Sample and hold acquisition time

I did a quick test on a home made sample and hold circuit to figure out how fast it settles (reaches a stable state).

The circuit consists of a 10nF capacitor and an opamp buffer, similar to what you will find in the Prophet 5 and the Jupiter 8.

The prophet 5 service manual says that they let the sample and hold settle for 30 to 40 uS. I connected the circuit to a function generator's square wave output and the output to a oscilloscope. Finding the maximum frequency was then only a matter of observing when the output no longer reached its peak within the time before the square wave changed polarity.

With a 10V p-p input, the s&h managed about 10kHz on the function generator. Since the s&h changes on both high and low parts of the wave, the s&h frequency is in fact 20kHz, meaning that the acquisition time is 50uS.

When changing the input to 5V p-p I was able, unsurprisingly, to reach 20kHz on the function generator, which means an acquisition time of 25uS.

Vertical resolution: 2V/square, Horizontal resolution: 20uS/square. Squares are input, slanted ones are output.

I am aiming for a 32 channel controller, which means that with a 40kHz S&H I get a maximum refresh rate per channel of 1250Hz. It remains to be seen if this is fast enough for pitch bends etc.

Old school din-based midi runs at 31.25 kbits/s. Each 14 bit pitch bend takes up 16 bits, which means that the maximum midi pitch bend rate is 1953Hz. Perhaps 1250Hz is enough, if not, I have to use a dual channel dac with two 16channel s&h, and update two and two at the same time.