates any noise when adjusted,
this node should be bypassed.
Figure 69 is a multiple exposure waveform photograph
that shows the action of this circuit. Two linear ramp video
test signals are shown in this photograph. The video is
faded from full amplitude to zero amplitude in six steps.
The sync waveform (lower center) remains unchanged. In
AN67 F69
Figure 69. Multiple Exposure Photo Showing Circuit Operation
LUMA KEYING WITH THE LT1203
VIDEO MULTIPLEXER
by Frank Cox
In video systems, the action of switching between two or
more active video sources is referred to as a “wipe” or a
“key.” When the decision to switch video sources is based
on an attribute of the active video itself, the action is called
keying. A wipe is controlled by a nonvideo signal such as
a ramp. The circuit presented in Figure 71 is referred to as
a “luma key” because it switches between two sources
when the luminance (“luma”) of a monochrome key signal
reaches a set level. It is also possible to key on the color
of the video source and this, not surprisingly, is called
“chroma keying.”
Figure 71’s operation is very straightforward. A monochrome video source is used to generate the key signal.
The LT1363 is used as a buffer and may not be needed in
all applications. If the key signal is to be used as one of the
switched signals, it is convenient to “loop through” the
input of this buffer. The LT1016 comparator switches
AN67-54
Figure 70, a single video line modulated with color
subcarrier is faded from full video amplitude to zero video
amplitude. The monitor will eventually lose color lock and
shut the color off as the amplitude of the color subcarrier
is reduced. This is not a problem in this application
because the color decoding circuits in the monitor are
designed to work with a variety of signals from tape or
broadcast, and so have a large dynamic range. Color
portions of the picture will remain after the luminance
portion is completely black.
AN67 F70
Figure 70. Photo Detailing a Single Video Line with Color
Subcarrier Faded to Zero Amplitude
when the video level exceeds the DC reference on its
inverting input, which is controlled by the “key sensitivity”
control. The TTL key signal controls an LT1203 video
multiplexer. Any two video sources may be connected to
the inputs of the LT1203, as long as they are gen locked
and within the common mode range (on ±5V supplies this
is ±3V over 0°C to 70°C) of the multiplexer. The LT1203’s
fast switching speed, low offset and clean switching make
it a natural for an active video switching application like
this one. Composite color signals can be used, but the best
results will be obtained if the key signal’s horizontal sync
is phase coherent with the color reference of the sources.
The key source video should be monochrome to prevent
the key comparator from switching on the color subcarrier.
Nonstandard video signals can be used for the inputs to
the LT1203. For instance, it is possible to select between
two DC input levels to construct a two-level image. Figure
72 is an example of an image constructed this way. A
monochrome video signal is sliced and used to key between black (0V) and gray (approximately 0.5V) to
Application Note 67
generate this image of a famous linear IC designer. An
image formed in this way is not a standard video output
until the blanking and sync intervals are reconstructed.
The second LT1203 blanks the video and an LT1363
circuit sums composite sync to the video and drives a
cable. For more information on this part of the circuit, see
KEY VIDEO
3
AN 57, page 7. A clamp is not used since the DC levels are
arbitrarily set by the inputs, but one could be used, as in the
figure on page 7 of AN57, if the sources were video. As
another option, Figure 73 shows the same key signal used
as one of the inputs to the multiplexer.
5
+
INPUT
6
LT1363
10k
2
2
15V
–
7
LT1016
3
10k
1000Ω
+
5
–
6
4
1000Ω
–5
LT1004 - 2.5
100k
KEY SENSITIVITY
1
SOURCE 1
IN1
75Ω
LOGIC
5
LT1203
2
GRD
OUT
IN2
EW
7
1
75Ω
SOURCE 2
1000Ω
3
6
2
NC
IN1
LOGIC
5
COMPOSITE BLANKING (TTL LEVELS)
LT1203
7
GRD
OUT
3
1000Ω
3
IN2
EW
6
NC
+
LT1363
2
–
6
75Ω
VIDEO
OUT
510Ω
1000Ω
74AC04
300Ω
9.3µH
5k
2.2k
510Ω
COMPOSITE SYNC (TTL LEVELS)
36.5pF
232pF
300Ω
AN67 F71
Figure 71. Luma Keyer Schematic Diagram
Figure 72. Two Level Image of IC Designer
Figure 73. Key Signal Used as Input to the MUX
AN67-55
Application Note 67
LT1251/LT1256 VIDEO FADER AND
DC GAIN CONTROLLED AMPLIFIER
by William H. Gross
A summary of the LT1251/LT1256 performance operating
on ±5V supplies in the configuration shown in Figure 74
is given in Table 1.
The Video Fader
Table 1. LT1251/LT1256 Performance Summary
Figure 74 shows the LT1251/LT1256 configured as a fader
with unity gain. A full-scale voltage of 2.5V is applied to Pin
12 and the control input drives Pin 3.
Figure 75 shows the true response of the control path. The
control path is fast enough for quick switching between
signals, as when keying on a color or luminance level. The
control path introduces only a small (50mV), short (50ns)
glitch when switched quickly.
IN1
1
14
LT1251/LT1256
+
2
–
2
1
+
–
IN2
Slew Rate (at ±2V, RL = 150Ω)
300V/µs
Full-Power Bandwidth (1VRMS)
30MHz
Small-Signal Bandwidth
30MHz
Differential Gain (NTSC, RL = 150Ω)
0.1%
Differential Phase (NTSC, RL = 150Ω)
0.1°
Total Harmonic Distortion (1kHz, K = 1)
(1kHz, K = 0.5)
(1kHz, K = 0.1)
Rise Time, Fall Time
Overshoot
0.001%
0.01%
0.4%
11ns
3%
Propagation Delay
10ns
Settling Time (0.1%, VO = 2V)
65ns
Quiescent Supply Current
13.5mA
13
CONTROL
0V TO 2.5V
CONTROL
IC
3
+
–
4
IC IFS
C
+
FS
–
5
NULL
V–
12
11
2.5V
REF
IFS
10
5k
6
5k
7
9
Applications
RF2
1.5k
RF1
1.5k
V+
8
AN67 F74
VOUT
Figure 74. Two Input Video Fader
10
8
VOLTAGE GAIN (dB)
6
VOLTAGE DRIVE RC
VC = GND
VS = ±5V
By operating one input stage in an inverting configuration
and the other in a noninverting configuration and driving
both inputs, the LT1256 becomes a 4-quadrant multiplier. Figure 77 shows the 4-quadrant multiplier being
used as a double-sideband, suppressed-carrier modulator. The LT1077 DC output nulling circuit could be added
if necessary.
4
2
0
–2
–4
–6
–8
– 10
0.01
0.1
10
1
FREQUENCY (MHz)
100
AN67 F75
Figure 75. LT1251/56 Control Path Bandwidth
AN67-56
Grounding IN2 of the LT1256 in Figure 74 results in a
2-quadrant multiplier. Figure 76 shows the 2-quadrant
multiplier being used as an AM modulator. The output will
deliver 10dBm into 50Ω. The LT1077 op amp senses the
LT1256 output DC and drives the Null pin, eliminating any
DC at the output. The Null pin voltage is nominally 100mV
above the negative supply and therefore the op amp output
must be able to swing within a few millivolts of the negative
supply. Without the LT1077, the worst-case DC output
voltage is 50mV.
The LT1251/LT1256 can be used to implement numerous
other functions, including voltage controlled filters, phase
shifters and oscillators. Squaring and limiting circuits can
be designed by feeding the output or input into the Control
pins. Gamma correction and other compression circuits
are created in a similar manner. The applications are
limited only by the designer’s imagination.
Application Note 67
50Ω
0.1µF
LT1256
1
1MHz
CARRIER
+
2
–
14
2
1
+
–
13
RF2
1.5k
RF1
1.5k
VOUT
CONTROL
0.1µF
3
AUDIO
MODULATION
220k
4
+
–
C
IC
+
FS
–
IFS
5
6
12
2.5V REF
11
10
5k
5k
9
NULL
V–
220k
7
V+
8
200k
0.1µF
V+
–
LT1077
+
AN67 F76
V–
Figure 76. AM Modulator with DC Output Nulling Circuit
LT1256
1
14
+
RG1
1.5k
2
MODULATION
–
+
1
2
–
13
RF2
1.5k
CONTROL
3
1MHz
CARRIER
0.1µF
10k*
4
+
–
C
IC
IFS
+
FS
–
5
6
12
7
10k
VOUT
2.5V REF
11
0.1µF
10
5k
5k
9
*TRIM FOR SYMMETRY
V–
RF1
1.5k
V+
8
AN67 F77
Figure 77. Four Quadrant Multiplier Uses as a Double Sideband, Suppressed Carrier Modulator
AN67-57
Application Note 67
V+
EXTENDING OP AMP SUPPLIES
TO GET MORE OUTPUT VOLTAGE
by Dale Eagar
We often hear of applications that require high output
voltage, low output impedance amplifiers. Here is a topology that allows you to extend an op amp’s output voltage
swing while still maintaining its short-circuit protection.
The trick is to suspend the op amp between two MOSFET
source followers so that the supply voltages track the op
amp’s output voltage (see Figure 78). The circuit shown in
Figure 78 will perform very nicely with any run-of-the-mill
ideal op amp. The problem is in the lead times of ideal op
amps—they just keep getting pushed out to later dates.
R6
100k
R5
100Ω
Q1
IRF610
C2
0.1µF
7
+
R2
R1
6
C1
0.1µF
Z1
15V
CR1
C4
IC1
–
Z2
15V
CR2
4
C3
0.1µF
R3
100Ω
Q2
IRF9610
R4
100k
AN67 F79
Nonideal op amps have realistic lead times and can be
made to work in the extended supply mode. They have
bandwidth limitations in both CMRR and PSRR. The
circuit shown in Figure 79 implements the extended supply as shown in Figure 78 and has several additional
components: C1 is added to decouple the supply, improving high frequency PSRR; R3 and R5 decouple the gates
of Q1 and Q2 from AC ground, preventing Q1 and Q2 from
running off together to redirect local air traffic; R1, R2 and
C4 form a snubber to de-Q the 2-pole system formed by
the Miller capacitance of Q1 and Q2 and the high frequency
CMRR of IC1; additionally, R4, R6, C2, C3, Z1 and Z2 form
the two 15V voltage sources (E1 and E2 in Figure 78); CR1
and CR2 are protection diodes that allow the output to be
instantaneously shorted to ground when the output is at
any output voltage.
The values of R1, R2 and C4 vary with the MOSFETs’ Miller
capacitance and with the high frequency CMRR of the op
amp used. They are selected to minimize the overshoot in
the step response of the amplifier.
