The power dissipation by the 5532 is due to the drop across the internal resistors and transistors during current limiting (it's trying to go to the ±15v rails and cannot due to the LEDs and low value resistor).
I was thinking about ways to reduce heating. Reducing the supply voltage to the last 5532 would do it. Doing as you suggested, moving the dissipation to the resistor would also do it. It only took a few minutes to wire it up and it satisfied my curiosity as to whether it would work. The 1k gate resistor significantly limits the transient current during rise and fall but I couldn't tell much different in heating with it or without it.
Side note, 2.96 amps idle current but it's not a completely healthy amp so I don't know if that's even relevant.
I was thinking about ways to reduce heating. Reducing the supply voltage to the last 5532 would do it. Doing as you suggested, moving the dissipation to the resistor would also do it. It only took a few minutes to wire it up and it satisfied my curiosity as to whether it would work. The 1k gate resistor significantly limits the transient current during rise and fall but I couldn't tell much different in heating with it or without it.
Side note, 2.96 amps idle current but it's not a completely healthy amp so I don't know if that's even relevant.
I wanted to go experiment with this today, but I'm a little snowed in at the moment. I will try to do it when the roads are passable. Finding higher power resistors is easier than finding new opamps.
Changing the power supplies of the last opamp is certainly one way to do it, but I don't see an easy way to do that. If memory serves, there are current limiting resistors on the supply legs, but it is shared between both opamps.
It seems like a poor design decision to rely on the opamp short circuit current protection to be the current limiter in the design. But, we have to deal with the cards we are dealt.
I found this quick little power dissipation calculator from analog that basically validates what we were saying/thinking. Power Dissipation vs Die Temp | Design Center | Analog Devices It also can help figure out values quickly.
Using the current NE5532 parameters of 40C ambient, 100 theta-jc, +/-15V supplies, 16mA quiescent current, 5V load output, 0V load ground, and 140 ohm (100 ohm resistor and 1.5V/40mA = 40ohm LED) load resistance, I get 124C junction temperature. The max is 150C according to the datasheet. If I keep everything the same and switch opamps to something with about 4mA quiescent, it goes down to 88C. If I keep the NE5532 and switch to 13V output voltage and 1000 load resistor, I get 91C. If I switch to an opamp with 4mA quiescent, 13V output, and 1000 ohm load, it goes all the way down to 56C.
With the 1000 ohm load resistor and 13V opamp output, I get an output current of 11.5mA. That is still well in the 3120's recommended min/max current of 7 to 16mA. It keeps the temperature more in check. I'll go try the 1000 ohm resistor when I can with the NE5532. I will probably stack 2 2k resistors on top of each other because the power handling of the resistors need to be higher than a standard 0805. On my next round of parts, I am going to get the OPA1662-Q1 opamp to try also. I'll experiment with this if I'm not satisfied with the 1k load.
Changing the power supplies of the last opamp is certainly one way to do it, but I don't see an easy way to do that. If memory serves, there are current limiting resistors on the supply legs, but it is shared between both opamps.
It seems like a poor design decision to rely on the opamp short circuit current protection to be the current limiter in the design. But, we have to deal with the cards we are dealt.
I found this quick little power dissipation calculator from analog that basically validates what we were saying/thinking. Power Dissipation vs Die Temp | Design Center | Analog Devices It also can help figure out values quickly.
Using the current NE5532 parameters of 40C ambient, 100 theta-jc, +/-15V supplies, 16mA quiescent current, 5V load output, 0V load ground, and 140 ohm (100 ohm resistor and 1.5V/40mA = 40ohm LED) load resistance, I get 124C junction temperature. The max is 150C according to the datasheet. If I keep everything the same and switch opamps to something with about 4mA quiescent, it goes down to 88C. If I keep the NE5532 and switch to 13V output voltage and 1000 load resistor, I get 91C. If I switch to an opamp with 4mA quiescent, 13V output, and 1000 ohm load, it goes all the way down to 56C.
With the 1000 ohm load resistor and 13V opamp output, I get an output current of 11.5mA. That is still well in the 3120's recommended min/max current of 7 to 16mA. It keeps the temperature more in check. I'll go try the 1000 ohm resistor when I can with the NE5532. I will probably stack 2 2k resistors on top of each other because the power handling of the resistors need to be higher than a standard 0805. On my next round of parts, I am going to get the OPA1662-Q1 opamp to try also. I'll experiment with this if I'm not satisfied with the 1k load.
Last edited:
It's nice to see someone willing to learn something new but from a purely 'repair' standpoint, we're both likely not doing much. This design has been around since (at least) 2000 and many thousands of amps have been produced. The reliability has been good.
To reduce the supply voltage, you could lift the supply legs, connect a zener across them (20v?) and insert suitable current limiting resistors. You may also need to add two 0.1uF caps from pins 4 and 8 to ground.
To reduce the supply voltage, you could lift the supply legs, connect a zener across them (20v?) and insert suitable current limiting resistors. You may also need to add two 0.1uF caps from pins 4 and 8 to ground.
I agree that this is all overkill for a repair that appears to be working again. It is possible that something else is still not quite right and we are just seeing that effect or something like that. I know what I am chasing is not uncommon, but as a few of the other threads here have shown, the issue is not unheard of. When I get back to the lab I'll try a couple of our ideas and report back. Maybe it can help someone in the future and make this particular amp a little more reliable.
I tried out a couple things in regards to this. I switched to a 1k resistor, and it dropped the temperature by about 15 degrees. Not a silver bullet, but certainly an improvement. The output of the opamp did jump up to +/- 14V. The input at the opto was +/-1.5V like it should be.
I found out I had done something I shouldn't have also during my initial repair. C412 and the corresponding and C417 were not present on the amp when I got it. I figured (erroneously) that these were used as bypass caps, so I used a standard 0.1uF 100V 1206 cap in those locations. I had always looked at the outputs of the optocouplers, not directly at the gates of the FETs (or right after the 100 ohm resistor). I did not want to hook up 2 probes before and do the trace math. I did that today and found this issue. The 100 resistor plus 0.1uF cap caused a pretty terrible response on the gates. This was probably one of the reasons for excess inductor heating. Bad waveforms below (one for each optocoupler).
After removing C412 and C417 again, I have decent square waves onthe gates again. Just one shown below Do you know what value these caps should be?
I didn't have the exact other opamp I wanted, but I did try a TLE2072 opamp in place of the NE5532. It has substantially less supply current, but is a FET input, not bipolar. I wasn't expecting the whole amp to act the same, I just wanted to see the output. It did actually work well enough to get an idea of the temperature drop. It dropped down to roughly 70C, which is MUCH better than the NE5532. The output of the opamp still looked good, but the overall amp started drawing 8 amps and the outputs were heating up. I'm guessing something with the feedback was causing something, but I only had it on for 5-10 seconds to get an idea of the opamp dissipation. I will still try a more compatible opamp later, but I do not have it on hand. Below is the output (pin 7) of that opamp going into the optocoupler. That alone looked OK.
Lastly, I did pull all 8 output filter caps and measured them. 6 of them measure within 5uF of their rated 470uF. Another was 455uF and the last one was 430uF. Those last 2 are not obviously bad, but I will replace them anyway as the other 6 were almost spot on. I put all of them back in for now. I plan to reinstall the rest of the outputs and actually load test it. I had checked the preamp board already and everything on there seems good.
I found out I had done something I shouldn't have also during my initial repair. C412 and the corresponding and C417 were not present on the amp when I got it. I figured (erroneously) that these were used as bypass caps, so I used a standard 0.1uF 100V 1206 cap in those locations. I had always looked at the outputs of the optocouplers, not directly at the gates of the FETs (or right after the 100 ohm resistor). I did not want to hook up 2 probes before and do the trace math. I did that today and found this issue. The 100 resistor plus 0.1uF cap caused a pretty terrible response on the gates. This was probably one of the reasons for excess inductor heating. Bad waveforms below (one for each optocoupler).
After removing C412 and C417 again, I have decent square waves onthe gates again. Just one shown below Do you know what value these caps should be?
I didn't have the exact other opamp I wanted, but I did try a TLE2072 opamp in place of the NE5532. It has substantially less supply current, but is a FET input, not bipolar. I wasn't expecting the whole amp to act the same, I just wanted to see the output. It did actually work well enough to get an idea of the temperature drop. It dropped down to roughly 70C, which is MUCH better than the NE5532. The output of the opamp still looked good, but the overall amp started drawing 8 amps and the outputs were heating up. I'm guessing something with the feedback was causing something, but I only had it on for 5-10 seconds to get an idea of the opamp dissipation. I will still try a more compatible opamp later, but I do not have it on hand. Below is the output (pin 7) of that opamp going into the optocoupler. That alone looked OK.
Lastly, I did pull all 8 output filter caps and measured them. 6 of them measure within 5uF of their rated 470uF. Another was 455uF and the last one was 430uF. Those last 2 are not obviously bad, but I will replace them anyway as the other 6 were almost spot on. I put all of them back in for now. I plan to reinstall the rest of the outputs and actually load test it. I had checked the preamp board already and everything on there seems good.
Update to this. I replaced the two output capacitors and the NE5532 with a OPA1662 (and 1k output resistor). After I did this, the opamp temperature did drop to about 90C total (from about 110-115) on the FLIR. The waveforms on the output of the optocoupers look the same and the gates of the output FET's still look the same. I don't think they really look all that great, but they match what Perry has as reference at the optocoupler output.
The input current dropped from 4 amps to 3.1 amps. I am not sure if it is the opamp or the output capacitors that helped this. The output inductors now get warm, but not scorching hot. With the preamp board bypassed, I was getting about 60V peak to peak on the output with a 15V peak to peak input. I was not able to load test it today, but I will do that soon.
The input current dropped from 4 amps to 3.1 amps. I am not sure if it is the opamp or the output capacitors that helped this. The output inductors now get warm, but not scorching hot. With the preamp board bypassed, I was getting about 60V peak to peak on the output with a 15V peak to peak input. I was not able to load test it today, but I will do that soon.
I do not think that is it. I had a 1k resistor in there before with the NE5532 and the original output capacitors. The current draw was still around 4 amps then. Using a 1k resistor with both the NE5532 and OPA1662 increased the voltage output of the opamp to its rail, but the effect is that the current into the LED of the optocoupler was reduced. The opamps were no longer relying on their short circuit current protection.
Hopefully this is the last update I have for this amp. I tested it up to ~450W RMS into my 4 ohm load resistor. I played constant for a minute with a 100Hz test tone at that power level. I had a nice unclipped signal with +/-60V peaks. I checked the thermals about 30 seconds into it and everything looked OK. The output inductor never got above 60C, and the opamp in question never got about 90C. The +15V reg was getting warm about 70C, but that is also because the fans start running. I wasn't too worried about it. I could not run it more than about a minute as my load resistors start getting obscenely hot.
I still have to re-epoxy the output inductors back to the board/carrier, but that is trivial. It has been a long time since the top cover has been screwed back onto this amp!
I still have to re-epoxy the output inductors back to the board/carrier, but that is trivial. It has been a long time since the top cover has been screwed back onto this amp!
It did not. I do not have the screen shot of it, but it stayed right around the 60kHz mark. When I checked it before, it was the same time scales as I used before. The opamp I picked was meant to be as close as I could find to the NE5532 specs with a lower quiescent current. When I randomly tried a TLE2072 earlier, that did change it. That opamp is quite a bit different than the NE5532's though.
As a side note, when I was putting out 450W RMS, I was pulling about 50 amps at 13.8V, so about 65% efficient.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.