Linear Amplifier Module Diagnosis

[Obi_Kwiet]

I'm trying to figure out what's wrong with one of the channels on a Klipsh Promedia 2.1 amplifier. This is really a academic exercise on my part, as I'm pretty awful at understanding discrete transistor networks, and I'm trying to get better. My strategy has been to poke around at this thing until I get stuck, and then see if someone can give me some pointers on what is going on and how to approach this kind of problem.

When I turn the volume up, one channel distorts. When I throw the output of the distorted channel on the oscilloscope, the output starts to clip on the positive side of the signal at if the signal level is driven past a certain amplitude. I've verified that the signal going into the module does not clip. The voltage levels on both sides of R17 are about 7V +-~1V, rather than the expected 0V +- ~1V that we expect and indeed see on the other channel. So, it's pretty obvious that the bias point of the output is off, but I don't understand the biasing method very well. Q6 is supposed to be setting the bais point, and it looks like the audio signal is driven in by Q3 and stabilized by the feedback loop. Q6 is being driven by a constant current source, but I don't really follow how it's supposed to be setting the bais point, so it's hard for me to come up with a good guess as to what might be wrong. R15 and R16 ring out correctly.

Any help would be appreciated!

[coromonadalix]

As you guessed 

Q6 is the bias control transistor
Q3 is a current regulator or an voltage regulator and Q4 Q5 would be a current regulator for the differential input transistors Q1 Q2

The feedback loop come from the output header pin #6 going to Q2 via C4, R7

I would remove  Q7 Q8 Q9 Q10 and test them for any leakage

Normally an biais transistor should be thermally coupled to the heatsink to do a good job ...

You should check all the supply voltages for stability and ripple, without any sound you should have near 0v dc at the output header pin 6

I dont see why r18 and r19 are 0 ohms 5 watts, they shoulde ex : 0.15, 0.18, 0.2  ohms  etc ...  not an zero value who is considered as a short, maybe a miss print


here's a trouble shooting link:
https://community.klipsch.com/index.php?/topic/129544-solutions-to-fixing-your-klipsch-promedia-2141-speaker-system-compiled-links/

[Obi_Kwiet]

R18 and R19 are indeed jumpers. I expect they are left over from an earlier revision of the board.

I guess I'll pull out the transistors. The power supply rails seem pretty crap; the negative rail is has 1V saw-tooth ripple in it, and there's an annoying amount of switching noise on both rails. However, the other channel seems to do ok.

How does that bias transistor work? I don't really understand the theory of operation there with it's emitter tied to the base of Q8 and the collector of Q3.

[coromonadalix]

No resistors value   thats total crap

here's some explanations :

https://www.electronics-tutorials.ws/amplifier/emitter-resistance.html

https://www.electronics-tutorials.ws/amplifier/class-ab-amplifier.html

https://www.electronics-tutorials.ws/amplifier/transistor-biasing.html

[David Hess]

There is no way the voltage across R17 could get that high without the output stage being damaged.  One or more of Q9, Q10, R18, and R19 are open.

A more interesting question is why they failed.  Make sure that the voltage from the collector to emitter of Vbe multiplier Q6 which sets the output bias is about 3 volts.

[Obi_Kwiet]

No resistors value   thats total crap

here's some explanations :

https://www.electronics-tutorials.ws/amplifier/emitter-resistance.html

https://www.electronics-tutorials.ws/amplifier/class-ab-amplifier.html

https://www.electronics-tutorials.ws/amplifier/transistor-biasing.html

Those explanations are good, but they are all pretty straightforward. The problem I am having is that is seems very difficult to break this network down into conceptual blocks that I understand. For example, Q6 is obviously biasing Q7 and Q8 somehow, but it's weird, because it's driven by a current rather than a voltage source. It's also getting a signal input from the emitter of Q3. Obviously it's supposed to be evenly spitting the current into the bases of Q7 and Q8, but I can't quite understand how that process interacts with the current going through it's bais voltage divider of R15 and R16, and collector of Q3.