V–
Figure 79. Detailed High Voltage Op Amp
High Voltage, High Frequency Amplifier
Using the LT1227 current feedback amplifier (CFA) in the
extended supply mode as shown in Figure 79, it is relatively easy to get a 1MHz power bandwidth at 100VP-P (see
Figure 80 for component values). This circuit has shortcircuit protection and is stable into all capacitive loads.
If One Is Good, Are Two Better?
Dual and quad op amps can also be configured with
extended supplies, although the design gets just a wee bit
tricky. When extending supplies of multiple stages and/or
complete circuits, some design rules need to change. Op
amp circuits generally require a ground against which to
reference all signals. The problem encountered when
R7
10k
R8
1k
IN
V+
2
N-CHANNEL
MOSFET
LT1227
3
E1 15V
–
OP AMP
+
–
+
OUT
6
OUT
+
R10
1k
R9
9.1k
E2 15V
SUPPLIES SUSPENDED
AS IN FIGURE 2 WITH
R1 = 200Ω , R2 = 1.6K,
AV =
= −11 C4 = INF
R8 R9 – R7 R10
–
P-CHANNEL
MOSFET
(
R8 R9 + R10
V–
AN67 F78
Figure 78. Block Diagram of Suspended Supply Op Amp
AN67-58
–
+
)
AN67 F80
Figure 80. High Speed Suspended Op Amp
Application Note 67
R11
R12
–
–
=
E1
+
E1
R11
LOAD
+
LOAD
R12
A. CONVENTIONAL
B. SUSPENDED
transformer, with the center tap hooked to 5V or 12V or
whatever.” At this point everyone is happy until the transformer comes in. After a few phone calls to make sure that
the transformer maker shipped the right one, the engineer
(face covered with egg) asks if anyone needs a rather large
paperweight. The engineer (still wiping egg from his face)
then decides to use switching power supply technology to
solve this “simple” problem.
AN67 F81
Figure 81. Inverting Amplifiers (A. Conventional, B. Suspended)
using extended supply mode is that “ground” is swinging
through the common mode range of the op amp and
beyond. This raises the following question: “If I cannot
reference the signals to ground, to what can I reference
them?” The answer? “Use the output as the signal reference.” This works for all stages except the last stage,
where using the output as the reference would simply
discard the signal. In the last stage, ground is effectively
the output and the feedback resistor is R12. This is shown
in Figures 81a and 81b. Figure 81a shows a conventional
inverting amplifier where the input and output signals are
referred to ground. Figure 81b shows the equivalent
circuit implemented in the extended supply mode.
Here are two rules for design in the extended supply mode,
which will be demonstrated in the next application:
Rule 1: When designing multiple stages in the extended
supply mode, reference the signals of all stages except the
last to the output of the last stage.
Rule 2: Invert the signal using the circuit in Figure 81b at
the last stage.
Ring-Tone Generator
Ring-tone generators are sine wave output, high voltage
inverters for the specific purpose of ringing telephone
bells. In decades past, the phone company generated their
ring tones with motor generator sets with the capacity to
ring numerous phones simultaneously. Often, ring tones
are 20Hz at 90V with less than 10mA per bell output
current capability. Since the power supplied is low one
would think that the task is minimal. This is not always so.
“It’s simple—no problem,” is often heard in response to
queries about ring tone generators. “Just hook a couple of
logic level FETs to two spare output bits of the microprocessor and hook their drains to the primary side of a
Here is a simple ring-tone generator that can be turned on
and off with a logic signal. It has a fully isolated output, is
short-circuit protected and can be powered by any input
voltage from 3V to 24V.
How It Works
Suspended along with the dual op amp in Figure 82 are two
voltage references and an oscillator. Keep in mind when
referring to Figure 82 that the node labeled “A” is the
output; this is the reference common for the references,
the oscillator and the first lowpass filter (U1a). The two
references, VR1 and VR2, produce ±2.5V. The oscillator
U2, running on the ±2.5V references, produces a 20Hz
square wave rail-to-rail. U1a is a 2nd order, Sallen and Key
lowpass filter that knocks off the sharp edges, presenting
the somewhat smoothed signal at point “B.”
Next comes the tricky stuff. U1b is a 2nd order, multiplefeedback (MFB) lowpass filter/amplifier that performs
four functions: first, it subtracts the voltage at point “A”
(its own output voltage) from the voltage at point “B” (the
incoming signal), forming a difference that is the signal;
second, it filters the difference signal with a 2-pole lowpass filter, smoothing out the last wrinkles in the signal;
third, it amplifies the filtered difference signal with a gain
of 34; and fourth, it references the amplified signal to
ground, forming the output.
Note that R99 shown in Figure 82 is there to protect the
input of U1b in the event that the output is shorted when
the output voltage is very high. This measure is necessary
because the bottom end of C99 is connected to ground,
and C99 could have up to 100V across it. When the output
is shorted to ground from a high voltage, R99 limits the
current into the input of U1b to an acceptable level.
This circuit, when coupled with the switching power
supply shown in Figure 83, implements a fully isolated
sine wave ring tone generator.
AN67-59
Application Note 67
0.047µF
110V
1N4001
3
A
+
A
100k
10k
51k
100Ω
0.01µF
IRF610
4
A
15V
0.1µF
VR1
LT1004
–2.5V
0.1µF
8
2k
6.8k
100V
0.047µF
4
15V
0.1µF
–
U1 = LT1078
LEFT TRIANGLE
SHOWS POWER
INPUT ONLY
0.15µF
51k
VCC
6
U2
CMOS
555
OUT
VR2
LT1004
–2.5V
B
8
R
A
U1
1
U1A
2
11k
A
360k
2
6
3
R99
10k 5
10k
VSS
–
U1B
7
+
1
C99
0.01µF
100V
100Ω
360k
0.33µF
IRF9610
100k
LOAD
10k
A
A
1N5817
AN67 F82
–110V
Figure 82. Ring Tone Generator: Oscillator, Filter and Driver
18 : 112
INPUT
3V TO 15V
MUR120
+
H = SHUTDOWN
110V
68µF
20V
1
10k
2N3904
VC
0.22µF
160V
5
VIN
LT1070
0.01µF
VSW
2
1N5817
MUR120
4
–110V
TRANSFORMER:
EFD25/1319 - 3C8 CORES
15MIL GAP ALL LEGS
MUR120
FB
510Ω
GND
3
0.22µF
160V
100Ω
1/4W
0.01µF
63V
PRI 50µH
18T
SEC
112T
SEC
112T
PRI 50µH
18T
Figure 83. High Voltage Power Supply for Ring Tone Generator
AN67-60
2
* #26
#35
#35
2
* #26
AN67 F83
Application Note 67
500
2.025
460
1.95
420
1.875
380
1.8
340
1.725
300
1.65
260
1.575
220
1.5
180
1.425
140
1.35
100
1.275
60
Although somewhat tricky at first, extended supply mode
is a valuable tool to get out of many tight places. There is
also a great deal of satisfaction to be gathered when
making it work, for those of you who love a technical
challenge.
POWER (W)
INPUT I (mA)
The input current and power versus input voltage for the
combination ring-tone generator (Figures 82 and 83) are
shown in Figure 84. The output waveform (loaded with one
bell) is shown in Figure 85 and the harmonic distortion is
shown in Figure 86.
1.2
3
6
9
12 15 18 21 24
INPUT VOLTAGE (V)
27
30
AN67 F84
Figure 85. Ring-Tone Generator Frequency Spectrum Plot
Figure 84. Input Current and Input Power vs Input Voltage While
Ringing One Bell for Figure 82 Circuit
AN67 F86
Figure 86. Sine Wave Output from Ring-Tone Generator
AN67-61
Application Note 67
USING SUPER OP AMPS TO PUSH TECHNOLOGICAL
FRONTIERS: AN ULTRAPURE OSCILLATOR
by Dale Eagar
The advent of high speed op amps allows the implementation of circuits that were impossible just a few years ago.
This article describes a new topology that makes use of
these new high speed circuits and makes astounding
improvements in its performance. An oscillator using
such op amps has distortion limits beyond our ability to
measure.
An Ultralow Distortion, 10kHz Sine Wave Source for
Calibration of 16-Bit or Higher A/D Converters
stage A1. Second, the input impedance of A1 can be made
very high, further improving both open-loop gain of U1
and the open-loop harmonic distortion of U1. Third, the
output voltage swing of U1 is decreased, keeping its
output circuitry in its lowest distortion area.
The composite circuit, A1, consists of three sections. The
first section, as seen in Figure 90, has the gain/phase plot
shown in Figure 92. Note the high gain at 10kHz (60dB)
and the gain of 6dB at 5MHz, with only 17 degrees of phase
contribution. In fact, this looks so nice that you might ask,
“why not use two?” and thus reduce your distortion by an
additional 60dB?
1
The trick used in this circuit is to build an amplifier that has
excessive gain where it is needed but no excess gain or
phase shift where it isn’t. In many applications the band
from DC to 100kHz requires the above mentioned high
gain; the gain should fall off when the open-loop gain falls
through unity (around 5MHz). How this is done in the flesh
(silicon) is shown here.
Circuit Operation and Circuit Evolution
AV
–
LT1007
+
U1
RL
AN67 F87
Figure 87. Conventional Inverting Op Amp Topology
180
TA = 25°C
VS = ±15V
RL = 2k
160
140
VOLTAGE GAIN (dB)
The path to low distortion in an amplifier or an oscillator
begins with amplifiers with the lowest possible open-loop
distortion and lots of excess open-loop gain in the frequency band of interest. The next step is closing the loop,
thereby reducing open-loop distortion by an amount approximately equal to the loop gain. This is not easy, as
certain stability criteria must be met by an amplifier that
isn’t an oscillator or by an oscillator that oscillates at a
specified frequency.
120
100
80
LT1037
60
40
LT1007
20
0
A standard inverting amplifier topology, as shown in
Figure 87, has a finite open-loop gain in the frequency
band of interest (see Figure 88), with some open-loop
harmonic distortion (about – 60dB) and an open-loop
output impedance of about 70Ω.
The amplifier shown in Figure 87 can achieve low distortion, but since the circuit has a limited loop gain, the
curative effects of feedback can only be taken so far. The
designer must also be careful to ensure that RL is many
times higher than the open-loop output impedance of U1.
Figure 89’s circuit makes several improvements over the
circuit of Figure 87. First, the open-loop gain of U1 is
multiplied by AV(f), the gain of the composite amplifier
AN67-62
–20
0.01 0.1
1
10 100 1k 10k 100k 1M 10M 100M
FREQUENCY (Hz)
AN67 F88
Figure 88. Voltage Gain vs Frequency
AV
1
AV(f)
–
U1
LT1007
A1
AMP
+
AN67 F89
Figure 89. LT1007 Followed by Composite Amplifier A1
Application Note 67
The second section, shown in Figure 91, has the gain/
phase plot shown in Figure 93. Note that here the gain
doesn’t change significantly, but the phase is positive just
IN
+
where we want it (1MHz to 5MHz) to allow a very stable
system to be built.
