Fast And Gate, 1 Ghz

Of course this is all 50 ohm differential stuff but I'm sure you can figure it out.

Mention my name ;-) ...Jim Thompson

Seems like a job for a single-slope integrating A/D. Integrate a current onto a cap while the pulse flies, then de-integrate at your leisure in-between pulses.

-- Cheers, James Arthur

How about a diode mixer? They come completely specified for DC drive on one port, and it'll sure be easy to buy one completely built and connectorized.

....

how about the ECLinPS family and its variants...?

Jure Z.

The Eclips Plus gates would be about right, if you really need gates. ON and Micrel make them.

But one of the Analog Devices fast (ecl/pecl) comparators might be all you need. It could discriminate the photodiode signal, and you could use its ECL output and a couple of transistors to steer a current into an integrating cap to give pulse width.

If you count a 2 GHz source, the resolution will be 500 ps, so you'll only get 8 to 40 ticks. An analog trick could give lots better resolution.

John

Fully agree, this really calls for the analog integrator. With a 2GHz source you'd also get tons of "amplitude jitter". Sometimes it catches three pulses, sometimes only two, stuff like that.

Hey, I can see your posts again!

John

I have no clue why they didn't make it through your ISP before. Not sure which news provider you use but sometimes ISPs or newsserver companies take the liberty to hose off posts according to what they feel. I guess we can thank companies like Google for that :-(

The most gross blanket hose-off happened a few years ago when the largest German ISP obviously chucked all Pacific Bell domain emails. I had calls from Germans, desperate for answers from American business colleagues that didn't get through. That's when it finally also dawned on them that ditching the fax machine was not a good idea.

How accurate is this integrator method with regards to repeatability? I would imagine any simple implementation would be rather inaccurate? (due to temperature issues, etc...)

That was my thought. Or just integrate a current into the cap while the pulse flies, then measure the resulting voltage with a plain ol' ADC.

If you did it carefully, I'd think you could get repeatability of 20 ps or less for a 20 ns full-scale. Steer a good current source with fairly fast transistors, and use an NPO cap, follow with a good low-bias-current opamp.

John

ds

etting

Sure. That's just as good and what with today's uC's and a/d's, possibly more convenient.

-- Cheers, James Arthur

That's the classic pulse stretcher, used to get picosecond resolution edge measurements. That suggests an interesting and cheap circuit to solve the OP's problem. But my camera broke so I can't post it now.

John

But I found it hard to convince folks to do it that way. Nowadays many seem almost hell-bent on doing it digitally. Maybe because the analog world is too close to voodoo for younger engineers ;-)

It's tough to do picosecond time measurements all-digitally, although it has been done.

How about a semi-logarithmic pulse width detector:

Input comparator steers a current blip into a capacitor. The cap has a resistor to ground and a comparator working against some small positive DC bias. An input pulse charges the cap, voltage proportional to pulse width. The comparator tells you that you had a pulse, and the comparator pulse width is a (nonlinear) reflection of the much faster input pulse width; measure that with a reasonable counter.

That's pretty simple and doesn't need a pretrigger or reset or separate pulse-start detector.

You could do the same thing with a constant-current discharge and a negative-swing catch diode and get a linear stretch, but the catcher should be temp compensated for best results.

---- | | in | |

-------------- ------------------------------------

________ / \__________ / \___________ / \_________ / / fast charge slow discharge /

-------------

___________________________________________ | | | comp output | _____________| |__

John

Yes, a "time stretcher" like this could get it into the realms of regular speed digital counters. FPGA and such.

However, since Steve wrote about 1-4Hz PRF I don't see anything wrong with the classic integrate and hold approach. You've got tons of time to read it out, issue a text message that a pulse has come, turn on the coffeemaker ... :-)

I'd do this analog -> ADC.

Sure. You'll need to trigger the ADC and discharge the ramp. That's all easy if there is a reasonable event pretrigger available.

John

Why does it need a pre-trigger event? Start two ramps when the leading edge of the pulse arrives. One stops integrating into the little cap when the short pulse is over. The other keeps integrating into another cap until a threshold is reach and that is then your "timer". Or just skip the whole timer thing and have the ADC triggered by the falling edge of the laser pulse and discharge the one cap if the caps don't need much tempco correction. Basically needs a simple delay circuit that makes sure the ADC has done its acquisition job before the cap is reset.