Connect with us

Trouble shooting push-pull amp

Discussion in 'Electronic Design' started by Terry Pinnell, May 12, 2004.

Scroll to continue with content
  1. This circuit is part of a very old project.
    http://www.terrypin.dial.pipex.com/Images/AudioPowerAmp1.gif

    Until a few hours ago it was working after a fashion. But while
    attempting to improve it I've clearly screwed something up. The
    voltage at the +ve terminal of the output cap is now at Vcc (24V),
    instead of around 12V. Adjusting the R2 100k preset makes no
    difference. Staring and fiddling has got me no closer to discovering
    why.

    Can anyone help isolate the likely cause please?

    (Also posted with attachment in alt.binaries.schematics.electronic)
     
  2. j.b. miller

    j.b. miller Guest

    Obvious items.... are C2 ,Q3,Q5 shorted ?
    Hey at least it's a start....
     
  3. Seems you have no trouble shooting it, as it has already happened ;-)

    Have you got a scope or at least a multimeter? What about measuring a
    few more voltages in the circuit? You're not exactly giving away much
    information. Could Q5 have melted down?
     
  4. Tim Wescott

    Tim Wescott Guest

    Measure the Vbe and Vce of each transistor, from the speaker backwards,
    and look for readings that make sense (i.e. 0 < Vbe < 0.7V, and Vce of
    around 12V). Also measure across your diodes. If that doesn't get you
    there, start measuring across resistors (oh joy).

    Chances are you've fried a transistor, you've probably made one fail
    short on the supply-rail end but you may have made one fail open on the
    ground. Failing that it could be a short or an open nearly anywhere in
    the circuit.

    Find the bad part and replace.
     
  5. legg

    legg Guest

    When you've found the dead semicoductor, relocate R8's connection to
    the emitter of Q6, not to the kelvin signal ground.

    RL
     
  6. Many thanks for all the helpful replies, here and in
    alt.binaries.schematics.electronic. First thing I'll check this
    morning is whether Q3 and Q5 are shorted. If so, I'll fix and then add
    some minimal protection along the lines Ban suggests.

    The very last actions I took seemed innocent enough, which is why I
    was at a loss about these symptoms. I had been getting a distorted
    output, and C1 +ve end was well above 12V. I'd tested all four small
    BJTs and the two power types. Therefore I began to mistrust my
    recollection and scruffy handwritten notes that it *used* to perform
    properly. (I built it 23 years ago.) So I speculated that R2 wasn't a
    large enough range to give me optimum adjustment, and replaced it with
    breadboard connections to 220k. It was at that stage that the C1
    voltage became fixed at about 24V.

    Thinking about it overnight, I suspect that I may heve been careless.
    I have that particular stripboard circuit unscrewed from a large case
    containing other sections. But it's still hard wired to the 2N3055s,
    which are fixed with usual insulation to rear of case. So it's
    temporarily on a piece of cardboard suspended on top of other wiring,
    with power temporarily from my bench supply. I'm careful to switch off
    that supply each time I make any tests. But a hefty charge must remain
    on C1, and one possibility is that some of it got into the wrong place
    while I considered the circuit 'unpowered'.

    Digressing, I often get into this loop. Temporarily testing or
    modifying a finished project, with some parts still in the case and
    others on breadboard, with power supply and probe wires etc across the
    bench - a scenario which risks adding *new* faults as well as those
    under investigation. (Oh, did I mention my natural impatience too?
    <g>)

    Finally, the simulation
    http://www.terrypin.dial.pipex.com/Images/AudioPowerAmp2.gif )
    prompts me to wonder whether another possibility might be that I
    simply haven't got R2 and R4 correctly adjusted. They do seem
    critical, and several I simulated end up placing C1 at the sort of
    high level I reported. Assuming that's not the cause, when I do have a
    'working circuit' again, what is the correct sequence for adjusting
    these two presets? Is it just a matter of trial and error, or is there
    an optimum quiescent current I should first set with R4 before
    adjusting R2 to get 12V at C1?
     
