Triggering on variable-height pulse

Okay, so I have this big nasty laser. It produces 20-ps pulses tunable over a wide range (420 nm-10 um), which is very nice. Two not-so nice features: the rep rate is 20 Hz, and the pulse-to-pulse amplitude variation is +- 5% or worse, which means it's close to 2:1 p-p. Using that with a sampling scope translates to a 25.6 second sweep time for

512 points, plus lots and lots of averaging to get the noise down to something reasonable. Say an hour per trace--frustrating.

A streak camera would be a beautiful solution--I could get a whole trace per pulse, and would need many fewer averages, but they're too expensive and take up too much room.

I have a nice Tek TDS7704 7 GHz scope coming, which will help some, although it's really still too slow--the budget didn't stretch as far as those 12-GHz monsters from Agilent.

The worst problem is how to trigger properly from a pulse train, in the face of the amplitude noise. Right now, I use a ~2 GHz photodetector(*) feeding a digital delay generator and a mixer to pick off the 8th pulse in the sequence, and stuff that into the trigger input of my 11801B sampling scope. This produces about 10 ps jitter, which is uncomfortably large given the pulse width. Small pulses get sampled later, which makes the amplitude variation worse.

I can look backward in time by about 50 ns due to optical path delay, so it might be possible to normalize the pulse height somehow before the trigger.

Any bright ideas?

Thanks,

Phil Hobbs

(*) It's a Thor Labs InGaAs photodiode running straight into a Mini-Circuits MMIC amp with no coupling cap.

Reply to
Phil Hobbs
Loading thread data ...

Hello Phil,

When I am doing this kind of stuff at clients we usually rent one from places such as Telogy.

Anyway, what exactly are you after? finding how to make the source more stable? A posted schematic would help.

--
Regards, Joerg

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

Phil, Usually these pulses are sampled with themselves. So, split it in half, delay one against the other with a either a cpu controlled linear stage or with a manually controlled linear stage and feed both beams focussed into a doubling cristal (wedged KTP or such) under a flat angle. The doubled light will exit in the center with a beautiful signal to noise. The setup is called autocorrelator. Yes, it needs some modest power in the pulse. After having read your publications about photodetectors, I can savely assume that the current solution has sufficient power in the pulse.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
Reply to
Rene Tschaggelar

Sorry, I was unclear. I have a pulse train that comes from the photodiode/MMIC, that looks like this.

^ (overall) ; ^ jit (detail)

| | | . . ___ | i | | . . / \\ | ; | | | . ./ \\ | . | | | | . / \\ | , | | | | | trg /. --^-- \\ lvl | , | | | | | |.... / -/ \\- \\ ..... | ; | | | | | | | /-/ \\-\\ L__|__|__|__|__|__|__|__\\ L___/_______________\\_____\\

0 5 10 20 30 35 30.0 30.01 30.02 30.03

t(ns) t(ns)

where the detail shows the problem with the varying pulse height--it turns into trigger jitter.

I can pick out the pulse that I want to trigger from, by using the laser's trigger strobe. I'd like to do something to the detected signal

--normalize its height, or give it an offset, or something like that, so that the trigger position is stable in time. I have probably 50 ns of delay budget in which to do this.

Thanks,

Phil Hobbs

Reply to
Phil Hobbs

Rene, thanks for the response. I don't think that an autocorrelator is quite what I want. I'm using the laser to excite my little IR antenna gizmos, so the signal I want to sample with the scope is already electrical. The problem is that the pulse train has the same nasty amplitude noise as the signal, so that the problem becomes quadratically worse--amplitude noise becomes timing jitter, which moves the trigger, which causes more amplitude noise.

I suppose there might be a threshold position in which the two partly cancel--big pulses cause early triggers which sample the pulses lower down their leading edge--but that depends on the pulse shape, which I'd like to avoid.

All my ideas to date involve peak detectors or gated integrators, which are time consuming to get right at these sorts of speeds, for me at least.

Cheers,

Phil Hobbs

Reply to
Phil Hobbs

Hello Phil,

Ok, the ability to pick out the desired pulse position covers half the rent already. What if you'd amplify the heck out of the pulse signal and then run it into a (fast) diode limiter in order not to blow out the scope's trigger input? This way the rising edge will become very steep, reducing the trigger uncertainty. It's what we do a lot in ultrasound when hunting undesired phase noise sources. Of course, you can't amplify too excessively so there remains sufficient noise margin to set the trigger level.

Also, isn't there a chance to tap off a digital trigger signal at the driver side of the laser?

--
Regards, Joerg

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

Ahhh then since I don't know about this stuff you have to talk to yourself about the predictability of the possible delays of the do-dad and something else to do with the amount of whatever that triggers the event and how much of whatever was involved in the first place and the general relationship of all that sort of stuff. Having spent some time doing that you will realise it was all quite simple in the first place.

Quite simple really.

DNA

Reply to
Genome

Sounds like you need the equivalent of a constant-fraction discriminator. The classic delay-line plus comparator CFD is just possible at your speeds, using one of the newish ADI ecl comparators.

