Hi,
I've built a Slone fig 11.4 amp. I built two channels which are almost identical. I stored them in a drawer to finish the protection PCB before I wired everything up. When I stored them in my drawer I made sure they both worked perfect.
When I finished the protection PCB I wanted to wire everything together. So I hooked up the first channel and it worked flawlessly. Then I hooked up the second amp and it gave a slowly rising DC offset and some smoke. Back at the test-bench it appeared to have started oscillating at +- 80kHz. I still works, as in, it amplifies everything at it's input, but with a nice 80kHz signal riding on top of it. The smoke probably came from the resistor damping the output inductor. There was no speakerload connected, so there was not much place to dump enough power to get something smoking (with transistors still operational). It also visually appeared to come from there, but the time it smoked was very short.
Now I wonder, why did one channel start oscillating, while the other is fine (I double checked that)? I attached a schematic with in green the peak-peak voltage of the oscillation referenced to ground. The input was shorted during this measurement and only my scope connected to the output. It seems to be somewhere in the output stage, since the input and VA stage are virtually free of oscillation. Only between C9 and C10 we see some of the oscillation, but that seems to come from the collector of Q6. It's also slightly shifted in phase compared to the other measurements.
Does anybody have some clue? I'm really desperate on this one.
Thanks in advance,
Remco Poelstra
I've built a Slone fig 11.4 amp. I built two channels which are almost identical. I stored them in a drawer to finish the protection PCB before I wired everything up. When I stored them in my drawer I made sure they both worked perfect.
When I finished the protection PCB I wanted to wire everything together. So I hooked up the first channel and it worked flawlessly. Then I hooked up the second amp and it gave a slowly rising DC offset and some smoke. Back at the test-bench it appeared to have started oscillating at +- 80kHz. I still works, as in, it amplifies everything at it's input, but with a nice 80kHz signal riding on top of it. The smoke probably came from the resistor damping the output inductor. There was no speakerload connected, so there was not much place to dump enough power to get something smoking (with transistors still operational). It also visually appeared to come from there, but the time it smoked was very short.
Now I wonder, why did one channel start oscillating, while the other is fine (I double checked that)? I attached a schematic with in green the peak-peak voltage of the oscillation referenced to ground. The input was shorted during this measurement and only my scope connected to the output. It seems to be somewhere in the output stage, since the input and VA stage are virtually free of oscillation. Only between C9 and C10 we see some of the oscillation, but that seems to come from the collector of Q6. It's also slightly shifted in phase compared to the other measurements.
Does anybody have some clue? I'm really desperate on this one.
Thanks in advance,
Remco Poelstra
Hi Thoru,
There must be something wrong with your schematic: as it is shown, the VAS is rather an attenuator. Q17 will try to keep the current through Q6 constant, and the signal will only cause minor perturbations.
Now, there is probably enough loop gain elsewhere to make it appear to work, but it's certainly not the way it is supposed to function.
I suggest you double check your schematic and your physical prototype to determine the actual configuration.
When you come back with the information, we will be able to make meaningful comments.
LV
There must be something wrong with your schematic: as it is shown, the VAS is rather an attenuator. Q17 will try to keep the current through Q6 constant, and the signal will only cause minor perturbations.
Now, there is probably enough loop gain elsewhere to make it appear to work, but it's certainly not the way it is supposed to function.
I suggest you double check your schematic and your physical prototype to determine the actual configuration.
When you come back with the information, we will be able to make meaningful comments.
LV
Hi,
Thanks for your reply.
That's very strange. Q17 shouldn't be conducting at all, it's there to protect the VAS stage for overcurrent. Q8 sends 7mA thru the collector of Q6. With 47Ohm, that gives 0.3V Vbe for Q17, so it should be off. I'll check whether it's conducting or not.
Regards,
Remco Poelstra
Thanks for your reply.
That's very strange. Q17 shouldn't be conducting at all, it's there to protect the VAS stage for overcurrent. Q8 sends 7mA thru the collector of Q6. With 47Ohm, that gives 0.3V Vbe for Q17, so it should be off. I'll check whether it's conducting or not.
Regards,
Remco Poelstra
Looking at the measurements in your schematic I wouldn't say that the VAS is free from oscillation. Also, the VAS is current driven so you wouldn't see much of a voltage swing on the collector of Q1 anyway (hence the 0 mVpp). I'm pretty certain your amplifier as a whole is oscillating and not just the output stage. The oscillation frequency seems to support this. For an oscillating output stage (as for example in a complimentary feedback design) one would expect to see much higher frequencies (tens or hundreds of MHz).
Maybe first check for bad solder connections around the compensation capacitors C9 and C10 and R13. And also around R11 and C6. If all seems ok then maybe the design just is unstable. But in that case one would expect the other channel to oscillate as well. Is it truly identical? PCB layout may play a significant role here. Also any capacitive load at the output before L1 could easily trigger oscillations (as e.g. in a coax test lead attached to the junction of R33 and R34).
Try to stabilize it by experimenting with the value of C9 (increase it). You can simplify the compensation scheme by removing C10 and R13 and placing C9 directly from the base of Q5 to the collector of Q6. Once you've found a value that works you can refine the compensation scheme by again adding C10 and R13 and tweaking it further. Don't be satisfied when the amplifier stops oscillating when there is no input signal, test it with a low level square wave input and check the output for ringing.
PS: looking at figure 11.4 in Slone's book I see a complimentary feedback pair while yours is a darlington configuration! Also the transistor types differ, which could also play a role.
Maybe first check for bad solder connections around the compensation capacitors C9 and C10 and R13. And also around R11 and C6. If all seems ok then maybe the design just is unstable. But in that case one would expect the other channel to oscillate as well. Is it truly identical? PCB layout may play a significant role here. Also any capacitive load at the output before L1 could easily trigger oscillations (as e.g. in a coax test lead attached to the junction of R33 and R34).
Try to stabilize it by experimenting with the value of C9 (increase it). You can simplify the compensation scheme by removing C10 and R13 and placing C9 directly from the base of Q5 to the collector of Q6. Once you've found a value that works you can refine the compensation scheme by again adding C10 and R13 and tweaking it further. Don't be satisfied when the amplifier stops oscillating when there is no input signal, test it with a low level square wave input and check the output for ringing.
PS: looking at figure 11.4 in Slone's book I see a complimentary feedback pair while yours is a darlington configuration! Also the transistor types differ, which could also play a role.
Thanks for your reply.
Yes, I modified the original design of fig 11.4. The EF stage is more stable and has a more constant output impedance in class B mode, so that's why I chose that one.
I'll try to get it stable with different capacitor values. The problem is that dismantling the PCB is a lot of work, since I sandwiched the output transtors between the solderside of the PCB and the heatsink. But well, it seems I can't avoid taking it all apart. I'm only afraid of turning the amplifier on without a heatsink. I turned the potentiometer in such a position that Vbias is minimal, but will it than still have thermal runaway without heatsink? It normally runs cold in this situation.
Will changing C6 help? Ot is it better left at 5pF?
Both PCB's are made from the same design. The oscillating one from a single-sided PCB and the not-oscillating one from a double sided PCB. I didn't remove all coper from the top side since that consumed to much etching liquid. But since the now-oscillating amp was once stable I supposed it wasn't that important.
I will let know how it all ends, it can take a while, due to the mechanical construction.
Regards,
Remco Poelstra
Yes, I modified the original design of fig 11.4. The EF stage is more stable and has a more constant output impedance in class B mode, so that's why I chose that one.
I'll try to get it stable with different capacitor values. The problem is that dismantling the PCB is a lot of work, since I sandwiched the output transtors between the solderside of the PCB and the heatsink. But well, it seems I can't avoid taking it all apart. I'm only afraid of turning the amplifier on without a heatsink. I turned the potentiometer in such a position that Vbias is minimal, but will it than still have thermal runaway without heatsink? It normally runs cold in this situation.
Will changing C6 help? Ot is it better left at 5pF?
Both PCB's are made from the same design. The oscillating one from a single-sided PCB and the not-oscillating one from a double sided PCB. I didn't remove all coper from the top side since that consumed to much etching liquid. But since the now-oscillating amp was once stable I supposed it wasn't that important.
I will let know how it all ends, it can take a while, due to the mechanical construction.
Regards,
Remco Poelstra
I don't think that changing C6 is going to solve your oscillation problem. It's purpose is to limit the bandwith of the amplifier. With the values shown it only affects much higher frequencies (> 1MHz).
C9/C10 and R13 form the amplifier's compensation scheme so those are the ones you should play with.
Best to leave the transitors attached to the heatsink as they can get hot when the amplifier breaks out into full blown oscillation. When the frequency of oscillation is high enough it can happen that the output transistors can't turn off in time. When that happens they both conduct for a moment during every cycle, resulting in a very large current and some VERY hot transistors. The amplifier can die in a hurry in such a case. Now, I don't think you need to worry about that at 80KHz, but still if you're playing around with the compensation caps it's wise to have the heatsinks in place.
Maybe you can rig somethig up with some (short) pieces of wire in place of C9 so you can try different values from the component side?
C9/C10 and R13 form the amplifier's compensation scheme so those are the ones you should play with.
Best to leave the transitors attached to the heatsink as they can get hot when the amplifier breaks out into full blown oscillation. When the frequency of oscillation is high enough it can happen that the output transistors can't turn off in time. When that happens they both conduct for a moment during every cycle, resulting in a very large current and some VERY hot transistors. The amplifier can die in a hurry in such a case. Now, I don't think you need to worry about that at 80KHz, but still if you're playing around with the compensation caps it's wise to have the heatsinks in place.
Maybe you can rig somethig up with some (short) pieces of wire in place of C9 so you can try different values from the component side?
Try changing to "regular" Miller compensation - remove c9, r13 and put C10 around 100pF.
At a guess, in the double sided one that is not oscillating, the blank layer is acting as a shield which prevents stray RF causing the oscillation. Is this surface connected to ground? Is there clearance around all the through-holes?
At a guess, in the double sided one that is not oscillating, the blank layer is acting as a shield which prevents stray RF causing the oscillation. Is this surface connected to ground? Is there clearance around all the through-holes?
Yes, the top layer is connected to GND and AGND. Corresponding to which components are on top of it. I divided the layer to reflect the two GND's.
I will see what I can do with headers to change the components.
So, first increase R13, then try single Miller compensation and if that works, try to get back to two pole compensation. Well, I don't think my evening will be long enough to try all that in one run
.
Regards,
Remco Poelstra
I will see what I can do with headers to change the components.
So, first increase R13, then try single Miller compensation and if that works, try to get back to two pole compensation. Well, I don't think my evening will be long enough to try all that in one run
.Regards,
Remco Poelstra
It's probable that all the fancy compensation was to make the CFP output stage work stable. I hate them - they are so unpredictable!
I don't get why there is a resistor in the base of Q4. Perhaps it is the current source that is oscillating? I have seen it happen with the two-transistor type.
I'm probably being quite harsh but I don't think much of this design of Slone's.
I don't get why there is a resistor in the base of Q4. Perhaps it is the current source that is oscillating? I have seen it happen with the two-transistor type.
I'm probably being quite harsh but I don't think much of this design of Slone's.
The compensation used by Slone is called "two pole compensation". It's a technique that gives you somewhat better distortion performance for the higher audio frequencies because it doesn't reduce the open loop gain as quickly as the regular miller compensation does. As such it doesn't have anything to do with the output stage configuration or stability thereof.
I've been wondering about the use of those weird base resistors as well. You see them in many places in the designs by Doug. Self and Slone (e.g. Q3, Q4, Q8, Q17). At some point in Self's book he explains that he adds R7/R15 to aid in troubleshooting the amplifier as without them it can sometimes be hard to diagnose which transistor has failed.
E.g. imagine what would happen to the circuit if the collector current trough Q3 would be interrupted for some reason. Without R7 the collector of Q4 would be pinned to about 0.7V. As a result Q8 would stop conducting as well. So you're left wondering if the fault is located in Q3's collector circuit or in Q8's collector circuit (because if Q8's collector current is interrupted for some reason you would see essentially the same symptoms).
Similarly imagine what would happen if Q8 fails with an emmitor-collector short circuit. Without R14 there would be nothing to limit the base current of Q4 and it would perish, possibly with consequences for Q3 as well. That leaves you with a bunch of broken transistors to figure out instead of just one. I think that basically is the reason for R14.
Anyway, that's my take on it. I have been known to be wrong though
I've been wondering about the use of those weird base resistors as well. You see them in many places in the designs by Doug. Self and Slone (e.g. Q3, Q4, Q8, Q17). At some point in Self's book he explains that he adds R7/R15 to aid in troubleshooting the amplifier as without them it can sometimes be hard to diagnose which transistor has failed.
E.g. imagine what would happen to the circuit if the collector current trough Q3 would be interrupted for some reason. Without R7 the collector of Q4 would be pinned to about 0.7V. As a result Q8 would stop conducting as well. So you're left wondering if the fault is located in Q3's collector circuit or in Q8's collector circuit (because if Q8's collector current is interrupted for some reason you would see essentially the same symptoms).
Similarly imagine what would happen if Q8 fails with an emmitor-collector short circuit. Without R14 there would be nothing to limit the base current of Q4 and it would perish, possibly with consequences for Q3 as well. That leaves you with a bunch of broken transistors to figure out instead of just one. I think that basically is the reason for R14.
Anyway, that's my take on it. I have been known to be wrong though
I think I've some sort of problem, but I don't know what it is.
I removed R13 and C10. It was just easiest in construction to get these out first. With my lab supply set to 5V, 0.04A I turned on the amp, just to see what happened. It stopped oscillating, a sinewave came out as a nice sinewave, but a triangle or rectangle waveform were not recognizable. The triangle looked most like a capacitor charging/discharging under constant current and the rectangle looked like some delta-pulses with fast recovery to 0VDC. (Frequency was 1kHz).
When I put back C10 and R13, the oscillation was back, but at lower amplitude.
Why is the amp not operating normal anymore with C10 and R13 removed (C10 shorted ofcourse)? It should work nice with the 150p from C9.
Well anyway, it seems it's indeed the whole amp that's oscillating.
Anybody any clues on why my amp is slowly dying?
Regards,
Remco Poelstra
I removed R13 and C10. It was just easiest in construction to get these out first. With my lab supply set to 5V, 0.04A I turned on the amp, just to see what happened. It stopped oscillating, a sinewave came out as a nice sinewave, but a triangle or rectangle waveform were not recognizable. The triangle looked most like a capacitor charging/discharging under constant current and the rectangle looked like some delta-pulses with fast recovery to 0VDC. (Frequency was 1kHz).
When I put back C10 and R13, the oscillation was back, but at lower amplitude.
Why is the amp not operating normal anymore with C10 and R13 removed (C10 shorted ofcourse)? It should work nice with the 150p from C9.
Well anyway, it seems it's indeed the whole amp that's oscillating.
Anybody any clues on why my amp is slowly dying?
Regards,
Remco Poelstra
Maybe first start by checking the DC voltages troughout the amplifier. The voltages acros R1 and R20 shoud be about 0.65V. The voltages acros R5 and R6 should be about 0.15V. The voltage acros R22 should be about 0.3V. The output voltage should be nicely centered at 0V and the voltages across R33/R34 should confirm the bias current you expect to see. If that all checks out then I think your transistors are all ok.
What kind of signal amplitudes are you using to test with? The voltage gain of this amplifier seems a bit high at 30x = (R10+R12)/R12. So an input signal of 100mVeff would already generate an output voltage of 3Veff. That might be pushing it with only a 5V supply. Is the amplitude of the sine wave what you expect at the output?
I'm not sure I understand your description of the signal waveforms you're seeing. A capacitor charging/discharging with constant current would give you a perfect triangle. The square you describe sounds like it has been high pass filtered (assuming you see spikes where the transitions are, followed by an exponential decay towards 0).
What kind of signal amplitudes are you using to test with? The voltage gain of this amplifier seems a bit high at 30x = (R10+R12)/R12. So an input signal of 100mVeff would already generate an output voltage of 3Veff. That might be pushing it with only a 5V supply. Is the amplitude of the sine wave what you expect at the output?
I'm not sure I understand your description of the signal waveforms you're seeing. A capacitor charging/discharging with constant current would give you a perfect triangle. The square you describe sounds like it has been high pass filtered (assuming you see spikes where the transitions are, followed by an exponential decay towards 0).
Yes, you're right about the capacitor. It's getting late I mean indeed a capacitor charged with a resistor in series. So a quick rise and than slowly to the top, the same for the discharge.
The rectangle indeed looks highpass filtered.
I'll check the biassing of the amp tomorrow, today I'm only going to make more errors.
The input voltage is really small, 0.02V if I remember correctly. Will double check that tomorrow too.
Regards,
Remco
I mean indeed a capacitor charged with a resistor in series. So a quick rise and than slowly to the top, the same for the discharge.The rectangle indeed looks highpass filtered.
I'll check the biassing of the amp tomorrow, today I'm only going to make more errors.
The input voltage is really small, 0.02V if I remember correctly. Will double check that tomorrow too.
Regards,
Remco
I replaced a resistor in the feedback loop and the amp is behaving normal now with only C9 in place. Probably it had a bad solder joint. I tried increasing R13, but that only shifts the oscillation frequency to a lower frequency. Isn't there some fine math to calculate the best R13 and C10 value? With C9 at 150p the amps seems fine, but I don't like just guessing some values for the other two components. Slone works too much with some "rule of thumb" for me on this one.
What I think is strange is that Slone suggest a 10xC9 value for C10 and then makes the value only twice as big in his final design.
Regards,
Remco Poelstra
What I think is strange is that Slone suggest a 10xC9 value for C10 and then makes the value only twice as big in his final design.
Regards,
Remco Poelstra
Tried something more after reading Self's version of two-pole compensation. Having only C9 and C10 in place (no R13), the amp is stable. If I then increase R13 to 10k the amp stays stable, but ringing appears in the rectangular waveform. As far as I could judge the frequency from my scope, the ringing was some 250kHz. Which I didn't expect since increasing R13 lowered the oscillation frequency.
So it remains a mystery why one amp oscillates and the other doesn't.
I hope you guys can give a new clue on what's going on here.
Regards,
Remco
So it remains a mystery why one amp oscillates and the other doesn't.
I hope you guys can give a new clue on what's going on here.
Regards,
Remco
The spike is relatively harmless and I remember it is mentioned somewhere in Slone's book.
Your other amp isn't identical. For starters you have a different PCB geometry (double sided versus single sided). Secondly there may be very wide tolerances in components like capacitors and certainly in many of the parasitics of the transistors themselves. If the design was "on the edge" to begin with this may easily explain things. Designing something that works is one thing, designing something that can be reliably build in large numbers is quite another thing. Your other amplifier may not oscillate, but check it for ringing.
If you still see ringing you should fiddle a bit more and see if you can eradicate it. Ringing indicates that the phase margin is still small and hence oscillation is still a possibility if circumstances change just a little bit.
Your other amp isn't identical. For starters you have a different PCB geometry (double sided versus single sided). Secondly there may be very wide tolerances in components like capacitors and certainly in many of the parasitics of the transistors themselves. If the design was "on the edge" to begin with this may easily explain things. Designing something that works is one thing, designing something that can be reliably build in large numbers is quite another thing. Your other amplifier may not oscillate, but check it for ringing.
If you still see ringing you should fiddle a bit more and see if you can eradicate it. Ringing indicates that the phase margin is still small and hence oscillation is still a possibility if circumstances change just a little bit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.