There is no way the voltage across R17 could get that high without the output stage being damaged.  One or more of Q9, Q10, R18, and R19 are open.

A more interesting question is why they failed.  Make sure that the voltage from the collector to emitter of Vbe multiplier Q6 which sets the output bias is about 3 volts.

R18 and R19 are not open for sure. They are jumpers and they are intact. I also rang out the terminals of Q9 and Q10 and they are not open. The output is actually correct until the output gets to a particular amplitude, at which point is clips on one side only. I feels more like the bais point is way off somehow.

[David Hess]

R18 and R19 are not open for sure. They are jumpers and they are intact. I also rang out the terminals of Q9 and Q10 and they are not open. The output is actually correct until the output gets to a particular amplitude, at which point is clips on one side only. I feels more like the bais point is way off somehow.

The bias point cannot be that far off without blowing up the output transistor circuit.

The Vbe bias circuit must be damaged as well since it would never allow that.

[Obi_Kwiet]

Well, I don't think it's the transistor bias itself, rather it's the input signal bias. It acts like the output is just hitting the +V rail. Maybe I should throw a probe on the 6 pin in DC coupling mode and see what it looks like. So far I'm just measuring the speaker output which almost certainly has a DC blocking capacitor somewhere in the signal path.

[nsd_c]

I have my own set of Klipsch ProMedia 2.1 speakers that I have also been trying to repair. In my case, the failure was catastrophic: Q9, Q10, and R17 blew up dramatically (significant amounts of smoke revealed when opening the cabinet). I naively tried to replace those components only, but this failed. I did not discover until a second repair attempt that Q6 had also failed (but in less visually noticeable fashion): its base and collector shorted together. I found an image online of another user's unit which failed in much the same way. (I linked to the Google search because the images themselves have disappeared but live on in Google's image cache... Original thread was here.)

So, it's pretty obvious that the bias point of the output is off, but I don't understand the biasing method very well. Q6 is supposed to be setting the bais point, and it looks like the audio signal is driven in by Q3 and stabilized by the feedback loop. Q6 is being driven by a constant current source, but I don't really follow how it's supposed to be setting the bais point, so it's hard for me to come up with a good guess as to what might be wrong.

Somewhere I found a user on another forum indicate that Q6 (and its neighboring components) form a circuit called a "rubber diode." That link would seem to do a pretty good job of explaining the circuit, but I can't say that I understand it all that well myself...

Normally an biais transistor should be thermally coupled to the heatsink to do a good job ...

Q6 is thermally epoxied to the rear panel of the cabinet (a piece of ~1/8" thick aluminum about 8x8"). The output transistors Q9 and Q10 are screwed to that same panel (on either side of Q6) with a thin layer of thermal paste (as well as mica insulators).

Well, I don't think it's the transistor bias itself, rather it's the input signal bias. It acts like the output is just hitting the +V rail. Maybe I should throw a probe on the 6 pin in DC coupling mode and see what it looks like. So far I'm just measuring the speaker output which almost certainly has a DC blocking capacitor somewhere in the signal path.

Be warned that the speaker output *does not* have a DC blocking capacitor in the signal path! In my case, when the output transistors blew, DC was applied to one speaker which damaged the voice coil on the mid-range driver! There are fuses between each HF board and each satellite speaker, but the fuse for the HF board that failed in my unit did not blow. I would suggest building a dummy load to test the bad channel with. I used three 2 ohm 30W power resistors in series as this is what I had lying around.

Btw, the satellite speakers do contain a cross-over circuit, which does contain a capacitor, but the capacitor is in parallel with the mid-range driver, hence it does not block DC. I've attached my reverse-engineered cross-over schematic. 'T+' and 'T-' go to the tweeter; 'M+' and 'M-' go to the mid-range driver. Note that I did not measure the inductor with an LCR meter, but it looked like "4 mH" was printed on the PCB in the silkscreen beneath it (so don't put too much faith in that value).