But I'm thinking that there's also a passive network that could go between the mmic output and the scope trigger input that would help a lot. I've done a pseudo-CFD using an R-L-C network to get pulse-peak triggers over a wide range of amplitudes and some range of pulse shapes, for a 2-d delay-line imager behind a microchannel plate, imaging individual ion hits. It was just...

in-------C-------+---------comparator/zero cross | | R | | L | | gnd

which is just a partial differentiator with a bit of cheating; the comparator threshold is nominally zero, but may benefit from a bit of offset tweaking, too. The scope trigger front-end is pretty good already, so it may pay to use it. Some sort of hacked coax thing might work at your speeds, or just teeny surface-mount parts.

That photodiode is pretty slow, though.

John

Reply to
John Larkin

Weellll, sort of. It comes from a 74HC14 driven by an LF351 transimpedance amp--which is doing quite well to get 200 ps jitter. That's the signal I'm using to pick out the desired pulse.

There's a 200 MHz active modelocker in the laser, so I've thought about tapping that off somehow and resynchronizing on it, but the laser pulse occurs at the peak of the modelocker waveform, so there's still liable to be a fair amount of slop there. It would also involve can-opening the laser, which I'd rather avoid.

Amplifying the pulse and then clipping it would work fine if it were bipolar and I wanted to trigger at the zero-crossing, but unfortunately it's unipolar. That's an interesting point, though...if I used a couple of fibres of slightly different lengths, and two photodiodes connected in opposite polarities, I could make a difference signal that _would_ be bipolar, and then do as you suggest--sort of like an FM discriminator, but in the time domain. That actually might work pretty well--definitely worth a try, if I can dig up another photodiode. The scope knows which slope to trigger on, so the S-shaped waveform should be no big problem.

Thanks,

Phil Hobbs

Reply to
Phil Hobbs

Run it thru a PECL stage used as a comparator to normalize the heights.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Hello Phil,

Ouch. Those things can produce phase noise like Niagara Falls. Any chance to use more hot stuff there? AUP logic at least?

The mode-locker will spoil the broth. No chance to tap in and get a reliable signal unless you take a crowbar to it and I can understand that you don't want to. But the mode-locker will pass any of the HC14 and LF351 phase noise right on.

You could also do it unipolar, provided that the SNR in those pulses is at least 20dB or so. If you amplify and clip then the transition will be much steeper and hence the point where it reaches, say, 50% of clipping level will be much less dependent on the pulse amplitude. Of course, amplifying at those frequencies is no small feat. You'd also have to make sure that the amps don't choke.

Differential would be the icing on the cake here.

--
Regards, Joerg

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

A classic circuit that I've never seen, cool. That's close to an electronic equivalent of what I was just describing.

Cute. That would have a bit of voltage gain, too, which I could probably use.

I could probably do it with a small shorted coaxial stub, hanging right off the scope's trigger input. The bad news would be that the delayed pulse would be smaller than the initial one, but since the scope is edge-sensitive, it would probably be OK.

A barrel connector, a patch cord, and a thumbtack--just the ticket. (Some days it's great to be a physicist...like when I have the urge to use a 100 G resistor, I can do it without guilt. ;)

Yeah, I know. I hacked it together in a big hurry one day and haven't got round to upgrading it. Rings like a SOB, too.

Cheers,

Phil Hobbs

Reply to
Phil Hobbs

Hello Jim,

It's a 20psec pulse. I am not sure but AFAIR the Micrel SY series won't go much above 10GHz. Even with MMIC this is a tall order.

--
Regards, Joerg

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

I recall gain numbers like 0.25 or so; hey, gain is cheap these days. The idea here is that an ideal R-C differentiator and zero-cross detector would find the peak, but the tau would have to be near zero, gain would suck, and noise would be magnified. So a sloppy R-C differentiator (tau roughly FWHM) is used, and the L adds the extra phase shift to make it a true peak detector for a reasonably-well-known pulse shape. Beyond that fuzzy thinking, I just simulated it. In my case, a lowpass filter earlier-on in the signal path added, well, something that allowed this rlc to be tuned to work over a range of pulse widths, too. In the case of the microchannel plate, space-charge effects make big pulses wider than little ones, so that helps. I wound up with about 10 ps rms jitter over a 5:1 amplitude range, with pulse widths varying from 3 to 6 ns. I grandly named it the Equalized Centroid Detector to impress the PhDs involved.

John

Reply to
John Larkin

I was with the original ECL group at Motorola in 1962... Narud, Seelbach, Capon, Phillips.

I always kidded those fellows, "Speed kills" ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Ok. So may I assume that the 20Hz reprate is from a regenerative amplifier ? Pulse picked and divided down from a fast modelocked seed laser ? If so, then derive your timing from there.

Otherwise, there are faster diodes, I've seen fibrecoupled photodiodes delivering a bandwidth of 25 to 40GHz. Only a sufficiently fast diode is able to deliver a decent trigger signal.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
Reply to
Rene Tschaggelar

Constant fraction discriminators are interesting but cranky.

formatting link

I presume John Larkin is thinking of something like this

formatting link

which looks absolutely lovely, if you can buy it, and keep it from blowing up once you've bought it.

--
Bill Sloman, Nijmegen
Reply to
bill.sloman

I use this one, same family...

formatting link

in several products, and haven't had any problems with it. But this puppy...

formatting link

is a lot more interesting... 150 ps prop delay and 35 ps edges.

John

Reply to
John Larkin

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.