12Vac would produce about 8 V on the pos rail and 5V out is about right. I think you can keep going, if you know the voltage where it clipped that is where you will see trouble if any. I suggest that you note the pos rail voltage and then the output as you bring up the Variac voltage. Get to know how a few points on the Variac dial associate with DC voltage.
Also, You could monitor the rail current by putting your meter on the amps scale in series with the rail fuse. Then bring up the voltage slowly. Or take one of your blown fuses and solder a 10 ohm resistor across it as a current sense, then read the voltage across the resistor. It will burn up if it is 1/2 W or less and something goes wrong.
Oh, you asked what does this mean. When the current source pulled with no load it could pull very close to the rail, and since no current was flowing to a load the output stage showed very little drop. Now, you have the 1K resistor pulling against the current source and it will drop about 4 volts down from the rail. So Vout should be about Vpos rail - 4.
Also note that you will be turning on the A/B bias in the output stage so it will be normal to measure a small voltage across the emitter resistors. Tell us what you read.
Last edited:
Oh, I mis-understood you earlier. You meant I should see the rail voltage minus 5 volts at the output.
I hooked up the resistor to C of Q7 again. Sure enough, as I approached the rail voltage where oscillation occurs (approaching 50v) I started to draw a lot of current (close to an amp) on the ammeter at the variac, so I shut it down.
Last edited:
OK, so something is not right. I don't think you'll do any more damage if you bring up the voltage to where it draws say .1 to .2 amp on the variac. Then measure the emitter resistor voltages. This will tell us if one or more output devices is bad. If the issue is balanced across the outputs then it is probably the drivers or bias circuit. You could just replace the drivers at this point if you want to avoid risking the outputs. I think you were planning to do it anyway since it is a suggested upgrade. Check the bias transistors also, they never see high voltage so a basic test should be good enough or use new devices if you have them.
We now know that the problem is in the output stage and only at higher voltage. Something on the negative side is turning on, it could be the driver which then turns on the outputs. Or it could be one or more of the outputs.
Edit: I was just thinking that it could be Q4 breaking down at this voltage which would cause clipping, but it would not cause high current. It is a sure thing that the negative side of the outputs is turning on to produce so much current, so not likely Q4.
Edit2: Thinking again - I take this back, if Q4 breaks down at the higher voltage it will no longer act as a 4 mA constant current source so the current in that leg will go up, causing the bias to go up and it could cause this issue. You could pull out Q4, then the resistor will pull it all the way to the rail so Vout will be about equal to the pos rail, the pos rail test should behave the same as when you did the neg rail.
My bets are on the driver since as I said earlier the drivers are stressed when a fuse is blown, could also be one or more outputs since when a driver goes often more is taken out. I would still pull Q4 to rule it out and it would be nice if you don't have to replace any outputs.
We now know that the problem is in the output stage and only at higher voltage. Something on the negative side is turning on, it could be the driver which then turns on the outputs. Or it could be one or more of the outputs.
Edit: I was just thinking that it could be Q4 breaking down at this voltage which would cause clipping, but it would not cause high current. It is a sure thing that the negative side of the outputs is turning on to produce so much current, so not likely Q4.
Edit2: Thinking again - I take this back, if Q4 breaks down at the higher voltage it will no longer act as a 4 mA constant current source so the current in that leg will go up, causing the bias to go up and it could cause this issue. You could pull out Q4, then the resistor will pull it all the way to the rail so Vout will be about equal to the pos rail, the pos rail test should behave the same as when you did the neg rail.
My bets are on the driver since as I said earlier the drivers are stressed when a fuse is blown, could also be one or more outputs since when a driver goes often more is taken out. I would still pull Q4 to rule it out and it would be nice if you don't have to replace any outputs.
Last edited:
Well this feels like progress. Thanks so much Pete and everybody. I'll measure those emitter voltages and post them soon.
Actually, suspecting the output stage sends me back to the pic I posted on post #1!
