Sample and Hold Circuit Design

A sample and hold circuit captures an input voltage and holds it steady while a converter or a multiplexed system uses it. It is the answer to a source that cannot settle in the time available, and its own errors are a different set from the converter it serves.

Why It Exists

A converter that has to sample several channels in sequence has little time for each. The source may be a sensor with a high impedance that cannot settle in that window, and the sample and hold separates the acquisition from the conversion.

The circuit also removes the dependence of the reading on the source impedance during the conversion. The converter sees the hold capacitor, which is a low impedance and a stable value, for the whole conversion.

The third reason is to capture a value at an instant. A measurement of a waveform that changes quickly needs the sample to be simultaneous with the reference event, and the hold preserves it while the conversion proceeds.

Acquisition time

acquisition time is the period during which the switch is closed and the hold capacitor charges to the input voltage. It is set by the source impedance, the switch resistance and the capacitor value.

The settling is exponential, and the number of time constants needed depends on the resolution. Ten time constants give an error of about fifty parts per million, which is adequate for a twelve bit converter and not for a sixteen bit one.

A buffer before the switch reduces the source impedance and shortens the acquisition time. The buffer’s own settling then becomes the limit, and it is specified for a full scale step in the same way.

droop rate and Hold Time

When the switch opens, the hold capacitor starts to discharge through the leakage of the switch, the amplifier input and the capacitor itself. The droop rate is the change in voltage per unit time, and it sets the maximum hold time.

A droop of a microvolt per microsecond gives an error of a milli volt after a second, which is small for a twelve bit system and significant for a sixteen bit one. The leakage of the capacitor and the cleanliness of the board both matter.

Where the hold time is long, a buffer with a low bias current is used and the board is kept clean. Surface contamination adds leakage that appears as droop and as a reading which changes when the humidity changes.

aperture jitter and Timing

aperture jitter is the uncertainty in the instant at which the switch opens. In a system measuring a changing waveform it produces an error proportional to the rate of change of the signal.

A jitter of ten picoseconds on a signal changing at a volt per microsecond gives an error of ten microvolts. On a fast changing signal the same jitter gives a much larger error, which is why the specification is quoted in time.

The switch drive has to be fast and clean to keep the jitter small. A slow drive makes the opening time depend on the signal level, which turns the jitter into a distortion rather than a noise.

charge injection and Switching Artifacts

When the switch opens it injects a small charge into the hold capacitor, which appears as a step in the held voltage. The charge is roughly constant, so the step is a fixed offset rather than a gain error.

charge injection varies with the signal level in a real switch, because the gate voltage relative to the channel changes. The variation is what turns it into a non linearity, and a differential arrangement with two matched switches cancels most of it.

A dummy switch driven from the inverse of the control signal can cancel the injection by adding an opposite charge. The technique depends on matching and it reduces the error by an order of magnitude rather than removing it.

Switch Technologies

A junction field effect transistor has a low leakage and a moderate on resistance, and it is used where the hold time is long. Its drive has to be arranged so that the gate does not conduct in either state.

A complementary metal oxide switch integrates well and has a low charge injection, and its leakage rises with temperature. The on resistance varies with the signal, which affects the acquisition time.

An integrated sample and hold amplifier includes the switch, the capacitor and the buffers with the timing arranged internally. It is easier to use and its performance is specified as a whole rather than as a set of parts.

Layout and Guarding

The hold capacitor has to be a low leakage type, and a polystyrene or a polypropylene part is used where the droop matters. A ceramic capacitor with a high dielectric constant has a leakage and a voltage coefficient that make it unsuitable.

Guard the hold node with a track at the same potential. The node is high impedance when the switch is open, and a guard reduces the effect of surface leakage and of the electric field around it.

Keep the switch drive away from the hold node. The drive signal is a fast edge and it couples capacitively into the node, which appears as a step in the held value. The layout measures that keep such a node clean are described in our guide to mixed signal board design.

Typical Applications

A data acquisition system with eight channels multiplexed into one converter is the classic application. The sample and hold gives every channel the same acquisition time regardless of how long the conversion takes.

Capturing a waveform for a digitizer is the second case. Several sample and hold circuits triggered together capture a set of points at the same instant, which is what a multi channel measurement of a changing signal requires.

A transient recorder uses the same principle with a much longer hold. The value is captured at the event and read out afterwards, and the hold time is measured in seconds rather than in microseconds.

Verification and Faults

Verify the acquisition time by applying a full scale step and observing when the held value reaches its final value within the required error. That measurement is the direct answer and it accounts for everything in the path.

Measure the droop by holding a value for the longest time the system uses and observing the change. It is a slow measurement and it is the only way to confirm the leakage of the assembled board.

A held value that changes when the board is touched is a leakage or a guard problem. The release checks that keep such a circuit consistent are collected in our PCB design release checklist, the assembly points in judging PCB quality, and the practices that keep a surface clean in our guide to conformal coating board protection.

FAQ

When do I need a sample and hold? When the source cannot settle in the converter sampling window, or when several channels must be captured at the same instant.

What limits the hold time? The droop rate, which comes from the leakage of the switch, the amplifier and the board surface.

What capacitor should I use? A low leakage film type. A high dielectric constant ceramic has too much leakage and voltage coefficient.