Help Me Solve This Power Amp Riddle

[Phostenix]

I've got a Crate amp that I'm looking for some assistance with. It came to me with the usual blown output transistors & emitter resistors. I bought parts off ebay a few years back (I didn't know about this counterfeit transistor thing) & just recently got back to this amp. The amp came up fine at first & I played through it for about a half hour at small room levels into a single 12" 8 ohm speaker. The xstrs were, of course, fakes & blew when I connected a 4 ohm load - 2 12" 8 ohm speakers in parallel - taking with them R9, R18, & the bias transistor Q11.

I've now replaced all the blown parts with stuff from Mouser, so I'm confident that is no longer an issue.

The amp comes up looking pretty normal as far as voltages go, but it plays more like a 5 watt amp than a 350 watt amp. During its short stint with the fake xstrs, it was very loud at low volume settings, as you would expect from this amp. Now, I can turn it all the way up and, although it sounds normal, the level is WAY down.

If I take the Line Out & connect it to another power amp & speaker, it works fine. If I insert the output of my ME-50 into the Line In, I get the same low level output. If I use a signal generator to insert a signal into the power amp, it will take a huge (~15v p-p) and reproduce that to the speaker out without clipping (into an 8 ohm resistive dummy load viewed on the scope), but normal guitar level signals in only get about 20v p-p on the output.

With the rails at +-50vDC, I can set the bias so that there is anywhere from 0V DC to over 30mV DC across R3-R5. but R6-R8 stay at about 3mV no matter what I do with the bias control. At this point, Q1-Q12 have been replaced with new parts. D1, D6, D3, & D5 have been replaced. R3-R5 have been replaced. Q13-Q15 test good out of circuit with a diode test. I don't think there's a part in the power amp circuit from Q13 on that I haven't checked (out of circuit, if necessary). DC offset is around 100mV. The voltage across R9 is a little less than a volt with no signal in.

With the bias set for ~30mA through Q1-Q3 (10mV DC across R3-R6), Vbe on Q1-Q3 and Q4-Q6 are all about .54v DC. Vbe on Q8 is 1.3v DC, but Vbe on Q7 is only .54v DC. I can increase the bias to the point where I get Vbe on Q7 to about 1 V DC (around 100mA through R3-R5), but the heatsink will heat quickly with no signal, and the currents in R6-R8 stays down about 9mA (3mV across the emitter resistors).

BTW, I currently have R10 & R15, D5 & D6, and Q9 & Q10 out of the circuit to remove the protection circuitry. I thought it might be activating, pulling down the drives, but it makes no difference with it out.

I've run out of ideas on this one and am looking for some help with what the bias setting should be. Any ideas?

[Phostenix]

Also, I can put an 800mV p-p 1KHz sine wave into the input of the amp & get a clean output into the dummy load - no crossover distortion (clean channel). The output into the load is right about 20v p-p. An increase in the input signal from there gets clipped at the 20v p-p level.

[Phostenix]

If I insert a 1KHz sine wave into the Line In, I can run it up about 15v p-p (the limit of my signal generator) and get a 50v p-p output into the dummy load - clean, no crossover or clipping.

[R.G.]

I believe that your failure may have blown something in the negative feedback gain setting resistors/capacitors.

Think of this thing as a big opamp. The gain is 1+Rf/Ri. Rf is R21||C9. Ri is C11 series to R25 and R31 provides some current feedback. If you happened to open C11 by the death/abuse, then the gain drops to one.

What you describe is simply low gain - which fits the hypothesis. R25 and/or R31 might be open as well, but C11 is more fragile.

[Phostenix]

I thought that, too and checked those. I'll do it again, but I do remember C11 being about 90uF.

What kind of voltage gain should I see through a power amp like this? I'll put a signal into the input again & trace it through the PA. I did this before, but it's all becoming a blur to me. I've got tons of time into this crazy thing now. I remember the signal dropping significantly at the input of the diff pair, but I assume that's normal, no?

[R.G.]

What kind of voltage gain should I see through a power amp like this? I'll put a signal into the input again & trace it through the PA.

Line in (something like 1V peak) produces full unclipped output.

I remember the signal dropping significantly at the input of the diff pair, but I assume that's normal, no?

No, that's not normal.

Pull out Q16 and see what happens.

[gbono]

What does Q16 do? It looks like it's tied into the feedback of the amp or is that a test point (#1 marker) on the drain?

[Phostenix]

Winner, Winner, Winner!! Genius! Thank you! I have full output now.

Here's how much time I've spent with this amp: I read your post as an email on my phone while we were out at dinner. I said to my wife, "Q16? That's the muting FET. You know, I never have tested that...."

I was working under the misconception that the differential pair acts like the inputs to an op amp, so I wasn't surprised that the input went away on the right side of R1....

