Appendix D: New Products Appendix to the RMS-to-DC Conversion Application Guide (October 2002)

Appendix D
October 2002
New Products Appendix to the RMS
to DC Conversion Application Guide
INTRODUCTION*
Both the AD736 and AD737 are optimized for use in
portable instruments; they consume less than 200 µA
of quiescent current and accept signal levels from
0 mV rms to 200 mV rms. The AD737 also has a powerdown input that allows the user to reduce its quiescent
current from 160 µA to 40 µA in portable applications.
Since the last printing of this Applications Guide in
1986, Analog Devices has introduced four new rms
products. These devices supplement ADI’s original
rms products: the AD536A, AD636, and AD637.
The AD736 and AD737 are low power, low cost rms
converters designed for use in portable instruments.
The AD637 should be chosen if the application requires
high accuracy and a quick response for large, abrupt
changes in signal level. The AD637’s settling time is
independent of signal level, while for a given value
of averaging capacitor, the settling times of the
AD536A, AD636, AD736, and AD737 will be longer
for low level signals and shorter for high level signals.
™
The AD8361 and AD8362 TruPwr RF power detectors are designed for accurate control of radiated power
levels in cellular, broadband, CATV, MMDS, and
LMDS communications equipment.
HOW TO SELECT AN RMS-TO-DC CONVERTER
Some monolithic rms converters use a sigma-delta
computational technique. This method can provide
wide bandwidth while operating at low supply current
levels but suffers from a very serious low frequency
“rumble.” In effect, as the input signal frequency
increases, a larger percentage of the input signal is aliased
down to dc, producing a low frequency modulation at
the output of the converter. This rumble is a low
frequency error that increases both with rms signal
level and signal frequency. When this type of rms
converter is used in a DVM, the lower digits on the
display will flicker and, as the rms level or frequency
increases, more and more display digits will become
erratic. Attempts to minimize this problem using a
low-pass “rumble filter” after the converter will
result in longer settling times between readings and
will, in any case, be generally ineffective in removing the
flicker. In contrast, the AD736/AD737 architecture
is continuous-time and not subject to this type of
unstable behavior.
Selecting an rms-to-dc converter means picking the
product whose attributes best match the requirements
of the application. Unfortunately, no one converter
fits every situation so trade-offs must be made between
accuracy, bandwidth, power consumption, input signal
level, crest factor, settling time, and cost.
The AD637 accepts input voltages as high as 7 V rms
and is Analog Devices’ most accurate and widest
bandwidth rms-to-dc converter. Its –3 dB bandwidth
is 8 MHz for a 1 V rms input. It has an auxiliary dB
output and a power-down feature that reduces its
quiescent current from 3 mA to 450 µA.
The AD536A and its 200 mV companion product,
the AD636, are designed for low cost, general-purpose
applications. These converters offer true rms conversion accuracy and will accurately process input signals
with high crest factors. The AD536A has a 2 V rms
full-scale range, while that of the AD636 is 200 mV rms.
Both products consume only about 1 mA of power
supply current. They also offer a dB output feature
with an output that is proportional to the logarithm of
the rms input signal.
*Portions of this appendix are excerpts from Application Note AN-268, RMS to DC Converters Ease Measurement Tasks, by Bob Clarke,
Mark Fazio, and Dave Scott: Analog Devices Publication.
TruPwr is a trademark of Analog Devices, Inc.
1
Table I provides a quick performance comparison between Analog Devices’ rms products.
