Hello again,
Sorry for yet another post but found this one a rather interesting scenario. But I was running this amp for an hour or 2, ran amazing without an issue until the short circuit detection kicked in. I turned off the amp and found a immediate drain of the caps. I turn the amp on after a bit, everything back to normal. Short circuit kicks in again this time the power supply was completely shorted considering the nice screeching sound before shutting off itself. The weird thing was only Q16/Q17 where the only devices damaged which is most likely an indication that the bias current somehow spiked to critical levels. The VBE servo and the power darlingtons where thermally coupled to a large aluminium heatsink.
Changes I have made:
Changed the VAS and CCS transistors for the output stage from KSA992/KSC1845 to 2SA1201/2SC2911 in case the VAS and CCS where getting hot considering they are only small BJTs with 6mA going through them
I tested the removed VAS and CCS transistors and they appear to still be transistors and not a resistor, diode, or jumper.
Sorry for yet another post but found this one a rather interesting scenario. But I was running this amp for an hour or 2, ran amazing without an issue until the short circuit detection kicked in. I turned off the amp and found a immediate drain of the caps. I turn the amp on after a bit, everything back to normal. Short circuit kicks in again this time the power supply was completely shorted considering the nice screeching sound before shutting off itself. The weird thing was only Q16/Q17 where the only devices damaged which is most likely an indication that the bias current somehow spiked to critical levels. The VBE servo and the power darlingtons where thermally coupled to a large aluminium heatsink.
Changes I have made:
Changed the VAS and CCS transistors for the output stage from KSA992/KSC1845 to 2SA1201/2SC2911 in case the VAS and CCS where getting hot considering they are only small BJTs with 6mA going through them
I tested the removed VAS and CCS transistors and they appear to still be transistors and not a resistor, diode, or jumper.
Remote troubleshooting can be difficult but I have a few observations/suggestions. RV1 is wired so that the lidle current should greatly decrease if the wiper goes resistive, so that probably is not the problem.
The output-protection circuitry protects the speakers but not the amplifier. Q14/Q15 and their associated circuitry are there to protect the output transistors -- but they didn't. I also note that the -3dB points of R21/C10 and R22/C11 are about 3.3KHz. That suggests something, in terms of the "screeching" sound you heard -- the transistors could be triggering on and off due an oscillation condition, but the circuit as shown looks reasonable in terms of stability. Therefore there's a fault or situation that is NOT on the schematic I would take a look at C4 to see if it's open or if one of its connections has gone resistive. A cold solder perhaps?
I'm puzzled by the presence of D1 and D2. If the output offset voltage exceeds their turn-on voltage, the amplifier basically goes open-loop. Wouldn't that be a bad thing??
The output-protection circuitry protects the speakers but not the amplifier. Q14/Q15 and their associated circuitry are there to protect the output transistors -- but they didn't. I also note that the -3dB points of R21/C10 and R22/C11 are about 3.3KHz. That suggests something, in terms of the "screeching" sound you heard -- the transistors could be triggering on and off due an oscillation condition, but the circuit as shown looks reasonable in terms of stability. Therefore there's a fault or situation that is NOT on the schematic I would take a look at C4 to see if it's open or if one of its connections has gone resistive. A cold solder perhaps?
I'm puzzled by the presence of D1 and D2. If the output offset voltage exceeds their turn-on voltage, the amplifier basically goes open-loop. Wouldn't that be a bad thing??
The emitter/base/collector connections on Q12 also are suspect, along with R13/14. Is Q12 board-mounted or connected via wires?
Think that's a current limiting circuit, which is normally off.
Your name is familiar somehow, maybe from Tektronix etc?
I'm skeptical of that. Look at what happens when one of the diodes turns on. The DC NFB goes from unity to 23 (worst case).
Yes, I did work at Tek. 1974 through 1994. Their IC fab I was working in was acquired by a different company then.
I performed failure analysis on IC's and discrete devices. It was a lot of fun -- on average, anyway. I got to take IC's/transistors apart using really expensive tools, and didn't have to put them back together again!
