Appendix D

October 2002

New Products Appendix to the RMS

to DC Conversion Application Guide
INTRODUCTION* Both the AD736 and AD737 are optimized for use in
Since the last printing of this Applications Guide in portable instruments; they consume less than 200 µA
1986, Analog Devices has introduced four new rms of quiescent current and accept signal levels from
products. These devices supplement ADI’s original 0 mV rms to 200 mV rms. The AD737 also has a power-
rms products: the AD536A, AD636, and AD637. down input that allows the user to reduce its quiescent
The AD736 and AD737 are low power, low cost rms current from 160 µA to 40 µA in portable applications.
converters designed for use in portable instruments. The AD637 should be chosen if the application requires
™
The AD8361 and AD8362 TruPwr RF power detec- high accuracy and a quick response for large, abrupt
tors are designed for accurate control of radiated power changes in signal level. The AD637’s settling time is
levels in cellular, broadband, CATV, MMDS, and independent of signal level, while for a given value
LMDS communications equipment. of averaging capacitor, the settling times of the
AD536A, AD636, AD736, and AD737 will be longer
HOW TO SELECT AN RMS-TO-DC CONVERTER for low level signals and shorter for high level signals.
Selecting an rms-to-dc converter means picking the Some monolithic rms converters use a sigma-delta
product whose attributes best match the requirements computational technique. This method can provide
of the application. Unfortunately, no one converter wide bandwidth while operating at low supply current
fits every situation so trade-offs must be made between levels but suffers from a very serious low frequency
accuracy, bandwidth, power consumption, input signal “rumble.” In effect, as the input signal frequency
level, crest factor, settling time, and cost. increases, a larger percentage of the input signal is aliased
The AD637 accepts input voltages as high as 7 V rms down to dc, producing a low frequency modulation at
and is Analog Devices’ most accurate and widest the output of the converter. This rumble is a low
bandwidth rms-to-dc converter. Its –3 dB bandwidth frequency error that increases both with rms signal
is 8 MHz for a 1 V rms input. It has an auxiliary dB level and signal frequency. When this type of rms
output and a power-down feature that reduces its converter is used in a DVM, the lower digits on the
quiescent current from 3 mA to 450 µA. display will flicker and, as the rms level or frequency
increases, more and more display digits will become
The AD536A and its 200 mV companion product, erratic. Attempts to minimize this problem using a
the AD636, are designed for low cost, general-purpose low-pass “rumble filter” after the converter will
applications. These converters offer true rms conver- result in longer settling times between readings and
sion accuracy and will accurately process input signals will, in any case, be generally ineffective in removing the
with high crest factors. The AD536A has a 2 V rms flicker. In contrast, the AD736/AD737 architecture
full-scale range, while that of the AD636 is 200 mV rms. is continuous-time and not subject to this type of
Both products consume only about 1 mA of power unstable behavior.
supply current. They also offer a dB output feature
with an output that is proportional to the logarithm of
the rms input signal.

*Portions of this appendix are excerpts from Application Note AN-268, RMS to DC Converters Ease Measurement Tasks, by Bob Clarke,
Mark Fazio, and Dave Scott: Analog Devices Publication.
TruPwr is a trademark of Analog Devices, Inc.

1
Table I provides a quick performance comparison between Analog Devices’ rms products.
Table I. RMS Converter Comparison Table