The DC offset went from over 100mV to around 5mV now.

So, now I have 2 follow ups:

1. What should the bias be set to on the output transistors? It can go as high as 100mA and I see a lot of output xstrs set that high, but I'm not sure if I need to run this that hot.

2. After it warms up, if I disconnect the leads to my signal generator (leaving them connected to the input of the amp), the amp will immediately break into oscillation with a 10kHz square wave on the speaker output (8 ohm resistive load) at full clipped output. This doesn't happen if I unplug the input from the amp (I haven't looked to see if there is a mute switch on the input) & it doesn't happen with an unterminated guitar cable plugged in. So, in the real world, the conditions for oscillation will probably never be met, but it's still very disconcerting.

The protection circuits are all re-installed now.

Thanks for the quick help on this R.G.! I'll check some of the other components around Q16, as well.

[Phostenix]

What does Q16 do? It looks like it's tied into the feedback of the amp or is that a test point (#1 marker) on the drain?

I'm not sure why ground is noted on the schematic with different numbers. I should look on the PCB to see if there's a grounding topology that is being referred to. I have noticed that there is a main star topology that converges near the output.

[Phostenix]

There are, in fact, 7 legs off the star ground on the PCB. The components that have the same number in the ground symbol on the schematic are all on the same leg of ground, generally speaking. There is a very short leg with 1 component & it gets labelled with a 3 in addition to the others on the "3" leg and 7 is actually an offshoot of 5.

[Phostenix]

I figured out the oscillation problem. I had a set of test leads across R9 so that I could monitor the voltage across it as I brought up the power with the variac. R9 burned on me previously, so I wanted to make sure that didn't happen again (although it is a flameproof resistor now). With those leads hanging off the inputs to the power transistors, the amp will oscillate with no input. All is well now without the leads.

Now, I'm just looking for some input on where the bias should be set.

[R.G.]

Here's how much time I've spent with this amp: I read your post as an email on my phone while we were out at dinner. I said to my wife, "Q16? That's the muting FET. You know, I never have tested that...."

I've been there. Someday I'll tell you the "Well, Duuuuh!" story.

I was working under the misconception that the differential pair acts like the inputs to an op amp, so I wasn't surprised that the input went away on the right side of R1....

It's not a misconception. They do act exactly like the inputs to an opamp. the + input (which is where the signal comes into this one) follows the input signal almost perfectly. I think you're remembering that the inverting (-) input that doesn't wiggle in an inverting opamp setup. But on a non-inverting amp, the + and - inputs both move with the input signal, almost exactly.

Now, I'm just looking for some input on where the bias should be set.

That's a tough one to describe. The right way is to hook it up to a distortion analyzer and increase the bias until the distortion residual hits a minimum. As you increase the bias current, the distortion first declines from where it was when biased too cold, hits a minimum, then rises again as you get into gm doubling. The only thing wrong with doing it that way is that you need a distortion analyzer.

Lack of a distortion analyzer is what has resulted in all the different ways to set bias in class AB solid state amps. Frankly, I like using my picoscope in spectrum analyzer mode, and watching the spurs of the second through seventh harmonics of an input sine wave. Since you don't care what number to put on the distortion, only hitting a relative minimum, this works OK.

It's the only good way I know of to avoid biasing too hot and getting into thermal runaway without setting, letting it heat up, resetting, letting it heat up, etc.

[Phostenix]

Well, I don't have a distortion analyzer or picoscope & my spectrum analyzer is line level in only, so I guess I have to go with setting & checking. I never get crossover distortion, even if I turn the bias adjustment all the way down, so I'm inclined to go with the mid point of the pot and go from there (the original setting was about 2 o'clock on the trim pot). Am I mainly looking for the bias current to remain rock solid even after the amp heats?

[R.G.]

Well, I don't have a distortion analyzer or picoscope & my spectrum analyzer is line level in only, so I guess I have to go with setting & checking. I never get crossover distortion, even if I turn the bias adjustment all the way down, so I'm inclined to go with the mid point of the pot and go from there (the original setting was about 2 o'clock on the trim pot). Am I mainly looking for the bias current to remain rock solid even after the amp heats?

Yep, that's why no one else does biasing that way either.

Yes, you're looking for the bias to stabilize and not keep drifting up. This avoids one issue with bias, but doesn't really address the gm-doubling thing. But it's sure better than having the amp go into thermal runaway.

I got a neat toy at Harbor Freight. It's an infrared no-contact thermometer. Takes temps by point-and-shoot. It's worth the $30.

[Euthymia]

Regarding setting the bias, the technique that Enzo taught me is to monitor the amplifier's overall current draw (ammeter on the mains) and slowly turn the bias pot until the current draw starts to jump sharply, then back off a hair.