Was it the 1K or something else? He shouldn't worry if it is the 1K because smoke is what indicates that something is wrong. If everything is fine there will be little if any drop across the 1K if there is a big drop then the power is large and it smokes. However, he's seeing an oscillation which is not predicted or expected so I'm thinking to take an even easier approach.
This would be to isolate the power transistors from everything and do an in place breakdown test. If it was not oscillating and the resistor smoked then the power transistor on the negative side is breaking down. But that transistor tested OK the last time he tested it so it could be a wiring error - don't know ...
He could remove them if we think he should redo the heat sink compound anyway.
They cannot oscillate when isolated.
A bad output could explain the oscillation, but if all the outputs test OK then we have
to question if it is the wiring inductance, or the bias stage that has no "Cdom" cap.
He could try 100 pF caps there since they should be operating at essentially DC.
This would be to isolate the power transistors from everything and do an in place breakdown test. If it was not oscillating and the resistor smoked then the power transistor on the negative side is breaking down. But that transistor tested OK the last time he tested it so it could be a wiring error - don't know ...
He could remove them if we think he should redo the heat sink compound anyway.
They cannot oscillate when isolated.
A bad output could explain the oscillation, but if all the outputs test OK then we have
to question if it is the wiring inductance, or the bias stage that has no "Cdom" cap.
He could try 100 pF caps there since they should be operating at essentially DC.
Last edited:
I'm back, sorry, busy weekend.
Yes, it was the 1k resistor that smoked. It was only 1/8 w resistor. What's really odd is that after I hooked up another one, it hasn't smoked when I turn the power up to the point at which it starts to oscillate. Maybe I just turned it up a little too quickly the first time and power surge at turn-on smoked it?
David, no worries, I have full documentation and was aware of the 1207/1270 typo. I put in a 1207. So that, at least, can be eliminated from consideration.
At this point I have replaced every single capacitor of every type on every board (not counting caps in the PS). I did find a "technician-induced error" in that I had put C2 in C4's position at first (so that C4 was half the value it should be), but the oscillation was present before I made that error and is still present with the correct values in both positions.
I'll do Pete's tests so that we can definitively rule out the output stages. But I keep coming back to this logic:
1. Initial fault in amp: blown fuse on negative output board, 81vdc at the speaker out. Fuse only blows under load, amp is stable unloaded (i.e. no oscillation).
2. Find Q4 is bad. I clean the boards and replace: Q4, Q7, Q8, R7, R8
3. With no load, circuit oscillates.
Ergo, I did something bad in the replacement process, or one of these parts is not good. Of course, all of these parts have been re-tested and seem good, but I have limited test equipment and skills.
I know this is simplistic, and not based on analyzing the function of the circuit, I just bring it up as food for thought.
Yes, it was the 1k resistor that smoked. It was only 1/8 w resistor. What's really odd is that after I hooked up another one, it hasn't smoked when I turn the power up to the point at which it starts to oscillate. Maybe I just turned it up a little too quickly the first time and power surge at turn-on smoked it?
David, no worries, I have full documentation and was aware of the 1207/1270 typo. I put in a 1207. So that, at least, can be eliminated from consideration.
At this point I have replaced every single capacitor of every type on every board (not counting caps in the PS). I did find a "technician-induced error" in that I had put C2 in C4's position at first (so that C4 was half the value it should be), but the oscillation was present before I made that error and is still present with the correct values in both positions.
I'll do Pete's tests so that we can definitively rule out the output stages. But I keep coming back to this logic:
1. Initial fault in amp: blown fuse on negative output board, 81vdc at the speaker out. Fuse only blows under load, amp is stable unloaded (i.e. no oscillation).
2. Find Q4 is bad. I clean the boards and replace: Q4, Q7, Q8, R7, R8
3. With no load, circuit oscillates.
Ergo, I did something bad in the replacement process, or one of these parts is not good. Of course, all of these parts have been re-tested and seem good, but I have limited test equipment and skills.
I know this is simplistic, and not based on analyzing the function of the circuit, I just bring it up as food for thought.
Hi Fred,
I think the burning of the resistor is from the remaining fault(s). This test procedure works fine in spice with not enough power in the 1k resistor to even burn a 1/8W. I think it was somewhere around 50 mW.
I think you should check carefully the rewiring you did when you put it all back together.
