Testing Dielectric Absorption as Memory in a Hold Capacitor

A sample-and-hold circuit may produce different readings for the same input depending on the previous sample. Acquisition time, switch charge injection, and leakage are familiar causes, but the hold capacitor itself can retain a history-dependent contribution. Dielectric absorption describes charge that reappears after a capacitor has been charged and then discharged.

Use voltage history as the test variable

Analog Devices discusses this mechanism in MT-090, Sample-and-Hold Amplifiers. For a board experiment, hold the desired test voltage constant while changing only the preceding voltage and its duration. Compare a long positive precharge, a long negative precharge, and a near-zero precharge before acquiring the same final input. A polarity-dependent residual is more informative than a single droop trace.

For example, an instrumentation channel may alternate between a large calibration voltage and a small sensor voltage. Characterize that exact transition as well as two successive small measurements. If the latter are clean but the former have a persistent tail, increasing the ADC delay alone may hide a memory problem rather than explain it. Repeat the sequence with the switch timing unchanged and a candidate capacitor of equal nominal value.

Separate absorption from other errors

Leakage usually produces a continuing voltage change during hold. Charge injection is associated with the switching event. Incomplete acquisition depends strongly on source resistance and available track time. Absorption depends on the capacitor’s preceding electrical history and may involve several time scales. These signatures can coexist, so do not fit every transient with one RC constant.

• Keep probe loading consistent; a probe can discharge the node being evaluated.
• Record the precharge level, precharge duration, discharge duration, and observation interval.
• Repeat with reversed polarity while staying within all component limits.
• Compare candidate dielectrics on the same guarded, clean fixture.
• Allow thermal conditions to stabilize before comparing small residuals.

Budget the effect at the sampling instant

The relevant error is the voltage present when the measurement is accepted, not the largest recovery voltage at an arbitrary later time. A larger capacitor can reduce charge-injection steps and leakage droop, yet require more acquisition current and time. Changing dielectric, reducing the preceding voltage excursion, or revising the channel sequence may solve different parts of the problem.

Combine the evidence with hold droop and acquisition budgeting and multiplexed channel settling tests. Document the worst preceding state in the acceptance procedure; testing only repeated identical samples leaves the history dependence untested.