AD781

REV. A

–7–

common pin should also be connected to the digital ground,

which is usually tied to analog common at the A-to-D converter.

Figure 4 illustrates the recommended decoupling and grounding

practice.

NOISE CHARACTERISTICS

Designers of data conversion circuits must also consider the

effect of noise sources on the accuracy of the data acquisition

system. A sample-and-hold amplifier that precedes the A-to-D

converter introduces some noise and represents another source

of uncertainty in the conversion process. The noise from the

AD781 is specified as the total output noise, which includes

both the sampled wideband noise of the SHA in addition to the

band limited output noise. The total output noise is the rms

sum of the sampled dc uncertainty and the hold mode noise. A

plot of the total output noise vs. the equivalent input bandwidth

of the converter being used is given in Figure 5.

300

0

1k

10M

200

100

10k

1M

100k

FREQUENCY – Hz

Figure 5. RMS Noise vs. Input Bandwidth of ADC

DRIVING THE ANALOG INPUTS

For best performance, it is important to drive the AD781 analog

input from a low impedance signal source. This enhances the

sampling accuracy by minimizing the analog and digital

crosstalk. Signals which come from higher impedance sources

(e.g., over 5 k

Ω) will have a relatively higher level of crosstalk.

For applications where signals have high source impedance, an

operational amplifier buffer in front of the AD781 is required.

The AD711 (precision BiFET op amp) is recommended for

these applications.

HIGH FREQUENCY SAMPLING

Aperture jitter and distortion are the primary factors which limit

frequency domain performance of a sample-and-hold amplifier.

Aperture jitter modulates the phase of the hold command and

produces an effective noise on the sampled analog input. The

magnitude of the jitter induced noise is directly related to the

frequency of the input signal.

A graph showing the magnitude of the jitter induced error vs.

frequency of the input signal is given in Figure 6.

The accuracy in sampling high frequency signals is also con-

strained by the distortion and noise created by the sample-and

hold. The level of distortion increases with frequency and re-

duces the “effective number of bits” of the conversion.

Measurements of Figures 7 and 8 were made using a 14-bit A/D

100 kSPS.

1%

1k

1M

0.1%

0.01%

10k

100k

FREQUENCY – Hz

APERTURE JITTER TYPICAL AT 50ps

1/2 BIT @

8 BITS

1/2 BIT @

10 BITS

1/2 BIT @

12 BITS

1/2 BIT @

14 BITS

Figure 6. Error Magnitude vs. Frequency

–65

–95

1M

–80

–90

1k

–85

100

–70

–75

100k

10k

FREQUENCY – Hz

Figure 7. Total Harmonic Distortion vs. Frequency

90

0

100k

20

10

1k

100

30

40

50

60

70

80

10k

FREQUENCY – Hz

Figure 8. Signal/(Noise and Distortion) vs. Frequency