Q9 and Q10 are the bias transistors, right? They are new. Q4 is also new.
I'll order up some MJE15032/33. In the meantime, can I use the 30/31's I have for testing?
Feeling hopeful!
Actually, suspecting the output stage sends me back to the pic I posted on post #1!
Q9 and Q10 are the bias transistors, right? They are new. Q4 is also new.
I'll order up some MJE15032/33. In the meantime, can I use the 30/31's I have for testing?
Feeling hopeful!
Last edited:
Yes, that is all correct. Yes you can use those drivers but I would not take the rails over 70V just to play it safe. This will give you a chance to determine if you need to order outputs as well.
Hi Pete,
Is there any way I can have a look at what you are seeing there? Just in case there are any other small changes that might explain things. I would like to give an accurate answer.
Hey! I just realized what you are looking at!
That component is a thermistor used for temperature compensation. There is no adjustment and it's found mounted on the heat sink near one thermal switch.
You really had me going there Pete! I can understand why you might think it's an adjustment. The symbol is supposed to have a circle around it with a small "T" to indicate what it is sensitive too.
You're giving really excellent support to Fred. Thank you - just 'cause! 🙂
Hi Fred,
Most digital multi-meters are not going to give you reasonable accuracy, never mind with low capacitance. The only hand held meter I might trust would be a Fluke 87 or equiv. The new Agilent meters are probably good for that as well. The rest of the meters out there probably mislead people.
At 39 pF, a higher test frequency is normally used. So high that it's not usable for electrolytic caps. I normally test these values at 10 KHz or 100 KHz, and null the leads first!. That's using a meter designed for testing capacitors. Electrolytic capacitors are normally tested at 100 Hz, 120 Hz or 1 KHz. It depends on what you are looking for.
-Chris
Is there any way I can have a look at what you are seeing there? Just in case there are any other small changes that might explain things. I would like to give an accurate answer.
Hey! I just realized what you are looking at!
That component is a thermistor used for temperature compensation. There is no adjustment and it's found mounted on the heat sink near one thermal switch.
You really had me going there Pete! I can understand why you might think it's an adjustment. The symbol is supposed to have a circle around it with a small "T" to indicate what it is sensitive too.
You're giving really excellent support to Fred. Thank you - just 'cause! 🙂
Hi Fred,
Most digital multi-meters are not going to give you reasonable accuracy, never mind with low capacitance. The only hand held meter I might trust would be a Fluke 87 or equiv. The new Agilent meters are probably good for that as well. The rest of the meters out there probably mislead people.
At 39 pF, a higher test frequency is normally used. So high that it's not usable for electrolytic caps. I normally test these values at 10 KHz or 100 KHz, and null the leads first!. That's using a meter designed for testing capacitors. Electrolytic capacitors are normally tested at 100 Hz, 120 Hz or 1 KHz. It depends on what you are looking for.
-Chris
Ah, yes thanks Chris, I've seen that thermistor, excellent memory there! You're right about the symbol; good so only one adjustment.
Fred's doing an excellent job hear with all the advice people are giving. I'll be ready to work on mine where we're done here and I appreciate the input from techs who have a lot of experience with them.
Fred's doing an excellent job hear with all the advice people are giving. I'll be ready to work on mine where we're done here and I appreciate the input from techs who have a lot of experience with them.
Last edited:
Well, yes we would want to know the driver Imax, but we would also want to know the IV curve as related to the SOA of the device. There will be a point where the IV is worst case on the SOA curve. People have made spread sheets to calculate this for the output stage but I'm not sure if any include the drivers.
But, again consider a shorted output. And for this worst case calculation we have to assume that the user hears no output so they just crank the volume in a way that saturates the front end. If we look at the path from the rail to the C of the driver, then the E to the B of the output where the E goes to the small emitter resistors (.82R) which are then grounded by the short, there is nothing to "give", no significant resistance. We have a big (huge) cap on the rail, the B-E junctions saturate at perhaps a few volts and even with 10A (40A total output) there will be only 8.2 V drop on the emitter resistors. The drivers will just try to pump as much current as they can, stil with 80V minus about 12V or 68V Vce where they can handle very little current and that current could shoot up only limited by beta times the base drive from the VAS, which in turn is limited by the VAS protection transistor.
