AD737
Table I. Error Introduced by an Average Responding Circuit When Measuring Common Waveforms
Waveform Type
1 Volt Peak
Amplitude
Crest Factor
(V
PEAK
/V rms)
True rms Value
Average Responding
Circuit Calibrated to
Read rms Value of
Sine Waves Will Read
% of Reading Error*
Using Average
Responding Circuit
Undistorted
Sine Wave
Symmetrical
Square Wave
Undistorted
Triangle Wave
Gaussian
Noise (98% of
Peaks <1 V)
Rectangular
Pulse Train
SCR Waveforms
50% Duty Cycle
25% Duty Cycle
1.414
0.707 V
0.707 V
0%
1.00
1.73
1.00 V
0.577 V
1.11 V
0.555 V
+11.0%
–3.8%
3
2
10
2
4.7
0.333 V
0.5 V
0.1 V
0.495 V
0.212 V
0.295 V
0.278 V
0.011 V
0.354 V
0.150 V
–11.4%
–44%
–89%
–28%
–30%
Mathematically, the rms value of a voltage is defined (using a
simplified equation) as:
V rms
=
Avg.(V
2
)
This involves squaring the signal, taking the average, and then
obtaining the square root. True rms converters are “smart recti-
fiers”: they provide an accurate rms reading regardless of the
type of waveform being measured. However, average responding
converters can exhibit very high errors when their input signals
deviate from their precalibrated waveform; the magnitude of the
error will depend upon the type of waveform being measured.
As an example, if an average responding converter is calibrated
to measure the rms value of sine-wave voltages, and then is used
to measure either symmetrical square waves or de voltages, the
converter will have a computational error 11% (of reading)
higher than the true rms value (see Table I).
AD737 THEORY OF OPERATION
As shown by Figure 16, the AD737 has four functional subsec-
tions: input amplifier, full-wave rectifier, rms core and bias sec-
tions. The FET input amplifier allows both a high impedance,
buffered input (Pin 2) or a low impedance, wide-dynamic-range
C
C
10 F
(OPTIONAL
input (Pin 1). The high impedance input, with its low input
bias current, is well suited for use with high impedance input
attenuators. The input signal may be either dc or ac coupled
to the input amplifier. Unlike other rms converters, the AD737
permits both direct and indirect ac coupling of the inputs. AC
coupling is provided by placing a series capacitor between the
input signal and Pin 2 (or Pin 1) for direct coupling and
between Pin 1 and ground (while driving Pin 2) for indirect
coupling.
The output of the input amplifier drives a full-wave precision
rectifier, which in turn, drives the rms core. It is in the core that
the essential rms operations of squaring, averaging and square
rooting are performed, using an external averaging capacitor,
C
AV
. Without C
AV
, the rectified input signal travels through the
core unprocessed, as is done with the average responding con-
nection (Figure 17).
A final subsection, the bias section, permits a “power down”
function. This reduces the idle current of the AD737 from 160
µA
down to a mere 30
µA.
This feature is selected by tying Pin
3 to the +V
S
terminal. In the average responding connection, all
of the averaging is carried out by an RC post filter consisting of
an 8 kΩ internal scale-factor resistor connected between Pins 6
and 8 and an external averaging capacitor, C
F
. In the rms cir-
cuit, this additional filtering stage helps reduce any output
ripple which was not removed by the averaging capacitor, C
AV
.
RMS MEASUREMENT – CHOOSING THE OPTIMUM
VALUE FOR C
AV
8k
C
C
1
AD737
FULL
WAVE
RECTIFIER
8k
RMS
CORE
8
COM
V
IN
2
INPUT
AMPLIFIER
BIAS
SECTION
7 +V
S
OUTPUT
C
F
10 F
(OPTIONAL)
POWER
3
DOWN
–V
S
4
6
V
OUT
5
C
AV
C
AV
33 F
POSITIVE SUPPLY
0.1 F
COMMON
NEGATIVE SUPPLY
0.1 F
–V
S
+V
S
Figure 16. AD737 True RMS Circuit
Since the external averaging capacitor, C
AV
, “holds” the recti-
fied input signal during rms computation, its value directly af-
fects the accuracy of the rms measurement, especially at low
frequencies. Furthermore, because the averaging capacitor ap-
pears across a diode in the rms core, the averaging time con-
stant will increase exponentially as the input signal is reduced.
This means that as the input level decreases, errors due to
nonideal averaging will
reduce
while the time it takes for the cir-
cuit to settle to the new rms level will
increase.
Therefore, lower
input levels allow the circuit to perform better (due to increased
averaging) but increase the waiting time between measure-
ments. Obviously, when selecting C
AV
, a trade-off between
computational accuracy and settling time is required.
–6–
REV. C