Amp keeps blowing reverb driver chips

[nickb]

No, they are not transposed, and no, there is no wiring issue (unless it was wired wrong at the factory, in which case it wouldn't have lasted as long as it did).

The complaint of "reverb stopped working" is a few years old. Customer didn't want to ship it to me, so he took it to some yahoo who said he couldn't figure it out, but didn't do anything to it (ha). Load of BS. Then other complaints emerged (all due to the 'nothing' yahoo did to the amp). When I finally got it back, it was clear that yahoo couldn't figure out the issue, but poked around extensively with a large, overly hot iron, burning various wires, parts, and traces right off the board. So I fixed all that, traced everything more than once to be sure it was right, etc.

I suspect that one of these driver chips blew to start with (given the description of "amp kinda hums now") I got back when the verb issue first surfaced.

SO what do I glean from this? At some point a few years ago, when the amp was several years out of my shop and probably 8 years or so old, one of those verb drivers went tits up. Then a whole lot of damage was caused by some boob poking around in it. The latter is fixed, but I'm left with, something caused that driver chip to crap out way back when, and that something is STILL apparently causing the same issue after x time, but what that something is, I can't seem to put my finger on.

Here's what it's NOT: 1) wiring 2) wrong parts 3) "transposed" parts. It's also not any particular "bad" part , at least not an op amp or driver, because multiple parts have been in those places.

If there is only 0.4V across the 6.8K resistors then there should be 0.4V from base to emitter and there is no way they will pass any current at idle yet you measure 40mA or so, this cannot be. Leave the 1 ohms in place for now as they will stop the transistors from baking.

Do this and report results:
1) Check the base to emitter voltage at the transistor pins
2) Check the voltage on the output, should be close to zero
3) Measure the voltage at the emitter of each transistor
4) Scope the output ( i.e input to the reverb) in case it's oscillating.


I think you misunderstood the what RG said. With no signal (we now know) there is 0.4V across the base to emitter and therefore no current will flow in the those transistors. Now, suppose the output goes positive and current starts to flow in the load. This current will flow through the top resistor as it supplies IC3 , increase the voltage drop across it and so turn on the top transistor so that now supplies the current. But, at idle the current is, well should be, zero

[wizard333]

BD136: Emitter: 14.76vdc Base 14.26vdc
BD135: Emitter: -15.16vdc Base -14.6vdc; across the 1ohm resistor: other side = -15.1vdc, so drop = .06v; 60 ish mA on that run; it fluctuates some as I measure it.

Reverb works as expected, but those transistors heat up quickly.

With TL061 removed, there is no voltage drop from the filter cap to the socket pin, so there can't be anything other than the chip pulling current (i.e. incorrect wiring, etc).

Perfect continuity from the filter caps to the emitters, and from bases to socket pins 4, 7.

Oh and VAC is 36mv on F-, 117mv on F+, so shouldn't be an issue.

There is no voltage change across the 1ohm resistor with the op amp removed, so no current in the driver.

[nickb]

BD136: Emitter: 14.76vdc Base 14.26vdc
BD135: Emitter: -15.16vdc Base -14.6vdc; across the 1ohm resistor: other side = -15.1vdc, so drop = .06v; 60 ish mA on that run; it fluctuates some as I measure it.

Reverb works as expected, but those transistors heat up quickly.

With TL061 removed, there is no voltage drop from the filter cap to the socket pin, so there can't be anything other than the chip pulling current (i.e. incorrect wiring, etc).

Perfect continuity from the filter caps to the emitters, and from bases to socket pins 4, 7.

Oh and VAC is 36mv on F-, 117mv on F+, so shouldn't be an issue.

There is no voltage change across the 1ohm resistor with the op amp removed, so no current in the driver.

You be done 'n' git yersel' a real headscratch'r thar

What does the 'scope say - any sign of oscillation?

[R.G.]

60 ish mA on that run

And there you have it. 60ma with 14-ish volts across each of the transistors is about 0.84W, and the transistor is going to be HOT. If not dead.

60ma in the transistors is way too much.

I suggest reducing R49 and R51 by paralleling them with other resistors. I'd start with about 47K, just eyeballing the resistor reduction, the currently existing voltage across them, and WAG-ing a value. See if you can reduce the resistors a bit and stop the overheating.

For comparison, big power amps in the 50-100W range, tend to run about 25mA through the opposing output devices - although those amps also have a different output stage topology, and likely other considerations.

But it's worth a try, and paralleling resistors on the top of other resistors is fairly easy to do.

[wizard333]

And there you have it. 60ma with 14-ish volts across each of the transistors is about 0.84W, and the transistor is going to be HOT. If not dead.