Table I. RMS Converter Comparison Table
Input Dynamic Range
Nominal RMS Full Scale
Peak Trans. Input
Max Total Error
No Ext. Trim
–3 dB Bandwidth
Full Scale
0.1 V rms
Error at Crest
Factor of 5
For V rms
Power Supply
Volts
Current
AD536AJ
AD637J
AD636J
AD736J
AD737J
7 V rms
2V
± 20 V
5 mV
± 0.5% rdg
7 V rms
2V
± 15 V
1 mV
± 0.5% rdg
1 V rms
200 mV
± 2.8 V
0.5 mV
± 1% rdg
1 V rms*
200 mV
± 2.7 V
0.5 mV
± 0.5% rdg
1 V rms*
200 mV
± 2.7 V
0.4 mV
± 0.5% rdg
2 MHz
300 kHz
8 MHz
600 kHz
1.3 MHz
800 kHz
190 kHz
170 kHz
190 kHz
170 kHz
–0.3%
@1 V
± 0.15%
@1 V
–0.5%
@200 mV
± 2.5%
@200 mV
± 2.5%
@200 mV
± 13 V
to ± 18 V
1 mA
± 13 V
to ± 18 V
2 mA
+2 V, –2.5V
to ± 12 V
800 µA
+2.8, –3.2 V
to ± 16.5 V
230 µA
+2.8, –3.2 V
to ± 16.5 V
170 µA
*± 5 V to ± 16.5 V dual-supply operation only. 200 mV rms under +5 V, –3 V single-supply conditions.
either as a true rms converter or as an average responding
(MAD) rectifier. The AD736 and AD737 converters
are optimized for low power operation, and their
averaging capacitor appears directly across a diode in
the rms core. Because of this, the averaging time constant
will increase as the rms input level decreases. Consequently, lower input levels allow the circuit to perform
better (due to increased averaging) but result in longer
settling times, requiring more time between readings.
THE AD736 AND AD737 LOW COST
RMS CONVERTERS
As shown in Figures 1 and 2, the AD736 and AD737
architectures are very similar. The major difference
between these two products is that the AD736 includes
an output buffer amplifier for general-purpose applications, while the AD737 is unbuffered for low power
operation. The AD737 also includes a power-down
feature that further reduces its standby current consumption to a mere 25 µA. Both products can be operated
8
COM
+VS
7
+VRMS
OUTPUT
6
CC
1
8k⍀
8k⍀
VIN
FET
OP AMP
IB<10pA
CF
3
CF
2
CURRENT
MODE
RECTIFIER
CAV
5
CAV
4
RMS TRANSLINEAR CORE
Figure 1. Simplified Schematic of the AD736
2
–VS
The input to the AD736 and AD737 is through a
FET input op amp connected as a unity-gain buffer.
This amplifier allows both a high impedance, buffered
input (Pin 2) or a low impedance input (Pin 1) that
provides a wider dynamic range. The high impedance
input, with its low input bias current, is well suited for
use with high impedance input attenuators.
The design of the AD737 is very similar to that of the
AD736. In order to reduce power consumption, the
AD737 does not have an output buffer amplifier.
Instead, it uses an NPN transistor to drive an 8 kΩ
internal load resistor.
The converter develops its output voltage by sinking
current through this resistor. The external averaging
capacitor (CAV) for the AD736 and AD737 is connected
between Pins 4 (–VS) and 5 (CAV), which places CAV
across a transistor’s base-emitter junction in the rms
core. This means that a diode is in parallel with the
averaging capacitor; the resulting time constant is
therefore inversely proportional to the rms value.
The output of the buffer drives a current mode
rectifier (absolute value circuit) that in turn drives
an rms core.
In the AD736 (Figure 1), the output of the rms core
drives the summing node of an inverting op amp connected as a current-to-voltage converter. Pin 3 gives
access to this node to connect a filter capacitor, CF,
in parallel with the 8 kΩ feedback resistor, to form a
one-pole low-pass filter.
8
COM
+VS
7
CC
1
8k⍀
8k⍀
VIN
6
OUTPUT
(–VRMS Out)
2
FET
Op Amp
IB<10pA
CURRENT
MODE
RECTIFIER
POWER
DOWN
3
CAV
BIAS
SECTION
5
4
–VS
RMS TRANSLINEAR CORE
Figure 2. Simplified Schematic of the AD737
3
CAV
1V
Because the external averaging capacitor, CAV, “holds”
the rectified input signal during rms computation, its
value directly affects the accuracy of the measurement—
especially at low frequencies. (The larger the value of
CAV, the lower the error.) Also, because the averaging
capacitor appears across a base-emitter junction in the
squarer/divider, the averaging time constant will
increase linearly as the input signal is reduced.
