Connect with us

Fast trigger schmitt

Discussion in 'Electronic Design' started by Opps, Apr 22, 2004.

Scroll to continue with content
  1. Opps

    Opps Guest

    I built a schmitt trigger with a LM360N comparator. Thic IC should have a
    time rensponse of 13ns (MAX25ns) so, I guess, should work up to 70MHz
    (MIN40MHz). Anyway make it working at higher frequency than 1MHz look
    difficult. Any suggestion about other faster comparator or about schematics
    that optimize the performances?

    I have a signal generator with output of 2Vpp and a laser that I need to
    drive with TTL logic and frequencies up to 150MHz. My idea was to build a
    trigger schmitt which input is offset (or reference voltage) so I can
    control duty cycle and convert the sinuswave into a TTL squarewave with
    adjustable duty cycle. It works... but not so fast. I should be gald to
    reach 80MHz, at least.

    Second question, almost more basic than the one before (hem... sorry :p ). I
    need to drive LEDs... yes I know all the people asks, BUT I use very bright
    (very current consuming) LEDs and at high frequencies (up to100MHz). I do
    not have major problem (I mean, I have the schematics) to drive with pulsed
    current at very high frequencies, but I would like to have a suggestione for
    a driver configuration with at least 500mA as output, as current source,
    controlled by sine voltages -2V +2V up to some tens of MHz...


    Well, ok.... thanks a lot :)
    Ale
     
  2. Joerg

    Joerg Guest

    Ale,

    As far as I understand you app you'll need a lot faster of a comparator. Check
    out the TL3016 to see if that would suit your needs:

    http://focus.ti.com/lit/ds/symlink/tl3016.pdf

    But even with that one it might be tough to get above 80MHz. Most important is
    to adhere to RF design techniques when laying out your circuit. Another
    alternative would be to do this discrete with fast transistors. The kind that is
    used in UHF television circuits.

    Regards, Joerg.
     
  3. mike

    mike Guest

    The solution depends on a more complete description of your issue.
    There's some confusion in my mind...always...but especially now...
    about the requirement for a 100 MHz. comparator to drive a tens of
    MHz. driver. What range of frequencies do you really, really need.
    Forget about pulling numbers out of the air and decide the min and max
    you REALLY need. Getting much duty factor change at 150 MHz. is going
    to take really fast risetimes and GREATLY complicate your task. And if
    that gets degraded by your driver...

    Absent full disclosure, I'm gonna say that you're probably better off
    if you build yourself a pulse generator and forget all about the
    comparator and sine wave. Schmitt triggers are often a liability
    at high freqencies. Anything with external feedback is gonna have a
    hole in the frequency response. Feedback internal to the IC can
    push this hole up past the bandwidth of the chip...sometimes...
    You can do some pretty fast and simple
    pulsers with ECL logic.

    If you're stuck on the sine wave and don't have to go too low in
    frequency, use a really fast discrete cascode diff-amp or Gilbert Cell
    and attenuate the input voltage way down from the 4V p-p.

    If you bias the diode on slightly so the voltage doesn't move, you
    can probably get 500mA out of a diff-amp driven into saturation...
    not to be confused with saturating the transistors, which is a bad thing.

    mike


    --
    Return address is VALID.
    Bunch of stuff For Sale and Wanted at the link below.
    Toshiba & Compaq LiIon Batteries, Test Equipment
    Honda CB-125S $800 in PDX
    Yaesu FTV901R Transverter, 30pS pulser
    Tektronix Concept Books, spot welding head...
    http://www.geocities.com/SiliconValley/Monitor/4710/
     
  4. John Larkin

    John Larkin Guest

    You might try

    http://focus.ti.com/lit/ds/symlink/sn65lvds2.pdf

    as a comparator. Scale the sine wave to swing from ground to maybe
    +2.5 volts at one input and vary the DC threshold at the other. Don't
    bother with positive feedback. Should be fast enough, and the output
    swings 3.3 volts at insane speeds, perfect to drive the gate of a
    power GaAs fet. $5 for the whole thing.