AD536AJ AD637J AD636J AD736J AD737J

Input Dynamic Range 7 V rms 7 V rms 1 V rms 1 V rms* 1 V rms*
Nominal RMS Full Scale 2V 2V 200 mV 200 mV 200 mV
Peak Trans. Input ± 20 V ± 15 V ± 2.8 V ± 2.7 V ± 2.7 V
Max Total Error 5 mV 1 mV 0.5 mV 0.5 mV 0.4 mV
No Ext. Trim ± 0.5% rdg ± 0.5% rdg ± 1% rdg ± 0.5% rdg ± 0.5% rdg
–3 dB Bandwidth
Full Scale 2 MHz 8 MHz 1.3 MHz 190 kHz 190 kHz
0.1 V rms 300 kHz 600 kHz 800 kHz 170 kHz 170 kHz
Error at Crest
Factor of 5 –0.3% ± 0.15% –0.5% ± 2.5% ± 2.5%
For V rms @1 V @1 V @200 mV @200 mV @200 mV
Power Supply
Volts ± 13 V ± 13 V +2 V, –2.5V +2.8, –3.2 V +2.8, –3.2 V
to ± 18 V to ± 18 V to ± 12 V to ± 16.5 V to ± 16.5 V
Current 1 mA 2 mA 800 µA 230 µA 170 µA
*± 5 V to ± 16.5 V dual-supply operation only. 200 mV rms under +5 V, –3 V single-supply conditions.

THE AD736 AND AD737 LOW COST either as a true rms converter or as an average responding
RMS CONVERTERS (MAD) rectifier. The AD736 and AD737 converters
As shown in Figures 1 and 2, the AD736 and AD737 are optimized for low power operation, and their
architectures are very similar. The major difference averaging capacitor appears directly across a diode in
between these two products is that the AD736 includes the rms core. Because of this, the averaging time constant
an output buffer amplifier for general-purpose applica- will increase as the rms input level decreases. Conse-
tions, while the AD737 is unbuffered for low power quently, lower input levels allow the circuit to perform
operation. The AD737 also includes a power-down better (due to increased averaging) but result in longer
feature that further reduces its standby current consump- settling times, requiring more time between readings.
tion to a mere 25 µA. Both products can be operated
8
COM

+VS
7
+VRMS
CC OUTPUT
1 6
8k⍀
8k⍀
VIN CF

2 3
CF
FET
OP AMP
IB<10pA CURRENT
MODE
RECTIFIER

CAV
5

CAV

4
RMS TRANSLINEAR CORE –VS

Figure 1. Simplified Schematic of the AD736

2
The input to the AD736 and AD737 is through a The design of the AD737 is very similar to that of the
FET input op amp connected as a unity-gain buffer. AD736. In order to reduce power consumption, the
This amplifier allows both a high impedance, buffered AD737 does not have an output buffer amplifier.
input (Pin 2) or a low impedance input (Pin 1) that Instead, it uses an NPN transistor to drive an 8 kΩ
provides a wider dynamic range. The high impedance internal load resistor.
input, with its low input bias current, is well suited for The converter develops its output voltage by sinking
use with high impedance input attenuators. current through this resistor. The external averaging
The output of the buffer drives a current mode capacitor (CAV) for the AD736 and AD737 is connected
rectifier (absolute value circuit) that in turn drives between Pins 4 (–VS) and 5 (CAV), which places CAV
an rms core. across a transistor’s base-emitter junction in the rms
In the AD736 (Figure 1), the output of the rms core core. This means that a diode is in parallel with the
drives the summing node of an inverting op amp con- averaging capacitor; the resulting time constant is
nected as a current-to-voltage converter. Pin 3 gives therefore inversely proportional to the rms value.
access to this node to connect a filter capacitor, CF,
in parallel with the 8 kΩ feedback resistor, to form a
one-pole low-pass filter.

8
COM

+VS
7

CC
8k⍀
1
8k⍀

VIN 6
2 OUTPUT
(–VRMS Out)
FET
Op Amp
IB<10pA
CURRENT
MODE
RECTIFIER

POWER
DOWN CAV
3 BIAS
SECTION 5

CAV
4
–VS
RMS TRANSLINEAR CORE

Figure 2. Simplified Schematic of the AD737

3
Because the external averaging capacitor, CAV, “holds” 1V
VS = ⴞ5V, CC = 22␮F, CF = 0␮F
the rectified input signal during rms computation, its
value directly affects the accuracy of the measurement—
especially at low frequencies. (The larger the value of 100mV
CAV = 100␮F
CAV, the lower the error.) Also, because the averaging

