I'm designing a low-cost system to record femtosecond laser pulses and characterize the average power and shot-to-shot deviations (of the laser pulses). A photodiode seems to be the economic solution, but I'm running into issues on the ADC side: the laser-induced pulse coming out of the photodiode is on the order of 10's of nanoseconds, requiring expensive ADC solutions to properly record. So I'm searching for analog circuits that might somehow temporally elongate these electrical pulses, allowing me to sample more points across the signal and accurately measure it. Simply adding some capacitance might work, but I'm looking into other solutions (peak/envelope detectors, etc.)

Currently, the pulses I get from my photodetector are about 30-40 ns wide and a few volts (up to 4V) in amplitude.

I'd like to be able to measure RMS/Average power of the laser (which is essentially measuring RMS/Average of the photodiode output) as well as shot-to-shot deviations (this is the real tough one, not sure its possible with my requirements). I want the lowest possible sampling rate (slowest ADC) to save on money. It will probably be in the 10-100 MS/s range

Thanks for your thoughts.

  • \$\begingroup\$ What is it about a capacitor (rather, low pass filter) solution that's unsatisfactory? Also, you might look into SAR ADCs, which will be better able to integrate (sum) the input from pulses shorter than your sampling period. \$\endgroup\$ Nov 6 '15 at 16:54
  • \$\begingroup\$ Hmm maybe a boxcar averager, (gated integrator). Do people still use those? \$\endgroup\$ Nov 6 '15 at 19:22
  • \$\begingroup\$ I know i've used resistors to slow the slew rate of a signal for SMPS design. \$\endgroup\$
    – Funkyguy
    Nov 6 '15 at 19:51
  • \$\begingroup\$ You use a current steering circuit. The pulse steers the current into a capacitor (which causes the capacitor to be a time to voltage converter). When the pulse ends, the current is steered somewhere else, so the capacitor voltage stops increasing. Then you sample the capacitor voltage at your leisure. Don't forget to discharge the capacitor before the next pulse comes along. I have never done anything like this, but I believe it is done in some pulse generators. \$\endgroup\$
    – mkeith
    Nov 7 '15 at 21:49

I'm sceptical that you'll be able to retain the pulse energy information through excessive amounts of low pass filtering. The more you slow down the pulse, the smaller the amplitude will get, and the lower the SNR of your sampling process will be.

As for how to measure this signal remember that the fastest scopes available actually have very low sample rates (on the order of 40 kHz). The trick is to use a fast sample-and-hold or track-and-hold circuit.

For a 1 ns aperture, you should be able to make a T/H circuit with just a few dollars worth of parts for a reasonable price. The challenge will be to syncronize the T/H circuit with your laser pulse. Pretty much any ECL logic family still available will have the timing performance needed for this, but the details of how to do it depend on what signals your laser produces to syncronize with.

  • \$\begingroup\$ Think "leaky integrator" --not "low pass filter" \$\endgroup\$ Nov 6 '15 at 19:52
  • \$\begingroup\$ That was exactly my concern with the capacitor/LP idea - though I may still have enough SNR to make it work... Thanks for the ideas! \$\endgroup\$ Nov 7 '15 at 18:15

For a crazy idea on catching super-fast pulses look up the (very old) Tektronix 545 analogue sampling scope that fires a pulse "backwards" up a delay line against the input signal coming the other way, it's mad genius. It's a 30MHz scope that can display GHz signals.

You could use a similar idea to fire a chain of cheap ADC's from a delay line or somesuch.

Edit: Can't find detail on the 545 but here's a link to Jim Williams explaining why "old [Tek] scopes are better" and a few specs: Reading Jim Williams - 3.9GHz bandwidth & 10uV per division sounds pretty groovy to me.


Phosphorescence is your friend (maybe). Wiki says: -

Phosphorescence is a specific type of photoluminescence related to fluorescence. Unlike fluorescence, a phosphorescent material does not immediately re-emit the radiation it absorbs. The slower time scales of the re-emission are associated with "forbidden" energy state transitions in quantum mechanics. As these transitions occur very slowly in certain materials, absorbed radiation may be re-emitted at a lower intensity for up to several hours after the original excitation.

The important thing here is that the energy taken in from the laser becomes translated into an "output" that lasts a lot longer - this should give you bags of time to sample it with a really slow photodiode and ADC. Think about old CRT screens that had long-persistance phosphur coatings.

  • \$\begingroup\$ That would seem to be a great solution, but I can't seem to find any phosphorescence-based sensors/detectors. Do you know on what scale these "slower time scales of the re-emission" are (microseconds, milliseconds, etc.)? \$\endgroup\$ Nov 7 '15 at 18:13
  • \$\begingroup\$ No I don't - I only thought about it because the old storage scopes of the pre-1970s used a long-persistance cathode ray tube for capturing one-off waveforms. It's time for you to brush up on your google powers methinks. Gut feeling is that there will be a material you can use but more than likely it'll be something you add between laser and photodiode. \$\endgroup\$
    – Andy aka
    Nov 7 '15 at 18:15

If your pulses are 30/40ns wide, then what about using quick ADCs, use this type of detect-peak-and-hold solution -- and resetting this one after you made your measurement?

Sure, circuit would still need to be adapted, but with quick enough ADCs, you may have an original (and flexible) answer.

  • \$\begingroup\$ This is a bad answer because you give a link to a website showing an empty box. Even if the box is not empty (probably requires javascript or custom plugins), it may go away at any time. Assuming it's a somewhat ordinary peak-and-hold, how would you measure the pulse time with such a solution? You also mention the the circuit has to be adapted. In what way? \$\endgroup\$
    – pipe
    May 3 '16 at 23:52
  • \$\begingroup\$ Thanks for asking for clarification. Thats indeed a peak-detect-hold, and since this solution was not presented so far, wanted to introduce it to OP. Link was a live simulation, though it could have been of interest. \$\endgroup\$
    – Kelu124
    May 4 '16 at 7:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.