Connect with us

Fast NPN/PNP transistors for a totempole driver

Discussion in 'Electronic Design' started by Klaus Kragelund, Jan 19, 2006.

Scroll to continue with content
  1. Hi

    I have a 74HCT14 hex inverter driving a NPN/PNP (BC847/BC857) totempole
    stage connected to GND and 5V

    The output is a small capacitor connected from the common node to
    ground

    The bottom PNP transistor has a current sense resistor to ground so I
    can measure the current when the output falls from 5V to zero

    My problem is I want to drive the totempole with about 10ns tr/tf and
    this causes an error current in the PNP transistor due to BE parasitic
    capacitances

    So I need a couple of fast transistors - better than the BC857 with
    regards to capacitance and ft and it needs to be affordable. Anyone got
    a direct hit?

    Regards

    Klaus
     
  2. Guest

    BFR92/BFT92 - 5GHz at 14mA. Check out the prices from Farnell - a euro
    or so IIRR.
     
  3. Hi Klaus,
    I would test 74ACT14.

    Marte
     
  4. Guest

    Yep. Or even parallel a few inveters for more current, all depending
    on how heavy the load is.

    James Arthur
     
  5. I did infact look at the ACT devices, but could not right away find info
    if its ok to parallel them since the device is a schmitt trigger and a
    slow changing input voltage would lead to much "cross" conduction
    between devices. Ofcourse I could use one device to sharpen the input
    edge and run the others off this first inverter. Is it ok to parallel if
    the cross conduction period is small (perhaps with output resistors on
    all outputs)?

    Moreover I read one this group a while ago about a special part rated
    for high currents (perhaps it was a 4000 series chip). I have searched
    but could not find the thread......

    Would it be possible to use another coupling to get the speed and low
    parasitic capacitance I'm after? And still being able to measure by
    means of a current sense resistor the transient current in an unknown
    capacitor?

    Regards

    Klaus
     
  6. Guest

    Yes, you can parallel them, and yes, you can expect them to
    cross-conduct for a fraction of a nS, drawing a nasty spike. It's not
    destructive though, so you just bypass and ground accordingly, or add
    current-sharing resistors if you really care.

    Squaring up the waveform with a preliminary inverter is a good idea,
    and I'd square it up with something before the 'ACT14, personally. I
    don't know about the 'ACT14 in particular, but I've seen grossly
    excessive dissipation in earlier devices when fed slow waveforms.

    Another possiblity is using a MOSFET gate driver. Those are
    basically very hefty CMOS inverters, some including bipolar buffers on
    their outputs. I've got some TPSxxxx devices from T.I. fitting that
    description.

    Cheers,
    James Arthur
     
  7. Robert Baer

    Robert Baer Guest

    Take a look at the schematic of the "standard" NPN/NPN totempole
    output used in the 74xx series of logic.
    There is a short period of time (up to 6nSec) when both the top and
    bottom NPNs conduct, creating the "sharp spike" that others mentioned
    elsewhere in this thread.
    That happens when the bottom NPN is supposed to turn off, and the top
    NPN is turned on.
    That happens *not* due to B-E capacitance that you refer to, but to
    stored charge in the base that must be removed for the NPN to turn off.
    Perhaps a similar problem exists inyour (discree?) circuit.
    The solution to the NPN/NPN totem pole problem is to 1) put a
    resistor in series with the base of that bottom NPN, and 2) add a PNP
    with its emitter connected to the NPN base and its base to the "input"
    end of the resistor.
    When the driver turns off, the PNP turns on and pulls the base charge
    out.
    This scheme can speed up the output from 2 to 5nSec and will
    completely eliminate the spike at most temperatures (in an IC, the spike
    is reduced by only a few orders of magnitude at 125C).
     
  8. Klaus,
    forget TTL, it was never designed to cope with 100MHz.
    Have a look at other families.
    What kind of voltage range on the output is required
    and what is the load.

    Rene
     
  9. The output load from the totempole point of view is >300Ohm and max
    100pF

    5V or any other voltage will be fine

    Regards

    Klaus
     
  10. Guest

    100pF raised 5 volts in 10nS will take 50mA, a very modest
    requirement. Paralleled 74ACxxxx CMOS gates will do the job quite
    easily, and make pretty rail-to-rail squarewaves too.

    Even the 74HCxxxx types might do--a single 74HC244 is guaranteed to
    drive 50pF in 12nS--but you'd need to parallel more gates than with the
    newer stuff.

    Regards,
    James Arthur
     
  11. I tend to use the 74F0037, a quad NAND driver with 30 Ohms output
    impedance. It rises to 2.5V in 2ns then becomes slower, say another
    4ns to 4V, all into 50 Ohms.

    Then these are MOS driversfrom various manufacturers.

    Rene
     
  12. Ken Smith

    Ken Smith Guest

    If you want faster edges, you can run the thing on 5.5V. They get a bit
    faster as you go up from the STD 5V supply.
     
  13. Jim Thompson

    Jim Thompson Guest

    I think this thread has gone astray. The OP wanted to measure load
    current, which was being obscured by totem-pole device capacitive
    currents.

    ...Jim Thompson
     
  14. Ken Smith

    Ken Smith Guest

    Not astray just digressing wildly.
     
  15. Guest

    Hmm... I took the OP's current-sense resistor to be for
    troubleshooting, but you could be right Jim.

    No worries, just move the sense resistor to the CMOS gate's Vss. Be
    sure to feed the CMOS driver-part fast edges though, and/or use a
    non-inverting part to avoid negative-feedback via Vss & such. (a small
    Vss resistor + output resistor adds a little input hysteresis to an
    ordinary, non-inverting CMOS gate.)

    I suggested paralleling CMOS gates because the OP indicated he's
    already using one as the driver.

    Cheers,
    James Arthur
     
  16. Fred Bloggs

    Fred Bloggs Guest

    Excuse me...but a PNP/NPN combo is not a "totempole." Who the hell knows
    what this person is asking? I hope he communicates to himself better
    than this- he will be stuck forever.

    [snip]
     
  17. Guest

    Yeah, you're right, it's not what we normally mean by 'totem pole,'
    but from this:

    "The bottom PNP transistor has a current sense resistor to ground so I
    can measure the current when the output falls from 5V to zero "

    I took the OP to mean this:

    .. Vcc
    .. -+-
    .. |
    .. |
    .. |/ Q1
    .. +---| BC847
    .. | |>.
    .. | |
    .. | o----> >---o----+
    .. |\ | | | |
    .. | \ | |<' C(L) --- .-.
    .. --| >O--o---| 100pF --- | | R(L) = 300 ohms
    .. | / |\ | | |
    .. |/ Q2 | === '-'
    .. BC857 | |
    .. 'HCT14 .-. ===
    .. | |
    .. | | Rs
    .. '-'
    .. |
    .. |
    .. ===
    .. GND

    Best Regards,
    James Arthur
     
  18. wrote...
    Right. Whereas a totem pole is this:
    Which is much more simple, but depends on the driver's pulldown
    capability for the sinking current. It's especially well suited
    for designs with high-performance NPN or N-channel devices.
     
  19. Yes, and this was infact my first solution back when I did tests with
    the CD4093 chip and it failed from sufficient output drive/speed

    Regards

    Klaus
     
  20. 100% correct :)

    Regards

    Klaus
     
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

-