TIA-PD compensation, by eye

Yes. The input impedance Re at the emitter of the cascode is 25/Ie, where Ie is transistor current in mA. So at 1 uA, the impedance is

So when the light hits, the emitter current is low, the Cpd*Re time constant is long, and the current builds up slowly as the emitter current creeps up. Adding Rb pulls the emitter down and keeps Re low.

You can add Rx, or something downstream of the TIA amp, to null out the current in Rb.

Here's an over-the-top version:

Booting is great for slowish (MHz range) high capacitance photodiodes. It's usually not a good idea for fast stuff.

In some alternate universe where 40 ns is blazingly fast. ;) UV ones are faster than that, and if you really bomb a high-current IR one, you can get the recombination to enhance enough to get them down to 100-ns territory.

LEDs are slow compared with even cleaved-cavity lasers, which will do 200 ps pretty easily. VCSELs are much faster than that.

Cheers

Phil Hobbs

The biased side of the PD needs an RF-compliant small ceramic cap to ground. 0.1uF in 0603 package or similar, very short path to the ground plane. Short meaning a via right next to its ground pad. The RF impedance of an electrolytic cap is like roulette.

They can be pulsed faster, it's mostly a matter of the drive circuit:

You'll also have to keep all the leads very short and close to each other, ideally use coax for everything.

The onset can be quite fast, just not the trailing edge:

When I pulsed LEDs I never tried it with the really old ones I still have from my teenage days (when I still had hair on top). Maybe I should. Those need at least 10mA to be reasonably visible during the day.

Phil is the expert on that stuff. He might know tricks to milk more BW at such high current/voltage ratios.

IMHO you should not force all that gain into one stage. It may be better to reduce to 1-5k in the TIA and then follow with a regular amplifier stage to do the rest.

This is the best we've done with a butterfly-packaged fiber-coupled laser:

To get that sort of pulse, you have to tease the pulse width and amplitude to get a sort of optical avalanche effect. Lots of lasers have so much inductance that they can't go fast. Or they were intended for telecom, and the impedances are wrong for pulsing.

Here's the driver, which has a built-on one-shot to make the pulse width:

Leave it at 1k and place an amp behind the TIA if more gain is required. Using high resistance in a fast TIA makes no sense.

True, _provided_ that the minimum photocurrent you care about is at least 100-200 uA.

Cheers

Phil Hobbs

I guess I'm talking/thinking about adding some driven shield plane around the inverting input, try to block some of the ~1.5pf of FB capacitance. 1MHz is fine, I could use a dual opamp.

George H .

Hmm, I can try adding a ceramic to the bias supply. There's no ground plane in this circuit. Well the whole thing lives in a grounded aluminum box, ~1"x1"x2" maybe a bit bigger. Inside, capacitance is not your friend at 1 meg ohm. (I've been measuring this outside of the Al box, I wonder if it will change by much inside?)

George H.

Right, that's what I'm doing at the moment, more gain down stream.

In theory you have the least noise with the most gain at the front end. In practice you don't always need the least noise.

George H.

Your 'scope pics show a ringing period of about 65nS, so, a 7-foot coax's lack of termination isn't the whole explanation.

Are you driving the coax with your TIA? It might not like that.

Cheers, James Arthur

Yikes! Then it's almost a wonder that a 200MHZ amp runs in there. Although a 40MHz scope won't necessarily show you if it does tarantella dances.

The RF behavior can't be ascertained well from outside. Or even inside, without a ground plane. Trying nanosecond stuff that way doesn't make much sense, I'd rebuild that with a ground plane. For prototyping a piece of copperclad will do.

With 1k in there the difference is so miniscule you couldn't even measure it.

Depends on the desired dynamic range, and use a low noise amp in the first stage. Most of all, build it with proper RF guidelines in mind and that's where this project could use some upgrading. Otherwise anyone with a GSM phone walking around could disturb the signal especially if the amp has BJT structures in its input area.

Hi Joerg, (others) Here's a couple of pics.

James, yeah the opamp TIA drives the coax too. It was oscillating happily when I had only 330 ohms of TIA R. :^) But it seems OK when terminated. So as you say, the ringing is due somewhat to the cable capacitance.. I guess, I'm not really sure how to think of 7' of coax.. transmission line or capacitance?

Ground planes and such. Joerg I'm not sure, what I'd like is a big ass photodiode at 1 M ohm gain and 1 MHz BW. (typically ~1 uA of current.)

I bought a couple of THS4361's so maybe I'll build up a pcb w/o the switch and see what that look like.

Oh the numbers I post upstream are not right. I'm a little bothered by the high gain end, cause I think I was getting more (HF) gain there with a slower opamp, so I need to redo this more carefully.

Gain R 3dB

George H.

Build on solid ground plane with good RF bypassing.

Use Manhattan for non-critical components.

Air wire the sensitive node.

Do you really need to drive the coax directly? Can you add a series resistor to reduce the loading? Any ground bounce from poor supply bypassing can give feedback to the inputs.

The datasheet Application Information section has a lot of useful information on feedback resistor selection, layout, bypassing, alternate configurations, etc.

They definitely do not recommend a two-sided pcb.

That is not likely going to work reliably. You can get away with double-sided PCB but one side has to be a solid or almost solid ground plane.

That is one of the telltale signs that the whole enchilada is teetering on the edge when it comes to stability.

For coax you need a staunch driver. I believe the lowest load resistance the THS4631 is supposed to drive is 100ohms.

You can leave open a small island around IN- but this sort of requirement asks for Phil's medicine. Cascode and such.

Yes, and with a solid ground plane.

At 1M it is bound to fall apart. You might also want to look for amps with lower input noise so you don't need such high feedbakc resistor values.