Well, my work in the audio amplifier field so far has left me devastated.
My latest design was working happily today until when I went to bias it properly, it exploded in my face. Blowing all eight of the output transistors, which are mj21193, and 21194. And it blew the drivers. I have to find out 'why' it did this.
I'm just so disappointed after I spent months working on this and it sounded so good.
Maybe I'll post a schematic soon and see if anyone can help me with the autopsy.
Anyone here who's had this happen?
My latest design was working happily today until when I went to bias it properly, it exploded in my face. Blowing all eight of the output transistors, which are mj21193, and 21194. And it blew the drivers. I have to find out 'why' it did this.
I'm just so disappointed after I spent months working on this and it sounded so good.
Maybe I'll post a schematic soon and see if anyone can help me with the autopsy.
Anyone here who's had this happen?
It's just awful.
It's not so much the value of the parts, but the amount of work that went into building it.
It's not so much the value of the parts, but the amount of work that went into building it.
I've had it happen to me, not entirely sure why. I think under one circumstance I was using a compound configuration in the output and just turning up the bias sent it into oscillation, something sort of easy to do with that circuit. Aother time I think I remember using a metal bladed screwdriver to adjust the pot which maybe sparked something, but also maybe the output was unstable anyway and just waiting to be turned on. These things happened with first time experimental circuits too, and I've long since used plastic tuning tools.
Since your circuit was already up and working, it's a head scratcher to me. Maybe you brushed up against something unintentionaly. If you used a metal blade to adjust, maybe ESD got you somehow. I have heard of amplifiers with certain output configurations behaving ok until one day, they blow, perhaps due to transistor(s) properties altering over time and thermal cycling. I'd be interested in knowing more myself.
Since your circuit was already up and working, it's a head scratcher to me. Maybe you brushed up against something unintentionaly. If you used a metal blade to adjust, maybe ESD got you somehow. I have heard of amplifiers with certain output configurations behaving ok until one day, they blow, perhaps due to transistor(s) properties altering over time and thermal cycling. I'd be interested in knowing more myself.
So you accidentally cranked the bias too high and a part blew
and took the rest of them with it. Happens to all of us.
and took the rest of them with it. Happens to all of us.
Well, this was by adjusting a pot with my bare hands. So no metal in there.
It has been working properly.
But I think the reason it blew was because of a few possibilities.
One of the most probable was that I had the wires from the board to the output transistors too close and I think two of them shorted because there was an amazing spark there. It had some real current behind it and was very loud.
Next, I know now that there is an inherent current sharing problem in my topology. I didn't implement the CFP properly. The way I had it caused feedback only from one of the transistors on each rail, instead of all of them.
And next on the list. My LTP didn't have the most matched transistors around, so the amplifier had a small offset of almost half a volt. I think that might be a real problem when rising the bias current don't you?
Anyway, I'm completely re-designing it with more output transistors (10 per channel instead of 8) and more stability, lower distortion, and power management for extra safety and clean power up and down.
It has been working properly.
But I think the reason it blew was because of a few possibilities.
One of the most probable was that I had the wires from the board to the output transistors too close and I think two of them shorted because there was an amazing spark there. It had some real current behind it and was very loud.
Next, I know now that there is an inherent current sharing problem in my topology. I didn't implement the CFP properly. The way I had it caused feedback only from one of the transistors on each rail, instead of all of them.
And next on the list. My LTP didn't have the most matched transistors around, so the amplifier had a small offset of almost half a volt. I think that might be a real problem when rising the bias current don't you?
Anyway, I'm completely re-designing it with more output transistors (10 per channel instead of 8) and more stability, lower distortion, and power management for extra safety and clean power up and down.
Duo what a shame but Its good to see that its not just the amateurs and complete beginners like me, that have problems from time to time. Nelson thanks for admitting that the professionals do sometimes as well.
Well, I do believe I've found out how my amplifier fried.
After some serious autopsy, and then a look at the schematic, I've realized that my Vbe multiplier bias circuit was not implemented correctly.
By turning the pot enough, you can easily cause the multiplier to lose bias and open wide. This in turn causes the output transistors to conduct as hard as they can as the current source in my VAS causes one rail to be pulled, and the VAS itself pulls the other.
After some serious autopsy, and then a look at the schematic, I've realized that my Vbe multiplier bias circuit was not implemented correctly.
