I desoldered the filter caps yesterday and checked them with multimeter in resistance mode. The same as Jac, the resistance was low at the beginning and climbed to 0.6MOhms for one cap and 1.6MOhms for the second. I measured 3.2MOhms resistance of a new cap.
I checked the capacity and the meter showed correct values.
It seems to me that the caps are OK. Would you agree?
It is all strange to me. Why the circuit was completely fine, playing music for about 3 minutes and then the short was developed resulting with the faulty 8A diodes.
I think the 3886 chip is well protected.
From the datasheet:
"SPiKe protection means that these parts are completely safeguarded at the output
against overvoltage, undervoltage, overloads, including shorts to the supplies,
thermal runaway, and instantaneous temperature peaks"
Do you think it is possible that I overheated the diodes during soldering and that resulted in the early failure?
The fuse I have used was not slo-blow. It was standard one.
I still don't have a clue what is the problem.
I think I will fit new caps and diodes and power on with the bulb tester wishing for the best.. Would it be better to remove the chip beforehand (I don't have spare one handy) ?
I checked the capacity and the meter showed correct values.
It seems to me that the caps are OK. Would you agree?
It is all strange to me. Why the circuit was completely fine, playing music for about 3 minutes and then the short was developed resulting with the faulty 8A diodes.
I think the 3886 chip is well protected.
From the datasheet:
"SPiKe protection means that these parts are completely safeguarded at the output
against overvoltage, undervoltage, overloads, including shorts to the supplies,
thermal runaway, and instantaneous temperature peaks"
Do you think it is possible that I overheated the diodes during soldering and that resulted in the early failure?
The fuse I have used was not slo-blow. It was standard one.
I still don't have a clue what is the problem.
I think I will fit new caps and diodes and power on with the bulb tester wishing for the best.. Would it be better to remove the chip beforehand (I don't have spare one handy) ?
I have custom made transformers for UK mains 240VAC +6% voltage (I live in UK).
The secondaries are 2x21V (for 4Ohm speakers) and 2x25v (for 8Ohms speakers).
Without a load I measure (If I remember correctly) 22v and 26.5v.
When the circuit failed it was connected to 8Ohms speakers and 25v secondaries.
The secondaries are 2x21V (for 4Ohm speakers) and 2x25v (for 8Ohms speakers).
Without a load I measure (If I remember correctly) 22v and 26.5v.
When the circuit failed it was connected to 8Ohms speakers and 25v secondaries.
Not really.
You are seeing 2x to 5x the leakage of a new one.
Measure the leakage current at the working voltage.
Compare with spec.
What is the ripple current spec?
Check the AC voltage on your working model and the DC rails. With 21 volt primaries, I would expect 30 volt rails.
Replace both caps. The voltage from your meter is at best 9 volts dc. It is already breaking down the cap that measures 0.6 M. The resistance may be zero at operating voltage.
Were these NOS caps?
hmm..
there are more threads to follow here.
One is the operating voltage in the system.
You say custom made 25 V. with 7-10% of transformer regulation value (typical) that is in worst case 27 -27.5 V in standby mode. Let's suppose a 10% line excess peak - that is 29.7V. The diode drop (2*.4V-- we have MUR820) brings that back to 29V.
After the filter caps that is 41V..
I'm calculating with minimum current losses, in standby mode, that is the most extreme situation for overvoltage stress.
And there is some, in effect.
Add to this a sligth oscillation in a channel, now that could become catastrophic..
Because one is getting close to the breakdown votage of the LM3886 chip.
Take into account that the given limits are for close to zero current.
If some oscillation is present then there is current flowing in the junction,
that brings down the safe threshold.
To George: he told he had his settings at the higher secondary voltage at the moment of failure.
Just guessing here... don't take me too seriously
there are more threads to follow here.
One is the operating voltage in the system.
You say custom made 25 V. with 7-10% of transformer regulation value (typical) that is in worst case 27 -27.5 V in standby mode. Let's suppose a 10% line excess peak - that is 29.7V. The diode drop (2*.4V-- we have MUR820) brings that back to 29V.
After the filter caps that is 41V..
I'm calculating with minimum current losses, in standby mode, that is the most extreme situation for overvoltage stress.
And there is some, in effect.
Add to this a sligth oscillation in a channel, now that could become catastrophic..
Because one is getting close to the breakdown votage of the LM3886 chip.
Take into account that the given limits are for close to zero current.