The third section, as you might guess, is the same as the
first. In sum, the gain/phase plot of the composite amplifier A1 is shown in Figure 94. Note the gain, which is in
LT1230
470pF
–
390pF
750Ω
390pF
0Ω
SOURCE
5.1k
750Ω
750Ω
750Ω
2200pF
AN67 F90
AN67 F91
Figure 91. Second Section of Composite Amplifier A1
Figure 90. First Section of Composite Amplifier A1
80
20
70
0
60
–20
50
–40
40
–60
30
–80
20
–100
10
PHASE (DEGREE)
PHASE
GAIN (dB)
∞Ω
LOAD
–120
GAIN
0
–140
–10
–160
–20
1k
10k
100k
1M
FREQUENCY (Hz)
10M
–180
100M
AN67 F92
Figure 92. Gain/Phase Response of Circuit Shown in Figure 90
4
50
2
40
GAIN
0
30
20
–4
10
PHASE
–6
0
–8
–10
–10
–20
–12
–30
–14
–40
–16
1k
10k
100k
1M
FREQUENCY (Hz)
10M
PHASE (DEGREE)
GAIN (dB)
–2
–50
100M
AN67 F93
Figure 93. Gain/Phase Response of Circuit Shown in Figure 91
AN67-63
Application Note 67
excess of 120db at 10kHz and the total phase contribution
of about – 20 degrees at 5MHz. The complete gain block is
shown in Figure 95.
1 billion). This means that the closed-loop harmonic distortion can easily be kept in the region of “parts per billion.”
A Wien bridge oscillator with harmonic distortion in the
parts per billion is shown in Figure 96. The super op amps
S1 and S2 are the previously described composite amplifiers as shown in Figure 95. Note that the output is taken
between the two outputs of S1 and S2. This topology gives
the best signal-to-noise ratio, in addition to balancing the
Super Gain Block Oscillator Circuitry
When A1, as described above is connected with U1, as
shown in Figure 89, the resulting circuit is not only unitygain stable but has open-loop gain of 180dB at 10kHz (yes,
160
100
140
50
120
0
100
– 50
GAIN (dB)
80
–100
60
–150
40
–200
20
PHASE (DEGREE)
PHASE
–250
GAIN
0
–300
– 20
–350
– 40
1k
10k
100k
1M
FREQUENCY (Hz)
10M
– 400
100M
AN67 F94
Figure 94. Gain/Phase Response of Composite Amplifier A1 (Shown in Figure 89)
Super gain-block (U1 + A1),
complete schematic shown below
–
50pF
AV = 180dB AT 10kHz
1k
– IN
470pF
–
+
LT1007
+
750Ω
750Ω
LT1230
10k
–
1N4148
1N4148
750Ω
–7.5V
2200pF
390pF
+
LT1230
–
750Ω
750Ω
750Ω
A1
Figure 95. Super Gain Block S1 and S2 Schematic Diagram
AN67-64
390pF
390pF
390pF
5.1k
U1
OUTPUT
5.1k
AN67 F95
Application Note 67
power supply currents and their harmonics. Taking the
output from one amplifier’s output to ground is also valid.
To align the circuit, first center the output amplitude
adjustment potentiometer. Next, adjust the gain trim for
oscillation while also adjusting the output amplitude for
5VP-P output (single ended). Next, adjust the gain trim to
1VP-P at the output of the LT1228. Finally, connect a
2k
0.02µF
49.9Ω
spectrum analyzer to the output of the LT1228 and adjust
the second harmonic trim potentiometer for a null in the
second harmonic of the oscillator frequency. The measurement of the harmonic distortion of this oscillator
defies all of our resources, but appears to be well into the
parts-per-billion range.
OUTPUT BNC
49.9Ω
1k
1k
0.01µF
1k
–
–
S2
S1
7.5
500k
100k
10k
GAIN
TRIM
SECOND HARMONIC
TRIM
–7.5
20k
50k
10k
10k
LT1228
3
1N4148
1N4148
+
1N4148
+
2
100Ω
4.7µF
20k
100k
100Ω
7.5k
20k
1
–
1k
5
8
–
6
–
–
LT1007
+
470Ω
3k
100Ω
LT1007
24k
+
OUTPUT
AMPLITUDE
ADJ.
1.6k
500k
–7.5V 1.5µF
2N3906
–
1k
LT1007
+
1.6k
1.6k
0.33µF
1µF
AN67 F96
Figure 96. Schematic Diagram: Wien Bridge Oscillator with Distortion in the Parts-per-Billion Range
AN67-65
Application Note 67
FAST VIDEO MUX USES LT1203/LT1205
by Frank Cox
To demonstrate the switching speed of the LT1203/LT1205,
the RGB MUX of Figure 97 is used to switch the inputs of
an RGB workstation with a 22ns pixel width. Figure 98a is
a photo showing the workstation output and RGB MUX
output. The slight rise time degradation at the RGB MUX
output is due to the bandwidth of the LT1260 current
feedback amplifier used to drive the 75Ω cable. In Figure
98b the LT1203 switches at the end of the first pixel to an
input at zero and removes the following pixels.
1.5k
RED 1
+1
RED 2
V–
+1
GREEN 1
+1
GREEN 2
V–
+1
BLUE 1
V+
CHANNEL
SELECT
–
75Ω
RED OUT
OUT
+
EN
75Ω
LOGIC
V+
OUT
+
EN
75Ω
GREEN OUT
LOGIC
LT1205
–
1.5k
V+
+1
+
EN
+1
75Ω
BLUE OUT
LOGIC
LT1203
75Ω
1.5k
OUT
BLUE 2
V–
1.5k
–
75Ω
LT1260
1.5k
1.5k
AN67 F97
Figure 97. Fast RGB MUX
“WORK
STATION”
OUTPUT
“WORK
STATION”
OUTPUT
RGB MUX
OUTPUT
RGB MUX
OUTPUT
AN67 F98a
Figure 98a. Workstation and RGB MUX Output
AN67-66
AN67 F98b
Figure 98b. RGB MUX Output Switched to Ground After One Pixel
Application Note 67
USING A FAST ANALOG MULTIPLEXER
TO SWITCH VIDEO SIGNALS FOR NTSC
“PICTURE-IN-PICTURE” DISPLAYS
by Frank Cox
Introduction
The majority of production1 video switching consists of
selecting one video source out of many for signal routing
or scene editing. For these purposes, the video signal is
switched during the vertical interval in order to reduce
visual switching transients. The image is blanked during
this time, so if the horizontal and vertical synchronization
and subcarrier lock are maintained, there will be no visible
artifacts. Although vertical interval switching is adequate
for most routing functions, there are times when it is
desirable to switch two synchronous video signals during
the active (visible) portion of the line to obtain picture-inpicture, key or overlay effects. Picture-in-picture or active
video switching requires signal-to-signal transitions that
are both clean and fast. A clean transition should have a
minimum of preshoot, overshoot, ringing or other aberrations commonly lumped under the term “glitching.”
OFF AIR VIDEO
SOURCE OR VIDEO
PATTERN
GENERATOR
SYNC STRIPPER,
SAMPLE PULSE
GENERATOR
SAMPLE PULSE
SCOPE
50%
75Ω
75Ω
75Ω
75Ω
LT1204
OUTPUT
MONITOR
LOOP
THROUGH
75Ω
INPUTS
AN67 F99
Figure 99. “Picture-in-Picture” Test Setup
Using the LT1204
A quality high speed multiplexer amplifier can be used with
good results for active video switching. The important
specifications for this application are small, controlled
switching glitch, good switching speed, low distortion,
good dynamic range, wide bandwidth, low path loss, low
channel-to-channel crosstalk and good channel-to-channel offset matching. The LT1204 specifications match
these requirements quite well, especially in the areas of
bandwidth, distortion and channel-to-channel crosstalk
(which is an outstanding 90dB at 10MHz). The LT1204
was evaluated for use in active video switching with the
test setup shown in Figure 99. Figure 100 shows the video
waveform of a switch between a 50% white level and a 0%
white level about 30% into the active interval and back
again at about 60% of the active interval. The switch
artifact is brief and well controlled. Figure 101 is an
expanded view of the same waveform. When viewed on a
monitor, the switch artifact is just visible as a very fine
AN67 F100
Figure 100. Video Waveform Switched from 50% White Level
to 0% White Level and Back
AN67 F101
Figure 101. Expanded View of Rising Edge of LT1204 Switching
from 0% to 50% (50ns Horizontal Division)
AN67-67
Application Note 67
Some Definitions—
“Picture in picture” refers to the production effect in
which one video image is inserted within the boundaries of another. The process may be as simple as
splitting the screen down the middle or it may involve
switching the two images along a complicated geometric boundary. In order to make the composite picture
stable and viewable, both video signals must be in
horizontal and vertical sync. For composite color signals the signals must also be in subcarrier lock.
AN67 F102
Figure 102. Expanded View of “Brand-X” Switch
0% to 50% Transition
line. The lower trace is a switch between two black level
(0V) video signals showing a very slight channel-tochannel offset, which is not visible on the monitor. Switching between two DC levels is a worst-case test, as almost
any active video will have enough variation to totally
obscure this small switch artifact.
“Keying” is the process of switching among two or
more video signals, triggering on some characteristic
of one of the signals. For instance, a chroma keyer will
switch on the presence of a particular color. Chroma
keyers are used to insert a portion of one scene into
another. In a commonly used effect, the TV weather
person (the “talent”) appears to be standing in front of
a computer generated weather map. Actually, the talent
is standing in front of a specially colored background;
the weather map is a separate video signal which has
been carefully prepared to contain none of that particular color. When the chroma keyer senses the keying
color, it switches to the weather map background.
Where there is no keying color, the keyer switches to the
talent’s image.
Video Switching Caveats
In a video processing system that has a large bandwidth
compared to the bandwidth of the video signal, a fast
transition from one video level to another (with a low
amplitude glitch) will cause minimal visual disturbance.
This situation is analogous to the proper use of an analog
oscilloscope. In order to make accurate measurement of
pulse waveforms, the instrument must have much more
bandwidth than the signal in question (usually five times
the highest frequency). Not only should the glitch be
small, it should be otherwise well controlled. A switching
glitch that has a long settling “tail” can be more troublesome (that is, more visible) than one that has more
amplitude but decays quickly. The LT1204 has a switching
glitch that is not only low in amplitude but well controlled
and quickly damped. Refer to Figure 102, which shows a
video multiplexer that has a long, slow-settling tail. This
sort of distortion is highly visible on a video monitor.