By turning the pot enough, you can easily cause the multiplier to lose bias and open wide. This in turn causes the output transistors to conduct as hard as they can as the current source in my VAS causes one rail to be pulled, and the VAS itself pulls the other.
one thing I find useful in experimenting with designs is to put a power resistor in serial with the rail(s). it acts like a current limiter. it worked like a charm for me.
I don't think it is that bad when I blow up the amp the first time. It's the second time that gets me.😀
Whenever I work on an amp, my (4 year old) son always tells me, "Dad, if you're going to blow something up, call me in first!" He's tired of just hearing a bang in the other room and missing all the fun. 😀
Sorry, what do you want to do? It happens.
Better if does to you than to me! 😉 🙂
What sucks even more, I think is to get through a whole project and realize the thing hums because of layout and/or PCB or that it sounds like crap. I think you were happy with the result so you just have to fix it.
Better if does to you than to me! 😉 🙂
What sucks even more, I think is to get through a whole project and realize the thing hums because of layout and/or PCB or that it sounds like crap. I think you were happy with the result so you just have to fix it.
Amp biasing
When I used to make class AB amps, I always biased them with a 100W light bulb in series with the AC line. Saved my butt many times. Now that I work almost exclusively with Class-D, I still use the bulb on initial start up.
When I used to make class AB amps, I always biased them with a 100W light bulb in series with the AC line. Saved my butt many times. Now that I work almost exclusively with Class-D, I still use the bulb on initial start up.
Duo said:
Duo very sorry for your loss...
The resistor trick mentioned works well because it limits the current from the caps to the OP devices. Something of biblical proportions must have happened for you to fry 8 MJ21193/94's.
Too late now but maybe next time you could "experiment" with only 2 pairs in circuit while you play with bias etc.
Lastly I have some spare MJ21195/96's if you are interested. Might get slightly better luck there... 😉
K-
Re: Amp biasing
Thats THE best insurance in terms of "why people don't do it beats me".
For larger Amps I also go to a 500 watt lamp, If memory serves, while trouble shooting oscillations in one Class-G amp, it would not oscillate till the rails were within 1 volt of spec.... and the 100 watt bulb would drop too many volts.
The disadvantage however is that because the bulb is on the mains, there is still enough juice in the caps to blow the devices.
K-
dmfraser said:
Thats THE best insurance in terms of "why people don't do it beats me".
For larger Amps I also go to a 500 watt lamp, If memory serves, while trouble shooting oscillations in one Class-G amp, it would not oscillate till the rails were within 1 volt of spec.... and the 100 watt bulb would drop too many volts.
The disadvantage however is that because the bulb is on the mains, there is still enough juice in the caps to blow the devices.
K-
I was lucky and got a surplus B&O vario transformer for US$ 40 when I was an apprentice. It’s a great help.
\Jens
\Jens
Duo said:
sometimes a low Z amp stage can produce a huge current spike which can melt the device inside in micro sec.
some suggestions.
---maybe the voltage swing caused one of the outputs to exceed its breakdown voltage. particularly in a no load situation.
---try using a current source to supply the bias current for the outputs.
---set your bias adj. to be more sensitive and limit to around 100mA max or less. (probably much less). this makes it so that both outputs cannot be turned on at the same time more than 100mA.
---Imparative that you make sure there are no power supply spikes that would cause the rails to increase and exceed Vce breakdown
just think how much current can flow if one device is on and the other suddenly shorted???
Well, I have discovered why the amplifier blew up so well.
My Vbe multiplier had something I missed. The way it was implemented, if I turned the pot a certain amount, the Vbe transistor lost bias, allowing full current to flow right across the output stage as the amplifier shorted out.
The voltage breakdown of the transistors was not reached in this situation, just too much current flowed.
My power supply was far more than enough to fry all those transistors.
Either way, I'm changing the design so almost no matter what happens to the bias pot, it will not make the amp short out. And it'll be limited so that the output stage will only go up to a certain current so no accidents happen.
My Vbe multiplier had something I missed. The way it was implemented, if I turned the pot a certain amount, the Vbe transistor lost bias, allowing full current to flow right across the output stage as the amplifier shorted out.
The voltage breakdown of the transistors was not reached in this situation, just too much current flowed.
My power supply was far more than enough to fry all those transistors.
Either way, I'm changing the design so almost no matter what happens to the bias pot, it will not make the amp short out. And it'll be limited so that the output stage will only go up to a certain current so no accidents happen.
- Status
- Not open for further replies.