INPUT LEVEL – rms

capacitor appears across a base-emitter junction in the
CAV = 33␮F
squarer/divider, the averaging time constant will 10mV
increase linearly as the input signal is reduced.
AD736/AD737 settling time versus rms input level
CAV = 10␮F
and CAV is shown in Figure 3. 1mV

Due to the varying time constant, as the input level

decreases, errors due to nonideal averaging will decrease,
while the time it takes for the circuit to settle to the 100␮V
1ms 10ms 100ms 1s 10s 100s
new rms level will increase. SETTLING TIME
Therefore, lower input levels allow the circuit to per-
Figure 3. Settling Time vs. RMS Input Level of
form better (due to increased averaging) but increase
the AD736 and AD737 for Various Values of CAV
the waiting time between measurements, because the
capacitor takes longer to charge or discharge. Thus, a
CALCULATING AD737 SETTLING TIME
trade-off between computational accuracy and settling
The graph of Figure 3 may be used to approximate
time is required.
the time required for the AD736 or AD737 to settle
Table II provides practical values of CAV and CF for when its input level is reduced in amplitude. The total
several common applications. time required for the rms converter to settle will be
the difference between two settling times extracted
from the graph—the initial settling time minus the final
settling time.

Table II. Practical Values for CAV and CF for AD736 and AD737

Low
Frequency Settling
RMS Cutoff Max Crest Time*
Application Input Level ( –3 dB) Factor CAV CF to 1%
General-Purpose 0 V–1 V 20 Hz 5 150 µF 10 µF 360 ms
RMS 200 Hz 5 15 µF 1 µF 36 ms
Computation 0 mV–200 mV 20 Hz 5 33 µF 10 µF 360 ms
200 Hz 5 3.3 µF 1 µF 36 ms
General-Purpose 0 V–1 V 20 Hz 33 µF 1.2 sec
Average 200 Hz 3.3 µF 120 ms
Responding 0 mV–200 mV 20 Hz 33 µF 1.2 sec
200 Hz 3.3 µF 120 ms
SCR Waveform 0 mV–200 mV 50 Hz 5 100 µF 33 µF 1.2 sec
Measurement 60 Hz 5 82 µF 27 µF 1.0 sec
0 mV–100 mV 50 Hz 5 50 µF 33 µF 1.2 sec
60 Hz 5 47 µF 27 µF 1.0 sec
Audio
Applications
Speech 0 mV–200 mV 300 Hz 3 1.5 µF 0.5 µF 18 ms
Music 0 mV–100 mV 20 Hz 10 100 µF 68 µF 2.4 sec
*Settling time is specified over the stated rms input level with the input signal increasing from zero. Settling times will be greater for decreas-
ing amplitude input.

4
As an example, consider the following conditions: Normally, the input offset errors in the traditional
a 33 µF averaging capacitor an initial rms input level monolithic rms converter will create a region of diode
of 100 mV and a final (reduced) input level of 1 mV. nonconduction at low level input voltages. That is,
From Figure 3, the initial settling time (where the any input voltages that are smaller than the input
100 mV line intersects the 33 µF line) is around 80 ms. offset voltage will not be rectified and a “dead zone”
The settling time corresponding to the new or final is created.
input level of 1 mV is about 8 seconds. Therefore, the However, the AD736 and AD737 are specifically
net time for the circuit to settle to its new value will be designed to eliminate this problem. The maximum
dominated by the final settling time. Figure 4 shows input offset voltage of these rms converters is 3 mV. If
the additional error versus the crest factor of the Pin 1 is directly grounded, this offset voltage will limit
AD736 and AD737 for various values of CAV. the converter’s low level resolution. However, as
shown in Figure 5, the use of capacitor CC between
6
CAV = 10␮F CAV = 33␮F Pin 1 and ground will ac couple the low impedance
input pin and “float” this input above ground. This
ADDITIONAL ERROR – % of Reading

