Photovoltaic isolator

The size of the cells required depends entirely on the intensity of the light hitting them. At "1 sun" illumination you get close to 40mA per cm2 area, but then you need probably 6 cells that size to get 3V output. The level of illumination producing an output power equivalent to what you get with sunlight might be quite a bit less than the power density of sunlight (which is about 1kW/m2), if you choose the wavelength that is used most efficiently by the PV cell.

The efficiency of ordinary silicon PV cells is pretty much independent of light level up to a couple of suns at least. By keeping the light concentrated (e.g. acrylic light pipe) you could use a very small cell, or with exactly the same light source you could let the light spread out and then you would need a bigger PV cell to collect the same amount of light.

Taken to an extreme, if money were no object then you could use a 3W laser diode coupled to an optical fibre (e.g. Osram SPL 2F81-2S) as your illumination source. Even a very small cell could be used to convert the light back to electricity, if you can find a cell that can handle such concentrated illumination. If you wanted to use ordinary "1-sun" solar cells then you would instead need to let the output of the optical fibre diverge until the intensity is in the range of a few suns or less.

It is possible to achieve somewhat higher efficiencies at more than 1 sun of illumination, if the cells are designed for that (e.g. with beefed up metal contacts for less resistance, and other optimisations). If the cells designed for 1-sun use have a lot of internal series resistance then this will drop too much voltage at high current densities (many suns of illumination). Some cells are designed for operation at tens or hundreds or even thousands of suns of concentrated illumination but they are probably hard to obtain.

At low intensities you could probably use BPW34 photodiodes as solar cells. I am not sure how much series resistance they have so you would need to check that to see what is the sensible maximum current you can use them at. Into a short-circuit they produce about 2.8mA under sunlight, but I have not tried to use them near the maximum power point, and the series resistance could limit the power available. The advantage of these photodiodes over solar cells is the small size and easy PCB mounting.

As others pointed out already, IR LEDs will be best. e.g. SFH4725S

If you use more than one PV cell in series so that you can get your 3V output, then the current available will be limited to that produced by the weakest (least-illuminated) cell. Therefore you might have to either over-illuminate them all a bit, or put some attention into the optical arrangement so that you will be illuminating the PV cells equally.

Also: If you want to simplify the PV cell part so that you don't need many PV cells in series but could just use a single PV cell, then you could modulate the IR LED, and shine that onto a single PV cell generating a modulated current at up to about 0.5V. You can then feed the PV cell output directly into a step-up transformer to whatever rail voltages you require. You could also have two IR LEDs and two solar cells connected in anti-parallel, if you wanted a push-pull current waveform driving the transformer primary.

I'd be very interested to see how well that one does.

Yeah, the about-half-a-volt per photovoltaic diode is a killer. Probably best to use a center-tapped transformer driven by two PV solar cells, and feed constant current through two IRLEDs in series, shorting one at a time. That way the output can be nearly constant, AND the light-source drain is also nearly constant.

Hopefully, the PV cells are not so capacitive that the frequency has to be kept low.

I vaguely recall polycrystalline cells being very slow. Someone might correct me though.

NT

Good point. I think the high efficiency cells like Sunpower have quite a bit of capacitance. If one wanted to use modulated IR and a transformer after the PV cells, that would be one reason to prefer an optical system that can apply concentrated light to the PV cells - then the PV cells could be small and therefore would have low capacitance.

Bearing in mind the issue of capacitance, perhaps 6 cells in series is a better idea.

The problem, then, is balancing the six cells in series so one of 'em isn't a weak link. With square cells, you could aim at a four-corners spot and infer balance from symmetry.

If efficiency is not an important consideration then it doesn't matter if some of the cells are over-illuminated, as long as the least-illuminated one receives enough light to supply the output current required. Therefore the problem of non-uniform illumination could be solved by increasing the current through the LED, or adding more LEDs.

Thanks, all, there's been some useful discussion here. If it proceeds, I'll post results.

Cheers