Photodiode TIA

The problem is that the cpacitance of your photo-diode and the reistance of your feedback resistor combine to produce a steadily risng noise gain about their RC product.

The higher the bandwidth of your amplifier (if it is above the RC product, which yours will be) the more the thermal noise of the amplifier will be amplified.

The Burr-Brown - now TI - OPA656 has a useful datasheet, including some discussion of this point, around page 11

Hope this helps.

--
Bill Sloman, Nijmegen

A lot depends on how much light you have. If it's microwatts, the noise budget will be tight. If's it's in the milliwatt range, you'll probably overload the Maxim and might be better off making your own TIA. I've had good results using a current-mode opamp as a TIA when there's lots of light.

What's the capacitance of the OPV480? That's a serious issue at high speeds.

John

Bill - thanks for the info, i'll check out that datasheet.

John - the VCSEL i'm using is

I've not used it before or anything but the spec looks good. The optical output is 1mW @ 5mA, but then the photodiode will be 1cm away.

Its actually a OPF380 Optek part (sorry, typo i should have spotted) and the capacitance is max 2pF at 5Vr so i was hoping it would drop right off when reverse biasing at 20 odd volts?

Thanks

Andrew

You don't usually seem to get much reduction in capacitance in going from 5V to 20V of reverse bias. Since you are starting off with only

2pF, the capacitance you need to plug into your equations is likely to be dominated by stray capacitances anyway - the OPA656 has about 2.8pF of common mode input capacitance.

--
Bill Sloman, Nijmegen

You forgot that Maxim only advertises stuff that is not available.

The capacitance will drop somewhat from 5 to 20 volts (check the datasheet, probably less than 2:1) but in fact higher voltages make the critters faster independent of the capacitance dropoff... carriers are swept out of the I region faster, I guess. 50 ohms and 2 pF is only a 100 ps tau, so capacitance isn't the speed limit here.

You could do a quick DC experiment to see how much photodiode current you're going to get, and then plan the amp. The Maxim TIA will probably work OK. I'm getting about 200 MHz bandwidth here,

using an Optek silicon PIN diode driving an AD8014 as a tia, with only

6 volts on the photodiode. I guess I should have built a little boost converter to get some more bias voltage. Oh well, next time.

John

But they always have samples. A few months ago, in deep trouble, I had all my engineers and some outside friends and family request Maxim dac samples. Turns out you can get 18 samples of this dac per request if you play with the part number options. We got enough to get out of trouble.

John

Sure no way to run an airline.....

There have been threads on this. The engineering concensus seems to be "don't design with Maxim parts."

If I browse my parts/purchasing database, most of the Maxim parts have a note like "LONG LEAD!" or "10 WEEK LEADTIME" and occasionally "OBSOLETE NO SUB"

Maybe they cater to a small number of huge-quantity OEM customers or something.

John

Sorry for sounding ignorant but what quick DC experiment would i do? Is this just turning the VCSEL on at the required working distance and seeing how much current i get? Can i just add a load resistor and check the voltage across it?

Lots of questions!

Andrew

Yup. Run the nominal DC current into the laser, set up the mechanics like the real thing, bias the PIN with a 9-volt battery and measure the pin current with a microammeter, or measure the voltage drop across a 10K or some such series resistor. Expect maybe a milliwatt from a cheap vcsel laser, and roughly 0.5 amps/watt (0.5 ma/mw) from the photodiode.

Now the issue becomes, how much of the laser's light is landing on the active area of the photodiode? If it were fiber-coupled, it would be close to 100%. In free air with no optics, it might be 1% or it might be 0.01%.

John

everything

on

use

the mechanics

OK so i'm back for more!

Experimentation shows i'm getting about 7uA at my 1cm free air with the stated 1mW output of the VCSEL. I ended up getting in the AD8011 as it was easy to get hold of and i'm trying to plan my TIA around it.

i'm going to need an Rf of standard resistance 620K. Now reading the suggested reading material on the OPA656 datasheet it goes into a calculation for Cf, but this requries the GBP which i can't decipher from the AD8011 datasheet. It suggests that for a gain of 10 the bandwidth is 57Mhz - does this mean the GBP is 570Mhz? Or am i missing something, i mainly ask because the GBP at G=+1 and G=+2 don't tally up with this at all.

Thanks

Andrew

snipped-for-privacy@yahoo.com a écrit :

the mechanics

AD8011 is a *current* feedback opamp and there's no such thing as GBW product for CFB opamps.

And you surely don't expect 4V signal in one stage, do you?

