Ampeg B100R power transistors saturating

[dwmorrin]

This Ampeg B100R came to me with 2 shorted power transistors (and a blown fuse). 1 TIP142, and 1 TIP147 shorted - I forgot to note which Q# the shorted ones were before I replaced them all.
I replaced all the power transistors with new TIP142s and TIP147s, and replaced all the other semiconductors in the power amp: (except the diodes which all read good) IC1 (TL074) and IC2 (NE5532), MPSA06, MPSA56, MPSA13.

I'm powering this up with a light bulb limiter, and the mV readings across the emitter resistors (R55-R58) goes up and up, except for R55, which stays low (3.4mV) (Q5's emitter resistor). All the others quickly start shooting towards 100mV and beyond, and the light bulb limiter starts to glow (after ~1 min).

Reading across the power transistors CE (1 probe to collector, 1 probe to emitter), the voltage is dropping.
Reading across the power transistors BE (1 probe to base, 1 probe to emitter), all the voltages are dropping except for Q5, which stays put at 1.1V. (Others are dropping below a volt.)

If I short the bases together (a suggestion from Enzo on another B100R thread) the amp settles down (all the transistors are cutoff), but I'm not sure where to go from here.
I've checked all the components in the bias circuit (Q9, R40, R45) - they check fine. Without the jumper, the PNP and NPN bases are 2.2V apart.
None of the "OPT" parts are installed. No bias trim, or optional resistors.

I've replaced all the emitter resistors (R55-R58) with fresh .33Ω resistors. The old ones didn't read bad, but I wasn't sure (the meter doesn't really read that low accurately).

Any suggestions on where to go from here?

[Enzo]

Yes, adjust the bias control so the amp runs colder. That is OPT8 near Q9 on the drawing. Set it for a larger resistance setting, whichever way you have to turn it for that.

[dwmorrin]

Thanks for the quick response, Enzo.
I should clarify that none of the "OPT" parts are installed. OPT8 is not there to turn, but I can install one. It appears to have left the factory without these components. OPT I'm assuming stands for "optional."

When I originally looked at this (it's been sitting in the "tough dog" pile), I tried installing the bias control, and other "OPT" resistors, but it did not relieve the problem, no matter where the control was set. I will try this again on Monday, perhaps with some measurements, and let you know what happens.

I'm curious if you think the fact that Q5 is just sitting there, not saturating, while the other 3 means anything? The amp used to work without a bias control, just for understanding's sake, I'm wondering why I need to install one now.

Also, should I just install OPT8 and OPT2, or is OPT3 necessary as well? None are currently installed.

-Addition --> Just realized that I probably needed to remove R15 when I installed the bias trim for it to work correctly, right? I think I will try that.

What voltage drop across that MPSA13 (Q9) would you expect to see for this amp? 2.2V is where it's at, and I know we're trying to lower it somewhere between that and zero, but any guess as to where it should usually be?

[mozwell]

while i am not familiar with this amp, 2.2V may be a bit high across Q9, if we look at 0.6V BE drop we end up with 0.5V across each of the 0.33 ohm emitter resistor, giving us 1.5A bias current, a tad high i think.
i would put the OPT parts in and start by adjusting the Q9 CE volts to just over 1.2V.
Measure the voltage across each 0.33 ohm resistor, i would expect 5 to 15mV DC only.

Short CE of Q9, and power up without the bulb limiter, and with no speaker attached.
Measure the power amp rails +/-40VDC, all ok ?
Measure the op amp rails IC1 +/-16V, ok ? and IC2 +/-20V ok ?
Check out R49, R54, D10, D14, D9, C25, C26, R47.
This seems to be a semi floating supply for IC2, as the outputs of IC2 directly drive the main output transistors.
R47, C25, C26, D9 appear to be a bootstrap power supply going up and down with the power amp output volts. If this is gradually floating high, this will cause your output to float high as well.
Is IC2 a NE5532 device, or has someone put something else in there.
If all ok,set OPT8 for max ohms & remove the short on Q9 CE and adjust bias for approx 10mV on the 0.33 ohm emitter resistors.

The sch lists test point voltages, check them all.

[flyingdutchman]

while i am not familiar with this amp, 2.2V may be a bit high across Q9, if we look at 0.6V BE drop we end up with 0.5V across each of the 0.33 ohm emitter resistor, giving us 1.5A bias current, a tad high i think.

The output devices are darlingtons, a Vbe of 1.1V would be about right.

I suspect the OP ran into a bad batch of replacement devices. Think chinese fakes. Reject devices get relabeled with any popular part number that corresponds with the given package so you may find any crappy TO3P transistor labeled as TIP142/147, even if they aren't darlingtons.

[mozwell]

That will teach me for not checking the transistor datasheet first..... (face-palm)
darlington Vbe approx 1.0V to 1.2V at idle , so 2.2V on Q9 CE will be approx correct.

Do these checks with CE of Q9 shorted & no bulb limiter.
DC output voltage, power supply voltages +/- 40V, +/-16V, and the power supply for IC2.
What do TP4, TP5, TP6 measure ?
Lets get it working without any bias from Q9, and then we can set the bias at the end.

Check Q9 is a MPSA13, and some other device is not installed, check the pin out from the data sheet & make sure it is installed the correct way.

Confirm you the power transistors you have are not fakes. Do you have any others from a different source you can try.
Where did you get the transistors from (evil bay ?) or somewhere reputable....

[dwmorrin]

Do these checks with CE of Q9 shorted & no bulb limiter.
-Following readings are with CE Q9 shorted, and no bulb limiter.

DC output voltage, power supply voltages +/- 40V, +/-16V, and the power supply for IC2.
DC output: 0V (reads -1mV on mV setting... seems "too good" - that's measuring at the speaker terminals on the PCB)
+40 reads: 38.03V
-40 reads: -38.01V
+16 reads: 16.60V
-16 reads: -16.58V
IC2 Vcc: 18.70
IC Vee: -18.64

What do TP4, TP5, TP6 measure ?
TP4: -15mV
TP5: 5.9V
TP6: -5.9V

Check Q9 is a MPSA13, and some other device is not installed, check the pin out from the data sheet & make sure it is installed the correct way.
-Q9 is definitely a MPSA13, installed correctly.

Confirm you the power transistors you have are not fakes. Do you have any others from a different source you can try.
Where did you get the transistors from (evil bay ?) or somewhere reputable....
-The power transistors were purchased from Mouser. They are all STMicroelectronics brand transistors in TO-247 cases.

Regarding previous post, NE5532 is a real NE5532, also purchased from Mouser.

Just got back to the bench, and posted the measurements. I will probably try removing R45, and installing OPT3, (effectively changing R45 from 15k to 150k). Remember, no OPT parts are currently installed (seems to have come from factory that way). I previously installed the OPT bias trim, but I think it didn't work because R45 was setting the max resistance to 15k, and we want more resistance there to calm the amp down, right?

Any other comments are appreciated. I'm still curious from a theoretical standpoint, why does 1 of the power transistor's Vce stay put (Q5), with just a few mV across its' Re, while the other 3 rapidly lose Vce and their Re mV rise quickly. This happens with the light bulb limiter... haven't tried letting Q9 loose (no short) without the limiter yet, as I'm afraid of burning out the transistors.

[52 Bill]

I'm still curious from a theoretical standpoint, why does 1 of the power transistor's Vce stay put (Q5), with just a few mV across its' Re, while the other 3 rapidly lose Vce and their Re mV rise quickly. This happens with the light bulb limiter... haven't tried letting Q9 loose (no short) without the limiter yet, as I'm afraid of burning out the transistors.

Q5 and Q6 are basically paralleled, separated only by their emitter resistors. Is it possible that there is broken circuit board trace somewhere between Q5 and Q6?

[dwmorrin]

Q5 and Q6 don't have any broken traces.

So, I removed R45, and installed OPT3, and got a the drops across all the emitter resistors to 0.5mV.
Adding in the OPT2 and OPT8 (trim limiter, and trim pot), and I was able to get all the emitter resistors between 1-2mV with about 2/3 rotation of the trim. (I should note: We are off the light bulb limiter at this point.)

Hopefully problem solved. Just need to add in signal and scope the output. If you don't hear from me again, then all is probably OK.

I guess I will chalk up the weird action of Q5 sitting while the other 3 turned on to small differences in the transistors being amplified when too much bias was applied. Maybe Q5 and Q6 are not very well matched.
How about this question:
When ordering a new set of output transistors from a retailer like Mouser, do you over-order and hand match transistors by beta? Would this depend on the topology of the amp? (As in, some designs need closer matched transistors than others.)
Or do you find that a batch from Mouser is usually close enough to just throw in?

[flyingdutchman]

Matching beta is not really required in reasonably modern amps. The only thing that matters is that you have enough of it. More than enough doesn't do any harm.

Also, darlingtons have built-in base-emitter resistors that help smooth out beta.

One thing you have to be careful about with darlington output stages is thermal runaway. The driver and the power transistor proper share the same chip, and as it gets warm, the driver Vbe and the power Vbe drop in unison. If the thermal contact with the heat sink is less than satisfactory, a darlington is much more prone to thermal runaway than a standard emitter follower with a separate driver transistor.

[dwmorrin]

The amp failed after a few minutes of playing a bass through it.
Fuse blown, and Q6 and Q8 shorted C to E.
With Q6's and Q8's legs pulled up, all the voltages seem OK.
The original set of transistors also had 2 blown, and two seemingly OK, so I think the shorted one's were also Q6 and Q8.

So, the original problem still exists. The bias had been set to 1-2mV across each emitter resistor before closing it up and testing it out.

Maybe one of you can clarify the Q3/Q4 circuits - are those protection circuits for Q5/Q7, or something else? Looks like they turn off Q5/Q7 if too high a voltage develops across R55/R57... am I close to the truth there?

So, where do I go next to figure out how to keep Q6/Q8 from blowing up? Set the bias even colder and see what happens?

[flyingdutchman]

Maybe one of you can clarify the Q3/Q4 circuits - are those protection circuits for Q5/Q7, or something else? Looks like they turn off Q5/Q7 if too high a voltage develops across R55/R57... am I close to the truth there?

These act as current limiters. Since under normal circumstances the current is shared more or less equally between paralleled output devices, sensing the voltage across one of the emitter resistors is usually sufficient, but I've seen versions which used a resistor network to sense the voltage across all the emitter resistors of paralleled devices. More complicated versions also sense the voltage across the output devices, the higher the voltage the less current the circuit will allow - this is called "SOA Protection", and refers to the Safe Operating Area of the output devices.

how to keep Q6/Q8 from blowing up?

When mounting new devices, check that the heatsink is clean, the mounting hole is free of burrs, and the insulating washer is clean too. Preferably use new mica washers and good quality silicone heatsink compound. Check that the device mounting tabs sit perfectly flat on the heatsink. Don't overtorque the mounting screws.
Extruded heatsinks rarely have flatness problems. But a heatsink or thermal coupler made out of aluminum sheet sometimes may not be perfectly flat, and this may impede correct heat transfer.

As I stated in my previous post, darlingtons are very prone to thermal runaway and I suspect that's exactly the problem you're dealing with here. If the thermal path from the device tab to the heatsink is anything less than perfect, there will be problems.

Use the light bulb limiter, connect the amp to a suitable load resistor, and let it work at a moderate signal level so you can keep an eye on it while you let it warm up gently. From time to time, turn the signal down to 0 and check the voltage across the emitter resistors.

[dwmorrin]

Aha, I had screws for the heatsink in the wrong way, and didn't notice that it wasn't clamped down correctly (the other PCB hold down screws fooled me).
I also had not replaced the insulators, which I should I have known better.

I am currently testing it as you suggested, flyingdutchman, and it appears to be OK.

[flyingdutchman]

and it appears to be OK

Good. Don't forget to check Q5 and Q7 too.

When the amp is cold, you can set the bias such that you have about 3-5mV across the emitter resistors. That would be 10 to 15 mA of quiescent current for each transistor. When the amp is warm, with no signal and the load disconnected, you might measure 15mV or so which is OK. Q9 senses the heatsink temperature, and as the heatsink gets warm, its Vbe drops a bit and this compensates for the drop in Vbe in the output devices when they heat up. Since the output devices in operation are always a bit hotter than the heatsink and Q9, this compensation is not perfect, therefore the amp's quiescent current always goes up a bit when it warms up.

Measuring the voltage across the emitter resistors, should always be done with the load disconnected. Most amps have a small DC offset voltage on the output, up to 50-100mV or so, and with a load connected this causes an additional current flow through the output devices, which screws up your measurement. Of course, the measurement should be done with zero signal.