Connect with us

Class D Audio

Discussion in 'Electronic Design' started by Joseph Roberts, Feb 2, 2004.

Scroll to continue with content
  1. Has anyone out there ever designed class D amplifiers? I have scrubbed out
    a basic design, but have no experience with class D at all. Any advice
    would be greatly appreciated. Thanks.
  2. Rolavine

    Rolavine Guest

    Subject: Class D Audio
    You have checked out TI, National Semi, and Zetex? They all make ICs for class
    D audio?

  3. N. Thornton

    N. Thornton Guest

    My only ssuggestion would be to ask less vague qs
  4. I believe Class D amplifiers have actually made it into real applications,
    which would suggest that yes, someone "out there" has had experience
    designing them.

    Who knows if any of them are here.

    Yes, the question is silly, because whatever he really wants to know
    might easily be answered by someone who doesn't have first hand experience.
    Or the question may be more general than he assumes, so someone with
    experience in another realm may still have the answer without first
    hand experience.

  5. Spajky®

    Spajky® Guest

    There is also a Phillips car amp IC for that ...

    -- Regards, SPAJKY ®
    & visit my site @
    "Tualatin OC-ed / BX-Slot1 / inaudible setup!"
    E-mail AntiSpam: remove ##
  6. Paul Mathews

    Paul Mathews Guest

    have a look here:

    Paul Mathews
  7. N. Thornton

    N. Thornton Guest


    I'm sure one could write a fair bit about class D, but not many folk
    here will write someone a book in response to a vague comment. App
    notes might be a place to start.

    I assume many folk here have done class D amps, or maybe I should say
    Class S. Theyre an effective way to considerably cut power
    consumption, and thus dissipation, and those are exactly whats needed
    in laptops for example. There are now filterless class D amps as well.

    Regards, NT

  8. Okay, let me try, since I have what I think is essentially the same question
    as the OP:

    I'd like to design class D audio amplifiers, capable of several hundred
    watts or possibly up to 1kW, with THD+N less than 0.1% at rated power. Of
    course I don't expect to get there in one step, but it's the eventual goal.
    Commercial manufacturers have obtained this, so it is possible.

    As an initial goal, I'd like to design a decent-sounding and reliable 20W
    (avg continuous 1kHz sinewave) audio amp with THD+N < 0.25% (1kHz,
    unweighted, 20Hz-80kHz) into 8 ohms. But I'd like to understand what I'm
    designing: I don't just want to plug in an eval board and call it done,
    because I won't have learned anything.

    However, I have no experience with class D at all.

    Are there any textbooks and/or good tutorials that cover the topics relevant
    to designing class D amps? I am aware that there are some manufacturer app
    notes for IC-based solutions, but I would prefer information that was more
    thorough in its coverage and less biased toward particular solutions.

    If there are not any such texts, can anyone **with practical experience in
    this area** recommend a preferred design approach, or some preferred
    manufacturers to start with? For instance, will I likely get farthest
    working with discrete components, or should I base my designs around one of
    the chips by Zetex or TI or such? What might the tradeoffs be, and are
    there particular dead-end alleys I should be careful to avoid?

    Thanks for any advice you can give me.

  9. Would appear so:
    The ST "DDX" STA304/6/8 + STA505 amplifier chips would appear to meet that
    I would go looking for the patents on Class D amplifiers and related tech.
    f.ex. DDX - Apogee Technology holds many of them - that may be a place to
    start. . Thomcast SA once took out a patent on a
    "Pulse-Step Modulator" that may be applicable too.

    There is also Bang & Olufson's "Ice Power".
    I think one will get further in less time with the chipsets and here is why:

    The digital amplifers achieve their low distortion by applying proprietary
    pre-distortion algorithms to the signal; The algorithms are likely to be
    described in the patents but - in my experience - algorithms are difficult
    to implement; With DSP's there are often hardware tricks/deficiencies on
    needs to know and the tools are $$$. Much better to buy the custom
    audio-processor chip.

    That leave the power stage *but* with the price of components being what it
    is it is way cheaper and easier to just buy the matching integrated device -
    sorry ;-)

    One can then spend time and effort on producing a decent power supply, a
    proper layout and adequate cooling - not to mention the software interface
    requirted to control these chips. This is a challenge in itself.
    Home construction is no fun anymore - in the good olde days one could
    actually build stuff that:

    1: cost a great deal less than the stuff one could buy
    2: performed much better
    3: where not even available "on the street", thus bringing Hack Value into
    the benefits.

    Sadly, this is not so. Perhaps within Robotics - but within "straight"
    electronics one cannot even obtain the parts at less than four times the
    costs of a commercial unit with good performance!
  10. Rolavine

    Rolavine Guest

    I think I saw a Sterophile recommended class D audio amp made by PS Audio
    within the last year. Was I dreaming?

  11. Ban

    Ban Guest

    you can design and build a good quality PWM-modulator like this. No special
    knowledge or patents needed. The principle is very simple and in fact you
    will find this incorporated into PICs and other MCUs.

    Triangle Generator Comparator Power stage Lowpass Filter
    | | +-----+ +-+ +---+ + +5V
    | /\ | | | | | | | |
    | / \ |--+ --+ +-+ +-----+ +---+ -5V
    | \ / | |
    | \/ | |
    | | | |\ .--------------.
    '-------------' +---|+\ | | ___ __ /|
    | >----| |----UUU---+--o---| | |
    +---|-/ | | | +-|__| |
    Analog In | |/ | | --- | \|
    o----------------+ '--------------' --- |
    | |
    === ===
    created by Andy´s ASCII-Circuit v1.24.140803 Beta

    Now I can help you with the circuits as well, but if you understand the
    principle, you will find out the required frequencies etc. yourself.

    ciao Ban

  12. Thanks, Ban. I understand that is the basis for class D. (I have a little
    knowledge, just no *experience*). But I believe that there are practical
    issues, such as how to keep the power stage from having both transistors on
    at the same time. I am sure the companies who make class-D audio amps have
    already discovered them... maybe I will have to rediscover them myself, the
    hard way.

