Connect with us

class AB biasing scheme with MOSFET's

Discussion in 'Electronic Design' started by Mac, Dec 9, 2003.

Scroll to continue with content
  1. Mac

    Mac Guest

    Hello, all.

    I want to design a 10-watt audio amplifier using an LM833 for the gain
    stages, and high power MOSFET's in a class AB configuration for the output
    stage. I have a circuit, which omits the 833, and shows my scheme for
    keeping the biasing regulated at between 100 and 200 mA. I plan to take
    feedback from the output back to the 833, though I haven't shown this in
    the schematic.

    Anyway, my concern is that the bias regulation scheme may not be stable. I
    am in the process of re-reading some of my textbooks, so I can re-learn
    how to analyze the circuit, but I thought I'd just post it here in the
    mean time.

    Oh, I'm assuming that Vth might be anywhere from 2 to 4 volts for both
    FET's, so the idea is that U1 (see below) will span a range of 3 V when
    the bias current spans a range of 100 mA. I haven't chosen a part for U1,
    but when I do I will select a precision type with input common mode down
    to V-.

    Any type of comments are welcome. For intellectual satisfaction, I would
    still like to know whether this bias scheme is stable (and how to think
    about it so that it is easy to see), even if someone suggests a better
    alternative.

    Use courier or some similar font to view:


    +15V +15V
    | |
    | \
    | /
    \ R10 \
    / 0.5 /
    R1 360K \ \
    / /
    \ |
    / ___|
    | |
    +---+------------------------| Q1 N-chan.
    | | |-->
    \ | |
    / | |
    R2 110K \ --- C1 large |
    / --- |
    \ | |
    / | |
    | | |
    LM833--> o---+---+ +---VOUT
    | | |
    \ | |
    / | |
    R3 110K \ --- C2 large |
    / --- |
    \ | |
    / | |
    | | |
    | | <--|
    | | |
    +---+------------------------| Q2 P-chan.
    | |___
    / U1 |
    R4 220K \ |
    / /| R7 R8 |
    \ / | 220 K 220 K |
    | / +|------+-/\/\/-+-/\/\/--+
    +---+ | | | |
    | \ -|---+ = C3 = C4 |
    / \ | | | 0.1uF | 0.1uF \
    R5 33K \ \| | | | / R9
    / | | | \ 0.5
    \ | | | /
    | | | | |
    +-----------+ | | |
    / | | |
    \ | | |
    R6 560 / | | |
    \ | | |
    | | | |
    +--+-------+--------+
    |
    -15V

    regards,
    Mac
     
  2. I'm not an expert, but my concern is that it won't work at all. You seem to
    be controlling the bias based on looking at the DC component of the current
    through the negative rail. But either rail's contribution is rectified,
    meaning that its DC component is directly proportional to the AC signal
    level. So as the signal level changes, the bias will be changing, which
    doesn't seem to be what you want.

    Might you be reinventing the wheel here? I mean, plenty of bias circuits
    have been presented and used successfully; this is admittedly one of the
    challenging parts of amp design, but the versions that are out there are
    really pretty reasonable and have stood the test of time.

    Maybe there's more to it than I saw, in which case my apologies.
     
  3. This will not work. It measures bias and load current.

    Use a bog standard circuit. Use an npn transistor with collector
    connected to top fet gate, emitter connected to bottom gate. connect a
    resistor from collector to base, and a resister from base to emitter.
    This is called a vbe multiplier. Set the resister to give you the
    required current. You will need to power up the amp an adjust the
    resisters by measuring the supply current.

    Kevin Aylward

    http://www.anasoft.co.uk
    SuperSpice, a very affordable Mixed-Mode
    Windows Simulator with Schematic Capture,
    Waveform Display, FFT's and Filter Design.

    http://www.anasoft.co.uk/replicators/index.html

    "Understanding" itself requires consciousness,
    therefore consciousness cannot be "understood"
    without referring to itself for the explanation,
    therefore the "hard problem" of consciousness,
    is intrinsically unsolvable as it is self referral.
     
  4. Sorry, your circuit won't work. The current through one power resistor
    tells nothing of the bias because the output power flows through it too.

    You can try something like this:
    http://www.pixelmemory.us/Photos/Nerd/AutoBiasAmp2.gif

    I didn't write down the final circuit because it never left the
    breadboard. The idea is that current must be flowing through both
    resistors at once and the current pulling the two gates together must be
    balanced.

    Some things to experiment with -

    It needs some resistors and capacitors to prevent ringing.

    The transistors on the left don't get much bias voltage because the EB
    junctions of the transistors on the right shorts the path. A voltage
    divider on each of the bases of the transistors on the right would turn
    them on later and solve that.

    The left half is designed to be connected to a simple constant current
    Class A differential amp or the fancy balanced current sources that
    you'll see in high-end audio circuits.

    The simple Class A differential can be seen in any op-amp chip
    schematic. This works perfectly in small and medium sized amps.

    The simple Class A diff amp rectifies high frequencies somewhat by
    slewing faster in one direction than the other, and that's a problem for
    a big high voltage amp that needs very high slew rates. Marshal Leach
    demonstrates how to produce a symmetrical and compensated drive:
    http://users.ece.gatech.edu/~mleach/papers/lowtim/feb76feb77articles.pdf
    http://users.ece.gatech.edu/~mleach/
     
  5. John Larkin

    John Larkin Guest

    As mentioned, this won't work. Why not just use a pot... set it once
    and you're done. You can get away with this with fets.

    I just did an amp - not audio! - that uses a uP to set output offset.
    It observes the signal level, waits for 'quiet', measures output
    offset with a precision amp/ADC, and tweaks the input stage using a
    DAC. A scheme like this could set idle current, too.

    John
     
  6. Genome

    Genome Guest

    No you can't, iriot!..... With the caveats, like I personally wouldn't claim
    to be certain but I did it once and it worked......... Enhancement mode
    mosfets operated in their linear reagion below some level of current have a
    negative temperature coefficient of gate threshold of something
    like -6.2mV/degC. That was old data from Siliconix and IR. Might be worth
    checking it with the manufacturer.
    I should think that even Phil Allison and his multiple bumholes has done
    this using a non-precision op-amp as a DC servo without the uP/ADC/DAC.
    Please, try not to argue about bandwidth and such stuff.
    I'm sure it could.

    Well.... you might have to think a bit harder about that one. You need to
    sense both devices operating current. Then you have to say 'Now!!!' when
    they are equal. Then you have to regulate what it was when you said Now!!!.
    Or you might sense the load current and say 'Now!!' when it is zero and do
    the same. Bear in mind that the load is most likely reactive so your 'Now!!'
    signal won't necessarily happen at zero volts output.

    Once you have figured out how to generate the Now!!! signal you still don't
    need all the other clutter.
    Oh, go on........ you were just taking the piss after all. Hang on the
    lights have gone out......

    Hmmm, do I replace the bulb or reset the embedded digital signal processor?

    DNA
     
  7. Genome wrote...
    Maybe he was thinking of the elegant Hitachi lateral FETs, which
    have nearly zero tempco at about 0.25A, made to order for the job.

    Thanks,
    - Win

    whill_at_picovolt-dot-com
     
  8. Genome

    Genome Guest

    Now..... I didn't know that. I knew something else..... but I didn't know
    that.

    DNA
     
  9. John Larkin

    John Larkin Guest

    Heck, it's just software. Software is cool because

    it never breaks

    it's free

    you don't have to solder it in

    you can fill it with goofy comments and pun-labels.

    John
     
  10. Now..... I didn't know that. I knew something else.....
    Hmm, neither did I till a few years back. Very popular
    among audio enthusiasts, those who are in the know.

    Thanks,
    - Win

    whill_at_picovolt-dot-com
     
  11. John Larkin

    John Larkin Guest

    No, I was thinking, hell, this is just audio, most any old quiescent
    current is OK.

    John
     
  12. Mac

    Mac Guest

    Doh! You (and several others) are right, of course. The circuit is a Bad
    Idea, reflecting wrong thinking on my part. In fact, it's a little bit
    embarrasing.
    No apology necessary, since you were right. Thanks for looking it over. I
    don't really know of any biasing arrangements for power MOSFET's, so I
    just did what seemed sensible to me. I did want to avoid having a bias
    adjust pot, but maybe I'll let that requirement drop.

    best regards,
    Mac
     
  13. John Larkin wrote...
    Maybe so. But sadly too often 10 to 15 years later, what is
    an elegant program? ... Too often just a memory and a claim
    to have written 10000 lines of code. Whatever that means ten
    years later. How melancholy.

    Thanks,
    - Win

    whill_at_picovolt-dot-com
     
  14. Mac

    Mac Guest

    Right. I think I originally thought of this biasing scheme for a class A
    amp, where the DC current is the bias current. It doesn't work at all for
    class AB because only half of the waveform flows through each transistor.
    I'll take a look.
    Thanks for looking at my circuit.

    Mac
     
  15. Mac

    Mac Guest

    So where does the BJT get biased from? I guess there is still a resistor
    from collector up to VCC and another one from emitter to VEE?

    And where is the input? It seems you would have to drive it at the
    emitter for it to work.

    By the way, I was not familiar with the expression "bog standard." I had
    to look it up.

    Mac
     
  16. Mac

    Mac Guest

    That's an interesting idea. It's essentially a signal-processing solution
    to offset drift.

    Thanks for looking at my circuit.
    Mac
     
  17. It can be driven by the collector of emitter. Its just in a bias chain.
    usually, you have a pnp on top and a npn on the bottom. Its a constant
    voltage ether way. It can help to put a capacitor across it.

    Kevin Aylward

    http://www.anasoft.co.uk
    SuperSpice, a very affordable Mixed-Mode
    Windows Simulator with Schematic Capture,
    Waveform Display, FFT's and Filter Design.

    http://www.anasoft.co.uk/replicators/index.html

    "Understanding" itself requires consciousness,
    therefore consciousness cannot be "understood"
    without referring to itself for the explanation,
    therefore the "hard problem" of consciousness,
    is intrinsically unsolvable as it is self referral.

    "I now understand how inanimate matter results in
    consciousness. What do you mean by "understand"?
    Understanding is that quality which results when
    one has consciousness".
     
  18. John Larkin

    John Larkin Guest

    It is interesting to go back 15 (30?) years and review your old
    schematics and read your old code. I try to keep some of this stuff
    around to remind myself of how bad I used to be. As far as the code
    goes, you've got to keep paper listings or be sure to transfer to more
    modern media (punchcards-hard disk packs-8" floppies-4"
    floppies-tape-3.5" floppies - CDs) as things get obsolete, and that's
    a risky path to keep up over the years.

    Some of my PDP-11 code was pretty elegant, come to think of it. Some
    of my logic design was execrable.

    John
     
  19. John Larkin wrote...
    Indeed, as was some old code I wrote. But what can you do with
    the old code, no matter how elegant? Often you can't run it, or
    wouldn't want to. Sadly nobody else wants to admire it. R.I.P.

    Thanks,
    - Win

    whill_at_picovolt-dot-com
     
  20. Jim Thompson

    Jim Thompson Guest

    What's even scarier is to go back and look at an old design, that you
    know worked, but that you now can't figure out why.

    I designed a ramp generator for a Chrysler ignition chip that dumped a
    capacitor until the (NPN) device saturated, then it released to charge
    back up.

    This was before computers, all hand designed.

    Took me a *week* of Algebra before I figured out how saturation was
    detected (VBE drops).

    ...Jim Thompson
     
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

-