What puzzles me the most is that these amps ran fine for months! then one day blammo! Maybe they have been running right on the edge and one the day of death the AC line voltage was just a bit high? I have been setting the variac right at 120V ac and maybe that's the issue?
It's either that or my other wild guess theory is that one of the caps in the ZL CLC power supply is funky and that is causing a spike at switch flip turn on? whatever the case is I am NOT turning on the other ZL with the good 040 mosfets until i have a 100% rock solid solution!
I emailed Nelson and asked (almost begged) him to look and see if by chance he has one more 040 left laying around. In the meantime I have another 20x 044's on the way...
BTW, what is the difference between the 044 and the 044N version? one has a lot lower input capacitance.
It's either that or my other wild guess theory is that one of the caps in the ZL CLC power supply is funky and that is causing a spike at switch flip turn on? whatever the case is I am NOT turning on the other ZL with the good 040 mosfets until i have a 100% rock solid solution!
I emailed Nelson and asked (almost begged) him to look and see if by chance he has one more 040 left laying around. In the meantime I have another 20x 044's on the way...
BTW, what is the difference between the 044 and the 044N version? one has a lot lower input capacitance.
Sounds like you might have been on the right track to begin with, a single moderately sized resistor on the primary with a timed relay would probably kill this issue with few other issues remaining to be resolved.
I've not experienced this sort of mosfet breakdown phenomena so far, but will keep it in mind for the day I potentially do.
If this is indeed the mechanism (and it seems possible) then the idea I suggested would tend to make it even worse..
I've not experienced this sort of mosfet breakdown phenomena so far, but will keep it in mind for the day I potentially do.
If this is indeed the mechanism (and it seems possible) then the idea I suggested would tend to make it even worse..
This is one of the things I alluded to in an earlier post, what if anything has changed?
- Cold resistance of varistors lower than than when new?
- New lower resistance line cords? (Lower IR drop during inrush?)
- Plugging into a different outlet closer to main panel (Lower IR drop during inrush?)
- New pole transformer, tap change? Line voltage marginally higher than before?
- Changed power sequencing (did you turn them on together in the past causing more line sag and a slower dvdt at power up?)
Hoping NP has a few more of the original fets so that when you have this resolved you can use the originals which presumably perform better..
I'd stick the digital scope on the supply and try to figure out what is happening during the initial inrush phase.. Might not be a bad idea to also look at the gate so that you can determine the gate to drain voltage differential over time.
For MOS devices, breakdown voltage of drain to gate is the same as breakdown voltage of source to gate. Gate oxide thickness is same over drain, source and channel (substrate).
IRFP044N appears to be attempt to duplicate IRFP044 using newer processing techniques that results in smaller geometries to achieve same Ids and Vto. Thus smaller capacitance is due to smaller gate area with perhaps same gate oxide thickness.
IRFP4321 is an even newer technology achieving some improvements over 044 for Ids. I would theorize that gate oxide thickness is 1.5 x of 044. Gate oxide area would also be 1.5 x, resulting in input capacitance ~ 044 input capacitance. But the big reward is gate to source and gate to drain breakdown voltage is increased from 20v to 30v minimum. This is the device I would be using. Thicker gate oxide also increases Vto about 1 volt, from 2v - 4v to 3v - 5v. You will need to change R1 and R2 accordingly.
Looks like a IRFP4321 also costs twice as much as a IRFP044N.
IRFP044N appears to be attempt to duplicate IRFP044 using newer processing techniques that results in smaller geometries to achieve same Ids and Vto. Thus smaller capacitance is due to smaller gate area with perhaps same gate oxide thickness.
IRFP4321 is an even newer technology achieving some improvements over 044 for Ids. I would theorize that gate oxide thickness is 1.5 x of 044. Gate oxide area would also be 1.5 x, resulting in input capacitance ~ 044 input capacitance. But the big reward is gate to source and gate to drain breakdown voltage is increased from 20v to 30v minimum. This is the device I would be using. Thicker gate oxide also increases Vto about 1 volt, from 2v - 4v to 3v - 5v. You will need to change R1 and R2 accordingly.
Looks like a IRFP4321 also costs twice as much as a IRFP044N.
Compared to IRFP040, IRFP4321 has higher Cgs but much lower Cgd.
Cgs is specified at 0v, but in actual circuit the voltage is ~ 4v. This reduces capacitance.
More importantly, Cgd is less. Cgd x circuit gain = miller capacitance. Therefore, miller capacitance is less with IRFP4321. Total input capacitance is Cgs plus miller capacitance.