60ma in the transistors is way too much.

I suggest reducing R49 and R51 by paralleling them with other resistors. I'd start with about 47K, just eyeballing the resistor reduction, the currently existing voltage across them, and WAG-ing a value. See if you can reduce the resistors a bit and stop the overheating.

For comparison, big power amps in the 50-100W range, tend to run about 25mA through the opposing output devices - although those amps also have a different output stage topology, and likely other considerations.

But it's worth a try, and paralleling resistors on the top of other resistors is fairly easy to do.

Given the parts are well in spec and there are quite a few of these out there running fine with those parts, and this is a PCB amp so, while there are some variations on the pcb (hit one low frequency oscillating late last year), if the parts were not a good value they'd be failing a lot. And they aren't.

So, if this one by chance was oscillating VHF, even if my scope can't see it, the snubbers would go where? Off pin 6? Pin 6 and pin 3? If it is oscillating VHF, and I do lower the resistor values, I'm band-aiding it. If I put snubbers in and it stops with the high current draw, that answers that fairly quickly.

[J M Fahey]

Given the parts are well in spec and there are quite a few of these out there running fine with those parts, and this is a PCB amp so, while there are some variations on the pcb (hit one low frequency oscillating late last year), if the parts were not a good value they'd be failing a lot. And they aren't.

You are giving an excuse for NOT checking something which has a high possibility of solving your problem. .... what would Enzo say about that?
Which to boot is very easy to test, just tack solder a couple resistors in parallel with R49 and R51.
Personally Iīd be more drastic and start with, say, 10K in parallel, which would bring effective value down to around 4000 ohms.

Another nagging question is that your posted voltage measurements do not look accurate enough:

R49 = 6.7K, voltage drop across R49 = ~.4vdc

R51 = 6.73K, voltage drop across R51 = ~.4vdc

such a low resolution makes me think you are using a too high scale to measure, say the 20VDC one or even the 200VDC one, while the proper one in this case is the 2VDC one, where I would expect to see, say, "419mVDC" or some similar 3 significant digit result.

To boot, you donīt seem to measure *across* those resistors which is the proper way but from each end to ground, and then substract values, which is a guarantee of inaccuracy, specially here where a few mV difference across the BE junction makes for a huge current difference, specially that close to transistor turn on voltage.
Which seems to be confirmed by:

BD136: Emitter: 14.76vdc Base 14.26vdc
BD135: Emitter: -15.16vdc Base -14.6vdc

In fact, even as shown above, BD136 would be showing 500mV VBE and BD135 even higher 560mV VCE , WAY higher than quoted

~.4vdc

So please set meter to 2VDC scale and reread voltage across R49 and R51.

Besides that, TL061 idle current consumption is 0.25mA so across 6k8 that would drop staggering 1700mV , not "about 0.4V" ... IF you had removed T1/T2

Going on with circuit analysis: since it "should" drop 1700 mV across R49/51 but such voltage is clamped to around 500/600mV by VBE junctions, then most of TL061 idle current straight goes into T1/T2 base:
0.5V across 6800 ohms accounts for around 0.074 mA so remaining 0.175mA go straight into T1/T2 base.

Now those transistors have current gain Hfe around 150 (typical value according to datasheet graph) with actual spread from 40 minimum to 250 maximum
So typical idle collector current "should" be around 0.175mA*150=26mA

IF you got a "high quality" batch with Hfe 250, then idle current would be 0.175*250=44mA
Looks like my little analysis was quite accurate, after all

As I now see it, Mr Koch developed his design around average BD135/136.
To boot, he might have bought a bucketful (literally) of them, all same batch , and so production for years might have used those, which would match those used in development.

But such a design is marginal (sorry) because it fully depends on Hfe , which as seen varies an amazing 6:1 range (40 to 250) so good design practice is making them relatively immune to part to part variation.
That or provide some bias adjustment trimmers ... exactly like in Power Amps ... which this reverb driver exactly is.
Or: use Hfe selected parts.
In fact, I bet if you bought a pair of BD transistors from Koch, they would send you a couple out of that hypotethic "bucket" and would work fine.

Hereīs the forum screen friendly .png version of .PDFfor quick reference:


As of:

Given the parts are well in spec and there are quite a few of these out there running fine with those parts, and this is a PCB amp so, while there are some variations on the pcb (hit one low frequency oscillating late last year), if the parts were not a good value they'd be failing a lot. And they aren't.

Yes, the parts may very well be in spec ... and yet those with highest Hfe will overheat and die.
Probably not with those picked from the factory stock batch but possible with *some* store bought ones ...as it seems to be the case here.
Mind you: not fakes or doubtful "Hong Kong" parts but the real, best quality ones, caused by a component sensitive design.

