Photodiodes

What would you do with them?

John

Endless applications. Mostly looking at flourescent medical imaging. There is also nano dot and nano flare technology which can enhance it even more.

Cameras, astronomy.

greg

Thats quantum dots, not nano dots.

greg

also

A lot will depend on the noise level. Gain is cheap.

The most interesting part is the wide spectral range.

Press-release breakthrough things like this have historically made it to market at maybe a 1% rate.

John

I sure hope not. It also includes much more powerfull solar cells.

What if the Hubble had 100 times the sensitivity ?

greg

Does it? I missed that part.

Impossible. A good CCD already catches and integrates most of the photons that hit it.

If you get excited about every press-release breakthrough like this, you'll go crazy.

John

also

You said it.

Black silicon is well known from trench DRAM, circa 1992--we used to get spikes that looked a lot like that in the bottoms of trench capacitors, mainly because of contamination. A bit of crud blocked the RIE, leaving a big long spike coming up from the bottom of the trench. This stuff is somewhat different because they actually make an alloy. The stuff in the puff piece is hilarious--impurity sites actually being so numerous as to form *bands!* (The crowd gasps.) Claims of low noise are totally unsubstantiated--all they show is a spectral responsivity curve, with comparisons to silicon and Ge/InGaAs PDs with no gain. (It wouldn't be as impressive if they'd put in APDs or electron-multiplying CCDs as well.)

Pluses:

It's really black and works over a very wide wavelength range--which could be exceptionally nice for some niches, because integrating III-V devices on silicon is very complicated (read expensive).

Photconductive gain (as in CdS cells) gives you more signal.

Its high responsivity can help people making infrared night lights. ;)

Definite minuses: Slow speed. There's a 1:1 tradeoff between gain and speed in a photoconductor, as I mentioned a few days ago in this very boutique.

Things to worry about:

Passivation issues. How in the world are you going to passivate those spikes?

Leakage. With the amount of crystal damage you have to have to do what they're doing, I would anticipate really horrible amounts of leakage.

Trapping. Ditto traps. This might lead to nasty nonlinearity at low light levels.

And most of all,

Noise. For practical purposes, all primary photogeneration processes exhibit exactly full shot noise, and photoconductors have twice the noise of photodiodes, because there the recombination is stochastic as well as the generation.

There are specially prepared quantum systems called "squeezed states", popular in the 1980s, which can get to, say, 3 dB below shot noise. The joker is that it only works if you shine the laser straight into the photodiode--once the total attenuation reaches 3 dB, the effect is essentially lost.

There are also classical fluctuations (i.e. Johnson noise) in the light itself--looking very hard at the red tail of the spectrum of a very blue thermal source (e.g. Rigel), you can measure noise about 1% above shot noise. Those fluctuations have important spatial correlations as well, which enabled Robert Hanbury Brown (a technical hero of mine) to measure stellar diameters by using two light buckets feeding PMTs, and cross-correlating the *noise*! (He called it the Intensity Interferometer.)

But otherwise, the instantaneous rate of photogeneration will exhibit exactly full shot noise, and that sets the absolute maximum achievable signal-to-noise ratio of the detector system in a given bandwidth. The rest of the circuitry is there solely to give you high response speed, a bigger signal swing, and more drive. The very best that circuitry can do for the SNR is to refrain from screwing it up--there's no way to improve it, even in principle, except to get more light, reduce the background light, or give up hope and narrow the bandwidth.

Silicon CCDs used correctly can have sub-electron noise levels. Yes, folks, that's less than one electron RMS readout noise per pixel per sample. That takes cooling and multiple readouts per shift cycle, but it isn't that hard to do.

Photoconductive gain amplifies the shot noise as well as the desired signal, so it sounds innocuous, but remember that the recombination in photoconductors is stochastic too--so each photoelectron gets two copies of the shot noise, once when the photocarrier is generated and once when it recombines. (Photodiodes don't have this effect because the carriers all recombine when they hit the pads, which isn't stochastic and so doesn't produce noise.)

And lastly, it really isn't difficult to get to the shot noise with a photodiode if you're willing to wait awhile--Gary Eppeldauer et al at NIST published a heroic study where they got *14 decades* of linear response out of a specially prepared Si photodiode. They had to settle for millihertz bandwidths, though. The tradeoff is the same as with the photoconductors: you lose bandwidth linearly as you raise the load resistance, just as you do by increasing the photoconductive gain by increasing the carrier lifetime.

So it might have some utility in niche applications, and if the passivation can be figured out, it would be really great for 1.0-1.5 micron wavelength cameras. I'd probably buy one of those for looking at diode laser beams in the lab, if they were cheap enough. You can't use that in consumer goods, though, because they can see through many kinds of clothes...remember the Panasonic debacle!

Moral of story: a photodiode is the most amazing transducer of anything in all of technology--the only thing arguably better is a good transformer, and they don't count.