Besides, what does it matter? Your preamplifier is driving this input capacitance through 348 ohms.
You do the math. I'm getting confused.
Cgs is specified at 0v, but in actual circuit the voltage is ~ 4v. This reduces capacitance.
More importantly, Cgd is less. Cgd x circuit gain = miller capacitance. Therefore, miller capacitance is less with IRFP4321. Total input capacitance is Cgs plus miller capacitance.
Besides, what does it matter? Your preamplifier is driving this input capacitance through 348 ohms.
You do the math. I'm getting confused.
AHHHHHH HAAAAA we have answers! more than expected actually! TADAAAA 
So my new tube of mosfets arrived today! I bought some NOS 044's and some 044N's First I used an 044N and placed a 15v zener from drain to gate and flipped the switch using the ZL power supply connected to the Bench Zen. and after a dozen switch flips, not a single mosfet failure. So to prove concept I disconnected the Zener and flipped the switch and BLAMMO dead mosfet instantly just as expected!
Then I connected a 044 and repeated the test and zener in, no failures. zener out....also no failures???? 2 dozen switch flips and not a single failure! So the 044's are MUCH tougher than the 044N's!
No combination of input or output shorts or power up cold, hot, supply drained, half drained etc etc etc produced a dead 044 mosfet! but if you blink wrong a 044N will blow up!
I can't leave the zener connected during operation as it definitely interferes with operation. and if I am going to redo the output relay circuit anyway. I can just add another relay that is tied to the speaker relay circuit that connects the zeners during start up and then after a few seconds disconnects the zeners as the amp is coming out of mute! this way the precious original 040 mosfets will be protected! No slow start needed, although it may still be nice to build a simpler version that allows remote start.
So I think I have the problem figured out and solved!
A big thank you to JamesBrennan4 for the D-G zener suggestion and generally for all the cool mosfet info! also a big thank you to everyone else for all the input!
Zc
So my new tube of mosfets arrived today! I bought some NOS 044's and some 044N's First I used an 044N and placed a 15v zener from drain to gate and flipped the switch using the ZL power supply connected to the Bench Zen. and after a dozen switch flips, not a single mosfet failure. So to prove concept I disconnected the Zener and flipped the switch and BLAMMO dead mosfet instantly just as expected!
Then I connected a 044 and repeated the test and zener in, no failures. zener out....also no failures????
2 dozen switch flips and not a single failure! So the 044's are MUCH tougher than the 044N's! No combination of input or output shorts or power up cold, hot, supply drained, half drained etc etc etc produced a dead 044 mosfet! but if you blink wrong a 044N will blow up!
I can't leave the zener connected during operation as it definitely interferes with operation. and if I am going to redo the output relay circuit anyway. I can just add another relay that is tied to the speaker relay circuit that connects the zeners during start up and then after a few seconds disconnects the zeners as the amp is coming out of mute! this way the precious original 040 mosfets will be protected! No slow start needed, although it may still be nice to build a simpler version that allows remote start.
So I think I have the problem figured out and solved!
A big thank you to JamesBrennan4 for the D-G zener suggestion and generally for all the cool mosfet info! also a big thank you to everyone else for all the input!
Zc
Last edited:
I think you've got it.
My suggestion of a 27v zener was based on belief that at 0v input, gate is at +4v and drain is at +12v. Zener would not conduct. With a negative input of -4v, gate would be at 0v and drain would go to +24v. Zener would just start to conduct.
At power-on, power supply starts at +55v, zener would force gate up to +28v, which turns on device hard which then reduces drain voltage.
Initial gate-drain stress would be 27v, gate-source stress would be 28v.
I would also predict that a 040 would be the best for surviving the voltage stress (I think the gate oxide thickness for the oldest device would also be the thickest as the way to reduce cost is to make area smaller for more devices per wafer, but smaller area requires thinner gate oxide).
Glad to help. Have fun.
My suggestion of a 27v zener was based on belief that at 0v input, gate is at +4v and drain is at +12v. Zener would not conduct. With a negative input of -4v, gate would be at 0v and drain would go to +24v. Zener would just start to conduct.
At power-on, power supply starts at +55v, zener would force gate up to +28v, which turns on device hard which then reduces drain voltage.
Initial gate-drain stress would be 27v, gate-source stress would be 28v.
I would also predict that a 040 would be the best for surviving the voltage stress (I think the gate oxide thickness for the oldest device would also be the thickest as the way to reduce cost is to make area smaller for more devices per wafer, but smaller area requires thinner gate oxide).
Glad to help. Have fun.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.