Do a point to point test with you ohm or diode function in your DVM. Use the schematic to pick the point. For example test from R17 to the base of Q12, emitter of Q12 to base of Q10 etc. This way if you made an error in wiring you will find it.
David.
I think the burning of the resistor is from the remaining fault(s). This test procedure works fine in spice with not enough power in the 1k resistor to even burn a 1/8W. I think it was somewhere around 50 mW.
I think you should check carefully the rewiring you did when you put it all back together.
Do a point to point test with you ohm or diode function in your DVM. Use the schematic to pick the point. For example test from R17 to the base of Q12, emitter of Q12 to base of Q10 etc. This way if you made an error in wiring you will find it.
David.
Just discovered that R16, which I had replaced, also burned up when the 1k resistor went. So it would appear that a lot of current came off the emitter of Q4? I didn't see any shorts, bridges or other soldering/assembly issues.
I'll replace it and look for more damage in that area.
I'll check the wiring as David suggests, but all looks correct visually.
I'll replace it and look for more damage in that area.
I'll check the wiring as David suggests, but all looks correct visually.
Last edited:
Hi Fred,
The current through R16 is set by the voltage across it. D5 and D6 set the voltage into the base Q4 through R15 minus the base emitter voltage, (in theory 0.7V). There will be some small voltage loss across R15 but the current here is small. The only way for this resistor R16 to see enough current to burn open is if one of the diodes D5 or D6 were open or Q4 is shorted and enough current is supplied to the collector of Q4.
Something besides a 1k resistor would have to be supplying the current to Q4's collector for this to happen. So what might that be? Q7 and Q12 are suspect. There maybe something wrong in those areas.
David.
The current through R16 is normally (according to my schematic) 4 mA + a small amount of base-emitter current.
The current through R16 is set by the voltage across it. D5 and D6 set the voltage into the base Q4 through R15 minus the base emitter voltage, (in theory 0.7V). There will be some small voltage loss across R15 but the current here is small. The only way for this resistor R16 to see enough current to burn open is if one of the diodes D5 or D6 were open or Q4 is shorted and enough current is supplied to the collector of Q4.
Something besides a 1k resistor would have to be supplying the current to Q4's collector for this to happen. So what might that be? Q7 and Q12 are suspect. There maybe something wrong in those areas.
David.
The current through R16 is normally (according to my schematic) 4 mA + a small amount of base-emitter current.
That transistor is a current source, it should not be possible for it to fail for any voltage available in the amp that you put on the collector. You did test it with T1 on, but even turned on it would have just supplied a current equal to V(one diode drop)/R16 or about .6/130 = 4.6 mA no matter what voltage you put on the collector. This is the definition of a current source. T1 being on, and R16 being the wrong value could cause higher current, but for R16 to actually overheat I believe that Q4 had to have failed with a C-E short and E-B open. I suggest that you remove, and test R16 and Q4 but leave them out. They can't fail if they are out and let's finish testing the output stage. I suggest that we test the diode string there before you put in new parts but after we finish the output stage. This was not the time to put in new parts when we have not tested all the parts, we just risk damaging the new parts when the old outputs were still in the amp. It is a learning experience so don't worry about it. A failed amp is a bit like a crime scene, we want to minimize the amount of changes or work done until we figure out who done it as far as failed components go. We need to finish the detective work.
I am a bit concerned that the drivers and outputs are oscillating due to stray wiring inductance, I'd say the probability is low but pulling the drivers should eliminate the oscillation.
OK, if I were more inexperienced I would have tested every transistor and probably diode before rebuilding and powering up the front end. I probably would have pulled all the outputs and tested them out of the amp, and checked the insulators but it is possible to test them in place. What do you prefer do you want to remove the outputs or test them in place? I'm not sure how difficult it is to desolder the outputs or what you are more comfortable with. I'll give you more instructions once I know which method you prefer.
I am a bit concerned that the drivers and outputs are oscillating due to stray wiring inductance, I'd say the probability is low but pulling the drivers should eliminate the oscillation.
OK, if I were more inexperienced I would have tested every transistor and probably diode before rebuilding and powering up the front end. I probably would have pulled all the outputs and tested them out of the amp, and checked the insulators but it is possible to test them in place. What do you prefer do you want to remove the outputs or test them in place? I'm not sure how difficult it is to desolder the outputs or what you are more comfortable with. I'll give you more instructions once I know which method you prefer.
