That looks the right place...
The recommended current is just 8 milliamps which would be 8 millivolts. Modern transistors have slightly different characteristics and that could put the bias setting out of range of the preset. If that has happened then you can reduce the value of R80, the 1k2 resistor between collector and emitter of the bias generator. Trial and error I'm afraid, 1k first, then 820 ohm if still out of range. Can't see it needing to be lower than that but nothing bad would happen.
On the channel that blows a fuse...
If you link collector and emitter of the bias generator as shown here then that forces zero bias. If it still blows a fuse then you have a genuine problem like a shorted output transistor etc. Make sure the outputs are correctly insulated from the heatsink i.e. no continuity from the tab to the heatsink.
If you use a bulb tester you will save blowing fuses and also save damaging output transistors in the event of a fault.
The recommended current is just 8 milliamps which would be 8 millivolts. Modern transistors have slightly different characteristics and that could put the bias setting out of range of the preset. If that has happened then you can reduce the value of R80, the 1k2 resistor between collector and emitter of the bias generator. Trial and error I'm afraid, 1k first, then 820 ohm if still out of range. Can't see it needing to be lower than that but nothing bad would happen.
On the channel that blows a fuse...
If you link collector and emitter of the bias generator as shown here then that forces zero bias. If it still blows a fuse then you have a genuine problem like a shorted output transistor etc. Make sure the outputs are correctly insulated from the heatsink i.e. no continuity from the tab to the heatsink.
If you use a bulb tester you will save blowing fuses and also save damaging output transistors in the event of a fault.
Well, it is good news that you have managed to get one channel working.
Since the bias transistor should be providing something like 1.8V then it is hard to see that the ratio for the 1.2k resistor and 1k bias pot would be far out, as we would expect the pot to be set at around 600 ohms. One of the design quirks this circuit has is the absence of a resistor in series with the adjusting pot, which would normally limit the maximum curent which can be set. Something in the region of 470 ohms which would give a nominal voltage just over three Vbe's would be my suggestion.
The absence of this limiting resistor suggests that the potentiometer could have been set to the wrong end (zero) and caused excessive current in the transistors. Excessive currents can partially damage transistors if they aren't blown open or short. This may mean that they leak. One candidate is of course the PNP driver, the others being the NPN driver and output pair. I would start with the bias regulator transistor and see if there are shorts around the base resistor and pot, or if the transistor still works.
While I am reluctant to suggest altering transistors (given your problems) perhaps you may want to remove the transistors and check them all individually. Do you have any means to test them at voltages up to their maximum, in at least a pulsed mode such as a curve tracer (not recommended for measuring Vceo these days but at least testing up to (but not exceeding) the rated value while checking the allowed maximum current). Or know someone who could?
Another problem you might want to check for is any leakage or shorts on the PCB tracks. While the transistors are out of the board do you measure any adverse leakage currents (resistances) anywhere whcih should not be?
Current spikes could have fused a PCB track which in some cases could evaporate the copper and spill over onto other tracks, but I suspect that would be visible. "Black copper" isn't always, though.
Once again, before this deeper investigation, I recommend Mooly's suggestion of a dim bulb. I suggested measuring the voltages on the board and I think this is the place to start if you can power it up with a lower than normal voltage. An alternative to a tungsten lamp in series with the input would be a variable current-limited power supply. Do you have one or know someone who could lend you one? Failing that a power resistor (on a heatsink) which could drop several volts in the event of high current - perhaps 100 ohms 25W - would only drop 1V at normal quiescent currents (10mA or so) but absorb most of the voltage in the event of a short somewhere.
Turning up the voltage with a current limted power supply would allow you to check some things without blowing them up.
Another check I would suggest you make with a power-limited supply is for signs of oscillation. Given the compensation scheme used, despite my simulation results, the circuit should be stable so I doubt that it is oscillating, but that needs to be checked.
If a capacitor has shorted that would set the voltages anywhere and -for example the feedback decoupling capacitor - cause the input PNP to reverse bias whcih would lead to incorrect voltages further down the chain.