Redesign: replace R49/51 by 2200 ohm resistors.
They will drop around 600mV with TL061 idle current and *barely* turn T1/T2 on. (meaning cool transistors).
Practically all of IC1 idle current will pass through those resistors and very little will go through base , yet when IC1 tries to drive the reverb tank through R52 that current will turn booster transistors ON for good, driving the tank as intended.
Good luck

PS: if you feel uncomfortable with such cold biasing, try 3k3 resistors instead, although that should not be an issue with a reverb tank driver power amp.
For minimal messing with the PCB, which has already suffered a lot, tack solder a couple 6k8 resistors in parallel with existing ones.

[nickb]

Given the parts are well in spec and there are quite a few of these out there running fine with those parts, and this is a PCB amp so, while there are some variations on the pcb (hit one low frequency oscillating late last year), if the parts were not a good value they'd be failing a lot. And they aren't.

So, if this one by chance was oscillating VHF, even if my scope can't see it, the snubbers would go where? Off pin 6? Pin 6 and pin 3? If it is oscillating VHF, and I do lower the resistor values, I'm band-aiding it. If I put snubbers in and it stops with the high current draw, that answers that fairly quickly.

Honestly,I doubt it's oscillating at VHF, but never say never. Let's step back and look at DC idle again. You measure 0.4V so those transistors cannot be on. If it were oscillating the average voltage across those 6.8K would be higher and your meter would show it.

A look a the TI datasheet for a TL061 shows a quiescent current of 200uA. Just 80uA thru the 6.8K is 0.6V and that is the point at which your transistors turn on. So there seem to be two issues

(1) The current production idle current of the TL061 is apparently higher than the original design - change the 6.8K to something like 2.7K or 2.2K should get you in the ballpark, IMHO you want those transistors barely on especially as there is no thermal compensation, I suggest <1mA.
(2) The reading of 0.4V is now very suspicious. For your meter check the battery, check it's on DCV and not ACV, check against something else of known accuracy.

[wizard333]

You are giving an excuse for NOT checking something which has a high possibility of solving your problem. .... what would Enzo say about that?
Which to boot is very easy to test, just tack solder a couple resistors in parallel with R49 and R51.
Personally Iīd be more drastic and start with, say, 10K in parallel, which would bring effective value down to around 4000 ohms.

Another nagging question is that your posted voltage measurements do not look accurate enough:

such a low resolution makes me think you are using a too high scale to measure, say the 20VDC one or even the 200VDC one, while the proper one in this case is the 2VDC one, where I would expect to see, say, "419mVDC" or some similar 3 significant digit result.

To boot, you donīt seem to measure *across* those resistors which is the proper way but from each end to ground, and then substract values, which is a guarantee of inaccuracy, specially here where a few mV difference across the BE junction makes for a huge current difference, specially that close to transistor turn on voltage.
Which seems to be confirmed by:

In fact, even as shown above, BD136 would be showing 500mV VBE and BD135 even higher 560mV VCE , WAY higher than quoted
So please set meter to 2VDC scale and reread voltage across R49 and R51.

Besides that, TL061 idle current consumption is 0.25mA so across 6k8 that would drop staggering 1700mV , not "about 0.4V" ... IF you had removed T1/T2

Going on with circuit analysis: since it "should" drop 1700 mV across R49/51 but such voltage is clamped to around 500/600mV by VBE junctions, then most of TL061 idle current straight goes into T1/T2 base:
0.5V across 6800 ohms accounts for around 0.074 mA so remaining 0.175mA go straight into T1/T2 base.

Now those transistors have current gain Hfe around 150 (typical value according to datasheet graph) with actual spread from 40 minimum to 250 maximum
So typical idle collector current "should" be around 0.175mA*150=26mA

IF you got a "high quality" batch with Hfe 250, then idle current would be 0.175*250=44mA
Looks like my little analysis was quite accurate, after all

As I now see it, Mr Koch developed his design around average BD135/136.
To boot, he might have bought a bucketful (literally) of them, all same batch , and so production for years might have used those, which would match those used in development.

But such a design is marginal (sorry) because it fully depends on Hfe , which as seen varies an amazing 6:1 range (40 to 250) so good design practice is making them relatively immune to part to part variation.
That or provide some bias adjustment trimmers ... exactly like in Power Amps ... which this reverb driver exactly is.
Or: use Hfe selected parts.
In fact, I bet if you bought a pair of BD transistors from Koch, they would send you a couple out of that hypotethic "bucket" and would work fine.

