What is charge injection error in sampling switches, and how can it be mitigated?

• A. Charge injection occurs when switching devices inject charge into the sampling capacitor during transition, causing offset; mitigations include dummy switches, bottom-plate sampling, and careful timing/layout.
• B. Charge injection is caused by thermal noise in the capacitor; mitigations include cooling.
• C. Charge injection only happens in digital switches; mitigations are irrelevant.
• D. Charge injection can be eliminated completely by using larger capacitance.

Answer: (A)

Charge injection error happens when the sampling switch changes state and the charge stored in the switch and its parasitic capacitances is dumped into the sampling capacitor. When the transistor turns off, the charge that's been held in the channel and in drain/gate/body capacitances has to go somewhere, and a portion ends up on the hold capacitor. That added charge produces a voltage error on the stored sample, roughly Q_injected divided by the value of the hold capacitor. The smaller the capacitor or the larger the injected charge, the bigger the error, and fast, abrupt switching tends to increase the injection.

To reduce this error, designers use techniques that balance or cancel the injection and minimize how much charge ends up on the hold node. A dummy switch mirrors the real switch so the injected charge largely cancels, lowering the net disturbance. Bottom-plate sampling keeps the bottom plate of the capacitor at a controlled potential during the switching event, which decouples the hold node from the injection. Careful timing and layout—such as matching devices, shielding parasitics, and routing switches symmetrically—also helps minimize the stray charges that reach the hold capacitor. Other approaches include using complementary or bootstrapped switches and, when feasible, increasing the hold capacitor size to reduce the relative impact of a given amount of injected charge.