In short, I think you need to be able to test the circuit with a power limiter of some form so that, as Mooly has siad, you can start to track down the problems. If not, then I would check all the transistors and PCB (this may be the umpteenth time for you) but obviously there is still something to be uncovered.
Since the bias transistor should be providing something like 1.8V then it is hard to see that the ratio for the 1.2k resistor and 1k bias pot would be far out, as we would expect the pot to be set at around 600 ohms. One of the design quirks this circuit has is the absence of a resistor in series with the adjusting pot, which would normally limit the maximum curent which can be set. Something in the region of 470 ohms which would give a nominal voltage just over three Vbe's would be my suggestion.
The absence of this limiting resistor suggests that the potentiometer could have been set to the wrong end (zero) and caused excessive current in the transistors. Excessive currents can partially damage transistors if they aren't blown open or short. This may mean that they leak. One candidate is of course the PNP driver, the others being the NPN driver and output pair. I would start with the bias regulator transistor and see if there are shorts around the base resistor and pot, or if the transistor still works.
While I am reluctant to suggest altering transistors (given your problems) perhaps you may want to remove the transistors and check them all individually. Do you have any means to test them at voltages up to their maximum, in at least a pulsed mode such as a curve tracer (not recommended for measuring Vceo these days but at least testing up to (but not exceeding) the rated value while checking the allowed maximum current). Or know someone who could?
Another problem you might want to check for is any leakage or shorts on the PCB tracks. While the transistors are out of the board do you measure any adverse leakage currents (resistances) anywhere whcih should not be?
Current spikes could have fused a PCB track which in some cases could evaporate the copper and spill over onto other tracks, but I suspect that would be visible. "Black copper" isn't always, though.
Once again, before this deeper investigation, I recommend Mooly's suggestion of a dim bulb. I suggested measuring the voltages on the board and I think this is the place to start if you can power it up with a lower than normal voltage. An alternative to a tungsten lamp in series with the input would be a variable current-limited power supply. Do you have one or know someone who could lend you one? Failing that a power resistor (on a heatsink) which could drop several volts in the event of high current - perhaps 100 ohms 25W - would only drop 1V at normal quiescent currents (10mA or so) but absorb most of the voltage in the event of a short somewhere.
Turning up the voltage with a current limted power supply would allow you to check some things without blowing them up.
Another check I would suggest you make with a power-limited supply is for signs of oscillation. Given the compensation scheme used, despite my simulation results, the circuit should be stable so I doubt that it is oscillating, but that needs to be checked.
If a capacitor has shorted that would set the voltages anywhere and -for example the feedback decoupling capacitor - cause the input PNP to reverse bias whcih would lead to incorrect voltages further down the chain.
In short, I think you need to be able to test the circuit with a power limiter of some form so that, as Mooly has siad, you can start to track down the problems. If not, then I would check all the transistors and PCB (this may be the umpteenth time for you) but obviously there is still something to be uncovered.
Last edited:
And another thought ...
one explanation for the apparent quiescent current would be a leakage path from the centre rail to ground.
You might check the output capacitor and output path. That would have the effect of ignoring the bias pot, turning the upper output section into a sort-of Class A.
It could be this as much as the PNP driver.
one explanation for the apparent quiescent current would be a leakage path from the centre rail to ground.
You might check the output capacitor and output path. That would have the effect of ignoring the bias pot, turning the upper output section into a sort-of Class A.
It could be this as much as the PNP driver.
...when the amplifier has been unpowered for several minutes, what is the resistance between the centre output rail and ground?
Should be fairly high (positive to centre rail from your meter). If there is a resistance through the output capacitor (with a speaker or dummy load connected) it might show a few mA.
That is not an exhaustive test but may be a quick check.
Sometimes transistors which have been subjected to stress leak at higher voltages, so it could still be transistor problems.
Should be fairly high (positive to centre rail from your meter). If there is a resistance through the output capacitor (with a speaker or dummy load connected) it might show a few mA.
That is not an exhaustive test but may be a quick check.
Sometimes transistors which have been subjected to stress leak at higher voltages, so it could still be transistor problems.
Left channel - centre output to ground varies constantly, but between around 80k and 160k. Across output cap, varies, but again high 60k - 80k
Right channel - centre output to ground around 330 ohm. Across output cap also around 330 ohm.
Time for cap change?
Right channel - centre output to ground around 330 ohm. Across output cap also around 330 ohm.
Time for cap change?
What DC voltage do you have on the junction of the 0.5 ohm resistors which are the main output node (so before the speaker coupling cap)?
What supply voltage do measure on TR8 collector?
What supply voltage do measure on TR8 collector?
That would be the last thing to do imo at this point.
With no speaker connected, I measure a constantly fluctuating -20 to +20 mV on the emitter resistors (both channels) and a constantly fluctuating -16 to + 16 mV on TR8 collector (both channels)
But I have just checked the left channel fuse, and it is blown now! I'll see if I can rig up a dim bulb...
From center output through R90-R91, through (opened up) TR5, through R86 (330Ω) adds up to some 330Ω.
Why/how is TR5 opened up? If at all. C87???
Values should be the same at both channels obviously.
Not an incandescent lamp in the house - we're all LED, and I took a box of unused old incandescents to be recycled about a month ago!
I took out TR5 (right channel) and it's shorted between all terminals. I've replaced it with a good one.
Now I measure very high resistance on both channels from centre output to ground and across the output caps.
C87 is still omitted.
I took out TR5 (right channel) and it's shorted between all terminals. I've replaced it with a good one.
Now I measure very high resistance on both channels from centre output to ground and across the output caps.
C87 is still omitted.
Look at the circuit...
The collector of TR8 goes to the positive supply and so you should see approx 39 volts DC here. If there really is only millivolts then make sure the fuse is good. The emitter resistors should bias up to around half supply voltage (so about 20 volts DC). All voltage here are measured with respect to ground.
The 330 ohms is what you would see with a short circuited Tr5. As I suggested in earlier posts, the PNP could have been shorted, so no, it does not appear to be the capacitor. You would have to disconnect the capacitor from the output to check whether that actually leaked. Normally it would show an increasing resistance as it charged up.
It now seems that you have found a shorted PNP.
If the resistances are now the same (high) on both channels you may be getting towards a working amp.
The left hand channel may have fused if the right hand caused a momentary short just before its fuse blew.
If you can't get a bulb try a resistor. I suspect (hope) the left hand channel is still functional (but I would leave the fuse out even if you confirm it works until the right hand is sorted).
It now seems that you have found a shorted PNP.
If the resistances are now the same (high) on both channels you may be getting towards a working amp.
The left hand channel may have fused if the right hand caused a momentary short just before its fuse blew.
If you can't get a bulb try a resistor. I suspect (hope) the left hand channel is still functional (but I would leave the fuse out even if you confirm it works until the right hand is sorted).
Last edited:
I took the previous readings before I realised that the fuse was blown. Now with a fuse (and TR5 replaced) I get around 40 volts at collector of TR8. Both channels working, although only one at a time as I only have one fuse. I have a feeling the right channel is at a slightly lower level than the left - maybe that will change when I get another BC237B for TR3 instead of the 2SC734. I had the left channel playing for an hour with no problem, and only slightly warm, so the right is playing now. I'll leave it for an hour and then I have to go out anyway. This is a real voyage of discovery - thank you!
There may be slight gain differences between channels. However the open loop gain is fairly high and while the feedback is not particularly high either (giving an overall gain of about 50) the resistors may not be matched, (R77- R73) and the values may have degraded with the stresses which have been imposed. If they are 5% tolerance, there may be a 10% change between the channels (worst case might even be 20% with R77 5% up on one, down on the other; and R73 the opposite) -I didn't check the whole circuit for a balance control but that should be able to offset this. These changes are relatively small but may just be detectable towards the extremes.
You might want to build a simple sinewave generator (a low distortion one is expensive and complicated; a simple one using an op-amp and diode-resistor shunted resistor to control the amplitude is pretty crude but can be sufficient for a gain test) and feed that into both channels, and then measure the outputs from each side, preferably into 8 ohm dummy loads. You will likely need a volume control on the output of the oscillator as the inputs probably only need tens of mV while typical oscillators may give 1V rms or so output. You would need a DMM to measure the input voltage - an old low impedance analogue meter might not have a high enough impedance and load a simple oscillator too much - but it is not essential to measure the oscillator, only the output fromt he amplifier channels.
A quick sim of the OLG suggests around 8000. Changing transistors should not affect the overall gain much (less I suspect than resistor variations) but I haven't checked this).
If you find the channels have different gains, you may want to use 1% resistors which are pretty standard these days for R77/R73.
@ Mooly - I have to say the design discussed here has some small differences from the Mullard circuit (which uses a similar compensation scheme) but improve the performance significantly.
You might want to build a simple sinewave generator (a low distortion one is expensive and complicated; a simple one using an op-amp and diode-resistor shunted resistor to control the amplitude is pretty crude but can be sufficient for a gain test) and feed that into both channels, and then measure the outputs from each side, preferably into 8 ohm dummy loads. You will likely need a volume control on the output of the oscillator as the inputs probably only need tens of mV while typical oscillators may give 1V rms or so output. You would need a DMM to measure the input voltage - an old low impedance analogue meter might not have a high enough impedance and load a simple oscillator too much - but it is not essential to measure the oscillator, only the output fromt he amplifier channels.
A quick sim of the OLG suggests around 8000. Changing transistors should not affect the overall gain much (less I suspect than resistor variations) but I haven't checked this).
If you find the channels have different gains, you may want to use 1% resistors which are pretty standard these days for R77/R73.
@ Mooly - I have to say the design discussed here has some small differences from the Mullard circuit (which uses a similar compensation scheme) but improve the performance significantly.
A mismatched volume control is most likely for that, particularly at lower levels. The two gangs track each other slightly differently. In fact I see a mono/stereo switch on the diagram after the volume pot. Try it in mono and see if the levels seem the same or different.
I haven't any of the Mullard circuits to hand I'm afraid. I remember a couple of small but rather glossy hardback Mullard books in the local library, one was of amp circuits and the other FET's as I recall.
Yes, you are probably correct on the volume control having the greatest mis-match. I'd overlooked that.
I tried it in Mono and the left channel is still significantly louder. In Stereo even at maximum the output is pretty weak, whereas it really pumps on the left.
I measured both faders - there are two separate sliders - which are 47 K, and the left measures 46.5 K at minimum setting, while the right is at around 43 K. Both measure 0 at maximum, so surely at that setting the channels should be equal?
R73 is 68 ohms on left and 67.7 ohms on right
R77 is 3.28 K ohms on left and 3.3 K ohms on right
I should be able to get a few more transistors tomorrow to match up both channels.
I measured both faders - there are two separate sliders - which are 47 K, and the left measures 46.5 K at minimum setting, while the right is at around 43 K. Both measure 0 at maximum, so surely at that setting the channels should be equal?
R73 is 68 ohms on left and 67.7 ohms on right
R77 is 3.28 K ohms on left and 3.3 K ohms on right
I should be able to get a few more transistors tomorrow to match up both channels.
The transistors should (will 😉) make no difference to the subjective gain at all.
The gain is determined by R77 and R73 (3k3 and 68 ohm) and equals (3300/68) + 1 which is approx '50'
So 0.1 volts at the power amp input should give 5 volts at the output and if we look at the simulation we can see that is the case. The numbers agree very closely. Those two resistors alone set the voltage gain. If you have a real imbalance then I would look at the single transistor stage before the power amp (TR1) and just make sure all is OK there. Also swap the speakers around just to be sure.
The gain is determined by R77 and R73 (3k3 and 68 ohm) and equals (3300/68) + 1 which is approx '50'
So 0.1 volts at the power amp input should give 5 volts at the output and if we look at the simulation we can see that is the case. The numbers agree very closely. Those two resistors alone set the voltage gain. If you have a real imbalance then I would look at the single transistor stage before the power amp (TR1) and just make sure all is OK there. Also swap the speakers around just to be sure.