I don't disagree about the rail fusing. I don't necessarily think that having potentially combustable insulation on the wiring is a problem (and also not AR's fault- that's an issue with the wire manufacturer), the bigger issue is that the wires were allowed to get that hot. PS rail fuses should be a fast blow, properly sized.
I don't know exactly how AR is using those ThermalTrak devices in this amp, so I can't make comments on substituting without a schematic. Something tells me you won't get AR to cough that up, but it's worth a try.
I've never been a big fan of the ThermalTrak devices for this reason- substitution is much more difficult, and there are other ways of making an amp thermally stable. One very highly performing amplifier did it by gluing a thermistor into the heatsink almost immediately behind one of the output devices, and they got it to stay within about 10% of the ideal bias current. That's pretty good in my book, especially for an output triple.
I don't know exactly how AR is using those ThermalTrak devices in this amp, so I can't make comments on substituting without a schematic. Something tells me you won't get AR to cough that up, but it's worth a try.
I've never been a big fan of the ThermalTrak devices for this reason- substitution is much more difficult, and there are other ways of making an amp thermally stable. One very highly performing amplifier did it by gluing a thermistor into the heatsink almost immediately behind one of the output devices, and they got it to stay within about 10% of the ideal bias current. That's pretty good in my book, especially for an output triple.
So, look at the schematic. Why would people even comment without looking? Too much hard work to Google?
I had to look up- ThermalTrak devices as being out of touch with up to date methods has left me at a disadvantage as I am more used to the old fashioned method as H713 mentioned.
Implementing Schematics and a ThermalTrak hybrid amplifer
What worries me when you scroll down is the implementation circuit diagram showing compensation/feedback capacitors= 150pf having to be attached directly to the signal path, can anybody reassure me they will have ZERO effect on the signal ?
Implementing Schematics and a ThermalTrak hybrid amplifer
What worries me when you scroll down is the implementation circuit diagram showing compensation/feedback capacitors= 150pf having to be attached directly to the signal path, can anybody reassure me they will have ZERO effect on the signal ?
Instead of rail fuses a thyristor-based crowbar could be used. E.g. in my Revox A740 power amp from the 70's the protection circuit triggers the crowbar if DC appears in output or heatsink temperature is too high. The crowbar thyristor shorts the output DC rails thus blowing the main fuse. In Revox A740 this is used as the last resort as the first-level protection shuts down output signal and quiescent current to outputs.
Typo in post above - the amp has no rail fusing - only a AC mains fuse - 8Amp slow blow.
Also I have schematic and parts list if any one wants it I'll email it to them, send me an email at pbnaudio@me.com
As far as the flammable wiring, ARC should have tested this before using it IMO
Good Listening
Peter
Also I have schematic and parts list if any one wants it I'll email it to them, send me an email at pbnaudio@me.com
As far as the flammable wiring, ARC should have tested this before using it IMO
Good Listening
Peter
The energy in the filter-caps is substantial, likely damaging the thyristor(s) themselves. I've heard reports of pcb traces vaporizing in sympathy with the fuses in such conditions too...
Better to disconnect the amp from its supplies than sabotage them! This can be done pretty simply with MOSFETs these days, unlike in the '70s
+1
Even easier is the use of a couple fast-blow fuses, but MOSFETs could switch fast enough to protect the output devices if they aren't the original fault.
Even if the rails had something high like 15A fast-blow fuses on them, they would still probably stop things from catching fire. By my math, a 65V rail with 200,000 uF of capacitance stores about 425 joules of energy. You could certainly see a pulse as high as 50 or 100 amps if shorted, especially with that many caps in parallel.
Did you test it or just listen to it? I bet there was some indication that it was going to fail- a rising current somewhere, some noise or instability- something that needed further investigation. Things can be surprisingly screwed up and still sound quite good.
I know I was voted down in my diagnosis about oscillation but have you considered that any power amp has a circuit board designed precisely for
that particular design which then uses a set value of compensation capacitor to set stabilization.
