Connect with us

Narrow pulse generation

Discussion in 'Electronic Design' started by Anthony Fremont, Mar 9, 2007.

Scroll to continue with content
  1. Hello,

    Anyone have good ideas on generating 5nS pulses into a 100ohm load, but the
    load can be anything from an open circuit to a dead short. 5 to 10 Volts
    would be great. It's not a beauty contest on the shape so I want to do this
    cheaply using generic parts I have laying around if at all possible. I
    would want to be able to trigger these repetively at up to say a few hundred
    Hz. It's for some TDR tinkering.

  2. Guest

    Seems to me any logic gate can drive faster than that, but you might
    not get the voltage swing you want. Raises the question of why you
    want such a high voltage? With such a slow rise, you might want to
    test very long cables?
    TDRing nearby typically uses 25pS rise-times with about 100mV swing.
    But I think with a typical logic gate you can see all the classic TDR
    cases with a few feet of coax for sure.
  3. Guest

    You can use the output of some waveform generator (or a astable
    multivibrator) to feed a high pass filter (a simple C-R with a diode
    in anti parallel to R to avoid negative pulses). Then the filter
    output feed to a schmitt trigger like 74hc14. The pulse wide is
    controled by the values of R and C. Easy and cheap!!

    Good luck
  4. Thanks for your reply.
    The voltage I chose was arbitrary, I was just thinking of something battery
    Those rise-times would be nice, but not really necessary for me AIUI. I'll
    just be using an average scope for the monitoring so anything should work
    for tinkering purposes.

    Since I'm not very clever with analog stuff, how can I make the logic gate
    switch on, wait a few (adjustable would be cool) nanoseconds and then switch
    back off on command?
  5. Andrew Holme

    Andrew Holme Guest
  6. Joerg

    Joerg Guest

    Figure 4 shows the basic method of how it's done:

    Of course you have to use much faster logic and then amplify it. Or use
    tons of parallel buffers and step up via a stripline transformer.
    Sourcing 5-10V into a dead short might turn out to be a challenge
    though. At least I'd ask the local uitility to spool up their peaker
    plant and warn the fire district before moving that big switch. A notice
    to the FAA would also be good so they can alert pilots that this big
    bright flash is not a UFO or meteorite that has crashed ;-)
  7. Rich Grise

    Rich Grise Guest

    You don't do TDR with a pulse - you do it with a step; you reset the step
    when the last reflection gets displayed (which will depend on the length
    of the line). That gives you a nice graph of the cable's impedance along
    its length. A pulse will just differentiate that and smear out all of the

    Generating the step is left as an exercise for the reader? Heck, it's
    just a step generator - you just go from 0 to 1, and the risetime will
    depend on the chip, when a trace is done, reset to 0 until all of the
    reflections are back, and give another step. At 1' per nanosecond, you
    probably won't have to wait long. :)

    Good Luck!
  8. John Larkin

    John Larkin Guest

    Electrical TDR usually uses a step, not a pulse. My TDR scopes all
    seem to use a square wave, 100 KHz ballpark, to generate a repetitive
    step, although you might go slower to shoot out really long cables.

    An AC-family logic gate should work.

  9. LOL You don't think I'll have trouble keeping the voltage up do you? ;-)
    I guess I better allow for that then. I would like the output impedance of
    the generator to be somewhere in the vicinity of 75 Ohms, adjustable from 50
    to 100 would be ideal. Thanks for the link. :)
  10. Joerg

    Joerg Guest

    That would certainly mean some paralleling or transformer-based
    combining so you can have fast logic and boost up the voltage. But as
    Rich already mentioned TDR is usually done with a transition, not a pulse.
  11. Joerg

    Joerg Guest

    I'd go all out and use PECL or something like that.
  12. Guest

    I don't know the details, but when I was doing video speed analog, we
    used HP ATE. Anyway, it tuned itself up using a pulse that they
    generated with a diode specifically designed for flat pulse
    generation. If you do a patent search, Tek and HP have many pulse
    generation circuits.

    This particular HP circuit was designed into every DUT board.
  13. That certainly sounds easy enough and I happen to have some 74AHC04s. Will
    those work ok?

    And now, how do cheap micro based TDRs measure those nanosecond response
    times? They obviously aren't using a CCP module to time them. There must
    be a cheezy trick of some sort. ;-)
  14. John Larkin

    John Larkin Guest

    Should. The source impedance of the tdr step should be 50 ohms. Put
    several gate sections in parallel to get a stiff source, and then put
    a series resistor, 47 ohms maybe, to make a net 50 ohm source. Run
    that to a tee connector at the scope input (assuming a hi-z scope) and
    the other tee port is the tdr thing.
    Dunno. The old Tek 1503 (?) handheld used a classic diode-bridge
    sampler. Nowadays you could just trigger a fast adc, once a tdr step,
    with the adc trigger slowly walked in time to make the range sweep.

  15. By 'fast adc,' I think you are addressing yourself to the sampling
    window (which needs to settle quickly) and not so much to the length
    of the actual conversion time. That may need to be made clear to

  16. John Larkin

    John Larkin Guest

    Right. You only need to take one sample each TDR edge, and slowly walk
    the sample delay to build the waveform. You can even take multiple
    samples at the same delay, over several shots, to average out noise.

    No rush.

  17. I can understand that, you just build up the information over many pulses
    one piece at a time. My question then becomes, how are you "walking" your
    samples so precisely? What is generating the time delays?
  18. Guest

    5nsec pulses don't sound too hard - we got 0.5nsec FWHM at 5V into 50R
    and 7.5V for anything longer.

    We did use a fairly expensive HP RF transistor as the output stage.

    On the way there I got 800psec FWHM out of a bunch of dead cheap BFR96
    wide-band transistors.

    Farnell doesn't stock them any more, but you might like to look at the
    Philips BFR106 or the BFG590, which are listed as limited stock in my
    Farnell catalogue. The BFQ68 looks like a keeper and is a lot sexier,
    but contains berylium oxide and costs a lot more.

    We drove our transistors in long-tailed pairs from 100k ECL using
    BFT92 or BFT93 PNP parts as level shifters - modern CMOS would be fast
    enough for your purposes, but note that the thin base-emitter
    junctions in wide-band transisors often break down when reversed
    biased by more than 2V (check the data sheet!).
  19. Joerg

    Joerg Guest

    For example, you can build an LC or RC section that scoots your bridge
    trigger time sliver by time sliver. You can use a varicap in the C or
    make the R variable. The latter is what I usually do. Early on with the
    SD5400 but when that got expensive I used dual gate FETs for TV tuners.
    Those cost only pennies but have very low inherent capacitances. Low
    stray capacitance is very important here. You can easily scoot a few
    hundred psec. Personally I avoid digitally adjustable delay lines for
    noise reasons.

    That SD5400 was very handy. Since contains four matched devices I could
    servo it an literally steer resistance. Very smooth. But all good things
    will some day end <sigh>.

    BTW the scope connection in Hi-Z that John mentioned needs one word of
    caution added: Some of the glitzy "modern" digital scopes have the nasty
    habit of reducing their bandwidth when cranking up the channel gain.
  20. _Sounds_ like a piece of cake, thanks. I should have most of what I need
    laying around here somewhere. ;-) I'm not sure my old Hitachi scope is
    going to be up for it, but I should see something with 100' piece of Cat-5.
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day