Need some pointers with finishing touches on mosfet amp repair

[Welcome]

Hey dudes.

I just successfully repaired some generic PA MOSFET amp. It had a blown channel and wouldn't come out of protection.

A small fire had developed on the board, ruining a bunch stuff. Replaced all that. Also, one pair of the MOSFETs were blown and I decided to just leave them out entirely.

Anyway. Two things:


1. DC offset is super high (adjustable beteen 600 - 700 mV). Q21 was a component I replaced. Should it be matched with Q20? It's hard to get to so that's why I'm asking instead of doing. How can I improve DC offset?

Idle current seems high. The repaired channel runs way hotter than the original. This is my first foray into MOSFET amps and I don't know where to measure this. The "emitter resistors" seem to give meaningless values even on the stock channel. Where do I measure idle current?

I would love some pointers. Thanks in advance.

Schematic:

 

[JonSnell Electronic]

VR3 is the off set. VR4 the quiescent adjust. Measure the voltage between the source of the top transistor and the drain of the lower one. The do not need to be matched pairs, they are Power MOS Fets.

 

[Welcome]

What I meant is, adjusting the DC offset is not good enough. It will only go from around 600 to 700 mV. How do I get the "base" offset down? The stock channel is easily adjustable to 0 mV.

The high idling current is a different issue. It seems way high because the heatsink gets much hotter than the stock channel.

So you're saying I should put the measuring probes on the middle leg (source) of the 2SK1058, and the other probe on the right (drain) leg of the accompanying 2SJ162? I'll try that. Thanks a lot. EDIT: Didn't work. That's just the rail voltage.

 

[Ian Finch]

0.1R is a very low value source resistor. Have you measured voltage across both resistors, e.g R71+R72 and then calculated the current for 0.2R?
With possibly 100mA bias per pair, an expected voltage measurement of 20 mV for each pair should be easy enough to read - unless it is unstable for some reason, if you say the measurements are meaningless values. Can you 'scope it?

 

[Welcome]

Ah-ha! The only thing I missed to measure, somehow. Measuring across both source resistors, I get around 20 mV on the stock channel. Should be fine then I suppose. I adjusted the fixed channel to the same value. Heat is fairly even now I think.

Now the only problem is getting the base DC offset down. Could a matched pair of Q20/Q21 help?

 

[Welcome]

No, fixed channel definitely runs hotter still. Burns to touch that part of the heatsink after a couple minutes of operation. Stock channel is hot, but not painfully so.

To clarify: The stock channel has four pairs of outputs, fixed channel only has three. Bias is 20 mV across 0.2R, so 100 mA per pair. I have left the source resistors of the removed pair intact. This may not be the smartest thing?

 

[JonSnell Electronic]

100mA is rather a lot!

 

[anatech]

Hi JonSnell Electronic,

100mA is rather a lot!

Actually, it's the higher end of normal and expected in many mosfet amplifiers.

-Chris

 

[mrshow4u]

From the description, it sounds like it could be a leaky device, Q22.
Since your offset adjustment isn't working enough, there could be a leakage current that is dominating the correction current.

From your description, I'm assuming a positive 600-700mV offset. If so, I'd expect a correction current coming from the differential stage. This correction could be measured across resistors, R55&R56. Those 2k2's. With a positive offset, and if the differential pair are doing their job, I'd expect a larger drop measured across R56. This would show that the differential pair is functioning correctly and trying to correct the offset.

If you do see this current difference, then I'd expect a leak in Q22 (or something else wrong with the current source). I'd tend to think that there's not an issue with Q20, since there is bias current (maybe on the high side, if not adjustable).

Good luck.

 

[Welcome]

I think I found the problem.

On the both channels there is about 500 mV DC before R54.

On the bad channel, there is about 600 mV DC after it (which is the same value I measure on the speaker output), but on the good one - nothing.

That resistor appears to be open on the bad channel.

It's hard to get to so I haven't replaced it yet.

 

[Welcome]

No, I was wrong. It's not open. But the DC offset starts after this resistor. Before it, everything seems to measure the same as the good channel. The above post mentioned R55 and R56, but both resistors have the same voltage drop (1.7V) on both channels.

Still stuck. Got any more tricks for me?

 

[Welcome]

From the description, it sounds like it could be a leaky device, Q22.
Since your offset adjustment isn't working enough, there could be a leakage current that is dominating the correction current.

From your description, I'm assuming a positive 600-700mV offset. If so, I'd expect a correction current coming from the differential stage. This correction could be measured across resistors, R55&R56. Those 2k2's. With a positive offset, and if the differential pair are doing their job, I'd expect a larger drop measured across R56. This would show that the differential pair is functioning correctly and trying to correct the offset.