Now think-----if somebody repairing that amplifier causes changes to be made to the circuit due to "improper " repairs --and some of you think that,then the circuit will become unstable as the compensation capacitor does not fulfill its design requirements therebye causing a slow rise in oscillation due to heat changes affecting stability even more .
I have come across this on many occasions where the safety margin of compensation was exceeded-----and watched long term as a small oscillation became a large one, that wont be shown by any virtual circuit but only by practical day to day experience.
that particular design which then uses a set value of compensation capacitor to set stabilization.
Now think-----if somebody repairing that amplifier causes changes to be made to the circuit due to "improper " repairs --and some of you think that,then the circuit will become unstable as the compensation capacitor does not fulfill its design requirements therebye causing a slow rise in oscillation due to heat changes affecting stability even more .
I have come across this on many occasions where the safety margin of compensation was exceeded-----and watched long term as a small oscillation became a large one, that wont be shown by any virtual circuit but only by practical day to day experience.
My initial schematic search was unsuccessful, though after making a hifiengine account I managed to find it. At least from what I can see, there should be no reason why the current production NJL3281DG and NJL1302DG devices aren't suitable. Costly to buy a bunch and match them? Sure. But again, this is a pretty expensive amplifier.
This amp has 200,000 uF of filter capacitance. That can do a serious amount of damage before the mains fuse ever feels anything.
This amp has 200,000 uF of filter capacitance. That can do a serious amount of damage before the mains fuse ever feels anything.
Once the repair is done, you simply have to test the amplifier with at least a resistive load at its outputs, tone generator, and an oscilloscope. Especially amps that have 8A slow blow AC fuse!!!
If you input rf into and amplifier the input low pass filter may not stop penetration. rf is used for welding plastics and it can cause untold damage with electronics. A friend of mine lost an amplifier due to an rf input being applied. It is plausible this resulted in mutual conduction of the output transistors which caused the overheating.
I agree but thyristor crowbar could be retrofitted to an amp already having a DC protection circuit such as this AR HD220. And there is no need to put it on a PCB. I have nothing against adding rail fuses but their main purpose is to prevent fire hazard not to protect the circuits or components. Same thing with the crowbar.
There are some issues with rail fuses so they are often left out and replaced with electronic protections. However the designers of AR HD220 clearly have made a shortcut in their protection circuit.
You know that crowbarring big filter caps will generate hundreds or even thousands of amps? The kind of thyristors that can survive that are big and expensive, and its not great for the caps either.
I'll say again this is a very unsatisfactory approach, like stopping a car engine by shorting out the battery...
I'll say again this is a very unsatisfactory approach, like stopping a car engine by shorting out the battery...
The crowbar thyristor should have a series resistor (e.g. 100mohms) and the caps also have series resistance so the current will definately not be thousands of amps. The peak surge current of even quite small thyristors is in the region of hundreds of amps.
BTW your analogy is familiar to me since the alternator diode bridge in my car shorted thus sorting the battery when I was driving. Actually a very mild event. Car (diesel powered) just stopped and did not start. No sparks or explosions 😉
BTW your analogy is familiar to me since the alternator diode bridge in my car shorted thus sorting the battery when I was driving. Actually a very mild event. Car (diesel powered) just stopped and did not start. No sparks or explosions 😉
This is all fascinating but entirely irrelevant to this particular amp. Did you even bother to look at the circuit? No comp caps whatsoever.
It is the only approach to protection without audible artefacts. Surprisingly, to some this matters.
May I re-open this thread to see if there are any thoughts on the design concepts involved with ThermalTrak bias stabilization?
Let me jump to the specific aspect on my mind before I launch into a probably too-long explanation of why I'm asking: It seems that the design center for ThermalTraks is to connect the sensing diodes in a series arrangement that produces a total Vforward that is appropriate for the Darlington configuration of the output stage. That is what is done in the ON Semi app note (link below), and also what is done in the ARC DH220.
https://www.onsemi.com/pub/Collateral/AND8196-D.PDF
But the thing is, how is this supposed to stabilize bias if one of the paralleled output transistors runs significantly hotter than the others? Only the one compensation diode in the hot-running part will have its Vforward decrease (initially, at least). That voltage shift will end up being shared over the set of paralleled outputs, so will it be enough to keep the hotter part from running away? Sure, once the hot-running part warms up the others through the heat sink thermal resistance, stability may be reached - but that seems like it is exactly the situation that ThermalTraks are supposed to eliminate.