VS = ⴞ5V, CC = 22␮F, CF = 0␮F
INPUT LEVEL – rms
100mV
AD736/AD737 settling time versus rms input level
and CAV is shown in Figure 3.
CAV = 100␮F
CAV = 33␮F
10mV
CAV = 10␮F
1mV
Due to the varying time constant, as the input level
decreases, errors due to nonideal averaging will decrease,
while the time it takes for the circuit to settle to the
new rms level will increase.
100␮V
1ms
Therefore, lower input levels allow the circuit to perform better (due to increased averaging) but increase
the waiting time between measurements, because the
capacitor takes longer to charge or discharge. Thus, a
trade-off between computational accuracy and settling
time is required.
10ms
100ms
1s
SETTLING TIME
10s
100s
Figure 3. Settling Time vs. RMS Input Level of
the AD736 and AD737 for Various Values of CAV
CALCULATING AD737 SETTLING TIME
The graph of Figure 3 may be used to approximate
the time required for the AD736 or AD737 to settle
when its input level is reduced in amplitude. The total
time required for the rms converter to settle will be
the difference between two settling times extracted
from the graph—the initial settling time minus the final
settling time.
Table II provides practical values of CAV and CF for
several common applications.
Table II. Practical Values for CAV and CF for AD736 and AD737
Application
RMS
Input Level
General-Purpose
RMS
Computation
0 V–1 V
General-Purpose
Average
Responding
0 V–1 V
SCR Waveform
Measurement
0 mV–200 mV
0 mV–200 mV
0 mV–200 mV
0 mV–100 mV
Audio
Applications
Speech
Music
0 mV–200 mV
0 mV–100 mV
Low
Frequency
Cutoff
( –3 dB)
Max Crest
Factor
CAV
CF
Settling
Time*
to 1%
20 Hz
200 Hz
20 Hz
200 Hz
5
5
5
5
150 µF
15 µF
33 µF
3.3 µF
10 µF
1 µF
10 µF
1 µF
360 ms
36 ms
360 ms
36 ms
33 µF
3.3 µF
33 µF
3.3 µF
1.2 sec
120 ms
1.2 sec
120 ms
20 Hz
200 Hz
20 Hz
200 Hz
50 Hz
60 Hz
50 Hz
60 Hz
5
5
5
5
100 µF
82 µF
50 µF
47 µF
33 µF
27 µF
33 µF
27 µF
1.2 sec
1.0 sec
1.2 sec
1.0 sec
300 Hz
20 Hz
3
10
1.5 µF
100 µF
0.5 µF
68 µF
18 ms
2.4 sec
*Settling time is specified over the stated rms input level with the input signal increasing from zero. Settling times will be greater for decreasing amplitude input.
4
Normally, the input offset errors in the traditional
monolithic rms converter will create a region of diode
nonconduction at low level input voltages. That is,
any input voltages that are smaller than the input
offset voltage will not be rectified and a “dead zone”
is created.
As an example, consider the following conditions:
a 33 µF averaging capacitor an initial rms input level
of 100 mV and a final (reduced) input level of 1 mV.
From Figure 3, the initial settling time (where the
100 mV line intersects the 33 µF line) is around 80 ms.
The settling time corresponding to the new or final
input level of 1 mV is about 8 seconds. Therefore, the
net time for the circuit to settle to its new value will be
dominated by the final settling time. Figure 4 shows
the additional error versus the crest factor of the
AD736 and AD737 for various values of CAV.
However, the AD736 and AD737 are specifically
designed to eliminate this problem. The maximum
input offset voltage of these rms converters is 3 mV. If
Pin 1 is directly grounded, this offset voltage will limit
the converter’s low level resolution. However, as
shown in Figure 5, the use of capacitor CC between
Pin 1 and ground will ac couple the low impedance
input pin and “float” this input above ground. This
prevents any dc currents from flowing through the
8 kΩ internal resistor and creating an input voltage
offset. Capacitor CC should be chosen to provide a
low frequency cutoff substantially below the lowest
signal input frequency.