Hereīs the forum screen friendly .png version of .PDFfor quick reference:
[ATTACH=CONFIG]42546[/ATTACH]

As of:
Yes, the parts may very well be in spec ... and yet those with highest Hfe will overheat and die.
Probably not with those picked from the factory stock batch but possible with *some* store bought ones ...as it seems to be the case here.
Mind you: not fakes or doubtful "Hong Kong" parts but the real, best quality ones, caused by a component sensitive design.

Redesign: replace R49/51 by 2200 ohm resistors.
They will drop around 600mV with TL061 idle current and *barely* turn T1/T2 on. (meaning cool transistors).
Practically all of IC1 idle current will pass through those resistors and very little will go through base , yet when IC1 tries to drive the reverb tank through R52 that current will turn booster transistors ON for good, driving the tank as intended.
Good luck

PS: if you feel uncomfortable with such cold biasing, try 3k3 resistors instead, although that should not be an issue with a reverb tank driver power amp.
For minimal messing with the PCB, which has already suffered a lot, tack solder a couple 6k8 resistors in parallel with existing ones.

I measured both across and to ground; the value I posted is ~ because the reading was never quite stable. It varied, but not by a ton. That's with T1/2 IN, with them out, there is no voltage drop period.

Nor am I "Making an excuse.....", it's just as easy to clip in some snubbers to see if that is the problem as to clip in some resistors, and while I'm at it, I can try both to see which addresses the issue, which seems to be either

1) bias of driver transistors,
2) VHF oscillation.

But since the historic placement/value of snubbers seems to be a state secret and cause cries of "heresy!" when inquired about, I'll figure it out.

[R.G.]

About state secrets -
It's not that the arrangement of small caps in circuits is a secret so much as it is random and non-reproduceable. Whenever I see several different small caps arrayed around in a circuit, it's a red flag that it once oscillated and the designer tossed in these caps till he tuned it out.

And it is a matter of tuning. RF oscillation is predictable, if you have any idea what parameters are there inside the devices to cause the phase shifts and impedance variations that sustain it. Problem is, EVERYTHING affects the circuit's impedances, including temperature, time and the phase of the moon.

[As a side note, they're not snubbers. Snubbers are usually RC circuits used to eat the energy from edges or transients. These are tuners. But I digress. ]

So unless you like RF puzzles (and I have some ham-radio-engineer friends that do), the right way to approach this is to eliminate the DC and low-frequency AC issues first, then go off into the tall grass chasing RF ghosts. I can assure you that if the more likely situation of DC bias problems is an issue, no amount of puttering around tagging 10pF tuners from anywhere to anywhere else will fix it. It is vastly more likely to be a DC problem. If for no other reason than economy of time, eliminate that issue first, then go off into Smith-chart land.

My best advice is to try lowering the resistors first. Once that's out of the way, go off into sprinkling in RF tuners. Well, actually, the best advice is to the original designer not to do the circuit that way, but I think I've missed that boat. .

[nickb]

But since the historic placement/value of snubbers seems to be a state secret and cause cries of "heresy!" when inquired about, I'll figure it out.

I deliberately avoided making any suggestions about stability for two reasons. First it's very unlikely and we have something that is much more likely (2.2K resistors) to try first. The second is that it would be very hard to do. Despite the apparent simplicity of the circuit there are many things that will cause phase shifts and therefore I don't there is a simple "add a cap here type solution". It's much more tricky than that. I'm just trying to keep you out of trouble.

[J M Fahey]

I measured both across and to ground; the value I posted is ~ because the reading was never quite stable. It varied, but not by a ton. That's with T1/2 IN, with them out, there is no voltage drop period.

Sorry but thatīs impossible.
IC1 idle current across R49/51 causes a voltage drop of 200/250uA*6800 ohms=1360/1700 mV , which is clamped down by BE junctions to about 600mV (notice I donīt even consider the 400mV measurement).
If you remove such clamps, meaning you remove T1/T2, voltage across R49/R51 will *rise* to above calculated value, never ever drop to 0.

Nor am I "Making an excuse.....", it's just as easy to clip in some snubbers to see if that is the problem as to clip in some resistors, and while I'm at it, I can try both to see which addresses the issue, which seems to be either

Sorry, nothing personal, itīs just that a popular Enzoism around here is: "never find an excuse NOT to check something".

As of overheating, and even more having *confirmed* that those transistors are passing 60mA so dissipating 0.84W , I *already* see a thermal issue strong enough to justify it.
Even worse, there is NO thermal compensation built in, so thermal runaway becomes a distinct possibility.

Can there be other possibilities?
Yes, maybe, but (sorry for the second Enzoism): "when I see hoof marks in the snow, I think horses, not zebras".