  7. Can you amplify (sorry!) on that please? Why would that be an
    improvement? The original 30-year old circuit on which I based mine
    presumably connected all low power signals to 'Signal earth', and only
    the final output return to 'Power earth'.

    BTW, that term 'kelvin signal ground' is unfamiliar to me. (A google
    news search on it turns up only your post here.) Could you explain its
    meaning please?
     
  8. Paul Burke

    Paul Burke Guest

    A Kelvin connection is a separate trace to the high- impedance sensing
    circuit, added to avoid voltage errors caused by a (relatively) high
    current through the track resistance. It's often used in sensor bridge
    circuits, particularly where the distance from the bridge to the
    conditioning circuit is large.

    Paul Burke
     
  9. Thanks, Paul. In that case, I don't see its relevance? Unless RL is
    using it to mean 'signal ground'.
     
  10. Terry Pinnell wrote:
    [...]
    If you were getting a distorted output all of a sudden, it is possible
    that the amp was oscillating at a high frequency, which frequently
    results in overheated output transistors. This may therefore have caused
    the death of Q5. As Ban already noted, your amp has absolutely no
    protection against any sort of mishap built in. I can't even see
    anything that would limit the bandwidth, which presumably makes it
    rather difficult to control a tendency to oscillate. I'm rather amazed
    it worked for 23 years :)
    By removing the circuit board from its case you changed the layout of
    the wiring, which may well have started the oscillations. No short or
    capacitor discharge into the wrong place is necessary to explain the
    outcome. It could have happened without any carelessness of your side.

    [...]
    Note that your simulation shows a tendency of the circuit to oscillate,
    which supports my argument.

    R4 controls the quiescent current in the output stage, hence it is very
    critical. If adjusted wrongly the output stage simply overheats,
    possibly in a very short time.

    If you simulate the circuit, have a look at the *currents* in the output
    stage, too. Looking at the voltages may not give you the real picture.
    Well, actually, the only thing to give you the real picture is the real
    circuit ;-)
     
  11. Still unresolved.

    The cause appears to be none of the suggestions made so far. All 6
    transistors are OK. Fresh from the workshop, attached are some actual
    voltage measurements which I propose studying this evening. Maybe with
    a glass of dry white...
    (Also posted in sed, and same illustration is at
    http://www.terrypin.dial.pipex.com/Images/AudioPowerAmp3.gif )
     
  12. Do you really have 24.7V and 80mV on the same track? If so, you
    probably have an ex-track.

    Cheers
     
  13. legg

    legg Guest

    The connection of R8 to signal ground, rather
    than to the emitter of Q6 introduces some
    discontinuity in the function of Q6. This
    acts to add to Q6 off-bias when Q5 is conducting
    through the load or through Q6, but produces a
    sudden increase in Q6 on bias when Q5 turns off.
    This does not come from the control circuit, so there
    will be a tendency to oscillate, unless R8 is connected
    to the emitter.

    R8 is not part of the 'low power signals'.

    Setup

    -R2 at midpoint
    -R4 at minimum value.
    -Apply power. VC1 should read ~6volts
    -Decrease R2 to set VC1 to 1/2VCC
    -Increase R4 until supply current starts
    to increase, then back it off.

    This design requires ClassB biasing.
    It cannot be safely biased as classAB.

    The quiescent voltage on C1 will also tend to
    drift down with increase in temperature of Q1.

    Q1, Q2 and Q3 form a signifigant silicon short
    if the control circuit tries to handle ANY quiescent
    current in Q5.

    Adding an emitter resistor to Q4, Q5 and Q6
    might allow the circuit to run in classAB.

    RL
     
  14. Yes - that was it! A couple of others over in
    alt.binaries.schematics.electronic spotted it too. Proved to be a hair
    line crack on a stripboard track.

    I hadn't really looked at those figures before I posted them. Like to
    think I'd have spotted that inconsistency after dinner <g>. Just glad
    I measured that node.

