The dissipation of Q12 is a bit high when it delivers 30 mA - 11 mA of base current to the saturated Q8, if that's what you mean.
I usually run VAS at 7ma and use 2n5551 instead of BD139. 2N5551 is pretty robust, near 350mW in side case without problem.
Is the BD139 the one blew up? If so, it might be its inferior voltage rating.
Is the BD139 the one blew up? If so, it might be its inferior voltage rating.
Yes! Q12 will dissipate about 800mW at clipping but is only rated for 625mW. I haven’t desoldered it yet, but it’s definitely the VAS area that’s blown up.
I originally had a 47 ohm resistor at the emitter of Q13, which would have limited the dissipation at Q12 to about 155mW, but I lowered it when chasing THD without recalculating the emitter current through Q12.
You live and learn! 😀
Seems like I was wrong. Both Q8 and Q12 are good. Those were the obvious candidates to fail on heavy clipping...
It's the drivers that are shot. At clipping, they will see the full rail-to-tail voltage, which looks like it's about 75V. That's technically within spec, but I guess when things are going a bit wild at clipping, there could be some overshoots. They both measure about 20ohm C-to-E. And R23 is pretty charred, so that would make sense.
Any recommendation for a higher voltage driver?
Any recommendation for a higher voltage driver?
During clipping, both Q8 and Q16 will saturate. The increased VAS current will charge C6, leading to increased bias in the driver and output transistors. That explains the blown drivers.
You need to prevent Q8 and Q16 from saturating.
ETA: Problem is really Q16 saturating.
Ed
You need to prevent Q8 and Q16 from saturating.
ETA: Problem is really Q16 saturating.
Ed
Last edited:
C6 value doesn't need to be that high, does it?
220nF should be enough. Not sure if this would mitigate the saturation?
220nF should be enough. Not sure if this would mitigate the saturation?
How so? When Q16 saturates, it will put B+ on the base of Q13, which will put rail-to-rail voltage (worst case) across CE on Q13. With the 37V rails I'm running, I'd be cutting it close to the max voltage for the drivers, but not exceeding it. Here's what it looks like in the sim. (Sorry for the designators in SPICE not matching the KiCAD schematic)
Okay, I see that Q16 comes out of saturation quickly enough not to affect C6. The bias is actually collapsing during clipping, which is a different problem.
I generally avoid saturation and long time constants.
Ed
I generally avoid saturation and long time constants.
Ed
I can certainly reduce C6 a bit. However, it turns out the problem wasn’t the drivers, but the IRFP9240 that had given up the ghost. It was shorted, which made the drivers look shorted when I measured them in circuit. I’m a bit bewildered over this… D4 had a bit of a questionable solder joint, but that should have affected the top MOSFET, not the bottom one. I guess it’s possible that the large C6 somehow gets the bias locked up at a too high level when Q16 saturates. I can certainly try a smaller cap.
D3-D6 will behave in similar way like speakers protecting fuses 🙂
I think usually mosfets get destroyed before these diodes kick in...
I think usually mosfets get destroyed before these diodes kick in...
They’re from Mouser. It’s a bit strange. I have to measure, but I think my rails are about 35V. Even at full clipping, the dissipation should be below 150W. Granted, my heatsinks aren’t huge, but it failed almost immediately when I cranked up the volume and I would have expected it to take longer for sharp drum beats to heat up the die enough that it fails.
What I might do is to hook it up to a lab supply with maybe 20V rails and a bit of current limiting and carefully drive it into clipping and check the waveforms around the VAS, drivers and MOSFETs. Maybe I’m getting some weird oscillation that throws things out of whack. I don’t see how that MOSFET should have failed from just the music power.
What I might do is to hook it up to a lab supply with maybe 20V rails and a bit of current limiting and carefully drive it into clipping and check the waveforms around the VAS, drivers and MOSFETs. Maybe I’m getting some weird oscillation that throws things out of whack. I don’t see how that MOSFET should have failed from just the music power.
Are the protection diodes ok? Might be opened up during the fault and cause the mosfet blowing up.
PS: you mentioned R23 is burned. That means there are significant voltages over it, the protection diodes might be damaged also.
PS: you mentioned R23 is burned. That means there are significant voltages over it, the protection diodes might be damaged also.
Of course, the bias might have jumped for some reason creating a dead short. I’m not sure I want to blame C6, thought. I jacked that up to 1000uF in SPICE without any effect on the bias. If anything, a large cap there should help keeping the bias stable.
Maybe investigate ringing, overshoot or oscillation of the drivers, allowing high AC current to shoot through C8, especially when clipping introduces high frequencies.
There's no base stopper resistors, and I'm eyeing the 47p capacitors, too.
There's no base stopper resistors, and I'm eyeing the 47p capacitors, too.
Probably not an issue but does the speaker have any large electrical phase angles? For example a 45° phase angle will double the dissipation. It's something I've been reading about so of course it's the first thing that comes to my mind.
The base current is limited by the current source. What do you have against the base capacitors?