    LEDs are usually slow, though. You need a laser to make really fast
    light pulses.

    John
     

  5. I can recommend the MC100EL16 ECL line receiver and the MC100ELT23 ECL
    to TTL/LVCMOS converter. It works with presettable hysteresis.


    Rene
     
  6. John Larkin

    John Larkin Guest

    Oh, of course you want the receiver, the one without the internal
    termination.

    John
     
  7. Bill Sloman

    Bill Sloman Guest

    For a slightly different problem, I built a Schmitt trigger with a
    pair of PNP 5GHz transistors.

    Ghiggino, K.P., Phillips, D., and Sloman, A.W. "Nanosecond pulse
    stretcher",
    Journal of Physics E: Scientific Instruments, 12, 686-687 (1979).

    I don't think you can buy those particular transistors any more, but
    Farnell stock a bunch of similar cheap devices in even smaller, faster
    packages. Check out the BFR92 (NPN) and the the BFT92 (PNP).

    The LM360 is a very old comparator, and not all that fast - 13nsec
    typical propagation delay. The Linear Technology LT1016 is rather
    faster, at about 10nsec, and rather easier to use.

    For faster parts, you are stuck with comparators that generate
    ECL-level outputs, mostly derivatives of the Advanced Micro Devices
    AM685, which originally offered a 6nsec propagation delay - be warned
    that its gain dropped very rapidly with frequency above 100MHz, which
    is why I had to use the 5GHz transsitors in my Schmitt trigger. Analog
    Devices subsequently produced the AD96685, with a 2.5nsec propagation
    delay, and when I last looked Fairchild had the fastest part around

    http://www.fairchildsemi.com/ds/SP/SPT9689.pdf

    with a 650psec propagation delay.

    Newark list them, but as special order items from $21 each.

    Maxim have some parts that are almost as fast.
     
  8. markp

    markp Guest

    I Have a suggestion: Get one of those fancy PLL clock delay chips used for
    clock distribution, create a phase delayed version of your fast clock, and
    AND with a fast logic gate. Vary duty cycle by changing the phase (e.g.
    select different delay elements in the feedback path to the PLL).

    Mark.
     
  9. Opps

    Opps Guest

    Well,
    I studied a bit your suggestions, thanks to all.
    Anyway I understood that there are components (and enough cheap) that I need
    (aspected from one point of view), but I relize that I am working in a
    region at which I start to have problem with the effect of RF. I need to
    work up to 100MHz.... for application up to 1MHz I basically solved my
    problems.


    The components specified to work that fast start to have socket to which I
    am not used, so I think I should need diferent soldering techinque... well,
    I will try. But now I have a doubt. I don't think I am the only one that
    need to convert a sinewave electrical signal into a TTL trigger... well also
    at that frequncies there should be something standard and commercial. Am I
    wrong?

    Instead to work the next 2 weaks (ordering, trying, retrying, reodrering
    something else take a while) to get some how a result I am wondering now if
    there is something to buy... may be more expensive that needed, but more
    fast to get (saving a lot of time-work).

    Thanks a lot
    Ale
     
  10. Opps

    Opps Guest

    Well,
    you suggested to use this very fast transistors, but they are surface
    mounting devices, and I need, if possible something more handy. OK.... there
    is the possibility that I don't have an easy solution to my problem, but I
    need to go up to 100MHz, not to the GHz and I think this make a bit of
    difference. But I got also that 10MHz-100MHz is the range at which the RF
    layouts and components start to be REQUIRED.

    So I have a basic question. Yep, another one :)
    Do you think is possible to use a couple of transistors with "normal"
    packages (well, with this I just mean no SMD/SOT....) that I can use in that
    region of frequencies? I mean, is it a without-hope-task, building a trigger
    schmitt with a couple of transistors that works up to 100MHz without
    stepping towards SMD techinques?

    I have here a table on the Horowitz-Hill. For example the couple
    2N3725/2N5022 can sink 1A up to 450MHz with a TO package. Well they have 50V
    ov Vceo... so probably it is better the use of a 2N5769/2N5771, up to
    500MHz, Vceo 12V and 200mA. Just for signal conditioning and later, I can
    use another transistor as driver if I need higher currents. Is it correct?

