Photodiode TIA

Look carefully at the scope trace. It sounds very much like you're oscillating at several frequencies simultaneously, which looks superficially like random noise. I've seen this several times with poor layouts.

If you can see any kind of patterning at all, you've almost certainly got oscillation at several frequencies, not noise. 200mv (as you mentioned previously) sounds like altogether too much for real noise with this little bit of amplification.

John Perry

Well, if you want to get all picky and technical on us, well, yes.

John

C'mon, Joerg, that's the key to low output noise.

Cheers,

Phil Hobbs

The diode is connected to conduct current ( Anode at +5V, Cathode at lower potential). This results in high currents with high current noise and a high input capacitance, resulting in high noise amplification.

That's probably not the key to low output noise...

It results in a totally railed amplifier, which will output only the supply rail noise. You have to have quiet supply rails in a sensitive front end, so the output noise would be low.

That's a joke, son. Jes' a joke.

Cheers,

Phil Hobbs

If there is a spectrum analyzer and nobody is using it (as if that ever happens...) Andrew could run that on his bench with a 1ft piece of wire at the input hanging in the air. That'll show when stuff starts to howl. Can be a cheap EMC analyzer. There'll be lots of TV stations etc. but you can see out of the corner of your eye when a "new station" pops up. If a wee hand waving makes it move one can be sure that this ain't the David Letterman show ;-)

Good point. I don't know which scope probes Andrew is using but most of them make a pretty good antenna. If that antenna gets to be too good the input of the TIA might snap at it like a mountain lion. Then there is probably a ground clip lead with lots of inductance of unknown magnitude.

Andrew: There are two good ways to measure delicate RF stuff in small spaces. The best one is a FET probe (I use the Philips). These come with a bunch of very tiny jacks that can be soldered into place, for example at the TIA output. This provide a nearly leak-free connection to the scope. If you want to be extra good slip a #43 toroid over it right near the tip.

If you don't have a FET probe and are in a hurry you could take thin

BTW it's good policy to provide your circuit and layout with such RF style test points where a 442ohm resistor or the like goes to a small pad, with a ground pad right next to it. Later when everything is debugged and cost really matters you could ECO it with those resistor spots vacant.

OK so i'm a little embarrassed, turns out that most of the "noise" went away after i added decoupling capacitors - yes i know they should have been there from the begining and i apologise for wasting time!

OK so now i have two stage amplification working quite well at DC frequencies, with no Cfb. It doesn't go into oscillation, but that might just be a fluke, i intend to lay down pads for Cfb in the final PCB. The first stage as per my last schematic (with photodiode other way round!) feeding an inverting amplifier with a gain of 1000, i'm getting a very healthy output (100's of mV).

My next question is how do you go about interfacing this output to Logic / PECL inputs. I've tried to find a comparator but most don't seem to work that fast, unless i'm missing something? My theory was to feed the signal into the comparator and have that interface directly to the logic - is this how it is normally done?

Thanks

Andrew

The gain-1000 stage can't work at 160 MHz. The amplifier gbw would have to be 160 GHz.

John

I'm still trying to fully figure out these CFB opamps (i'm using another AD8011 as the second stage), i've read that the bandwidth depends only on the feedback resistor, but i'm not sure by what relationship. Will my initial transimpedance stage using an AD8011 go up to 160MHz? Which parameter do you need to determine this?

I guess the easy way out is to use a very high bandwidth as the second stage to get my gain. A gain of 100 actually gives me around 200mV with the devices aligned properly at 1cm. I've seen some opamps with bandwidths up to 4GHz (OPA847) that could help as the second stage?

Thanks again

Andrew

Rather read essentially...

Have a look at National, TI , AD websites. They do have some good appnotes about the CFB concept. There's nothing difficult about those, just a few important things to grasp.

With a 4GHz opamp your considering 'hot' parts, and layout, bypass,... are much more critical that with a 8011. I wouldn't even try without a carefully designed PCB. Plus I don't think you'd have to resort to that sort of expedient.

I'd first consider carefully studying the appnotes and only then decide what you want to do. And should you decide to give the 847 a try TI probably have a test board. If you're for a one off, this is probably a wise solution.

Or get someone on board that knows RF stuff, at least for a few hours. RF engineers can make hotrod opamps work on copper clad but don't expect it to look pretty.

I was just drooling over the OPA847 myself but now drifting towards the THS4021. Fast 'nuff and it looks like it takes the +/-12V rails without having to regulate down.

CFB amps have a very low input resistance on the inverting input, which changes the gain vs feedback behaviour. If you imagine taking a (say)

Yup these are really nice. I've used them as a front end of a high precision 100kHz IF amplifier for a precision VNA and it really did wonders.

of

he

e=2E

nt

to

Thanks for the advice - i've been reading this app note

Which goes into quite a lot of detail as to the differences. Anyway it also states the bandwidth (or corner frequency) of a CFB is only dependant on the feedback resistor and two internal parameters of the device. Now on the front of the AD8011 datasheet it shows with a Rf of

Andrew

For a 100kHz amp? That's almost like using a Maserati to go grocery shopping ;-)

I even had internal self-calibration and other niceties. I said precision; that was for sub milli-degree phase accuracy.