6
ADDITIONAL ERROR – % of Reading
CAV = 10␮F
CAV = 33␮F
5
3ms BURST of 1kHz,
= 3 CYCLES
200mV RMS SIGNAL
VS = ⴞ5V
CC = 22␮F, CF = 100␮F
4
3
CAV = 100␮F
The 3 db roll-off frequency of CC =
2
1
1
2 π (8 ,000 Ω) (CC in Farads)
CAV = 250␮F
For most applications, a value of 10 µF
(FCUTOFF = 2 Hz) will suffice. A good rule of thumb
is to use a value of CC approximately equal to onethird that of CAV.
0
1
2
3
4
CREST FACTOR – V PEAK /V RMS
5
Figure 4. Additional Error vs. Crest Factor of the
AD736 and AD737 for Various Values of CAV
CC
AC Coupling Design Considerations
The AD736 and AD737 rms converters offer the designer
the option of ac coupling both the input signal and
the dc offset voltages on the rms converter’s input stage.
CC
CIN
VIN
CC
2
RIN
3
8k⍀
AD736/AD737
1
2
3
4
COM
4
8
VIN
VIN
INPUT
AMPLIFIER
BIAS
SECTION
FULL
WAVE
RECTIFIER
7
RMS CORE
6
AD736/AD737
COM
8
VIN
1M⍀
CC
8k⍀
1
INPUT
AMPLIFIER
BIAS
SECTION
FULL
WAVE
RECTIFIER
7
RMS CORE
6
5
Figure 6. AC Coupling Using Capacitor CIN
Figure 6 shows ac input coupling for Pin 2. Capacitor
CIN is necessary if the input signal is an ac waveform
riding on a dc voltage, or if the rms converter is operating from a single-supply voltage. In this case, Pin 1 will
be “floating” above ground and CIN is needed to prevent
the rms converter from full-wave rectifying the differential voltage between Pins 1 and 2, which often will
result in input overload. A resistor is needed between
Pin 2 of the rms converter and ground to provide a dc
return path for input bias currents. Note that capacitor
CC is still needed to prevent input offset voltage errors.
5
Figure 5. Using Capacitor CC to Block Internal
Offset Voltage Errors
5
APPLICATIONS OF THE AD736 AND AD737
AD736 as Precision Rectifier
Figure 8 shows the AD736’s performance as a precision
rectifier.
Building a precision rectifier from discrete components
requires two op amps, two diodes, and a handful of
matched resistors. An easy way to replace all these parts
and save some board space is to use an rms-to-dc
converter IC. Just omit the averaging capacitor and
disconnect the feedback; this uses only the converter’s
internal precision rectifier (Figure 7), which, being
monolithic, has inherently matched diodes.
200mV
100mV
2µS
100mV
200mV
10µS
100
90
CC
10
0%
VIN
1
CC
2
VIN
1M⍀
3
CF
4
–VS
COMMON 8
+VS 7
AD736
OUTPUT
6
0.1␮F
+5V
VOUT
CAV 5
0.1␮F
100
90
–5V
Figure 7. AD736 Connected as a Precision Rectifier
A precision rectifier circuit must provide enough gain to
forward bias its rectifier diodes. So, as the input signal
gets smaller in amplitude, more gain is needed. The
traditional circuit uses an op amp to provide this gain.
However, it will usually have a fixed gain/bandwidth
product, which means that the rectifier’s bandwidth
will change with the input signal level. In contrast to a
discrete circuit, the internally trimmed, monolithic
design of the AD736 and AD737 greatly helps to
minimize this effect.
10
0%
Figure 8. Performance of AD736 Precision
Rectifier at 1 kHz (Top) and 19 kHz (Bottom)
6
R7 and R8 form a voltage divider to allow operation
from a single-supply voltage or battery. Capacitors
C4 and C5 bypass any signal currents on VS or VS/2
to ground.
True RMS and Average Value Circuit
Figure 9 shows a circuit that measures both the true
rms value and the rectified average value of an ac signal.
