TIA Photodiode Bootstrap at 10MHz

I have been told by the project engineer who has given me this assignment the capacitance is a parallel one. It is a custom detector device so there is no data sheet I can link to. I have not even been given a hard copy of one. So I can only go by what he has told me. Do you have knowledge of photoelectromotive force detector physics that causes you to disagree?

Right now my main concern is just getting the bandwidth. Once one or more ways to do that are established I can be concerned about minimizing noise.

I have not until now considered using another op amp instead of an amp composed of discrete devices to do the bootstrapping. Doing it this way would reduce added capacitance on the virtual ground.

Regarding my earlier posting that was 18.8GHz not 18.8THz.

ay

one

t,

et

es

in

te

ce.

's

or.

ng

be

"I have been told by the project engineer who has given me this

That sounds much worse. It's like some photo-resistor.? It will always be leaking as you bias it. (and leaking as the light shines on it.) ughh.

George H.

Similar, but unlike the passive photoresistor, the electromotive force detector converts photons into electrical energy. In a TIA circuit it can be modeled as a current source that does not require a bias. The current is proportional to the motion of a beam light across it. It is very good at detecting changes in a laser light's speckled pattern produced when it reflects off any but the most even of surfaces.

Right now my plan is get two of the lowest noise JFET or CMOS input op amps I can find that satisfy the bandwidth requirement. One will be used for the feedback resistor and the other will do the bootstrap.

I just had the exact same requirement for an IR CO2 sensor: Cust: we've made something, but it's way too noisy and we don't understand why (I was said it has 40R leak resistance). Can you improve on that? And no, we don't have the sensor spec... Oh, and you'd get paid only if the outcome please us. Oh, and we want you cheap...

Me: OK, no way for me to work with you. Good luck :-)

They've finally found someone accepting that.

And no, it isn't really a small firm. Rather, a very big one...

way

s one

put,

uiet

is

odes

by

e in

ce

eate

the

to

the

ance.

th,

at

of

at's

n,

ld

ctor.

sing

in

es a

ay be

ng

You might look at the ADA4817 from analog devices. They show a TIA circuit that almost meets your bandwidth spec with 50k ohm of 'gain'. At 1uA that would give you 50mV of signal which is starting to get the shot noise above the johnson noise of the resistor. And perhaps you can do better with your bootstrap or using a cascode front end ala Phil H. I've got a sample of one of these in a drawer... whenever I find the time to try it. There is a nasty gain peak out beyond

George H.

Photodiodes and photoconductors are completely different.

The

I'd never heard of a photoelectromotive detector, but Google had. Interesting. From what I gather from the free literature, it's basically a GaAs 1-D lateral effect photodiode run at zero bias, with the output taken across the ends of the top layer.

I haven't read any really detailed papers, because they all cost $30, but it relies on the carrier density responding slowly to changes in the illumination pattern. It's made of GaAs rather than silicon, which I'm guessing is because the hole mobility in GaAs is so very low, which essentially glues the charge density in place. (Silicon would respond much faster.)

For static illumination, the slow carrier diffusion makes the junction E field flatten out locally, so the charges all recombine, and the output current is zero. When the illumination changes, though, you have photocarriers being generated in regions where the net current is suddenly nonzero. These carriers produce a displacement current immediately (not when they arrive at the electrodes), and the carriers still flowing in the previously illuminated region produce a net displacement current of the opposite sign. The two sum to zero, but they're spatially offset from each other, which produces a net current from the outputs. Like everything else in nature, it's linear for small displacements.

It's a cute technique, which reminds me a bit of a photorefractive phase conjugate mirror, in which the beam interference pattern forms a slowly-varying hologram that corrects for phase and pointing errors.

A few things I'm not clear on: (1) There has to be a junction in there somewhere, because it works at zero bias and can produce current in both directions--which in a symmetric structure requires a third electrode.

(2) Splitting the photocarriers between the ends seems to require a high-resistance epi layer, which would tend to make the response slow or else very noisy, like a lateral effect cell--the full Johnson noise current of the epi layer appears in the output.

(3) The diffusion and recombination currents both have full shot noise, and the forward-biased regions will have a very low resistance. This will seemingly contribute a lot of noise to the output--the noise will be that of the full beam's photocurrent, not merely that of the small difference current.

So in general this doesn't sound like a low-noise device, although it's certainly convenient for speckle measurements--the usual alternative method is TV interferometry, in which you cross-correlate the speckle patterns before and after some change (e.g. pressurizing a tank). TV interferometry is slow, since it takes a minimum of two video frame times.

It would be worth finding out what the noise level of the device itself is. If I'm mistaken and it's actually quiet, you'd be better off making the bootstrap out of a single-ended BF862 JFET rather than an op amp. Otherwise a couple of op amps is probably fine (one of my favourites is the ADA4817-1).

Cheers

Phil Hobbs

Thanks for the hint Phil, A search for (photo-emf detector) brings up a nice article from Newport.

George H.

This project had a break for awhile but is now revived. I am given new specifications to model the emf detector with. These are 100kOhm with a parallel capacitance of 1pf. This is a much easier task.

There was discussion in this thread about other methods of noise reduction. The goal of using a bootstrap was primarily to get the bandwidth. The low impedance of the emf detector meant bandwidth challenges because to a first approximation that neglects capacitances the bandwidth will be the GBW of the op amp divided by the noise gain.

I like the ADA4817-1. Using it may make possible doing this without bootstrapping.

Given the 1pF parallel capacitance in the model a bootstrap would simply add as much capacitance as it removes and add noise.

The bootstrap would also increase the equivalent resistance of the emf detector. But the resistance in the model is high enough now, and the GBW of the ADA4817-1 is high enough, that I do not see a need.

The manager wants a 10MHz bandwidth and 2 Mohm transimpedance. I am not going to be able to do this in one stage. Even before considering the capacitances from the virtual ground to actual ground the parallel capacitance of 2Mohm to get a 10MHz bandwidth would have to be 8fF. That is not going to happen.

Whats more, because of the capacitances on the virtual ground the feedback resistor must have a parallel capacitor for stability.

To reduce inductances of resistors and capacitors I intend to use wide packages such as 0508 along with wide pcb traces.

Your information about emf detector physics is of much interest to me. Thanks.

es.

"The manager wants a 10MHz bandwidth and 2 Mohm transimpedance. I am not going to be able to do this in one stage."

This sounds like a stretch no matter how many stages you use. Sounds like you may have to educate your manager about capacitance. Even with only 1pF in the detector you've still got a few pF on the input to the opamp, plus the PCB trace connecting the two.

"> To reduce inductances of resistors and capacitors I intend to use wide

Well on the photodetector side of the circuit I'd keep traces short and thin to reduce the capacitance. And be careful of where you put the ground plane.

(But Phil H. is the expert here. BTW have you bought his book yet?)

George H.

Wider traces reduce inductance and increase capacitance.

I have not bought his book yet but I will.

Trace inductances won't matter in a 10 MHz TIA. They barely matter at

John