Datasheet
Data Sheet AD636
Rev. E | Page 11 of 16
slightly more restricted than in the dual supply connection. The
load resistor, R
L
, is necessary to provide current sinking capability.
C2
3.3µF
AD636
ABSOLUTE
V
ALUE
SQUARER
DIVIDER
10kΩ
10kΩ
CURRENT
MIRROR
–
C
A
V
BUF
+
20kΩ
NONPOLARIZED
39kΩ
0.1µF
0.1µF
+V
S
V
OUT
R
L
1kΩ
T
O 10kΩ
V
IN
00787-007
1
2
3
4
5
6
7
14
13
12
11
10
9
8
V
IN
NC
–V
S
C
A
V
dB
BUF OUT
BUF IN
NC
NC
NC
COM
R
L
I
OUT
NC = NO CONNECT
+
–
Figure 11. Single-Supply Connection (See Text)
CHOOSING THE AVERAGING TIME CONSTANT
The AD636 computes the rms of both ac and dc signals. If the
input is a slowly varying dc voltage, the output of the AD636
tracks the input exactly. At higher frequencies, the average
output of the AD636 approaches the rms value of the input
signal. The actual output of the AD636 differs from the ideal
output by a dc (or average) error and some amount of ripple, as
demonstrated in Figure 12.
TIME
IDEAL
E
O
DC ERROR = E
O
– E
O
(IDEAL)
AVERAGE E
O
= E
O
DOUBLE-FREQUENCY
RIPPLE
E
O
00787-008
Figure 12. Typical Output Waveform for Sinusoidal Input
The dc error is dependent on the input signal frequency and the
value of C
AV
. Figure 13 can be used to determine the minimum
value of C
AV,
which yields a given % dc error above a given
frequency using the standard rms connection.
The ac component of the output signal is the ripple. There are
two ways to reduce the ripple. The first method involves using a
large value of C
AV
. Because the ripple is inversely proportional
to C
AV
, a tenfold increase in this capacitance effects a tenfold
reduction in ripple. When measuring waveforms with high crest
factors (such as low duty cycle pulse trains), the averaging time
constant should be at least ten times the signal period. For example,
a 100 Hz pulse rate requires a 100 ms time constant, which
corresponds to a 4 μF capacitor (time constant = 25 ms per μF).
INPUT FREQUENCY (Hz)
100
0.01
1
10
0.1
1
10
100
0.1
0.01
0.01% ERROR
0.1% ERROR
*% dc ERROR + % RIPPLE (PEAK)
1% ERROR
FOR 1% SETTLING TIME IN SECONDS
MULTIPLY READING BY 0.115
REQUIRED C
AV
(µF)
1 10 100 1k 10k 100k
VALUES FOR C
AV
AND
1% SETTLING TIME FOR
STATED % OF READING
AVERAGING ERROR*
ACCURACY ±20% DUE TO
COMPONENT TOLERANCE
10% ERROR
00787-009
Figure 13. Error/Settling Time Graph for Use with the Standard RMS
Connection
The primary disadvantage in using a large C
AV
to remove ripple
is that the settling time for a step change in input level is
increased proportionately. Figure 13 shows the relationship
between C
AV
and 1% settling time is 115 ms for each microfarad
of C
AV
. The settling time is twice as great for decreasing signals
as for increasing signals (the values in Figure 13 are for decreasing
signals). Settling time also increases for low signal levels, as
shown in Figure 14.
rms
INP
UT LEVEL
10
.0
7.5
0
10m
V 100mV
1
.0
5.0
2.5
1V1m
V
SETTLING TIME RELATIVE TO
SETTLING TIME @ 200mV rms
00787-010
Figure 14. Settling Time vs. Input Level
A better method for reducing output ripple is the use of a post-
filter. Figure 15 shows a suggested circuit. If a single-pole filter
is used (C3 removed, R
X
shorted), and C2 is approximately
5 times the value of C
AV
, the ripple is reduced, as shown in
Figure 16, and the settling time is increased. For example, with
C
AV
= 1 µF and C2 = 4.7 μF, the ripple for a 60 Hz input is
reduced from 10% of reading to approximately 0.3% of reading.
The settling time, however, is increased by approximately a
factor of 3. The values of C
AV
and C2 can therefore be reduced
to permit faster settling times while still providing substantial
ripple reduction.