Connect with us

Fast rise time?

Discussion in 'Electronic Design' started by RST Engineering \(jw\), Jun 1, 2006.

Scroll to continue with content
  1. Back in '98, Electronic Design had a design tip on a homemade time domain
    reflectometer. I'd sort of like to build one up, but this one used the 74AC
    series of logic and the best it would do is a rise time of 5 ns. I'm only
    needing something to go 50 feet (15m) or so, but would like glitch
    resolution to a couple of inches. Is there a logic family readily available
    that can do 500ps to 1ns or so?

    Jim
     
  2. John Larkin

    John Larkin Guest


    DS90C402, 650 ps maybe, fairly clean step. I could post a pic to abse.

    MC10EP89 is a lot faster but the ecl levels are a bit of a nuisance.

    What sort of scope do you have?

    John
     
  3. Joerg

    Joerg Guest

    Hello Jim,

    Sure, for example the ONSemi MC10EL89. It is an ECL coax driver and its
    transition time is under 400psec. Of course, they aren't cheap ($3-4).

    You could also roll your own with ordinary BFS17A. Those can go down to
    700-800psec or so.
     
  4. Guest

    Why stop at the 2.8Ghz BFS17? Farnell's range of low power wideband
    transistors goes up to the 9GHz BFR40 which I've never used, but I did
    use the 5GHz NPN BFR93 and the 5GHz PNP BFT93. Provided that you
    remembered to put a "base-stopper" resistor (22R to 33R) in series with
    the base, and close up against it, they were pretty well behaved, and
    good for risetimes around 500psec.
     
  5. Joerg

    Joerg Guest

    Hello Bill,

    Well, yes, you can get 40GHz+ versions at Infineon. But a BFS17A is
    something that can often be found right in the lab part bins and it's
    fast enough for this job.
     
  6. Well duh, John, I hadn't gotten that far yet in my thinking (sometimes my
    desires get ahead of my brain). Seems the best I can do is 50 nS/div, so
    getting picosecond times with nanosecond oscilloscopes isn't going to buy me
    a thing. Thanks for the reality check.

    BTW, Joerg, ordered the TI little USB uC development kit this morning.

    Jim
     
  7. Jim Thompson

    Jim Thompson Guest

    I'm doing some design work for AZMicrotek and just complained that I'd
    need to increase the power to get a good sine wave at 600MHz.

    The owner replied...

    ECL = Energy Consumption Logic.

    ...Jim Thompson
     
  8. John Larkin

    John Larkin Guest

    My experience is that bipolars, even the 40 GHz SiGe parts, don't
    switch very fast, even with absurd drive. Figure that if Ft = 40G and
    beta=200, the -3 dB point on current gain is merely 200 MHz. It's
    easier to use a fast CMOS gate or some ECL.

    (and Farnell yet again?!)


    John
     
  9. Mike Monett

    Mike Monett Guest

    You may be thinking of "Build Your Own Cable Radar TDR Multivibrator", by
    Gunnar Englund,

    http://www.elecdesign.com/Articles/Index.cfm?ArticleID=6260

    If you really need fast edges, the 100EP16 will do about 200ps. See Fig. 3
    in

    http://www3.sympatico.ca/add.automation/sampler/binsamp.htm

    A simplified circuit is shown in Fig. 1 of

    http://www3.sympatico.ca/add.automation/sampler/design.htm

    But here's two gotchas:

    1. Two cascaded 100EP16's are needed to convert the ~5ns risetime signal
    from the 1MHz clock oscillator down to 200ps. One is not enough.

    2. Due to cable losses, the 200ps risetime is rapidly degraded after going
    just a few feet in RG-58 coax. If you are going to go 50 feet, might as
    well stay with 74AC. For example, see the risetime of the reflection at 20
    meters at a sweep rate of 200ps/div in

    http://www.elecdesign.com/Files/29/6260/Figure_02.gif

    Regards,

    Mike Monett
     
  10. Joerg

    Joerg Guest

    Hello John,
    At those speeds I use "semi-absurd" drive, expecting no more than
    10-15dB per stage. The good thing is they are really cheap. I had the
    BFS17A do well under a nsec.

    Of course ECL is a lot easier if it isn't for a mass product and cost
    doesn't matter much.
     
  11. Joerg

    Joerg Guest

    Hello Jim,
    Plus things get toasty rather fast. Those are the reasons why I prefer
    two, three discretes for a single pulse application like this.
     
  12. Joerg

    Joerg Guest

    Hello Jim,

    Guess then you'll have to build a little "single event correlator" that
    sees the return pulse. Or a "walking sample gate" that looks for it and
    would show you the time directly. Sampling at sub-nsec windows is not
    for the faint of heart. You'd have to use diode bridges and RF
    transformers or use something from Mini-Circuits.

    Great! You can actually go ahead and load the latest IAR Kickstart from
    TI's web site. Then you don't have to wait.
     
  13. Joerg

    Joerg Guest

    Hello Mike,
    Sometimes it's easier to just look for resonances that are inconsistent
    with cable length. That's how I once found a bubble in a coax cable.
    Being in college that didn't help much as I did not have the funds for a
    new cable run. Or let's say those funds had been invested in cerveza. So
    I kept ham radio going, hoping it'll last just one more month. Kaboom!
     
  14. Mike Monett

    Mike Monett Guest

    Sometimes it's easier to just look at the cable visually and feel it with
    your fingers. You can spot dings, cuts, crimps, pinches, staples, nails,
    flat spots, broken and cut shielding, some types of corrosion, and many
    other kinds of defects that affect performance.

    The nice thing is you already have the instrumentation. As long as you
    don't have too much cerveza:)

    Regards,

    Mike Monett
     
  15. Joerg

    Joerg Guest

    Hello Mike,
    True. However, when the cable is behind paneling and you'd have to tear
    that off for an inspection that's another story. In my case it was in
    kind of a concrete raceway, hard to pull out (I'd have had to pull all
    of them out for that).

    The "instrumentation" consisted of a grid dip meter. Heathkit, $80 or so
    as a kit. Still have it.
     
  16. You'll want to download it anyway, the version on the CD is not as
    new. Interesting that mine was marked "Made in Germany". Two solder
    bridges between pins on the large MSP controller chip. 8-(


    Best regards,
    Spehro Pefhany
     
  17. John  Larkin

    John Larkin Guest

    Yeah, I like that part... it makes a good current-mode NIM driver, and
    it's fast but not so fast that it screams too much. But a 40 GHz
    transistor isn't 20x as fast as a 2 GHz one. Gaasfets and especially
    phemts are really fast.
    We use some GigaLogic parts. A NOR gate costs $35.

    John
     
  18. Guest

    Your experience is a little unspecific. Mine was - as I posted - that
    we got about 500psec for 10% to 90% transition times with 5GHz parts.
    So who is your favourite broad-line distributor? Farnell owns Newark,
    and you can get their range in the USA even if your own tame
    distributor doesn't stock them.

    I know about the Infineon parts, but I'm less sure about being able to
    buy them - if Farnell stocked them I could have been reasonably
    confident that they were readily available
     
  19. Guest

    Is that Gigabit Logic, or some descendant? We were using their parts
    back in 1988 and the prices were much the same back then. ECLinPS is a
    lot cheaper and easier to get hold of, and almost as fast.
     
  20. My God, you too? I've got an original Eico tube dip meter and one of the
    Heathkit tunnel dippers. Still use the Heath; was going to convert the Eico
    to solid state and life got in the way.

    Jim
     
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

-