Photodiode Capacitance versus Reverse Voltage

The capacitance curve gets very, but not entirely, flat as voltage increases. The equation can be made to do that.

I sometimes fiddle a diode's Spice parameters to get a C-V curve that looks like the data sheet. That avoids a lot of hard thinking.

Photodiode people are weird. They specify an abs-max voltage (2V in your case) then give curves that go way beyond that. I use one part that's spec'd abs max 12 volts but characterized at 75.

Photodiodes are usually run at constant voltage, so you may not need to model the CV curve. Just run it at 2 volts or something. Or 4.

More voltage makes PD's faster, even when capacitance has leveled off.

We'll have to wait for Phil Hobbs for a definitive response, but it's probably worth noting that the real capacitance will probably include some voltage insensitive contribution from the stray electric fields outside the silicon.

You might have to decapsulate one to see how it was put together to get some idea about that.

It's likely to be small, but it can mess up curve fitting.

Trust the datasheet. The capacitance variation depends on the depletion region of the diode, and a photodiode typically has a P-I-N structure, with P doping, intrinsic, and N doped layers. So, after the first few volts of bias, the intrinsic region (a fairly large volume, because that's how you get quantum efficiency in a photodetector) is the depletion region, and subsequent bias only slowly moves the region boundary through the more heavily doped P and N parts.

The equation that uses "K" and explains that it depends on the doping, is correct for a variety of diodes, but the special doping profiles of PIN photodiodes (and, for that matter, voltage-variable-capacitor diodes) are not well characterized by a single constant, K.

In theory it goes as 1/sqrt(V), this good.

I'm not sure what the 'built-in' voltage is.

George H.

The C(V) curve depends entirely on the doping profile. The physics is that with a doping density rho, depleting an additional dz of material produces a sheet of charge

sigma = rho dz.

That sheet increases the E field in the depletion zone in the usual way for sheet charges. Avalanche photodiodes have a highly doped buried layer that creates an isolated region of high E field. That way you get good multiplication without a lot of leakage and surface state crap.

A PIN diode with a fairly narrow intrinsic layer and highly-doped (low resistivity) substrate and epi will behave like yours. These are usually blue-enhanced devices, i.e. with a thin epi and an AR coating centred in the blue.

Diodes like that have higher capacitance than those built on high-resistivity substrates, but are a bit faster since the transit time is less.

Then you have to worry about diffusion delays in the epi layer, since the current has to flow across the diode to get to the metal contact. Since diffusion gets quadratically slow with distance, both RC and diffusion delays increase like the area, which makes them harder to tell apart. The shape of the rise is one clue--when diffusion dominates, the rise is initially fast but gets much slower after about 50%, like a long cable.

There are a lot of subtleties in photodiodes.

Cheers

Phil Hobbs