150MHz BW over a 620K resistor is, hem, 1.7aF (yes 0.0017pF) You'll never get, even remotely, down to this figure. Your resistor parasitics will be one the 0.1pF order, and you'll have to account for the PCB too. There's a nice trick to overcome this (to some extend). Google for "RCR trick win hill" without the quotes, but this won't get you where you want.

Also a 600K FB resistor will give you a very low loop gain you have to proceed in 2 stages or with a composite opamp.

4 volts is a lot of signal too. Why do you thing you need that much?

Depending on your application CFB opamps might not be the best choice due to their higher current noise, but you have to check this. One interesting point of theirs is their low input impedance, which make your PD parasitics less critical regarding loop stability. However this unfortunately change nothing about the en.C.w noise, which will begin to dominate around 5E-12/(2E-9*5E-12*2*pi) = 80MHz (2pF for your PD + 2p for the opamp + 1p for PCB).

--
Thanks,
Fred.

everything

on

i

use

the mechanics

Hi, Andrew,

To work at 160 MHz, with the 7 uA pd current, you're going to need multiple gain stages. The stray capacitance across a 600K resistor would squash the gain, and the opamp doesn't have enough gbw anyhow.

An official tia, like one of the Maxims, would be a good first stage. They can achieve effective transresistance in the kilohms and still keep the bendwidth up. Or a current-mode opamp like the AD8011 with a feedback resistor in the < 1K range, with maybe a pF of shunt capacitance if you need it for stability. That will give you under 7 mV of signal, so you'd need later gain stages to get up to the volts level. This much gain-bandwidth desperately wants to be an oscillator, so the input node needs to be well shielded from the output.

Improving the optics, with a lens somewhere, could increase your pd signal by a huge factor and simplify the situation. 7 uA ain't a lot of signal at these speeds.

John

but i

everything

622Mbps,

Reading on

if i

to use

straight

transimpedance

up the mechanics

So it looks like i've got a lot to learn. Fred the main reason i wanted in the order of 4v signal is so i can directly couple it to ecl logic. But as you say i can accomplish this with two stages.

I've tried setting up the AD8011 with a 820R feedback. It doesn't seem to need the shunt capacitor - at least it doesn't seem to be oscillating however whenever i attach the photodiode even with a very short single core connection, all i can see is around 200mv of noise on the output.

+5v | V 820R - |-|___|-| | | | | | |+5v| | | |\\| | |-----o-|-\\ | | >--o---- 0v -|+/ AD8011 |/| | |-5v

(created by AACircuit v1.28.6 beta 04/19/05

Is all i've got - is there something i'm missing?

John - you mention lenses do you know where i can get my hands on any. I've done a bit of digging and have found GRIN lenses that seem to be what i want but they also seem expensive from only a couple of places i've been able to find them - is there any cheaper alternatives?

Thanks for you help

Andrew

Hi, Andy,

What do you mean by "single core connection"? If you mean coax, the capacitance will wreck things. This opamp will not like capacitance from its - input to ground, and the pd should be no more than 2 pF itself, preferably less. The pd capacitance adds a pole in the feedback loop and makes things unstable. Tne easy fix is to add maybe a pF across Rf, and that slows it down.

What's your photodiode capacitance?

The lead from the pd into the opamp should be as short as possible. Even a pcb thru-hole pad will add nasty capacitance.

John

Ah no, by single core i mean single core wire, just a single strand of tinned copper connecting the photodiode to opamp - its about 1cm long, but it is in a hole so maybe i'll change that. The photodiode capacitance is 2pF max at 5v. I've added two 1.8pF caps in series across Rf and it doesn't seem to help anything. Maybe i need to add more F?

Thanks

Andrew

Maybe: Isn't the photodiode in there the wrong way?

--
Regards, Joerg

http://www.analogconsultants.com

I've done almost exactly the same thing: AD8014, 820 ohms feedback with 1.1 pf in shunt (two 2.2's in series) driven by a ~~2 pF pin diode running at 5 volts bias. It works fine, about 180 MHz net bw after a couple more gain stages. You need a very solid layout - good ground plane, good layout and bypassing - to keep this sort of thing stable. But you are dealing with a very small signal, and need lots of gain, so the whole thing is non-trivial.

You could dump the pd current into a mmic, like an ERA-5 maybe. 7 uA dc pd output corresponds to an ac signal current around 2.5 uA RMS, which dumped into a 50-ohm amp gives about 125 uV equivalent input. The thermal noise of the mmic will be roughly 20 uV RMS (bandlimited to 160 MHz), so the s/n ratio would be about 6:1, possibly usable. The mmics are very stable and easy to work with, but AC-coupled of course. And running the pd through a coax to the mmic is now ok, since it's a matched system. One of the Sirenza SiGe mmics would be even better.

Tuning (bandpass filtering) would help a lot. And optics would help!

John