This design uses two low cost ICs in SOIC packages
and consumes only 180 µA quiescent current. Operating from a 5 V single supply, this circuit has an input
dynamic range from below 30 mV to greater than 3 V
rms. Sine wave accuracy is quite good (see performance data below) and bandwidth is approximately
100 kHz, depending on the input level. The circuit can
also measure a 1 V rms, crest factor of 5 pulse train
with less than 1% of reading error.
The rms converter IC has two inputs: a high impedance (1012 Ω) input (at Pin 2) and an 8 kΩ, wide
dynamic range input via Pin 1. The rms converter’s
full-scale input range is normally 200 mV. This can
be greatly increased by adding an external resistance,
in this case resistor R1 and trimpot R2, between the
signal input and Pin 1. This has the added advantage
of increasing the circuit’s input impedance.
Average responding measurements and rms have
traditionally used different circuits. However, in some
cases, it may be extremely useful to know both the
rms and rectified average value of an ac waveform.
The ratio of rms to rectified average value is one way
to determine the characteristics of a particular waveform without actually seeing it on an oscilloscope.
For example, the rms/average value ratio for a 1 V
peak undistorted sine wave is 0.707 V/0.636 V or 1.11,
a symmetrical square wave is 1.0, a triangular wave is
1.155, and Gaussian noise is 1.253.
The AD737JR measures the true rms value when
switch SW1 connects its averaging capacitor, CAV, to
Pin 5. The averaging capacitor performs the “mean”
portion of the rms (root-mean-square) function.
Removing CAV, by opening SW1, converts the circuit
to rectified average value operation. Resistor R6 allows
a small leakage current to flow past the switch, keeping
the capacitor charged and preventing any large surge
currents from flowing into or out of CAV when the
switch is closed.
The rms value of a sine wave is 0.707 V peak while the
rectified average value is 0.636 V peak. This ratio of
0.707 V/0.636 V is equivalent to an 11% scale factor
difference between the two measurement methods. If it is
Circuit Operation
As shown in Figure 9, an AD737 rms converter IC
drives an AD8541AR micropower op amp. Resistors
INPUT SCALEFACTOR ADJ
+5V
C1
0.47␮F
R1
69.8k⍀ 1%
R2
5k⍀
INPUT
C2
0.47␮F
CF
CC
COM
VIN
+VS
0.01␮F
R3
78.7k⍀
AD737JR
CF
80.6k⍀
R4
5k⍀
R5
OUTPUT ZERO ADJ
+5V
0.01␮F
OUTPUT
OUTPUT
AD8541AR
CAV
–VS
C3
0.01
␮F
SW1
33␮F
+
RMS
AVER
CAV
+
2.2␮F
C4
R6
100k⍀
+
–3dB BW
CAV
CF*
10Hz
68␮F
0.82␮F
20Hz
33␮F
0.47␮F
100Hz
6.8␮F
0.1␮F
*CAV IS DISCONNECTED
IN THE AVERAGE VALUE MODE.
THEREFORE, THE OUTPUT RIPPLE
WILL BE NOTICEABLY HIGHER AT VERY
LOW FREQUENCIES. SIMPLY INCREASE
THE VALUE OF CF TO REDUCE RIPPLE
TO THE DESIRED LEVEL.
1␮F
C5
+5V
R7
100k⍀
+2.5V
R8
100k⍀
Figure 9. An RMS/Average Rectified Value Measurement Circuit
7
Measured Performance Data
desired to have this circuit accurately read the rms value
for sine waves in the rectified average value mode, SW1
can be a two-pole switch. The second pole can connect
a 523 kΩ 1% resistor in parallel with R1 to increase
the scale factor by 11% in the average value mode.
1 kHz Sine Wave Accuracy
VIN is in ac volts rms as monitored by Keithley 191
DVM in ac mode. 5 VDC supply.
The AD737JR drives the AD8541AR op amp with a
negative flowing output current. The op amp operates
as a current-to-voltage converter and also inverts the
signal, providing an output voltage that swings more
positive with increasing input level. Resistor R5’s
value of 80 kΩ matches the effective input resistance
of the AD737 (R1 + R2 + 8 kΩ) so that input/output
scaling is 1:1. Resistor R3 and trimpot R4 cause a
current to flow from the supply to the op amp summing junction. This offsets the op amp output such
that the circuit’s output is approximately zero with no
voltage applied. Note that this circuit has a maximum
supply voltage limit of 5.5 V; operation may be extended
up to 12 V by substituting an OP-196GS op amp for
the AD8541AR.