If you do see this current difference, then I'd expect a leak in Q22 (or something else wrong with the current source). I'd tend to think that there's not an issue with Q20, since there is bias current (maybe on the high side, if not adjustable).

Good luck.

My previous two posts can probably be ignored.

I now took a better look at R55 and R56. They both have the exact same voltage drop - 1.7 V. On the good channel, R56 has a drop of 1.72 whereas R55 has a drop of 1.67, so there is clearly a difference here.

But I don't know what to make of your post. It says Q22 is broken if there IS a difference, but what about now - there is none. What do I do now?

More measuring: Q22. On the bad channel, I can measure -60 mV between ground and collector, if the DC offset is set to 600 mV. If I set it to 900 mV, this measurement becomes -360 mV. On the good channel, that measurement gives 570 mV with zero offset. Hint?

 

[diy didi]

Q18 &Q19 need to be matched. The same with Q20 and Q21. You can match them at low current. 2mA or so will be good enough.

 

[diy didi]

Match for Vbe and gain as there are no current mirrors present.

 

[AndrewT]

The two input transistors are an LTP. Ideally they should be matched, but the production versions probably use selected pairs to ensure the equipment meets specification.
Similarly the second stage LTP should ideally also use a matched pair.

measure voltages (to audio ground) around the circuit and add them to your sch. Post those results.
Both sides of:
R50, Vr3, R56, R55, R65, R46, R47, R66, R82, R75, R68

 

[AndrewT]

Match for Vbe and gain as there are no current mirrors present.

That stupid diode D7 is supposed to mimic a current mirror.

I can't understand the penny pinchers trying to save 0.3p (0.003GBP)

 

[AndrewT]

.................

I now took a better look at R55 and R56. They both have the exact same voltage drop - 1.7 V. On the good channel, R56 has a drop of 1.72 whereas R55 has a drop of 1.67, so there is clearly a difference here...............

If the LTP is passing EXACTLY 50% through each half, then the voltages across the collector resistors should be identical. The only error would be a slight difference in base current sent to the second stage. If the second stage is also balanced, then that base current error must be zero.

50mV error in the good channel is probably good enough to meet specification. I would be looking for much closer equality (D.Self suggests <2%).
But first one needs to accurately select collector resistor values if one wants to use them to monitor accuracy of the LTP balance.

 

[mrshow4u]

Hi Welcome, I think the prognosis, if I may be so bold as to repeat what others are alluding to, is that the differential pair (LTP Long Tailed Pair) doesn't seem to be doing enough to correct for the offset.

With the same/similar drops across R55 and R56, the diff pair seem happy as pie to leave the offset where it is, .......with an offset, which is bad. If I recall correctly, this amp had an offset control potentiometer, that basically fed a DC offset current into the non-inverting input. You can check that just to make sure that it does, in fact, feed a positive and negative current by exercising the control.
If the control and offset circuit do supply a positive or negative offset current, then you can surmise that the diff pair are not matched well. I think we can assume that the current source, for the diff pair, is fine as there is a reasonable voltage drop across R's 55&56.

Assuming that this DC offset does not result in high bias current when the load is disconnected:
I don't have the schematic infant of me, one other thing to check is if there is an electrolytic capacitor in the shunt leg of the feedback network, make sure that it is not leaky. A leaky cap here will divert some of the offset current to ground instead of letting the inverting input of the diff pair "see" it.

 

[mrshow4u]

Hi Welcome, I just took a look at the schematic. There is no input coupling cap, which for this design, seems like an oversight. ....I mean, they put in a very low current offset control, yet the input has a 2.2k Ohm resistor to the outside world which may have it's own DC component.

So, I'd recommend that you measure the DC offset when the input is shorted. I also noticed a non polar cap to ground in the offset circuit. Keep that is mind when you test the offset circuit for it's ability to apply a DC offset current.

 

[Welcome]

Hi again, thanks for your replies. I will come back with the voltages AndrewT asked for later, but first...

There is actually a capacitor at the input, but it's on the earlier stages of the amp, not on the schematic. When I measure between ground and input, I get about 500 mV offset: on both channels. This voltage is present pretty much everywhere except after R54, where it is gone on the good channel, but even higher (again, about 600) on the bad channel.

I tried to short the input. This happens now: 1.1V offset, and the adjustment pot no longer does much of anything (before it was going from 900-600 mV one end to the other).

Does this say anything useful?