    Unfortunatly my electronics is nowadays very rusty, I step back to the
    "try-and-error" working manner :) but I need so some hint about the
    components to order and on the possibility that a certain idea can work :)

    Thanks again
    Ale
     
  11. Opps

    Opps Guest

    well I have two requirements/applications.

    I have signal generator that drive the gain of a ICCD and modulate the
    intensity of a laser by a AOM by which I make cross-correlation of a
    fluorescence signal... well... just to describe in a few words.

    Now I want to change light source and I have two options:
    -solid-state diode
    -LED

    The laser diode is directly modulatable and accept TTL trigger input. For
    this I want to keep my previous signal generators that are well integrated
    and computer controlled in my system. I need just to convert this sinewave
    in a square-wave TTL-compliant manner in order to drive the laser diode. The
    trigger schmitt solution look interesting because I can easily change also
    tha duty-cycle. Anyway I am working also with a 7404 IC. Well as I said the
    LM360 look more slow than aspected, but now I relize I used too high
    resistor values at the input ot the IC, probably this deteriorate the
    performance of the system. Stepping also from the prototyping board to the
    final layout I will probably able to overtake the 10MHz... BUT... I need
    100MHz. I am going to buy a 74ALS04, probably will be usefull.

    The LED. I will drive them by two configurations.
    1) Pulsed excitation. Overdriving the LED with pulsed currents signal I can
    use them in the MHz region with ns pulses. I found enough material in
    literature.
    2) Sinewave excitation. Will be more difficult to get more than some couple
    of MHz. At the moment, with a Bias-T I got 10MHz. Now I want to build a
    transistor driver. Well I suppose was a easy solution but one of the
    electronic workshop at which I asked the circuit had problem to bulid it...
    so now I start alone.
    Well i principle should be a easy task, but my electronics is nowadays very
    rasty, so I am waiting to understand which transistors are the best
    solutions.... in the mean while I am prototyping with the components that I
    have.

    Well this is the complete overview of the problem... in the meanwhile I am
    reading my old books of electronics and looking on the internet for fast
    solutions :)))))

    Thanks
    Ale
     
  12. mike

    mike Guest

    To square up a sinewave over a wide range of frequencies, you don't
    have a lot of options. If you have low noise, you can overdrive
    a diff-amp and multiply the slew rate by the amount of gain you
    have at that frequency.

    Typically, people try to build schmitt triggers. To do that you need
    gain and 360 degrees of phase shift to feed back. At high frequencies,
    the delay thru the two-port gain block adds to the phase shift. At
    some frequency, the total phase shift gets to be 540 degrees and you
    have negative feedback, and a hole in your frequency response around
    that region. If your gain and feedback are both INSIDE the IC, the
    hole CAN be high enough to be ignored. Building one with discrete
    gain blocks and external feedback usually provides disappointing results
    at high frequencies.

    One "simple" solution to this problem is to use a ONE-port device, like
    a tunnel diode. That pushes your response hole WAY up in frequency.
    You then amplify back up to TTL or embed the TD into your gain block.
    A TD can give you REALLY fast risetimes over a wide range of frequencies.

    mike

    --
    Return address is VALID.
    Bunch of stuff For Sale and Wanted at the link below.
    Toshiba & Compaq LiIon Batteries, Test Equipment
    Honda CB-125S $800 in PDX
    Yaesu FTV901R Transverter, 30pS pulser
    Tektronix Concept Books, spot welding head...
    http://www.geocities.com/SiliconValley/Monitor/4710/
     
  13. Bill Sloman

    Bill Sloman Guest

    I imagine that slower transistors in conventional packages could be
    made to work, though 100MHz might be difficult. Bear in mind that it
    took about a nanosecond from the time my pair of 5GHz transistors were
    driven into the region where they'd start switching to the time when
    they had switched.

