Need help to design a transimpedance amplifier

djA

se

It's avalanche gain, which probably means that it's sensitive to everything , including the phase of the moon.

I did a little googing on avalanche gain as function of electric field, and couldn't find anything useful.

"Zener" diodes with breakdown voltages above about 10V are actually avalanc he diodes, and the sharpness of the knee in these parts illustrates the vol tage sensitivity of the process. The temperature cofficient of breakdown vo ltage is positive and of the order of 0.1% of the breakdown voltage per deg ree K fpr voltages well above 10V. It's listed for some NXP parts

Sometimes the noise path is not visible on the schematic; it's capacitive coupling through the air, or ground loops, or magic. Switchers should be far away from low-noise amps.

We recently tested a fiber link intended for clock distribution with femtosecond jitter. The wideband noise floor was below -170 dBc, but it has half a dozen birdies in the 5-300KHz range, most of which we can't account for. If we productize this one, the most serious part of the design is going to be the metal boxes.

Of course we can't measure 180 fS RMS jitter. We have to inferr it from a spectral phase-noise analysis.

Yup, and they should have nice smooth slow edges, a tight layout, generously-sized toroidal inductors, and electrostatic shields. (Magnetic coupling is tougher to get rid of, but I haven't had a problem with 150-kHz switchers and toroids, as long as nothing saturates.)

One thing that really helps is making sure the board is grounded to the box in several places, especially near the SMPS. I'm using some little L brackets in a one-piece Hammond box with die-cast end plates and a conductive gasket. Works nicely, though it's a bit on the fiddly side.

It's the hardest to simulate, for sure.

Fortunately that's pretty easy to do!

Electro-optic sampling can measure numbers like that standing on its head. Besides cost, the big drawback is that you have to sync to the laser, it can't sync to you.

Cheers

Phil Hobbs

Not knowing what you are doing is fine by me. (It's how we all start.) The common mistake (I think we've probably all made at some time.) is to assume you know what the problem is. (wrong opamp) And then ask for that. Rather than laying out the whole problem. I guess it can be hard for some people to say, "I don't know what's wrong here."

George (not shamed by my ignorance) Herold

Hah, I meant to comment up-thread on putting a (gain/R) switch in a TIA. The first (or maybe second) time I did a PD TIA I put the rotary switch in backwards. Too much capacitance. You need to put the pole on the inverting input.

George H.

Sure, I did too, and judging from the IIT Madras logo block, he's probably a grad student. What I don't understand is getting all the way through schematic capture, prototyping, testing, and making Powerpoint slides without ever estimating anything. I'd have expected a bit of professorial input there at the very least.

Well, T. Obulesu did lay it out very well in the PP deck. No problem there, except for a lot of wasted effort. (Not wasted if he learns to estimate stuff ahead of time, of course.)

He eventually did that, too, so he deserves full points there as well. Doing the estimate is the hat trick. ;)

I'm considering doing a small book on the apparently lost art of doing photon budgets. I can mostly string together stuff I've done over the years, so it wouldn't be a huge schlepp. Maybe next year after I get my third edition finished.

Cheers

Phil Hobbs

Yup. For really high-Z stuff I've used relays with bootstrapped coils to get rid of that last 0.4 pF. Tough to get many ranges that way, though. For my first few products I'm going to use a single range.

Cheers

Phil Hobbs

apparently it has a name:

-Lasse

The gain-range switch is in the wrong place no matter where you put it.

I did one box that had two TIAS, with 100:1 gain difference, and switched the photodiode between them with the cascode transistors, a spin on Phil's bootstrap cascode. That wasn't as awful as some other ideas.

It was a brilliant product, but the customer decided to "design" their own, after I showed them how to do it.

Sometimes that happens. When a customer has an in-house design team, they usually manage to win.

Oh, _those_ guys.

Cheers

Phil Hobbs

In message , Lasse Langwadt Christensen writes

There's a parallel thread going in the LtSpice Yahoo group with the same OP and topic. It's interesting to compare and contrast.

I designed a Laser Rangefinder Receiver with an RF Switcher PSU* on the same board, but I had the luxury of being able to switch it off and rely on a FGC to hold the charge during the ranging gate. When that was over all hell broke loose.

Brian

• Based loosely on this note :-

The first paragraph is what the OP is experiencing .....