TIA-PD compensation, by eye

Hi all, months ago I bought this screaming fet opamp (THS463) to gin up an old photodiode circuit. Put it in today, and the step response doesn't look right. Here's 1 k ohm of gain w/ 10 pF across it.

formatting link
A lot of wiggles, but the amplitude isn't big enough.

Here's the uncompensated step.

formatting link
(note time base change)

It's only a 40 MHz 'scope so I'll try something faster. Maybe it's just the LED I'm blinking isn't fast enough... Chan. 1 (upper) is the voltage into the LED... Or my voltage step isn't fast enough.

I could try a step with a 74HC14... hmm there was some flavor that was faster than the 'HC14?

George H.

Reply to
George Herold
Loading thread data ...

Georrggggeeee, have you been using the white solderless breadboard again?

Grins, James Arthur

Reply to
dagmargoodboat

I was going to say ground lead oscillations, but that is a low frequency.

Cheers

Reply to
Martin Riddle

If this is a classic vanilla TIA, it may be the photodiode capacitance at the summing point making it ring. A bigger feedback cap, or a cascode thing, can fix that. I don't know why the amplitude should change.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

They aren't exactly my definition of screaming fast. Screaming starts around 2GHz.

Then something fundamental must be off. Wrong resistor?

Yes, 74AUC1G14, it's a single gate version with sports exhaust and all.

It could be something like that or maybe there is a tortilla-sized circuitry connected to IN-. George, can you post of photo of that whole lab bench setup?

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Grin.. no but its a PD circuit I did ~15 years ago, big switch capacitance. I'm tuning out that and the PD capacitance.

George H.

Reply to
George Herold

Right, Sorry, classic vanilla TIA-PD. I'm just picking the tuning cap by eye... shape of the step response. But the shape doesn't look right. I've tuned a few TIA's with slower opamps and maybe more C on the inverting input. It has this classic two pole shape. You can add more feeed back C and pick your Q by eye. This doesn't, so it's my light pulse or my electronics. I'm mostly stuck with the old pcb layout so I'm just trying to make sure my light source is not the problem, or fix it if it is.

Instead of an led, I could wiggle a diode laser, that's less convenient. Right now I'm driving the led from a

20 MHz Rigol func. gen. Maybe that's the slow spot? George H.
Reply to
George Herold

Well it's all relative, the previous opamp was a opa124.

I don't suppose that comes in a dip? There was a TDR paper Steve posted with a fast(ish) '14 in a dip. (I'll find it.)

Grin, well switch sized. At least I got the right side of the rotary switch connected to the inverting input. (sometimes it's enough to be lucky.) It's a hack/mod of a pcb I did years ago. I'll post a pic. I'm driving a long piece of coax too... (details details, It's sometimes hard to know what's important.)

George H.

Reply to
George Herold

Save you some time:

formatting link

Youtube video:

formatting link

A 40 MHz scope is useless for these frequencies. You need 1 GHz or more. A Rigol goes for 8k or so, but a good sampling scope on eBay can be much cheaper if you are lucky. Or maybe you could pick up a used PicoScope 6000 series.

2ns risetime is definitely 50 ohm coax territory. 6" scope probe grounds will ring like crazy. A 499 ohm resistor in series with 50 ohm coax, terminated at the scope with 50 ohms, will give a 20 dB attenuator which will have negligible effect on the signal. Ground the coax shield at the source. A ferrite toroid over the coax will also help reduce common mode ground bounce.

This is also ground plane territory. It is a bit frightening to glue a $50 ic to copperclad, but do it live bug so you can salvage it later. Really good grounding and bypassing is necessary, along with zero length leads.

Read the Application Information section of the datasheet:

formatting link

Reply to
Steve Wilson

Thanks Steve, I knew I had that page bookmarked at work. Well I built a fast(er) edge with a 74HC14. No change. But the problem was operator error. My ~7 feet of coax was un-terminated. 'Duh' Dope slap!

George H.

George H.

Reply to
George Herold

Don't kill yourself. You are a physicist. There is no way any of us could compare with you in your own field. Electronics has a bunch of gotchas that take years to learn. This is why we are here, so you've come to the right place.

On high frequency oscilloscopes, there might be another method that may interest you. The Rigol DS1052E is a 50 MHz two channel scope that sells for around $349. A german hacker found that removing the input bandwidth filter in front of the ADC chip would raise the bandwidth to around 450 MHz. Unfortunately I did not save his page, and I'm having trouble locating it again. I happen to have two of these scopes along with a bunch of others, and it would be very interesting to do the mod.