Composite video systems, such as NTSC, are inherently
band limited and thus edge-rate limited. In a sharply band
limited system, the introduction of signals that contain
significant energy higher in frequency than the filter cutoff
will cause distortion of transient waveforms (see Figure
103). Filters used to control the bandwidth of these video
AN67-68
systems should be group-delay equalized to minimize this
pulse distortion. Additionally, in a band limited system, the
edge rates of switching glitches or level-to-level transitions should be controlled to prevent ringing and other
pulse aberrations that could be visible. In practice, this is
usually accomplished with pulse-shaping networks (Bessel
filters are one example). Pulse-shaping networks and
delay equalized filters add cost and complexity to video
systems and are usually found only on expensive equipment. Where cost is a determining factor in system
design, the exceptionally low amplitude and brief duration
of the LT1204’s switching artifact make it an excellent
choice for active video switching.
1
Video production, in the most general sense, means any purposeful manipulation of the video
signal, whether in a television studio or on a desktop PC.
Application Note 67
1
fC
RISE TIME ≈
DELAY ≈
1
2fC
N
(WHERE N IS ORDER OF FILTER)
2fC
AN67 F103
Figure 103. Pulse Response of an Ideal Sharp Cutoff Filter at Frequency fC
APPLICATIONS FOR THE LT1113 DUAL JFET OP AMP
by Alexander Strong
Figure 104 shows a low noise hydrophone amplifier with
a DC servo. Here one half (A) of the LT1113 is configured
in the noninverting mode to amplify a voltage signal from
the hydrophone, and the other half (B) of the LT1113 nulls
errors due to voltage and current offsets of amplifier A and
to null out DC errors of the hydrophone. The value of C1
depends on the capacitance of the hydrophone, which can
range from 200pF to 8000pF. The time constant of the
servo should be larger than the time constant of the
hydrophone capacitance and the 100M source resistance.
This will prevent the servo from canceling the low frequency signals from the hydrophone.
3.9k
C1
V+
2
100M
200Ω
3
8
A
1/2 LT1113
–
+
1
OUTPUT
4
V–
1µF
CT
HYDROPHONE
100M
1M
100k
–
7
6
1M
5
1M
B
1/2 LT1113
C1 = CT = 200pF TO 8000pF
DC OUTPUT ≤ 4mV AT TA < 70°C
OUTPUT VOLTAGE NOISE = 130nV/√Hz AT 1kHz
POWER SUPPLY RANGE = ±5V TO ±15V
+
AN67 F104
Figure 104. Low Noise Hydrophone Amplifier with DC Servo
AN67-69
Application Note 67
Another popular charge-output transducer is the accelerometer. Since precision accelerometers are charge-output devices, the inverting mode is used to convert the
transducer charge to an output voltage. Figure 105 is an
example of an accelerometer with a DC servo. The charge
from the transducer is converted to a voltage by C1, which
should equal the transducer capacitance plus the input
capacitance of the op amp. The noise gain will be 1 + C1/
CT. The low frequency bandwidth of the amplifier will
depend on the value of R1 • C1 (or R1 (1 + R2/R3) for a Tee
network). As with the hydrophone example, the time
constant of the servo (1/R5C5) should be larger than the
time constant of the amplifier (1/R1C1).
C1
1250pF
R1
100M
OUTPUT = 0.8µV/pC*
= 8.0mV/G**
R2
18k
R3
2k
7
DC OUTPUT ≤ 2.7mV
OUTPUT NOISE = 6µV/√Hz AT 1kHz
C4
2µF
*PICOCOULOMBS
**EARTH’S GRAVITATIONAL CONSTANT
–
6
R4
20M
+
5
R5
20M
A
1/2 LT1113
C5
2µF
8
2
–
B
ACCELEROMETER
1/2 LT1113
3
CT
1
+
AN67 F105
4
Figure 105. Accelerometer Amplifier with DC Servo
INPUT
1µF
15V
1µF
+
Although the wide bandwidth and high output drive capabilities of the LT1206 make it a natural for video circuits,
these characteristics are also useful for audio applications. Figure 106 shows the LT1206 combined with the
LT1115 low noise amplifier to form a very low noise, low
distortion audio buffer with a gain of 10. With a 32Ω load
and a 5VRMS output level (780mW), the THD + noise for
the circuit is 0.0009% at 1kHz, rising to 0.004% at 20kHz.
The frequency response is flat to 0.1dB from DC to
600kHz, with a – 3dB bandwidth of 4MHz. The circuit is
stable with capacitive loads of 250pF or less.
15V
+
+
LT1206 AND LT1115 MAKE LOW NOISE
AUDIO LINE DRIVER
by William Jett
+
LT1115
–
1µF
LT1206
+
–
–15V
1µF
68pF
0.01µF
+
–15V
560Ω
560Ω
909Ω
100Ω
AN67 F106
Figure 106. Low Noise × 10 Buffered Line Driver
AN67-70
Application Note 67
DRIVING MULITPLE VIDEO CABLES WITH THE LT1206
by William Jett
The combination of a 60MHz bandwidth, 250mA output
current capability and low output impedance makes the
LT1206 ideal for driving multiple video cables. One concern when driving multiple transmission lines is the effect
of an unterminated (open) line on the other outputs. Since
the unterminated line creates a reflected wave that is
incident on the output of the driver, a nonzero amplifier
output impedance will result in crosstalk to the other lines.
Figure 107 shows the LT1206 connected as a distribution
amplifier. Each line is separately terminated to minimize
the effect of reflections. For systems using composite
video, the differential gain and phase performance are also
important and have been considered in the internal design
of the device. The differential phase and differential gain
performance versus supply is shown in Figures 108 and
109 for 1, 3, 5 and 10 cables. Figure 110 shows the output
impedance versus frequency. Note that at 5MHz the output impedance is only 0.6Ω.
0.10
RF = RG = 560Ω
N PACKAGE
+
eIN
75Ω
75Ω CABLE
LT1206
–
75Ω
75Ω
RF
DIFFERENTIAL GAIN (%)
0.08
RL = 15Ω
(10 CABLES)
0.06
RL = 30Ω (5 CABLES)
RL = 50Ω
(3 CABLES)
0.04
0.02
75Ω
RL = 150Ω (1 CABLE)
0.00
RG
7
5
75Ω
RF = RG = 560
9
11
13
SUPPLY VOLTAGE (±V)
AN67 F107
Figure 107. LT1206 Distribution Amplifier
AN67 F109
Figure 109. Differential Gain vs Supply Voltage
0.50
100
RL = 15Ω (10 CABLES)
OUTPUT IMPEDANCE (Ω)
DIFFERENTIAL PHASE (DEG)
RF = RG = 560Ω
N PACKAGE
0.40
0.30
RL = 30Ω (5 CABLES)
0.20
RL = 50Ω (3 CABLES)
0.10
15
VS = ±15V
RF = RG = 560Ω
N PACKAGE
10
1
0.1
RL = 150Ω (1 CABLE)
0.00
5
7
9
11
13
SUPPLY VOLTAGE (±V)
15
AN67 F108
Figure 108. Differential Phase vs Supply Voltage
0.01
100k
1M
10
FREQUENCY (Hz)
100
AN67 F110
Figure 110. Output Impedance vs Frequency
AN67-71
Application Note 67
OPTIMIZING A VIDEO GAIN CONTROL STAGE
USING THE LT1228
by Frank Cox
Notice that the signal is attenuated 20:1 by the 75Ω
attenuator at the input of the LT1228, so the voltage on the
input (Pin 3) ranges from 0.028V to 0.090V. This is done to
limit distortion in the transconductance stage. The gain of
this circuit is controlled by the current into the ISET terminal,
Pin 5 of the IC. In a closed-loop AGC system the loop control
circuitry generates this current by comparing the output of
a detector2 to a reference voltage, integrating the difference
and then converting to a suitable current. The measured
performance for this circuit is presented in Table 1.
Video automatic gain control (AGC) systems require a
voltage- or current-controlled gain element. The performance of this gain-control element is often a limiting
factor in the overall performance of the AGC loop. The gain
element is subject to several, often conflicting, restraints.
This is especially true of AGC for composite color video
systems such as NTSC, which have exacting phase and
gain distortion requirements. To preserve the best possible signal-to-noise ratio (S/N),1 it is desirable for the
input signal level to be as large as practical. Obviously, the
larger the input signal the less the S/N will be degraded by
the noise contribution of the gain control stage. On the
other hand, the gain control element is subject to dynamic
range constraints and exceeding these will result in rising
levels of distortion.
Table 1. Measured Performance Data (Uncorrected)
INPUT
(V)
ISET
(mA)
DIFFERENTIAL
GAIN
DIFFERENTIAL
PHASE
S/N
0.03
1.93
0.5%
2.7°
55dB
0.06
0.90
1.2%
1.2°
56dB
0.09
0.584
10.8%
3.0°
57dB
All video measurements were taken with a Tektronix
1780R video measurement set, using test signals generated by a Tektronix TSG 120. The standard criteria for
characterizing NTSC video color distortion are the differential gain and the differential phase. For a brief explanation of these tests see “Differential Gain and Phase” page
74. For this design exercise the distortion limits were set
at a somewhat arbitrary 3% for differential gain and 3° for
differential phase. Depending on conditions, this should
be barely visible on a video monitor.
Linear Technology makes a high speed transconductance
(gm) amplifier, the LT1228, which can be used as a quality,
inexpensive gain control element in color video and some
lower frequency RF applications. Extracting the optimum
performance from video AGC systems takes careful attention to circuit details.
As an example of this optimization, consider the typical
gain control circuit using the LT1228 shown in Figure 111.
The input is NTSC composite video, which can cover a
10dB range, from 0.56V to 1.8V. The output is to be 1V
peak-to-peak into 75Ω. Amplitudes were measured from
peak negative chroma to peak positive chroma on an NTSC
modulated ramp test signal (see page 74).
Figures 112 and 113 plot the measured differential gain
and phase, respectively, against the input signal level (the
1 Signal to noise ratio, S/N = 20 ×
the colorburst for NTSC is 40% of the peak luminance) with a sample-and-hold and peak detector.
750Ω
82.5Ω
BIAS
GENERATOR
3.75k
TEKTRONIX
TSG 120
75
8
2
LT1228
20:1
6
1
gm
3
75Ω
ATTENUATOR
–
–
37.5Ω
VARIABLE
ATTENUATOR
+
5
+
RSET
365Ω
75Ω
TEKTRONIX
1780R VIDEO
MEASUREMENT
75Ω SET
75Ω
AN67 F111
2k
VARIABLE ISET
GENERATOR
Figure 111. Schematic Diagram
AN67-72
log(RMS signal/RMS noise).