Cheers,

Phil Hobbs

What's the gain mechanism in a CdS cell? It behaves like a resistor, completely different from an avalanche diode.

John

Depends on whether it has a "Larkin gain factor" or not ;-)

Seriously, I think photons can easily "knock loose" carriers in CdS .

Big problems with CdS: repeatability, slo-o-o-ow recovery time constant after light is removed.

I used them in an automatic head-lamp dimmer design for GM's Guide Lamp Division back in the mid '60's.

...Jim Thompson

Don't try to make obnoxious personal jokes, mainly because you're not very good at it.

John

Oooooh! Score one for the Jimby ;-)

...Jim Thompson

In a photodiode, once you generate a carrier pair, you immediately get a current pulse of 1 electron per transit time coming out the leads, which lasts until the electron and hole hit their respective pads and recombine. Thus the transit time is an upper limit as well as a lower limit on the length of the pulse, and the integrated charge is the same as the number of photocarriers:

q = eN

In a photoconductor, for solid-statey reasons that I don't adequately understand, the carrier lifetime is *not* limited by the transit time--it can be either greater or less. The instantaneous current in the pulse is the same, but it lasts for the carrier lifetime, not the transit time, so the integrated charge is

q = eN(tau/t_t)

which can be either gain or loss. CdS has a lot of gain because its carrier lifetime is very long, which also means that it's very slow.

Cheers,

Phil Hobbs

I was just reading about avalance photodiodes. I was not aware of gain increases with large reverse bias. I have seen this before trying to make a geiger counter. I see reverse bias to decrease capacitance. I have been using Hamamatsu 256 diode camera arrays. They come prebiased. On later models of the $15K camera, they changed the bias supply creating extra noise from a faulty chip circuit. Its centered around 400 Hz as I remember. That was pain to deal with. Now they stopped production. I don't think anybody is making arrays now, but that company has arrays in their thinking. We got two camera systems with 512 channels and adjustable gain and filtering on each channel.

greg

make a

AFAIK TI still makes LLLCCDs.

Cheers

Phil Hobbs

here is also

lly

ter.)

Wow, thanks Phil I totally enjoyed that post! A few questions follow,

"> Noise. For practical purposes, all primary photogeneration processes

Just a clarification here. Reversed biased photodiodes are sometimes referred to by the manufactures as being used in "photoconductive mode", but this is NOT to be confused with the above mentioned photoconductors.

About Hanbury Brown (and Twiss) and the excess noise caused by the spatial extent of the source. Have you every considered trying to do these measurements in the lab with 'cheap' photodiodes? (I guess by cheap I'd like to exclude $2k APD's and photomultipliers as used by HB&T)

"because they can see through many kinds of clothes...remember the Panasonic debacle!"

What was the Panasonic debacle? There really were "X-ray" glasses? Can you give me a link or seach phrase?

Take care, George Herold

Right. The SiOnyx gizmos are photoconductor photoconductors. Whoever came up with the term 'photoconductive mode' for back-biased photodiodes ought to be staked out on an anthill for a few weeks to reconsider.

I've wanted to build an intensity interferometer for ages. Before my antenna-coupled tunnel junction project got cancelled a couple of weeks ago, I was hoping to use the II to make an IR imaging system that would cover a whole hemisphere with no lenses at all, using a CO2 laser as the LO of a heterodyne array. The idea was to use the II approach (fourth order statistics) for most of the array elements, and a few regular phase-sensitive interferometers with really fast ADCs. Since the II measures the autocorrelation of the image, all optical phase information is lost; I wanted to use the few phase sensitive pixels to seed the phase retrieval of the rest of the image.

Sic transit gloria mundi.

It was actually the Sony Nightshot camcorder feature, allowing near-IR illumination to be used at night--but it also worked in the daytime, where it rendered many kinds of light clothing almost invisible. That was around 1998, according to Google.

Cheers

Phil Hobbs

Right. The SiOnyx gizmos are photoconductor photoconductors. Whoever came up with the term 'photoconductive mode' for back-biased photodiodes ought to be staked out on an anthill for a few weeks to reconsider.

I've wanted to build an intensity interferometer for ages. Before my antenna-coupled tunnel junction project got cancelled a couple of weeks ago, I was hoping to use the II to make an IR imaging system that would cover a whole hemisphere with no lenses at all, using a CO2 laser as the LO of a heterodyne array. The idea was to use the II approach (fourth order statistics) for most of the array elements, and a few regular phase-sensitive interferometers with really fast ADCs. Since the II measures the autocorrelation of the image, all optical phase information is lost; I wanted to use the few phase sensitive pixels to seed the phase retrieval of the rest of the image.

Sic transit gloria mundi.

It was actually the Sony Nightshot camcorder feature, allowing near-IR illumination to be used at night--but it also worked in the daytime, where it rendered many kinds of light clothing almost invisible. That was around 1998, according to Google.

Cheers

Phil Hobbs