REV. E

–5–

AD637

OPTIONAL TRIMS FOR HIGH ACCURACY

The AD637 includes provisions to allow the user to trim out

both output offset and scale factor errors. These trims will result

in significant reduction in the maximum total error as shown in

Figure 4. This remaining error is due to a nontrimmable input

offset in the absolute value circuit and the irreducible non-

linearity of the device.

The trimming procedure on the AD637 is as follows:

put from Pin 9. Alternatively R1 can be adjusted to give the

or a calibrated ac signal, trim R3 to give the correct output at

Pin 9, i.e., 1 V dc should give l.000 V dc output. Of course, a

2 V peak-to-peak sine wave should give 0.707 V dc output.

Remaining errors are due to the nonlinearity.

INPUT LEVEL – Volts

5.0

2.5

5.0

0

2.0

0.5

1.0

0

2.5

1.5

AD637K MAX

INTERNAL TRIM

AD637K

EXTERNAL TRIM

AD637K: 0.5mV

0.2%

0.25mV

0.05%

EXTERNAL

Figure 4. Max Total Error vs. Input Level AD637K

Internal and External Trims

BUFFER

AD637

SQUARER/DIVIDER

BIAS

SECTION

FILTER

25k

25k

1

2

3

4

5

6

7

14

13

12

11

10

9

8

V rms

OUT

R4

147

+

R3

1k

SCALE FACTOR ADJUST,

2%

R2

1M

R1

50k

OUTPUT

OFFSET

ADJUST

ABSOLUTE

VALUE

Figure 5. Optional External Gain and Offset Trims

CHOOSING THE AVERAGING TIME CONSTANT

The AD637 will compute the true rms value of both dc and ac

input signals. At dc the output will track the absolute value of

the input exactly; with ac signals the AD637’s output will ap-

proach 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/µF of averaging capacitance). As shown in

Figure 6, the averaging error is defined as the peak value of the

ac component, ripple, plus the value of the dc error.

The peak value of the ac ripple component of the averaging er-

ror is defined approximately by the relationship:

50

6.3

τf in % of reading where (t > 1/f)

DC ERROR = AVERAGE OF OUTPUT–IDEAL

DOUBLE-FREQUENCY

RIPPLE

TIME

AVERAGE ERROR

IDEAL

Figure 6. 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 net-

work 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:

1

0.16

in % of reading

time that the rms converter “holds” the input signal during

computation, the magnitude of the dc error is determined only

SINEWAVE INPUT FREQUENCY – Hz

100

0.1

1.0

10

10k

1k

100

10

DC ERROR

PEAK RIPPLE

Figure 7. Comparison of Percent DC Error to the Percent

Peak Ripple over Frequency Using the AD637 in the Stan-

dard RMS Connection with a 1

× µF C

AV

The ac ripple component of averaging error can be greatly

reduced by increasing the value of the averaging capacitor.

There are two major disadvantages to this: first, the value of the

averaging capacitor will become extremely large and second, the

settling time of the AD637 increases in direct proportion to the

value of the averaging capacitor (Ts = 115 ms/

µF of averaging

capacitance). A preferable method of reducing the ripple is

through the use of the post filter network, shown in Figure 8.

This network can be used in either a one or two pole configura-

tion. For most applications the single pole filter will give the

best overall compromise between ripple and settling time.