References for high-speed (>GHz) testing of photodetectors

(My apologies if this shows up twice. My system glitched as I was sending it, and I'm not sure it went out.)

I need to do some high-speed (>GHz) testing of some visible-light photodetectors, including building the test setup.

Can anybody give me some pointers to papers, book chapters, application notes, whatever...

I've tried doing some literature searches, but everything I've found so far has consisted of a paragraph or two in a paper somebody's written about their new high-speed photodetector design. I'm hoping for something a little more specific than that.

Bob Pownall

Reply to
Bob Pownall
Loading thread data ...

Building Electro-Optical Systems, Philip C.C. Hobbs, ISBN0-471-24681-6

Reply to
Paul Mathews

I don't recall if Phil includes this or not, but there was a very clever technique employed to test the high-frequency/fast time response of photodetectors in early days of lasers, when fast light modulators, scopes, and related instrumentation were not yet available.

Incoherent light from an incoherent source (i.e., a thermal source or discharge lamp) can equally well be described as a monochromatic light signal randomly modulated by random phase and amplitude modulation components at frequencies extending up to the bandwidth of the spectral distribution of the light, right?

Even a very narrow bandwidth incoherent light beam -- unless it's

*really* narrow -- thus contains a continuous distribution of noise AM and FM components extending up to well beyond the GHz regime.

Shine a light source like this onto your photodetector; feed the photodetector output into an rf spectrum analyzer (and those were generally available at the time); look at how the output from the spectrum analyzer (integrated over some suitable integration time) falls off at higher frequencies; and you've got the amplitude response you want.

Chopping the light source at some frequency down in the audio range and using a phase sensitive detector/integrator on the spectrum analyser output lets you subtract out the shot noise component of the dark current in the photodetector or the noise background in any subsequent amplifier. Doing a reasonably accurate calibration of the passband and sensitivity of the spectrum analyzer using coherent rf sources lets you make quantitative measurements of the photodetector response.

Can't cite a reference off hand, but I seem to recall Peter W. Smith being one of the people who published on this, some time in the 1960s or

1970s.
Reply to
AES

Thanks! Am I correct in assuming that's the Phil Hobbs who posts here occasionally?

My library has a copy of the book (Yay!), but it's listed as "Missing". (Bummer.)

I'm go Prof. Siegman - thanks for the pointer. I'll follow up on the reference, to see if it's a technique I can use in my application.

(BTW, I've got your book! I keep hoping I'll run into you sometime - maybe I can get it autographed!)

Bob Pownall

Reply to
Bob Pownall

So, basically, the apparent spectrum of the shot noise is the frequency response of the receiver? Pretty simple!

Reply to
John Devereux

Yup. Wish I could find the reference.

Reply to
AES

John Devereux schrieb:

Yes, if your source is shot noise limited (that's probably always the case at such high sideband frequencies) and if it's running in a single mode.

Additionally of course the shot noise must be a lot bigger than the electronic noise of your amplifier. If you cannot guarantee that, you will need two lasers at slightly different frequencies and sweep one against the other. If you avoid backreflections from one laser to the other so that they do not injection lock, the resultant beat signal is of constant amplitude and it's envelope will show your receiver response.

Cheers, Jürgen

Reply to
Jürgen Appel

I was thinking in terms of using a non-laser source, like an incandescent lamp or perhaps a LED. The RIN of a laser could be an additional noise source likely to mess up this measurement.

Reply to
John Devereux

I like shot noise for calibration very much, because there's a simple relationship based on fundamental physics between the dc photocurrent and the ac shot noise. For computing the power spectral response and SNR of a photodiode, TIA, and subsequent stuff, it's a great technique. Using chopping to take out the effects of subsequent amplifiers etc is probably useful sometimes, but I've never done it that way--I'm usually insisting on being shot noise limited whenever possible.

The problem with using that for high speed photodetectors is that (at least for data applications) there are other things involved than just the (power) frequency response. People are usually interested in eye diagrams, and phase funnies and low-frequency artifacts can really make a mess of those. These show up especially in eye diagrams with long sequences (2^24-1 to 2^32-1 or longer).

For instance, if your photodetector is AC-coupled with corner frequency f_HP, then at a data rate B, you'll start having problems with sequence lengths of around B/f_HP or longer. Phase funnies give rise to ringing, slow transitions, and long tails, any of which will mess up an eye diagram really fast but won't be visible in a power spectrum plot.

Cheers,

Phil Hobbs

Reply to
Phil Hobbs

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.