There is another hack that involves changing the firmware to convince the DS1052E that it is a different scope. This changes the input filter bandwidth to 100 MHz and sets the maximum timebase to 2 ns instead of 5 ns.

A different version raises the bandwidth to 150 MHz.

The hack is a bit dangerous since you can brick the scope if you make a mistake. And current revisions have to be downgraded to an earlier version before applying the hack, then upgraded back to the new version. More opportunity to brick the scope. So I probably won't try that one.

But I'll let you know if I find the 450 MHz page.

Reply to
Steve Wilson

Grin, I don't like compliments, mostly 'cause they're not true. Physics is no harder or easier than other fields.. it's mostly just a choice of what you like and are good at. I'm sorta a mid grade physicist, plenty are smarter and others less smart. I guess the one thing I pride myself on is trouble shooting, which I think I do well. Forgetting to terminate a fast signal, then seeing ringing, (a big hint) deserves a dope slap. It's fine though, my other talent is making mistakes. :^)

I've got a 50 MHz Rigol, and I've seen the hack that can make it 100 MHz, I haven't bothered. We have a 200 MHz TEK if I need it. If I ever get around to playing with the TDR stuff, I might need a faster 'scope.

Speaking of 'scopes I noticed the new low end Rigol's now have separate knobs for gain and offset for each channel. (Good move)

formatting link

And 0.5mV/div gain, up to 8,000 number of averages. Those would be nice. Maybe next year.

George H.

Reply to
George Herold

You may be stuck with an old pcb layout, but a simple cascode transistor can really isolate the pd capacitance from the TIA amplifier input, which can radically increase bandwidth and reduce the feedback cap value.

That is half of Phil's bootstrap cascode.

I recently tried a simple NPN cascode between a photodiode and a fast opamp. Risetime was 450 ps, about 5x better than the vanilla circuit.

I also tried a cascode into a mini-circuits MMIC. That didn't work.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

LEDs are ok. They can be pulsed blazingly fast.

That would be slow. Just run it through a fast gate driver or the Schmitt single inverter I posted about, that'll speed it up. RF transistors work as well by slamming the current from the generator into the base. The gain that this provides can results in more than an order of magnitude of speed-up but the turn-off will be sluggish unless you use a gold-doped part like a 2N2369. Since fast transistors are usually available in a parts rack or can be pilfered from scrapped gear you could save the wait time for an order.

However, you already found the ringing issue in your unterminated coax canle. Later you might work on that rotary switch connected ti IN- which probably adds a lot of capacitance.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Which leaves the question which photodiode you used and how much you pre-biased it. That can have a major impact on overall bandwidth.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Once I terminated the coax, I didn't need any FB caps!(?) gain from 0.33k to 10M ~15MHz BW w/1k, but the gain dropped off 'too fast' above 33k ohm I'm reproducing this data from memory... data book's at work, (I'm 1/2 making up the numbers above 100k where it dropped off as a single pole.. the 'stray' feed back capacitance I assume.)

Gain R 3dB

1k 15M 3.3k 9M 10k 7M 33k 4M 100k 1.5M 333k 400k 1M 130k? 3.3M 40k 10M 15k?

Putting in some numbers it looks like maybe

1 pF a bit more, (I might have made a small math mistake.) ..but that's about what I would guess.

I should do a new layout... I've been sending off a bunch of proto's to advanced circuits. I could stick a little pcb in the corner or on the end.

'thinking about my pcb layout.' I might have some traces to the feed back R's lying over each other... a razor blade might make it better too!

Right, I figure I'd try the bootstrap 1/2 first. (I've got 100 of Phil's fav. jfet that's eol.) Does the cascode do anything weird at low currents, (besides no work as well.)

George H.

Reply to
George Herold

Some visible ones are not so not so fast. The NIR led I used says 14 ns rise and fall times, (870 nm?) I'm sure there are faster ones.

George H.

Reply to
George Herold

Thanks Joerg, I took pic's of my setup/pcb, but didn't post them. I did make a fast edge with a 74HC14, dip, all six in parallel. (now I've got two led /PD test jigs. :^) I'd mostly like more BW at high gain 1M ohm*, 100k is ok. seems like stray feedback capacitance is the current problem.

George H.

Reply to
George Herold

One normally adds a resistor to donate a bit of extra current into the cascode emitter to keep the emitter impedance down at low pd currents. That makes a DC error at the TIA output, but that's easy to subtract out.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

osi-optoelectronics 44pin (there might be some letters in there. or pin44) IIRC ~100 pF at 10V, mines biased to 12, via 10 ohm and

100uF Al to ground... that's hacked in too... Well, you're going to have to wait till Monday for the pics.

George H.

Reply to
George Herold

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.