As a quick calculation, and now we have to use best case numbers to get the worst case current, let's say the VAS is protected at i=.7/32 or about 18 mA (after subtracting 4ma for the CCS), let's assume a driver beta of 100 so that the Ic is then 1.8 A with roughly 68V driver Vce, where the small drivers can only handle 70 ma and the replacements 400 mA without temp de-rating. If we assume a driver beta of 200 then the peak current is even worse at 3.6A where we see that both cases grossly stress the devices.
What is needed are base stoppers to the outputs, something to limit the max current, 10 ohm resistors would drop 18V with the 1.8A drive reducing the driver Vce to 50V. We could go to the SOA curve for the MJE15032 and look for the 1.8A limit which is about 40V. We then need to drop 68-40 = 28V on the base stopper, and with 1.8A then R = 28/1.8 = 15.6 ohms. What probably saves the amp in most cases with this larger driver transistor is that even though the fuses take some time to blow they will start dropping voltage as they are heating up, and the drivers will probably handle twice the current that they are rated for as design margin - on a good day! The base stoppers will reduce the max output current into low impedance loads and this is the disadvantage of using them. A compromise is to use half the value, say 8 to 10 ohms and if they are half watt they will burn up acting as fuses to also protect the amp. Under the fault condition above with 28V * 1.8A they dissipate 50W so a 2W-5W part is probably called for. These need to only handle that power until the fuse blows and it is better for them not to act as fuses since it means a repair is required.
Obviously, other forms of current limiting could be used.
Now, why did the VAS fail? I just noticed that a shorted output can explain this also. Note that if we ground the output, the VAS collector sees two base junctions, the .82R to ground, and the small 68.2R (R17) from the collector to ground with the emitter essentially at the rail. Even with the current limited by the protection transistor to 22 mA, under this shorted condition it will see the driver Vce minus one more Vbe drop and the insignificant R17 drop, roughly 60 - 65V. There is no SOA curve for this part, however simply based on power .022 * 60 = 1.32W and the part is rated for 1W in free air without temp or SOA derating, probably more like .5W after derating.
Edit: I believe that a better part here would be the Fairchild KSA1142 with excellent SOA characteristics being dissipation limited all the way out to 70V such that it will handle 100 mA at 70V, all other specs are also excellent:http://www.fairchildsemi.com/ds/KS/KSA1142.pdf
Contrast this SOA curve with the KSA1381 which seems to have a more traditional curve: http://www.fairchildsemi.com/ds/KS/KSA1381.pdf
Last edited:
Just noticed that the KSA1142 curve is missing the Ta=25C line making it look much different. Both parts are excellent showing minimal secondary breakdown. A heat sink should be used with the KSA1142 to maintain good SOA under the worst conditions. The amp would have to be checked for stability however this part is very similar to the stock part. A heatsink on the original part might be good enough.
Last edited:
That's funny, the first time I looked at the schematic I thought that was a pot too. It does bring up a question, though. On my amp, the thermistor is glued to the driver. I'm sure I can carefully get it off with a razor blade and patience, but what do I use to glue it back on? I assume all I can do to test it is check the resistance.
Re: cap testing. My meter is a B&K 2707a. I have no idea if it compares to the Fluke 87. I'll just have to go on faith that the new parts are good (I did at least check that they weren't shorted). I guess I need to add a capacitor tester to the the never-ending list of things my bench needs. I think I also need a bigger bench!
And needless to say, many, many thanks to everyone for the amazing support. This is a great community.
Since it is actually not raining today in Portland, I'm going outdoors to ride my bike, but this afternoon I'll swap in the temporary drivers and install new Q1 and Q2. Then I'll test to see if the oscillation is gone and report back.
