low-cost sensitive silicon photodiodes.

Win,

How low is low? Do you need PIN or is PN okay?

You can often come out ahead by tiling with solar cells--the dark noise goes up as the area while the signal goes as area**2, and you can't beat the price. Depends on the BW, of course.

Cheers,

Phil Hobbs

These are a bit lower cost than $50: $28 ea. from Digikey:

$14.34 ea.

These are better (lower dark current), but they just raised the price:

If you need just a few, not a reliable source, you might take a look at the surplus stuff at Electronic Goldmine, like:

Don't forget that as the area increases, the junction capacitance also goes up making the devices a little tricky to keep stable. Burr Brown (TI) has some application notes on this -- why not try one of their amplifer incorporated diodes? I have also used the Texas Advanced Optical "light to frequency" converters -- these obviate the need of an ADC -- just count the pulse width. (It helps to shape the pulse a bit, however.)

Silonex do large-area planar photodiode elements.

Or a collimating lens onto an ordinary photodiode.

Large-area diodes will never be cheap, as they are relatively specialist, and have an inherently large silicon area and hence manufacturing cost. A lens plus a smaller diode will probably be cheaper if your application allows.

Give me your address, and i will send one or two of the old Fairchild FPT 100s; they were transistors *designed* to be used as photodetectors. None of the industry standard flim-flam of putting a glass lens at the top of a metal canned 2N2222. Now all i gotta do is find them!

Speed requirement?

Quantum efficiency doesn't vary that much among silicon diodes. The more limiting variable is dark current. Larger diodes may or may not be more sensitive, because dark current may increase as much as photocurrent for given radiant power density or luminous flux density. If speed and/or bandwidth is an issue, then noise or D* is also important.

I've used photodiodes with 1 mm^2 die under starlight condx. They were made by Fairchild back in the '70's, weren't very expensive.

I recall this 100 mm^2 device from the RS catalogue:

But it is 149 UKP(!)

Even allowing for RS expensiveness, I bet they are not cheap from other places either. :-<

Sounds like there's the common conflict of needing high quality and low price.

I'd suggest el-cheapo solar cells, but I guess they won't be that good.

What's the application?

Only 85 UKP, but still too dear.

I forgot about these.

If you can tile an area with several smaller diodes and parallel them, these are hard to beat for cost to performance ratio. About $1 each, with 7 mm^2 area and dark current around a nA at room temperature with

Did you really calculate what you gain by a larger area photodiode with higher leakage currents against a feedback resistor with higher value and an amplifier with lower Bias?

If the photodiode is operated at zero bias, does the leakage not go to zero?

. .---R---. . | | . | |\\ | . .------+--|-\\__|_____ . | .---|+/ . v | |/ . - | . |PD | . -+-----+-----------------

I think Tony already mentioned Silonex, these were the lowest cost that I found, too. They make large area "bare" photodiodes (look like solar cells) as well as the lowest cost "packaged" diodes I found.

Thanks for those excellent leads.

Thanks for the kind offer, but I already have a few of those golden oldies. Anyway, they're too small, unless we use a lens, which is a mechanical pain, but allows almost any sensor to work.

Luminescent e-coli cells, cheap high-school student experiment.

Super slow.

The cell's dark resistance should be shown across the diode. UDT's PIN-10DI spec is 200M typ at 10mV. NSC's LMC6061 opamp has 100uV typ voltage offset, which should create a false dark current of 0.5pA with 200M. A 1000M feedback resistor will let one easily read this out with 0.1pA resolution on a low-cost 3.5-digit DVM.

Everything in this setup is cheap, except the detector. Actually, it's not that far out of reach, and the PIN-10s with their elegant 1" dia case and handy BNC connector make a worthwhile investment. I got my first one as a graduate student almost 40 years ago when they first came out. Every high-school lab should have one.

I'll investigate them, thanks.

How does the signal go up as the area**2?

Iphoto goes as the area, and P=i**2 R. Background noise power goes up linearly, since shot noise current goes as sqrt(i). Capacitance goes up as the area, so you have to watch out for the multiplied voltage noise of the op amp. You really can win big by tiling with solar cells.

Many moons ago, a colleague and I built an endpoint sensor for post-exposure bake of chemically-amplified photoresist. It used seven 1-inch by 3-inch solar cells arranged in a heptagonal prism, looking at the diffraction pattern of light from a red LED shining down on the wafer from inside a black cardboard cone. The diffraction pattern was hexagonally symmetric, so we always got at least two good diffraction orders of opposite sign. (Taking linear combinations of the two diffracted orders made it possible to reject diffraction signals from underlying structures.)

The solar cell approach was about 20 dB more sensitive than our original setup using a lens and a CCD camera, not to mention about 10 dB cheaper. The reason was that we detected all the diffracted light, and pi steradians is dramatically bigger than what you get with a lens.

Cheers,

Phil Hobbs