Data Converter Support Circuits.

DATA CONVERTER SUPPORT CIRCUITS
ANALOG-DIGITAL CONVERSION
1. Data Converter History
2. Fundamentals of Sampled Data Systems
3. Data Converter Architectures
4. Data Converter Process Technology
5. Testing Data Converters
6. Interfacing to Data Converters
7. Data Converter Support Circuits
7.1 Voltage References
7.2 Low Dropout Linear Regulators
7.3 Analog Switches and Multiplexers
7.4 Sample-and-Hold Circuits
8. Data Converter Applications
9. Hardware Design Techniques
I. Index
ANALOG-DIGITAL CONVERSION
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
CHAPTER 7
DATA CONVERTER SUPPORT CIRCUITS
SECTION 7.1: VOLTAGE REFERENCES
Walt Jung, Walt Kester, James Bryant
Reference circuits and linear regulators actually have much in common. In fact, the latter
could be functionally described as a reference circuit, but with greater current (or power)
output. Accordingly, almost all of the specifications of the two circuit types have great
commonality (even though the performance of references is usually tighter with regard to
drift, accuracy, etc.). This section discusses voltage references, and the next section
covers linear regulators, with emphasis on their low dropout operation for highest power
efficiency.
Precision Voltage References
Voltage references have a major impact on the performance and accuracy of analog
systems. A ±5-mV tolerance on a 5-V reference corresponds to ±0.1% absolute
accuracy—only 10 bits. For a 12-bit system, choosing a reference that has a ±1-mV
tolerance may be far more cost effective than performing manual calibration, while both
high initial accuracy and calibration will be necessary in a system making absolute 16-bit
measurements. Note that many systems make relative measurements rather than absolute
ones, and in such cases the absolute accuracy of the reference is not important, although
noise and short-term stability may be. Figure 7.1 summarizes some key points of the
reference selection process.
Temperature drift or drift due to aging may be an even greater problem than absolute
accuracy. The initial error can always be trimmed, but compensating for drift is difficult.
Where possible, references should be chosen for temperature coefficient and aging
characteristics which preserve adequate accuracy over the operating temperature range
and expected lifetime of the system.
Noise in voltage references is often overlooked, but it can be very important in system
design. It is generally specified on data sheets, but system designers frequently ignore the
specification and assume that voltage references do not contribute to system noise.
There are two dynamic issues that must be considered with voltage references: their
behavior at start-up, and their behavior with transient loads. With regard to the first,
always bear in mind that voltage references do not power up instantly (this is true of
references inside ADCs and DACs as well as discrete designs). Thus it is rarely possible
to turn on an ADC and reference, whether internal or external, make a reading, and turn
off again within a few microseconds, however attractive such a procedure might be in
terms of energy saving.
7.1
ANALOG-DIGITAL CONVERSION
Regarding the second point, a given reference IC may or may not be well suited for
pulse-loading conditions, dependent upon the specific architecture. Many references use
low power, and therefore low bandwidth, output buffer amplifiers. This makes for poor
behavior under fast transient loads, which may degrade the performance of fast ADCs
(especially successive approximation and flash ADCs). Suitable decoupling can ease the
problem (but some references oscillate with capacitive loads), or an additional external
broadband buffer amplifier may be used to drive the node where the transients occur.
References, like almost all other ICs today, are fast migrating to such smaller packages
such as SO-8 and MSOP, and the even more tiny SOT-23 and SC-70, enabling much
higher circuit densities within a given area of real estate. In addition to the system size
reductions these steps bring, there are also tangible reductions in standby power and cost
with the smaller and less expense ICs.
Tight Tolerance Improves Accuracy, Reduces System Costs
Temperature Drift Affects Accuracy
Long-Term Stability, Low Hysteresis Assures Repeatability
Noise Limits System Resolution
Dynamic Loading Can Cause Errors
Power Consumption is Critical to Battery Systems
Tiny Low Cost Packages Increase Circuit Density
Figure 7.1: Choosing Voltage References for High Performance Systems
Types of Voltage References
In terms of the functionality of their circuit connection, standard reference ICs are often
only available in series, or three-terminal form (VIN, Common, VOUT), and also in
positive polarity only. The series types have the potential advantages of lower and more
stable quiescent current, standard pre-trimmed output voltages, and relatively high output
current without accuracy loss. Shunt, or two-terminal (i.e., diode-like) references are
more flexible regarding operating polarity, but they are also more restrictive as to
loading. They can in fact eat up excessive power with widely varying resistor-fed voltage
inputs. Also, they sometimes come in non-standard voltages. All of these various factors
tend to govern when one functional type is preferred over the other.
Some simple diode-based references are shown in Figure 7.2. In the first of these, a
current driven forward biased diode (or diode-connected transistor) produces a voltage,
Vf = VREF. While the junction drop is somewhat decoupled from the raw supply, it has
numerous deficiencies as a reference. Among them are a strong TC of about –0.3%/°C,
some sensitivity to loading, and a rather inflexible output voltage: it is only available in
600-mV jumps.
7.2
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
By contrast, these most simple references (as well as all other shunt-type regulators) have
a basic advantage, which is the fact that the polarity is readily reversible by flipping
connections and reversing the drive current. However, a basic limitation of all shunt
regulators is that load current must always be less (usually appreciably less) than the
driving current, ID.
+VS
RS
D1
+VS
ID
RZ
VREF
IZ
D1
VREF
D2
FORWARD-BIASED
DIODE
ZENER (AVALANCHE)
DIODE
Figure 7.2: Simple Diode Reference Circuits
In the second circuit of Figure 7.2, a Zener or avalanche diode is used, and an appreciably
higher output voltage realized. While true Zener breakdown occurs below 5 V, avalanche
breakdown occurs at higher voltages and has a positive temperature coefficient. Note that
diode reverse breakdown is referred to almost universally today as Zener, even though it
is usually avalanche breakdown. With a D1 breakdown voltage in the 5- to 8-V range, the
net positive TC is such that it equals the negative TC of forward-biased diode D2,
yielding a net TC of 100 ppm/°C or less with proper bias current. Combinations of such
carefully chosen diodes formed the basis of the early single package "temperaturecompensated Zener" references, such as the 1N821-1N829 series.
The temperature-compensated Zener reference is limited in terms of initial accuracy,
since the best TC combinations fall at odd voltages, such as the 1N829's 6.2 V. And, the
scheme is also limited for loading, since for best TC the diode current must be carefully
controlled. Unlike a fundamentally lower voltage (<2 V) reference, Zener diode based
references must of necessity be driven from voltage sources appreciably higher than 6-V
levels, so this precludes operation of Zener references from 5-V system supplies.
References based on low TC Zener (avalanche) diodes also tend to be noisy, due to the
basic noise of the breakdown mechanism. This has been improved greatly with
monolithic Zener types, as is described further below.
At this point, we know that a reference circuit can be functionally arranged into either a
series or shunt operated form, and the technology within may use either bandgap based or
7.3
ANALOG-DIGITAL CONVERSION
Zener diode based circuitry. In practice there are all permutations of these available, as
well as a third major technology category. The three major reference technologies are
now described in more detail.
Bandgap References
The development of low voltage (<5 V) references based on the bandgap voltage of
silicon led to the introductions of various ICs which could be operated on low voltage
supplies with good TC performance. The first of these was the LM109 (Reference 1), and
a basic bandgap reference cell is shown in Figure 7.3.
+VS
IZ
R2 ∆V
BE
R3
R2
6kΩ
R1
600Ω
Q2
VR = VBE +
R2 ∆V
BE
R3
Q3
Q1
R3
600Ω
∆VBE
VBE
Figure 7.3: Basic Bandgap Reference
This circuit is also called a "∆VBE" reference because the differing current densities
between matched transistors Q1-Q2 produces a ∆VBE across R3. It works by summing
the VBE of Q3 with the amplified ∆VBE of Q1-Q2, developed across R2. The ∆VBE and
VBE components have opposite polarity TCs; ∆VBE is proportional-to-absolutetemperature (PTAT), while VBE is complementary-to-absolute-temperature (CTAT). The
summed output is VR, and when it is equal to 1.205 V (silicon bandgap voltage), the TC
is a minimum.
The bandgap reference technique is attractive in IC designs because of several reasons;
among these are the relative simplicity, and the avoidance of Zeners and their noise.
However, very important in these days of ever decreasing system power supplies is the
fundamental fact that bandgap devices operate at low voltages, i.e., <5 V. Not only are
they used for stand-alone IC references, but they are also used within the designs of many
other ICs, such as ADCs and DACs.
7.4
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
Buffered forms of 1.2-V two terminal shunt bandgap references, such as the AD589 IC,
remain stable under varying load currents. The AD589 (introduced in 1980), a 1.235-V
reference, handles 50 µA to 5 mA with an output impedance of 0.6 Ω, and TCs ranging
between 10 and 100 ppm/°C. The more recent and functionally similar AD1580, a
1.225-V shunt reference, is in the tiny SOT-23 package and handles the same nominal
currents as the AD589, with TCs of 50 and 100 ppm/°C. The ADR510 shunt reference
supplies 1.000 V, and the ADR512 supplies 1.200 V.
However, the basic designs of Figure 7.3 suffer from load and current drive sensitivity,
plus the fact that the output needs accurate scaling to more useful levels, i.e., 2.5 V, 5 V,
etc. The load drive issue is best addressed with the use of a buffer amplifier, which also
provides convenient voltage scaling to standard levels.
An improved three-terminal bandgap reference, the AD580 (introduced in 1974) is shown
in Figure 7.4. Popularly called the "Brokaw Cell" (see References 2 and 3), this circuit
provides on-chip output buffering, which allows good drive capability and standard
output voltage scaling. The AD580 was the first precision bandgap based IC reference,
and variants of the topology have influenced further generations of both industry standard
references such as the REF01, REF02, and REF03 series, as well as more recent ADI
bandgap parts such as the REF19x series, the AD680, AD780, the AD1582-85 series, the
ADR38x series, the ADR39x series, and recent SC-70 and SOT-23 offerings of improved
versions of the REF01, REF02, and REF03 (designated ADR01, ADR02, and ADR03).
The AD580 has two 8:1 emitter-scaled transistors Q1-Q2 operating at identical collector
currents (and thus 1/8 current densities), by virtue of equal load resistors and a closed
loop around the buffer op amp. Due to the resultant smaller VBE of the 8× area Q2, R2 in
series with Q2 drops the ∆VBE voltage, while R1 (due to the current relationships) drops a
PTAT voltage V1:
V1 = 2 ×
R1
× ∆VBE .
R2
Eq. 7.1
The bandgap cell reference voltage VZ appears at the base of Q1, and is the sum of VBE
(Q1) and V1, or 1.205 V, the bandgap voltage:
VZ = VBE(Q1) + V1
R1
× ∆VBE
R2
J1
R1 kT
= VBE(Q1) + 2 ×
×
× ln
J2
R2 q
R1 kT
= VBE(Q1) + 2 ×
×
× ln8
R2 q
= 1.205V .
= VBE(Q1) + 2 ×
Eq. 7.2
Eq. 7.3
Eq. 7.4
Eq. 7.5
Eq. 7.6
Note that J1 = current density in Q1, J2 = current density in Q2, and J1/J2 = 8.
7.5
ANALOG-DIGITAL CONVERSION
+VIN
R8
R7
+
I2 ≅ I1
VOUT = 2.5V
R4
Q2
8A
∆VBE
Q1
VZ = 1.205V
A
VBE
(Q1)
R2
R5
A = TRANSISTOR
AREA
R1
V1 = 2
R1 ∆V
BE
R2
COM
Figure 7.4: AD580 Precision Bandgap Reference Uses Brokaw Cell (1974)
However, because of the presence of the R4/R5 (laser trimmed) thin film divider and the
op amp, the actual voltage appearing at VOUT can be scaled higher, in the AD580 case
2.5 V. Following this general principle, VOUT can be raised to other practical levels, such
as for example in the AD584, with taps for precise 2.5-, 5-, 7.5-, and 10-V operation. The
AD580 provides up to 10-mA output current while operating from supplies between
4.5 and 30 V. It is available in tolerances as low as 0.4%, with TCs as low as 10 ppm/°C.
Many of the recent developments in bandgap references have focused on smaller package
size and cost reduction, to address system needs for smaller, more power efficient and
less costly reference ICs. Among these are several recent bandgap-based IC references.
The AD1580 (introduced in 1996) is a shunt mode IC reference which is functionally
quite similar to the classic shunt IC reference, the AD589 (introduced in 1980) mentioned
above. A key difference is the fact that the AD1580 uses a newer, small geometry
process, enabling its availability within the tiny SOT-23 package. The very small size of
this package allows use in a wide variety of space limited applications, and the low
operating current lends itself to portable battery powered uses. The AD1580 circuit is
shown in simplified form in Figure 7.5.
In this circuit, transistors Q1 and Q2 form the bandgap core, and are operated at a current
ratio of 5 times, determined by the ratio of R7 to R2. An op amp is formed by the
differential pair Q3-Q4, current mirror Q5, and driver/output stage Q8-Q9. In closed loop
equilibrium, this amplifier maintains the bottom ends of R2-R7 at the same potential.
7.6
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
V+
+
R6
R5
R1
Q5
V1
Q8
R7
R2
–
Q9
Q3
+
Q4
R3
∆VBE
Q2
–
+
Q1
R4
V–
VBE
–
Figure 7.5: AD1580 1.2-V Shunt Type Bandgap Reference
has Tiny Size in SOT-23 Footprint
As a result of the closed loop control described, a basic ∆VBE voltage is dropped across
R3, and a scaled PTAT voltage also appears as V1, which is effectively in series with
VBE. The nominal bandgap reference voltage of 1.225 V is then the sum of Q1's VBE and
V1. The AD1580 is designed to operate at currents as low as 50 µA, also handling
maximum currents as high as 10 mA. It is available in grades with voltage tolerances of
±1 or ±10 mV, and with corresponding TCs of 50 or 100 ppm/°C. Newer members of the
Analog Devices' family of shunt regulators are the ADR510 (1.000 V), and the ADR512
(1.200 V).
The ADR520 (2.048 V), ADR525 (2.500 V), ADR530 (3.000 V), ADR540 (4.096 V),
ADR545 (4.5 V), and ADR550 (5.0 V) are the latest in the shunt regulator family, with
initial accuracies of 0.2%, and available in either SC-70 or SOT-23 packages.
The AD1582-AD1585 series comprises a family of series mode IC references, which
produce voltage outputs of 2.5, 3.0, 4.096 and 5.0 V. Like the AD1580, the series uses a
small geometry process to allow packaging within an SOT-23. The AD1582 series
specifications are summarized in Figure 7.6.
The circuit diagram for the series, shown in Figure 7.7, may be recognized as a variant of
the basic Brokaw bandgap cell, as described under Figure 7.4. In this case Q1-Q2 form
the core, and the overall loop operates to produce the stable reference voltage VBG at the
base of Q1. A notable difference here is that the op amp's output stage is designed with
push-pull common-emitter stages. This has the effect of requiring an output capacitor for
stability, but it also provides the IC with relatively low dropout operation.
7.7
ANALOG-DIGITAL CONVERSION
The low dropout feature means essentially that VIN can be lowered to as close as several
hundred mV above the VOUT level without disturbing operation. The push-pull operation
also means that this device series can actually both sink and source currents at the output,
as opposed to the classic reference operation of sourcing current (only). For the various
output voltage ratings, the divider R5-R6 is adjusted for the respective levels.
VOUT : 2.500, 3.000, 4.096, & 5.000V
2.7V to 12V Supply Range (200mV Headroom)
Supply Current : 65µA max
Initial Accuracy: ±0.1% max
Temperature Coefficient: 50 ppm/°C max
Noise: 70µV p-p (0.1Hz - 10Hz)
Noise: 50µV rms (10Hz - 10kHz)
Long-Term Drift: 100ppm/1khrs
High Output Current: ±5mA min
Temperature Range –40°C to +85°C
Low Cost SOT-23 Package
Figure 7.6: AD1582-AD1585 2.5-V to 5-V Series-Type
Bandgap Reference Specifications
The AD1582-series is designed to operate with quiescent currents of only 65 µA
(maximum), which allows good power efficiency when used in low power systems with
varying voltage inputs. The rated output current for the series is 5 mA, and they are
available in grades with voltage tolerances of ±0.1 or ±1% of VOUT, with corresponding
TCs of 50 or 100 ppm/°C.
Because of stability requirements, devices of the AD1582 series must be used with both
an output and input bypass capacitor. Recommended optimum values for these are shown
in the hookup diagram of Figure 7.8. For the electrical values noted, it is likely that
tantalum chip capacitors will be the smallest in size.
ADR38x and ADR39x-series are low dropout (300 mV) bandgap references in SOT-23
packages. Noise is typically 5-µV p-p in the 0.1 Hz to 10 Hz bandwidth. Quiescent
current is typically 100 µA, and the ADR39x-series have a shutdown pin (shutdown
current < 3 µA) as well as a "sense" pin for Kelvin sensing. A connection diagram for the
ADR39x series is shown in Figure 7.9, and key specifications for the family are shown in
Figure 7.10. The ADR38x and ADR39x-series do not require an output capacitor for
stability, regardless of the load conditions. However, at least a 1-µF capacitor is
recommended to filter out noise. Larger capacitors may be desirable to act as a source of
stored energy for transient loads.
7.8
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
VIN
R3
R4
VOUT
+
R5
Q1
Q2
VBG
∆VBE
R6
R2
R1
V1 = 2
R1 ∆V
BE
R2
GND
Figure 7.7: AD1582-AD1585 2.5-V to 5-V Series-Type
Bandgap References in SOT-23 Footprint
1
VIN
VOUT
+
+
1µF
COUT
3
+
2
4.7µF
AD1582-1585: COUT REQUIRED FOR STABILITY
ADR380, ADR381: COUT RECOMMENDED TO ABSORB TRANSIENTS
Figure 7.8: AD1582-AD1585 Series Connection Diagram
7.9
ANALOG-DIGITAL CONVERSION
1
SHDN
VIN
+
4.7µF
2
5
ADR390, 2.048V
ADR391, 2.500V
ADR392, 4.096V
ADR395, 5.000V
VOUT FORCE
VOUT SENSE
3
4
+
COUT
COUT RECOMMENDED TO ABSORB TRANSIENTS
Figure 7.9: ADR390, ADR391, ADR392, ADR395
Connection Diagram
VOUT : 2.048, 2.500, 4.096, & 5.000V
2.3V to 15V Supply Range (300mV Headroom)
Supply Current : 120µA max
Initial Accuracy: ±6mV max
Temperature Coefficient: 25 ppm/°C max
Noise: 5µV p-p (0.1Hz - 10Hz)
Long-Term Drift: 50ppm/1khrs
High Output Current: +5mA min
Temperature Range –40°C to +85°C
Shutdown Feature: <3µA max
Kelvin Sensing (Force and Sense Pins)
Low Cost SOT-23 (5 pin) Package
Figure 7.10: ADR390-ADR395 2.048-V to 5-V Series-Type
Bandgap Reference Specifications
7.10
1µF
RL
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
Buried Zener References
In terms of the design approaches used within the reference core, the two most popular
basic types of IC references consist of the bandgap and buried Zener units. Bandgaps
have been discussed, but Zener based references warrant some further discussion.
In an IC chip, surface operated diode junction breakdown is prone to crystal
imperfections and other contamination, thus Zener diodes formed at the surface are more
noisy and less stable than are buried (or sub-surface) ones. ADI Zener based IC
references employ the much preferred buried Zener. This improves substantially upon the
noise and drift of surface-mode operated Zeners (see Reference 4). Buried Zener
references offer very low temperature drift, down to the 1-2 ppm/°C (AD588 and
AD586), and the lowest noise as a percent of full-scale, i.e., 100 nV/√Hz or less. On the
downside, the operating current of Zener type references is usually relatively high,
typically on the order of several mA.
An important general point arises when comparing noise performance of different
references. The best way to do this is to compare the ratio of the noise (within a given
bandwidth) to the dc output voltage. For example, a 10-V reference with a 100-nV/√Hz
noise density is 6-dB more quiet in relative terms than is a 5-V reference with the same
noise level.
XFET® References
A third and relatively new category of IC reference core design is based on the properties
of junction field effect (JFET) transistors. Somewhat analogous to the bandgap reference
for bipolar transistors, the JFET based reference operates a pair of junction field effect
transistors with different pinchoff voltages, and amplifies the differential output to
produce a stable reference voltage. One of the two JFETs uses an extra ion implantation,
giving rise to the name XFET® (eXtra implantation junction Field Effect Transistor) for
the reference core design.
The basic topology for the XFET reference circuit is shown in Figure 7.11. J1 and J2 are
the two JFET transistors, which form the core of the reference. J1 and J2 are driven at the
same current level from matched current sources, I1 and I2. To the right, J1 is the JFET
with the extra implantation, which causes the difference in the J1-J2 pinchoff voltages to
differ by 500 mV. With the pinchoff voltage of two such FETs purposely skewed, a
differential voltage will appear between the gates for identical current drive conditions
and equal source voltages. This voltage, ∆VP, is:
∆ VP = VP1 - VP2 ,
Eq. 7.7
where VP1 and VP2 are the pinchoff voltages of FETs J1 and J2, respectively.
7.11
ANALOG-DIGITAL CONVERSION
VIN
I2
I1
R2 + R 3 

VOUT = ∆VP  1 +
 + IPTAT • R 3

R1 
+
VOUT
V=0
J2
R1
J1
∆VP
IPTAT
R2
R3
Figure 7.11: ADR290-ADR293 2.048-V to 5-V XFET® References Feature High
Stability and Low Power
Note that, within this circuit, the voltage ∆VP exists between the gates of the two FETs.
We also know that, with the overall feedback loop closed, the op amp axiom of zero input
differential voltage will hold the sources of the two JFETs at the same potential. These
source voltages are applied as inputs to the op amp, the output of which drives feedback
divider R1-R3. As this loop is configured, it stabilizes at an output voltage from the R1R2 tap which does in fact produce the required ∆VP between the J1-J2 gates. In essence,
the op amp amplifies ∆VP to produce VOUT,where
R2 + R3 