[nsd_c]

The power supply rails seem pretty crap; the negative rail is has 1V saw-tooth ripple in it, and there's an annoying amount of switching noise on both rails.

If you study the power supply schematic (attached--copied from here), note that the HF board rails (+/-V_HF) are unregulated; there's just smoothing caps following a bridge rectifier. So, all of the switching noise that you're seeing is probably ~100 kHz noise from the LF amplifier board's square-wave generator, which is part of its class-D amplifier.

There doesn't seem to be any filtering to remove this noise from elsewhere on other boards, but I guess the designers figured that it wasn't necessary because 100 kHz is beyond the range of human hearing (and is probably hugely attenuated by the drivers).

[Obi_Kwiet]

Look like Q9 is bad. The Vbe is 7V. It seems weird to me that it's failed open, but that's the only thing that I can think to explain it.

Oh, and for no obvious reason, the V+ and V- rails are now at +/-27V. I'd like to fix the transistor just to see if that was the problem, but I'm sketched out by the power supply. I don't trust it not to catch on fire or something.

[nsd_c]

[...] for no obvious reason, the V+ and V- rails are now at +/-27V. [...] I'm sketched out by the power supply. I don't trust it not to catch on fire or something.

Somewhere in the neighborhood of +/- 27V is correct:

Bridge rectifiers create +/-26V for the HF (satellite) amplifiers [...]

Look at the power supply schematic. 40VAC goes in, then is bridge rectified and filtered. 40VAC * sqrt(2) ~= 40 * 1.414 ~= 56.6 Vpk-pk; 56.6/2 = 28.3Vpk. Subtract out a little bit for the bridge rectifier diode drops and half the ripple amplitude, and factor in some sloppy tolerances, and you'd get ~26-27V.

Also, I don't think that there's anything to really be sketched out by on the power supply itself; as I've indicated before, it's pretty straight-forward for the +/- V_HF rails and V_LF rails; these are just bridge rectified & cap-filtered transformer secondaries.

What is a little sketchy is how the +/- 15V rails are generated for U1 (quad op amp) on the LF (subwoofer amp) board--see my attached image. +V_LF (+50V) is dropped to +15V through a 1k 2W resistor (R16), which gives (50-15)^2/1000 ~= 1.23W(!) power dissipation. The same thing is done for the negative 15V rail. I measured R16 and R17 using a thermal camera, and these get to ~260 deg. F (130C). This is technically within spec both in terms of the resistors' power handling capacity and what a typical 2W THT resistor is rated, as max temp ratings for 2W 1000 ohm resistors seem to go from 155C on the low end to 235C on the high end on Digi-Key. But, it would be less sketchy if the designers had perhaps mounted these resistors to the rear panel for some heatsinking, but instead, because these are mounted off of the PCB, they conduct heat to the PCB through their leads, which is why the LF boards on all of these units look darkened.

Look like Q9 is bad. [...] I'd like to fix the transistor just to see if that was the problem

I would strongly suggest that you test other transistors on the amplifier for shorts. You can do this in circuit as long as your DMM puts out a relatively low voltage in resistance or continuity modes. The classic failure mode seems to be that Q6 fails, which I presume then is what takes out one or both of the power transistors Q9 and Q10. I've already "been there and done that" with my own unit; I've tried replacing just Q9/Q10 and quickly found that the new Q9/Q10 would blow again immediately after applying power. I tried this (stupidly) a couple of times (with the same result each time) only to end up damaging the PCB to the point of no repair, at which point I constructed a new (copy) PCB to replace the old.

[Obi_Kwiet]

I would strongly suggest that you test other transistors on the amplifier for shorts. You can do this in circuit as long as your DMM puts out a relatively low voltage in resistance or continuity modes. The classic failure mode seems to be that Q6 fails, which I presume then is what takes out one or both of the power transistors Q9 and Q10. I've already "been there and done that" with my own unit; I've tried replacing just Q9/Q10 and quickly found that the new Q9/Q10 would blow again immediately after applying power. I tried this (stupidly) a couple of times (with the same result each time) only to end up damaging the PCB to the point of no repair, at which point I constructed a new (copy) PCB to replace the old.

