Those, like most of them, show constant-power curves. It is an extrapolation based on a thermal model only - it just shows when you reach maximum Tj for a given V, I and time. S/B limits are significantly inside of this. S/B is voltage dependent and happens long before the maximum Tj is reached. Well, what happens is it goes up to say 300C in some tiny spot on the die rather than spread the heat/current out like the thermal model assumes that it does. Device physics is a bitch.
It was well recognized in bipolars from the start because germanium was SO BAD about this. Silicon got better, and modern audio devices better still. All lateral MOS and most early vertical MOS was better still, handling full at higher voltages. Leading many to believe that MOS tech is completely immune to it. It is not and people found out the hard way. And the “better” they make modern MOSFETs and IGBTs for the intended switching applications, the worse the effect is becoming in each generation of devices. Some data sheets actually show the truth - and many only handle full “power“ to say 10 or 20 volts, falling like a stone above that.
It was well recognized in bipolars from the start because germanium was SO BAD about this. Silicon got better, and modern audio devices better still. All lateral MOS and most early vertical MOS was better still, handling full at higher voltages. Leading many to believe that MOS tech is completely immune to it. It is not and people found out the hard way. And the “better” they make modern MOSFETs and IGBTs for the intended switching applications, the worse the effect is becoming in each generation of devices. Some data sheets actually show the truth - and many only handle full “power“ to say 10 or 20 volts, falling like a stone above that.
A typical SOA is definitely not constant power - check it out.
It really is the data you need to stay out of 2ndBD trouble.
But you are right, MOS devices are not immune to it, due to the Spirito-effect.
I did link to a NASA report about it earlier.
Jan
It really is the data you need to stay out of 2ndBD trouble.
But you are right, MOS devices are not immune to it, due to the Spirito-effect.
I did link to a NASA report about it earlier.
Jan
While I appreciate all the inputs and suggestions on the possible causes of failure and selections of MOSFET, I have been listening to the amp without the MOSFET. I simply rearranged the filter circuit to CRCRC with 220uF/82 ohm/220uF/37.5 ohm/220uF. As I said in my original post, speaker output noise level was 0.5 mV with the filter and it is not audible with my 92 dB/watt speaker.
So do I even need to replace the MOSFET? I am sure the circuit can reduce hum to an even lower level. But will it make the amp sound better? If not, I am happy with the CRCRC arrangement.
So do I even need to replace the MOSFET? I am sure the circuit can reduce hum to an even lower level. But will it make the amp sound better? If not, I am happy with the CRCRC arrangement.
The SOA in Post #37 does not include a "DC" operating point. This almost always means it's designed to be a switch and will fail as a linear device no matter how thermally limited it is (within reason).
Compare to the IXYS IXTH10N100D2: There is clearly an adjustment for secondary breakdown, no?
Personally, I wouldn't go over 300V B+ @ 500mA with this = 150W on a device rated for 695W. Seems legit? The heatsink would be the size of a toaster though.
Compare to the IXYS IXTH10N100D2: There is clearly an adjustment for secondary breakdown, no?
Personally, I wouldn't go over 300V B+ @ 500mA with this = 150W on a device rated for 695W. Seems legit? The heatsink would be the size of a toaster though.
I have been talking to IXYS guys about using their high-voltage (>3kV) devices for linear operation.
They wouldn't recommend even 10mA linear operation at that voltage.
Jan
They wouldn't recommend even 10mA linear operation at that voltage.
Jan
Probably better to do a MOScode/MOStode/MOScade (think cascode but MOSFETs) of some sort with 1kV parts?
Yeah, I've noticed that from EXICON's Lateral MOSFET. But testing them by ourselves requires a lot of expensive instrument and the test itself could be dangerous.
Models from IXYS and Toshiba appear to have another transition in SOA diagram over higher Vds. It seems safer choosing parts from them.
In a linear power supply, you just load test it under all probable operating conditions. Full load, possible fault, startup, etc. Just use a load you don’t care about too much - like sand box resistors instead of tubes. If it dies you need a different type. Don’t try to find the ones with the lowest possible on resistance - those are the ones that will give you grief as linear amplifiers. When you find one that holds up buy a tube of 100 before they go EOL.
This is an excellent and very informative thread! I had a similar issue with a cap multiplier and now I have a much better idea why was it killing mosfets like there was no tomorrow.
I’ve used 3 x 230 V halogene light bulbs (50 Watt each) in series for testing my 600 VDC power supply with linear mosfets.
Works great!
Regards, Gerrit
Works great!
Regards, Gerrit
Not trying to convert anybody, just stating personal preferences:
1) simpler is better
2) have as much OCD as anybody around here, but not wasted on über-regulating everything within a micro-Volt or so.
1 and 2 are symbiotic, one helps achieve the other and viceversa.
