I have an F5 I built a few years ago (no P3, 24 V rails, single pair of MOSFETs on each channel, current protection circuit, legit Toshiba JFETs, 0.47 ohm source resistors). It was biased at about 1 amp. As in the original schematic, I used 4 X 100 ohm feedback resistors and had a gain (voltage) of about 6. The amp worked fine.
I decided I would like to double the gain of the amplifier. I checked the gain using a sine wave input (1 volt RMS) and 8 ohm load on speakers with my scope and had a perfect waveform out with a gain of 6 on the left channel. I didn't check gain on the right channel, but removed both boards and changed out the 100 ohm resistors for 220 ohm 3W resistors.
I confirmed that both boards had good soldering, and resistance across the set of new feedback resistors read about 58 ohms, which was about what was expected. I then reinstalled the boards, confirmed good voltage from the power supply on both sides. Neither board required any change in settings of p1 or p2, and offset remained < 2 mV, and bias was at 1 amp.
So....I checked gain on the left channel using the same conditions as before changing resistors, and the gain was slightly over 12, just as expected. Great waveform.
But then I checked gain on the right channel, and found it to be 17.4 with a perfect waveform. Triple checked. Powered down the amp and rechecked resistance across the feedback resistor network and it was fine. Took each resistor out of circuit and each measured fine. Repowered up and again had a gain of 17. I added two more 220 ohm resistors in parallel with jumper leads and gain fell to 12. I removed the two extra resistors and it rose back to 17. This was followed shortly by a sudden deterioration in waveform on the output of the right channel (sine wave in, strange-looking non-sine wave out) with no gain whatsoever. The waveform was not showing oscillation (that I could recognize). No sparks. No smoke. Source resistors not hot. MOSFETs were not hot. I turned off the amp and removed the board. Source resistors look discolored but measured fine. MOSFETs checked OK out of circuit. JFETs checked OK out of circuit. Feedback resistors look fine and measure 58 ohms across all four in parallel (as does the good board). I rechecked voltage on the power supply and it is fine and smooth DC on the scope.
I'm not sure exactly what failed, but I am simply building a new board, as I have all the components and some spare boards. But I was looking for some opinions.
1.Have others encountered unexpected increased gain prior to failure of an amplifier circuit, or is this highly unusual?
2. Could the gain have always been high (say, 9), but I had never checked it before, and thus was unaware of it? If so, how could this have been if all resistor values were normal and I had legit Toshiba JFETs.
3. Would increasing gain, alone, in this amplifier with 24 volt rails be expected to cause instability and oscillation?
4. If it was oscillating prior to failure, could that have accounted for the high gain, despite what looked like an excellent wave form (and source resistors and MOSFETs that did not feel hot - measured at 94F)?
5. Any suggestions as to what happened?
If I need to make a modification, such as addition of a low value cap across feeback resistors, I'd rather do it now, but really didn't anticipate this would be a problem. Of note, I have another F5 with 32 V rails and 4 devices per channel (cascoded) with a gain of 12 (4 X 220 ohm feedback resistors) that works great without any suggestion of instability.
Thanks,
Steve
Phoenix
I decided I would like to double the gain of the amplifier. I checked the gain using a sine wave input (1 volt RMS) and 8 ohm load on speakers with my scope and had a perfect waveform out with a gain of 6 on the left channel. I didn't check gain on the right channel, but removed both boards and changed out the 100 ohm resistors for 220 ohm 3W resistors.
I confirmed that both boards had good soldering, and resistance across the set of new feedback resistors read about 58 ohms, which was about what was expected. I then reinstalled the boards, confirmed good voltage from the power supply on both sides. Neither board required any change in settings of p1 or p2, and offset remained < 2 mV, and bias was at 1 amp.
So....I checked gain on the left channel using the same conditions as before changing resistors, and the gain was slightly over 12, just as expected. Great waveform.
But then I checked gain on the right channel, and found it to be 17.4 with a perfect waveform. Triple checked. Powered down the amp and rechecked resistance across the feedback resistor network and it was fine. Took each resistor out of circuit and each measured fine. Repowered up and again had a gain of 17. I added two more 220 ohm resistors in parallel with jumper leads and gain fell to 12. I removed the two extra resistors and it rose back to 17. This was followed shortly by a sudden deterioration in waveform on the output of the right channel (sine wave in, strange-looking non-sine wave out) with no gain whatsoever. The waveform was not showing oscillation (that I could recognize). No sparks. No smoke. Source resistors not hot. MOSFETs were not hot. I turned off the amp and removed the board. Source resistors look discolored but measured fine. MOSFETs checked OK out of circuit. JFETs checked OK out of circuit. Feedback resistors look fine and measure 58 ohms across all four in parallel (as does the good board). I rechecked voltage on the power supply and it is fine and smooth DC on the scope.