VIN
VOUT rms
VOUT Rectified
Average
Value
3V
1V
0.3 V
0.1 V
0.03 V
2.9999
1.0027
0.30201
0.10082
0.02960
2.6762
0.8947
0.2698
0.09947
0.02956
Error versus Crest Factor. +5 VDC Supply,
1 V rms, 100 ␮s pulse.
Duty Cycle Varied for Desired Crest Factor.
Circuit calibration:
Crest Factor
% of Reading Error
3
5
10
0.67%
0.98%
4.7%
1. Adjust trimpot R4 to midscale and set SW1 for rms.
2. Apply a 2.000 V rms, 1 kHz sine wave input signal.
Extending the AD736 and AD737 Full-Scale
Input Ranges
3. Adjust R2 until the circuit’s output voltage is
2.000 V dc.
The high impedance input (Pin 2) of the AD736 and
AD737 allows simple resistive attenuators (Figure 10)
to be used to extend their input range. Without input
attenuation, both the AD736 and AD737 can accurately measure input signals as large as 200 mV rms
with crest factors of 1 to 3.
4. Reduce the input to 100 mV rms and adjust offset
trimpot R4 for a reading of 100 mV dc.
5. Repeat Step 3.
As the dc offset circuitry is ratiometric, it will remain
calibrated with modest variations in supply voltage. The
measured PSRR of this circuit (over a 4.5 V to 5.5 V
supply range) is approximately 61 dB.
C1
C3
0.01␮F
1kV
VIN
10␮F
200mV
+
+5V
9M⍀
1N4148
2V
47k⍀
1W
900k⍀
20V
1 CC
2
VIN
90k⍀
200V
10k⍀
1N4148
3 CF
COMMON 8
U1
AD736
C4
0.1␮F
OUTPUT 6
VRMS
CAV 5
4 –VS
–5V
+VS 7
+5V
C2
0.1␮F
+
CAV 33␮F
C1 AND THE RESISTIVE
DIVIDER FORM A 1.6Hz (–3dB)
HIGH-PASS FILTER
+
CF 10␮F
Figure 10. By Using an External Input Attenuator, the Measurement Range of the AD736
and AD737 Can be Extended
8
The external attenuator simply reduces the full-scale
input to the 200 mV rms input range of the AD736 or
AD737. For a maximum 7 V rms input (10 V peak), for
example, the attenuator should be a 35:1 (7 V/200 mV)
voltage divider. The reading of the converter should be
scaled by the factor of attenuation used. An external
attenuator can also be used with the converter’s low
impedance input (Pin 1), as shown in Figure 10.
Figures 11 and 12 show the recommended connections
for external offset and scale factor.
DC-COUPLED
+
CC
10␮F
AC-COUPLED
CC
(OPTIONAL)
COM
8k⍀
AD736
1
FULL
WAVE
RECTIFIER
VIN
VIN
2
+VS
8k⍀
7
INPUT
AMPLIFIER
CF
OUTPUT
3
6
BIAS
SECTION
–VS
+VS
8
OUTPUT
AMPLIFIER
RMS
CORE
4
CAV
5
39M⍀
1M⍀
–VS
+
OUTPUT
VOS
ADJUST
33␮F
CAV
+
CF
10␮F
(OPTIONAL)
Figure 11. AD736 External VOS Adjustment
+VS
OFFSET ADJUST
500k⍀
–VS
1M⍀
1k⍀
CC
AD737
8k⍀
FULL
WAVE
RECTIFIER
VIN
VIN
COM
8
1
2
+VS
8k⍀
7
499⍀
1k⍀
SCALE
FACTOR
ADJUST
INPUT
AMPLIFIER
3
6
VOUT
Figure 12. AD737 DC-Coupled VOS and Scale Factor Adjustments
9
10
11
12
PRINTED IN U.S.A.
G03133–0–10/02(0)
Similar pages