2 One way to do this is to sample the colorburst amplitude (the nominal peak-to-peak amplitude of
Application Note 67
11
2.0
10
1.8
TRANSCONDUCTANCE (µA/mV)
DIFFERENTIAL GAIN (%)
9
8
7
6
A, UNCORRECTED
5
4
3
2
B, CORRECTED
1.6
V S = ±2V TO ±15V
ISET = 100µA
1.4
–55°C
1.2
1.0
25°C
0.8
0.6
125°C
0.4
0.2
1
0
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
VIDEO INPUT LEVEL (V)
0
50 100 150 200
–200 –150 –100 –50 0
INPUT VOLTAGE (mVDC)
AN67 F112
Figure 112. Differential Gain vs Input Level
DIFFERENTIAL PHASE (DEG)
3.0
A, UNCORRECTED
2.5
2.0
B, CORRECTED
1.0
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
VIDEO INPUT LEVEL (V)
AN67 F113
Figure 113. Differential Phase vs Input Level
curves labeled “A” show the uncorrected data from
Table 1). The plots show that increasing the input signal
level beyond 0.06V results in a rapid increase in the gain
distortion, but comparatively little change in the phase
distortion. Further attenuating the input signal (and consequently increasing the set current) would improve the
differential gain performance but degrade the S/N. What
this circuit needs is a good tweak.
Optimizing for Differential Gain
Referring to the small-signal transconductance versus
DC input voltage graph (Figure 114), observe that the
transconductance of the amplifier is linear over a region
centered around 0V.3 The 25°C gm curve starts to become quite nonlinear above 0.050V. This explains why
the differential gain (see Figure 112, curve A) degrades so
quickly with signals above this level. Most RF signals do
3Notice
Figure 114. Small-Signal Transconductance vs DC Input Voltage
not have DC bias levels but the composite video signal is
mostly unipolar.
3.5
1.5
AN67 F114
also that the linear region expands with higher temperature. Heating the chip has been
suggested.
Video is usually clamped at some DC level to allow easy
processing of sync information. The sync tip, the chroma
reference burst and some chroma signal information swing
negative, but 80% of the signal that carries the critical color
information (chroma) swings positive. Efficient use of the
dynamic range of the LT1228 requires that the input signal
have little or no offset. Offsetting the video signal so that the
critical part of the chroma waveform is centered in the
linear region of the transconductance amplifier allows a
larger signal to be input before the onset of severe distortion. A simple way to do this is to bias the unused input (in
this circuit the inverting input, Pin 2) with a DC level.
In a video system, it might be convenient to clamp the sync
tip at a more negative voltage than usual. Clamping the
signal prior to the gain-control stage is good practice
because a stable DC reference level must be maintained.
The optimum value of the bias level on Pin 2 used for this
evaluation was determined experimentally to be about
0.03V. The distortion tests were repeated with this bias
voltage added. The results are reported in Table 2 and
Figures 112 and 113 (curves B). The improvement to the
differential phase is inconclusive, but the improvement in
the differential gain is substantial.
Table 2. Measured Performance Data (Corrected)
INPUT
BIAS
(V)
VOLTAGE
ISET
(mA)
DIFFERENTIAL
GAIN
DIFFERENTIAL
PHASE
S/N
0.03
0.03
1.935
0.9%
1.45°
55dB
0.06
0.03
0.889
1.0%
2.25°
56dB
0.09
0.03
0.584
1.4%
2.85°
57dB
AN67-73
Application Note 67
Differential Gain and Phase
Differential gain and phase are sensitive indications of
chroma signal distortion. The NTSC system encodes
color information on a separate subcarrier at
3.579545MHz. The color subcarrier is directly summed
to the black and white video signal. (The black and white
information is a voltage proportional to image intensity
and is called luminance or luma.) Each line of video has
a burst of 9 to 11 cycles of the subcarrier (so timed that
it is not visible) that is used as a phase reference for
demodulation of the color information of that line. The
color signal is relatively immune to distortions, except
for those that cause a phase shift or an amplitude error
to the subcarrier during the period of the video line.
Differential gain is a measure of the gain error of a linear
amplifier at the frequency of the color subcarrier. This
distortion is measured with a test signal called a modulated ramp (shown in Figure 115). The modulated ramp
consists of the color subcarrier frequency superimposed
on a linear ramp (or sometimes on a stair step). The ramp
has the duration of the active portion of a horizontal line
of video. The amplitude of the ramp varies from zero to
the maximum level of the luminance, which in this case,
is 0.714V. The gain error corresponds to compression or
expansion by the amplifier (sometimes called “incremental gain”) and is expressed as a percentage of the full
amplitude range. An appreciable amount of differential
gain will cause the luminance to modulate the chroma,
producing visual chroma distortion. The effect of differential gain errors is to change the saturation of the color
being displayed. Saturation is the relative degree of
dilution of a pure color with white. A 100% saturated
color has 0% white, a 75% saturated color has 25%
white, and so on. Pure red is 100% saturated, whereas
pink is red with some percentage of white and is therefore
less than 100% saturated.
Differential phase is a measure of the phase shift in a
linear amplifier at the color subcarrier frequency when
the modulated ramp signal is used as an input.
The phase shift is measured relative to the colorburst on
the test waveform and is expressed in degrees. The
visual effect of the distortion is a change in hue. Hue is
that quality of perception which differentiates the frequency of the color, red from green, yellow-green from
yellow, and so forth.
AN67-74
Three degrees of differential phase is about the lower
limit that can unambiguously be detected by observers.
This level of differential phase is just detectable on a
video monitor as a shift in hue, mostly in the yellowgreen region. Saturation errors are somewhat harder to
see at these levels of distortion—3% of differential gain
is very difficult to detect on a monitor. The test is
performed by switching between a reference signal,
SMPTE (Society of Motion Picture and Television Engineers) 75% color bars and a distorted version of the
same signal, with matched signal levels. An observer is
then asked to note any difference.
In professional video systems (studios, for instance)
cascades of processing and gain blocks can reach hundreds of units. In order to maintain a quality video signal,
the distortion contribution of each processing block must
be a small fraction of the total allowed distortion budget4
(the errors are cumulative). For this reason, high quality
video amplifiers will have distortion specifications as low
as a few thousandths of a degree for differential phase and
a few thousandths of a percent for differential gain.
4 From
the preceding discussions, the limits on visibility are about 3° differential phase, 3%
differential gain. Please note that these are not hard and fast limits. Tests of perception can be
very subjective.
0.714V
100% WHITE
0V
BLANKING
–0.286V
0.1429V
0V
–0.1429V
3.58 MHz COLOR
SUBCARRIER SUMMED
TO LINEAR RAMP
0µs
7µs
10µs
11.5µs
AN67 F115
Figure 115. NTSC Test Signal
Application Note 67
Introduction
The LT1190 series op amps combine bandwidth, slew rate
and output drive capability to satisfy the demands of many
high speed applications. This family offers up to 350MHz
gain bandwidth product and slew rates of 450V/µs while
driving 150Ω (75Ω, double terminated) loads. In 50Ω
systems, the LT1190 family can deliver 13.5dBm to a
double terminated load. These parts are based on the
familiar, easy-to-use, voltage mode feedback topology.
frequency rises, the error increases because capacitor
charging time decreases. During this time the overdrive
becomes a very small portion of a sine wave cycle. Finally,
at approximately 4MHz, the error rises rapidly owing to the
slew-rate limitation of the op amp. For comparison purposes, the error of an LM118 is also plotted for VIN = 2VP-P.
10
AV = +1
RL = 1k
8
LT1191
6
4
2
2dB/DIV
LT1190 FAMILY ULTRAHIGH SPEED
OP AMP CIRCUITS
by John Wright and Mitchell Lee
0
–2
LT1190
–4
Small-Signal Performance
Figures 116 and 117 show the small-signal performance
of the LT1190 and LT1191 when configured for gains of +1
and – 1. The noninverting plots show peaking at 130MHz,
which is characteristic of the socketed test fixture and
supply bypass components. A tight PC board layout would
reduce the LT1190 peaking to 2dB. The small-signal
performance of an LM118 is shown for comparison.
–6
LM118
–8
–10
100k
1M
10M
100M
TIME (s)
AN67 F116
Figure 116. Small-Signal Response AV = +1. 130MHz Peaking
Due to Socket and Bypass Components
8
Fast Peak Detectors
AV = –1
RL = RFB = 1k
6
4
Fast peak detectors place unusual demands on amplifiers.
The output stage must have a high slew rate in order to
keep up with the intermediate stages of the amplifier. This
condition causes either a long overload or DC accuracy
errors. To maintain a high slew rate at the output, the
amplifier must deliver large currents into the capacitive
load of the detector. Other problems include amplifier
instability with large capacitive loads and preservation of
output voltage accuracy.
2dB/DIV
2
LT1191
0
–2
–4
LM118
–6
–8
LT1190
–10
–12
100k
1M
10M
100M
TIME (s)
AN67 F117
The LT1190 is the ideal candidate for this application, with
a 450V/µs slew rate, 50mA output current and 70° phase
margin. The closed-loop peak detector circuit of Figure
118 uses a Schottky diode inside the feedback loop to
obtain good accuracy. A 20Ω resistor (RO) isolates the
10nF load and prevents oscillation.
DC error with a sine wave input is plotted in Figure 119 for
various input amplitudes. The DC value is read with a DVM.
At low frequencies, the error is small and is dominated by
the decay of the detector capacitor between cycles. As
Figure 117. Small-Signal Response AV = – 1
5V
3
VIN
RS
51Ω
+
7
LT1190
2
–
1N5712
6
RO
20Ω
VO DC
CL
10nF
4
– 5V
AN67 F118
Figure 118. Closed-Loop Peak Detector
AN67-75
Application Note 67
50
LM118
VIN = 2VP-P
40
DC DETECTOR ERROR (%)
DC DETECTOR ERROR (%)
50
SLEW RATE LIMIT
30
LT1190
VIN = 4VP-P
20
LT1190
VIN = 2VP-P
10
LT1190
VIN = 6VP-P
0
10k
100k
1M
40
LM118
VIN = 2VP-P
SCHOTTKY
VIN = 2VP-P
LT1190
30 SCHOTTKY
VIN = 4VP-P
VIN = 2VP-P
VIN = 4VP-P
20
SCHOTTKY
VIN = 6VP-P
VIN = 6VP-P
10
0
10M
10
1
FREQUENCY (Hz)
100
FREQUENCY (MHz)
AN67 F119
AN67 F121
Figure 121. Open-Loop Peak Detector Error vs Frequency
Figure 119. Closed-Loop Peak Detector Error vs Frequency
VIN
+
LT1190
RS
51Ω
2
–
D1
1N5712
7
4
6
VIN
D2
1N5712
CFB
10nF
5V
RI
1k
5V
3
RL
51k
CL
1nF
3
CI
20pF
RS
51Ω
+
7
LT1190
2
–
4
– 5V
–5V
6
D1
1N5712
D2
1N5712
CL
1nF
– 5V
–5V
RB
51k
RL
10k
RB
10k
AN67 F120
AN67 F122
– 5V
– 5V
Figure 120. Open-Loop, High Speed Peak Detector
Figure 122. Fast Pulse Detector
A Schottky diode peak detector can be built with a 1nF
capacitor and a 10kΩ pulldown. Although this simple
circuit is very fast, it has limited usefulness because of the
error of the diode threshold and its low input impedance.