1) simpler is better
2) have as much OCD as anybody around here, but not wasted on über-regulating everything within a micro-Volt or so.
1 and 2 are symbiotic, one helps achieve the other and viceversa.
That’s why I use a single “big” IXYS MOSFET (mentioned earlier) and forget about it, because it has worked properly ever since.
Regards, Gerrit
Regards, Gerrit
Simpler is better.
Solid state and quite a few components is simple?
Brute Force CLC or LCRC is as simple as it gets.
Designed with the proper parts, the regulation % is almost as good as your power company.
If the mains power is not very well regulated, then perhaps your power company will fix that, or put you on a better feed.
If your power company does not fix the problem, then the solution is no longer simple, you need to move to a better location.
Solid state and quite a few components is simple?
Brute Force CLC or LCRC is as simple as it gets.
Designed with the proper parts, the regulation % is almost as good as your power company.
If the mains power is not very well regulated, then perhaps your power company will fix that, or put you on a better feed.
If your power company does not fix the problem, then the solution is no longer simple, you need to move to a better location.
Last edited:
L is large and expensive, Mosfet is cheap.
Regulator can be quiet with smaller capacitor.
Mosfet can be dialed into the voltage you want (and sometimes you need two or more different V’s) and still maintain regulation (resistor can’t). So can a tube, but not as easily or simply (Need that blasted isolated elevated heater).
SS regulator is often the right answer, even for simplicity. Just takes a little experience to make one that wont blow up.
Regulator can be quiet with smaller capacitor.
Mosfet can be dialed into the voltage you want (and sometimes you need two or more different V’s) and still maintain regulation (resistor can’t). So can a tube, but not as easily or simply (Need that blasted isolated elevated heater).
SS regulator is often the right answer, even for simplicity. Just takes a little experience to make one that wont blow up.
My T-reg is simple (well, maybe not) and ot expensive and has 0.1R Zout and noise that hard to measure.
0 - 500V, max 400mA, fully overload protected. The pass MOSFET is 2 bucks or so.
https://diyaudiostore.com/collections/power-supply-kits/products/linear-audio-t-reg
Jan
0 - 500V, max 400mA, fully overload protected. The pass MOSFET is 2 bucks or so.
https://diyaudiostore.com/collections/power-supply-kits/products/linear-audio-t-reg
Jan
Never had a blowout of my Brute Force real world filters.
Simple. Works
No heat sink, no Silicone grease or composite rubber-like insulator for the T0-220 part (external heat sinks are more efficient than internal heat sinks, so the TO-220 has to be insulated). An internal heat sink only serves to heat up the rest of the internal parts.
Use enough capacitance, and put the Telarc's 1812 overture recording of the full scale 6Hz Cannon. Take a clip of that and do an endless repeat.
Yes, the voltage will finally droop, if your amplifier output stages do not break first.
(Who ever listens to music that way, this is a Hi Fi amplifier, it is not a servo amplifier for a shake platform to find the breaking point of an equipment under test).
Simple. Works
I am not talking about the overloaded guitar amplifier with the bass player breaking all 4 strings one after the other.
Is anyone worried if the sound changes as the B+ droops?
My amplifier designs require the B+ to be relatively steady under real world changing dynamic signal conditions.
But they do not perform badly just because the power mains varies and B+ voltage varies by +/- 2.5%.
Other design solutions do work.
For any engineering problem, there are at least 100 solutions; of which at least 3 will actually work reliably . . .
But only if they are implemented properly, and they use really good parts.
I guess the original poster's problem was caused because it was not one of the 3 reliable solutions.
Just my opinions
Simple. Works
No heat sink, no Silicone grease or composite rubber-like insulator for the T0-220 part (external heat sinks are more efficient than internal heat sinks, so the TO-220 has to be insulated). An internal heat sink only serves to heat up the rest of the internal parts.
Use enough capacitance, and put the Telarc's 1812 overture recording of the full scale 6Hz Cannon. Take a clip of that and do an endless repeat.
Yes, the voltage will finally droop, if your amplifier output stages do not break first.
(Who ever listens to music that way, this is a Hi Fi amplifier, it is not a servo amplifier for a shake platform to find the breaking point of an equipment under test).
Simple. Works
I am not talking about the overloaded guitar amplifier with the bass player breaking all 4 strings one after the other.
Is anyone worried if the sound changes as the B+ droops?
My amplifier designs require the B+ to be relatively steady under real world changing dynamic signal conditions.
But they do not perform badly just because the power mains varies and B+ voltage varies by +/- 2.5%.
Other design solutions do work.
For any engineering problem, there are at least 100 solutions; of which at least 3 will actually work reliably . . .
But only if they are implemented properly, and they use really good parts.
I guess the original poster's problem was caused because it was not one of the 3 reliable solutions.
Just my opinions