I'm not sure exactly what failed, but I am simply building a new board, as I have all the components and some spare boards. But I was looking for some opinions.
1.Have others encountered unexpected increased gain prior to failure of an amplifier circuit, or is this highly unusual?
2. Could the gain have always been high (say, 9), but I had never checked it before, and thus was unaware of it? If so, how could this have been if all resistor values were normal and I had legit Toshiba JFETs.
3. Would increasing gain, alone, in this amplifier with 24 volt rails be expected to cause instability and oscillation?
4. If it was oscillating prior to failure, could that have accounted for the high gain, despite what looked like an excellent wave form (and source resistors and MOSFETs that did not feel hot - measured at 94F)?
5. Any suggestions as to what happened?
If I need to make a modification, such as addition of a low value cap across feeback resistors, I'd rather do it now, but really didn't anticipate this would be a problem. Of note, I have another F5 with 32 V rails and 4 devices per channel (cascoded) with a gain of 12 (4 X 220 ohm feedback resistors) that works great without any suggestion of instability.
Thanks,
Steve
Phoenix
Hi Zen and sangram,
I had already checked offset and bias before the failure, and while getting a gain of about 17, and they were fine. I wasn't checking offset or bias at the time of the sudden loss of function of that channel several minutes later. I'll post some waveform photos I took shortly before shutting down later as well as though when wave form looked good.
It is quickest to simply rebuild a board from scratch since I have the parts and can't find anything defective when testing out of circuit, and the board is almost done. Quicker than changing a part, back on heatsinks for testing, changing another part, back on heatsinks, etc. But if I really needed to know exactly which component failed, I would need to do what you, Zen, suggested, I guess. Every single resistor measures right on the mark out of circuit (I have disassembled the bad board), and both JFETs and MOSFETs behave identically as new devices when checked with an ohm meter and diode tester, but I know that is not 100% reliable by any means. Of course, there are no capacitors. All the tracts on the board look great. Again, the most confusing part to me was unexpectedly high gain prior to failure.
When I get the new board installed, I run it through some square and triangle waves to be sure I'm not seeing overshoot, as well as carefully look for high frequency stuff on all wave forms. If I find something, I'll add a small cap across feedback resistors.
Steve
I had already checked offset and bias before the failure, and while getting a gain of about 17, and they were fine. I wasn't checking offset or bias at the time of the sudden loss of function of that channel several minutes later. I'll post some waveform photos I took shortly before shutting down later as well as though when wave form looked good.
It is quickest to simply rebuild a board from scratch since I have the parts and can't find anything defective when testing out of circuit, and the board is almost done. Quicker than changing a part, back on heatsinks for testing, changing another part, back on heatsinks, etc. But if I really needed to know exactly which component failed, I would need to do what you, Zen, suggested, I guess. Every single resistor measures right on the mark out of circuit (I have disassembled the bad board), and both JFETs and MOSFETs behave identically as new devices when checked with an ohm meter and diode tester, but I know that is not 100% reliable by any means. Of course, there are no capacitors. All the tracts on the board look great. Again, the most confusing part to me was unexpectedly high gain prior to failure.
When I get the new board installed, I run it through some square and triangle waves to be sure I'm not seeing overshoot, as well as carefully look for high frequency stuff on all wave forms. If I find something, I'll add a small cap across feedback resistors.
Steve
Some waveforms
I attached some photos. First is baseline on the L channel with input (yellow) on channel 1 and output (blue) on channel 2. Gain = 5.95 (2.93 / .492). Unfortunately, I never checked gain at baseline on the R channel, but a couple years ago it was 6, as expected.
Second photo is L channel after changing from 100 to 220 ohm feedback resistors - note that channels have reversed. Input in blue on channel 2 and output in yellow on channel 1. Gain = 11.76 (12/1.02).
Third photo is R channel after changing resistors. Input of 1V in blue and output 16.6V in yellow for gain of 16.6. Bias and offset were fine.
Fourth photo is R channel several minutes later, after which I powered down the amplifier.
I attached some photos. First is baseline on the L channel with input (yellow) on channel 1 and output (blue) on channel 2. Gain = 5.95 (2.93 / .492). Unfortunately, I never checked gain at baseline on the R channel, but a couple years ago it was 6, as expected.
Second photo is L channel after changing from 100 to 220 ohm feedback resistors - note that channels have reversed. Input in blue on channel 2 and output in yellow on channel 1. Gain = 11.76 (12/1.02).
Third photo is R channel after changing resistors. Input of 1V in blue and output 16.6V in yellow for gain of 16.6. Bias and offset were fine.
Fourth photo is R channel several minutes later, after which I powered down the amplifier.
Attachments
Hi
Vpp on the output is noted as 50V in the third shot, what rails are you running? 50V doesn't seem possible from a 48V supply unless something is terribly wrong or you have a custom supply voltage.
