As part of a phosphorescence detector, [lcamtuf] has been working with photodiodes. The components, like all diodes, have some capacitance at the junction, and this can limit performance. That’s why [lcamtuf] turned to bootstrapping to make that parasitic capacitance almost disappear.
The technique appears in several Analog Devices datasheets that presents a mystery. An op amp circuit that would normally limit changes to about 52 kHz has an unusually-placed JFET and claims to boost the bandwidth to 350 kHz.
The JFET turns out to be in a voltage-follower configuration. The photodiode sees approximately the same voltage on both terminals, so the internal capacitor can’t charge and, therefore, doesn’t impose any limits on rate of change.
Of course, a better way to think of it is that tiny changes cause an immediate response to counteract them, and so the capacitor’s charging and discharging are kept to a minimum.
It really isn’t important that the capacitor is not charged, but rather that the capacitor doesn’t increase or decrease charge. This leads to a second design, which imposes a DC bias voltage on the diode but prevents any signal from causing the capacitor to change from its precharged value.
Photodiodes seem exotic, but honestly, all semiconductor diodes are photodiodes if you let the light get to them. It seems that capacitors and op amps are always at loggerheads.
A similar issue applies to charge amplifiers used for measuring piezoelectric sensors. The piezo element generates a small charge, which leaks back through the device almost instantly. The other issue is the capacitance of the cable that connects the sensor to the amplifier, which eats up the small amount of charge you’re trying to measure. As a solution, the charge amplifier is built in a similar fashion to keep the voltage across the piezo at zero (trimmed for op-amp input offset etc.) so any charge generated is removed and put into the integrating capacitor (Cf in diagram). Of course there’s still leakages, so it cannot do steady state (DC) but when well adjusted the circuit can work down to milli-Hertz without too much distortion. I once made an anti-alias filter for such a setup and found the -3 dB point at around 0.1 Hz out of the charge amp.
For photodiodes, the scheme is slightly different because there’s a threshold voltage that has to be overcome for the diode to push out current, so that’s why the Vref in the diagram and the “pre-charge” over the diode can’t be zero. It can be, but then it won’t respond very quickly and the voltage follower output has to dip below ground.
All semiconductor diodes are photodiodes if you shine enough light on them, and all diodes are LED’s if you give them enough current…
Surely you mean incandescent light bulbs.