1) bias of driver transistors,
2) VHF oscillation.

But since the historic placement/value of snubbers seems to be a state secret and cause cries of "heresy!" when inquired about, I'll figure it out.

#1 is he standard "horses" answer; as of #2, *maybe* zebras they escaped from a circus touring the North (or South) Pole , not impossible but less likely.

As a side note, if itīs a State Secret, maybe they kill all who find it, just to keep it so
What will Alex Jones, Rense, Snowden or Wikileaks say about that ?

And if truly the answer, makes me wonder about all those Koch Twintone all over the World happily running without snubbers.

Besides, TL061 is sssslllllloooooowwwwwww .... like all low power Op Amps ; way slower than TL07x
Oscillating/unstable?
Yes, why not?
Frequency so high that even a 20MHz scope shows no trace of it?
Not a chance.

[wizard333]

Sorry but thatīs impossible.
IC1 idle current across R49/51 causes a voltage drop of 200/250uA*6800 ohms=1360/1700 mV , which is clamped down by BE junctions to about 600mV (notice I donīt even consider the 400mV measurement).
If you remove such clamps, meaning you remove T1/T2, voltage across R49/R51 will *rise* to above calculated value, never ever drop to 0.

Sorry, nothing personal, itīs just that a popular Enzoism around here is: "never find an excuse NOT to check something".

As of overheating, and even more having *confirmed* that those transistors are passing 60mA so dissipating 0.84W , I *already* see a thermal issue strong enough to justify it.
Even worse, there is NO thermal compensation built in, so thermal runaway becomes a distinct possibility.

Can there be other possibilities?
Yes, maybe, but (sorry for the second Enzoism): "when I see hoof marks in the snow, I think horses, not zebras".

#1 is he standard "horses" answer; as of #2, *maybe* zebras they escaped from a circus touring the North (or South) Pole , not impossible but less likely.

As a side note, if itīs a State Secret, maybe they kill all who find it, just to keep it so
What will Alex Jones, Rense, Snowden or Wikileaks say about that ?

And if truly the answer, makes me wonder about all those Koch Twintone all over the World happily running without snubbers.

Besides, TL061 is sssslllllloooooowwwwwww .... like all low power Op Amps ; way slower than TL07x
Oscillating/unstable?
Yes, why not?
Frequency so high that even a 20MHz scope shows no trace of it?
Not a chance.

Not sure what to tell you on part 1; when I remove the op amp, there is 0 voltage drop across those resistors. At least, none large enough for my meter to measure, so it's significantly smaller than a millivolt if it's there.

I experimented with different resistor values; at 10k paralleled the drivers get warm but not scalding; at 6.8k in parallel they are room temperature. In either case the verb seems to still work. How cold is too cold? When the verb stops working, has audible crossover distortion, or at some point before that?

I'm not seeing an oscillation issue on the scope, so we'll assume it's not there.

[R.G.]

Not sure what to tell you on part 1; when I remove the op amp, there is 0 voltage drop across those resistors. At least, none large enough for my meter to measure, so it's significantly smaller than a millivolt if it's there.

With the opamp out, there's no current pulled through them. The collector base junction on the silicon bipolars doesn't leak enough to show.

I experimented with different resistor values; at 10k paralleled the drivers get warm but not scalding; at 6.8k in parallel they are room temperature. In either case the verb seems to still work.

Hold your bids, we have a winner!

How cold is too cold? When the verb stops working, has audible crossover distortion, or at some point before that?

It's quite difficult to tell a good stopping point other than audible distortion. A good distortion analyzer might help, but if there's no audible distortion, just pick a place.

It is difficult in such current dumping schemes to say one point's worse than another. Fact is, the opamp itself can supply several ma of current, and that current, pulled through the bias resistors and Vbes is plenty to get the transistors to dump current in the load until they get the output within range of the opamp's current output. So picking "correct" points is difficult.

I think this is the point where you make the resistors such that the transistors get just barely thermally warm enough to notice, then declare victory.

I'm not seeing an oscillation issue on the scope, so we'll assume it's not there.

If the resistors cured the heating, it's not RF oscillation. You're done.

Where's the party, and who's buying?

[J M Fahey]

Congratulations, you solved it

Personally, Iīd use the cold setting (around 3k3 final value) ,simply because we have been burnt (in this case literally) before.
Donīt think you have noticeable crossover distortion, if any at all, because as RG said, the first few mA are supplied by the Op Amp itself and transistors start working only beyond that .

All is well that ends well

[Enzo]

Plus any such crossover distortion would have to be apparent AFTER leaving the spring mechanism.