Connect with us

Transistor Selection

Discussion in 'Electronic Design' started by Tim Williams, Jul 14, 2007.

Scroll to continue with content
  1. Tim Williams

    Tim Williams Guest

    It just occurred to me, how the hell do you select transistors?

    Digikey has a couple brazillion in stock and no way to select more than the
    most fundamental characteristics of their stock! Mouser is no more helpful.
    Where do you go?

    In particular, I thought I'd look for high speed BJTs in the 300-1000 MHz fT
    range, with V, I and package similar to a 3904 and such.

  2. DJ Delorie

    DJ Delorie Guest

    I'm usually not picky, so... narrow it down to the Ice you want and
    basic functionality (NPN? dual? built-in bias resistors?). Then I
    usually pick the smallest SMT package of what's left, then pick
    whatever Digikey stocks the most of (i.e. most popular).
  3. Fred Bloggs

    Fred Bloggs Guest

    You use something called a *selector* *guide*, the OEMs like ON,
    Zetex, and Fairchild have them.

    guide: n. Something that serves to direct or indicate.

    select: To take as a choice from among several; pick out.

    So you want to use this "something that serves to direct or indicate" to
    select as in "to take as a choice from among several; pick out" the one
    you deem best suited for the job. You may even SPICE it in the
    prospective circuit.
  4. I would be surprised if you had a very big list to search
    with that package and fT. TO-92 isn't much use at 1000 MHz,
    is it? I think you also have few choices that include both
    that high fT range and the voltage capability of a 2N3904.
    most of the fast transistors have low voltage and current
    capability. I doubt you will find a 40V 100 mA rated TO-92
    with an fT>300MHz.

    i.e. Data/MPS918, MPS3563.pdf Data/MPSH17.pdf Data/MPS5179.pdf
  5. You get on here and ask a few thousand of your closest friends what they
    use. Most of us use the least expensive device that will do the job.
    My low power 50-500 MHz. RF device of choice is the gold-doped 2N5770. You
    will be pushing it very hard at a gig with the TO-92 package but if you know
    what you are doing with matching devices it isn't too difficult ... and the
    usual PCB layout with splayed 1/2" leads isn't going to cut it much above
    quarter of a gig. The base lead is the bitch of the bunch as you want
    minimum L for maximum bandwidth -- component side short and fat traces; the
    collector isn't quite as bad because you will probably be having some L in a
    tank circuit there anyway. The emitter is even easier, as you simply choose
    an emitter bypass to either give you gain less than maximum, or for maximum
    gain, simply series resonate the lead inductance with the package inductance
    with the capacitor lead inductance at the frequency of choice.

    The venerable 2N918 isn't too bad either, and the 2N5179 comes in a rather
    distant third. Them's my choices.

    Did that help?

  6. Guest

    Work out what you want the transistor to do, read a few selection
    guides, and see if Farnell stocks anything that looks as if it might
    do the job.
    Manufacturer's data sheets, though they can be a bit sparse for
    individual transistors. National Semiconductor's transistor data book
    had an interesting section at the back, which gave more detailed
    information about various transistor processes - you could find out
    the process used to make a specific transistor and use the process
    data sheet to fill out the information that wasn't specified on the
    device data sheet.
    What do you want to use it for? Why do you want to buy a high
    frequency part in the antediluvian low frequency package used for the
  7. Tim Williams

    Tim Williams Guest

    A little bit, thanks.

    For example, I hobbled together a 20~50MHz differential cascode stage using
    ancient 2N3646's -- the mushroom shaped things. They happen to have a date
    code of 71xx or so. Evidently they are somewhat fast, 350MHz compared to
    the 100-300MHz of 3904, etc. So that's about 1/10th fT, which I would
    certainly consider fast (especially for a mere breadboarded circuit).

    A shame I have to make estimations of bandwidth based on rise time using a
    square wave that's not particularly fast, so my estimate is 20-50...

    I can push 4401/4403 into the 30ns range, but that's no good if I'm to
    explore something like the FM band. It'd also be fun, probably more out of
    bragging rights than anything remotely practical, to say I can make discrete
    logic in the single digit nanosecond range, too.

    I've pulled some UHF transistors from a bunch of high-def monitors. These
    things used a cascode/emitter follower video amp stage: PH2369 / Philips
    2N2369 in TO-92 for the bottom, BFQ225 (I think), a TO-202, 100V transistor,
    and a complementary (two diode biased ;) emitter follower of BFQ221/241 to
    drive the grid or cathode or whatever it was. I've tried making a similar
    circuit out of these things, wiring as carefully as I can, point-to-point,
    but they're a bitch to stabilize, and it's hard to even see how they're
    oscillating -- I suspect they are fond of 500MHz or so, which is a shame as
    my scope is only 200MHz!

    So yeah, I could look up other transistors as these, but what if I want
    other ratings, or packages, or pricing? That's why I want a good

  8. Tim Williams

    Tim Williams Guest

    V, I and package, no problem. fT, who said anything about fT? Like I said,
    those vendors don't have any speed specs to pick from.
    No, probably not -- though see my reply to RST.
    Why is that anyway, hFE vs. collector capacitance, or perhaps more to the
    point, hFE per collector area or somesuch? For sure, most of the GHz range
    ones are those icky black lumps with flat leads coming off and are rated in
    mA up to 12 or 25V.

    And what makes an RF power amplifier special, anyway? A lot of the
    transistors I see for that kind of duty are rated for a specific frequency
    range. I might guess that's due to peculiarities in their performance
    curves, but still, a transistor is a transistor, why couldn't I use a 10W
    VHF amp to make a fast-assed DC power switch?

    And speaking of long lists of questions, what keeps a MOSFET from doing VHF
    work? Lead inductance and gate capacitance, sure, but an IRF540 should
    still be good out to, erm, at least 20MHz I would guess. And lead L and
    gate C are just circuit constants, those could be varied, if not by the
    circuit then by the manufacturer, to get a certain passband of interest...
    Any leaded package, really. TO-18 and TO-39 are common too...

  9. Eeyore

    Eeyore Guest

    NPN / PNP (assuming bipolar)
    Darlington ?
    Voltage rating
    Current rating
    Current gain @ low / medium current
    Current gain at higher currents
    Power dissipation / thermal resistance / max operating temp
    Saturation voltage where an issue
    Package and leadout
    Secondary breakdown characteristics where an issue
    Noise figure where an issue

  10. Eeyore

    Eeyore Guest

    I hate to think what the gate drive power would be !

  11. Eeyore

    Eeyore Guest

  12. Tim Williams wrote:

    In a nut shell, you get a fast transistor by making it very
    small, for the current and power rating (to minimize
    capacitances), and use the thinnest possible junctions to
    minimize the carrier transport time (but you also get low
    break down voltage). Those things go against high power
    ratings. Fast transistors are fragile, compared to slower
    devices, where bigger chunks of silicon can be used to carry
    the same current and drop larger voltages.
  13. D from BC

    D from BC Guest

    I get realy lazy and just order the transistors (models) found in
    spice programs. ex: switchercad III
    D from BC
  14. You could try the TO-92 C3355.
    Vcbo 20V
    Vceo 12V
    Vebo 3V
    Ic 100mA
    hFE 50-300

    Cob 0.65pF typ
    ft @ 20mA 6.5GHz typ

    Best regards,
    Spehro Pefhany
  15. I'll admit to being gabberflasted. 6 gig in a TO-92? I'll bet that was an
    extrapolated number from the gain at a gig or so.

  16. whit3rd

    whit3rd Guest

    RF performance is limited by 'base spreading resistance' .
    Both current gain and breakdown voltage get lower when
    you increase base doping to decrease the spreading resistance.
  17. linnix

    linnix Guest

    Just out of curiosity, I picked the first RF NPN from mouser.

    Fairchild SS9018
    1100 MHz Typ
    30V Vce
    50 mA Ic
    $0.04 each

    Sound good enough for me. You can gamble on it for 4 pennies.
  18. Tim Williams wrote:

    There is no reason, except that your design has to contain
    all that high speed gain without accidentally producing any
    instability all the way up into the gHz region. So you have
    to be a UHF designer, or be very lucky to get a DC amplifier
    to function without also being a UHF transmitter, because an
    inch of trace is also an inductor and delay line at a
    frequency where the transistor still has lots of gain.
  19. joseph2k

    joseph2k Guest

    gain * bandwidth product does seem to be remarkably consistent at the device
    level and in simple circuits. It even works for opamps.
  20. LVMarc

    LVMarc Guest


    good question! I start with a search on the VCE and gain bandwidth
    product. then from this set the lowest cos most common for prodcution runs.

    If there is are any other circuit requirments than they enter the
    selection and filtering process. IN summary the circit rquiremtn take
    the ga zilion choice down to a sensible few..

    Best regards,

    marc Popek
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