Photovoltaic Capabilities of LEDs

One of the follow-up questions on optocoupler bias generation was if optical LEDs would be a decent candidate for generating a higher voltage output without needing to handle the 8+ volt output from the isolated gate driver photocouplers. The short answer is no, probably not even in comparison to stacking multiple 0.5 V output optocouplers.

Energized green SMD LED driving another unit of the same part number 

Unfortunately it seems that LEDs are too efficient these days with high output intensity from extremely small dice. Hence how indicator and even some low-end illumination LEDs can be mere cents. I started with the LTST-C191KGKT commonly used on designs at the office.

LiteOn LTST-C191KGKT 10 mA drive for 760 nA photocurrent

Even the fairly bright 10 mA drive doesn't elicit much response. Certainly not enough output current to scale up to a usable level (10s of uA) with additional drive current no matter how nice the output voltage might be.

Close-up on the SMD LED assembly and measurement fixturing

In addition, these were vertical-mount LEDs so bodging them horizontal onto 0.1" pitch protoboard was a difficult effort. It's very easy to destroy the lens bond to the package taking the bond wire or even the die with it. There's also not a good process to achieve automated assembly capable footprints. That would probably require lining up the devices in the correct polarity and somehow gluing their lenses together before reflow.

Maybe going bigger would help? There were some 3mm through-hole LEDs also sitting around the lab.

Energized blue TH LED driving another unit of the same part number

While more successful, the achievable output of this device fails to meet the bar within absolute maximum input drive while also being incredibly mechanically sensitive. Building the fixture to hold the LEDs in the right alignment would easily cost more than the optocouplers necessary even discounting that TH is more expensive than SMD.

Kingbright WP710A10LVBC/D 10 mA drive for 1.9 uA photocurrent

So there might be some more combinations of components to test, but the die size of the receiver will be key for achieving the target level of 10-50 uA photocurrent. I was able to beat that for 100 uA with some white LEDs. Those cost 10s of cents each and would be operating at a drive level requiring PCB thermal management on the emitter, though!

If you have a promising candidate, please reach out so I can consider adding it to the characterization.