I did that as one of the first steps. I wasn't able to find any shorts. It's possible that I just had a hard time with the conformal coating on the back, but I dug the probes in pretty good. It should have worked.

I suppose I must have simply forgotten what I measured the voltage rails to be originally. It seemed really unlikely that it wouldn't blow something with that much additional voltage. However, The negative rail has a really nasty sawtooth ripple.

I tested Q6 several times, since that looked like an obvious failure point, and that seemed to be the consensus of people in the thread, but there were no shorts. The voltage across it even looked good with the circuit energized.

I think with no current going into the base of Q9, that somehow caused the overlap region of the output to expand significantly, so Q10 was basically acting as a class A amplifier for low volumes.

Either way, the thing has spent too many summer cooking in a shed, and I don't really want to go fooling with the power supply to deal with that sawtooth ripple. I think I'm going to call it good at this point. Thanks for the help though!

[nsd_c]

The negative rail has a really nasty sawtooth ripple.

What's the amplitude of ripple that you are seeing...? Again, the V_HF supply is unregulated, so there will always be some amount of "sawtooth ripple." Are you sure you weren't just seeing 100 kHz noise from the square wave generator in the class D subwoofer amplifier...? There's only two caps IIRC on those rails, so those would be easy enough to replace if they're failing.

[Obi_Kwiet]

It about 2v p-p IIRC, at a lower frequency than the switching. The switching ripple was riding on top of it, and it only presented on the negative rail.

[nsd_c]

It about 2v p-p IIRC, at a lower frequency than the switching.

Was it at your mains frequency (i.e., 50 or 60 Hz)? That would confirm for certain if that ripple was just simply filter caps charging and discharging.

I decided to play around in LTspice (out of curiosity) to see what sort of load would be required to obtain a 2Vpp ripple, and arrived at an 18 ohm equivalent load dissipating an average of ~39W and drawing ~1.5A. If I have the time tomorrow, I could compare this ripple to what I observe on my own (repaired) unit. Note that there is a 2A fuse on the primary side, so 2A*120VAC = 240W total  consumption before that blows. According to Dale Thompson's site, the sub nominally puts out 50W max, which makes 40W for the satellites seem high...

But perhaps that rail is loaded more heavily as a result of your damaged channel? What if you were to desolder that amplifier board and measure the ripple again and see if the ripple on +V_HF looks now similar to -V_HF? (Ofc., don't forget the mica insulators if you re-assemble the unit, and also don't turn the volume up too high if the TO-220's aren't secured to the case--I made both of those mistakes and blew fuses and transistors both as a result.)

[nsd_c]

Perhaps your memory was foggy; re-reading, I saw you said this earlier:

[...] the negative rail is has 1V saw-tooth ripple in it [...]

1V sawtooth ripple gives a load resistance of ~40 ohms dissipating 18W and drawing ~670mA. This seems more reasonable. Perhaps if Q9 really has failed open, +V_HF is not loaded as heavily as it otherwise would be; replacing Q9 might bring the ripple amplitude of +V_HF in line with what you saw on -V_HF.

[Obi_Kwiet]

Yeah, I did forget. That does seem possible.

I'd like to have tested it, but I just don't have the room to keep it around. Too many other projects on the healing bench! I also busted some traces trying to be lazy, so, eh. I'm not really feeling trying to re-do the board.

[soldar]

Those resistors are not 0 Ohm - 5 W, they are originally 0.5 Ohm although I suppose they could have been replaced with a value of zero.

Their main purpose is to improve feedback and lower distortion. You could probably test with and without and you would probably find lower amplification and lower distortion with the resistors in place.

Q6 is the bias control transistor
Q3 is a current regulator or an voltage regulator

Yup. This design was pretty standard forty or so years ago and I built and repaired dozens. Normally those were adjustable with pots but I guess with very tight tolerances you could do away with the pots.

Have a look at the attached schematic. It is very similar.