Last edited:
Doner Transistors and Test Results
The amp that I have for doner parts is a Soundcraftsman by the way. I pulled 4 outputs, they are socketed so it is easy, not that I like sockets. Soundcraftsman uses the same Toshiba outputs but bins them for beta, and relabels them. One channel has been repaired with regular Toshibas both are O beta grade.
One of the factory types was shorted C-E that must have been the bad channel. The other 3 tested fine on the IT-18 and then I did a full voltage leakage test. Briefly, I used a 10K ohm resistor as a current limiter and current sense device. All I did was to put the full ~160 V from C to E in the right direction, be sure to get it right, with the 10K resistor in series. Two had zero volts across the resistor so the leakage was zero, and one bounced between zero and 1 mV across the sense resistor which is fine. The data sheet says that the allowable leakage current is 50 mA (actually another data sheet says .1A or 100 mA) and this is a typo, that is above the bias point for the amp, it should have read 50 uA or one thousand times smaller. 50 uA through a 10K resistor produces .5V so you can see that the 1 mV was essentially nothing. These parts have good beta and essentially no leakage at full voltage so I'd say they are good. We can do this test in or out, it is up to you, but let me give you more detail before you proceed.
I could also do a Collector to Base leakage test, but if there was any leakage, it would be multiplied by beta, and show up as leakage in the Collector to Emitter test. No need to do the Collector to Base test.
The amp that I have for doner parts is a Soundcraftsman by the way. I pulled 4 outputs, they are socketed so it is easy, not that I like sockets. Soundcraftsman uses the same Toshiba outputs but bins them for beta, and relabels them. One channel has been repaired with regular Toshibas both are O beta grade.
One of the factory types was shorted C-E that must have been the bad channel. The other 3 tested fine on the IT-18 and then I did a full voltage leakage test. Briefly, I used a 10K ohm resistor as a current limiter and current sense device. All I did was to put the full ~160 V from C to E in the right direction, be sure to get it right, with the 10K resistor in series. Two had zero volts across the resistor so the leakage was zero, and one bounced between zero and 1 mV across the sense resistor which is fine. The data sheet says that the allowable leakage current is 50 mA (actually another data sheet says .1A or 100 mA) and this is a typo, that is above the bias point for the amp, it should have read 50 uA or one thousand times smaller. 50 uA through a 10K resistor produces .5V so you can see that the 1 mV was essentially nothing. These parts have good beta and essentially no leakage at full voltage so I'd say they are good. We can do this test in or out, it is up to you, but let me give you more detail before you proceed.
I could also do a Collector to Base leakage test, but if there was any leakage, it would be multiplied by beta, and show up as leakage in the Collector to Emitter test. No need to do the Collector to Base test.
Possible Further Damage to the Front End
We will test the voltage across D5 and D6 at a later time, if they are blown, then Q3 is probably also blown - I doubt it, but just to play it safe we will test that. I had suggested and for all further testing keep P1 to it's minimum resistance - ALWAYS. It protects Q9 and Q10 in the min position. Since there was high current through Q4 and the additional 1K resistor, there might have been high current through P1 and Q9, Q10 if the pot was not at the min setting. These transistors and P1 could be damaged. I would remove Q9 and Q10 and replace them, they are dirt cheap and if you have good outputs in the amp we don't want them to blow. Leave them out if you take them out now. While they are out, test resistors around them as they are in place. Test P1 by hooking up your ohm meter and watching for jumps as you move it from end to end.
My guess is that R16 was the wrong value, better to have found this out with the old outputs and test drivers, than with new parts.
We will test the voltage across D5 and D6 at a later time, if they are blown, then Q3 is probably also blown - I doubt it, but just to play it safe we will test that. I had suggested and for all further testing keep P1 to it's minimum resistance - ALWAYS. It protects Q9 and Q10 in the min position. Since there was high current through Q4 and the additional 1K resistor, there might have been high current through P1 and Q9, Q10 if the pot was not at the min setting. These transistors and P1 could be damaged. I would remove Q9 and Q10 and replace them, they are dirt cheap and if you have good outputs in the amp we don't want them to blow. Leave them out if you take them out now. While they are out, test resistors around them as they are in place. Test P1 by hooking up your ohm meter and watching for jumps as you move it from end to end.
My guess is that R16 was the wrong value, better to have found this out with the old outputs and test drivers, than with new parts.
Last edited:
Hi Pete,
You are right in your analysis of Q4, D5 and D6. They shouldn't be damaged in an amplifier failure. Of course, if any failed (Q4 short or leaky C-E, the diodes open), this would also damage the rest of the amplifier. I haven't looked at the schematic today to analyze these failure modes, but I can say with certainty that these parts are normally not harmed in any way when an amplifier channel fails.
Removing the outputs is a pain in the butt on these amps. I normally bite the bullet and do this work on a channel failure type repair. Not fun.
Now as far as installing incorrect parts, or even answering the question of what was done for sure is something that writing out that list I recommended would have helped with. Remember, observations only - no conclusions allowed at all. Just test results and observations (just the facts 'mam).
As for working on your own amplifier, label each variable resistor as to what channel it came from. Don't adjust it at all, but you can measure it. Reinstall it once you are done building and have confirmed the control is good. Your bias current setting should be close to correct now. Remember that the Adcom bias procedure involves getting the output stage hot, then allowing it to settle into it's rest state from there. The bias reading will not be correct until this is done, so fight off the urge to measure the bias and correct it right away. This should save you time in setting it back up.
Best, Chris
You are right in your analysis of Q4, D5 and D6. They shouldn't be damaged in an amplifier failure. Of course, if any failed (Q4 short or leaky C-E, the diodes open), this would also damage the rest of the amplifier. I haven't looked at the schematic today to analyze these failure modes, but I can say with certainty that these parts are normally not harmed in any way when an amplifier channel fails.
Removing the outputs is a pain in the butt on these amps. I normally bite the bullet and do this work on a channel failure type repair. Not fun.
Now as far as installing incorrect parts, or even answering the question of what was done for sure is something that writing out that list I recommended would have helped with. Remember, observations only - no conclusions allowed at all. Just test results and observations (just the facts 'mam).
As for working on your own amplifier, label each variable resistor as to what channel it came from. Don't adjust it at all, but you can measure it. Reinstall it once you are done building and have confirmed the control is good. Your bias current setting should be close to correct now. Remember that the Adcom bias procedure involves getting the output stage hot, then allowing it to settle into it's rest state from there. The bias reading will not be correct until this is done, so fight off the urge to measure the bias and correct it right away. This should save you time in setting it back up.
Best, Chris
That sounds like a good plan. There's no point in testing R16, it's most definitely toast.
Well, I have tested every part that I've removed/replaced, and written down the results of all those tests. But I have replaced old, tested parts with new tested, ones, even if they tested good. I did this because:
* I lack confidence in my ability to test components completely
* The old parts are really, really old
* I am, probably excessively, concerned about the impact of desoldering/re-soldering on old parts.
* It seemed advisable to replace components in the area of known failures.
I see now that was the wrong approach. What I need to learn is how to read a schematic better, so I can know what can be tested in-circuit, and how to take things out of circuit without actually removing every component.
So yeah, it's a learning project, and that includes learning what I need to learn. That said, I won't be sad to see this particular learning experience come to a swift and successful conclusion!
Now that I'm more experienced, I'll be more careful about testing things before rebuilding.
That said, Chris is right, pulling the outputs is a pain, though I can do it. So my preference would be to test them in place. Instructions for doing so, much appreciated.
Well, I did replace D5 and D6 with tested parts, so maybe I did damage in installation. I have checked all my work and there are no errors of placement or value.
Initially, bias was at the factory setting (I must have missed your advice to leave it turned down all the time, rats.) So, I'll remove Q9 and 10, test them and leave them out, and test nearby resistors. I'll test P1 again.
Well, good timing here. I was about to write that the sort of shot gun approach is good when it works but otherwise it is better to learn how to debug the amp, which is the plan that we should move to now. I'm going to suggest an approach that should avoid any further burnt parts that I used probably for the first 20 years of my debugging without a Variac. Take 2 blown fuses and solder 1K resistors straight along parallel with them, just tack solder it on. These will go into the open frame fuse holder with the resistor facing out - I think they will fit. These will limit the current to the amp and are large enough to protect the front end. If you have no dead fuses you could wrap the leads around the fuse holder ends I suppose. This should even work with the output stage in place as long as the bias control is all the way down - as long as there is no leakage. The front end draws less than 10 mA from the supplies and this will drop 10V across the resistors which is fine for testing - you will probably see less than this. I'd set up another pair of blown fuses with 100 ohm resistors across them that will protect the output stage when you are ready to bring up the bias. Keep the amp on the Variac for now, since we want to be extra careful from here on out. The 1K resistors will burn up if there is something drastically wrong with the amp and you go to full voltage right away.
We might as well test what is left of the front end first so that more doesn't burn out as you test the output stage. The 1K current limiters are good also in case you slip with a jumper lead next when you go to test the outputs.
We need:
1. the drivers Q11 and Q12 pulled out
2. T2 thermister with at least one lead open
3. All the output device emitter resistors with one lead open
4. All output device collector screws out
5. Confirm that Q1, Q2, Q4, Q9, and Q10 are out
6. P1 set to minimum resistance until further notice
With these conditions the outputs cannot draw current, the drivers are out and cannot draw current. The VAS is off because there is no diff amp transistor to turn it on. The VAS clamp is off for the same reason. The clip indicator is off because the other diff amp transistor is out. All that is left is the DC current source Q6 used to bias D5 and D6 and in turn Q3's base circuit. This is what we want to test.
After doing all of the above, confirm that you can bring up the Variac, take it to +/- 50V on the rails and measure the voltage on the 1K current limiters in the fuse holders. If that is reasonable (<10V) we will next probe around Q6 and the diode strings.
No current should flow in the output stage with this setup and we can minimally test the input board.
We might as well test what is left of the front end first so that more doesn't burn out as you test the output stage. The 1K current limiters are good also in case you slip with a jumper lead next when you go to test the outputs.
We need:
1. the drivers Q11 and Q12 pulled out
2. T2 thermister with at least one lead open
3. All the output device emitter resistors with one lead open
4. All output device collector screws out
5. Confirm that Q1, Q2, Q4, Q9, and Q10 are out
6. P1 set to minimum resistance until further notice
With these conditions the outputs cannot draw current, the drivers are out and cannot draw current. The VAS is off because there is no diff amp transistor to turn it on. The VAS clamp is off for the same reason. The clip indicator is off because the other diff amp transistor is out. All that is left is the DC current source Q6 used to bias D5 and D6 and in turn Q3's base circuit. This is what we want to test.
After doing all of the above, confirm that you can bring up the Variac, take it to +/- 50V on the rails and measure the voltage on the 1K current limiters in the fuse holders. If that is reasonable (<10V) we will next probe around Q6 and the diode strings.
No current should flow in the output stage with this setup and we can minimally test the input board.
What I do is remove all the output transistors and wire the VAS output back into the LTP. This has never in my experience blown a fuse.
It allows testing of the LTP. VAS and bias without risking output transistors.
Only when convinced that the driver stage is working and the bias is set to zero do I even consider connecting in the output transistors and even then I just risk one pair.
He has already smoked parts on the driver board. The VAS and one output are the only failed parts from the in use failure. It is always best to play it safe unless you like replacing parts! I do mostly agree with what you're saying.
I said much earlier to take the voltage measurements referenced to ground. This is where I usually start to get the big picture. Most of the voltages of interest are small so here we will measure first with the positive rail (B+) as the reference so connect the red meter lead there.
Measure:
1. Q6 B to B+
2. Q6 C to B+
3. Q6 E to B+
Now put the Black meter lead on the B- supply and measure:
4. Q3 B to B-
5. if Q3 B to B- is less than 1.5 V, connect Q3 C to ground, and measure Q3 E to B-
6. What is the drop across the B+ 1K resistor
7. And across the B- 1K resistor
I will explain how to calculate the bias currents based on these readings next.
Measure:
1. Q6 B to B+
2. Q6 C to B+
3. Q6 E to B+
Now put the Black meter lead on the B- supply and measure:
4. Q3 B to B-
5. if Q3 B to B- is less than 1.5 V, connect Q3 C to ground, and measure Q3 E to B-
6. What is the drop across the B+ 1K resistor
7. And across the B- 1K resistor
I will explain how to calculate the bias currents based on these readings next.
Last edited:
From Post #449
Craig,
Just wondering if your fakes have the exact same properties (1-4 front and 1-2 rear) as the description on the ESP Site. THe most obvious way to tell a fake appears to be that the lot numbers are not centered, they are offset, did the fake Sankens have the lot numbers offset or centered?
(sorry to continue to be off topic...)
Craig,
Just wondering if your fakes have the exact same properties (1-4 front and 1-2 rear) as the description on the ESP Site. THe most obvious way to tell a fake appears to be that the lot numbers are not centered, they are offset, did the fake Sankens have the lot numbers offset or centered?
(sorry to continue to be off topic...)
Last edited: