Hi everyone, I recently picked up a Hafler 9180 that is sending HV DC to the right output during startup; amp is in great physical shape otherwise. I'm hoping to get some opinions on my debugging before throwing new parts at it since this is my first experience with a soft start.
From what I can tell, the voltage is coming from something turning on too early, shown by the first dip in the yellow trace at node 1 in the schematic versus the one big voltage drop in the purple trace at node 3 when the soft start comparator trips and turns everything on. Interestingly, the drain of the Q22 PMOS that turns on the input stage drifts down toward the - rail when the amp is turned off as does the cascode stage shown in the blue trace, while the drain stays close to 0 V; these both stay close to 0 in the left channel. I'm thinking this may point to an issue in Q7 or 8, but they seem to be testing okay, so my next thought was there might be too much channel resistance in Q22. Pretty much every component I have probed so far, including passives, tests the same as the working left channel. This amp doesn't have the substrate bias passives since it has the discrete components not the dual JFETs. The diodes between Q14 and 15 seem to be holding the bases ~3.8 V apart, which is a bit higher than the left channel. Adjusting the bias from ~260 mA to the recommended 230 mA for both channels didn't noticeably change anything.
This one's got me a bit stumped, so if anyone sees something I may not have looked at yet I'm all ears to test anything to get a better idea of what's going on before throwing new components at it. Cheers!
From what I can tell, the voltage is coming from something turning on too early, shown by the first dip in the yellow trace at node 1 in the schematic versus the one big voltage drop in the purple trace at node 3 when the soft start comparator trips and turns everything on. Interestingly, the drain of the Q22 PMOS that turns on the input stage drifts down toward the - rail when the amp is turned off as does the cascode stage shown in the blue trace, while the drain stays close to 0 V; these both stay close to 0 in the left channel. I'm thinking this may point to an issue in Q7 or 8, but they seem to be testing okay, so my next thought was there might be too much channel resistance in Q22. Pretty much every component I have probed so far, including passives, tests the same as the working left channel. This amp doesn't have the substrate bias passives since it has the discrete components not the dual JFETs. The diodes between Q14 and 15 seem to be holding the bases ~3.8 V apart, which is a bit higher than the left channel. Adjusting the bias from ~260 mA to the recommended 230 mA for both channels didn't noticeably change anything.
This one's got me a bit stumped, so if anyone sees something I may not have looked at yet I'm all ears to test anything to get a better idea of what's going on before throwing new components at it. Cheers!
Attachments
Welcome to the forum!
That you are able adjust bias current is encouraging. And bias current should be nearly independent of DC offset voltage, so a lack of interaction is not, in its self, surprising.
That said, I'm having trouble making sense of the scope traces. Any chance the scope ground is tied the the amp chassis? Do the traces change if you connect scope ground to speaker ground binding post?
If I understand your description:
ch. 1 (yellow) is node 1 in your attached schematic, base of Q9 ?
ch. 2 (blue) is node 2, base of Q14 ?
ch. 3 (purple) is node 3, base Q10 ?
ch. 4 (green) is node 4, source of Q22?
Would you confirm or correct? Would you note where 0V is positioned for each trace? What is time/div of sweep?
Q22 is an N-channel MOSFET. When its gate is driven to -15V, the FET is opened, and bias currents in all the transistors should drop to near 0.
The voltage span between bases Q14 and Q15 is set by P1 (in concert with Q13) and should be expected to vary with variations in output FETs. CR3 and CR4 are current limit clamps that shunt drive to the outputs when current out gets too large.
Please avoid "throwing new parts" in a an attempt to repair.
Good luck!
That you are able adjust bias current is encouraging. And bias current should be nearly independent of DC offset voltage, so a lack of interaction is not, in its self, surprising.
That said, I'm having trouble making sense of the scope traces. Any chance the scope ground is tied the the amp chassis? Do the traces change if you connect scope ground to speaker ground binding post?
If I understand your description:
ch. 1 (yellow) is node 1 in your attached schematic, base of Q9 ?
ch. 2 (blue) is node 2, base of Q14 ?
ch. 3 (purple) is node 3, base Q10 ?
ch. 4 (green) is node 4, source of Q22?
Would you confirm or correct? Would you note where 0V is positioned for each trace? What is time/div of sweep?
Q22 is an N-channel MOSFET. When its gate is driven to -15V, the FET is opened, and bias currents in all the transistors should drop to near 0.
The voltage span between bases Q14 and Q15 is set by P1 (in concert with Q13) and should be expected to vary with variations in output FETs. CR3 and CR4 are current limit clamps that shunt drive to the outputs when current out gets too large.
Please avoid "throwing new parts" in a an attempt to repair.
Good luck!
The scope ground was connected to the right output terminal ground for those traces, I can double check but I believe it was just a small offset when connected to chassis ground, no noticeable difference in the behavior.
The traces/nodes you mentioned are indeed how I had them hooked up, I had to start numbering the nodes to keep track of what I had looked at. In that picture, all of the traces have 0 V at the center division with 20 V/div, and the time there is 400 ms/div. The amp is turned on at the sharp rise in the yellow trace and switched off again when it starts the linear descent, so that negative voltage on the source of Q22 persists through the startup.
Additionally, node 5 at the base of Q15 does the same thing as node 2 during startup, only offset below by a few volts. If I let the supply caps completely discharge before switching on the amp, the negative voltage at Q22 is reduced and the output will only swing up to about 20 V or so, but with any voltage remaining on the caps it will swing basically to the rail.
Thanks for the response and hope this info helps!
The traces/nodes you mentioned are indeed how I had them hooked up, I had to start numbering the nodes to keep track of what I had looked at. In that picture, all of the traces have 0 V at the center division with 20 V/div, and the time there is 400 ms/div. The amp is turned on at the sharp rise in the yellow trace and switched off again when it starts the linear descent, so that negative voltage on the source of Q22 persists through the startup.
Additionally, node 5 at the base of Q15 does the same thing as node 2 during startup, only offset below by a few volts. If I let the supply caps completely discharge before switching on the amp, the negative voltage at Q22 is reduced and the output will only swing up to about 20 V or so, but with any voltage remaining on the caps it will swing basically to the rail.
Thanks for the response and hope this info helps!
Thanks, pinetronics. Very helpful.
Knowing that AC power was removed makes the waveforms much more reasonable looking. Would you advise about when the AC was removed? I might guess at about 5.6 divisions re left edge, but perhaps at 7.2 divisions re left edge. Please advise.
I hope there is no smoke produced if you allow the AC power to remain present. You should confirm about +/- 15V at supply pins of U1. Assuming no active smoke, a useful troubleshooting experiment would install a temporary short between pin 4 and pin 7 of U1B. This should cause the amp to remain in the disabled state, so that all devices should be held in 0 current bias. (You might add a ~22k resistor at the output terminals to provide a light load for off-state leakage currents.) Ideally, the output voltage should read near 0V and defective devices should be easier to locate.
Let us know what you observe and good luck!
Knowing that AC power was removed makes the waveforms much more reasonable looking. Would you advise about when the AC was removed? I might guess at about 5.6 divisions re left edge, but perhaps at 7.2 divisions re left edge. Please advise.
I hope there is no smoke produced if you allow the AC power to remain present. You should confirm about +/- 15V at supply pins of U1. Assuming no active smoke, a useful troubleshooting experiment would install a temporary short between pin 4 and pin 7 of U1B. This should cause the amp to remain in the disabled state, so that all devices should be held in 0 current bias. (You might add a ~22k resistor at the output terminals to provide a light load for off-state leakage currents.) Ideally, the output voltage should read near 0V and defective devices should be easier to locate.
Let us know what you observe and good luck!
AC is removed at 7.2 div re left edge in that case. Nothing bad has happened when leaving it on - the output does eventually stabilize to ~0 VDC within a few mV if AC is left on. When adjusting the bias I let the amp thermally stabilize each time for at least 10 minutes, and it was still stable then. Removing AC doesn’t create the same effect, so it does seem to only be a startup issue.
I will try shorting those pins and see how it behaves, good point. Hopefully something will look suspicious!
I will try shorting those pins and see how it behaves, good point. Hopefully something will look suspicious!
I've been playing around with shorting those pins and removing the jumper to see what happens. I'm noticing that on the left channel, point B on the schematic shows that -6 to -8 V offset at 1 div and turns on too early corresponding to the first dip in node 1 in the first image, versus the right channel staying at 0 V and turning on at the right time in the second image. The green trace is at B and the purple is on the other side of CR7.
Measuring DC voltages while the comparator is shorted, the only thing I can find is that Q22 source and drain are at ~-1.5 and 1.5 V respectively on the left channel, whereas the right has a bigger drop at -6 and 3.5 V respectively. I checked R42,43, CR7, and C20 and they all test within margin, and from what I can tell the same current is flowing through R8 and 9 on both channels.
I noticed R214 and 215 aren't present on the PCB despite that schematic being a scan from the manual I got with this amp, so at least I'm pretty sure the startup circuit can be ruled out since the exact same signal is being sent to both channels in this case.
Measuring DC voltages while the comparator is shorted, the only thing I can find is that Q22 source and drain are at ~-1.5 and 1.5 V respectively on the left channel, whereas the right has a bigger drop at -6 and 3.5 V respectively. I checked R42,43, CR7, and C20 and they all test within margin, and from what I can tell the same current is flowing through R8 and 9 on both channels.
I noticed R214 and 215 aren't present on the PCB despite that schematic being a scan from the manual I got with this amp, so at least I'm pretty sure the startup circuit can be ruled out since the exact same signal is being sent to both channels in this case.