Long live the MLJ21195 and MJL4281. They will probably be the last two survivors of the final E.L.E. (before whole world of linear amplifiers goes dark).
They also don’t make good audio amplifiers. They are not free from secondary breakdown type of effects. You may be able to run one with say 100 watts of average power dissipation under RF drive, but put that much DC into it and KABANG, even mounted on a cold plate. High frequency devices are only good for that - high frequency.
DamnI know that datasheet SOAR curves from the late 1990's are completely unreliable, because the MOSFET manufacturers were not aware of the issue yet. Anything with pulses up to 10 ms and no longer is indeed also unusable, as 10 ms is not >> the thermal time constants.
Anyway, I read the link, and while reading it remembered I had read it before.
In my opinion, those NASA people misinterpret the SOAR curves and use poor design practices. You shouldn't design your circuit to keep VGS constant, but rather to keep ID constant as implied in the SOAR curves. You can then still run into thermal runaway issues, but only when temperature differences inside the silicon itself become large enough to cause a too high loop gain, the heating up of the entire die then has nothing to do with it anymore.
Basically what those NASA people are doing with their constant VGS and VDS is like using a power BJT, biasing it at a constant VBE without any emitter resistor or anything else to stabilize the collector current and then complain that it is thermally unstable.
Regarding the issue of manufacturers changing their products at random moments without updating the part number: changing AEC-Q101 qualified components is not allowed without prior approval from the customer, but I guess that only applies when you buy millions of them straight from the manufacturer.
Well, at least we still have valves.
Last edited:
It seems like 2SC5200/2SA1943 are still available for now.
I also wonder if any of the Chinese versions can be trusted...
https://datasheet.lcsc.com/lcsc/2001071233_SPTECH-2SC5200_C480027.pdf
https://datasheet.lcsc.com/lcsc/2001071232_SPTECH-2SA1943_C480004.pdf
https://datasheet.lcsc.com/lcsc/2001071234_SPTECH-MJE15032_C480200.pdf
https://datasheet.lcsc.com/lcsc/2001071234_SPTECH-MJE15033_C480201.pdf
I also wonder if any of the Chinese versions can be trusted...
https://datasheet.lcsc.com/lcsc/2001071233_SPTECH-2SC5200_C480027.pdf
https://datasheet.lcsc.com/lcsc/2001071232_SPTECH-2SA1943_C480004.pdf
https://datasheet.lcsc.com/lcsc/2001071234_SPTECH-MJE15032_C480200.pdf
https://datasheet.lcsc.com/lcsc/2001071234_SPTECH-MJE15033_C480201.pdf
Only works if the manufacturer actually notifies the customer about the change in behaviour. I see very few datasheets with the Spirito slopes, but nobody has made general purpose single cell parts like the original Siliconix VMOS for about 30 years. This is a very real weakness and has got worse since the original NASA papers came out, as a couple of process shrinks have happened since then. This has been a dirty secret for a very long time
ISC copies are not as good. For now the Toshiba versions are still available - although the original 2S versions are drying up. The TTA/TTC versions replace them. And they too have undergone a die shrink. They still advertise the same SOA, which may actually be *true*. In the case of bipolars, smaller dies do handle a larger fraction of their thermal limits to higher voltage.
My guess is the Fairchild versions will outlive the Toshiba as far as product life cycle goes. My guess is that Toshiba will give up on them first. But they will eventually go away altogether, considering that the MJL4281 can be used to replace them, as well as almost any other “fast” type. The whole ON Semi line has been slowly consolidating over the years.
I think that you are correct, different behaviour at constant power, but different Id/Vds.
Interesting, a big variation in temperature coefficient. This is something that has been noticed in choosing parts for designs based on the SKA amplifier
I have spent an hour today searching for datasheets that do show Spirito in the SOAR and they have all vanished from the Internet with manufacturer consolidation in the last few years. I know the parts have actually got more fragile, so they are hiding it.
There is no way to prevent this, distributed source resistors would not balance current at lower Vgs/Id
Interesting, a big variation in temperature coefficient. This is something that has been noticed in choosing parts for designs based on the SKA amplifier
I have spent an hour today searching for datasheets that do show Spirito in the SOAR and they have all vanished from the Internet with manufacturer consolidation in the last few years. I know the parts have actually got more fragile, so they are hiding it.
There is no way to prevent this, distributed source resistors would not balance current at lower Vgs/Id
Last edited:
Well I've tried to answer the question with a measurement, and it looks like the answer is "no it doesn't show up on this measurement". The little FET I've tested is a TrenchFET, so there is no way it will work in linear mode... of course the datasheet shows a nice "DC" safe operating area which is probably bogus. No difference in transient thermal measurement at the same power between Vds=10V and 30V. Ah well, it was worth trying.
https://www.diyaudio.com/forums/sol...ed-transistor-measurements-2.html#post6698719
https://www.diyaudio.com/forums/sol...ed-transistor-measurements-2.html#post6698719
I spent five minutes, see figure 10 of:
https://nl.mouser.com/datasheet/2/196/Infineon_IRFP7430_DataSheet_v01_01_EN-1732904.pdf
Just look for insanely low on resistances.
MJL4281 - 100 volts at 1 amp, 200 volts at 200 mA - for one second. Trashes pretty much any mosfet (even the real IRFP240, which is better than the new ones).
I wonder what the REAL SOA is on some of these modern IGBTs, if they are better or worse than mosfets. ALL the data sheets show thermally limited SOA, and that can’t be true. But
if they are closer to BJT behavior, it could be a net win. Even switching bipolars have S/B slopes that start at 20 or 30 volts instead of only two or three, and fall off more gradually at high voltage than many Mosfets do.
I wonder what the REAL SOA is on some of these modern IGBTs, if they are better or worse than mosfets. ALL the data sheets show thermally limited SOA, and that can’t be true. But
if they are closer to BJT behavior, it could be a net win. Even switching bipolars have S/B slopes that start at 20 or 30 volts instead of only two or three, and fall off more gradually at high voltage than many Mosfets do.
Even "linear" parts have it a bit
OptiMOS™' 'Linear' 'FET' '100V'/'150V'/'200V - Infineon Technologies
I think this is the only honest manufacturer
OptiMOS™' 'Linear' 'FET' '100V'/'150V'/'200V - Infineon Technologies
I think this is the only honest manufacturer
There's always IXYS L2... Linear L2
I built an SE source follower amp using 500V 6A Depfets from IXYS . Biased at 25V VDS, and ~2A Iq. They work fine (since the addition of a bidirectional TVS from gate to source) even at 100°C.
IXTH6N50D2
I built an SE source follower amp using 500V 6A Depfets from IXYS . Biased at 25V VDS, and ~2A Iq. They work fine (since the addition of a bidirectional TVS from gate to source) even at 100°C.
IXTH6N50D2
Member
Joined 2009
Paid Member
the issue is that when the transistors we like are no longer viable they won’t simply stop producing them, they will abandon the tooling and processes too. These aren’t the kinds of tooling, materials and processes that a small business can acquire and run. When these parts are gone, they are really gone for ever. Ironically, vacuum tubes although not trivial to manufacture are less reliant on such tools and processes, they will outlive the transistors that relegated them to niche applications.