'optical' feedback resistor

The phototransistor optoisolators, if you short B-E, are photodiodes. The late generations of LEDs have good linearity, but you might want to be selective; an isolator with a poor optical coupling would allow your LED current to be rather high, and at high currents the linearity is better than at microamps.

b-e? I thought that was the photo-junction. short b-c maybe? My only experience with photo-transistors is they are slow.

Yeah I wonder what wavelength/ process is best? IR telecom do some nice high power. Linear from 1 uA to 1 mA would be nice. I saw some article abstract that used a laser diode below threshold.

George H.

ght

or.

It's not a bad idea. The leading problems are nonlinearity in the LEDs and the 3 dB shot noise penalty inherent in cancelling two photocurrents.

Some years back JL, his colleague Jonathan Dufour, and I did a nanowatt pho toreceiver of that general description, called the PH200 (since discontinue d due to poor sales). It got over the above two problems by (a) using two p hotodiodes in series to reduce the shot noise penalty to 1.6 dB rather than 3 dB, and (b) using a third PD plus an op amp TIA to form the output direc tly from the LED's output, which eliminated the LED as a source of error.

Cheers

Phil Hobbs

George, Drop me an email and I'll send you the paper

Sierra Romeo Oscar Bravo Echo Romeo Tango Sierra Six Three Two Eight @ gmail.com

srob**ts6328 @ gmail.com

Steve

There's nothing especially interesting in the 1987 Forsberg IEEE article. As usual, node capacitance and e_n C_in noise is a big issue for s/n with fast TIAs (see Aoe III, 537-555), and you don't want to add too much additional PD capacitance. Bootstrap tricks are good. Most LED-PD combinations are quite linear at low currents. Photo-transistors need not apply.

It's the depletion zone where photoelectric effect results in net base current, and B-C is the deeper depletion region because it's biased. I think.

This reference confirms:

I've done it, but it was a nuisance.

And a photodiode is going to have more capacitance than a resistor.

Thanks Steve, It's not all that important that I get the article.

Here's something more recent. from RSI

George H.

Hi Win, Well I was playing around with an old PD design, and found that it was the switch capacitance (about 5 pF) across the feedback resistor that was limiting my BW. This trick might get around that. It would also be awesome* (I think) at very low light levels. Crazy idea, I could stick an optical attenuator between the feedback LED and feedback PD to 'increase' the 'resistance'. (More voltage out for less current.)

I need to find a linear LED.

George H.

I can imagine the nuisance part. You've got optical alignment issues for one.

Right, but the capacitance is to ground and not across the feedback R. In high gain TIA's I'm thinking it's often that C which kills the speed.

Does anyone make fast opto-isolators? (They might also have parallel capacitance issues.)

George H.

light

ctor.

d the 3 dB shot noise penalty inherent in cancelling two photocurrents.

hotoreceiver of that general description, called the PH200 (since discontin ued due to poor sales). It got over the above two problems by (a) using two photodiodes in series to reduce the shot noise penalty to 1.6 dB rather th an 3 dB, and (b) using a third PD plus an op amp TIA to form the output dir ectly from the LED's output, which eliminated the LED as a source of error.

Huh, discontinued, that stinks. You need a market! Have you ever talked t o Thor labs and offered to do a 'better' photo detector with them? I was thinking of the LED linearization trick... well you see it in opto-isolators. But that starts to sound like a lot of TIA's.

It does seem like this would allow a gain switch on the TIA. That's really useful. (at least for me.)

George H.

The one I did with the Highland folks had a CTR of 0.01% between the LEDs and the current feedback photodiodes, and about 10% from the LEDs to a larger PD + vanilla op amp TIA that formed the actual output. Sort of a poor-man's PMT.

The series-connected CFB PDs were kept balanced by a BF862 diff pair that shunted the feedback current to two LEDs, one per PD. Interestingly that provides a 3-dB SNR improvement just as parallelling them would, but without doubling the photocurrent and with half the diode capacitance.

Cheers

Phil Hobbs

Well, Hobbs ElectroOptics now sells the QL01, which has slightly better specs anyway.

Cheers

Phil Hobbs

Huh, is it a free month at RSI? I can click on all their articles and get a download.

George H.