    For 500MHz transistors this could easily be 10 nanoseconds - more if
    the base resistance is high - which isn't all that compatible with
    your 100MHz.

    The data sheets for fast switching transistors usually give turn-on
    and turn-off times under more or less realistic driving conditions -
    they are a lot longer than 1/Ft.

    A fast (ECL-output) comparator followed by an ECL-to-TTL converter
    sounds like an easier way to go If your electronics *is* rusty.
    Getting to grips with what goes on in a Schmitt trigger as it switches
    isn't all that easy or quick.\
     
  14. Bill Sloman

    Bill Sloman Guest

    That is - use a comparator. The difference between a comparator and an
    op amp is that the comparator will have two or three stages of gain,
    and no frequency compensation. As you go on to say, in the context of
    a differential amplifier, the propagation delay through these iteratd
    gain stages delays any positive feedback to the input, which can be
    awkward.

    There is also the point that the comparator gain doesn't fall off in
    direct proportion to frequency, in the way that is engineered into an
    op amp, but much more rapidly, as each gain stage runs out of
    bandwidth at much the same frequency. You need a very fast comparator
    to get gain at 100MHz.

    The data sheet of the moderately fast (10nsec propagation delay)
    Linear Technology LT1016 includes a couple of oscillators, and the
    circuits aren't recommended for frequencies above 10MHz.
     
  15. mike

    mike Guest

    Ok, you can use a comparator if you want, BUT I didn't say op amp OR
    comparator. I said diff-amp as in two
    fast-as-hell transistors with their emitters tied together thru a
    resistor somewhere between zero and a lot, optimized for gain, input
    impedance, overdrive recovery etc. depending on the drive signal
    characteristics. If you can't drive it hard enough, cascode helps.

    You could call this an op amp or a comparator, but those labels lead
    you to incorrect assumptions, as you have shown, about what the
    relevant issues might be.

    Sticking a tunnel diode in the mix makes for some
    interesting wide-band schmitt-like characteristics.
    The whole idea is that delay is irrelevant because there's no feedback
    from a delayed signal to the input. Use one-port bistable devices
    like a TD with the appropriate bias and load line.

    mike



    --
    Return address is VALID.
    Bunch of stuff For Sale and Wanted at the link below.
    Toshiba & Compaq LiIon Batteries, Test Equipment
    Honda CB-125S $800 in PDX
    Yaesu FTV901R Transverter, 30pS pulser
    Tektronix Concept Books, spot welding head...
    http://www.geocities.com/SiliconValley/Monitor/4710/
     
  16. Bill Sloman

    Bill Sloman Guest

    That is exactly what I understood you to be saying. The OP would be
    better off to buy the same transistors integrated into a comparator
    designed for the job - because it gives him less chance to make a mess
    of the design and the layout.

    Not that laying out the components around an integrated circuit
    comparator doesn't give plenty of opportunity to create a disaster.
    Trying to clean up someone else's crappy layout that was driving
    comparators into oscillation has taken up too much of my professional
    life.
    Since they can - and frequently do - oscillate at frequencies that you
    can only see on a very fast and expensive oscilloscope, this is
    usually very bad advice. I'm well aware of what you can do with a
    tunnel diode in a controlled impedance environment, but my sole
    attempt to use one to produce a sub-nanosecond edge did not work,
    where a step-recovery diode worked fine.

    IIRR Tektronix did use a tunnel diode in the trigger circuit for one
    of their faster 'scopes, back in the late 1970's, but it got
    engineered out very early on.
    If you can get it to work.
     
  17. John Larkin

    John Larkin Guest

    TDs do make superb schmitt-like gadgets, as in the collector of one
    transistor in a diff pair. Too bad nobody makes them any more (at
    least the kind that switch well... back diodes don't count.) You can
    buy old Tek plugins (or entire 547 scopes) for peanuts on ebay, and
    recover a pile of TDs from each.

    I just saw an article (EDN maybe) that somebody has fabbed a
    modern-process silicon TD... can't wait to get some.

    But the OP should use the LVDS receiver thingie.

    John
     
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

-