    An example of the sort of practical issue that I don't understand: I have
    read (here on s.e.d) that it is preferable to use a linear power supply for
    a class-D amp, because it is hard to stabilize a SMPS in this application (I
    don't know why). I have also read the opposite assertion. I wonder what
    the people who actually make class-D amps have discovered, and whether and
    how the problems have been remedied. But that is just an example - I am
    sure there are many such issues.

    I suspect many of the issues of class D audio are the same as they are for
    switch-mode power supplies, and I have a few textbooks on the theory and
    practical design of those. And of course there are quite a few good books
    on the theory and practical design of linear audio amps. But it is mildly
    surprising that there seems to be nothing at all written on practical
    class-D audio.

    Anyway, it seems I've gotten as good an answer as is available; thanks to

  13. The Phantom

    The Phantom Guest

  14. If you take that circuit, and use a diff amp output for the slice voltage, and 2
    comparators, you can slice + and - side, and get 2 180 degree apart pulses
    that each can drive an output transistor.
    Even then it is strongly recommended to have some I measurement in the output
    transistors, that will limit the 'on' times.
    It is indeed a lot like switch mode design, it IS in fact switcht mode design with
    a fast variable reference (the audio).
  15. N. Thornton

    N. Thornton Guest

    Thats a basic class D amp, and the quality wont be good. Open loop
    amps like that are wide open to PSU V variations and PSU hum, and
    suffer from distortion caused by imperfect switching, nonlinear output
    inductors, and relatively high output impednace. Even the simplest
    linear amps OTOH have none of those problems.

    Its a good starting point to learn, to make something like the above
    with an opamp and a couple of output trs, but dont expect hifi.

    Regards, NT
  16. N. Thornton

    N. Thornton Guest

    That one is quite easy. Arrange your trs so they only switch on when
    drive voltage is close to the power rail, ie use pnp on top and npn on
    bottom. Now drive them off one common connection: it is now impossible
    for both to be on together.

    All digital signals slew, so however you do it just picture the drive
    signal slewing and ensure each tr is turned off before the other one
    starts to conduct.

    Class D has no inherent PSU rejection, whereas linear amps have. So
    you either need a very clean very stable PSU, or else error correction
    in the amp to work round the modulation by the vagaries of the PSU.
    SMPS will work, linear supplies will also work but with much less
    grace, and tend to cause junk modulation in the amps output.

    Actually there are a list of issues with Class D that dont really crop
    up with SMPS or linear audio. Most of them boil down to distortion,
    which is quite tough to minimise with class D. There are also
    questions of output filter design, EMI, operating frequency, and
    efficiency, but the big one is distortion.

    Regards, NT
  17. Ban

    Ban Guest

    when you look at commercially made amps they mostly do have conventional
    power supplies. And I do not know why, because you throw away form factor
    and efficiency. But then a 1kW supply with PFC and wide range input voltage
    and synchronized switching frequency is a beast by itself, so maybe with
    time this is gonna change.

    It is also not possible or too expensive to go beyond 1MHz switching
    frequency, hardly any drivers for MosFets available that fast, let alone
    BJTs switching times.
    I use 768kHz with the HIP4081A H-Bridge Driver and STP19NB20 Fets in a
    bridge configuration. The driver takes care of shoot-thru current in this
    The triangle generator can be made by charging/discharging a capacitor
    with current sources, which are referenced by the supply voltage, this gives
    some suppression for this hum-problem that usually comes up because of the
    missing PSSR. The varying resulting freq. is not a problem, as long as both
    channels use the same DC-supply.
    I'm still struggeling with a sample by sample integrator feedback from
    the output to the quantisation stage, as I want to drive the amp also
    digitally with a toslink cable.

    Well, I gave you a few hints, but didn't reveal all the secrets (yet). If
    you are a little further in your design, drop me a mail and I try to answer

    ciao Ban
    Bordighera, Italy
  18. Audio designers can generally design linear power supplies, but SMPS are
    a different art.
    It probably depends on the relative costs of the two technologies. Size,
    efficiency and mass are sometimes important, sometimes not.

  19. In a linear amp, global negative feedback addresses many of those issues.
    Is the problem with Ban's picture that the L-C output filter introduces too
    much phase shift to be useful as NFB? (I've read that was an issue with
    tube hifi amps, trying to use feedback to correct distortion in the output

    More broadly, how come you guys know this stuff? I mean that literally, not
    rhetorically: that is, for those of you responding with practical specifics,
    did you get the answers because you've designed class D amps, or because
    you've read about it in industry periodicals, or texts, or because you're
    surprisingly good at predicting real-world complications of theoretical
    designs, or...? I suspect that I'm missing out on some sources of
  20. I read in that Walter Harley
    It was an issue, but very much tied up with the details of transformer
    construction, the sort of thing than many amplifier designers left to
    transformer manufacturers. Basically, the 'too much phase-shift' was
    caused by excessive self- and inter-winding capacitances, and could be
    substantially eliminated by attention to insulation thicknesses and
    winding plans.
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