    Thanks to all here for the help with the detective work. With that
    mystery solved, I can return to getting a decent output now! I'll come
    back soon to seek further help in minimising distortion.
     
  15. Thanks. They *look* OK, but from waht you say I suppose it could be
    that in the earlier stages of my testing they were reforming. Latest
    results seem to be OK as you probably saw.

    Interesting that my '1000uF' is actually about 700uF!
     
  16. Ken

    Ken Guest

    My Philips TV is made in early 1976 and still going strong.
    I use it every day and all capacitors is original!
    World record?
     
  17. Dunno if world record, I have some old Philips electrolitics from that time
    in the box, and they are fine.
    Electrolitics did often go bad also in Philips TVs.
    I have an old Crown transistor BW portable, from the seventies I think, it
    works, but needs an other component transplant now, vertical is jumpy.
    JP

    JP
     
  18. Anyone remember Vitrohm composite carbon resistors?
    Anyone remember Wima capacitors?
    Wima still exists BTW, but I am still scared to use these...
    The resistors would go high in TV sets, or crack open, or go up in flames..
    Those caps would short... crack, fail.
    In those old days (sixties) if we foud those resistors in a TV set,
    we would start by replacing ALL of these, some job, no PCBs, all wires with
    the wires through the holes in the tube sockets, and we HAD to get the new
    wire through the hole, and then solder it.
    JP
     
  19. Terry Given

    Terry Given Guest

    If its 23 years old, your electrolytic capacitors are in all likelihood past
    their best. Electrolyte eventually "dries up" and the hotter the cap gets
    the faster it does so - this is why smps caps crap out all the time, the
    ripple current is high and they cook - in your amp C1 is the only candidate
    for this failure mode. Other things happen when the cap is not charged for
    long periods of time (google electrolytic capacitor re-forming if you want
    to find out what, how and why). The actual capacitance can decrease hugely,
    leakage current increases - I have replaced caps that measured 10s of ohms.
    From your perspective, just replace them as a matter of course (most TV
    techs do this before looking for faults).

    Terry
     
  20. Terry Given

    Terry Given Guest

    I shot myself in the foot with a lousy electrolytic once (metaphorically
    speaking - I did however destroy the caps in $3,000,000 worth of ac drives
    once but thats another story).

    The circuit was a 5V voltage reference using the humble TL431, running from
    12V. I had a 5k resistor, so about 1.4mA, and 2 x 10k resistors for the
    voltage divider. As you may well know, TL431's will oscillate with a certain
    load capacitance - between 1nF and 100nF is the danger area (TI have some
    nice pictures in their data sheets). So I used a 10uF smt electrolytic cap.
    And the damn thing sang like a canary. Why? well, the 10uF 25V electrolytic
    cap I used had 27 ohms (yep, no decimal point). I measured it to be so,
    using a step-voltage from a 50 ohm source, edge showed step to 1/3 of Vstep,
    then RC charge to Vstep (thanks Howard Johnson). I got an OSCON cap (0.02
    Ohms) and a series resistor, and found >= 10R was where oscillation started
    (froze it and baked it too, with freeze spray & heat gun). Actual solution
    was to buy a better cap, around 1R ESR. and forever more, look at the ESR of
    every electrolytic I ever use. In general, pick 105C caps - the sort of caps
    you get from hobbyist stores tend to be really crap, and if the data sheet
    doesnt mention ESR (or ripple current), you can be sure its very high (very
    low).

    I used to repair monitors and smps for a living before I went to uni and
    became an engineer. After about 6 months, I learned to replace all
    suspect-looking electro's - swollen tops (those grooves are so the case
    ruptures there), goo leaking out the bottom, etc. If i had a hard-to-find
    fault in a particular area, I would replace all the electros there, THEN
    look for the fault. between that and fixing all the broken tracks/solder
    joints, probably 75% of all faults had gone by the time I looked for them.
    caps are cheap. time is not.


    Cheers
    Terry
     
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

-