The accuracy of this simple detector can be improved with
the LT1190 circuit of Figure 120.
The DC error with a sine wave input, as read with a DVM,
is plotted in Figure 121. For comparison purposes the
LM118 error is plotted as well as the error of the simple
Schottky detector.
In this open loop design, D1 is the detector diode and D2
is a level shifting or compensating diode. A load resistor,
RL, is connected to – 5V and an identical bias resistor, RB,
is used to bias the compensating diode. Equal value
resistors ensure that the diode drops are equal. Low
values of RL and RB (1k to 10k) provide fast response, but
at the expense of poor low frequency accuracy. High
values of RL and RB provide good low frequency accuracy
but cause the amplifier to slew rate limit, resulting in poor
high frequency accuracy. A good compromise can be
made by adding a feedback capacitor, CFB, which enhances the negative slew rate on the (–) input.
A fast pulse detector can be made with the circuit of Figure
122. A very fast input pulse will exceed the amplifier’s slew
rate and cause a long overload recovery time. Some
amount of dV/dt limiting on the input can help this overload condition; however, it will delay the response.
AN67-76
Pulse Detector
Figure 123 shows the detector error versus pulse width.
Figure 124 is the response to a 4VP-P input pulse that is
80ns wide. The maximum output slew rate in the photo is
70V/µs. This rate is set by the 70mA current limit driving
1nF. As a performance benchmark, the LM118 takes 1.2µs
to peak detect and settle, given the same amplitude input.
Application Note 67
10k
90
VIN = 4VP-P
dV/dt LIMITING = 1kΩ, 20pF
80
5V
10k
DC DETECTOR ERROR (%)
70
60
VIN
2
3
10k
VCM
120VP-P
30
7
LT1192
100Ω
50
40
–
+
99Ω
20
6
4
– 5V
AN67 F125
10
Figure 125. 3.5MHz Instrumentation Amplifier Rejects 120VP-P
0
0
20
40
60
80
100
PULSE WIDTH (ns)
AN67 F123
Figure 123. Detector Error vs Pulse Width
2V/DIV
A
1V/DIV
AN67 F126
200ns/DIV
B
1V/DIV
Figure 126. Open-Loop Peak Detector Response
AN67 F124
20ns/DIV
Figure 124. Open-Loop Peak Detector Response
This slower response is due in part to the much lower slew
rate and lower phase margin of the LM118.
Instrumentation Amplifier Rejects High Voltage
Instrumentation amplifiers are normally used to process
slowly varying outputs from transducers, rather than fast
signals. However, it is possible to make an instrumentation amplifier that responds very quickly, with good common mode rejection. For the circuit of Figure 125, an
LT1192 is used to obtain 50dB of CMRR from a 120VP-P
signal. In this application, the CMRR is limited by the
matching of the resistors, which should match to better
than 0.01%.
An LT1192 is used in this application because the circuit
has a noise gain of 100 and because the higher gain
bandwidth of the LT1192 allows a – 3dB bandwidth of
3.5MHz. Note also that the 100:1 attenuation of the common mode signal presents a common mode voltage to the
amplifier of only 1.2VP-P. Figure 126 shows the amplifier
output for a 1MHz square wave riding on a 120VP-P, 60Hz
signal. The circuit exhibits 50dB rejection of the common
mode signal.
Crystal Oscillator
Op amps have found wide use in low frequency (≤ 100kHz)
crystal oscillator circuits, but just haven’t had the bandwidth to operate successfully at higher frequencies. The
LT1190 and LT1191 make excellent gain stages for highfrequency Colpitts oscillators. A practical implementation
is shown in Figure 127.
Gain limiting is provided by two Schottky diodes, which
maintain the output at approximately +11dBm—sufficient
to directly drive +7 or +10dBm diode-ring mixers. Outputstage clipping is not recommended as a means of gain
AN67-77
Application Note 67
5V
+
75pF
75pF
–
3.579545MHz
–5V
1k
330pF
TO NETWORK
ANALYZER
(ZIN = 50Ω)
51Ω
LT1190
100k
1N5711
1N5711
AN67 F127
Figure 127. High Frequency Colpitts Oscillator
20
limiting, as this increases distortion and allows internal
nodes to be overdriven. The recovery time would add
excessive phase shift in the oscillator loop, degrading
frequency stability.
10
OUTPUT POWER (dBm)
0
Distortion performance is good, considering that the oscillator consists of one stage and can deliver useful output
power. Figure 128 shows a spectral plot of the oscillator’s
output. The second harmonic is approximately 37dB down,
limited primarily by the clipping action of the Schottky
diodes. Power supply rejection is excellent, showing a
frequency sensitivity of approximately 0.1ppm/V. The
LT1190 gives acceptable performance to 10MHz, while the
LT1191 extends the circuit’s operating range to 20MHz.
AN LT1112 DUAL OUTPUT BUFFERED REFERENCE
by George Erdi
A dual output buffered reference application is shown in
Figure 129.
Figure 129 works on two AA batteries, which can be
discharged to ±1.3V. With two equal 20k resistors, two
equal but opposite sign reference voltages are available.
Changing the ratio of the two 0.1% resistors allows for
other values: one positive and one negative.
–10
–20
–30
–40
–50
–60
–70
–80
3
4
5
6
7
8
9
10 11 12 13
FREQUENCY (MHz)
AN67 F128
Figure 128. Oscillator Output Spectrum
1.5V
RX (OPTIONAL)
15k
+
0.617V
1/2 LT1112
–
TOTAL SUPPLY CURRENT = 700µA.
LT1004-1.2
20k
0.1%
100pF
–
+
–1.5V
AT ±1.5V: MAXIMUM LOAD
CURRENT = 800µA; CAN BE
INCREASED WITH OPTIONAL RX, RY;
AT RX = RY = 750Ω LOAD
CURRENT = 2mA.
TEMPERATURE COEFFICIENT LIMITED
BY REFERENCE = 20ppm/°C.
1/2 LT1112
RY (OPTIONAL)
WORKS WITH BATTERIES
DISCHARGED TO ±1.3V.
20k
0.1%
– 0.617V
AN67 F129
Figure 129. Dual Output Reference Operates on Two AA Cells
AN67-78
Application Note 67
THREE OP AMP INSTRUMENTATION AMP
USING THE LT1112/LT1114
by George Erdi
4. Low noise: 0.32µV peak-to-peak, 0.1Hz to 10Hz
5. Supply current is 400µA max per amplifier
The LT1112/LT1114 are dual and quad universal precision
op amps. All important precision specifications have been
maintained:
1. Microvolt offset voltage; the low cost grades (including
the small outline, 8-pin surface mount package) are
guaranteed to 75µV.
2. Drift guaranteed to 0.5µV/°C (0.75µV/°C low cost grades)
6. Voltage gain is in excess of one million
The LT1112/LT1114 also provide a full set of matching
specifications, facilitating their use in such matching
dependent applications as the three op amp instrumentation amplifier shown in Figure 130. The performance of
this instrumentation amplifier depends only on the matching parameters not the specifications of the individual
amplifiers.
3. Bias and offset currents are in the picoampere range,
even at 125°C
INPUT –
R4
100Ω
0.5%
+
1/2 LT1112
OR
1/4 LT1114
– A
R6
10k
0.5%
R1
10k
1%
R3
2.1k
1%
–
R8
200Ω
–
INPUT +
1/2 LT1112
OR
1/4 LT1114
+ D
C1
33pF
R10 LT1097
1M OR 1/4 LT1114
B OR C
OUTPUT
+
GAIN = 1000
R2
10k
1%
R5
100Ω
0.5%
TRIM R8 FOR GAIN
TRIM R9 FOR DC COMMON MODE REJECTION
TRIM R10 FOR AC COMMON MODE REJECTION
R7
9.88k
0.5%
R9
200Ω
TYPICAL PERFORMANCE OF THE
INSTRUMENTATION AMPLIFIER:
INPUT OFFSET VOLTAGE = 40µV
OFFSET VOLTAGE DRIFT = 0.3µV/°C
INPUT BIAS CURRENT = 80pA
INPUT OFFSET CURRENT = 100pA
INPUT RESISTANCE = 800GΩ
INPUT NOISE = 0.5µVP-P
AN67 F130
Figure 130. Three Op Amp Instrumetation Amp with Gain = 100
AN67-79
Application Note 67
ULTRALOW NOISE, THREE OP AMP
INSTRUMENTATION AMPLIFIER
by George Erdi and Alexander Strong
+
–
Op amp instrumentation amplifiers usually have op amps
with a fixed gain greater than one at the input stage (Figure
131). At low frequencies, decompensated op amps work
well, but at high frequencies and with one input grounded,
the virtual ground begins to lose its integrity. As the
frequency of the input signal increases, the amplitude at
the virtual ground increases, making the virtual ground
look inductive, eventually requiring a unity-gain stable
amplifier. The LT1028 can be made stable under these
conditions with bypass capacitors and a little experimenting, but the LT1128 is unconditionally stable.
300Ω
10k
LT1128
AC SIGNAL INCREASES WITH
FREQUENCY AT THIS NODE
–
820
–
LT1037
56Ω
+
GAIN = 1000
820
300Ω
LT1128
AC VIN
+
10k
AN67 F131
INPUT REFERRED NOISE = 1.5nV/√Hz at 1kHz
WIDE BAND NOISE = 1.4µVRMS
IF BAND LIMITED TO DC TO 100kHz = 0.6µVRMS
GAIN BANDWIDTH PRODUCT = 400MHz
Figure 131. Three Op Amp, Ultralow Noise
Instrumentation Amplifier
A TEMPERATURE COMPENSATED, VOLTAGECONTROLLED GAIN AMPLIFIER USING THE LT1228
by Frank Cox
Table 1. Characteristics of Example
It is often convenient to control the gain of a video or
intermediate frequency (IF) circuit with a voltage. The
LT1228, along with a suitable voltage-to-current converter circuit, forms a versatile gain control building block
ideal for many of these applications. In addition to gain
control over video bandwidths this circuit can add a
differential input and has sufficient output drive for 50Ω
systems.
