9.1.4 Sampling Dynamic Signals

Sample error to moving input signals probably causes more confusion among sample-and-hold users than any other parameter. The primary reason for this is that many users make the assumption that the sample and hold amplifier is truly locked on to the input signal while in the sample mode. In actuality, there are finite phase delays through the circuit creating an input-output differential for fast moving signals. In addition, although the output may have settled, the hold capacitor has an additional lag due to the 300-Ω series resistor on the chip. This means that at the moment the hold command arrives, the hold capacitor voltage may be somewhat different than the actual analog input. The effect of these delays is opposite to the effect created by delays in the logic which switches the circuit from sample to hold. For example, consider an analog input of 20 Vp–p at 10 kHz. Maximum dV/dt is 0.6 V/µs. With no analog phase delay and 100-ns logic delay, one could expect up to (0.1 µs) (0.6V/µs) = 60 mVerror if the hold signal arrived near maximum dV/dt of the input. A positive-going input would give a
60-mV error. Now assume a 1-MHz (3-dB) bandwidth for the overall analog loop. This generates a phase delay of 160 ns. If the hold capacitor sees this exact delay, then error due to analog delay will be (0.16 µs) (0.6 V/µs) = –96 mV. Total output error is 60 mV (digital) –96 mV (analog) for a total of –36 mV. To add to the confusion, analog delay is proportioned to hold capacitor value while digital delay remains constant. A family of curves (dynamic sampling error) is included to help estimate errors.

Figure 1 has been included for sampling conditions where the input is steady during the sampling period, but may experience a sudden change nearly coincident with the hold command. This curve is based on a 1-mV error fed into the output.

Figure 6 indicates the time required for the output to settle to 1 mV after the hold command.