Optical feedback to reduce noise in precision zeners

[5065AGuru]

Have enjoyed seeing the pictures showing the light emitted from a precision zener like the LTZ1000 or the ADL1000.
Wondered if you sensed and scaled that optical signal and fed it back into an op amp on the zeners output if you could cancel out any noise?
This would depend on the optical signal actually being correlated with the devices output noise.
Does the LM399H family also show that optical signal? Cheaper to experiment with with first!

Cheers,

Corby

[Conrad Hoffman]

I suspect it would be tough to detect the signal and not add a huge amount of noise, enough to swamp out the effect you're looking for. I used to use optical devices for position sensing and getting the noise down wasn't trivial. Phototransistors were terrible. Photodiodes were OK, but just barely good enough for my application.

[ArdWar]

Dark current alone would be PITA to cancel out reliably. Most light detectors are also excellent IR detectors, and one thing about ovenized zeners is that they're hot.

[magic]

Is it practical at all, even in theory, with noiseless light detector and so on?

Say you have a DC+AC signal which is correlated with noise, what can you do with it?
Subtract the DC offset and add the rest to the output?
How would you do it, using a separate reference? What about its drift and noise?

RC filter? Might as well put it on the normal output of the reference chip...


In practice...
Dark current of photodiodes increases with temperature - obvious source of drift.
The light output of zeners is weak, so the signal you are looking for (if it even exist) may be swamped by noise.

[Kleinstein]

There can be an indirect link between the light emissions and the popcorn noise. However this is ideirect and hardly usable to reduce the noise: with a higher votlage there are some hot electron produced and these can excite states at the oxide interface. These higher energy states can than cause light emissions or if there are long lived ones at sensitive positions they can effect the electrical properties and produce random telegraph type noise. Howeve the amount of light shows that there is way more light than extra noise. So there is just a good chance to have a similar mechanism and parts that produce a lot of light may also be noisy. Still the light is likely not directly correlated with noise and thus no "correction" possible. Even if there is a small effect chances are one would add more noise than remove.