Frequency Range
The transconductance of the LT1228 is inversely proportional to absolute temperature at a rate of – 0.33%/°C. For
circuits using closed-loop gain control (i.e., IF or video
automatic gain control) this temperature coefficient does
not present a problem. However, open-loop gain control
circuits that require accurate gains may require some
compensation. The circuit described here uses a simple
thermistor network in the voltage-to-current converter to
achieve this compensation. Table 1 summarizes the circuit’s
performance.
AN67-80
Input Signal Range
Desired Output Voltage
Operating Temperature Range
Supply Voltages
Output Load
Control Voltage vs Gain Relationship
Gain Variation Over Temperature
0.5V to 3.0V pk
1.0V pk
0Hz to 5MHz
0°C to 50°C
±15V
150Ω (75Ω + 75Ω)
0V to 5V Min to Max Gain
±3% from Gain at 25°C
Figure 132 shows the complete schematic of the gain
control amplifier. Please note that these component choices
are not the only ones that will work nor are they necessarily
the best. This circuit is intended to demonstrate one
approach out of many for this very versatile part and, as
always, the designer’s engineering judgment must be fully
engaged. Selection of the values for the input attenuator,
gain-set resistor and current feedback amplifier resistors
is relatively straightforward, although some iteration is
usually necessary. For the best bandwidth, remember to
keep the gain-set resistor, R1, as small as possible and the
set current as large as possible (with due regard for gain
Application Note 67
15V
4.7µF
+
+
R3A
10.7k
3
7
+
1
gm
R2A
10.7k
R3
274
R2
274
–
5
R1
806
4
+
CFA
8
6
ROUT
75
RLOAD
75
–
–15V
4.7µF
VCON
+
–
2
R4
2k
RF
750
RG
82.5
ISET
AN67 F132
Figure 132. Differential Input, Variable Gain Amplifier
compression). The voltage-controlled current source (ISET)
is detailed in the boxed section.
5
Several of these circuits have been built and tested using
various gain options and different thermistor values. Test
results for one of these circuits are shown in Figure 133.
The gain error versus temperature for this circuit is well
within the limit of ±3%. Compensation over a much wider
range of temperatures or to tighter tolerances is possible,
but would generally require more sophisticated methods,
such as multiple thermistor networks.
3
The VCCS is a standard circuit with the exception of the
current set resistor R5, which is made to have a temperature coefficient of – 0.33%/°C. R6 sets the overall gain and
is made adjustable to trim out the initial tolerance in the
LT1228 gain characteristic. A resistor (RP) in parallel with
the thermistor will tend, over a relatively small range, to
linearize the change in resistance of the combination with
temperature. RS trims the temperature coefficient of the
network to the desired value.
4
ERROR (%)
GAIN = –6dB
2
GAIN = 3dB
1
0
–1
GAIN = 6dB
–2
–3
–25 –12.5 0
12.5 25 37.5 50
62.5 75
TEMPERATURE (°C)
AN67 F133
Figure 133. Gain Error for the Circuit in Figure 132 Plus the
Temperature Compensation Circuit Shown in Figure 134
(Normalized to Gain at 25°C)
AN67-81
Application Note 67
Voltage Controlled Current Source (VCCS)
with a Compensating Temperature Coefficient
R6
266k
2.2k3A1
RT
VR
RS
4320
RP
1780
–
R8
150k
LT1006
+
ISET
50pF
R7
2.26M
VCON
ISET (MAX)
R5
2N3906
ISET
ISET =
R6 VC VR
+
R5 R8 R7
VR = REF VOLTAGE
ISET (MIN)
AN67 F134
5
0
Figure 134. Voltage Controlled Current Source (VCCS) with a
Compensating Temperature Coefficient
VCON (V)
AN67 F135
Figure 135. Voltage Control of ISET with Temperature
Compensation
VCCS Design Steps
1. Measure or obtain from the data sheet the thermistor
resistance at three equally spaced temperatures (in
this case 0°C, 25°C and 50°C). Find RP from:
RP =
(R0 • R25 + R25 • R50 – 2 • R0 • R50)
(R0 + R50 – 2 • R25)
2. Resistor RP is placed in parallel with the thermistor.
This network has a temperature dependence that is
approximately linear over the range given (0°C to
50°C).
3. The parallel combination of the thermistor and RP
(RP||RT) has a temperature coefficient of resistance
(TC) given by:
)
))
AN67-82
)
R0||RP – R50||RP
100
R25||RP
THIGH – TLOW
where: THIGH = the high temperature
TLOW = the low temperature
RT = the thermistor
)
)
TC RP||RT
(RP||R25) – (RP||RT)
–0.33
5. R6 contributes to the resultant temperature and so is
made large with respect to R5.
where: R0 = thermistor resistance at 0°C
R25 = thermistor resistance at 25°C
R50 = thermistor resistance at 50°C
TC RP||RT =
4. The desired temp. co. to compensate the LT1228 gain
temperature dependence is – 0.33%/°C. A series resistance (RS) is added to the parallel network to trim
its TC to the proper value. RS is given by:
6. The other resistors are calculated to give the desired
range of ISET.
This procedure was performed using a variety of thermistors (one possible source is BetaTHERM Corporation—phone 508-842-0516). Figure 5 shows typical
results reported as errors normalized to a resistance with
a – 0.33%/°C temperature coefficient. As a practical
matter, the thermistor need only have about a 10%
tolerance for this gain accuracy. The sensitivity of the
gain accuracy to the thermistor tolerance is decreased by
the linearization network, in the same ratio as is the
temperature coefficient; the room temperature gain may
be trimmed with R6. Of course, particular applications
require analysis of aging stability, interchangeability,
package style, cost, and the contributions of the tolerances of the other components in the circuit.
14
4
12
2
0
10
ERROR (%)
RESISTANCE (kΩ)
Application Note 67
8
COMPENSATED NETWORK
6
4
–4
–6
–8
2
–10
THERMISTOR
0
–10
–2
0
10
20
30
40
50
60
–12
–60 –40
70
–20
TEMPERATURE (°C)
20
0
40
60
80
TEMPERATURE (°C)
AN67 F137
Figure 136. Thermistor and Thermistor Network Resistance
vs Temperature
THE LTC1100, LT1101 AND LT1102: A TRIO OF
EFFECTIVE INSTRUMENTATION AMPLIFIERS
by George Erdi
Next to the universally used op amp, perhaps the most
useful linear IC building block is the instrumentation amp,
or “IA.” Using IAs effectively can in some ways be more
challenging than selecting op amps, because IAs have
different specs and can also use different topologies.
However, the basic task is a fixed gain, differential input,
single-ended output amplifier, the definition of an IA. The
differential signal typically rides on top of a common mode
signal; the differential input is amplified and the common
mode voltage is rejected by the IA.
The instrumentation amplifier can be implemented with
dedicated IA designs, or with one to three op amps to
realize the gain function, and a minimum of four ratiomatched precision resistors configured as two like ratio
pairs.
The most familiar IA type is the single op amp variety,
usually called a difference amplifier and shown in Figure
138. Using just two parts (one op amp and one resistor
network), this IA is the height of simplicity and utility. For
modest requirements it is built with just a general purpose
op amp and four precision resistors. A drawback to this
type of IA is that the resistor bridge loads the source. The
three op amp configuration uses seven resistors and has
high input impedance. It is obviously more difficult to
AN67 F137
Figure 137. Thermistor Network Resistance Normalized to a
Resistor with Exact – 0.33%/°C Temperature Coefficient
implement than the single op amp version. A nice compromise between these two approaches is illustrated in Figure
139. This IA design uses two op amps to buffer the signal
inputs and requires only four resistors. The use of two op
R1
–
R2
–
INPUT
A
+
OUTPUT
R3
+
R4
REF
G = GAIN = R2
R1
R2 ≡ R4
R1 R3
AN67 F138
Figure 138. Basic Single Op Amp Instrumentation Amplifier
REF
R1
R2
R3
R4
OUTPUT
–
–
A
–
+
B
+
INPUT
+
G = GAIN = 1 +
R4
WHEN R4 ≡ R1
R3
R3 R2
AN67 F139
Figure 139. Buffered Dual Op Amp Instrumentation Amplifier
AN67-83
Application Note 67
amps with modern dual devices causes no penalty, and in
fact this arrangement has real virtues over the more basic
setup of Figure 138.
points available for user connections. The key specifications of these three devices are summarized in Table 1.
Table 1. LTC Instrumentation Amplifier Specifications1
This IA architecture presents minimum loading to the
differential source, namely the bias current of the op amp
used, which is balanced between the two inputs. The
resistor network needs very precise trimming for high
common mode rejection (CMRR) and gain accuracy. The
trimming is noninteractive; first the R4/R3 ratio is trimmed
for gain accuracy then the R1/R2 ratio is trimmed for high
CMRR. Trimming compensates not only for resistor inaccuracies, but also for the finite gain and CMRR of the op
amps. The amplified difference appears between the output terminal and the voltage applied to the REF terminal
(normally grounded).
LTC1100C
LT1101C/I/M
LT1102C/I/M
Available Gains
1002
10/100
10/100
Gain Error (%)
0.01
0.01
0.01
Gain Nonlinearity (ppm)
3
3
7
Gain Drift (ppm/°C)
2
2
10
VOS (µV)
VOS Drift (µV/°C)
IB (pA)
IOS (pA)
60
200
0.5
3
2.5
6000
4
10
150
4
1.9µVP-P
0.9µVP-P
20nV/(Hz)1/2
(DC to 10Hz) (0.1Hz to 10Hz)
(at 1kHz)
en
As a basic building block, this IA can be performance
optimized for various applications by a choice of op amps.
LTC has taken this step with the LTC1100, LT1101 and
LT1102, an instrumentation amplifier series offered in an
8-pin footprint with connections as shown in Figure 140.
As illustrated, the gain of these IAs is user programmed by
taps on the resistor array, for pre-trimmed precision gains
of either 10 or 100 for the LT1101 and LT1102. The 8-pin
LTC1100 has a fixed gain of 100, but makes the summing
1
0.005
CMRR (dB)
110
112
98
PSRR (dB)
130
114
102
4V to 18V
(Single/Dual)
1.8V to 44V
(Single/Dual)
10V to 44V
(Dual)
2.4
0.09
3.4
Gain Bandwidth (MHz)
2
0.37
35
SR (V/µs)
4
0.1
VS (Total, Mode)
IS (mA)
for LT1101/LT1102 and ±5V for LTC1100.
