So I salvaged two large Peavey CS800 power amps. They are old school class A/B quasi complimentary output with a 50lb power transformer.
I kept blowing transistors until I realized the speaker protection crowbar triac had gone dead short permanently.
This is the second time this triac has gone short circuit. I think what happens is it get's triggered on power up because one power rail rises quicker than the other.
As usual, like any semiconductor, it's failure mode is short circuit.
I really question the value of a crowbar. I guess the philosophy is save the speakers by sacrificing the amp.
One thing I noticed is the filter circuit for firing the triac has a capacitor half the value of what is listed on the schematic. I speculate I could easily triple or quadrupole that value as listed on the schematic and still save speakers.
Maybe I'll just nix the crowbar.
Thoughts?
I kept blowing transistors until I realized the speaker protection crowbar triac had gone dead short permanently.
This is the second time this triac has gone short circuit. I think what happens is it get's triggered on power up because one power rail rises quicker than the other.
As usual, like any semiconductor, it's failure mode is short circuit.
I really question the value of a crowbar. I guess the philosophy is save the speakers by sacrificing the amp.
One thing I noticed is the filter circuit for firing the triac has a capacitor half the value of what is listed on the schematic. I speculate I could easily triple or quadrupole that value as listed on the schematic and still save speakers.
Maybe I'll just nix the crowbar.
Thoughts?
Take the crowbar OUT and install a traditional speaker protect relay. Either a big mechanical one that can break 20+ amps of DC or an SSR (The little thumb nail size PCB mount ones on typical DC protect kits are often^H^H^H^H^H usually insufficient). If you build a DIY SSR, it’s a good thing you only need 100V mosfets - the 200 volt are hard to get these days (most are out of stock). Off the shelf DC protect chips will work fine, as will a couple of op amps wired as comparators with an RC at the input, set to trip the relay in 20 milliseconds after applying full rail voltage (a little less than half a cycle at 20 Hz - or faster if extended LF response isn’t required). You can get fancy and add delayed turn-on, thermal sensors to disconnect the load on overheat, or even tie it in with the SOA protection or clip detection circuits. Can’t do any of that with a crowbar.
You might also want to check out the protection circuit in the amp - it is SUPPOSED to stay within the transistors SOA even under shorted conditions, at least for short periods of time before everything gets blazing hot. That’s is, if you’re using MJ15024’s. Fake output trannies will have a hard time with it, though.
The crowbar is useful if the output trannies have already gone short - but any other time it’s more of a nuisance.
You might also want to check out the protection circuit in the amp - it is SUPPOSED to stay within the transistors SOA even under shorted conditions, at least for short periods of time before everything gets blazing hot. That’s is, if you’re using MJ15024’s. Fake output trannies will have a hard time with it, though.
The crowbar is useful if the output trannies have already gone short - but any other time it’s more of a nuisance.
I decided to to put the driver board on the bench and power up with my bipolar bench supply. I monitored the voltage at the speaker output while slowly raising the V on the bench supply. From 0 to +/=15v the speaker follows the plus rail. After 15V the speaker output rests where it is supposed to be at 0v.
No wonder this thing blows up. (by design it appaers). Perhaps it's a characteristic of quasi complimentary design. The transistors on the plus rail dominate until you get to 15V. The design probably depends on the filtering being sufficient such that it never triggers the triac. Hmmmmm.... it's outa there.
No wonder this thing blows up. (by design it appaers). Perhaps it's a characteristic of quasi complimentary design. The transistors on the plus rail dominate until you get to 15V. The design probably depends on the filtering being sufficient such that it never triggers the triac. Hmmmmm.... it's outa there.
“Sticking to the rail till you get sufficient supply voltage” isn’t a function of quasi-comp output. It’s a function of the asymmetrical drive from the op amp to the VAS. The more modern version that uses a push pull VAS and a symmetrical splitter stage to drive it don’t do that. The Phase Linears are even worse about this - since as stock they don’t even have the local feedback to the stage in between the op amp and VAS. Big mistake. The turn on transient can get nasty, and faults downright deadly. And since they don’t have crowbars, they leave a string of blown speaker voice coils in their wake. ALL of these amps need a muting relay.
You could probably delay the VAS CCS until the VCC > 15V so that the VAS becomes active first +last. It might just require a resistor to divide the CCS bias.
A function of a speaker protection is turn-on delay. You may still get a turn-off thud. 15V should not trigger a crowbar since that is only about 30 Watts. Look at the crowbar trigger and maybe it needs adjusting, or maybe the trigger delay caps have dried up.
It would help us if you posted the schematic so we can look at what you are talking about. Sometimes we can look it up and sometimes not and sometimes we just won't bother.
A function of a speaker protection is turn-on delay. You may still get a turn-off thud. 15V should not trigger a crowbar since that is only about 30 Watts. Look at the crowbar trigger and maybe it needs adjusting, or maybe the trigger delay caps have dried up.
It would help us if you posted the schematic so we can look at what you are talking about. Sometimes we can look it up and sometimes not and sometimes we just won't bother.
I thought I had it licked by doubling up the value of the triac filer cap - no, waisted another couple transistors. Here's the schematic.
The idea of a crowbar isnt to destroy the amp but protect it by shorting input power supply and blowing fuses.
If its destroying the amp then something else is wrong somewhere.
If its destroying the amp then something else is wrong somewhere.
If the VI protection works properly then you should be able to short the output without any damage. But another hazard that the VI protection does not help with is instability. If the amp oscillates or is driven hard at 10KHz+, the outputs go into shoot-through current. And a TRIAC will trigger without any gate current from high dV/dt on the main terminals. Perhaps a suitable TRIAC is chosen for that reason and perhaps not. So check the Zobel network R and C.
Looking at this schematic, I see trouble:
1. You have BJT outputs operating on a total of 162V, which is a SOA problem for the best of BJTs. Class G or H is a much better idea if you need that kind of voltage.
2. You have an op-amp IPS, which is a recipe for instability. This is worse than a "Phase linear"; same bad design strategy pushed even further.
What you see here is a complicated circuit that is the result of the search for ways to make the strategy work. It could probably be made to work, but it would take a lot of work by an expert, and probably the latest and greatest transistors. Some people are converting Phase Linears to fully complimentary and that improves the stability issues of the CFP side of a quasi, but faster transistors and redesigning the feedback + compensation is probably more important.
As far as the crowbar goes, I would probably replace it with relay/MOSFET output protection. As I said before, I would fix the startup thump if nothing else.
Looking at this schematic, I see trouble:
1. You have BJT outputs operating on a total of 162V, which is a SOA problem for the best of BJTs. Class G or H is a much better idea if you need that kind of voltage.
2. You have an op-amp IPS, which is a recipe for instability. This is worse than a "Phase linear"; same bad design strategy pushed even further.
What you see here is a complicated circuit that is the result of the search for ways to make the strategy work. It could probably be made to work, but it would take a lot of work by an expert, and probably the latest and greatest transistors. Some people are converting Phase Linears to fully complimentary and that improves the stability issues of the CFP side of a quasi, but faster transistors and redesigning the feedback + compensation is probably more important.
As far as the crowbar goes, I would probably replace it with relay/MOSFET output protection. As I said before, I would fix the startup thump if nothing else.
Alas the reality is crowbars blow up PCB traces, blow themselves up, cause all sorts of grief. The problem is they can dump all the energy in the PSU filter caps very quickly indeed (triacs can turn on on microsecond timescales), which means very high currents and powers (1000's of amps and 10's to 100's of kW are possible I reckon). The fusewire will be an explanding plasma ball during most of this process, protecting nothing as all the action is over before the plasma can expand and quench.
Fuses are there to prevent wiring fires, not protect semiconductors (they are too slow to do that in most situations).
The VI protection should limit the current to something the transistors can handle, much less than the fuse rating, much, much less than the traces can handle. Unfortunately, at +/-81V, the SOA of BJTs is not much and once they have shorted then fuses are next,...
The amplifier worked for 30 plus years though.
And went through likely hundreds/thousands of on/off cycles without triggering the triac.
So assume there is likely cause that needs repair to remove the offset at startup.
Abundance of diodes and semiconductors likely damaged from the original fault.
Or even carbon tracking if anything burned up. Can assume damaged capacitors or diodes in the multi slope VI limiters.
Removing the triac for testing, I'm sure has crossed your mind. I'd personally blow a emotional fuse myself. After blowing new semiconductors.
And went through likely hundreds/thousands of on/off cycles without triggering the triac.
So assume there is likely cause that needs repair to remove the offset at startup.
Abundance of diodes and semiconductors likely damaged from the original fault.
Or even carbon tracking if anything burned up. Can assume damaged capacitors or diodes in the multi slope VI limiters.
Removing the triac for testing, I'm sure has crossed your mind. I'd personally blow a emotional fuse myself. After blowing new semiconductors.
For AC, 50% duty cycle signals. Shorting DC might well be outside the SOA.
And actually this design is quite a bit more stable than the Phase Linears. When I rebuild a PL I incorporate some of the features and they give me less trouble. You want start up offsets? Try a stock PL series 2 with an op amp that has an offset that sends its output to the + rail instead of the - when open loop. It has more trouble searching for the initial Q point. You need to select the op amp to go negative open loop to minimize the thump.
VI protection?
I see no current sensing for the sum of the transistor currents by means of a low ohm resistor so I assume there is none.
There is a V limiter up front.
I see no current sensing for the sum of the transistor currents by means of a low ohm resistor so I assume there is none.
There is a V limiter up front.
Not sure if this has been stated yet, the triac is there to protect your speakers from DC. If C1 charges to the Diac (Q2) trigger voltage of about 30v (because of DC on the output) it will avalanche ON through the Triac Q1 gate and snap it on quick time. Probably destroying the output transistors. So prolonged output DC is the issue. It will do this even when no speaker is connected. Really is a crazy design but I guess we now have power mosfets to do the job.
R31 provides a voltage component to the VI limit but it is complicated by R23 and R33, CR16 and CR17. Capacitors C18 and C20 filter the voltage component for stability during limiting. Dried up C18 or C20 will make the limiting unstable because the voltage component is positive feedback, leading to oscillations and OP failure.
Only one of the OPs is current sampled (R8, R9) assuming the emitter resistors balance the current so that all OPs conduct the same, i.e. the total current is not summed.
I am tempted to simulate this circuit VI limiting in spice but sorry, I'm not that interested. If I owned the amp myself, I definitely would. The voltage component is additive so failed the C18, C20 will allow much higher current sooner than if the AC component was attenuated by C18,C20. Even if C18, C20 are healthy, the voltage component moves the DC current limit higher when you have a high DC voltage output. A simulation would be very interesting.
So, it may be worth replacing C18 and C20. Caps like these, large value + low voltage, are perhaps the most prone to drying out and that could be the source of your problems. Of course, you want to check Q7 and Q8, and all the associated diodes.
Only one of the OPs is current sampled (R8, R9) assuming the emitter resistors balance the current so that all OPs conduct the same, i.e. the total current is not summed.
I am tempted to simulate this circuit VI limiting in spice but sorry, I'm not that interested. If I owned the amp myself, I definitely would. The voltage component is additive so failed the C18, C20 will allow much higher current sooner than if the AC component was attenuated by C18,C20. Even if C18, C20 are healthy, the voltage component moves the DC current limit higher when you have a high DC voltage output. A simulation would be very interesting.
So, it may be worth replacing C18 and C20. Caps like these, large value + low voltage, are perhaps the most prone to drying out and that could be the source of your problems. Of course, you want to check Q7 and Q8, and all the associated diodes.