The Vas tranny is blown? what junction?
I'm trying really hard to follow this repair thread. Do we know what parts have tested bad so far and what has been replaced with what? Does it make sense to list a quick summary at this point?
Edit> I don't like "shot gunning" with all these designers hanging around.LOL
I'm trying really hard to follow this repair thread. Do we know what parts have tested bad so far and what has been replaced with what? Does it make sense to list a quick summary at this point?
Edit> I don't like "shot gunning" with all these designers hanging around.LOL
Last edited:
Yes test the thermistor and post the result.
Not sure, it seems they used super glue? Thermal epoxy would be a nightmare for rework, as those thermistors might be difficult to find.
It seems that they were compensating for some thermal difference between the two driver polarities. I wonder if the MJE parts call for the same compensation? Most traditional designs have no thermistor there.
I wonder if this was some attempt at thermal current limiting. Going negative the base drive to the output is limited by the setting of the current source to about 4 mA, but the VAS current limits only by the protection transistor at a much higher 22 mA, and this current would turn on the positive side driver, especially under short circuit conditions. Could this be crude thermal limiting? I would expect it to have a long time constant given the thermal path. The driver needs to be de-rated as it heats up so perhaps this is the purpose. Perhaps they knew that the smaller drivers were marginal.
I wonder if this was some attempt at thermal current limiting. Going negative the base drive to the output is limited by the setting of the current source to about 4 mA, but the VAS current limits only by the protection transistor at a much higher 22 mA, and this current would turn on the positive side driver, especially under short circuit conditions. Could this be crude thermal limiting? I would expect it to have a long time constant given the thermal path. The driver needs to be de-rated as it heats up so perhaps this is the purpose. Perhaps they knew that the smaller drivers were marginal.
Hi Infinia,
A summary is always a good idea. In the software development world, when you are using an agile development model, every "scrum" is often followed by "lean" documentation that recaps everything and literally gets everyone back on the same page. If this thread doesn't qualify as a scrum I don't know what does! 😀
Look at post #94. That summarizes the initial fault and what was done to that point.
Picking up from that summary, the current situation is as follows:
1. The overheating on R22 was determined to be from oscillation on the positive side of an input sine wave as the amp comes to about 50v on the rails (this is the point at which it the circuit becomes stable and outputs a proper wave-form, as seen on the good channel). There are photos of this oscillation at posts #173 and #190.
2. To find the cause of the oscillation, I first tested the power supply. Voltage readings and 'scope traces showed it to be functioning well, with very little ripple. There are photos of the traces at post #160.
3. Trying to find the cause of the oscillation, the bias components and compensating caps were checked. Nothing tested outright bad, but some components were out of spec (see 4. below) and replaced. Electrolytic caps were replaced because the originals were 25 years old.
4. Altogether, the following parts have been tested/replaced:
Q1, Q2, Q4, Q7, Q9, Q10
R7, R8, R21
C3, C4, C5
Details and tests as follows:
a. Q9 and Q10. They both passed the diode test with my DVM and the IT-18 leakage test. Their gain was borderline (203, 179hfe) so I replaced them both with tested 2240 and 970 transistors with in-spec gains (390, 379hfe).
b. Tested R21, it was dead on. Replaced it with a tested new resistor since it was unsoldered.
c. Removed and tested C5, it was correct at 4.84uf. Replaced it with a new electrolytic (tested at 5.01uf).
d. Q4 tested good, but was replaced to match Q7. No sub was made.
e. Q7 tested bad, with an obvious B-E short. It was replaced. No sub was made.
f. C4 tested to only 10pf. It was replaced with a 39pf 300v silver mica cap. (note: cap tests on my DVM may not be very accurate.)
g. Q1 tested good, but had low gain (85hfe). It was replaced.
h. Q2 tested good, but had low gain (53hfe). It was replaced.
i. R17, R18, R19, R20 all test good to spec'd values.
j. C3 was replaced, substituting a 100v electrolytic for the original 10v. (Adcom made this change in later versions of the amp.)