Self puts those diodes (some times two in series in each direction) across the feedback cap to prevent damage to it in fault conditions. That's because he uses a low voltage polar cap as it usually sees negligible DC and AC voltage across it. Personally I always use / replace it with a bipolar Nichicon Muse ES rated to at least rail voltage. Bulkier and more expensive but more elegant, I think.
I decided to rig up the VI Limiter to allow a bit more time for short circuit detection to pick-up and open the relay. So the threshold is a bit more higher than the desired output power. The VI limiters did allow the output stage to last a lot longer before completely failing to the suspected over current condition. I'm starting to think it's an oscillation seeing as the KSC1845 can act as a VAS despite the high Vce without heating. I believe a calculated around 240mW which is under the 500mW rating of the part. Considering the 2SC2911 also gets hot as a VAS an oscillation is the only other reason I can think of for excessive heating in turn raising the VAS current and the bias current to the danger zone. I am most likely wrong and correct me if I am. I did have the amp circuit on a breadboard - the worst thing to use in a sensitive application. The amp was stable and no VAS or CCS heating. I will do more digging and tests seeing as I have yet to obtain an oscilloscope.
Food for thought:
- R2 seems to be two high in value, typ used for RF supperation. The audio gain reduction of 20% seems high.
- The bias circuit Q12 should have a small CB cap and CE 1-10uF bypass so it will never oscillate.
- Jumper RV1 pin 2 and 3.
- R9 should have a pole & zero RC?
- Why are current limiters on the VAS different? R17 & R18 should be equal.
Ive just performed some tests and found some rather interesting results.
So I ran the amp with the bias very low and the amp seemed fine, obviously there would be +100% crossover distortion but the amp stayed at a constant 1.2mV drop across the emitter resistor. The VAS and CCS even stayed cool. Its when I apply bias that the issue arrises. The system even with a bit of bias climbs and eventually turns into chernobyl. I will record a couple of videos but it was rather interesting. So it makes me wonder what kind of oscillation is present or if I screwed something up in the VAS stage.
So I ran the amp with the bias very low and the amp seemed fine, obviously there would be +100% crossover distortion but the amp stayed at a constant 1.2mV drop across the emitter resistor. The VAS and CCS even stayed cool. Its when I apply bias that the issue arrises. The system even with a bit of bias climbs and eventually turns into chernobyl. I will record a couple of videos but it was rather interesting. So it makes me wonder what kind of oscillation is present or if I screwed something up in the VAS stage.
Normally fast changing bias indicates bad thermals. Maybe this has something to do with it. Now to fix the VAS heating. Censored Profanity
It's not unusual for a solution to one problem to cause a problem in a different scenario. In the case where the diodes turn on to prevent damage to the capacitors, the speaker could suffer. What priority should the designer consider???
Yes you need to fix that problem, the output transistor(s?) need to be well-coupled to the heat sink and Q12. Good catch there.
In my designs I didn't worry about CB caps on the bias-setting transistors (although it wouldn't hurt). I just made sure there was a good-sized capacitor across the C-E nodes. With an output load there's enough variation in the output Q's base currents to modulate their bias current -- after all, the Vbe multiplier's Zout isn't zero -- and that can result in higher distortion. Simulations on some more-recent doodlings of mine confirm that.
Out of curiosity, is this a kit? It sounds to me like you are in the process of moving the amp from a breadboard to a packaged-up version.
In my designs I didn't worry about CB caps on the bias-setting transistors (although it wouldn't hurt). I just made sure there was a good-sized capacitor across the C-E nodes. With an output load there's enough variation in the output Q's base currents to modulate their bias current -- after all, the Vbe multiplier's Zout isn't zero -- and that can result in higher distortion. Simulations on some more-recent doodlings of mine confirm that.
Out of curiosity, is this a kit? It sounds to me like you are in the process of moving the amp from a breadboard to a packaged-up version.
I plan to use this amp to replace my 5.1 chipamp board. Going up to 7.1 with an embedded sub. The sub amp works first go and already have a solid 12 hours on that PCB, the VAS and CCS get a little bit warm on that but not toasy like this board. Both models share the exact same circuit but different output stages and VAS currents. If anything I might end up scrapping this board and creating another revision and go all out on it. e.g. cascode differential.
If anything I should be attaching a couple of photos and full schematics of both amp protatypes and see if I missed anything.
If anything I should be attaching a couple of photos and full schematics of both amp protatypes and see if I missed anything.
The main issue with these amps is that they are locked at 45-0-45 due the LTP fixed current resistor instead of a diode/s. This is also a requirement of the UPC1237's used in the safety features.
Well, in this case what could cause that cap to see a high DC voltage would be the output sticking to one rail, at which point you would expect some form of speaker DC protection (which he also includes in his designs) to have been triggered. I guess his idea is that, whatever the fault is, you don't want to have to deal with capacitor entrails covering your PCB as well...
I was speaking generically, and I know how I would not do it. I would not include those diodes.
BTW the 22K resistor would limit the current flow into the cap. It wouldn't explode. I agree that it could leak, certainly an undesirable thing. But in this kind of failure scenario there's a good chance there's badness elsewhere on the board -- some other capacitor has leaked or exploded, a resistor or transistor burned up etc....none of which would be prevented by a speaker-protection relay. Beyond using best design practices -- properly-rated components, good board design, good thermal management -and so on - there's not much you can do. Sometimes things just go wrong....
BTW the 22K resistor would limit the current flow into the cap. It wouldn't explode. I agree that it could leak, certainly an undesirable thing. But in this kind of failure scenario there's a good chance there's badness elsewhere on the board -- some other capacitor has leaked or exploded, a resistor or transistor burned up etc....none of which would be prevented by a speaker-protection relay. Beyond using best design practices -- properly-rated components, good board design, good thermal management -and so on - there's not much you can do. Sometimes things just go wrong....
Ah I am more than happy to start over, learning experience. May I ask for some transistor recommendations for the VAS for 6mA and 12mA VAS stages? I am gonna upgrade both amps and see what happens.
The amp still goes thermal runaway I'll have to get mica pads and see if i can better mount the transistors, or make up some copper plates as a heatspreader
You just about have to have a thermal-resistance problem between your transistors and heat sink. If you drilled/tapped the mounting holes in the HS perhaps you didn't go deep enough so you can't properly secure one or more of your transistors to the HS. To make it easier to get them started, taps have a taper on their end so in the case of a blind hole (it's not through-drilled) the threads near the bottom aren't completely formed-- which might prevent the screw(s) from going down far enough There's a type of tap called a "bottoming tap" with an almost nonexistent taper that can address that. They're used after the one with the long taper is used. They may not be commonly-available for smaller taps. I've seen them for Imperial threads but don't recall seeing them for metric threads....but they just about HAVE to be available somewhere. If all else fails, grinding a long-taper tap's end down will get you one.....but don't totally get rid of that taper or it may be impossible to get the tap properly started into the hole.
If the hole is thru-drilled, you still could have a problem with the threads if you didn't run the tap down far enough. I have done that to myself. There are numerous ways to shoot yourself in the foot here.
Another suggestion, something I've found handy on occasion in situations when I suspect excess temperatures may be present. It's less painful than touching stuff. Use an IR thermometer. I used mine very recently when I was seeing an oscillation problem with my headphone amplifier (I was using REW to check its harmonic distortion). The .5W dummy load resistor was heating up FAST. The problem turned out to be due to signal leakage in the USB sound card I was using, due to its built-in headphone amplifier. Fortunately, turning it off fixed the problem.
I bought one from Harbor Freight, possibly not around in your locale -- but there has to be an equivalent. Or Amazon. Or ebay.
Another possible issue is the thermal lag between the output transistors and Q12. If Q12 is too far away, thermally speaking, the relatively-long delay plus an attendant temperature gradient might be a problem. It's not clear how different your bench test setup and final packaging system are. Perhaps worth thinking about.