
AD637
Rev. G | Page 10 of 20
INPUT LEVEL (V)
5.0
2.5
5.0
0
2.0
0.5
ERRO
R
(
mV)
1.0
0
2.5
1.5
AD637K MAX
INTERNAL TRIM
AD637K
EXTERNAL TRIM
AD637K: 0.5mV ± 0.2%
0.25mV ± 0.05%
EXTERNAL
00788-007
Figure 7. Maximum Total Error vs.
Input Level AD637K Internal and External Trims
1
25k
DEN
IN
BUF IN
BUFFER
OUT
SQUARER/
DIVIDER
–VS
CS
DB OUT
3COM
BIAS
2NC
4
OUT
OFF
R2
1M
5
25k
6
7
8
9
10
+VS
11
12
NC
13
14
NC
VIN
VOUT =VIN2
VIN
CAV
+
AD637
ABSOLUTE
VALUE
00788-008
–VS
+VS
CAV
R4
147
SCALE FACTOR TRIM
R1
50k
OUTPUT
OFFSET
TRIM
4.7k
–VS
+VS
R3
1k
Figure 8. Optional External Gain and Offset Trims
CHOOSING THE AVERAGING TIME CONSTANT
The AD637 computes the true rms value of both dc and ac
input signals. At dc, the output tracks the absolute value of the
input exactly; with ac signals, the AD637’s output approaches
the true rms value of the input. The deviation from the ideal
rms value is due to an averaging error. The averaging error is
comprised of an ac and dc component. Both components are
functions of input signal frequency f and the averaging time
constant τ (τ: 25 ms/F of averaging capacitance).
Figure 9shows that the averaging error is defined as the peak value of
the ac component, ripple, and the value of the dc error.
The peak value of the ac ripple component of the averaging
error is defined approximately by the relationship
()f
1
τ
τf
6.3
50
>
where
reading
of
%
in
DC ERROR = AVERAGE OF OUTPUT – IDEAL
DOUBLE-FREQUENCY
RIPPLE
EO
TIME
AVERAGE ERROR
IDEAL
EO
00788-
009
Figure 9. Typical Output Waveform for a Sinusoidal Input
This ripple can add a significant amount of uncertainty to the
accuracy of the measurement being made. The uncertainty can
be significantly reduced through the use of a post filtering
network or by increasing the value of the averaging capacitor.
The dc error appears as a frequency dependent offset at the
output of the AD637 and follows the equation
reading
of
%
in
f
2
4
6
16
0
1
τ
+ .
.
Since the averaging time constant, set by CAV, directly sets the
time that the rms converter holds the input signal during
computation, the magnitude of the dc error is determined only
by CAV and is not affected by post filtering.
SINEWAVE INPUT FREQUENCY (Hz)
100
0.1
1.0
10
10k
DC
ERRO
R
O
R
RI
PPL
E
%
O
F
READI
NG
1k
100
10
DC ERROR
PEAK RIPPLE
00788-010
Figure 10. Comparison of Percent DC Error to the
Percent Peak Ripple over Frequency Using the
AD637 in the Standard RMS Connection with a 1 × F CAV
The ac ripple component of averaging error is greatly reduced
by increasing the value of the averaging capacitor. There are two
major disadvantages to this: the value of the averaging capacitor
becomes extremely large and the settling time of the AD637
increases in direct proportion to the value of the averaging
capacitor (Ts = 115 ms/F of averaging capacitance). A
preferable method of reducing the ripple is through the use of
the post filter network, as shown in
Figure 11. This network can
be used in either a one-pole or two-pole configuration. For
most applications, the single pole filter gives the best overall
compromise between ripple and settling time.