VOUT = ∆VP 1 +
 + (I PTAT )(R3) .
R1 

Eq. 7.8
As can be noted, this expression includes the basic output scaling (leftmost portion of the
right terms), plus a rightmost temperature dependent term including IPTAT. The IPTAT
portion of the expression compensates for a basic negative temperature coefficient of the
XFET core, such that the overall net temperature drift of the reference is typically in a
range of 3 to 8 ppm/°C.
During manufacture, the R1-R3 scaling resistance values are adjusted to produce the
different voltage output options of 2.048, 2.5, 4.096 and 5.0 V for the ADR290, ADR291,
ADR292 and ADR293 family (ADR29x). This ADR29x family of series mode
references is available in 8 pin packages with a standard footprint. They operate from
supplies of VOUT plus 500 mV to 15 V, with a maximum quiescent current of 12 µA, and
output currents of up to 5 mA. A summary of specifications for the family appears in
Figure 7.12.
7.12
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
VOUT : 2.048, 2.500, 4.096, & 5.000V
2.7V to 15V Supply Range (0.5V Headroom)
Supply Current : 12µA max
Initial Accuracy: ±0.1%
Temperature Coefficient: 8 ppm/°C max
Low-Noise: 6µV p-p (0.1 - 10Hz)
Wideband Noise: 420nV/√Hz @ 1kHz
Long-Term Drift: 50ppm/1000 hours
High Output Current: 5mA min
Temperature Range –40°C to +125°C
Standard REF02 Pinout
8-Lead Narrow Body SOIC, 8-Lead TSSOP
Figure 7.12: ADR290-ADR293 XFET® Series Specifications
The ADR43x-series are the second generation of low noise, low drift XFET references.
Standard voltage outputs are 2.048, 2.500, 3.000, 4.096, and 5.000 V. These devices
operate from supplies of VOUT + 1 V to 18 V with quiescent currents of 0.5-mA
maximum and output currents of ±10 mA. Temperature drift is 3-ppm/°C maximum. The
0.1-Hz to 10-Hz noise is an incredibly low 1.5-µV p-p. This ADR43x family of series
mode references is available in 8 pin packages with a standard footprint. Key
specifications for the family are summarized in Figure 7.13.
VOUT : 2.048, 2.500, 3.000, 4.096, & 5.000V
3V to 18V Supply Range (1V Headroom)
Supply Current : 500µA
Initial Accuracy: ±0.05%
Temperature Coefficient: 3 ppm/°C max
Low-Noise: 1.75µV p-p (0.1 - 10Hz)
Wideband Noise: 60nV/√Hz @ 1kHz
Long-Term Drift: 50ppm/1000 hours
High Output Current: ±10mA min
Temperature Range –40°C to +125°C
8-Lead MSOP, 8-Lead TSSOP
Figure 7.13: ADR430-ADR439 XFET® Series Specifications
7.13
ANALOG-DIGITAL CONVERSION
The XFET architecture offers performance improvements over bandgap and buried
Zener references, particularly for systems where operating current is critical, yet drift and
noise performance must still be excellent. XFET noise levels are lower than bandgap
based bipolar references operating at an equivalent current, the temperature drift is low
and linear at 3-8 ppm/°C (allowing easier compensation when required), and the series
has lower hysteresis than bandgaps. Thermal hysteresis is a low 50 ppm over a –40°C to
+125°C range, less that half that of a typical bandgap device. Finally, the long-term
stability is excellent, typically only 50 ppm/1000 hours.
Figure 7.14 summarizes the pro and con characteristics of the three reference
architectures; bandgap, buried Zener, and XFET.
BANDGAP
BURIED ZENER
XFET®
< 5V Supplies
> 5V Supplies
< 5V Supplies
High Noise
@ High Power
Low Noise
@ High Power
Low Noise
@ Low Power
Fair Drift and
Long Term Stability
Good Drift and
Long Term Stability
Excellent Drift and
Long Term Stability
Fair Hysteresis
Fair Hysteresis
Low Hysteresis
Figure 7.14: Characteristics of Reference Architectures
Modern IC references come in a variety of styles, but series operating, fixed output
positive types do tend to dominate. These devices can use bandgap-based bipolars,
JFETs, or buried Zeners at the device core, all of which has an impact on the part's
ultimate performance and application suitability. They may or may not also be low
power, low noise, and/or low dropout, and be available within a certain package. Of
course, in a given application, any single one of these differentiating factors can drive a
choice, thus it behooves the designer to be aware of all the different devices available.
Figure 7.15 shows the standard footprint for such a series type IC positive reference in an
8 pin package (Note that pin numbers shown refer to the standard pin for that function).
There are several details which are important. Many references allow optional trimming
by connecting an external trim circuit to drive the references' trim input pin (5). Some
bandgap references also have a high impedance PTAT output (VTEMP) for temperature
sensing (pin 3). The intent here is that no appreciable current be drawn from this pin, but
it can be useful for such non-loading types of connections as comparator inputs, to sense
temperature thresholds, etc.
7.14
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
+VS
2
OUTPUT (+) LEAD
SHORT, HEAVY TRACE
VIN
C1
VOUT
+
C2
6
0.1µF
3
TEMP
TRIM
10µF
RTRIM
5
RPAD
VTEMP
VREF
GND
+
VOUT
RL
COUT
4
Iq
POWER COMMON
Figure 7.15: Standard Positive Output Three Terminal
Reference Hookup (8-pin DIP Pinout)
All references should use decoupling capacitors on the input pin (2), but the amount of
decoupling (if any) placed on the output (pin 6) depends upon the stability of the
reference's output op amp with capacitive load. Simply put, there is no hard and fast rule
for capacitive loads here. For example, some three terminal types require the output
capacitor for stability (i.e., REF19x and AD1582-85 series), while with others it is
optional for performance improvement (AD780, REF43, ADR29x, ADR43x, AD38x,
AD39x, ADR01, ADR02, ADR03). Even if the output capacitor is optional, it may still
be required to supply the energy for transient load currents, as presented by some ADC
reference input circuits. The safest rule then is that you should use the data sheet to verify
what are the specific capacitive loading ground rules for the reference you intend to use,
for the load conditions your circuit presents.
Voltage Reference Specifications
Tolerance
It is usually better to select a reference with the required value and accuracy and to avoid
external trimming and scaling if possible. This allows the best TCs to be realized, as tight
tolerances and low TCs usually go hand-in-hand. Tolerances as low as approximately
0.04% can be achieved with the AD586, AD780, REF195, and ADR43x-series, while the
AD588 is 0.01%. If and when trimming must be used, be sure to use the recommended
trim network with no more range than is absolutely necessary. When/if additional
external scaling is required, a precision op amp should be used, along with ratio-accurate,
low TC tracking thin film resistors.
7.15
ANALOG-DIGITAL CONVERSION
Drift
The XFET and buried Zener reference families have the best long term drift and TC
performance. The XFET ADR43x-series have TCs as low as 3 ppm/°C. TCs as low as
1-2 ppm/°C are available with the AD586 and AD588 buried Zener references, and the
AD780 bandgap reference is almost as good at 3 ppm/°C.
The XFET series achieve long terms drifts of 50 ppm/1000 hours, while the buried Zener
types come in at 25 ppm/1000 hours. Note that where a figure is given for long term drift,
it is usually drift expressed in ppm/1000 hours. There are 8766 hours in a year, and many
engineers multiply the 1000-hour figure by 8.77 to find the annual drift—this is not
correct, and can in fact be quite pessimistic. Long term drift in precision analog circuits is
a "random walk" phenomenon and increases with the square root of the elapsed time
(this supposes that drift is due to random micro-effects in the chip and not some overriding cause such as contamination). The 1 year figure will therefore be about √8.766 ≈ 3
times the 1000-hour figure, and the ten year value will be roughly 9 times the 1000-hour
value. In practice, things are a little better even than this, as devices tend to stabilize with
age.
The accuracy of an ADC or DAC can be no better than that of its reference. Reference
temperature drift affects fullscale accuracy as shown in Figure 7.16. This table shows
system resolution and the TC required to maintain ½ LSB error over an operating
temperature range of 100°C. For example, a TC of about 1 ppm/°C is required to
maintain ½ LSB error at 12 bits. For smaller operating temperature ranges, the drift
requirement will be less. The last three columns of the table show the voltage value of
½ LSB for popular full scale ranges.
½ LSB WEIGHT (mV)
10, 5, AND 2.5V FULLSCALE RANGES
BITS
8
REQUIRED
DRIFT (ppm/ºC)
19.53
10V
19.53
5V
9.77
2.5V
4.88
9
9.77
9.77
4.88
2.44
10
4.88
4.88
2.44
1.22
11
2.44
2.44
1.22
0.61
12
1.22
1.22
0.61
0.31
13
0.61
0.61
0.31
0.15
14
0.31
0.31
0.15
0.08
15
0.15
0.15
0.08
0.04
16
0.08
0.08
0.04
0.02
Figure 7.16: Reference Temperature Drift Requirements for Various System
Accuracies (1/2 LSB Criteria, 100°C Span)
7.16
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
Supply Range
IC reference supply voltages range from about 3 V (or less) above rated output, to as high
as 30 V (or more) above rated output. Exceptions are devices designed for low dropout,
such as the REF19x, AD1582-AD1585, ADR38x, ADR39x series. At low currents, the
REF195 can deliver 5 V with an input as low as 5.1 V (100-mV dropout). Note that due
to process limits, some references may have more restrictive maximum voltage input
ranges, such as the AD1582-AD1585 series (12 V), the ADR29x series (15 V), and the
ADR43x series (18 V).
Load Sensitivity
Load sensitivity (or output impedance) is usually specified in µV/mA of load current, or
mΩ, or ppm/mA. While figures of 70 ppm/mA or less are quite good (AD780, REF43,
REF195, ADR29x, ADR43x), it should be noted that external wiring drops can produce
comparable or worse errors at high currents, without care in layout. Load current
dependent errors are minimized with short, heavy conductors on the (+) output and on the
ground return. For the highest precision, buffer amplifiers and Kelvin sensing circuits
(AD588, AD688, ADR39x) are used to ensure accurate voltages at the load.
The output of a buffered reference is the output of an op amp, and therefore the source
impedance is a function of frequency. Typical reference output impedance rises at
6 dB/octave from the dc value, and is nominally about 10 Ω at a few hundred kHz. This
impedance can be lowered with an external capacitor, provided the op amp within the
reference remains stable for such loading.
Line Sensitivity
Line sensitivity (or regulation) is usually specified in µV/V, (or ppm/V) of input change,
and is typically 25 ppm/V (–92 dB) in the REF43, REF195, AD680, AD780, ADR29x,
ADR39x, and ADR43x. For dc and very low frequencies, such errors are easily masked
by noise.
As with op amps, the line sensitivity (or power supply rejection) of references degrades
with increasing frequency, typically 30 to 50 dB at a few hundred kHz. For this reason,
the reference input should be highly decoupled (LF and HF). Line rejection can also be
increased with a low dropout pre-regulator, such as one of the ADP3300-series parts.
Figure 7.17 summarizes the major reference specifications along with typical values
available.
7.17
ANALOG-DIGITAL CONVERSION
Tolerance:
AD588
ADR43x, AD780, REF195
0.01%
0.04%
Drift (TC):
AD586, AD588
AD780, ADR42x, ADR43x,
1-2ppm/°C
3 ppm/°C
ADR01, ADR02, ADR03
Drift (long term):
ADR29x,ADR42x, ADR43x
AD588
50 ppm/1000 hours
25 ppm/1000 hours
Supply Range:
REF19x, ADR38x, ADR39x,
VOUT plus 0.3V-15V
AD158x, AD780
Load Sensitivity
70ppm/mA (350mΩ @ 5V)
Line Sensitivity
25ppm/V (–92 dB @ 5V)
Figure 7.17: Voltage Reference DC Specifications
(Typical Values Available)
Noise
Reference noise is not always specified, and when it is, there is not total uniformity on
how it is measured. For example, some devices are characterized for peak-to-peak noise
in a 0.1-Hz to 10-Hz bandwidth, while others are specified in terms of wideband rms or
peak-to-peak noise over a specified bandwidth. The most useful way to specify noise (as
with op amps) is a plot of noise voltage spectral density (nV/√Hz) versus frequency.
Low noise references are important in high resolution systems to prevent loss of
accuracy. Since white noise is statistical, a given noise density must be related to an
equivalent peak-to-peak noise in the relevant bandwidth. Strictly speaking, the peak-topeak noise in a gaussian system is infinite (but its probability is infinitesimal).
Conventionally, the figure of 6.6 × rms is used to define a practical peak value—
statistically, this occurs less than 0.1% of the time. This peak-to-peak value should be
less than ½ LSB in order to maintain required accuracy. If peak-to-peak noise is assumed
to be 6 times the rms value, then for an N-bit system, reference voltage fullscale VREF,
reference noise bandwidth (BW), the required noise voltage spectral density En (V/√Hz)
is given by:
En ≤
VREF
12 ⋅ 2 N ⋅ BW
.
Eq. 7.9
For a 10-V, 12-bit, 100-kHz system, the noise requirement is a modest 643 nV/√Hz.
Figure 7.18 shows that increasing resolution and/or lower fullscale references make noise
7.18
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
requirements more stringent. The 100-kHz bandwidth assumption is somewhat arbitrary,
but the user may reduce it with external filtering, thereby reducing the noise. Most good
IC references have noise spectral densities around 100 nV/√Hz, so additional filtering is
obviously required in most high resolution systems, especially those with low values of
VREF.
NOISE DENSITY (nV/√Hz) FOR
10, 5, AND 2.5V FULLSCALE RANGES
BITS
10V
5V
2.5V
12
643
322
161
13
322
161
80
14
161
80
40
15
80
40
20
16
40
20
10
Figure 7.18: Reference Noise Requirements for Various System Accuracies
(1/2-LSB / 100-kHz Criteria)
Some references, for example the AD587 buried Zener type have a pin designated as the
noise reduction pin (see data sheet). This pin is connected to a high impedance node
preceding the on-chip buffer amplifier. Thus an externally connected capacitor CN will
form a low pass filter with an internal resistor, to limit the effective noise bandwidth seen
at the output. A 1-µF capacitor gives a 3-dB bandwidth of 40 Hz. Note that this method
of noise reduction is by no means universal, and other devices may implement noise
reduction differently, if at all.
There are also general purpose methods of noise reduction, which can be used to reduce
the noise of any reference IC, at any standard voltage level. The reference circuit of
Figure 7.19 (References 5 and 6) is one such example. This circuit uses external filtering
and a precision low-noise op amp to provide both very low noise and high dc accuracy.
Reference U1 is a 2.5-, 3.0-, 5-, or 10-V reference with a low noise buffered output. The
output of U1 is applied to the R1-C1/C2 noise filter to produce a corner frequency of
about 1.7 Hz. Electrolytic capacitors usually imply dc leakage errors, but the bootstrap
connection of C1 causes its applied bias voltage to be only the relatively small drop
across R2. This lowers the leakage current through R1 to acceptable levels. Since the
filter attenuation is modest below a few Hertz, the reference noise still affects overall
performance at low frequencies (i.e., <10 Hz).
7.19
ANALOG-DIGITAL CONVERSION
+15V
+15V
100Ω
3
2
VIN
VO 6
R1
1kΩ
0.1µF
U1
+
+
R2
10kΩ
GND
4
+
6
U2
C1
100µF
25V
7
100Ω
2
DIODES:
1N4148
4
1.1kΩ
C2
100µF
25V
+
3.3Ω
100µF,25V
+
U1: AD586, AD587, ADR01
ADR02, ADR03, AD42x,
AD43x, AD29x
U2: OP113, OP27
AD797, OP184
10µF
25V
Figure 7.19: Combining Low-noise Amplifier with Extensive Filtering Yields
Exceptional Reference Noise Performance of (1.5 to 5 nV) /√Hz @ 1 kHz
The output of the filter is then buffered by a precision low noise unity-gain follower, such
as the OP113EP. With less than ±150-µV offset error and under 1-µV/°C drift, the buffer
amplifier's dc performance will not seriously affect the accuracy/drift of most references.
For example, an ADR292E for U1 will have a typical drift of 3 ppm/°C, equivalent to
7.5 µV/°C, higher than the buffer amplifier.
Almost any op amp will have a current limit higher than a typical IC reference. Further,
even lower noise op amps are available for 5- to 10-V use. The AD797 offers 1-kHz
noise performance less than 2 nV/√Hz in this circuit, compared to about 5 nV/√Hz for the
OP113. With any amplifier, Kelvin sensing can be used at the load point, a technique
which can eliminate I×R related output voltage errors.
Scaled References
A useful approach when a non-standard reference voltage is required is to simply buffer
and scale a basic low voltage reference diode. With this approach, a potential difficulty is
getting an amplifier to work well at such low voltages as 3 V. A workhorse solution is the
low power reference and scaling buffer shown in Figure 7.20. Here a low current 1.2-V
two terminal reference diode is used for D1, which can be either a 1.200-V ADR512,
1.235-V AD589, or the 1.225-V AD1580. Resistor R1 sets the diode current in either
case, and is chosen for the diode minimum current requirement at a minimum supply of
2.7 V. Obviously, loading on the unbuffered diode must be minimized at the VREF node.
7.20
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
+3V OR
MORE
R1
27.4kΩ (AD589, AD1580)
15kΩ (AD510, AD512)
C1
0.1µF
VOUT = VREF
OR
VOUT = VREF × (1 + R2/R3)
+
U1
D1:
ADR510 (+1.000V)
ADR512 (+1.200V)
AD589 (+1.235V)
AD1580 (+1.225V)
R3
U1: SEE TEXT
R2
VREF
(UNBUFFERED)
Figure 7.20: Rail-to-rail Output Op Amps Allow Greatest Flexibility in Low
Dropout References
The amplifier U1 both buffers and optionally scales up the nominal 1.0 or 1.2-V
reference, allowing much higher source/sink output currents. Of course, a higher op amp
quiescent current is expended in doing this, but this is a basic tradeoff of the approach.
Quiescent current is amplifier dependent, ranging from 45 µA/channel with the
OP196/296/496 series to 1000-2000 µA/channel with the OP284 and OP279. The former
series is most useful for very light loads (<2 mA), while the latter series provide device
dependent outputs up to 50 mA. Various devices can be used in the circuit as shown, and
their key specs are summarized in Figure 7.21.
DEVICE*
Iq, mA
Vsat (+)
Vsat (–)
Isc, mA
per channel
V (min @ mA)
V (max @ mA)
min
OP281/481
0.003
4.93 @ 0.05
0.075 @ 0.05
± 3.5
OP193/293
0.017
4.20 @ 1
0.280 @ 1 (typ)
±8
OP196/296/496
0.045
4.30 @ 1
0.400 @ 1
± 4 (typ)
AD8541/42/44
0.045
4.97 @ 1
0.025 @ 1
± 60
OP777
0.220
4.91 @ 1
0.126 @ 1
± 10
AD820/822
0.620
4.89 @ 2
0.055 @ 2
± 15
OP184/284/484
1.250**
4.85 @ 2.5
0.125 @ 2.5
± 7.5
AD8531/32/34
1.400
4.90 @ 10
0.100 @ 10
± 250
* Typical device specifications @ Vs = +5V, TA = 25°C, unless otherwise noted
** Maximum
Figure 7.21: Op Amps Useful in Low Voltage Rail-Rail References and
Regulators
7.21
ANALOG-DIGITAL CONVERSION
In Figure 7.20, without gain scaling resistors R2-R3, VOUT is simply equal to VREF. With
the use of the scaling resistors, VOUT can be set anywhere between a lower limit of VREF,
and an upper limit of the positive rail, due to the op amp's rail-rail output swing. Also,
note that this buffered reference is inherently low dropout, allowing a +4.5-V (or more)
reference output on a +5-V supply, for example. The general expression for VOUT is
shown in the figure, where VREF is the reference voltage.
Amplifier standby current can be further reduced below 20 µA, if an amplifier from the
OP181/281/481 or the OP193/293/493 series is used. This choice will be at some expense
of current drive, but can provide very low quiescent current if necessary. All devices
shown operate from voltages down to 3 V (except the OP279, which operates at 5 V).
Voltage Reference Pulse Current Response
The response of references to dynamic loads is often a concern, especially in applications
such as driving some ADCs and DACs. Fast changes in load current invariably perturb
the output, often outside the rated error band. For example, the reference input to a
sigma-delta ADC may be the switched capacitor circuit shown in Figure 7.22. The
dynamic load causes current spikes in the reference as the capacitor CIN is charged and
discharged. As a result, noise may be induced on the ADC reference circuitry.
RIN
SIGMA-DELTA ADC
VREF IN
+
CEXT
CIN
~ 10pF
AGND
Figure 7.22: Switched Capacitor Input of Sigma-Delta ADC Presents a Dynamic
Load to the Voltage Reference
Although sigma-delta ADCs have an internal digital filter, transients on the reference
input can still cause appreciable conversion errors. Thus it is important to maintain a low
noise, transient free potential at the ADC's reference input. Be aware that if the reference
source impedance is too high, dynamic loading can cause the reference input to shift by
more than 5 mV.
A bypass capacitor on the output of a reference may help it to cope with load transients,
but many references are unstable with large capacitive loads. Therefore it is quite
7.22
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
important to verify that the device chosen will satisfactorily drive the output capacitance
required. In any case, the converter reference inputs should always be decoupled—with at
least 0.1 µF, and with an additional 5-50 µF if there is any low frequency ripple on its
supply. See Figure 7.15 (again).
Since some references misbehave with transient loads, either by oscillating or by losing
accuracy for comparatively long periods, it is advisable to test the pulse response of
voltage references which may encounter transient loads. A suitable circuit is shown in
Figure 7.23. In a typical voltage reference, a step change of 1 mA produces the transients
shown. Both the duration of the transient, and the amplitude of the ringing increase when
a 0.01-µF capacitor is connected to the reference output.
TOP TRACE: NO LOAD (CL = 0)
50mV/div.
VIN
1mA to 2mA STEP
SCOPE
REFERENCE
UNDER
TEST
CL
RL
BOTTOM TRACE: CL = 0.01µF
200mV/div.
PULSE
GENERATOR
BOTH TRACES: 5µs/div.
Figure 7.23: Make Sure Reference is Stable with Large Capacitive Loads
Where possible, a reference should be designed to drive large capacitive loads. The
AD780 is designed to drive unlimited capacitance without oscillation, it has excellent
drift and an accurate output, in addition to relatively low power consumption. Other
references which are useful with output capacitors are the REF19x, the AD1582-AD1585
series, the ADR29x-series, and the ADR43x-series.
As noted above, reference bypass capacitors are useful when driving the reference inputs
of successive-approximation ADCs. Figure 7.24 illustrates reference voltage settling
behavior immediately following the "Start Convert" command. A small capacitor
(0.01 µF) does not provide sufficient charge storage to keep the reference voltage stable
during conversion, and errors may result. As shown by the bottom trace, decoupling with
a ≥ 1-µF capacitor maintains the reference stability during conversion.
Where voltage references are required to drive large capacitances, it is also critically
important to realize that their turn-on time will be prolonged. Experiment may be needed
to determine the delay before the reference output reaches full accuracy, but it will
certainly be much longer than the time specified on the data sheet for the same reference
in a low capacitance loaded state.
7.23
ANALOG-DIGITAL CONVERSION
VIN
SCOPE
START
CONVERT
VREF
AD780
CB
SAR
ADC
CB = 0.01µF
CB = 0.22µF
CB = 1µF
START CONVERT
SCOPE
TOP TRACE VERTICAL SCALE: 5V/div.
ALL OTHER VERTICAL SCALES: 5mV/div.
HORIZONTAL SCALE: 1µs/div.
Figure 7.24: Successive Approximation ADCs Can Present a Dynamic Transient
Load to the Reference
Low Noise References for High Resolution Converters
High resolution converters (both sigma-delta and high speed ones) can benefit from
recent improvements in IC references, such as lower noise and the ability to drive
capacitive loads. Even though many data converters have internal references, the
performance of these references is often compromised because of the limitations of the
converter process. In such cases, using an external reference rather than the internal one
often yields better overall performance. For example, the AD7710-series of 22-bit ADCs
has a 2.5-V internal reference with a 0.1-Hz to 10-Hz noise of 8.3-µV rms
(2600 nV/√Hz), while the AD780 reference noise is only 0.67 µV rms (200 nV/√Hz).
The internal noise of the AD7710-series in this bandwidth is about 1.7-µV rms. The use
of the AD780 increases the effective resolution of the AD7710 from about 20.5 bits to
21.5 bits.
Figure 7.25 shows the low noise ADR431 used as the +2.5-V reference for the AD77xxseries ADCs. Optimally, the use of the ADR433 (3-V output) enhances the dynamic
range of the ADC, while lowering overall system noise as described above. In addition,
the ADR43x-series allow a large decoupling capacitor on its output thereby minimizing
conversion errors due to transients.
There is one possible but yet quite real problem when replacing the internal reference of a
converter with a higher precision external one. The converter in question may have been
trimmed during manufacture to deliver its specified performance with a relatively
inaccurate internal reference. In such a case, using a more accurate external reference
with the converter may actually introduce additional gain error! For example, the early
AD574 had a guaranteed uncalibrated gain accuracy of 0.125% when using an internal
10-V reference (which itself had a specified accuracy of only ±1%). It is obvious that if
such a device, having an internal reference which is at one end of the specified range, is
used with an external reference of exactly 10 V, then its gain will be about 1% in error.
7.24
DATA CONVERTER SUPPORT CIRCUITS
7.1 VOLTAGE REFERENCES
+5V (ANALOG)
+
VIN
1µF
+2.5V
VO
AVDD
REF IN(+)
ADR431
+
GND
10µF
AD77xx
Σ∆ ADC
REF IN (–)
AGND
NOTE: ONLY REFERENCE
CONNECTIONS SHOWN
Figure 7.25: The AD431 XFET Reference is Ideal for Driving
Precision Sigma-Delta ADCs
7.25
ANALOG-DIGITAL CONVERSION
REFERENCES:
7.1 VOLTAGE REFERENCES
1.
Bob Widlar, "New Developments in IC Voltage Regulators," IEEE Journal of Solid State Circuits,
Vol. SC-6, February, 1971.
2.
Paul Brokaw, "A Simple Three-Terminal IC Bandgap Voltage Reference," IEEE Journal of Solid
State Circuits, Vol. SC-9, December, 1974.
3.
Paul Brokaw, "More About the AD580 Monolithic IC Voltage Regulator," Analog Dialogue, 9-1,
1975.
4.
Dan Sheingold, Section 20.2 within Analog-Digital Conversion Handbook, 3d. Edition, PrenticeHall, 1986.
5.
Walt Jung, "Build an Ultra-Low-Noise Voltage Reference," Electronic Design Analog Applications
Issue, June 24, 1993.
6.
Walt Jung, "Getting the Most from IC Voltage References," Analog Dialogue, 28-1, 1994, pp. 13-21.
7.26
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
SECTION 7.2: LOW DROPOUT LINEAR
REGULATORS
Walt Jung
Introduction
Linear IC voltage regulators have long been standard power system building blocks.
After an initial introduction in 5-V logic voltage regulator form, they have since
expanded into other standard voltage levels spanning from 1.5 to 24 V, handling output
currents from as low as 100 mA (or less) to as high as 5 A (or more). For several good
reasons, linear style IC voltage regulators have been valuable system components since
the early days. One reason is the relatively low noise characteristic vis-à-vis the switching
type of regulator. Others are a low parts count and overall simplicity compared to discrete
solutions. But, because of their power losses, these linear regulators have also been
known for being relatively inefficient. Early generation devices (of which many are still
available) required 2 V or more of unregulated input above the regulated output voltage,
making them lossy in power terms.
More recently however, linear IC regulators have been developed with more liberal (i.e.,
lower) limits on minimum input-output voltage. This voltage, known more commonly as
dropout voltage, has led to what is termed the low drop out regulator, or more popularly,
the LDO. Dropout voltage (VMIN) is defined simply as that minimum input-output
differential where the regulator undergoes a 2% reduction in output voltage. For example,
if a nominal 5.0-V LDO output drops to 4.9 V (–2%) under conditions of an input-output
differential of 0.5 V, by this definition the LDO's VMIN is 0.5 V.
As will be shown in this section, dropout voltage is extremely critical to a linear regulator
stage's power efficiency. The lower the voltage allowable across a regulator while still
maintaining a regulated output, the less power the regulator dissipates as a result. A low
regulator dropout voltage is the key to this, as it takes this lower dropout to maintain
regulation as the input voltage lowers. In performance terms, the bottom line for LDOs is
simply that more useful power is delivered to the load and less heat is generated in the
regulator. LDOs are key elements of power systems that must provide stable voltages
from batteries, such as portable computers, cellular phones, etc. This is simply because
they maintain their regulated output down to lower points on the battery's discharge
curve. Or, within classic mains-powered raw dc supplies, LDOs allow lower transformer
secondary voltages, reducing system susceptibility to shutdown under brownout
conditions as well as allowing cooler operation.
Linear Voltage Regulator Basics
A brief review of three terminal linear IC regulator fundamentals is necessary to
understanding the LDO variety. As it turns out, almost all LDOs available today, as well
as many of the more general three terminal regulator types, are positive leg, series style
regulators. This simply means that they control the regulated voltage output by means of
a pass element which is in series with the positive side of the unregulated input.
7.27
ANALOG-DIGITAL CONVERSION
This is shown more clearly in Figure 7.26, which is a hookup diagram for a hypothetical
three terminal style regulator. To reiterate what was said earlier in the chapter about
reference ICs, in terms of their basic functionality, many standard voltage regulator ICs
are available in the series three-terminal form as is shown here (VIN, GND or Common,
VOUT).
VIN (6V)
VMIN = VIN – VOUT = 1V
IN
OUT
VOUT (5V)
THREE
TERMINAL
REGULATOR
IL
(1A)
GND
RL (5Ω)
IGROUND
(1mA)
COMMON
Figure 7.26: A Basic Three Terminal Voltage Regulator
This diagram also allows some statements to be made about power losses in the regulator.
There are two components to power which are dissipated in the regulator, one a function
of VIN – VOUT and IL, plus a second which is a function of VIN and Iground. If we call the
total power PD, this then becomes:
PD = ( VIN − VOUT )( I L ) + ( VIN )( I ground ) .
Eq. 7.10
Obviously, the magnitude of the load current and the regulator dropout voltage both
greatly influence the power dissipated. However, it is also easy to see that for a given IL,
as the dropout voltage is lowered, the first term of PD is reduced. With an intermediate
dropout voltage rating of 1 V, a 1-A load current will produce 1 W of heat in this
regulator, which will require a heat sink for continuous operation. It is this first term of
the regulator power which usually predominates, at least for loaded regulator conditions.
The second term, being proportional to Iground (typically only 1-2 mA, sometimes even
less) usually only becomes significant when the regulator is unloaded, and the regulator's
quiescent or standby power then produces a constant drain on the source VIN.
7.28
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
However, it should be noted that in some types of regulators (notably those which have
very low β pass devices such as lateral PNP transistors) the Iground current under load can
actually run quite high. This effect is worst at the onset of regulation, or when the pass
device is in saturation, and can be noted by a sudden Iground current "spike", where the
current jumps upward abruptly from a low level. All LDO regulators using bipolar
transistor pass devices which can be saturated (such as PNPs) can show this effect. It is
much less severe in PNP regulators using vertical PNPs (since these have a higher
intrinsic β) and doesn't exist to any major extent in PMOS LDOs (since PMOS transistors
are controlled by voltage level, not current).
In the example shown, the regulator delivers 5 V × 1 A, or 5 W to the load. With a
dropout voltage of 1 V, the input power is 6-V times the same 1 A, or 6 W. In terms of
power efficiency, this can be calculated as:
P
PEFF (%) = 100 × OUT ,
PIN
Eq. 7.11
where POUT and PIN are the total output and input powers, respectively.
In these sample calculations, the relatively small portion of power related to Iground will be
ignored for simplicity, since this power is relatively small. In an actual design, this
simplifying step may not be justified.
In the case shown, the efficiency would be 100 × 5/6, or about 83%. But by contrast, if an
LDO were to be used with a dropout voltage of 0.1 V instead of 1 V, the input voltage
can then be allowed to go as low as 5.1 V. The new efficiency for this condition then
becomes 100 × 5/5.1, or 98%. It is obvious that an LDO can potentially greatly enhance
the power efficiency of linear voltage regulator systems.
A more detailed look within a typical regulator block diagram reveals a variety of
elements, as is shown in Figure 7.27.
In this diagram virtually all of the elements shown can be considered to be fundamentally
necessary, the exceptions being the shutdown control and saturation sensor functions
(shown dotted). While these are present on many current regulators, the shutdown feature
is relatively new as a standard function, and certainly isn't part of standard three-terminal
regulators. When present, shutdown control is a logic level controllable input, whereby a
digital HIGH (or LO) is defined as regulation active (or vice-versa).
The error output, ERR , is useful within a system to detect regulator overload, such as
saturation of the pass device, thermal overload, etc. The remaining functions shown are
always part of an IC power regulator.
7.29
ANALOG-DIGITAL CONVERSION
VIN
VDROPOUT = VMIN = VIN – VOUT
CURRENT
LIMIT
ERR
OVERLOAD
SATURATION
SENSOR
OVERTEMP
SENSOR
SD
SHUTDOWN
CONTROL
PASS
DEVICE
VOUT
IREF
R1
R1 

