Measure Analog Switch Charge Injection in the Actual Signal Network

A short pulse at an analog switch output can be easy to miss on a slow display and still disturb a sampled measurement. Start by identifying when the pulse occurs relative to the switch command. Then separate the initial disturbance from the time required for the connected network to recover.

Relate charge to the affected node

Texas Instruments’ discussion of multiplexer charge injection describes unwanted charge associated with switching and the importance of the surrounding source impedance. A datasheet charge value is useful, but its test conditions need to be compared with the real circuit.

As an illustrative first estimate, 1 pC deposited onto an isolated 100 pF node corresponds to a 10 mV change using voltage change equal to charge divided by capacitance. This estimate omits the source, load and time-dependent switching behavior. Use it to identify a potentially significant disturbance, not to predict the entire measured waveform.

Map the operating cases

List signal level, source impedance, load capacitance, switch state before the event and control transition. Include channels whose source impedance changes with operating mode. A buffered voltage source and a high-impedance sensor can produce different recovery even when connected through the same switch.

Determine when the downstream circuit uses the signal. For a sampled system, mark the acquisition interval on the timing diagram. For an analog control path, identify whether the transient reaches a sensitive input directly. The relevant acceptance criterion is the system’s allowed error during its actual observation window.

Measure without hiding the event

Trigger from the switch-control transition and capture both the signal node and the command. Document probe loading, bandwidth and acquisition settings. Additional probe capacitance can change a small capacitive node enough to alter the event being measured.

Repeat the capture with representative signal levels and source conditions. Compare the pulse amplitude, polarity and recovery rather than retaining only the most attractive waveform. If the disturbance appears before or after the expected switching event, investigate coupling and timing paths instead of attributing everything to charge injection.

Evaluate the tradeoff of a change

Adding capacitance may reduce an initial voltage step while extending settling. A lower-impedance driver, different switch or changed sampling time can also affect other requirements. Evaluate each candidate against the same timing and error criteria, including the required operating range.

Record the selected solution with measured conditions and the point at which the signal becomes usable. Avoid claiming that a smaller transient alone proves better system accuracy. The practical result is a signal network that meets its error requirement when observed, with enough documentation to reproduce that conclusion after a component or timing change.