5 prevents any dc currents from flowing through the

3ms BURST of 1kHz,
= 3 CYCLES 8 kΩ internal resistor and creating an input voltage
4 200mV RMS SIGNAL offset. Capacitor CC should be chosen to provide a
VS = ⴞ5V
CC = 22␮F, CF = 100␮F low frequency cutoff substantially below the lowest
3 CAV = 100␮F
signal input frequency.
The 3 db roll-off frequency of CC =
2
1
1 2 π (8 ,000 Ω) (CC in Farads)
CAV = 250␮F

For most applications, a value of 10 µF

0
1 2 3 4 5 (FCUTOFF = 2 Hz) will suffice. A good rule of thumb
CREST FACTOR – V PEAK /V RMS is to use a value of CC approximately equal to one-
third that of CAV.
Figure 4. Additional Error vs. Crest Factor of the CC
AD736 and AD737 for Various Values of CAV
AC Coupling Design Considerations
The AD736 and AD737 rms converters offer the designer
CC AD736/AD737
the option of ac coupling both the input signal and 8k⍀ COM
1 8
the dc offset voltages on the rms converter’s input stage.
CIN VIN FULL
WAVE
CC VIN 2 RECTIFIER 7
RIN INPUT
AMPLIFIER
1M⍀
BIAS
3 SECTION RMS CORE 6

CC AD736/AD737 COM
8k⍀ 4 5
1 8

VIN FULL
WAVE
VIN 2 RECTIFIER 7
INPUT
AMPLIFIER
Figure 6. AC Coupling Using Capacitor CIN
3
BIAS
SECTION RMS CORE 6 Figure 6 shows ac input coupling for Pin 2. Capacitor
CIN is necessary if the input signal is an ac waveform
4 5
riding on a dc voltage, or if the rms converter is oper-
ating from a single-supply voltage. In this case, Pin 1 will
be “floating” above ground and CIN is needed to prevent
the rms converter from full-wave rectifying the differ-
Figure 5. Using Capacitor CC to Block Internal ential voltage between Pins 1 and 2, which often will
Offset Voltage Errors result in input overload. A resistor is needed between
Pin 2 of the rms converter and ground to provide a dc
return path for input bias currents. Note that capacitor
CC is still needed to prevent input offset voltage errors.
5
APPLICATIONS OF THE AD736 AND AD737 Figure 8 shows the AD736’s performance as a precision
AD736 as Precision Rectifier rectifier.
Building a precision rectifier from discrete components
requires two op amps, two diodes, and a handful of
matched resistors. An easy way to replace all these parts 200mV 100mV 2µS
and save some board space is to use an rms-to-dc
100
converter IC. Just omit the averaging capacitor and 90

disconnect the feedback; this uses only the converter’s

internal precision rectifier (Figure 7), which, being
monolithic, has inherently matched diodes.

CC

10

0%

1 CC COMMON 8

VIN 2 VIN +VS 7 +5V

AD736 0.1␮F
1M⍀
3 CF OUTPUT 6 VOUT

100mV 200mV 10µS

4 –VS CAV 5

0.1␮F 100
90

–5V

Figure 7. AD736 Connected as a Precision Rectifier

A precision rectifier circuit must provide enough gain to
forward bias its rectifier diodes. So, as the input signal 10
gets smaller in amplitude, more gain is needed. The 0%

traditional circuit uses an op amp to provide this gain.

However, it will usually have a fixed gain/bandwidth
product, which means that the rectifier’s bandwidth
will change with the input signal level. In contrast to a Figure 8. Performance of AD736 Precision
discrete circuit, the internally trimmed, monolithic Rectifier at 1 kHz (Top) and 19 kHz (Bottom)
design of the AD736 and AD737 greatly helps to
minimize this effect.

6
True RMS and Average Value Circuit R7 and R8 form a voltage divider to allow operation
Figure 9 shows a circuit that measures both the true from a single-supply voltage or battery. Capacitors
rms value and the rectified average value of an ac signal. C4 and C5 bypass any signal currents on VS or VS/2
This design uses two low cost ICs in SOIC packages to ground.
and consumes only 180 µA quiescent current. Operat- The rms converter IC has two inputs: a high imped-
ing from a 5 V single supply, this circuit has an input ance (1012 Ω) input (at Pin 2) and an 8 kΩ, wide
dynamic range from below 30 mV to greater than 3 V dynamic range input via Pin 1. The rms converter’s
rms. Sine wave accuracy is quite good (see perfor- full-scale input range is normally 200 mV. This can
mance data below) and bandwidth is approximately be greatly increased by adding an external resistance,
100 kHz, depending on the input level. The circuit can in this case resistor R1 and trimpot R2, between the
also measure a 1 V rms, crest factor of 5 pulse train signal input and Pin 1. This has the added advantage
with less than 1% of reading error. of increasing the circuit’s input impedance.
Average responding measurements and rms have The AD737JR measures the true rms value when
traditionally used different circuits. However, in some switch SW1 connects its averaging capacitor, CAV, to
cases, it may be extremely useful to know both the Pin 5. The averaging capacitor performs the “mean”
rms and rectified average value of an ac waveform. portion of the rms (root-mean-square) function.
The ratio of rms to rectified average value is one way Removing CAV, by opening SW1, converts the circuit
to determine the characteristics of a particular wave- to rectified average value operation. Resistor R6 allows
form without actually seeing it on an oscilloscope. a small leakage current to flow past the switch, keeping
For example, the rms/average value ratio for a 1 V the capacitor charged and preventing any large surge
peak undistorted sine wave is 0.707 V/0.636 V or 1.11, currents from flowing into or out of CAV when the
a symmetrical square wave is 1.0, a triangular wave is switch is closed.
1.155, and Gaussian noise is 1.253.
The rms value of a sine wave is 0.707 V peak while the
Circuit Operation rectified average value is 0.636 V peak. This ratio of
As shown in Figure 9, an AD737 rms converter IC 0.707 V/0.636 V is equivalent to an 11% scale factor
drives an AD8541AR micropower op amp. Resistors difference between the two measurement methods. If it is
INPUT SCALEFACTOR ADJ
+5V 0.47␮F
C1
R1
0.47␮F CF
69.8k⍀ 1% R2
CC COM
INPUT 5k⍀
R3
78.7k⍀ R4 80.6k⍀
VIN +VS
5k⍀ R5
C2 0.01␮F
AD737JR OUTPUT ZERO ADJ +5V 0.01␮F

CF OUTPUT
OUTPUT
AD8541AR
–VS CAV

C3 0.01
␮F

SW1

33␮F + AVER RMS

CAV
+ +5V
R6 2.2␮F
C4 R7
100k⍀ 100k⍀

+2.5V
+
1␮F R8
*CAV IS DISCONNECTED 100k⍀
–3dB BW CAV CF* C5
IN THE AVERAGE VALUE MODE.
10Hz 68␮F 0.82␮F THEREFORE, THE OUTPUT RIPPLE
WILL BE NOTICEABLY HIGHER AT VERY
20Hz 33␮F 0.47␮F LOW FREQUENCIES. SIMPLY INCREASE
100Hz 6.8␮F 0.1␮F THE VALUE OF CF TO REDUCE RIPPLE
TO THE DESIRED LEVEL.