Okay, reason for asking: I've got an ARC HD220 that I'm being asked to repair after some previous worker made a rather large misstep. I'll spare you the details - the key thing is that the present repair will need to replace virtually every semiconductor on the output stage PCBs. The ThermalTrak outputs ("G" versions) are in hand.
I'm got an almost-fully-rebuilt output PCB on the bench running off a benchtop supply. All the parts that are meant to be mounted to the heat sink are so mounted. The ThermalTrak output transistors were matched for Vbe with a multimeter diode test function, to about +/- 1mV. The benchtop supply is stout enough to bias things up, but not much more. There is no external load on the audio output. I wanted to investigate bias stability before I attempt to test with any appreciable audio output.
What I observe is that if I trim the bias so that the average idle current in each output transistor is about 20mA, it stays nicely stable. I would be well satisfied with that in many amplifiers, but this one is apparently meant to idle with closer to 100mA per output transistor, per the schematic notes. (I assume the intended high idle current is at least partly related to the lack of an overall NFB loop encompassing the output stage.)
If I trim the bias for much more than 20 mA per output transistor, I observe that one transistor in each of the + and - halves will start to heavily dominate the total idle current, and their currents will continue to rise until I'm too scared to continue testing. I wondered, are the high-current transistors simply not matched well enough to their peers? Perhaps, but... if I cut the collector connections to the high-current transistors and try the same test, the apparent runaway just moves to some other transistors.
Many thanks for your thoughts - or even if you've read this far!
chazix
Let me jump to the specific aspect on my mind before I launch into a probably too-long explanation of why I'm asking: It seems that the design center for ThermalTraks is to connect the sensing diodes in a series arrangement that produces a total Vforward that is appropriate for the Darlington configuration of the output stage. That is what is done in the ON Semi app note (link below), and also what is done in the ARC DH220.
https://www.onsemi.com/pub/Collateral/AND8196-D.PDF
But the thing is, how is this supposed to stabilize bias if one of the paralleled output transistors runs significantly hotter than the others? Only the one compensation diode in the hot-running part will have its Vforward decrease (initially, at least). That voltage shift will end up being shared over the set of paralleled outputs, so will it be enough to keep the hotter part from running away? Sure, once the hot-running part warms up the others through the heat sink thermal resistance, stability may be reached - but that seems like it is exactly the situation that ThermalTraks are supposed to eliminate.
Okay, reason for asking: I've got an ARC HD220 that I'm being asked to repair after some previous worker made a rather large misstep. I'll spare you the details - the key thing is that the present repair will need to replace virtually every semiconductor on the output stage PCBs. The ThermalTrak outputs ("G" versions) are in hand.
I'm got an almost-fully-rebuilt output PCB on the bench running off a benchtop supply. All the parts that are meant to be mounted to the heat sink are so mounted. The ThermalTrak output transistors were matched for Vbe with a multimeter diode test function, to about +/- 1mV. The benchtop supply is stout enough to bias things up, but not much more. There is no external load on the audio output. I wanted to investigate bias stability before I attempt to test with any appreciable audio output.
What I observe is that if I trim the bias so that the average idle current in each output transistor is about 20mA, it stays nicely stable. I would be well satisfied with that in many amplifiers, but this one is apparently meant to idle with closer to 100mA per output transistor, per the schematic notes. (I assume the intended high idle current is at least partly related to the lack of an overall NFB loop encompassing the output stage.)
If I trim the bias for much more than 20 mA per output transistor, I observe that one transistor in each of the + and - halves will start to heavily dominate the total idle current, and their currents will continue to rise until I'm too scared to continue testing. I wondered, are the high-current transistors simply not matched well enough to their peers? Perhaps, but... if I cut the collector connections to the high-current transistors and try the same test, the apparent runaway just moves to some other transistors.
Many thanks for your thoughts - or even if you've read this far!
chazix