LT1101 AND LT1102
8
99R
GROUND 1
(REF)
OUTPUT
8
90R
99R
2
7
9R
9R
R
R
–
–
INVERTING 3
INPUT
V–
+
4
–
A
6 NONINVERTING
INPUT
5
INVERTING 3
INPUT
V–
V+
+
G = 10, SHORT PIN 2 TO PIN 1
SHORT PIN 7 TO PIN 8
}
B
6 NONINVERTING
INPUT
+
5
4
G = 100, NO ADDITIONAL CONNECTIONS
LT1101 AND LT1102 ONLY
R ≈ 1.8k FOR LTC1100 AND LT1102
R ≈ 9.2k FOR LT1101
Figure 140. Instrumentation Amplifiers in 8-Pin Packages
AN67-84
R
–
B
+
OUTPUT
90R
2
7
A
A = 25°C. VS = ±15V
2 A gain option of 10/100 is available in LTC1100CS (16-Lead SW)
LTC1100
GROUND 1
(REF)
R
30
1 Unless otherwise stated all specifications are typical at T
AN67 F140
V+
Application Note 67
With these three IA choices, the user can optimize performance for a variety of factors. The LTC1100 operates with
dual or single supplies ranging from 4V to 18V, whereas
the LTC1101 accepts a supply range of from 1.8V to 40V.
In addition, the LT1101 consumes only 100µA standby
current. For applications that require very low offset
voltage and drift, the LTC1100 excels with 1µV of offset
and 5nV/°C drift. Where both high speed and low bias
current are important, the LT1102 is the IA of choice, albeit
at a cost of slightly higher power consumption and dual
supplies. As can be seen from the table, all of these devices
are outstanding with regard to gain accuracy, linearity and
stability. The LTC1100, which is based on a dual chopper
amplifier prototype (the LTC1051), is by far the best in
terms of offset and drift. Either the LTC1100 or the LT1102
could be the unit of choice in terms of lowest bias current,
with the LT1102 gaining an edge at higher temperatures.
Applications Considerations
While this IA type is generally outstanding in terms of
performance and simplicity, independent of the op amps,
some caveats apply to using it most effectively. One
concern is AC CMRR. As noted in Figure 140, the first op
amp (A) is configured for unity gain, and the second op
amp (B) provides all of the voltage gain. This has the effect
of making the respective CMRR’s frequency mismatched,
since the CMRR of the higher gain, “B” side, corners at a
much lower frequency. The resulting differential CMRR
will therefore degrade more quickly with frequency than
that of a topology with better AC balance. On the LT1102
this problem is resolved by decompensating amplifier B to
gain-of-ten stability. This increases slew rate and bandwidth and also matches the CMRR rolloff with the frequencies of the two op amps when G = 10. At a gain of 100, this
rolloff match no longer holds. However, connecting an
18pF capacitor between Pins 1 and 2 matches the CMRRs
120
COMMON MODE REJECTION RATIO (dB)
It is apparent from Table 1 that for these three IAs, there
are no output contributions to input errors. With dedicated
IA’s or with the three op amp configuration there are
separate specifications for input and output offset voltage,
input and output drift and noise, and input and output
power supply rejection ratio. To calculate system errors
these input and output terms must be combined. With the
LTC1100/01/02 these error calculations are simple.
G = 100
100
G = 10
80
G = 100
18pF PIN 1 TO PIN 2
60
40
20
0
0
10
100
1k
10K
100K
1M
FREQUENCY (Hz)
AN67 F141
Figure 141. LT1102 Common Mode Rejection Ratio vs Frequency
of the two sides and improves CMRR by an order of
magnitude in the 300Hz to 30kHz range (Figure 141). As
shown on the LTC1100 and LTC1101 data sheets, similar
improvements can be obtained from those devices by
connecting external capacitors.
The LTC1100 and LT1101 also present some important
usage considerations because of their single supply abilities, i.e., when operating with the V – terminal tied to
ground. In this configuration, these devices handle CM
inputs near ground and voltage swings to ground and their
reference terminals can be tied to ground. One of the most
common uses of these two IAs is as bridge amplifiers in
conjunction with single supply powered DC strain gauges.
As such, these IAs have a unique ability to deliver high gain
with precision, while operating with a 1/2 supply voltage
CM input. At first glance, it appears that a dual supply IA
could operate, for example, on a 9V battery supply with
4.5V common mode input, but its output will not swing to
ground and its reference terminal cannot be tied to ground.
For SPICE simulation purposes, a model for the LT1101 is
included in the LTC macromodel library. The model is
configured as the resistor network shown for the LT1101,
combined with a model for the LT1078. A similar model for
the LTC1100 can be made by scaling the four resistors
appropriately, and using an LTC1051 model from the
same library. A close model approximation for the LT1102
can be made with the LT1102 resistor values, combined
with an LT1057 model for the “A” side, and a LT1022
model for the “B” side (both also in the library).
AN67-85
Application Note 67
Miscellaneous Circuits
sensitive to loading, the crystal current represents essentially a filtered version of the voltage waveform and is
relatively tolerant of loading effects.
DRIVING A HIGH LEVEL DIODE RING MIXER
WITH AN OPERATIONAL AMPLIFIER
by Mitchell Lee
The impedance, and therefore the voltage at the bottom of
the crystal, is kept low by injecting the current into the
summing node of an LT1206 current feedback amplifier.
Loop gain reduces the input impedance to well under 1Ω.
Oscillator bias is adjustable, allowing control of the mixer
drive. This also provides a convenient point for closing an
output power servo loop.
One of the most popular RF building blocks is the diode
ring mixer. Consisting of a diode ring and two coupling
transformers, this simple device is a favorite with RF
designers anywhere a quick multiplication is required, as
in frequency conversion, frequency synthesis or phase
detection. In many applications these mixers are driven
from an oscillator. Rarely does anyone try building an
oscillator capable of delivering 7dBm for a “minimum
geometry” mixer, let alone one of higher level. One or more
stages of amplification are added to achieve the drive level
required by the mixer. The new LT1206 high speed amplifier makes it possible to amplify an oscillator to 27dBm in
one stage.
Operating from ±15V supplies, the LT1206 can deliver
32dBm to a 50Ω load, and with a little extra headroom (the
absolute maximum supply voltage is ±18V), it can reach
2W output power into 50Ω. Peak guaranteed output
current is 250mA.
Shown in Figures 144 to 148 are spectral plots for various
combinations of single and double termination at power
levels ranging from +17dBm to +27dBm — not bad for an
inductorless circuit. Double termination may be used to
present a 50Ω source impedance to the mixer, or to isolate
two or more mixers driven simultaneously from one
LT1206 amplifier.
Figure 143 shows the complete circuit diagram for a
crystal oscillator, LT1206 op amp/buffer and diode-ring
mixer. Most of the components are used in the oscillator
itself, which is of the Colpitts class. Borrowing from a
technique used in Hewlett Packard’s Unit Oscillator, the
current of the crystal is amplified rather than the voltage.
There are several advantages to this method, the most
important of which is low distortion. Although the voltages
present in this circuit have poor wave shape and are
10nF
15V
Although a 10MHz example has been presented here, the
LT1206’s 65MHz bandwidth makes it useful in circuits up
to 30MHz. In addition, the shutdown feature can be used
to interrupt drive to the mixer. When the LT1206 is shut
15V
100nF
1k
PN4416
24k
10nF
SHUTDOWN
100nF
62pF
120k
10MHz
2
110nF
+
4
LT1206
10k
1
–
6
3
OPERATE
7
51Ω, 2W
OPTIONAL DOUBLE
TERMINATION RESISTOR
RF
LO
IF
5 10nF
620Ω
10nF
5k
100nF
–15V
–15V
OSCILLATOR (TYPICAL)
BUFFER/AMPLIFIER
Figure 143. Oscillator Buffer Drives +17dBm to + 27dBm Double Balanced Mixers
AN67-86
AN67 F143
DIODE RING MIXER
Application Note 67
making it ideal for this application. Another nice feature is
the LT1206’s ability to drive heavy capacitive loads while
remaining stable and free of spurious oscillations.
down, the oscillator will likely stop, since the crystal then
sees a series impedance of 620Ω and the mixer itself.
Upon re-enabling the LT1206 there will be some time
delay before the oscillator returns to full power. The circuit
works equally well with an LC version of the oscillator.
For mixers below +17dBm, the LT1227 is a lower cost
alternative, featuring 140MHz bandwidth in combination
with the shutdown feature of the LT1206.
Note that the current feedback topology is inherently
tolerant of stray capacitive effects at the summing node,
44
34
+
24
–
14
OUTPUT (dBm)
+30dBm
LT1206
50Ω
620Ω
4
–6
–16
–26
–36
–46
–56
5
15
25
35
45
55
65
75
85
95
FREQUENCY (MHz)
105
AN67 F144
Figure 144. Spectrum Plot of Figure 143’s Circuit Driving + 30dBm into a 50Ω Load (Single Termination)
40
30
+
20
–
10
OUTPUT (dBm)
+27dBm
LT1206
50Ω
620Ω
0
–10
–20
–30
–40
–50
–60
5
15
25
35
45
55
FREQUENCY (MHz)
65
75
85
95
105
AN67 F145
Figure 145. Spectrum Plot of Figure 143’s Circuit Driving + 27dBm into a 50Ω Load
Information furnished by Linear Technology Corporation is believed to be accurate and reliable.
However, no responsibility is assumed for its use. Linear Technology Corporation makes no representation that the interconnection of its circuits as described herein will not infringe on existing patent rights.
AN67-87
Application Note 67
40
30
+
20
–
10
OUTPUT (dBm)
+23dBm
LT1206
50Ω
620Ω
0
–10
–20
–30
–40
–50
–60
5
15
25
35
45
55
65
75
85
95
FREQUENCY (MHz)
105
AN67 F146
Figure 146. Spectrum Plot of Figure 143’s Circuit Driving + 23dBm into a 50Ω Load
40
30
+
20
–
10
OUTPUT (dBm)
50Ω
+27dBm
LT1206
50Ω
620Ω
0
–10
–20
–30
–40
–50
–60
5
15
25
35
45
55
65
75
85
95
FREQUENCY (MHz)
105
AN67 F147
Figure 147. Spectrum Plot of Figure 143’s Circuit Driving + 27dBm into a 50Ω, Double Terminated Load
40
30
+
20
–
10
OUTPUT (dBm)
50Ω
+17dBm
LT1206
50Ω
620Ω
0
–10
–20
–30
–40
–50
–60
5
15
25
35
45
55
FREQUENCY (MHz)
65
75
85
95
105
AN67 F148
Figure 148. Spectrum Plot of Figure 143’s Circuit Driving + 17dBm into a 50Ω, Double Terminated Load
Linear Technology Corporation
McCarthy Blvd., Milpitas, CA 95035-7417
AN67-88 1630
(408) 432-1900
: (408) 434-0507
: 499-3977
●
FAX
●
TELEX
LT/GP 0996 5K • PRINTED IN USA
 LINEAR TECHNOLOGY CORPORATION 1996