VOUT = VREF  1 +


R 2
ERROR
AMP
R2
VREF
COMMON
Figure 7.27: Block Diagram of a Voltage Regulator
In operation, a voltage reference block produces a stable voltage VREF, which is almost
always a bandgap based voltage, typically ~1.2 V, which allows output voltages of 3 V or
more from supplies as low as 5 V. This voltage is presented to one input of an error
amplifier, with the other input connected to the VOUT sensing divider, R1-R2. The error
amplifier drives the pass device, which in turn controls the output. The resulting
regulated voltage is then simply:
R1 

VOUT = VREF 1 +
.
R2 

Eq. 7.12
With a typical bandgap reference voltage of 1.2 V, the R1/R2 ratio will be approximately
3/1 for a 5-V output. When standby power is critical, several design steps will be taken.
The resistor values of the divider will be high, the error amplifier and pass device driver
will be low power, and the reference current IREF will also be low. By these means the
regulator's unloaded standby current can be reduced to a mA or less using bipolar
technology, and to only a few µA in CMOS parts. In regulators which offer a shutdown
mode, the shutdown state standby current will be reduced to a µA or less.
Nearly all regulators will have some means of current limiting and over temperature
sensing, to protect the pass device against failure. Current limiting is usually by a series
sensing resistor for high current parts, or alternately by a more simple drive current limit
to a controlled β pass device (which achieves the same end). For higher voltage circuits,
this current limiting may also be combined with voltage limiting, to provide complete
load line control for the pass device.
7.30
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
All power regulator devices will also have some means of sensing over-temperature,
usually by means of a fixed reference voltage and a VBE-based sensor monitoring chip
temperature. When the die temperature exceeds a dangerous level (above ~150°C), this
can be used to shutdown the chip, by removing the drive to the pass device. In some
cases an error flag output may be provided to warn of this shutdown (and also loss of
regulation from other sources).
Pass Devices and their Associated Tradeoffs
The discussion thus far has not treated the pass device in any detail. In practice, this
major part of the regulator can actually take on quite a number of alternate forms.
Precisely which type of pass device is chosen has a major influence on almost all major
regulator performance issues. Most notable among these is the dropout voltage, VMIN.
Figure 7.28a through 7.28e illustrates a number of pass devices which are useful within
voltage regulator circuits, shown in simple schematic form. On the figure is also listed the
salient VMIN for the device as it would typically be used, which directly indicates its
utility for use in an LDO. Not shown in these various mini-figures are the remaining
circuits of a regulator.
It is difficult to fully compare all of the devices from their schematic representations,
since they differ in so many ways beyond their applicable dropout voltages. For this
reason, the chart of Figure 7.29 is useful.
VIN
VIN
VMIN ≅ 2V
VMIN ≅ 1V
VOUT
VOUT
(a) SINGLE NPN
(b) DARLINGTON NPN
VIN
VIN
VMIN ≅ VCE(SAT)
Q1
VOUT
VMIN ≅ 1.5V
Q1
Q2
Q2
(d) PNP/NPN
(c) SINGLE PNP
VIN
VMIN ≅ RDS(ON) × IL
VOUT
P1
VOUT
Q1
(e) PMOS
Figure 7.28: Pass Devices Useful in Voltage Regulators
7.31
ANALOG-DIGITAL CONVERSION
A
B
SINGLE
NPN
DARLINGTON
NPN
C
D
E
SINGLE
PNP
PNP/NPN
PMOS
V MIN ~ 1V
V MIN ~ 2V
V MIN ~ 0.1V
VMIN ~ 1.5V
V MIN ~ R DS(ON) × I L
IL < 1A
IL > 1A
IL < 1A
IL > 1A
IL > 1A
Follower
Follower
Inverter
Inverter
Inverter
Low Z OUT
Low Z OUT
High Z OUT
High Z OUT
High Z OUT
Wide BW
Wide BW
Narrow BW
Narrow BW
Narrow BW
C L Immune
C L Immune
C L Sensitive
C L Sensitive
C L Sensitive
Figure 7.29: Pros and Cons of Voltage Regulator Pass Devices
This chart compares the various pass elements in greater detail, allowing easy
comparison between the device types, dependent upon which criteria is most important.
Note that columns A-E correspond to the schematics of Figure 7.28a-7.28e. Note also
that the pro/con comparison items are in relative terms, as opposed to a hard specification
limit for any particular pass device type.
For example, it can be seen that the all NPN pass devices of columns A and B have the
attributes of a follower circuit, which allows high bandwidth and provides relative
immunity to cap loading because of the characteristic low ZOUT. However, neither the
single NPN nor the Darlington NPN can achieve low dropout, for any load current. This
is because the VBE(s) of the pass device appears in series with the input, preventing its
saturation, and thus setting a VMIN of about 1 or 2 V.
By contrast, the inverting mode device connections of both columns C and E do allow the
pass device to be effectively saturated, which lowers the associated voltage losses to a
minimum. This single factor makes these two pass device types optimum for LDO use, at
least in terms of power efficiency.
For currents below 1 A, either a single PNP or a PMOS pass device is most useful for
low dropout, and they both can achieve a VMIN of 0.1 V or less at currents of 100 mA.
The dropout voltage of a PNP will be highly dependent upon the actual device used and
the operating current, with vertical PNP devices being superior for saturation losses, as
well as minimizing the Iground spike when in saturation. PMOS pass devices offer the
potential for the lowest possible VMIN, since the actual dropout voltage will be the
product of the device RDS(ON) and IL. Thus a low RDS(ON) PMOS device can always be
chosen to minimize VMIN for a given IL. PMOS pass devices are typically external to the
LDO IC, making the IC actually a controller (as opposed to a complete and integral
LDO). PMOS pass devices can allow currents up to several amps or more with very low
dropout voltages. The PNP/NPN connection of column D is actually a hybrid hookup,
intended to boost the current of a single PNP pass device. This it does, but it also adds the
7.32
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
VBE of the NPN in series (which cannot be saturated), making the net VMIN of the
connection about 1.5 V.
All of the three connections C/D/E have the characteristic of high output impedance, and
require an output capacitor for stability. The fact that the output cap is part of the
regulator frequency compensation is a most basic application point, and one which needs
to be clearly understood by the regulator user. This factor, denoted by "CL sensitive",
makes regulators using them generally critical as to the exact CL value, as well as its ESR
(equivalent series resistance). Typically this type of regulator must be used only with a
specific size as well as type of output capacitor, where the ESR is controlled with respect
to both time and temperature to fully guarantee regulator stability. Fortunately, some
recent Analog Devices LDO IC circuit developments have eased this burden on the part
of the regulator user a great deal, and will be discussed below in further detail.
Some examples of standard IC regulator architectures illustrate the points above
regarding pass devices, and allow an appreciation of regulator developments leading up
to more recent LDO technologies.
The classic LM309 5-V/1-A three-terminal regulator (see Reference 1) was the originator
in a long procession of regulators. This circuit is shown in much simplified form in
Figure 7.30, with current limiting and over temperature details omitted. This IC type is
still in standard production today, not just in original form, but in family derivatives such
as the 7805, 7815 etc., and their various low and medium current alternates. Using a
Darlington pass connection for Q18-Q19, the design has never been known for low
dropout characteristics (~1.5-V typical), or for low quiescent current (~5 mA). It is
however relatively immune to instability issues, due to the internal compensation of C1,
and the buffering of the emitter follower output. This helps make it easy to apply.
The LM109/309 bandgap voltage reference actually used in this circuit consists of a more
involved scheme, as opposed to the basic form which was described with Figure 7.3.
Resistor R8 drops a PTAT voltage, which drives the Darlington connected error
amplifier, Q9-Q10. The negative TC VBEs of Q9-Q10 and Q12-Q13 are summed with
this PTAT voltage, and this sum produces a temperature-stable 5-V output voltage.
Current buffering of the error amplifier Q10 is provided by PNP Q11, which drives the
NPN pass devices.
Later developments in references and three-terminal regulation techniques led to the
development of the voltage adjustable regulator. The original IC to employ this concept
was the LM317 (see Reference 2), which is shown in simplified schematic form in Figure
7.31. Note that this design does not use the same ∆VBE form of reference as in the
LM309. Instead, Q17-Q19, etc. are employed as a form of a Brokaw bandgap reference
cell (see Figure 7.4 again, and Reference 3).
7.33
ANALOG-DIGITAL CONVERSION
Q17
VIN
Q18
Q19
Q2
VOUT
Q13
Q3
R9
R1
Q11
Q12
R2
Q4
C1
R3
R8
Q9
Q6
Q7
Q10
Q5
R4
R7
R5
R6
COMMON
Q8
Figure 7.30: Simplified Schematic of LM309 Fixed 5-V/1-A
Three-Terminal Regulator
VIN
Q18
Q16
Q25
Q26
Q12
VOUT
Q19
Q17
R15
VREF = 1.25V
R14
R1
50µA
ADJ
R2

VOUT = VREF  1 +
 + 50µA × R 2

R1
R2
Figure 7.31: Simplified Schematic of LM317 Adjustable Three-terminal Regulator
7.34
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
This adjustable regulator bootstraps the reference cell transistors Q17-Q19 and the error
amplifier transistors Q16-Q18. The output of the error amplifier drives Darlington pass
transistors Q25-Q26, through buffer Q12. The basic reference cell produces a fixed
voltage of 1.25 V, which appears between the VOUT and ADJ pins of the IC as shown.
External scaling resistors R1 and R2 set up the desired output voltage, which is:
R2 

VOUT = VREF 1 +
 + 50µA × R 2 .
R1 

Eq. 7.13
As can be noted, the voltage output is a scaling of VREF by R2-R1, plus a small voltage
component which is a function of the 50-µA reference cell current. Typically, the R1-R2
values are chosen to draw >5 mA, making the rightmost term relatively small by
comparison. The design is internally compensated, and in many applications will not
necessarily need an output bypass capacitor.
Like the LM309 fixed voltage regulator, the LM317 series has relatively high dropout
voltage, due to the use of Darlington pass transistors. It is also not a low power IC
(quiescent current typically 3.5 mA). The strength of this regulator lies in the wide range
of user voltage adaptability it allows.
Subsequent variations on the LM317 pass device topology modified the method of output
drive, substituting a PNP/NPN cascade for the LM317's Darlington NPN pass devices.
This development achieves a lower VMIN, 1.5 V or less (see Reference 4). The
modification also allows all of the general voltage programmability of the basic LM317,
but at some potential increase in application sensitivity to output capacitance. This
sensitivity is brought about by the fundamental requirement for an output capacitor for
the IC's frequency compensation, which is a differentiation from the original LM317.
Low Dropout Regulator Architectures
As has been shown thus far, all LDO pass devices have the fundamental characteristics of
operating in an inverting mode. This allows the regulator circuit to achieve pass device
saturation, and thus low dropout. A by-product of this mode of operation is that this type
of topology will necessarily be more susceptible to stability issues. These basic points
give rise to some of the more difficult issues with regard to LDO performance. In fact,
these points influence both the design and the application of LDOs to a very large degree,
and in the end, determine how they are differentiated in the performance arena.
A traditional LDO architecture is shown in Figure 7.32, and is generally representative of
actual parts employing either a PNP pass device as shown, or alternately, a PMOS
device. There are both dc and ac design and application issues to be resolved with this
architecture, which are now discussed.
7.35
ANALOG-DIGITAL CONVERSION
VIN
VOUT
Q1
PNP (OR PMOS)
PASS DEVICE
R1
CL
RL
–
Q2
ESR
gm
+
IGROUND
CCOMP
+
R2
VREF
Figure 7.32: Traditional LDO Architecture
In dc terms, perhaps the major issue is the type of pass device used, which influences
dropout voltage and ground current. If a lateral PNP device is used for Q1, the β will be
low, sometimes only on the order or 10 or so. Since Q1 is driven from the collector of
Q2, the relatively high base current demanded by a lateral PNP results in relatively high
emitter current in Q2, or a high Iground. For a typical lateral PNP based regulator operating
with a 5-V/150-mA output, Iground will be typically ~18 mA, and can be as high as 40 mA.
To compound the problem of high Iground in PNP LDOs, there is also the "spike" in Iground,
as the regulator is operating within its dropout region. Under such conditions, the output
voltage is out of tolerance, and the regulation loop forces higher drive to the pass device,
in an attempt to maintain loop regulation. This results in a substantial spike upward in
Iground, which is typically internally limited by the regulator's saturation control circuits.
PMOS pass devices do not demonstrate a similar current spike in Iground, since they are
voltage controlled. But, while devoid of the Iground spike, PMOS pass devices do have
some problems of their own. Problem number one is that high quality, low RON, low
threshold PMOS devices generally aren't compatible with many IC processes. This makes
the best technical choice for a PMOS pass device an external part, driven from the
collector of Q2 in the figure. This introduces the term "LDO controller", where the LDO
architecture is completed by an external pass device. While in theory NMOS pass devices
would offer lower RON choice options, they also demand a boosted voltage supply to turn
on, making them impractical for a simple LDO. PMOS pass devices are widely available
in low both RON and low threshold forms, with current levels up to several amperes. They
offer the potential of the lowest dropout of any device, since dropout can always be
lowered by picking a lower RON part.
The dropout voltage of lateral PNP pass devices is reasonably good, typically around
300 mV at 150 mA, with a maximum of 600 mV. These levels are however considerably
bettered in regulators using vertical PNPs, which have a typical β of ~150 at currents of
200 mA. This leads directly to an Iground of 1.5 mA at the 200-mA output current. The
7.36
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
dropout voltage of vertical PNPs is also an improvement vis-à-vis that of the lateral PNP
regulator, and is typically 180 mV at 200 mA, with a maximum of 400 mV.
There are also major ac performance issues to be dealt with in the LDO architecture of
Figure. 7.32. This topology has an inherently high output impedance, due to the operation
of the PNP pass device in a common-emitter (or common-source with a PMOS device)
mode. In either case, this factor causes the regulator to appear as a high source impedance
to the load.
The internal compensation capacitor of the regulator, CCOMP, forms a fixed frequency
pole, in conjunction with the gm of the error amplifier. In addition, load capacitance CL
forms an output pole, in conjunction with RL. This particular pole, because it is a second
(and sometimes variable) pole of a two-pole system, is the source of a major LDO
application problem. The CL pole can strongly influence the overall frequency response
of the regulator, in ways that are both useful as well as detrimental. Depending upon the
relative positioning of the two poles in the frequency domain, along with the relative
value of the ESR of capacitor CL, it is quite possible that the stability of the system can be
compromised for certain combinations of CL and ESR. Note that CL is shown here as a
real capacitor, which is actually composed of a pure capacitance plus the series parasitic
resistance ESR.
Without a heavy duty exercise into closed-loop stability analysis, it can safely be said that
LDOs, like other feedback systems, need to satisfy certain basic stability criteria. One of
these is the gain-versus-frequency rate-of-change characteristic in the region approaching
the system's unity loop gain crossover point. For the system to be closed loop stable, the
phase shift must be less than 180° at the point of unity gain. In practice, a good feedback
design needs to have some phase margin, generally 45° or more to allow for various
parasitic effects. While a single pole system is intrinsically stable, two pole systems are
not necessarily so—they may in fact be stable, or they may also be unstable. Whether or
not they are stable for a given instance is highly dependent upon the specifics of their
gain-phase characteristics.
If the two poles of such a system are widely separated in terms of frequency, stability
may not be a serious problem. The emitter-follower output of a classic regulator like the
LM309 is an example with widely separated pole frequencies, as the very low ZOUT of the
NPN follower pushes the output pole due to load capacitance far out in frequency, where
it does little harm. The internal compensation capacitance (C1 of Fig. 7.30, again) then
forms part of a dominant pole, which reduces loop gain to below unity at the much higher
frequencies where the output pole does occur. Thus stability is not necessarily
compromised by load capacitance in this type of regulator.
Figure 7.33 summarizes the various dc and ac design issues of LDOs.
7.37
ANALOG-DIGITAL CONVERSION
DC
AC
Lateral PNP Pass Device:
High IGROUND
Two Pole Compensation
System
Vertical PNP Pass Device:
Low IGROUND
Low VMIN
CL ESR Critical to Stability
PMOS Pass Device:
Lowest IGROUND Variation
Low VMIN
Ampere Level Output
Currents
Requires Large CL
Requires"Zoned" CL ESR
(Max/Min ESR Limits Over Time
and Temperature)
Figure 7.33: DC and AC Design Issues in Low Dropout Regulators
By their nature however, LDOs simply can't afford the luxury of emitter follower outputs,
they must instead operate with pass devices capable of saturation. Thus, given the
existence of two or more poles (one or more internal and a second formed by external
loading) there is the potential for the cumulative gain-phase to add in a less than
satisfactory manner. The potential for instability under certain output loading conditions
is, for better or worse, a fact-of-life for most LDO topologies.
However, the output capacitor which gives rise to the instability can, for certain
circumstances, also be the solution to the same instability. This seemingly paradoxical
situation can be appreciated by realizing that almost all practical capacitors are actually
as shown in Fig. 7.32, a series combination of the capacitance CL and a parasitic
resistance, ESR. While load resistance RL and CL do form a pole, CL and its ESR also
form a zero. The effect of the zero is to mitigate the de-stabilizing effect of CL for certain
conditions.
For example, if the pole and zero in question are appropriately placed in frequency
relative to the internal regulator poles, some of the deleterious effects can be made to
essentially cancel, leaving little or no problematic instability (see Reference 5). The basic
problem with this setup is simply that the capacitor's ESR, being a parasitic term, is not at
all well controlled. As a result, LDOs which depend upon output pole-zero compensation
schemes must very carefully limit the capacitor ESR to certain zones, such as shown by
Figure 7.34.
7.38
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
A zoned ESR chart such as this is meant to guide the user of an LDO in picking an output
capacitor which confines ESR to the stable region, i.e., the central zone, for all operating
conditions. Note that this generic chart is not intended to portray any specific device, just
the general pattern. Unfortunately, capacitor facts of life make such data somewhat
limited in terms of the real help it provides. Bearing in mind the requirements of such a
zoned chart, it effectively means that general purpose aluminum electrolytic are
prohibited from use, since they deteriorate in terms of ESR at cold temperatures. Very
low ESR types such as OS-CON or multi-layer ceramic units have ESRs which are too
low for use. While they could in theory be padded up into the stable zone with external
resistance, this would hardly be a practical solution. This leaves tantalum types as the
best all around choice for LDO output use. Finally, since a large capacitor value is likely
to be used to maximize stability, this effectively means that the solution for an LDO such
as Fig. 7.32 must use a more expensive and physically large tantalum capacitor. This is
not desirable if small size is a major design criteria.
100
UNSTABLE
10
CAPACITOR
ESR (Ω)
STABLE
1
UNSTABLE
0.1
0
IOUT (mA)
1000
Figure 7.34: Zoned Load Capacitor ESR Can Make
an LDO Applications Nightmare
The anyCAP® Low Dropout Regulator Family
Some novel modifications to the basic LDO architecture of Fig. 7.32 allow major
improvements in terms of both dc and ac performance. These developments are shown
schematically in Figure 7.35, which is a simplified diagram of the Analog Devices
ADP330x, and ADP333x-series LDO regulator family. These regulators are also known
as the anyCAP® family, so named for their relative insensitivity to the output capacitor in
terms of both size and ESR. They are available in power efficient packages such as the
Thermal Coastline (discussed below), in both stand-alone LDO and LDO controller
forms, and also in a wide span of output voltage options.
7.39
ANALOG-DIGITAL CONVERSION
VIN
VOUT
Q1
R1
CCOMP
NONINVERTING
WIDEBAND
DRIVER
+
gm
–
PTAT
VOS
R3 D1
CL
R1||R2
×1
RL
R4
IPTAT
R2
GND
Figure 7.35: ADP330x and ADP333x anyCAP® Topology Features Improved
DC & AC Performance Over Traditional LDOs
Design Features Related to DC Performance
One of the key differences in the ADP330x/ADP333x-series is the use of a high gain
vertical PNP pass device, with all of the advantages described above with Figs. 7.32 and
7.33 (also, see Reference 6). This allows the typical dropout voltages for the series to be
on the order of 1 mV/mA for currents of 200 mA or less.
It is important to note that the topology of this LDO is distinctly different from that of the
generic form in Fig. 7.32, as there is no obvious VREF block. The reason for this is the fact
that the ADP330x/ADP333x-series uses what is termed a "merged" amplifier-reference
design. The operation of the integral amplifier and reference scheme illustrated in Fig.
7.35 can be described as follows.
In this circuit, VREF is defined as a reference voltage existing at the output of a zero
impedance divider of ratio R1/R2. In the figure, this is depicted symbolically by the
(dotted) unity gain buffer amplifier fed by R1/R2, which has an output of VREF. This
reference voltage feeds into a series connection of (dotted) R1||R2, then actual
components D1, R3, R4, etc.
The error amplifier, shown here as a gm stage, is actually a PNP input differential stage
with the two transistors of the pair operated at different current densities, so as to produce
a predictable PTAT offset voltage. Although shown here as a separate block VOS, this
offset voltage is inherent to a bipolar pair for such operating conditions. The PTAT VOS
causes a current IPTAT to flow in R4, which is simply:
I PTAT =
7.40
VOS
.
R4
Eq. 7.14
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
Note that this current also flows in series connected R4, R3, and the Thevenin resistance
of the divider, R1||R2, so:
VPTAT = I PTAT ( R 3 + R 4 + R1 || R 2) .
Eq. 7.15
The total voltage defined as VREF is the sum of two component voltages:
VREF = VPTAT + VD1 ,
Eq. 7.16
where the IPTAT scaled voltages across R3, R4, and R1||R2 produce a net PTAT voltage
VPTAT, and the diode voltage VD1 is a CTAT voltage. As in a standard bandgap reference,
the PTAT and CTAT components add up to a temperature stable reference voltage of
1.25 V. In this case however, the reference voltage is not directly accessible, but instead
it exists in the virtual form described above. It acts as it would be seen at the output of a
zero impedance divider of a numeric ratio of R1/R2, which is then fed into the R3-D1
series string through a Thevenin resistance of R1||R2 in series with D1.
With the closed loop regulator at equilibrium, the voltage at the virtual reference node
will be:
 R2 
VREF = VOUT 
.
 R1 + R 2 
Eq. 7.17
With minor re-arrangement, this can be put into the standard form to describe the
regulator output voltage, as:
R1 

VOUT = VREF 1 +
.
R2 

Eq. 7.18
In the various devices of the ADP330x/ADP333x-series, the R1-R2 divider is adjusted to
produce various standard output voltages from 1.5 V to 5.0 V.
As can be noted from this discussion, unlike a conventional reference setup, there is no
power wasting reference current such as used in a conventional regulator topology (IREF
of Fig. 7.27). In fact, the Fig. 7.35 regulator behaves as if the entire error amplifier has
simply an offset voltage of VREF volts, as seen at the output of a conventional R1-R2
divider.
Design Features Related to AC Performance
While the above described dc performance enhancements of the ADP330x series are
worthwhile, the most dramatic improvements come in areas of ac-related performance.
These improvements are in fact the genesis of the anyCAP® series name.
Capacitive loading and the potential instability it brings is a major deterrent
to easily applying LDOs. While low dropout goals prevent the use of emitter follower
type outputs, and so preclude their desirable buffering effect against cap loading, there is
an alternative technique of providing load immunity. One method of providing a measure
7.41
ANALOG-DIGITAL CONVERSION
of insusceptibility against variation in a particular amplifier response pole is called pole
splitting (see Reference 8). It refers to an amplifier compensation method whereby two
response poles are shifted in such a way so as to make one a dominant, lower frequency
pole. In this manner the secondary pole (which in this case is the CL related output pole)
becomes much less of a major contributor to the net ac response. This has the desirable
effect of greatly de-sensitizing the amplifier to variations in the output pole.
A Basic Pole-Splitting Topology
A basic LDO topology with frequency compensation as modified for pole splitting is
shown in Figure 7.36. Here the internal compensation capacitor CCOMP is connected as an
integrating capacitor, around pass device Q1 (C1 is the pass device input capacitance).
While it is true that this step will help immunize the regulator to the CL related pole, it
also has a built in fatal flaw. With CCOMP connected directly to the Q1 base as shown, the
line rejection characteristics of this setup will be quite poor. In effect, when doing it this
way one problem (CL sensitivity) will be exchanged for another (poor line rejection).
VIN
C1
VOUT
Q1
CCOMP
R1
CL
+
gm
–
VREF
ESR
RL
+
R2
Figure 7.36: The Solution to CL Sensitivity: Pole Split Compensation
(Wrong Way Example!)
The anyCAP® Pole-Splitting Topology
Returning to the anyCAP series topology, (Fig. 7.35, again) it can be noted that in this
case CCOMP is isolated from the pass device's base (and thus input ripple variations), by
the wideband non-inverting driver. But insofar as frequency compensation is concerned,
because of this buffer's isolation, CCOMP still functions as a modified pole splitting
capacitor (see Reference 9), and it does provide the benefits of a buffered, CL independent
single-pole response. The regulator's frequency response is dominated by the internal
compensation, and becomes relatively immune to the value and ESR of load capacitor
CL. Thus the name anyCAP for the series is apt, as the design is tolerant of virtually any
output capacitor type.
7.42
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
The benefits of the anyCAP topology are summarized by Figure 7.37. As can be noted,
CL can be as low as 0.47 µF, and it can also be a multi-layer ceramic capacitor (MLCC)
type. This allows a very small physical size for the entire regulation function, such as
when a SOT-23 packaged anyCAP LDO is used, for example the ADP3300 device.
Because of the in-sensitivity to CL, the designer needn't worry about such things as ESR
zones, and can better concentrate on the system aspects of the regulator application.
Internal CCOMP Dominates Response Rolloff
CL Can Range from 0.47µF(min) to Infinity
Low and Ultra-Low CL ESR is OK
MLCC Types for CL Work, is Physically Smallest Solution
No ESR Exclusion Zones
Fast Load Transient Response and Good Line Rejection
Figure 7.37: Benefits of anyCAP® LDO Topology
The anyCAP® LDO series devices
The major specifications of the ADP330x-series anyCAP LDO regulators are
summarized in Figure. 7.38. The devices include both single and dual output parts, with
current capabilities ranging from 50 to 200 mA. Rather than separate individual
specifications for output tolerance, line and load regulation, plus temperature, the
anyCAP series devices are rated simply for a combined total accuracy figure. For the
ADP3300, ADP3301, ADP3302, ADP3303, and ADP3307, this accuracy is either 0.8%
at 25°C, or 1.4% over the temperature range with the device operating over an input
range of VOUT +0.3 V (or 0.5 V), up to 12 V. The ADP3308 and ADP3309 are similarly
specified for a total 25°C accuracy of 1.1% and 2.2% over temperature. With total
accuracy being covered by one clear specification, the designer can then achieve a higher
degree of confidence. It is important to note that this method of specification also
includes operation within the regulator dropout range (unlike some LDO parts specified
for higher input-output voltage difference conditions).
The ADP333x-series are a newer family of anyCAP LDOs designed for 200-mA and
higher output currents with very low quiescent current, IQ. For instance, the ADP3330
has a typical no-load current of only 35 µA. The ADP333x-series are available in
thermally enhanced packages, and Figure 7.39 shows the key specifications for the
family.
7.43
ANALOG-DIGITAL CONVERSION
Part
Number
VMIN @ IL
IL
Accuracy
Package
Comment
(V, typ)
(mA)
(Total over
Temp, %)
(All SO-8 are
Thermal
coastline)
(Singles have
NR, SD, ERR;
Dual no NR)
ADP3300
0.08
50
1.4
SOT-23-6
Single
ADP3301
0.10
100
1.4
SO-8
Single
ADP3302
0.10
100
1.4
SO-8
Dual
ADP3303
0.18
200
1.4
SO-8
Single
ADP3307
0.13
100
1.4
SOT-23-6
Single
ADP3308
0.08
50
2.2
SOT-23-5
Single
ADP3309
0.12
100
2.2
SOT-23-5
Single
Figure 7.38: anyCAP® Series LDO Regulators
Part
Number
VMIN @ IL
IL
Accuracy
Package
(V, typ)
(mA)
(Total over
(SOT23-6
Temp, %)
are Chip on
Comment
Lead)
ADP3330
0.14
200
1.4
SOT-23-6
Single
ADP3331
0.14
200
1.4
SOT-23-6
Single
ADP3333
0.14
300
1.8
MSOP-8
Single
ADP3334
0.20
500
1.8
SO-8
Single
ADP3335
0.20
500
1.8
MSOP-8
Single
ADP3336
0.20
500
1.8
MSOP-8
Single
ADP3338
0.19
1000
1.4
SOT-223
Single
ADP3339
0.23
1500
1.5
SOT-223
Single
Figure 7.39: anyCAP® Series Low IQ LDOs
7.44
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
Functional Diagram and Basic 50-mA LDO Regulator
A functional diagram common to the various devices of the anyCAP series LDO
regulators is shown by Figure 7.40. Operation of the various pins and internal functions is
discussed below.
IN
Q1
THERMAL
PROTECTION
ERR
Q2
OUT
CC
DRIVER
R1
+
gm
–
+
SD
R2
BANDGAP
REF
GND
Figure 7.40: anyCAP® Series LDO Regulators Functional Diagram
In application, the use of the anyCAP series of LDOs is simple, as shown by a basic
50-mA ADP3300 regulator, in Figure 7.41. This circuit is a general one, to illustrate
points common to the entire device series. The ADP3300 is a basic LDO regulator
device, designed for fixed output voltage applications while operating from sources over
a range of 3 to 12 V and a temperature range of –40°C to +85°C. The actual ADP3300
device ordered would be specified as ADP3300ART-YY, where the "YY" is a voltage
designator suffix such as 2.7, 3, 3.2, 3.3, or 5, for those respective voltages. The "ART"
portion of the part number designates the SOT-23 6-lead package. The example circuit
shown produces 5.0 V with the use of the ADP3300-5.
In operation, the circuit will produce its rated 5 V output for loads of 50 mA or less, and
for input voltages above 5.3 V (VOUT + 0.3 V), when the shutdown ( SD ) input is in a
HIGH state. This can be accomplished either by a logic HIGH control input to the
SD pin, or by simply tying this pin to VIN. When SD is LOW (or tied to ground), the
regulator shuts down, and draws a quiescent current of 1 µA or less.
The ADP3300 and other anyCAP series devices maintain regulation over a wide range of
load, input voltage and temperature conditions. However, when the regulator is
overloaded or entering the dropout region (for example, by a reduction in the input
voltage) the open collector ERR pin becomes active, by going to a LOW or conducting
state. Once set, the ERR pin's internal hysteresis keeps the output low, until some
7.45
ANALOG-DIGITAL CONVERSION
margin of operating range is restored. In the circuit of Fig. 7.41, R1 is a pullup resistor
for the ERR output, EOUT. This resistor can be eliminated if the load being driven
provides a pullup current.
VIN
5
IN
NR
SD
GND
3
C3
0.01µF
OUT 4
ADP3300-5
C1
0.47µF
2
ERR
6
VOUT = 5V
R1
330kΩ
C2
0.47µF
1
ON
OFF
Figure 7.41: A Basic ADP3300 50-mA LDO Regulator Circuit
The ERR function can also be activated by the regulator's over temperature protection
circuit, which trips at 165°C. These internal current and thermal limits are intended to
protect the device against accidental overload conditions. For normal operation, device
power dissipation should be externally limited by means of heat sinking, air flow, etc. so
that junction temperatures will not exceed 125°C.
A capacitor, C3, connected between pins 2 and 4, can be used for an optional noise
reduction (NR) feature. This is accomplished by ac-bypassing a portion of the regulator's
internal scaling divider, which has the effect of reducing the output noise ~10 dB. When
this option is exercised, only low leakage 10- to 100-nF capacitors should be used. Also,
input and output capacitors should be changed to 1- and 4.7-µF values respectively, for
lowest noise and the best overall performance. Note that the noise reduction pin is
internally connected to a high impedance node, so connections to it should be carefully
done to avoid noise. PC traces and pads connected to this pin should be as short and small
as possible.
LDO Regulator Thermal Considerations
To determine a regulator's power dissipation, calculate it as follows:
(
)
PD = (VIN − VOUT )(IL ) + (VIN ) Iground ,
Eq. 7.19
where IL and Iground are load and ground current, and VIN and VOUT are the input and
output voltages respectively. Assuming IL= 50 mA, Iground = 0.5 mA, VIN = 8 V, and VOUT
= 5 V, the device power dissipation is:
7.46
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
PD = (8 – 5)(0.05) + (8)(0.0005) = 0.150 + .004 = 0.154 W.
Eq. 7.20
To determine the regulator's temperature rise, ∆T, calculate it as follows (assume the θJA
of the regulator is 165°C/W):
∆T = TJ – TA = PD × θJA = 0.154W × 165°C/W = 25.4°C.
Eq. 7.21
With a maximum junction temperature of 125°C, this yields a calculated maximum safe
ambient operating temperature of 125°C – 25.4°C, or just under 100°C. Since this
temperature is in excess of the device's rated temperature range of 85°C, the device will
then be operated conservatively at an 85°C (or less) maximum ambient temperature.
These general procedures can be used for other devices in the series, substituting the
appropriate θJA for the applicable package, and applying the remaining operating
conditions. For reference, a complete tutorial section on thermal management is
contained in Chapter 9.
In addition, layout and PCB design can have a significant influence on the power
dissipation capabilities of power management ICs. This is due to the fact that the surface
mount packages used with these devices rely heavily on thermally conductive traces or
pads, to transfer heat away from the package. Appropriate PC layout techniques should
then be used to remove the heat due to device power dissipation. The following general
guidelines will be helpful in designing a board layout for lowest thermal resistance in
SOT-23 and SO-8 packages:
1. PC board traces with large cross sectional areas remove more heat. For optimum
results, use large area PCB patterns with wide and heavy (2 oz.) copper traces,
placed on the uppermost side of the PCB.
2. Electrically connect dual VIN and VOUT pins in parallel, as well as to the
corresponding VIN and VOUT large area PCB lands.
3. In cases where maximum heat dissipation is required, use double-sided copper planes
connected with multiple vias.
4. Where possible, increase the thermally conducting surface area(s) openly exposed to
moving air, so that heat can be removed by convection (or forced air flow, if
available).
5. Do not use solder mask or silkscreen on the heat dissipating traces, as they increase
the net thermal resistance of the mounted IC package.
7.47
ANALOG-DIGITAL CONVERSION
A real life example visually illustrates a number of the above points far better than words
can do, and is shown in Figure 7.42, a photo of the ADP3300 1.5" square evaluation
PCB. The boxed area on the board represents the actual active circuit area.
TOTAL
BOARD SIZE:
1.5" X 1.5"
10µF / 16V
TANTALUM
CAPACITOR
(KEMET T491C
SERIES)
Figure 7.42: ADP3300 Evaluation Board:
Capacitor Size Can Make a Difference!
In this figure, a large cross section conductor area can be seen associated with pin 4 and
VOUT, the large "U" shaped trace at the lower part within the boxed outline.
Also, the effect of the anyCAP design on capacitor size can be noted from the tiny size of
the C1 and C2 0.47-µF input and output capacitors, near the upper left of the boxed area.
For comparison purposes, a 10-µF/16-V tantalum capacitor (Kemet T491C-series) is also
shown outside the box, as it might be used on a more conventional LDO circuit. It is
several times the size of output capacitor C2.
Recent developments in packaging have led to much improved thermal performance for
power management ICs. The anyCAP LDO regulator family capitalizes on this most
effectively, using a thermally improved leadframe as the basis for all 8 pin devices. This
package is called a "Thermal Coastline" design, and is shown in Figure 7.43. The
foundation of the improvement in heat transfer is related to two key parameters of the
leadframe design, distance and width. The payoff comes in the reduced thermal resistance
of the leadframe based on the Thermal Coastline, only 90°C/W versus 160°C/W for a
standard SO-8 package. The increased dissipation of the Thermal Coastline allows the
anyCAP series of SO-8 regulators to support more than one watt of dissipation at 25°C.
7.48
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
1
8
1
8
2
7
2
7
3
6
3
6
4
5
4
5
STANDARD LEADFRAME SOIC
THERMAL COASTLINE SOIC
θJA = 160ºC/W
θJA = 90ºC/W
Figure 7.43: anyCAP® Series Regulators in SO-8 Use
Thermal Coastline Packages
Additional insight into how the new leadframe increases heat transfer can be appreciated
by Figure. 7.44. In this figure, it can be noted how the spacing of the Thermal Coastline
paddle and leads shown on the right is reduced, while the width of the lead ends are
increased, versus the standard leadframe, on the left.
STANDARD FRAME
Lead 1
Paddle
THERMAL COASTLINE FRAME
Face-to-face distance,
Lead 1
from lead to paddle
reduced by a factor of
1.5 to 2
Lead 2
Lead 2
Center of
Package
Width of adjoining
faces increased by
factor of 2 to 2.5
Paddle
Center of
Package
Figure 7.44: Details of Thermal Coastline Package
The ADP3330 and ADP3331 are 200-mA anyCAP LDOs packaged in a 6-lead SOT-23
package which utilizes a proprietary Chip-on-Lead™ packaging technique for thermal
enhancement. In a standard SOT-23, the majority of the heat flows out of the ground pin.
This new package uses an electrically isolated die attach that allows all pins to contribute
to heat conduction. This technique reduces the thermal resistance to 165°C/W on a
4-layer board as compared to >230°C/W for a standard SOT-23 leadframe. Figure 7.45
shows the difference between the standard SOT-23 and the Chip-on-Lead leadframes.
7.49
ANALOG-DIGITAL CONVERSION
165°C/W vs. >230°C/W for Standard SOT-23
Figure 7.45: Thermally Enhanced Chip-on-Lead™ SOT-23-6 Package
The ADP3333 (300 mA), ADP3335 (500 mA) and ADP3336 (500 mA) anyCAP LDOs
use a patented "paddle-under-lead" package design to ensure the best thermal
performance in an MSOP-8 footprint. This package uses an electrically isolated die attach
that allows all pins to contribute to heat conduction. This technique reduces the thermal
resistance to 110°C/W on a 4-layer board as compared to >160°C/W for a standard
MSOP-8 leadframe. Figure 7.46 shows the standard physical construction of the MSOP-8
(left) and the thermally enhanced paddle-under-lead leadframe (right).
110°C/W vs. >160°C/W for Standard MSOP-8
Figure 7.46: Thermally Enhanced "Paddle-Under-Lead" 8-Lead MSOP Package
The ADP3338 (1 A) and ADP3339 (1.5 A) anyCAP LDOs are packaged in a thermally
enhanced SOT-223 package as shown in Figure 7.47. The SOT-223's thermal resistance,
θJA, is determined by the sum of the junction-to-case and the case-to-ambient thermal
resistances. The junction-to-case thermal resistance, θJC, is determined by the package
design and specified at 26.8°C/W. However, the case-to-ambient thermal resistance is
determined by the printed circuit board design. As shown in Figure 7.47A-C, the amount
of copper the ADP3338/ADP3339 is mounted to affects the thermal performance. When
mounted to 2 oz. copper with just the minimal pads (Figure 7.47A), the θJA is 126.6°C/W.
7.50
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
By adding a small copper pad under the ADP3338 (Figure 7.47B), the θJA is reduced to
102.9°C/W. Increasing the copper pad to 1 square inch (Figure 7.47C), reduces the θJA
even further to 52.8°C/W. Note that both pin 2 and pin 4 (tab) are the LDO output and
are internally connected.
4
≈ 7 mm
1
2
3
ADDED COPPER
≈1 square inch
≈ 6.5 mm
(A)
θJA = 126.6°C/W
(B)
θJA = 102.9°C/W
(C)
θJA = 52.8°C/W
Figure 7.47: Reducing SOT-223 Package θJA
LDO Regulator Controllers
To complement the anyCAP series of standalone LDO regulators, there is also the LDO
regulator controller. The regulator controller IC picks up where the standalone regulator
IC is no longer useful in either load current or power dissipation terms, and uses an
external PMOS FET for the pass device. The ADP3310 is a basic LDO regulator
controller device, designed for fixed output voltage applications while operating from
sources over a range of 3.8 to 15 V and a temperature range of –40 to +85°C. The actual
ADP3310 device ordered would be specified as ADP3310AR-YY, where the "YY" is a
voltage designator suffix such as 2.8, 3, 3.3, or 5, for those respective voltages. The "AR"
portion of the part number designates the SO-8 Thermal Coastline 8-lead package. A
summary of the main features of the ADP3310 device is listed in Figure 7.48.
7.51
ANALOG-DIGITAL CONVERSION
Controller drives external PMOS power FETs
User FET choice determines IL and VMIN performance
Small, 2 chip regulator solution handles up to 10A
Advantages compared to integrated solutions
High accuracy (1.5%) fixed voltages; 2.8, 3, 3.3, or 5V
User flexibility (selection of FET for performance)
Small footprint with anyCAPTM controller and SMD FET
Kelvin output sensing possible
Integral, low-loss current limit sensing for protection
Figure 7.48: anyCAP® ADP3310 LDO Regulator Controller Features
Regulator Controller Differences
An obvious basic difference of the regulator controller versus a stand alone regulator is
the removal of the pass device from the regulator chip. This design step has both
advantages and disadvantages. A positive is that the external PMOS pass device can be
chosen for the exact size, package, current rating and power handling which is most
useful to the application. This approach allows the same basic controller IC to be useful
for currents of several hundred mA to more than 10 A, simply by choice of the FET.
Also, since the regulator controller IC's Iground of 800 µA results is very little power
dissipation, its thermal drift will be enhanced. On the downside, there are two packages
now used to make up the regulator function. And, current limiting (which can be made
completely integral to a standalone IC LDO regulator) is now a function which must be
split between the regulator controller IC and an external sense resistor. This step also
increases the dropout voltage of the LDO regulator controller somewhat, by about
50 mV.
A functional diagram of the ADP3310 regulator controller is shown in Figure 7.49. The
basic error amplifier, reference and scaling divider of this circuit are similar to the
standalone anyCAP regulator, and will not be described in detail. The regulator controller
version does share the same cap load immunity of the standalone versions, and also has a
shutdown function, similarly controlled by the EN (enable) pin.
The main differences in the regulator controller IC architecture is the buffered output of
the amplifier, which is brought out on the GATE pin, to drive the external PMOS FET. In
addition, the current limit sense amplifier has a built in 50-mV threshold voltage, and is
designed to compare the voltage between the VIN and IS pins. When this voltage exceeds
50 mV, the current limit sense amplifier takes over control of the loop, by shutting down
the error amplifier and limiting output current to the preset level.
7.52
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
VIN
+
50mV
BIAS
EN
–
SOURCE
(IS)
–
VREF
GATE
+
VOUT
GND
Figure 7.49: Functional Block Diagram of anyCAP Series
LDO Regulator Controller
A Basic 5-V/1-A LDO Regulator Controller
An LDO regulator controller is easy to use, since a PMOS FET, a resistor and two
relatively small capacitors (one at the input, one at the output) is all that is needed to form
an LDO regulator. The general configuration is shown by Figure 7.50, an LDO suitable
as a 5-V/1-A regulator operating from a VIN of 6 V, using the ADP3310-5 controller IC.
VIN = 6V MIN
RS
50mΩ
NDP6020P OR NDB6020P
(FAIRCHILD)
VOUT = 5V @ 1A
+
IS
VIN
+
CIN
1µF
CL
10µF
GATE
VOUT
ADP3310-5
EN
GND
Figure 7.50: A Basic ADP3310 PMOS FET 1-A LDO Regulator Controller Circuit
7.53
ANALOG-DIGITAL CONVERSION
This regulator is stable with virtually any good quality output capacitor used for CL (as is
true with the other anyCAP devices). The actual CL value required and its associated
ESR depends on the gm and capacitance of the external PMOS device. In general, a
10-µF capacitor at the output is sufficient to ensure stability for load currents up to 10 A.
Larger capacitors can also be used, if high output surge currents are present. In such
cases, low ESR capacitors such as OS-CON electrolytics are preferred, because they offer
lowest ripple on the output. For less demanding requirements, a standard tantalum or
aluminum electrolytic can be adequate. When an aluminum electrolytic is used, it should
be qualified for adequate performance over temperature. The input capacitor, CIN, is only
necessary when the regulator is several inches or more distant from the raw dc filter
capacitor. However, since it is a small type, it is usually prudent to use it in most
instances, located close to the VIN pin of the regulator.
Selecting the Pass Device
The type and size of the pass transistor are determined by a set of requirements for
threshold voltage, input-output voltage differential, load current, power dissipation, and
thermal resistance. An actual PMOS pass device selected must satisfy all of these
electrical requirements, plus physical and thermal parameters. There are a number of
manufacturers offering suitable devices in packages ranging from SO-8 up through
TO-220 in size.
To ensure that the maximum available drive from the controller will adequately drive the
FET under worst case conditions of temperature range and manufacturing tolerances, the
maximum drive from the controller, VGS(DRIVE), to the pass device must be determined.
This voltage is calculated as follows:
(
)
VGS( DRIVE) = VIN − VBE − I L( MAX ) (R S ) ,
Eq. 7.22
where VIN is the minimum input voltage, IL(MAX) is the maximum load current, RS the
sense resistor, and VBE is a voltage internal to the ADP3310 (~ 0.5 V @ high temp, 0.9 V
cold, and 0.7 V at room temp). Note that since IL(MAX)× RS will be no more than 75 mV,
and VBE at cold temperature ≅0.9 V, this equation can be further simplified to:
VGS( DRIVE) ≅ VIN − 1 V .
Eq. 7.23
In the Figure 7.50 example, VIN = 6 V and VOUT = 5 V, so VGS(DRIVE) is 6 – 1 = 5 V.
It should be noted that the above two equations apply to FET drive voltages which are
less than the typical gate-to-source clamp voltage of 8 V (built into the ADP3310, for the
purposes of FET protection).
An overall goal of the design is to then select an FET which will have an RDS(ON)
sufficiently low so that the resulting dropout voltage will be less than VIN – VOUT, which
in this case is 1 V. For the NDP6020P used in Fig. 7.50 (see Reference 10), this device
achieves an RDS(ON) of 70-milliohms (max) with a VGS of 2.7 V, a voltage drive
appreciably less than the ADP3310's V GS(DRIVE) of 5V. The dropout voltage VMIN of this
7.54
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
regulator configuration is the sum of two series voltage drops, the FET's drop plus the
drop across RS, or:
(
)
VMIN = I L( MAX ) R DS(ON ) + R S .
Eq. 7.24
In the design here, the two resistances are roughly comparable to one another, so the net
VMIN will be 1 A × (50 + 70 milliohms) = 120 mV.
For a design safety margin, use a FET with a rated VGS at the required RDS, with a
substantial headroom between the applicable ADP3310 VGS(DRIVE) and the applicable VGS
rating for the FET. In the case here, there is ample margin, with 5 V of drive and a VGS of
2.7 V. It should be borne in mind that the FET's VGS and RDS(ON) will change over
temperature, but for the NDP6020P device even these variations and a VGS of 4.5 V are
still possible with the circuit as shown. With a rated minimum dc input of 6 V, this means
that the design is conservative with 5-V output. In practice, the circuit will typically
operate with input voltage minimums on the order of VOUT plus the dropout of 120 mV,
or ~ 5.12 V. Since the NDP6020P is also a fairly low threshold device, it will typically
operate at lower output voltages, down to about 3 V.
In the event the output is shorted to ground, the pass device chosen must be able to
conduct the maximum short circuit current, both instantaneously and longer term.
Thermal Design
The maximum allowable thermal resistance between the FET junction and the highest
expected ambient temperature must be taken into account, to determine the type of FET
package and heat sink used (if any).
Whenever possible to do so reliably, the FET pass device can be directly mounted to the
PCB, and the available PCB copper lands used as an effective heat sink. This heat sink
philosophy will likely be adequate when the power to be dissipated in the FET is on the
order of 1-2 W or less. Note that the very nature of an LDO helps this type of design
immensely, as the lower voltage drop across the pass device reduces the power to be
dissipated. Under normal conditions for example, Q1 of Figure 7.52 dissipates less than
1 W at a current of 1 A, since the drop across the FET is less than 1 V.
To use PCB lands as effective heat sinks with SO-8 and other SMD packages, the pass
device manufacturer's recommendations for the lowest θJA mounting should be followed
(see References 11 and 12). In general these suggestions will likely parallel the 5 rules
noted above, under "LDO regulator thermal considerations" for SO-8 and SOT-23