5. Tests of the output with Q1 and Q2 removed (see post #205) strongly indicate the oscillation is caused by a fault in the output stage, either drivers or power. I will be replacing the drivers with a test pair to see if it is them or the power transistors.
FYI, I have the following test gear available:
I have the following test gear:
Heathkit IT-18 Transistor Checker
B+K 2707A DVM
Elenco XP-720 Power Supply
Elenco FG-500 Function Generator
Velleman K7000 Signal Tracer/Injector
Velleman K8065 Audio signal generator
Tektronix 422 oscilloscope
Variac
Lastly, do not miss post #157, which has Chris's cinematic blow-by-blow description of an amplifier failing. Two thumbs up!
Hi what do you think happened to them?
I think Chris aka anatech already explained in great detail to you on this.
Those diodes are V bias for 2 current sources ie Q3 and Q4. Similar to some classic current sources types on semi IC design. Besides D5 and D6 only see small V and currents ie no stress to specs.
Last edited:
Hi Wahab,
D5 and D6 should be fine. There is more of a risk in removing them for test, or measuring them in circuit. If either diode shorted or became leaky, the tail current would drop. This would have revealed itself in earlier measurements. If one had opened up, or a solder connection failed, the currents would have hit the roof. Fred would have been picking TO-92 case parts out of his head, and the rail fuses would have opened also.
In years of servicing these and similar amps, I have yet to come across these parts in a damaged condition. Can it happen that there is something wrong there? Sure, but there is no additional damage that would accompany a failure such as this.
Hi Pete,
I'm not sure why the thermistor exists as I have never worried about it. The diff pair will maintain the bias currents for PNP and NPN sides to be close to equal anyway, so reducing the sensitivity of one driver will lower the bias current in the entire section.
I have had to remove these and place them back into the proper position. Remove by using a straight blade angled in to the transistor. Give that an impulse type shock. If you only need to test the transistor, unsolder both components and remove as a unit. To install the thermistor in the event it was removed from the transistor, just glue it back on with super glue. You can also use silicone over the part and transistor. That Wakefield thermal bond epoxy is fabulous stuff. I can't afford it anymore because it sits too long and goes solid. That stuff would be my first choice.
-Chris
D5 and D6 should be fine. There is more of a risk in removing them for test, or measuring them in circuit. If either diode shorted or became leaky, the tail current would drop. This would have revealed itself in earlier measurements. If one had opened up, or a solder connection failed, the currents would have hit the roof. Fred would have been picking TO-92 case parts out of his head, and the rail fuses would have opened also.
In years of servicing these and similar amps, I have yet to come across these parts in a damaged condition. Can it happen that there is something wrong there? Sure, but there is no additional damage that would accompany a failure such as this.
Hi Pete,
I'm not sure why the thermistor exists as I have never worried about it. The diff pair will maintain the bias currents for PNP and NPN sides to be close to equal anyway, so reducing the sensitivity of one driver will lower the bias current in the entire section.
I have had to remove these and place them back into the proper position. Remove by using a straight blade angled in to the transistor. Give that an impulse type shock. If you only need to test the transistor, unsolder both components and remove as a unit. To install the thermistor in the event it was removed from the transistor, just glue it back on with super glue. You can also use silicone over the part and transistor. That Wakefield thermal bond epoxy is fabulous stuff. I can't afford it anymore because it sits too long and goes solid. That stuff would be my first choice.
-Chris
Infinia, i m not dumb....i perfectly know what is the purpose of these
diodes.....
If one had correctly read the thread, he would see that the current source dedicated tranny lacks about 0.6V in his emitter, thus the reference diodes
are only providing half of the voltage to the base of the said bjt.
This tell that there might be a diode that is in short circuit.
Thanks to the differential, the amp manage to keep control of the voltage.
I m amazed than no one did catch this.
Anatech, "should be" is not "to be"..
Last edited: