Well it happens... because it happened to me:
https://www.diyaudio.com/community/threads/test-your-ears-in-my-new-abx-test.314218/post-5238536
https://www.diyaudio.com/community/threads/output-relays.191449/post-2646972
The Zobels are R7-C20.
Anyway, I'm scratching my head about those voltages at the power devices' bases, which calculate to just 3.5 mA current through the drivers. The power devices themselves are zero biased, conduction begins at a bit more than 4.3 mA driver current. I also guess a faulty power transistor.
Best regards!
Anyway, I'm scratching my head about those voltages at the power devices' bases, which calculate to just 3.5 mA current through the drivers. The power devices themselves are zero biased, conduction begins at a bit more than 4.3 mA driver current. I also guess a faulty power transistor.
Best regards!
Thanks, I had overlooked the R20- C7 Zobels.
Do both channels behave identically? Some distortion is inevitable with no bias established at T101,T102. There is some class A bias delivered via R16 and R19, but it’s miniscule. Perhaps that’s why signal appears clean on scope until relay closes.
Those two transistors don’t begin to conduct until output current delivered to load rises above about 6mA. As an experiment, you could connect a load from the output to one of the 30V rails to force larger class A bias. If you observe an improvement in crossover distortion, that would buttress the suspicion that inadequate output bias is the culprit. (For example, 1.5k between output and 30V would provide about 20mA class A bias at crossover. Resistor dissipation would be about 0.6W. Make resistor value lower for a more emphathic test.) If confirmed as cause, what to do as remedy would be next challenge.
Other notion is to consider relay distortion; install short across relay contacts as experiment. Use resistor load or sacrificial speaker. I doubt the fix will be that simple.
Do both channels behave identically? Some distortion is inevitable with no bias established at T101,T102. There is some class A bias delivered via R16 and R19, but it’s miniscule. Perhaps that’s why signal appears clean on scope until relay closes.
Those two transistors don’t begin to conduct until output current delivered to load rises above about 6mA. As an experiment, you could connect a load from the output to one of the 30V rails to force larger class A bias. If you observe an improvement in crossover distortion, that would buttress the suspicion that inadequate output bias is the culprit. (For example, 1.5k between output and 30V would provide about 20mA class A bias at crossover. Resistor dissipation would be about 0.6W. Make resistor value lower for a more emphathic test.) If confirmed as cause, what to do as remedy would be next challenge.
Other notion is to consider relay distortion; install short across relay contacts as experiment. Use resistor load or sacrificial speaker. I doubt the fix will be that simple.
Anyway, I'd replace both R8's by Vbe multipliers with their transistors thermally connected to the power device heatsink(s) and adjust an idle current of about 20 mA (= 10 mV across either R17 or R18 or 20 mV across both) through the power transistors. I don't see any reason for having omitted them.
Best regards!
Best regards!
I think I've misread this as meaning 'before the output relay' rather than before it engages.
I agree. A more conventional Vbe multipliier would have better regulation and probably easier implementation.
That would be as expected with the outputs not properly biased. With the amp not delivering any current, you are actually seeing the output of the drivers, which are biased on. Go ahead and add a 82 ohm resistor in series with R8, making the total resistance 412 ohms. .. or add a forward biased diode (1N4148) in series with R8. Should make a significant improvement. Assuming it helps, we'll work at improving that bias network.
Hi Ylli,
I agree with your suggestion, but my caution would be that the added diode and T5 may not well coupled thermally to the heat sink. Perhaps susceptible to thermal runaway?
I agree with your suggestion, but my caution would be that the added diode and T5 may not well coupled thermally to the heat sink. Perhaps susceptible to thermal runaway?
If you want to have any chance at a stable class A bias region, you can't do much worse than the present configuration. A Vbe multiplier is clearly indicated. If you are worried about thermal tracking, set the bias to just barely enter class A. I personally would not just add a diode to the present biasing scheme except possibly to do an experiment.
Well, Mooly, not unexpectedly your sim tells us that the amp should work as is 
, despite all our objections. Anyway, we're trying to locate the issue's origin, hence have to look in possible design flaws whose results may lead to the symptoms given in the OP if any of the components fail.
Although the component tester showed no abnormalities, the component(s) might show completely another behaviour in real life. To be honest, I'm still meandering... 🙄
Best regards!
, despite all our objections. Anyway, we're trying to locate the issue's origin, hence have to look in possible design flaws whose results may lead to the symptoms given in the OP if any of the components fail.Although the component tester showed no abnormalities, the component(s) might show completely another behaviour in real life. To be honest, I'm still meandering... 🙄
Best regards!
Alright, thank you all very much. I will try your suggestions and let you know. Once again, Thank you very much.
It is worth @Tom Vyborny just looking at the simulation and what we see on the base of T3.
This shows the signal at this point is distorted because of the lack of output stage bias current. The output is still good though:
And the output into 8 ohm.
Now lets link the 330 ohm out and force zero bias.
T3 base:
And the output:
There is the smallest hint of visible crossover distortion now but the amp would still sound 'reasonable'.
This shows the signal at this point is distorted because of the lack of output stage bias current. The output is still good though:
And the output into 8 ohm.
Now lets link the 330 ohm out and force zero bias.
T3 base:
And the output:
There is the smallest hint of visible crossover distortion now but the amp would still sound 'reasonable'.
Isn't your 1st graph some faint hint (besides the glitches) to what the scope showed us in #1? So my guess is a faulty driver and/or power device. But first I'd disconnect both VI limiters as a try.
Best regards!
Best regards!
If you have got a burned relay contact, it can behave like a semi conductor. Because of any contact wearing over time, I always use two relays in parallel on amp builds. So maybe measure at a point before the relay?
If this is not the cause, why don't you have it running at a low load on both channels and check any stage of the amp with the scope, from input to output, left to right alternating. There must be a change in voltage if the wave form changes.
I had one type of amp that often developed a kind of scratching noise, but still played well. It allways was a transistor of the input stage that went part defect. Very unsual, as you expect them to work or fail, 100% OK or zero %. Anything in between is unknown. Changing this input transistor cured it. I had no advanced transistor tester at that time, so they simply went into the bin. So maybe just change it.
The NPN BC546/BC547 should be able to replace the KC237A. Very common and cheap universal NPN
If this is not the cause, why don't you have it running at a low load on both channels and check any stage of the amp with the scope, from input to output, left to right alternating. There must be a change in voltage if the wave form changes.
I had one type of amp that often developed a kind of scratching noise, but still played well. It allways was a transistor of the input stage that went part defect. Very unsual, as you expect them to work or fail, 100% OK or zero %. Anything in between is unknown. Changing this input transistor cured it. I had no advanced transistor tester at that time, so they simply went into the bin. So maybe just change it.
The NPN BC546/BC547 should be able to replace the KC237A. Very common and cheap universal NPN
Last edited:
I’m casting about for things that don’t make sense.
Was customer happy prior to deterioration, then ask for repair? Is problem present in both channels?
Schematic indicates -0.7V on base of of T2, but I am very skeptical that’s correct—- I think it should be near 0V. Would you check? What is DC output voltage before relay?
AC gain should be about 18.7 V/V, output/input. What do you observe? With 1V pp as you mentioned in OP, should be about 18.7V pp at output. Consistent with observation?
I’m trying to understand the large distortion seen in scope at base of T3 vs. what’s predicted in Mooly’s simulation. Was the pic taken with a x10 probe at T3 base? (X1 load capacitance from scope might compromise open loop gain. Big guess.)
What amplitude seen at base of T3 vs. output, real vs. sim, at like drive levels? Thanks, Tom and Mooly!
Was customer happy prior to deterioration, then ask for repair? Is problem present in both channels?
Schematic indicates -0.7V on base of of T2, but I am very skeptical that’s correct—- I think it should be near 0V. Would you check? What is DC output voltage before relay?
AC gain should be about 18.7 V/V, output/input. What do you observe? With 1V pp as you mentioned in OP, should be about 18.7V pp at output. Consistent with observation?
I’m trying to understand the large distortion seen in scope at base of T3 vs. what’s predicted in Mooly’s simulation. Was the pic taken with a x10 probe at T3 base? (X1 load capacitance from scope might compromise open loop gain. Big guess.)
What amplitude seen at base of T3 vs. output, real vs. sim, at like drive levels? Thanks, Tom and Mooly!
Yes, and that is why I posted it. It shows the signal on the base of that stage is not a sine. Also, just as the op says, if you remove the load it becomes a sine... well it does.
I don't think we have a seen a real scope shot of the actual output of the amp. That is the next most important thing. Is the output visibly distorted under load? We might be chasing a non existent fault.
This is what I was trying to get at earlier in (post #22. This is a very real possibility and when I experienced this it sounded just like the midrange driver was faulty when fed with test tones.
The sim shows around -340mv and an offset of -118mv.
Hi everyone,
So I decided to ty to change all the transistors and diodes. It seems that it solved my issue. Thank you very much for your help.
So I decided to ty to change all the transistors and diodes. It seems that it solved my issue. Thank you very much for your help.