If some oscillation is present then there is current flowing in the junction,
that brings down the safe threshold.
To George: he told he had his settings at the higher secondary voltage at the moment of failure.
Just guessing here... don't take me too seriously
Last edited:
Cool transformer.
I don't think this would be an issue, but it might be worth checking that each of the primaries and secondaries still have appropriate resistance and are isolated from each other. Transformers are pretty tough, but there is a small possibility of damage at high current.
Regarding cap leakage testing, I have tried it yet myself, but here are a couple of links that might help.
Testing Capacitors: How To Measure Leakage Current
Tips and Techniques for Capacitor Testing | EEWeb Community
Jac
The transformer regulation value was specified as 4%.
I expected the idle voltage to be closer to 25V. 26.5v seems to be indeed high to me.
Maybe the maufacturer ignored my comment of 240V+6% voltage. Last time I measured 250V in the wall socket:-/
I will check the transformer again when back home.
I expected the idle voltage to be closer to 25V. 26.5v seems to be indeed high to me.
Maybe the maufacturer ignored my comment of 240V+6% voltage. Last time I measured 250V in the wall socket:-/
I will check the transformer again when back home.
Ok, I get it, but that would mean everything worst case. I ran a quick and dirty estimate on Duncan's PSU 2 using 25 VAC nominal and 26.5 VAC at no load. That's 6% regulation and 36 VDC at the load assuming a very high (1 Meg) load impedance at stand by.
With a 94V voltage differential max (no signal) for the LM3886, the 72V I estimated should be safe. I'm still guessing the LM3886 is safe. But I've been wrong many times before.
Which reminds me of a famous Bob Lutz saying. " Often wrong, never in doubt"
Jac
Jac, yes, me too I have / use 25V windings. And measure regularly 37V effective operating voltages which I consder safe.
I was just doing a worst case guess in presence of excess line voltage, which Greg ~~half confirmed just now..
Then don't forget the oscillation part of the hypothesis. Which has a very simple cure; ground the heatsink locally, for the tests. All this if indeed there are no errors in the compensation scheme.
I was just doing a worst case guess in presence of excess line voltage, which Greg ~~half confirmed just now..
Then don't forget the oscillation part of the hypothesis. Which has a very simple cure; ground the heatsink locally, for the tests. All this if indeed there are no errors in the compensation scheme.
Another verification: just use the tester in the circuit, if I'm right then the chip should be in short (between + and -, principally. )
A short thinking: true that normaly it (the LM3886) should blow up first.. but if there had been a shot through, and the output went short rapidly, then there is no more power dissipated in the dice.. >> then it is the turn of the diodes..
Ciao! G
A short thinking: true that normaly it (the LM3886) should blow up first.. but if there had been a shot through, and the output went short rapidly, then there is no more power dissipated in the dice.. >> then it is the turn of the diodes..
Ciao! G
I'd power up the caps with a milli-ammeter (multimeter) in series. Or even a resistor like 100ohms in series and measure the volts across it. That sounds safer for a cap I don't trust?
It's normally quite hard to blow up diodes. A cap that goes short might do it. Some self-repair and hide the fault until next time.
I replaced the filter caps and diodes over the weekend. This time I used the bulb tester when powering on. The bulb flashed for a very short moment (initial charging). No smoke, even the LED1 was lit (but only for maybe 15-20s).
I did two tests. One with 22v secondaries (-28v/28v after diodes) and one with 26.8v (-34v/34v after diodes).
I measured -9.8V on -14V rail and 0V-0.4V on 14V rail - no voltage after IC101 (LM317).
My understanding is that -14V/14V rails are for IC1 (LM318M) and the fault must be devloped there (or maybe the LM317 itself?).
My version is EvoMod which means that R39 (470K) is omitted.
What would be your suggestions? To solder new LM318M and LM317 and to install R39?
I wish I could avoid resoldering LM318M - it is so small...
I did two tests. One with 22v secondaries (-28v/28v after diodes) and one with 26.8v (-34v/34v after diodes).
I measured -9.8V on -14V rail and 0V-0.4V on 14V rail - no voltage after IC101 (LM317).
My understanding is that -14V/14V rails are for IC1 (LM318M) and the fault must be devloped there (or maybe the LM317 itself?).
My version is EvoMod which means that R39 (470K) is omitted.
What would be your suggestions? To solder new LM318M and LM317 and to install R39?
I wish I could avoid resoldering LM318M - it is so small...