packaged anyCAP LDOs. For lowest possible thermal resistance, also connect multiple
FET pins together, as follows:
Electrically connect multiple FET source and drain pins in parallel, as well as to the
corresponding RS and VOUT large area PCB lands.
Using 2-oz. copper PCB material and one square inch of copper PCB land area as a
heatsink, it is possible to achieve a net thermal resistance, θJA, for mounted SO-8 devices
7.55
ANALOG-DIGITAL CONVERSION
on the order of 60°C/W or less. Such data is available for SO-8 power FETs (see
Reference 11). There are also a variety of larger packages with lower thermal resistance
than the SO-8, but still useful with surface mount techniques. Examples are the DPAK
and D2PAK, etc.
For higher power dissipation applications, corresponding to thermal resistance of 50°C/W
or less, a bolt-on external heat sink is required to satisfy the θJA requirement. Compatible
package examples would be the TO-220 family, which is used with the NDP6020P
example of Fig. 7.52.
Calculating thermal resistance for VIN = 6.7 V, VOUT = 5 V, and IL = 1 A:
θJA =
TJ − TA( MAX )
VDS( MAX ) ⋅ I L
,
Eq. 7.25
where TJ is the pass device junction temperature limit, TA(MAX ) is the maximum ambient
temperature, VDS(MAX ) is the maximum pass device drain-source voltage, and IL(MAX ) is
the maximum load current.
Inserting some example numbers of 125°C as a max. junction temp for the NDP6020P, a
75°C expected ambient, and the VDS(MAX ) and IL(MAX) figures of 1.7 V and 1 A, the
required θJA works out to be (125°C – 75°C) /1.7 = 29.4°C/W. This can be met with a
very simple heat sink, which is derived as follows.
The NDP6020P in the TO-220 package has a junction-case thermal resistance, θJC, of
2°C/W. The required external heatsink's thermal resistance, θCA, is determined as
follows:
θCA = θJA − θJC ,
Eq. 7.26
where θCA is the required heat sink case-to-ambient thermal resistance, θJA is the
calculated overall junction-to-ambient thermal resistance, and θJC is the pass device
junction-to-case thermal resistance, which in this case is 2°C/W typical for TO-220
devices, and NDP6020P.
θCA = 29.4°C/W – 2°C/W = 27.4°C/W.
Eq. 7.27
For a safety margin, select a heatsink with a θCA less than the results of this calculation.
For example, the Aavid TO-220 style clip on heat sink # 576802 has a θCA of 18.8°C/W,
and in fact many others have performance of 25°C/W or less. As an alternative, the
NDB6020P D2PAK FET pass device could be used in this same design, with an SMD
style heat sink such as the Aavid 573300 series used in conjunction with an internal PCB
heat spreader.
Note that many LDO applications like the above will calculate out with very modest heat
sink requirements. This is fine, as long as the output never gets shorted! With a shorted
output, the current goes to the limit level (as much as 1.5 A in this case), while the
7.56
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
voltage across the pass device goes to VIN (which could also be at a maximum). In this
case, the new pass device dissipation for short circuit conditions becomes 1.5 A × 6.7 V,
or 10 W. Supporting this level of power continuously will require the entire heat sink
situation to be re-evaluated, as what was adequate for 1.7 W will simply not be adequate
for 10 W. In fact, the required heat sink θCA is about 3°C/W to support the 10 W safely on
a continuous basis, which requires a much larger heat sink.
Sensing Resistors for LDO Controllers
Current limiting in the ADP3310 controller is achieved by choosing an appropriate
external current sense resistor, RS, which is connected between the controller's VIN and IS
(source) pins. An internally derived 50-mV current limit threshold voltage appears
between these pins, to establish a comparison threshold for current limiting. This 50 mV
determines the threshold where current limiting begins. For a continuous current limiting,
a foldback mode is established, with dissipation controlled by reducing the gate drive.
The net effect is that the ultimate current limit level is a factor of 2/3 of maximum. The
foldback limiting reduces the power dissipated in the pass transistor substantially.
To choose a sense resistor for a maximum output current IL, RS is calculated as follows:
RS =
0.05
.
KF ⋅ IL
Eq. 7.28
In this expression, the nominal 50-mV current limit threshold voltage appears in the
numerator. In the denominator appears a scaling factor KF, which can be either 1.0 or 1.5,
plus the maximum load current, IL. For example, if a scaling factor of 1.0 is to be used for
a 1-A IL, the RS calculation is straightforward, and 50 milliohms is the correct RS value.
However, to account for uncertainties in the threshold voltage and to provide a more
conservative output current margin, a scaling factor of KF = 1.5 can alternately be used.
When this approach is used, the same 1-A IL load conditions will result in a 33-milliohm
RS value. In essence, the use of the 1.5 scaling factor takes into account the foldback
scheme's reduction in output current, allowing higher current in the limit mode.
The simplest and least expensive sense resistor for high current applications such as
Figure 7.50 is a copper PCB trace controlled in both thickness and width. Both the
temperature dependence of copper and the relative size of the trace must be taken into
account in the resistor design. The temperature coefficient of resistivity for copper has a
positive temperature coefficient of +0.39%/°C. This natural copper TC, in conjunction
with the controller's PTAT based current limit threshold voltage, can provide for a
current limit characteristic which is simple and effective over temperature.
The table of Figure 7.51 provides resistance data for designing PCB copper traces with
various PCB copper thickness (or weight), in ounces of copper per square foot area. To
use this information, note that the center column contains a resistance coefficient, which
is the conductor resistance in milliohms/inch, divided by the trace width, W. For
example, the first entry, for 1/2 ounce copper is 0.983 milliohms/inch/W. So, for a
reference trace width of 0.1", the resistance would be 9.83 milliohms/inch. Since these
7.57
ANALOG-DIGITAL CONVERSION
are all linear relationships, everything scales for wider/skinnier traces, or for differing
copper weights. As an example, to design a 50 milliohm RS for the circuit of Fig. 7.50
using 1/2 ounce copper, a 2.54" length of a 0.05" wide PCB trace could be used.
Copper Thickness
Resistance Coefficient,
milliohms / inch/ W
(trace width W in
inches)
milliohms / inch
0.983 / W
9.83
2
0.491 / W
4.91
2
0.246 / W
2.46
2
0.163 / W
1.63
1/2 oz / ft
1 oz / ft
2 oz / ft
3 oz / ft
Reference 0.1
inch wide trace,
2
Figure 7.51: Printed Circuit Copper Resistance Design
for LDO Controllers
To minimize current limit sense voltage errors, the two connections to RS should be made
four-terminal style, as is noted in Figure 7.50 (again). It is not absolutely necessary to
actually use four-terminal style resistors, except for the highest current levels. However,
as a minimum, the heavy currents flowing in the source circuit of the pass device should
not be allowed to flow in the ADP3310 sense pin traces. To minimize such errors, the VIN
connection trace to the ADP3310 should connect close to the body of RS (or the resistor's
input sense terminal), and the IS connection trace should also connect close to the resistor
body (or the resistor's output sense terminal). Four-terminal wiring is increasingly
important for output currents of 1 A or more.
Alternately, an appropriate selected sense resistor such as surface mount sense devices
available from resistor vendors can be used (see Reference 13). Sense resistor RS may not
be needed in all applications, if a current limiting function is provided by the circuit
feeding the regulator. For circuits that don't require current limiting, the IS and VIN pins
of the ADP3310 must be tied together.
PCB-Layout Issues
For best voltage regulation, place the load as close as possible to the controller device's
VOUT and GND pins. Where the best regulation is required, the VOUT trace from the
ADP3310 and the pass device's drain connection should connect to the positive load
terminal via separate traces. This step (Kelvin sensing) will keep the heavy load currents
in the pass device's drain out of the feedback sensing path, and thus maximize output
accuracy. Similarly, the unregulated input common should connect to the common side of
the load via a separate trace from the ADP3310 GND pin.
7.58
DATA CONVERTER SUPPORT CIRCUITS
7.2 LOW DROPOUT LINEAR REGULATORS
A 2.8-V / 8-A LDO Regulator Controller
With seemingly minor changes to the basic 1-A LDO circuit used in Fig. 7.50, an 8-A
LDO regulator controller can be configured, as shown in Figure 7.52. This circuit uses an
ADP3310-2.8, to produce a 2.8-V output. The sense resistor is dropped to 5 milliohms,
which supports currents of up to 10 A (or about 6.7 A, with current limiting active). Fourterminal wiring should be used with the sense resistor to minimize errors.
The most significant change over the more generic schematic of Fig. 7.50 is the use of
multiple, low ESR input and output bypass capacitors. At the output, C2 is a bank of
4 × 220-µF OS-CON type capacitors, in parallel with 2 × 10-µF MLCC chip type
capacitors. These are located right at the load point with minimum inductance wiring,
plus separate wiring back to the VOUT pin of the ADP3310 and the drain of the pass
device. This wiring will maximize the dc output accuracy, while the multiple capacitors
will minimize the transient errors at the point-of-load. In addition, multiple bypasses on
the regulator input in the form of C1 minimizes the transient errors at the regulator's VIN
pin.
RS
5mΩ
VIN
C1
Q1, NDP6020P OR NDB6020P
(FAIRCHILD)
+
+
2 × 220µF
OS-CON
+
2×10µF MLC
2.8V @ 8A
IS
GATE
VIN
VOUT
C2
4 × 220µF
OS-CON
+
2×10µF MLC
ADP3310-2.8
EN
GND
Figure 7.52: A 2.8 V/8 A LDO Regulator Controller
Heat sink requirements for the pass device in this application will be governed by the
loading and input voltage, and should be calculated by the procedures discussed above.
7.59
ANALOG-DIGITAL CONVERSION
REFERENCES:
7.2 LOW DROPOUT LINEAR REGULATORS
1.
Bob Widlar, "New Developments in IC Voltage Regulators," IEEE Journal of Solid State Circuits,
Vol. SC-6, February, 1971.
2.
Robert C. Dobkin, "3-Terminal Regulator is Adjustable," National Semiconductor AN-181, March,
1977.
3.
Paul Brokaw, "A Simple Three-Terminal IC Bandgap Voltage Reference," IEEE Journal of Solid
State Circuits, Vol. SC-9, December, 1974.
4.
Frank Goodenough, "Linear Regulator Cuts Dropout Voltage," Electronic Design, April 16, 1987.
5.
Chester Simpson, "LDO Regulators Require Proper Compensation," Electronic Design, November 4,
1996.
6.
Frank Goodenough, "Vertical-PNP-Based Monolithic LDO Regulator Sports Advanced Features,"
Electronic Design, May 13, 1996.
7.
Frank Goodenough, "Low Dropout Regulators Get Application Specific," Electronic Design, May 13,
1996.
8.
Jim Solomon, "The Monolithic Op Amp: A Tutorial Study." IEEE Journal of Solid State Circuits,
Vol. SC-9, No.6, December 1974.
9.
Richard J. Reay, Gregory T.A. Kovacs, "An Unconditionally Stable Two-Stage CMOS Amplifier,"
IEEE Journal of Solid State Circuits, Vol. SC-30, No.5, May 1995.
10. NDP6020P / NDB6020P P-Channel Logic Level Enhancement Mode Field Effect Transistor,
Fairchild Semiconductor data sheet, September 1997, http://www.fairchildsemi.com.
11. Alan Li, et all, "Maximum Power Enhancement Techniques for SO-8 Power MOSFETs," Fairchild
Semiconductor application note AN1029, April 1996, http://www.fairchildsemi.com.
12. Rob Blattner, Wharton McDaniel, "Thermal Management in On-Board DC-to-DC Power Conversion,"
Temic application note, http://www.temic.com.
13. "S" series surface mount current sensing resistors, KRL/Bantry Components, 160 Bouchard Street,
Manchester, NH, 03103-3399, (603) 668-3210.
7.60
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
SECTION 7.3: ANALOG SWITCHES AND
MULTIPLEXERS
Walt Kester
Introduction
Solid-state analog switches and multiplexers have become an essential component in the
design of electronic systems which require the ability to control and select a specified
transmission path for an analog signal. These devices are used in a wide variety of
applications including multi-channel data acquisition systems, process control,
instrumentation, video systems, etc.
One of the first commercial analog multiplexers is shown in Figure 7.53, the MOSES-8
from the Pastoriza Division of Analog Devices in 1969. This PC board multiplexer
consisted of 8 MOSFET switches and 8 switch drivers. The part had a switching time of
100 ns, on-resistance of 500 Ω. Selling price in 1969 was $320. For ±5-V inputs, the
multiplexer operated on ±15-V supplies, but for ±10-V inputs, it required not only the
+15 V but also a –28-V supply. Today, the ADG725/ADG726/ADG731/ADG732 family
offers a 32-channel multiplexer with 4-Ω on-resistance, 20-µA quiescent current, and
packaged in 7 mm × 7 mm chip scale (CSP) or thin plastic quad flatpack (TQFP). The
price is less than $5.
TODAY:
ADG725,ADG726,
ADG731, ADG732:
8 Channels
Switching time: 100 ns
On Resistance: 500Ω
Off Resistance: > 100MΩ
$320
32 Channels
Switching Time: 30ns
On Resistance: 4Ω
7 mm2 CSP or TQFP
< $5
Figure 7.53: "MOSES-8" MOSFET Analog Multiplexer
Analog Devices' Pastoriza Division, 1969
7.61
ANALOG-DIGITAL CONVERSION
With the development of CMOS processes (yielding good PMOS and NMOS transistors
on the same substrate), switches and multiplexers rapidly gravitated to integrated circuit
form in the mid-1970s, with product introductions such as the Analog Devices' popular
AD7500-series (introduced in 1973). A dielectrically-isolated family of these parts
introduced in 1976 allowed input overvoltages of ±25 V (beyond the supply rails) and
was insensitive to latch-up.
These early CMOS switches and multiplexers were typically designed to handle signal
levels up to ±10 V while operating on ±15-V supplies. In 1979, Analog Devices
introduced the popular ADG200-series of switches and multiplexers, and in 1988 the
ADG201-series was introduced which was fabricated on a proprietary linear-compatible
CMOS process (LC2MOS). These devices allowed input signals to ±15 V when operating
on ±15-V supplies.
A large number of switches and multiplexers were introduced in the 1980s and 1990s,
with the trend toward lower on-resistance, faster switching, lower supply voltages, lower
cost, lower power, and smaller surface-mount packages.
Today, analog switches and multiplexers are available in a wide variety of configurations,
options, etc., to suit nearly all applications. On-resistances less than 0.5 Ω, picoampere
leakage currents, signal bandwidths greater than 1 GHz, and single 1.8-V supply
operation are now possible with modern CMOS technology.
Although CMOS is by far the most popular IC process today for switches and
multiplexers, bipolar processes (with JFETs) and complementary bipolar processes (also
with JFET capability) are often used for special applications such as video switching and
multiplexing where the high performance characteristics required are not attainable with
CMOS. Traditional CMOS switches and multiplexers suffer from several disadvantages
at video frequencies. Their switching time is generally not fast enough, and they require
external buffering in order to drive typical video loads. In addition, the small variation of
the CMOS switch on-resistance with signal level (RON modulation) can introduce
unwanted distortion in differential gain and phase. Multiplexers based on complementary
bipolar technology offer better solutions at video frequencies—with obvious power and
cost increases above CMOS devices.
CMOS Switch Basics
The ideal analog switch has no on-resistance, infinite off-impedance and zero time delay,
and can handle large signal and common-mode voltages. Real CMOS analog switches
meet none of these criteria, but if we understand the limitations of analog switches, most
of these limitations can be overcome.
CMOS switches have an excellent combination of attributes. In its most basic form, the
MOSFET transistor is a voltage-controlled resistor. In the "on" state, its resistance can be
less than 1 Ω, while in the "off" state, the resistance increases to several hundreds of
megohms, with picoampere leakage currents. CMOS technology is compatible with logic
circuitry and can be densely packed in an IC. Its fast switching characteristics are well
controlled with minimum circuit parasitics.
7.62
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
MOSFET transistors are bilateral. That is, they can switch positive and negative voltages
and conduct positive and negative currents with equal ease. A MOSFET transistor has a
voltage controlled resistance which varies nonlinearly with signal voltage as shown in
Figure 7.54.
PMOS
NMOS
ALTERNATE SYMBOLS
Figure 7.54: MOSFET Switch ON-Resistance Versus Signal Voltage
The complementary-MOS process (CMOS) yields good P-channel and N-channel
MOSFETs. Connecting the PMOS and NMOS devices in parallel forms the basic
bilateral CMOS switch of Figure 7.55. This combination reduces the on-resistance, and
also produces a resistance which varies much less with signal voltage.
SWITCH
SWITCH
DRIVER
Figure 7.55: Basic CMOS Switch Uses Complementary Pair to
Minimize RON Variation due to Signal Swings
7.63
ANALOG-DIGITAL CONVERSION
Figure 7.56 shows the on-resistance changing with channel voltage for both N-type and
P-type devices. This nonlinear resistance can causes errors in dc accuracy as well as ac
distortion. The bilateral CMOS switch solves this problem. On-resistance is minimized,
and linearity is also improved. The bottom curve of Figure 7.56 shows the improved
flatness of the on-resistance characteristic of the switch.
COMBINED TRANSFER
FUNCTION
Figure 7.56: CMOS Switch ON-Resistance Versus Signal Voltage
The ADG8xx-series of CMOS switches are specifically designed for less than 0.5-Ω onresistance and are fabricated on a sub-micron process. These devices can carry currents
up to 400 mA, operate on a single 1.8-V to 5.5-V supply, and are rated over an extended
temperature range of –40°C to +125°C. On-resistance over temperature and input signal
level is shown in Figure 7.57.
INPUT SIGNAL LEVEL - V
Figure 7.57: ON-Resistance Versus Input Signal for
ADG801/ADG802 CMOS Switch, VDD = +5 V
7.64
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Error Sources in the CMOS Switch
It is important to understand the error sources in an analog switch. Many affect ac and dc
performance, while others only affect ac. Figure 7.58 shows the equivalent circuit of two
adjacent CMOS switches. The model includes leakage currents and junction
capacitances.
Figure 7.58: Equivalent Circuit of Two Adjacent CMOS Switches
dc errors associated with a single CMOS switch in the on state are shown in Figure 7.59.
When the switch is on, dc performance is affected mainly by the switch on-resistance
(RON) and leakage current (ILKG). A resistive attenuator is created by the RG-RON-RLOAD
combination which produces a gain error. The leakage current, ILKG, flows through the
equivalent resistance of RLOAD in parallel with the sum of RG and RON. Not only can RON
cause gain errors—which can be calibrated using a system gain trim—but its variation
with applied signal voltage (RON modulation) can introduce distortion—for which there is
no calibration. Low resistance circuits are more subject to errors due to RON, while high
resistance circuits are affected by leakage currents. Figure 7.59 also gives equations that
show how these parameters affect dc performance.
When the switch is OFF, leakage current can introduce errors as shown in Figure 7.60.
The leakage current flowing through the load resistance develops a corresponding voltage
error at the output.
7.65
ANALOG-DIGITAL CONVERSION
Figure 7.59: Factors Affecting DC Performance for
ON Switch Condition: RON, RLOAD, and ILKG
Leakage current creates error voltage at VOUT equal to:
VOUT = ILKG × RLOAD
Figure 7.60: Factors Affecting DC Performance for
OFF Switch Condition: ILKG and RLOAD
Figure 7.61 illustrates the parasitic components that affect the ac performance of CMOS
switches. Additional external capacitances will further degrade performance. These
capacitances affect feedthrough, crosstalk and system bandwidth. CDS (drain-to-source
capacitance), CD (drain-to-ground capacitance), and CLOAD all work in conjunction with
RON and RLOAD to form the overall transfer function.
7.66
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Figure 7.61: Dynamic Performance Considerations:
Transfer Accuracy Versus Frequency
In the equivalent circuit, CDS creates a frequency zero in the numerator of the transfer
function A(s). This zero usually occurs at high frequencies because the switch onresistance is small. The bandwidth is also a function of the switch output capacitance in
combination with CDS and the load capacitance. This frequency pole appears in the
denominator of the equation.
The composite frequency domain transfer function may be re-written as shown in Figure
7.62 which shows the overall Bode plot for the switch in the on state. In most cases, the
pole breakpoint frequency occurs first because of the dominant effect of the output
capacitance CD. Thus, to maximize bandwidth, a switch should have low input and output
capacitance and low on-resistance.
Figure 7.62: Bode Plot of CMOS Switch Transfer
Function in the ON State
7.67
ANALOG-DIGITAL CONVERSION
The series-pass capacitance, CDS, not only creates a zero in the response in the ON-state,
it degrades the feedthrough performance of the switch during its OFF state. When the
switch is off, CDS couples the input signal to the output load as shown in Figure 7.63.
OFF Isolation is Affected
by External R and C Load
Figure 7.63: Dynamic Performance Considerations: Off Isolation
Large values of CDS will produce large values of feedthrough, proportional to the input
frequency. Figure 7.64 illustrates the drop in OFF-isolation as a function of frequency.
The simplest way to maximize the OFF-isolation is to choose a switch that has as small a
CDS as possible.
Figure 7.64: Off Isolation Versus Frequency
Figure 7.65 shows typical CMOS analog switch OFF-isolation as a function of frequency
for the ADG708 8-channel multiplexer. From dc to several kilohertz, the multiplexer has
nearly 90-dB isolation. As the frequency increases, an increasing amount of signal
reaches the output. However, even at 10 MHz, the switch shown still has nearly 60-dB of
isolation.
7.68
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Figure 7.65: OFF-Isolation Versus Frequency for
ADG708 8-Channel Multiplexer
Another ac parameter that affects system performance is the charge injection that takes
place during switching. Figure 7.66 shows the equivalent circuit of the charge injection
mechanism.
VDD
VSS
Step waveforms of ± (VDD – VSS) are applied to CQ,
the gate capacitance of the output switches.
Figure 7.66: Dynamic Performance Considerations:
Charge Injection Model
7.69
ANALOG-DIGITAL CONVERSION
When the switch control input is asserted, it causes the control circuit to apply a large
voltage change (from VDD to VSS, or vice versa) at the gate of the CMOS switch. This
fast change in voltage injects a charge into the switch output through the gate-drain
capacitance CQ. The amount of charge coupled depends on the magnitude of the gatedrain capacitance.
The charge injection introduces a step change in output voltage when switching as
shown in Figure 7.67. The change in output voltage, ∆VOUT, is a function of the amount
of charge injected, QINJ (which is in turn a function of the gate-drain capacitance, CQ) and
the load capacitance, CL.
0V
Figure 7.67: Effects of Charge Injection on Output
Another problem caused by switch capacitance is the retained charge when switching
channels. This charge can cause transients in the switch output, and Figure 7.68 illustrates
the phenomenon.
Assume that initially S2 is closed and S1 open. CS1 and CS2 are charged to –5 V. As S2
opens, the –5 V remains on CS1 and CS2, as S1 closes. Thus, the output of Amplifier A
sees a –5-V transient. The output will not stabilize until Amplifier A's output fully
discharges CS1 and CS2 and settles to 0 V. The scope photo in Figure 7.69 depicts this
transient. The amplifier's transient load settling characteristics will therefore be an
important consideration when choosing the right input buffer.
7.70
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
–5V
Figure 7.68: Charge Coupling Causes Dynamic Settling Time
Transient When Multiplexing Signals
Figure 7.69: Output of Amplifier Shows Dynamic Settling Time
Transient Due to Charge Coupling
Crosstalk is related to the capacitances between two switches. This is modeled as the CSS
capacitance shown in Figure 7.70.
7.71
ANALOG-DIGITAL CONVERSION
Figure 7.70: Channel-to-Channel Crosstalk Equivalent
Circuit for Adjacent Switches
Figure 7.71shows typical crosstalk performance of the ADG708 8-channel CMOS
multiplexer.
Figure 7.71: Crosstalk Versus Frequency for ADG708 8-Channel Multiplexer
Finally, the switch itself has a settling time that must be considered. Figure 7.72 shows
the dynamic transfer function. The settling time can be calculated, because the response
is a function of the switch and circuit resistances and capacitances. One can assume that
this is a single-pole system and calculate the number of time constants required to settle
to the desired system accuracy as shown in Figure 7.73.
7.72
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Settling time is the time required for the switch output
to settle within a given error band of the final value.
Figure 7.72: Multiplexer Settling Time
RESOLUTION,
# OF BITS
LSB (%FS)
# OF TIME
CONSTANTS
6
1.563
4.16
8
0.391
5.55
10
0.0977
6.93
12
0.0244
8.32
14
0.0061
9.70
16
0.00153
11.09
18
0.00038
12.48
20
0.000095
13.86
22
0.000024
15.25
Figure 7.73: Number of Time Constants Required to Settle to
1 LSB Accuracy for a Single-Pole System
Applying the Analog Switch
Switching time is an important consideration in applying analog switches, but switching
time should not be confused with settling time. ON and OFF times are simply a measure
of the propagation delay from the control input to the toggling of the switch, and are
largely caused by time delays in the drive and level-shift circuits (see Figure 7.74). The
tON and tOFF values are generally measured from the 50% point of the control input
leading edge to the 90% point of the output signal level.
7.73
ANALOG-DIGITAL CONVERSION
tON, tOFF
tON and tOFF should not be confused with settling time.
tON and tOFF are simply a measure of the propagation delay
from control input to operation of the analog switch. It is caused
by time delays in the drive / level-shifter logic circuitry.
tON and tOFF are measured from the 50% point of the control
input to the 90% point of the output signal level.
Figure 7.74: Applying the Analog Switch: Dynamic
Performance Considerations
We will next consider the issues involved in buffering a CMOS switch or multiplexer
output using an op amp. When a CMOS multiplexer switches inputs to an inverting
summing amplifier, it should be noted that the on-resistance, and its nonlinear change as
a function of input voltage, will cause gain and distortion errors as shown in Figure 7.75.
If the resistors are large, the switch leakage current may introduce error. Small resistors
minimize leakage current error but increase the error due to the finite value of RON.
10kΩ
10kΩ
∆VSWITCH = ±10V
∆RON caused by ∆VIN , degrades linearity of VOUT relative to VIN.
∆RON causes overall gain error in VOUT relative to VIN .
Figure 7.75: Applying the Analog Switch: Unity
Gain Inverter with Switched Input
7.74
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
To minimize the effect of RON change due to the change in input voltage, it is advisable
to put the multiplexing switches at the op amp summing junction as shown in Figure
7.76. This ensures the switches are only modulated with about ±100 mV rather than the
full ±10 V—but a separate resistor is required for each input leg.
10kΩ
10kΩ
10kΩ
10kΩ
10kΩ
∆VSWITCH = ±100mV
Switch drives a virtual ground.
Switch sees only ±100mV, not ±10V, minimizes ∆RON .
Figure 7.76: Applying the Analog Switch:
Minimizing the Influence of ∆RON
It is important to know how much parasitic capacitance has been added to the summing
junction as a result of adding a multiplexer, because any capacitance added to that node
introduces phase shift to the amplifier closed loop response. If the capacitance is too
large, the amplifier may become unstable and oscillate. A small capacitance, C1, across
the feedback resistor may be required to stabilize the circuit.
The finite value of RON can be a significant error source in the circuit shown in Figure
7.77. The gain-setting resistors should be at least 1,000 times larger than the switch onresistance to guarantee 0.1% gain accuracy. Higher values yield greater accuracy but
lower bandwidth and greater sensitivity to leakage and bias current.
A better method of compensating for RON is to place one of the switches in series with the
feedback resistor of the inverting amplifier as shown in Figure 7.78. It is a safe
assumption that the multiple switches, fabricated on a single chip, are well-matched in
absolute characteristics and tracking over temperature. Therefore, the amplifier is closedloop gain stable at unity gain, since the total feedforward and feedback resistors are
matched.
7.75
ANALOG-DIGITAL CONVERSION
1MΩ
1MΩ
∆RON is small compared to 1MΩ switch load.
Effect on transfer accuracy is minimized.
Bias current and leakage current effects are now very important.
Circuit bandwidth degrades.
Figure 7.77: Applying the Analog Switch: Minimizing Effects
of ∆RON Using Large Resistor Values
RF
10kΩ
±10V
±10V
10kΩ
10kΩ
10kΩ
±10V
Figure 7.78: Applying the Analog Switch:Using "Dummy" Switch
in Feedback to Minimize Gain Error Due to ∆RON
The best multiplexer design drives the non-inverting input of the amplifier as shown in
Figure 7.79. The high input impedance of the non-inverting input eliminates the errors
due to RON.
7.76
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Figure 7.79: Applying the Analog Switch: Minimizing the Influence of ∆RON
Using Non-Inverting Configuration
CMOS switches and multiplexers are often used with op amps to make programmable
gain amplifiers (PGAs). To understand RON's effect on their performance, consider Figure
7.80, a poor PGA design. A non-inverting op amp has 4 different gain-set resistors, each
grounded by a switch, with an RON of 100-500 Ω. Even with RON as low as 25 Ω, the gain
of 16 error would be 2.4%, worse than 8-bit accuracy! RON also changes over
temperature, and from switch-to-switch.
RF = 10kΩ
–
625Ω
1.43kΩ
3.33kΩ
VOUT
10kΩ
+
G = 16
G=8
G=4
G=2
VIN
Gain accuracy limited by switch's on-resistance RON
and RON modulation
RON typically 1 - 500Ω for CMOS or JFET switch
For RON = 25Ω, there is a 2.4% gain error for G = 16
RON drift over temperature limits accuracy
Must use very low RON switches
Figure 7.80: A Poorly Designed PGA Using CMOS Switches
7.77
ANALOG-DIGITAL CONVERSION
To attempt "fixing" this design, the resistors might be increased, but noise and offset
could then be a problem. The only way to improve accuracy with this circuit is to use
relays, with virtually no RON. Only then will the few mΩ of relay RON be a small error
vis-à-vis 625 Ω.
It is much better to use a circuit insensitive to RON! In Figure 7.81, the switch is placed in
series with the inverting input of an op amp. Since the op amp input impedance is very
large, the switch RON is now irrelevant, and gain is now determined solely by the external
resistors. Note—RON may add a small offset error if op amp bias current is high. If this is
the case, it can readily be compensated with an equivalent resistance at VIN.
VIN
+
VOUT
–
G=1
500Ω
G=2
1kΩ
1kΩ
RON is not in series with gain setting resistors
RON is small compared to input impedance
Only slight offset errors occur due to bias
current flowing through the switches
Figure 7.81: Alternate PGA Configuration Minimizes the Effects of RON
1-GHz CMOS Switches
The ADG918/ADG919 are the first switches using a CMOS process to provide high
isolation and low insertion loss up to and exceeding 1 GHz. The switches exhibit low
insertion loss (0.8 dB) and relatively high off isolation (37 dB) when transmitting a
1-GHz signal. In high frequency applications with throughput power of +18 dBm or less
at 25°C, they are a cost-effective alternative to gallium arsenide (GaAs) switches. A
block diagram of the devices are shown in Figure 7.82 along with isolation and loss
versus frequency plots given in Figure 7.83.
7.78
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
ABSORPTIVE SWITCH
REFLECTIVE SWITCH
Figure 7.82: 1-GHz CMOS 1.65-V to 2.75-V 2:1 Mux/SPDT Switches
ISOLATION (dB) VS. FREQUENCY
LOSS (dB)VS. FREQUENCY
Figure 7.83: Isolation and Frequency Response of
AD918/AD919 1-GHz Switch
The ADG918 is an absorptive switch with 50-Ω terminated shunt legs that allow
impedance matching with the application circuit, while the ADG919 is a reflective switch
designed for use where the terminations are external to the chip. Both offer low power
consumption (<1 µA), tiny packages (8-lead MSOP and 3 mm × 3 mm lead frame chip
scale package), single-pin control voltage levels that are CMOS/LVTTL compatible,
making the switches ideal for wireless applications and general-purpose RF switching.
7.79
ANALOG-DIGITAL CONVERSION
Video Switches and Multiplexers
In order to meet stringent specifications of bandwidth flatness, differential gain and
phase, and 75-Ω drive capability, high speed complementary bipolar processes are more
suitable than CMOS processes for video switches and multiplexers. Traditional CMOS
switches and multiplexers suffer from several disadvantages at video frequencies. Their
switching time (typically 50 ns or so) is not fast enough for today's applications, and they
require external buffering in order to drive typical video loads. In addition, the small
variation of the CMOS switch on-resistance with signal level (Ron modulation)
introduces unwanted distortion in differential gain and phase. Multiplexers based on
complementary bipolar technology offer a better solution at video frequencies. The
tradeoffs, of course, are higher power and cost.
Functional block diagrams of the AD8170/8174/8180/8182 bipolar video multiplexer are
shown in Figure 7.84. The AD8183/AD8185 video multiplexer is shown in Figure 7.85.
These devices offer a high degree of flexibility and are ideally suited to video
applications, with excellent differential gain and phase specifications. Switching time for
all devices in the family is 10ns to 0.1%. The AD8186/8187 are single-supply versions of
the AD8183/8185. Note that these bipolar multiplexers are not bi-directional.
2:1
MUX
2:1
MUX
4:1
MUX
DUAL
2:1
MUX
Figure 7.84: AD8170/8174/8180/8182 Bipolar Video Multiplexers
The AD8170/8174 series of muxes include an on-chip current feedback op amp output
buffer whose gain can be set externally. Off channel isolation and crosstalk are typically
greater than 80 dB at 5 MHz for the entire family.
7.80
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Figure 7.86 shows an application circuit for three AD8170 2:1 muxes, where a single
RGB monitor is switched between two RGB computer video sources.
TRIPLE
2:1
MUX
AD8183: G = +1
(AD8186)
AD8185: G = +2
(AD8187)
(AD8186/AD8187:
SINGLE +5V SUPPLY
VERSIONS)
Figure 7.85: AD8183/AD8185 Triple 2:1 Video Multiplexers
CHANNEL
SELECT
COMPUTER
R
G
B
IN0
R
IN1
IN0
G
IN1
IN0
MONITOR
B
IN1
R
G
B
COMPUTER
THREE AD8170 2:1 MUXES
(OR 1 AD8183/AD8185/AD8186/AD8187
TRIPLE 2:1 MUX)
Figure 7.86: Dual Source RGB Multiplexer Using Three 2:1 Muxes
7.81
ANALOG-DIGITAL CONVERSION
In this setup, the overall effect is that of a three-pole, double-throw switch. The three
video sources constitute the three poles, and either the upper or lower of the video
sources constitute the two switch states. Note that the circuit can be simplified by using a
single AD8183, AD8185, AD8186, or AD8187 triple dual input multiplexer.
The AD8174 or AD8184 4:1 mux is used in Figure 7.87, to allow a single high speed
ADC to digitize the RGB outputs of a scanner.
R
SCANNER
G
B
AD8174 , AD8184
IN0
IN1
IN2
4:1
MUX
ADC
A1
A0
IN3
CHANNEL SELECT
Figure 7.87: Digitizing RGB Signals with One ADC and a 4:1 Mux
The RGB video signals from the scanner are fed in sequence to the ADC, and digitized in
sequence, making efficient use of the scanner data with one ADC.
Video Crosspoint Switches
The AD8116 extends the multiplexer concepts to a fully integrated, 16×16 buffered video
crosspoint switch matrix (Figure 7.88). The 3-dB bandwidth is greater than 200 MHz,
and the 0.1-dB gain flatness extends to 60 MHz. Channel switching time is less than
30 ns to 0.1%. Channel-to-channel crosstalk is −70 dB measured at 5 MHz. Differential
gain and phase is 0.01% and 0.01° for a 150-Ω load. Total power dissipation is 900 mW
on ±5 V.
The AD8116 includes output buffers that can be put into a high impedance state for
paralleling crosspoint stages so that the off channels do not load the output bus. The
channel switching is performed via a serial digital control that can accommodate "daisy
chaining" of several devices. The AD8116 package is a 128-pin 14 mm × 14 mm LQFP.
Other members of the crosspoint switch family include the AD8108/AD9109 8 × 8
crosspoint switch; the AD8110/AD8111, 260-MHz, 16 × 8, buffered crosspoint switch;
the AD8113 audio/video 60-MHz, 16 × 16 crosspoint switch; and the AD8114/AD8115
low cost 225-MHz, 16 × 16, crosspoint switch.
7.82
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Figure 7.88: AD8116 16×16 200-MHz Buffered Video Crosspoint Switch
Digital Crosspoint Switches
The AD8152 is a 3.2-Gbps 34 × 34 asynchronous digital crosspoint switch designed for
high speed networking (see Figure 7.89). The device operates at data rates up to 3.2 Gbps
per port, making it suitable for Sonet/SDH OC-48 with Forward Error Correction (FEC).
The AD8152 has digitally programmable current mode outputs that can drive a variety of
termination schemes and impedances while maintaining the correct voltage level and
minimizing power consumption. The part operates with a supply voltage as low as
+2.5 V, with excellent input sensitivity. The control interface is compatible with LVTTL
or CMOS/TTL.
As the lowest power solution of any comparable crosspoint switch, the AD8152
dissipates less than 2 W at 2.5-V supply with all I/Os active and does not require external
heat sinks. The low jitter specification of less than 45 ps makes the AD8152 ideal for
high speed networking systems. The AD8152's fully differential signal path reduces jitter
and crosstalk while allowing the use of smaller single-ended voltage swings. It is offered
in a 256-ball SBGA package that operates over the industrial temperature range of 0°C to
+85°C.
7.83
ANALOG-DIGITAL CONVERSION
Figure 7.89: AD8152 3.2-Gbps Asynchronous
Digital Crosspoint Switch
Switch and Multiplexer Families from Analog Devices
Selecting the right switch or multiplexer for a particular application can be a difficult task
in light of the large number of devices currently offered. Selection guides from Analog
Devices can be invaluable in this process. Figure 7.90 summarizes the generic families of
CMOS switches and multiplexers, starting with the higher voltage devices and working
downward to the newer lower voltage parts. While certainly not all-inclusive, this listing
can be useful in getting an overall idea of available choices. Figure 7.91 summarizes the
bipolar switch and multiplexer families. The ADG32xx-series of Bus Switches are
discussed in more detail in Chapter 9.
7.84
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
ADG2xx, ADG4xx, ADG5xx: ±15V
ADG508F, ADG509F, ADG528F, ADG438F, ADG439F:
Fault-protected ±15V family
ADG12xx: ±15V, Low RON (2Ω)
ADG14xx: ±15V , Low CON (2pF)
ADG6xx: Single +5V (some lower) or ±5V
3Ω RON family
1pC charge injection family
125°C family
ADG7xx: Single +5V (some as low as +1.8V)
Some as low as 2.5Ω RON
Some in CSP
3-5pC charge injection
ADG8xx: Single +1.8V to +5.5V
<0.5Ω RON
ADG9xx: Single +1.65V to +2.75V, > 1 GHz RF switches
ADG3xxx: Bus Switches and Logic Level Shifters
Figure 7.90: CMOS Switches and Multiplexer Families from Analog Devices
Video switches and multiplexers:
AD8074, AD8075, AD8170, AD8174, AD8180, AD8182,
AD8184, AD8185, AD8186, AD8187
Video crosspoint switches:
AD8108, AD8109, AD8110, AD8111, AD8114, AD8115,
AD8116
Audio and Video crosspoint switch:
AD8113
Digital crosspoint switches:
AD8150, AD8151, AD8152, ADSX34
Figure 7.91: High-Speed Bipolar Switches and Multiplexers from ADI
Parasitic Latchup in CMOS Switches and Muxes
Because multiplexers are often at the front-end of a data acquisition system, their inputs
generally come from remote locations—hence, they are often subjected to overvoltage
conditions. Although this topic is treated in more detail in Chapter 9, an understanding of
the problem as it relates to CMOS devices is particularly important. Although this
discussion centers around multiplexers, it is germane to nearly all types of CMOS parts.
7.85
ANALOG-DIGITAL CONVERSION
Most CMOS analog switches are built using junction-isolated CMOS processes. A crosssectional view of a single switch cell is shown in Figure 7.92. Parasitic SCR (silicon
controlled rectifier) latchup can occur if the analog switch terminal has voltages more
positive than VDD or more negative than VSS. Even a transient situation, such as poweron with an input voltage present, can trigger a parasitic latchup. If the conduction current
is too great (several hundred milliamperes or more), it can damage the switch.
–VSS
+VDD
Figure 7.92: Cross-Section of a Junction-Isolation CMOS Switch
The parasitic SCR mechanism is shown in Figure 7.93. SCR action takes place when
either terminal of the switch (source or the drain) is either one diode drop more positive
than VDD or one diode drop more negative than VSS. In the former case, the VDD terminal
becomes the SCR gate input and provides the current to trigger SCR action. In the case
where the voltage is more negative than VSS, the VSS terminal becomes the SCR gate
input and provides the gate current. In either case, high current will flow between the
supplies. The amount of current depends on the collector resistances of the two
transistors, which can be fairly small.
+VDD
–VSS
Figure 7.93: Bipolar Transistor Equivalent Circuit for CMOS
Switch Shows Parasitic SCR Latch
7.86
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
In general, to prevent the latchup condition, the inputs to CMOS devices should never be
allowed to be more than 0.3 V above the positive supply or 0.3 V below the negative
supply. Note that this restriction also applies when the power supplies are off (VDD = VSS
= 0 V), and therefore devices can latchup if power is applied to a part when signals are
present on the inputs. Manuracturers of CMOS devices invariably place this restriction in
the data sheet table of absolute maximum ratings. In addition, the input current under
overvoltage conditions should be restricted to 5-30 mA, depending upon the particular
device.
In order to prevent this type of SCR latchup, a series diode can be inserted into the VDD
and VSS terminals as shown in Figure 7.94. The diodes block the SCR gate current.
Normally the parasitic transistors Q1 and Q2 have low beta (usually less than 10) and
require a comparatively large gate current to fire the SCR. The diodes limit the reverse
gate current so that the SCR is not triggered.
CR1
+VDD
CR2
–VSS
Diodes CR1 and CR2 block base current drive to Q1 and Q2
in the event of overvoltage at S or D.
Figure 7.94: Diode Protection Scheme for CMOS Switch
If diode protection is used, the analog voltage range of the switch will be reduced by one
VBE drop at each rail, and this can be inconvenient when using low supply voltages.
As noted, CMOS switches and multiplexers can also be protected from possible
overcurrent by inserting a series resistor to limit the current to a safe level as shown in
Figure 7.95, generally less than 5-30 mA. Because of the resitive attenuator formed by
RLOAD and RLIMIT, this method works only if the switch drives a relatively high
impedance load.
7.87
ANALOG-DIGITAL CONVERSION
+VDD
RLIMIT
CMOS SWITCH
OR MUX
OUTPUT
INPUT
RLOAD
LIMIT OVERCURRENT TO 5-30mA
–VSS
Figure 7.95: Overcurrent Protection Using External Resistor
A common method for input protection is shown in Figure 7.96 where Schottky diodes
are connected from the input terminal to each supply voltage as shown. The diodes
effectively prevent the inputs from exceeding the supply voltage by more than 0.3-0.4 V,
thereby preventing latchup conditions. In addition, if the input voltage exceeds the supply
voltage, the input current flows through the external diodes to the supplies, not the
device. Schottky diodes can easily handle 50-100 mA of transient current, therefore the
RLIMIT resistor can be quite low.
+VDD
RLIMIT
CMOS SWITCH
OR MUX
OUTPUT
INPUT
RLOAD
–VSS
Figure 7.96: Input Protection Using External Schottky Diodes
Most CMOS devices have internal ESD-protection diodes connected from the inputs to
the supply rails, making the devices less susceptible to latchup. However, the internal
diodes begin conduction at 0.6 V, and have limited current-handling capability, thus
adding the external Schottky diodes offers an added degree of protection. However, the
effects of the diode leakage and capacitance must be considered.
Note that latchup protection does not provide overcurrent protection, and vice versa. If
both fault conditions can exist in a system, then both protective diodes and resistors
should be used.
7.88
DATA CONVERTER SUPPORT CIRCUITS
7.3 ANALOG SWITCHES AND MULTIPLEXERS
Analog Devices uses trench-isolation technology to produce its LC2MOS analog
switches. The process reduces the latchup susceptibility of the device, the junction
capacitances, increases switching time and leakage current, and extends the analog
voltage range to the supply rails.
Figure 7.97 shows the cross-sectional view of the trench-isolated CMOS structure. The
buried oxide layer and the side walls completely isolate the substrate from each transistor
junction. Therefore, no reverse-biased PN junction is formed. Consequently the
bandwidth-reducing capacitances and the possibility of SCR latchup are greatly reduced.
Figure 7.97: Trench-Isolation LC2MOS Structure
The ADG508F, ADG509F, ADG528F, ADG438F, and ADG439F are ±15-V trenchisolated LC2MOS multiplexers which offer "fault protection" for input and output
overvoltages between –40 V and + 55 V. These devices use a series structure of three
MOSFETS in the signal path: an N-channel, followed by a P-channel, followed by an Nchannel. In addition, the signal path becomes a high impedance when the power supplies
are turned off. This structure offers a high degree of latchup and overvoltage protection—
at the expense of higher RON (~300 Ω), and more RON variation with signal level. For
more details of this protection method, refer to the individual product data sheets.
7.89
ANALOG-DIGITAL CONVERSION
NOTES:
7.90
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
SECTION 7.4: SAMPLE-AND-HOLD CIRCUITS
Walt Kester
Introduction and Historical Perspective
The sample-and-hold amplifier, or SHA, is a critical part of most data acquisition
systems. It captures an analog signal and holds it during some operation (most commonly
analog-digital conversion). The circuitry involved is demanding, and unexpected
properties of commonplace components such as capacitors and printed circuit boards may
degrade SHA performance.
When the SHA is used with an ADC (either externally or internally), the SHA
performance is critical to the overall dynamic performance of the combination, and plays
a major role in determining the SFDR, SNR, etc., of the system.
Although today the SHA function has become an integral part of the sampling ADC,
understanding the fundamental concepts governing its operation is essential to
understanding ADC dynamic performance.
When the sample-and-hold is in the sample (or track) mode, the output follows the input
with only a small voltage offset. There do exist SHAs where the output during the sample
mode does not follow the input accurately, and the output is only accurate during the hold
period (such as the AD684, AD781, and AD783). These will not be considered here.
Strictly speaking, a sample-and-hold with good tracking performance should be referred
to as a track-and-hold circuit, but in practice the terms are used interchangeably.
The most common application of a SHA is to maintain the input to an ADC at a constant
value during conversion. With many, but not all, types of ADC the input may not change
by more than 1 LSB during conversion lest the process be corrupted—this either sets very
low input frequency limits on such ADCs, or requires that they be used with a SHA to
hold the input during each conversion.
From a historial perspective, it is interesting that the ADC described by A. H. Reeves in
his famous PCM patent of 1939 (Reference 1) was a 5-bit 6-kSPS counting ADC where
the analog input signal drove a vacuum tube pulse-width-modulator (PWM) directly—the
sampling function was incorporated into the PWM. Subsequent work on PCM at Bell
Labs led to the use of electron-beam encoder tubes and successive approximation ADCs;
and Reference 2 (1948) describes a companion 50-kSPS vacuum tube sample-and-hold
circuit based on a pulse transformer drive circuit.
There was increased interest in sample-and-hold circuits for ADCs during the period of
the late 1950s and early 1960s as transistors replaced vacuum tubes. One of the first
analytical treatments of the errors produced by a solid-state sample-and-hold was
published in 1964 by Gray and Kitsopolos of Bell Labs (Reference 3). Edson and
Henning of Bell Labs describe the results of experimental work done on a 224-Mbps
PCM system, including the 9-bit ADC and a companion 12-MSPS sample-and-hold.
References 4, 5 , and 6 are representative of work done on sample-and-hold circuits
during the 1960s and early 1970s.
7.91
ANALOG-DIGITAL CONVERSION
In 1969, the newly acquired Pastoriza division of Analog Devices offered one of the first
commercial sample-and-holds, the SHA1 and SHA2 as shown in Figure 7.98. The
circuits were offered on PC boards, and the SHA1 had an acquisition time of 2 µs to
0.01%, dissipated 0.9 W, and cost approximately $225. The faster SHA2 had an
acquisition time of 200 ns to 0.01%, dissipated 1.7 W, and cost approximately $400.
They were designed to operate with 12-bit successive approximation ADCs also offered
on PC boards.
Acquisition Time: 2µs to 0.01% (SHA1), 200ns to 0.01% (SHA2)
Power: 900mW (SHA1), 1.7W (SHA2)
$225 (SHA1), $400 (SHA2)
Figure 7.98: "SHA1 and SHA2" Sample-and-Holds from Analog Devices'
Pastoriza Division, 1969
Modular and hybrid technology quickly made the PC board sample-and-holds obsolete,
and the demand for sample-and-holds increased as IC ADCs, such as the industrystandard AD574, came on the market. In the 1970s and into the 1980s, it was quite
common for system designers to purchase separate sample-and-holds to drive such
ADCs, because process technology did not allow integrating them together onto the same
chip. IC SHAs such as the AD582 (4-µs acquisition time to 0.01%), AD583 (6-µs
acquisition time to 0.01%), and the AD585 (3-µs acquisition time to 14-bit accuracy)
served the lower speed markets of the 1970s and 1980s.
Hybrid SHAs such as the HTS-0025 (25-ns acquisition time to 0.1%), HTC-0300 (200-ns
acquisition time to 0.01%), and the AD386 (25-µs acquisition time to 16-bits) served the
high-speed, high-end markets. By 1995, Analog Devices offered approximately 20
sample-and-hold products for various applications, including the following high-speed
ICs: AD9100/AD9101 (10-ns acquisition time to 0.01%), AD684 (quad 1-µs acquisition
time to 0.01%) and the AD783 (250-ns acquisition time to 0.01%).
However, ADC technology was rapidly expanding during the same period, and many
ADCs were being offered with internal SHAs (i.e., sampling ADCs). This made them
easier to specify and certainly easier to use. Integration of the SHA function was made
7.92
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
possible by new process developments including high-speed complementary bipolar
processes and advanced CMOS processes. In fact, the proliferation and popularity of
sampling ADCs has been so great that today (2003), one rarely has the need for a
separate SHA.
The advantage of a sampling ADC, apart from the obvious ones of smaller size, lower
cost, and fewer external components, is that the overall dc and ac performance is fully
specified, and the designer need not spend time ensuring that there are no specification,
interface, or timing issues involved in combining a discrete ADC and a discrete SHA.
This is especially important when one considers dynamic specifications such as SFDR
and SNR.
Although the largest applications of SHAs are with ADCs, they are also occasionally
used in DAC deglitchers, peak detectors, analog delay circuits, simultaneous sampling
systems, and data distribution systems.
Basic SHA Operation
Regardless of the circuit details or type of SHA in question, all such devices have four
major components. The input amplifier, energy storage device (capacitor), output buffer,
and switching circuits are common to all SHAs as shown in the typical configuration of
Figure 7.99.
Figure 7.99: Basic Sample-and-Hold Circuit
The energy-storage device, the heart of the SHA, is a capacitor. The input amplifier
buffers the input by presenting a high impedance to the signal source and providing
current gain to charge the hold capacitor. In the track mode, the voltage on the hold
capacitor follows (or tracks) the input signal (with some delay and bandwidth limiting).
In the hold mode, the switch is opened, and the capacitor retains the voltage present
before it was disconnected from the input buffer. The output buffer offers a high
impedance to the hold capacitor to keep the held voltage from discharging prematurely.
The switching circuit and its driver form the mechanism by which the SHA is alternately
switched between track and hold.
7.93
ANALOG-DIGITAL CONVERSION
There are four groups of specifications that describe basic SHA operation: track mode,
track-to-hold transition, hold mode, hold-to-track transition. These specifications are
summarized in Figure 7.100, and some of the SHA error sources are shown graphically in
Figure 7.101. Because there are both dc and ac performance implications for each of the
four modes, properly specifying a SHA and understanding its operation in a system is a
complex matter.
SAMPLE MODE
STATIC:
Offset
Gain Error
Nonlinearity
DYNAMIC:
Settling Time
Bandwidth
Slew Rate
Distortion
Noise
SAMPLE-TO-HOLD
TRANSITION
STATIC:
Pedestal
Pedestal
Nonlinearity
DYNAMIC:
Aperture Delay Time
Aperture Jitter
Switching Transient
Settling Time
HOLD MODE
HOLD-TO-SAMPLE
TRANSITION
STATIC:
Droop
Dielectric
Absorption
DYNAMIC:
Feedthrough
Distortion
Noise
DYNAMIC:
Acquisition Time
Switching
Transient
Figure 7.100: Sample-and-Hold Specifications
APERTURE
JITTER
ERROR
Figure 7.101: Some Sources of Sample-and-Hold Errors
Track Mode Specifications
Since a SHA in the sample (or track) mode is simply an amplifier, both the static and
dynamic specifications in this mode are similar to those of any amplifier. (SHAs which
have degraded performance in the track mode are generally only specified in the hold
mode.) The principle track mode specifications are offset, gain, nonlinearity, bandwidth,
7.94
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
slew rate, settling time, distortion, and noise. However, distortion and noise in the track
mode are often of less interest than in the hold mode.
Track-to-Hold Mode Specifications
When the SHA switches from track to hold, there is generally a small amount of charge
dumped on the hold capacitor because of non-ideal switches. This results in a hold mode
dc offset voltage which is called pedestal error as shown in Figure 7.102. If the SHA is
driving an ADC, the pedestal error appears as a dc offset voltage which may be removed
by performing a system calibration. If the pedestal error is a function of input signal
level, the resulting nonlinearity contributes to hold-mode distortion.
Figure 7.102: Track-to-Hold Mode Pedestal, Transient,
and Settling Time Errors
Pedestal errors may be reduced by increasing the value of the hold capacitor with a
corresponding increase in acquisition time and a reduction in bandwidth and slew rate.
Switching from track to hold produces a transient, and the time required for the SHA
output to settle to within a specified error band is called hold mode settling time.
Occasionally, the peak amplitude of the switching transient is also specified.
Perhaps the most misunderstood and misused SHA specifications are those that include
the word aperture. The most essential dynamic property of a SHA is its ability to
disconnect quickly the hold capacitor from the input buffer amplifier. The short (but nonzero) interval required for this action is called aperture time. The various quantities
associated with the internal SHA timing are shown in the Figure 7.103.
The actual value of the voltage that is held at the end of this interval is a function of both
the input signal and the errors introduced by the switching operation itself. Figure 7.104
shows what happens when the hold command is applied with an input signal of arbitrary
slope (for clarity, the sample to hold pedestal and switching transients are ignored). The
value that finally gets held is a delayed version of the input signal, averaged over the
7.95
ANALOG-DIGITAL CONVERSION
aperture time of the switch as shown in Figure 7.104. The first-order model assumes that
the final value of the voltage on the hold capacitor is approximately equal to the average
value of the signal applied to the switch over the interval during which the switch
changes from a low to high impedance (ta).
Figure 7.103: SHA Circuit Showing Internal Timing
Figure 7.104: SHA Waveforms
The model shows that the finite time required for the switch to open (ta) is equivalent to
introducing a small delay in the sampling clock driving the SHA. This delay is constant
and may either be positive or negative. It is called effective aperture delay time, aperture
delay time, or simply aperture delay, (te) and is defined as the time difference between
7.96
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
the analog propagation delay of the front-end buffer (tda) and the switch digital delay (tdd)
plus one-half the aperture time (ta/2). The effective aperture delay time is usually
positive, but may be negative if the sum of one-half the aperture time (ta/2) and the switch
digital delay (tdd) is less than the propagation delay through the input buffer (tda). The
aperture delay specification thus establishes when the input signal is actually sampled
with respect to the sampling clock edge.
Aperture delay time can be measured by applying a bipolar sinewave signal to the SHA
and adjusting the synchronous sampling clock delay such that the output of the SHA is
zero during the hold time. The relative delay between the input sampling clock edge and
the actual zero-crossing of the input sinewave is the aperture delay time as shown in
Figure 7.105.
Figure 7.105: Effective Aperture Delay Time
Aperture delay produces no errors, but acts as a fixed delay in either the sampling clock
input or the analog input (depending on its sign). If there is sample-to-sample variation in
aperture delay (aperture jitter), then a corresponding voltage error is produced as shown
in Figure 7.106. This sample-to-sample variation in the instant the switch opens is called
aperture uncertainty, or aperture jitter and is usually measured in picoseconds rms. The
amplitude of the associated output error is related to the rate-of-change of the analog
input. For any given value of aperture jitter, the aperture jitter error increases as the input
dv/dt increases.
Measuring aperture jitter error in a SHA requires a jitter-free sampling clock and analog
input signal source, because jitter (or phase noise) on either signal cannot be
distinguished from the SHA aperture jitter itself—the effects are the same. In fact, the
largest source of timing jitter errors in a system is most often external to the SHA (or the
ADC if it is a sampling one) and is caused by noisy or unstable clocks, improper signal
routing, and lack of attention to good grounding and decoupling techniques. SHA
aperture jitter is generally less than 50-ps rms, and less than 5-ps rms in high speed
devices. Details of measuring aperture jitter of an ADC can be found in Chapter 5.
7.97
ANALOG-DIGITAL CONVERSION
Figure 7.106: Effects of Aperture or Sampling Clock
Jitter on SHA Output
Figure 7.107 shows the effects of total sampling clock jitter on the signal-to-noise ratio
(SNR) of a sampled data system. The total rms jitter will be composed of a number of
components, the actual SHA aperture jitter often being the least of them.
tj = 50fs
120
tj = 0.1ps
SNR = 20log 10
1
2π ft j
tj = 1ps
100
18
16
14
tj = 10ps
80
SNR
(dB)
tj = 100ps
60
12
10
8
tj = 1ns
40
6
4
20
1
3
100
10
30
FULL-SCALE SINEWAVE ANALOG INPUT FREQUENCY (MHz)
Figure 7.107: Effects of Sampling Clock Jitter on SNR
7.98
ENOB
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
Hold Mode Specifications
During the hold mode there are errors due to imperfections in the hold capacitor, switch,
and output amplifier. If a leakage current flows in or out of the hold capacitor, it will
slowly charge or discharge, and its voltage will change. This effect is known as droop in
the SHA output and is expressed in V/µs. Droop can be caused by leakage across a dirty
PC board if an external capacitor is used, or by a leaky capacitor, but is most usually due
to leakage current in semiconductor switches and the bias current of the output buffer
amplifier. An acceptable value of droop is where the output of a SHA does not change by
more than ½ LSB during the conversion time of the ADC it is driving, although this value
is highly dependent on the ADC architecture. Where droop is due to leakage current in
reversed biased junctions (CMOS switches or FET amplifier gates), it will double for
every 10°C increase in chip temperature—which means that it will increase a thousand
fold between +25°C and +125°C. Droop can be reduced by increasing the value of the
hold capacitor, but this will also increase acquisition time and reduce bandwidth in the
track mode. Differential techniques are often used to reduce the effects of droop in
modern IC sample-and-hold circuits that are part of the ADC.
Figure 7.108: Hold Mode Droop
Even quite small leakage currents can cause troublesome droop when SHAs use small
hold capacitors. Leakage currents in PCBs may be minimized by the intelligent use of
guard rings. A guard ring is a ring of conductor which surrounds a sensitive node and is
at the same potential. Since there is no voltage between them, there can be no leakage
current flow. In a non-inverting application, such as is shown in Figure 7.109, the guard
ring must be driven to the correct potential, whereas the guard ring on a virtual ground
can be at actual ground potential (Figure 7.110). The surface resistance of PCB material
is much lower than its bulk resistance, so guard rings must always be placed on both
sides of a PCB—and on multi-layer boards, guard rings should be present in all layers.
7.99
ANALOG-DIGITAL CONVERSION
Figure 7.109: Drive the Guard Shield with the Same Voltage as the Hold
Capacitor to Reduce Board Leakage
Figure 7.110: Using a Guard Shield on a Virtual
Ground SHA Design
Hold capacitors for SHAs must have low leakage, but there is another characteristic
which is equally important: low dielectric absorption. If a capacitor is charged, then
discharged, and then left open circuit, it will recover some of its charge as shown in
Figure 7.111. The phenomenon is known as dielectric absorption, and it can seriously
degrade the performance of a SHA, since it causes the remains of a previous sample to
contaminate a new one, and may introduce random errors of tens or even hundreds of
mV.
7.100
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
Figure 7.111: Dielectric Absorption
Different capacitor materials have differing amounts of dielectric absorption—
electrolytic capacitors are dreadful (their leakage is also high), and some high-K ceramic
types are bad, while mica, polystyrene and polypropylene are generally good.
Unfortunately, dielectric absorption varies from batch to batch, and even occasional
batches of polystyrene and polypropylene capacitors may be affected. It is therefore wise
to pay 30-50% extra when buying capacitors for SHA applications and buy devices
which are guaranteed by their manufacturers to have low dielectric absorption, rather
than types which might generally be expected to have it.
Stray capacity in a SHA may allow a small amount of the ac input to be coupled to the
output during hold. This effect is known as feedthrough and is dependent on input
frequency and amplitude. If the amplitude of the feedthrough to the output of the SHA is
more than ½ LSB, then the ADC is subject to conversion errors.
In many SHAs, distortion is specified only in the track mode. The track mode distortion
is often much better than hold mode distortion. Track mode distortion does not include
nonlinearities due to the switch network, and may not be indicative of the SHA
performance when driving an ADC. Modern SHAs, especially high speed ones, specify
distortion in both modes. While track mode distortion can be measured using an analog
spectrum analyzer, hold mode distortion measurements should be performed using digital
techniques as shown in Figure 7.112. A spectrally pure sinewave is applied to the SHA,
and a low distortion high speed ADC digitizes the SHA output near the end of the hold
time. An FFT analysis is performed on the ADC output, and the distortion components
computed.
7.101
ANALOG-DIGITAL CONVERSION
Figure 7.112: Measuring Hold Mode Distortion
SHA noise in the track mode is specified and measured like that of an amplifier. Peak-topeak hold mode noise is measured with an oscilloscope and converted to an rms value by
dividing by 6.6. Hold mode noise may be given as a spectral density in nV/√Hz, or as an
rms value over a specified bandwidth. Unless otherwise indicated, the hold mode noise
must be combined with the track mode noise to yield the total output noise. Some SHAs
specify the total output hold mode noise, in which case the track mode noise is included.
Hold-to-Track Transition Specifications
When the SHA switches from hold to track, it must reacquire the input signal (which may
have made a full scale transition during the hold mode). Acquisition time is the interval of
time required for the SHA to reacquire the signal to the desired accuracy when switching
from hold to track. The interval starts at the 50% point of the sampling clock edge, and
ends when the SHA output voltage falls within the specified error band (usually 0.1% and
0.01% times are given). Some SHAs also specify acquisition time with respect to the
voltage on the hold capacitor, neglecting the delay and settling time of the output buffer.
The hold capacitor acquisition time specification is applicable in high speed applications,
where the maximum possible time must be allocated for the hold mode. The output buffer
settling time must of course be significantly smaller than the hold time.
Acquisition time can be measured directly using modern digital sampling scopes (DSOs)
or digital phosphor scopes (DPOs) which are insensitive to large overdrives.
SHA Architectures
As with op amps, there are numerous SHA architectures, and we will examine a few of
the most popular ones. The simplest SHA structure is shown in Figure 7.113. The input
signal is buffered by an amplifier and applied to the switch. The input buffer may either
be open- or closed-loop and may or may not provide gain. The switch can be CMOS,
FET, or bipolar (using diodes or transistors) and is controlled by the switch driver circuit.
The signal on the hold capacitor is buffered by an output amplifier. This architecture is
sometimes referred to as open-loop because the switch is not inside a feedback loop.
7.102
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
Notice that the entire signal voltage is applied to the switch, therefore it must have
excellent common-mode characteristics.
Figure 7.113: Open-Loop SHA Architecture
An implementation of this architecture is shown in Figure 7.114, where a simple diode
bridge is used for the switch. In the track mode, current flows through the bridge diodes
D1, D2, D3, and D4. For fast slewing input signals, the hold capacitor is charged and
discharged with the current, I. Therefore, the maximum slew rate on the hold capacitor is
equal to I/CH. Reversing the bridge drive currents reverse biases the bridge and places the
circuit in the hold mode. Bootstrapping the turn-off pulses with the held output signal
minimizes common-mode distortion errors and is key to the circuit. The reverse bias
bridge voltage is equal to the forward drops of D5 and D6 plus the voltage drops across
the series resistors R1 and R2. This circuit is extremely fast, especially if the input and
output buffers are open-loop followers, and the diodes are Schottky ones. The turn-off
pulses can be generated with high frequency pulse transformers or with current switches
as shown in Figure 7.115. This circuit can be used at any sampling rate, because the
diode switching pulses are direct-coupled to the bridge. Variations of this circuit have
been used since the mid 1960s in high speed PC board, modular, hybrid, and IC SHAs.
I
I
D1
D2
R1
D5
D3
D6
R2
D4
CH
I
I
BOOTSTRAP
Figure 7.114: Open-Loop SHA Using Diode Bridge Switch
7.103
ANALOG-DIGITAL CONVERSION
+15V
I
Q1
Q2
R1
6.2V
6.2V
D1
D3
D5
A=1
D2
6.2V
CH
D6
R2
6.2V
T
(ECL)
D4
A=1
H
H
Q3
Q4
I
T
(ECL)
–15V
Figure 7.115: Open-Loop SHA Implementation
The SHA circuit shown in Figure 7.116 represents a classical closed-loop design and is
used in many CMOS sampling ADCs. Since the switches always operate at virtual
ground, there is no common-mode signal across them.
Switch S2 is required in order to maintain a constant input impedance and prevent the
input signal from coupling to the output during the hold time. In the track mode, the
transfer characteristic of the SHA is determined by the op amp, and the switches do not
introduce dc errors because they are inside the feedback loop. The effects of charge
injection can be minimized by using the differential switching techniques shown in
Figure 7.117.
7.104
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
Figure 7.116: Closed-Loop SHA Based on Inverting
Integrator Switched at the Summing Point
Figure 7.117: Differential Switching Reduces Charge Injection
Internal SHA Circuits for IC ADCs
CMOS ADCs are quite popular because of their low power and low cost. The equivalent
input circuit of a typical CMOS ADC using a differential sample-and-hold is shown in
Figure 7.118. While the switches are shown in the track mode, note that they open/close
at the sampling frequency. The 16-pF capacitors represent the effective capacitance of
switches S1 and S2, plus the stray input capacitance. The CS capacitors (4 pF) are the
7.105
ANALOG-DIGITAL CONVERSION
sampling capacitors, and the CH capacitors are the hold capacitors. Although the input
circuit is completely differential, this ADC structure can be driven either single-ended or
differentially. Optimum performance, however, is generally obtained using a differential
transformer or differential op amp drive.
CH
16pF
CP
4pF
S4
CS
S1
VINA
4pF
+
S3
S2
VINB
A
CS
-
4pF
CP
16pF
S6
CH
S5
S7
4pF
SWITCHES SHOWN IN TRACK MODE
Figure 7.118: Simplified Input Circuit for a Typical Switched
Capacitor CMOS Sample-and-Hold
In the track mode, the differential input voltage is applied to the CS capacitors. When the
circuit enters the hold mode, the voltage across the sampling capacitors is transferred to
the CH hold capacitors and buffered by the amplifier A (the switches are controlled by the
appropriate sampling clock phases). When the SHA returns to the track mode, the input
source must charge or discharge the voltage stored on CS to a new input voltage. This
action of charging and discharging CS, averaged over a period of time and for a given
sampling frequency fs, makes the input impedance appear to have a benign resistive
component. However, if this action is analyzed within a sampling period (1/fs), the input
impedance is dynamic, and certain input drive source precautions should be observed.
The resistive component to the input impedance can be computed by calculating the
average charge that is drawn by CH from the input drive source. It can be shown that if CS
is allowed to fully charge to the input voltage before switches S1 and S2 are opened that
the average current into the input is the same as if there were a resistor equal to 1/(CSfS)
connected between the inputs. Since CS is only a few picofarads, this resistive component
is typically greater than several kΩ for an fS = 10 MSPS.
Figure 7.119 shows a simplified circuit of the input SHA used in the AD9042 12-bit,
41-MSPS ADC introduced in 1995 (Reference 7). The AD9042 is fabricated on a high
speed complementary bipolar process, XFCB. The circuit comprises two independent
SHAs in parallel for fully differential operation—only one-half the circuit is shown in the
figure. Fully differential operation reduces the error due to droop rate and also reduces
second-order distortion. In the track mode, transistors Q1 and Q2 provide unity-gain
buffering. When the circuit is placed in the hold mode, the base voltage of Q2 is pulled
negative until it is clamped by the diode, D1. The on-chip hold capacitor, CH, is
7.106
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
nominally 6 pF. Q3 along with CF provide output current bootstrapping and reduce the
VBE variations of Q2. This reduces third-order signal distortion. Track mode THD is
typically –93 dB at 20 MHz. In the time domain, full-scale acquisition time to 12-bit
accuracy is 8 ns. In the hold mode, signal-dependent pedestal variations are minimized by
the voltage bootstrapping action of Q3 and the A = 1 buffer along with the low
feedthrough parasitics of Q2. Hold mode settling time is 5 ns to 12-bit accuracy. Holdmode THD at a clock rate of 50 MSPS and a 20-MHz input signal is –90 dB.
I
Q2
Q1
CIRCUIT SHOWN
IN TRACK MODE
CH
Q3
D1
I
A=1
FULLY DIFFERENTIAL,
ONLY ONE-HALF SHOWN
I
Q5
H
Q4
CF
T
H
T
2I
Figure 7.119: SHA Used in AD9042 12-Bit, 41 MSPS ADC Introduced in 1995
Figure 7.120 shows a simplified schematic of one-half of the differential SHA used in the
AD6645 14-bit, 105-MSPS ADC recently introduced (Reference 9) gives a complete
description of the ADC including the SHA). In the track mode, Q1, Q2, Q3, and Q4 form
a complementary emitter follower buffer which drives the hold capacitor, CH. In the hold
mode, the polarity of the bases of Q3 and Q4 is reversed and clamped to a low
impedance. This turns off Q1, Q2, Q3, and Q4, and results in double isolation between
the signal at the input and the hold capacitor. As previously discussed, the clamping
voltages are boostrapped by the held output voltage, thereby minimizing nonlinear
effects.
Track mode linearity is largely determined by the VBE modulation of Q3 and Q4 when
charging CH. Hold mode linearity depends on track mode linearity plus nonlinear errors
in the track-to-hold transitions caused by imbalances in the switching of the base voltages
of Q3 and Q4 and the resulting imbalance in charge injection through their base-emitter
junctions as they turn off.
7.107
ANALOG-DIGITAL CONVERSION
H+
T+
T+
H+
Q5
Q6
Q3
Q2
FULLY DIFFERENTIAL,
ONLY ONE-HALF SHOWN
CLAMP
RCH
CH
Q1
Q4
Q7
T–
H–
Q8
H–
T–
Figure 7.120: SHA Used in AD6645 14-Bit, 105 MSPS ADC
SHA Applications
By far the largest application of SHAs is driving ADCs. Most modern ADCs designed for
signal processing are sampling ones and contain an internal SHA optimized for the
converter design. Sampling ADCs are completely specified for both dc and ac
performance and should be used in lieu of discrete SHA/ADC combinations wherever
possible. In a very few selected cases, especially those requiring wide dynamic range and
low distortion, there may be advantages to using a discrete combination.
A similar application uses a low distortion SHA to minimize the effects of codedependent DAC glitches as shown in Figure 7.121. Just prior to latching new data into
the DAC, the SHA is put into the hold mode so that the DAC switching glitches are
isolated from the output. The switching transients produced by the SHA are not codedependent, occur at the update frequency, and are easily filterable. This technique may be
useful at low frequencies to improve the distortion performance of DACs, but has little
value when using high speed low-glitch low distortion DACs designed especially for
DDS applications where the update rate is several hundred MHz.
Rather than use a single ADC per channel in a simultaneous sampled system, it is often
more economical to use multiple SHAs followed by an analog multiplexer and a single
ADC (Figure 7.122). Similarly, in data distribution systems multiple SHAs can be used to
route the sequential outputs of a single DAC to multiple channels as shown in Figure
7.123; although this is not as common, as multiple DACs usually offer a better solution.
7.108
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
Figure 7.121: Using a SHA as a DAC Deglitcher
Figure 7.122: Simultaneous Sampling Using Multiple
SHAs and a Single ADC
7.109
ANALOG-DIGITAL CONVERSION
Figure 7.123: Data Distribution System Using Multiple
SHAs and a Single DAC
A final application for SHAs is shown in Figure 7.124, where SHAs are cascaded to
produce analog delay in a sampled data system. SHA 2 is placed in hold just prior to the
end of the hold interval for SHA 1. This results in a total pipeline delay greater than the
sampling period T. This technique is often used in multi-stage pipelined subranging
ADCs to allow for the conversion delays of successive stages. In pipelined ADCs, a 50%
duty cycle sampling clock is common, thereby allowing alternating clock phases to drive
each SHA in the pipeline (see Chapter 3 for more details of the pipelined ADC
architecture and the use of SHAs for analog delay).
Figure 7.124: SHAs Used for Analog Pipelined Delay
7.110
DATA CONVERTER SUPPORT CIRCUITS
7.4 SAMPLE-AND-HOLD CIRCUITS
REFERENCES:
7.4 SAMPLE-AND-HOLD CIRCUITS
1.
Alec Harley Reeves, "Electric Signaling System," U.S. Patent 2,272,070, filed November 22, 1939,
issued February 3, 1942. Also French Patent 852,183 issued 1938, and British Patent 538,860 issued
1939. (the classic patents on PCM including descriptions of a 5-bit, 6-kSPS vacuum tube ADC and
DAC).
2.
L. A. Meacham and E. Peterson, "An Experimental Multichannel Pulse Code Modulation System of
Toll Quality," Bell System Technical Journal, Vol 27, No. 1, January 1948, pp. 1-43. (describes the
culmination of much work leading to this 24-channel experimental PCM system. In addition, the
article describes a 50-kSPS vacuum tube sample-and-hold based on a pulse transformer driver).
3.
J. R. Gray and S. C. Kitsopoulos, "A Precision Sample-and-Hold Circuit with Subnanosecond
Switching," IEEE Transactions on Circuit Theory, CT11, September 1964, pp. 389-396. (an
excellent description of a solid-state transformer-driven diode bridge SHA, along with a detailed
mathematical analysis of the circuit and associated errors).
4.
J. O. Edson and H. H. Henning, "Broadband Codecs for an Experimental 224Mb/s PCM Terminal,"
Bell System Technical Journal, Vol. 44, pp. 1887-1940, Nov. 1965. (summarizes experiments on
ADCs based on the electron tube coder as well as a bit-per-stage Gray code 9-bit solid state ADC. The
electron beam coder was 9-bits at 12MSPS, and represented the fastest of its type).
5.
D. J. Kinniment, D. Aspinall, and D.B.G. Edwards, "High-Speed Analogue-Digital Converter," IEE
Proceedings, Vol. 113, pp. 2061-2069, Dec. 1966. (a 7-bit 9MSPS three-stage pipelined error
corrected converter is described based on recircuilating through a 3-bit stage three times. Tunnel
(Esaki) diodes are used for the individual comparators. The article also shows a proposed faster
pipelined 7-bit architecture using three individual 3-bit stages with error correction. The article also
describes a fast bootstrapped transformer-driven diode-bridge sample-and-hold circuit).
6.
O. A. Horna, "A 150Mbps A/D and D/A Conversion System," Comsat Technical Review, Vol. 2, No.
1, pp. 39-72, 1972. (a description of a subranging ADC including a detailed analysis of the sampleand-hold circuit).
7.
Roy Gosser and Frank Murden, "A 12-bit 50MSPS Two-Stage A/D Converter," 1995 ISSCC Digest
of Technical Papers, p. 278. (a description of the AD9042 error corrected subranging ADC using
MagAMP stages for the internal ADCs).
8.
Carl Moreland, "An 8-bit 150 MSPS Serial ADC," 1995 ISSCC Digest of Technical Papers, Vol. 38,
p. 272. (a description of an 8-bit ADC with 5 folding stages followed by a 3-bit flash converter,
including a discussion of the sample-and-hold circuit).
9.
Carl Moreland, Frank Murden, Michael Elliott, Joe Young, Mike Hensley, and Russell Stop, "A 14-bit
100-Msample/s Subranging ADC," IEEE Journal of Solid State Circuits, Vol. 35, No. 12, December
2000, pp. 1791-1798. (describes the architecture used in the 14-bit, 105MSPS AD6645 ADC and also
the sample-and-hold circuit).
7.111
ANALOG-DIGITAL CONVERSION
NOTES:
7.112