Figure 9. An RMS/Average Rectified Value Measurement Circuit

7
desired to have this circuit accurately read the rms value Measured Performance Data
for sine waves in the rectified average value mode, SW1 1 kHz Sine Wave Accuracy
can be a two-pole switch. The second pole can connect VIN is in ac volts rms as monitored by Keithley 191
a 523 kΩ 1% resistor in parallel with R1 to increase DVM in ac mode. 5 VDC supply.
the scale factor by 11% in the average value mode.
VOUT Rectified
The AD737JR drives the AD8541AR op amp with a Average
negative flowing output current. The op amp operates VIN VOUT rms Value
as a current-to-voltage converter and also inverts the
signal, providing an output voltage that swings more 3V 2.9999 2.6762
positive with increasing input level. Resistor R5’s 1V 1.0027 0.8947
value of 80 kΩ matches the effective input resistance 0.3 V 0.30201 0.2698
of the AD737 (R1 + R2 + 8 kΩ) so that input/output 0.1 V 0.10082 0.09947
scaling is 1:1. Resistor R3 and trimpot R4 cause a 0.03 V 0.02960 0.02956
current to flow from the supply to the op amp sum- Error versus Crest Factor. +5 VDC Supply,
ming junction. This offsets the op amp output such 1 V rms, 100 ␮s pulse.
that the circuit’s output is approximately zero with no
voltage applied. Note that this circuit has a maximum Duty Cycle Varied for Desired Crest Factor.
supply voltage limit of 5.5 V; operation may be extended Crest Factor % of Reading Error
up to 12 V by substituting an OP-196GS op amp for
the AD8541AR. 3 0.67%
5 0.98%
Circuit calibration:
10 4.7%
1. Adjust trimpot R4 to midscale and set SW1 for rms.
2. Apply a 2.000 V rms, 1 kHz sine wave input signal. Extending the AD736 and AD737 Full-Scale
Input Ranges
3. Adjust R2 until the circuit’s output voltage is The high impedance input (Pin 2) of the AD736 and
2.000 V dc. AD737 allows simple resistive attenuators (Figure 10)
4. Reduce the input to 100 mV rms and adjust offset to be used to extend their input range. Without input
trimpot R4 for a reading of 100 mV dc. attenuation, both the AD736 and AD737 can accu-
rately measure input signals as large as 200 mV rms
5. Repeat Step 3.
with crest factors of 1 to 3.
As the dc offset circuitry is ratiometric, it will remain
calibrated with modest variations in supply voltage. The
measured PSRR of this circuit (over a 4.5 V to 5.5 V
supply range) is approximately 61 dB.

C1
0.01␮F C3
1kV 10␮F
VIN 200mV +
+5V
9M⍀
2V 1N4148
1 CC COMMON 8 +5V
900k⍀ 47k⍀
1W U1
20V 2 VIN +VS 7
AD736 C4
90k⍀ 0.1␮F
1N4148 3 CF OUTPUT 6
200V VRMS
10k⍀ –5V 4 –VS CAV 5

C2
0.1␮F +
CAV 33␮F
C1 AND THE RESISTIVE
DIVIDER FORM A 1.6Hz (–3dB) +
HIGH-PASS FILTER CF 10␮F

Figure 10. By Using an External Input Attenuator, the Measurement Range of the AD736
and AD737 Can be Extended

8
The external attenuator simply reduces the full-scale scaled by the factor of attenuation used. An external
input to the 200 mV rms input range of the AD736 or attenuator can also be used with the converter’s low
AD737. For a maximum 7 V rms input (10 V peak), for impedance input (Pin 1), as shown in Figure 10.
example, the attenuator should be a 35:1 (7 V/200 mV)
Figures 11 and 12 show the recommended connections
voltage divider. The reading of the converter should be
for external offset and scale factor.

DC-COUPLED

CC
+ 10␮F
AC-COUPLED (OPTIONAL)
CC COM
8k⍀
1 AD736 8

VIN FULL +VS

WAVE 8k⍀
VIN 2 RECTIFIER 7
INPUT
AMPLIFIER
CF OUTPUT
3 BIAS 6
SECTION OUTPUT
AMPLIFIER CAV
–VS RMS
+VS CORE
4 5

39M⍀
1M⍀
OUTPUT
+
VOS CAV 33␮F
–VS ADJUST
+
CF 10␮F
(OPTIONAL)

Figure 11. AD736 External VOS Adjustment

OFFSET ADJUST
500k⍀
+VS –VS

1M⍀
1k⍀

CC COM
8k⍀ AD737 499⍀
1 8
1k⍀
FULL SCALE
VIN +VS
WAVE FACTOR
VIN 2 RECTIFIER 8k⍀ 7 ADJUST
INPUT
AMPLIFIER
VOUT
3 6

Figure 12. AD737 DC-Coupled VOS and Scale Factor Adjustments

9
10
11
12
PRINTED IN U.S.A. G03133–0–10/02(0)