4

The goal is to drive a low power LED at a very high frequency, with pulses of ~10-100ns width in order to produce a very narrow burst of light. The pulses do not need to be closely packed together, for instance a 1Hz square signal with a very small duty cycle (to match a ~10-100ns pulse) would be ideal.

I have checked with the manufacturer of the LED and the rise/fall time is roughly 2ns, which I think is fast enough for my application. I accept the LED won't be fully powered for the full duration of the pulse but this is okay as long as the light begins and ends within a very short timespan within my requirements.

Some other characteristics of my LED of choice are:

  • 20mA of max. forward current (100mA peak for short pulses).
  • ~6V of forward voltage.
  • Fairly expensive so any extra protection is probably justified.

I think my problem can be broken down into two separate problems:

  1. A fast LED driver/amplifier capable to follow a very narrow pulse.
  2. A way to generate this pulse.

One can always use an appropriate signal generator to solve (2.) but any ideas to lower cost in this aspect would also be appreciated given I only need a fixed pulse width and fixed pulse repetition rate (ideally ~1Hz).

This post is mainly concerned with point (1.). My current planned solution for this is to use a op amp with very fast slew rate (such as this one which I found referenced on this paper) in a similar fashion to what is described in this post.

I would love to hear any criticisms of this approach, about an alternative approach, or about anything I may be overlooking as I have never tackled a problem like this before.

Edit: After it was requested on the comments, the LED I want to drive is the following. The linked diode cannot be reverse biased according to the specs, but other lower power models such as this can.

lobis
  • 41
  • 3
  • 3
    Use a fast logic gate (74LVC or similar). – sh- Dec 28 '22 at 23:09
  • Could you just use an adequately fast transistor with a current limiting resistor? – floppydisk Dec 28 '22 at 23:14
  • 2
    Do you have data on the LED? 6V is not a forward voltage of any LED color I'm aware of, and that points to a multi-junction package, which might imply geometrically caused capacitance that makes turning off your LED in 10s of ns impossible. Datasheet kinda needed! Also, – Marcus Müller Dec 28 '22 at 23:29
  • All you have is a collection of source, with no explanation of how you intend to use them. Show us an explanation and schematic diagram of your approach so we don't have to guess. – RussellH Dec 28 '22 at 23:31
  • lobis, I recall reading a long summary on the state of affairs with LED driver speeds. It was ... maybe 2 years ago? At the time, even moving all the logic onto the same die with the LED couldn't get the pulse width below a microsecond or two. There are areas where the need is high and folks are spending real money on it, from what I gathered. So perhaps by now? But I think you are wading out into the deep end of the pool. – jonk Dec 29 '22 at 00:04
  • 2
    I'd have to check my notes, but I've done roughly 25 ns pulses with normal red and blue LEDs using a function generator. The trick is to pulse on to a very high voltage (>>Vf) and then quickly reverse bias the diode to very quickly shut it off. The output power will not be very uniform though, so it'll be some work if you were hoping for a rect or gaussian light pulse. – user1850479 Dec 29 '22 at 00:14
  • 1
    @user1850479 From memory the article drew out pictures [like this](https://i.stack.imgur.com/YQGuH.png). It would be very interesting to see the reverse-bias approach. How did you measure the optical output pulse? – jonk Dec 29 '22 at 00:21
  • 1
    @jonk I was using a fast photodiode and looking at the results on a scope. I think the driver in that figure is going for relatively uniform pulses, whereas if you do a short 10 or 20v pulse followed by a negative pulse you can very briefly charge the diode's capacitance above the forward voltage and get a short, irregular pulse. In my case I didn't have a pulsed laser on hand and needed to quickly test some large area photodiodes so I was testing random LEDs to see if any could be used for impulse response measurements. – user1850479 Dec 29 '22 at 00:39
  • I would also be very interested in the reverse-bias approach @user1850479! – lobis Dec 29 '22 at 00:46
  • @jonk in which ways do regular LEDs behave differently in terms of turning on/off than solid state laser diodes? Cheapest laser diode modules these days achieve gigabit/s in on/off keying! – Marcus Müller Dec 29 '22 at 00:53
  • 1
    @MarcusMüller Good question. Makes sense to me. So I'm not sure. Everything I've done is with 'regular LEDs.' And none it has been attempting to use them at the sub- (or even near-) microsecond pulse rate. It was just a recent article discussing the difficulties in some area of high speed imaging, I think. Perhaps I missed a constraint while reading it! – jonk Dec 29 '22 at 01:01
  • 1
    @MarcusMüller LEDs are different in that you don't have to achieve population inversion in the junction to have light emission, in theory a single electron could generate a photon. The output of a laser diode is therefore a lot more nonlinear, with a sudden transition from low or negligible spontaneous emission to intense laser emission. This is why you can have picosecond class gain switched diode lasers but not LEDs. – user1850479 Dec 29 '22 at 01:51
  • @user1850479 ah that makes sense! You turn off the lasing quite directly. – Marcus Müller Dec 29 '22 at 08:50

0 Answers0