Sometimes parts will test good with a meter and not work correctly when in the circuit, so no real way to tell except one by one. Obviously it is something to do with the feedback itself.
It just seems to me the amp was running open-loop (a dry solder, maybe, or a damaged via?) and the only reason you saw 50Vpp was because it was basically the limit of the voltage supplies. Perchance do you have 29-30V rails?
Vpp on the output is noted as 50V in the third shot, what rails are you running? 50V doesn't seem possible from a 48V supply unless something is terribly wrong or you have a custom supply voltage.
Sometimes parts will test good with a meter and not work correctly when in the circuit, so no real way to tell except one by one. Obviously it is something to do with the feedback itself.
It just seems to me the amp was running open-loop (a dry solder, maybe, or a damaged via?) and the only reason you saw 50Vpp was because it was basically the limit of the voltage supplies. Perchance do you have 29-30V rails?
Hi
Not sure about what is up with that!
I wouldn't worry too much about the voltage measurement if you took it off an oscilloscope. The ADC resolution would be lower (and less accurate) at lower input voltage (much like DACs, actually - think of it as bit loss but in reverse). 12 x 1.4 = 16.97, which is 33.9Vpp and the oscilloscope is reading 34.8V. At a lower input you can expect the reading to be significantly less accurate. I understand that you may not have been able to fit it in a 20V/div window, which is why we don't use digital scopes to make measurements.
A True RMS meter would give you a better 'picture' of the actual output voltage. Obviously it is not a small difference but not one I am startled about. Maybe it's an unrelated fault?
12V RMS is also a lot of output, close to -3dB of total amplifier power. If no dummy load was connected the output devices are put through a lot of stress. You didn't indicate if there was some load or not and what it was.
Sorry I'm not more helpful. These things are troublesome to figure out but I'm sure you eventually will.
Not sure about what is up with that!
I wouldn't worry too much about the voltage measurement if you took it off an oscilloscope. The ADC resolution would be lower (and less accurate) at lower input voltage (much like DACs, actually - think of it as bit loss but in reverse). 12 x 1.4 = 16.97, which is 33.9Vpp and the oscilloscope is reading 34.8V. At a lower input you can expect the reading to be significantly less accurate. I understand that you may not have been able to fit it in a 20V/div window, which is why we don't use digital scopes to make measurements.
A True RMS meter would give you a better 'picture' of the actual output voltage. Obviously it is not a small difference but not one I am startled about. Maybe it's an unrelated fault?
12V RMS is also a lot of output, close to -3dB of total amplifier power. If no dummy load was connected the output devices are put through a lot of stress. You didn't indicate if there was some load or not and what it was.
Sorry I'm not more helpful. These things are troublesome to figure out but I'm sure you eventually will.
I finally got back to this amplifier. I built a new board with completely new components, but used the same 220 ohm feedback resistors, and installed it back in the amplifier (first photo - please excuse thermal paste on the heat sink). It biased up fine after an initial power-up with a light bulb in series (as always). I biased both channels to 1.2A without difficulty, and they held quite steady.
Wave forms on both channels were excellent (without the addition of a capacitor across feedback resistors), and I've shown the new channel in the photos, below. These were at 1 KHz, 1 watt output (2.84 V RMS into 8 ohm 100 Watt dummy load).
I then measured gain at both 1 KHz and at 60 Hz using digital volt meters and using the scope. Both methods gave identical results. Voltage gain in both channels was identical at 11.4.
I still am not sure what was wrong with that bad board, but it must be a JFET or MOSFET I would imagine, though I suppose it could have been a problem in the current limiting circuit. The board looked fine and all resistors measured fine when I disassembled the board. All transistors measured identically to new, identical transistors, using a diode checker and ohm meter. When I have some time, I'll put those transistors on a couple different transistor testers I have and see how they perform, but I don't have a curve tracer.
Thanks, everyone, for the advice.
Steve
Wave forms on both channels were excellent (without the addition of a capacitor across feedback resistors), and I've shown the new channel in the photos, below. These were at 1 KHz, 1 watt output (2.84 V RMS into 8 ohm 100 Watt dummy load).
I then measured gain at both 1 KHz and at 60 Hz using digital volt meters and using the scope. Both methods gave identical results. Voltage gain in both channels was identical at 11.4.
I still am not sure what was wrong with that bad board, but it must be a JFET or MOSFET I would imagine, though I suppose it could have been a problem in the current limiting circuit. The board looked fine and all resistors measured fine when I disassembled the board. All transistors measured identically to new, identical transistors, using a diode checker and ohm meter. When I have some time, I'll put those transistors on a couple different transistor testers I have and see how they perform, but I don't have a curve tracer.
Thanks, everyone, for the advice.
Steve
