Development of 500Vpp Linear Amplifier

[MrAl]

Exactly, these are switching transistors. And one shouldn't even need to read the datasheet to know that SOT223 can't handle the power dissipation of an HV output stage...

Well my question was not if the two transistors were suitable, my question was can you use bipolars.
Apparently you do not think it is possible.

I believe that those are probably not useable for linear operation with any significant power, so you'd have to do your own search and make your own conclusions.  It also depends how much output power you can live with.

It is probably possible to use NPN transistors, again if you can find any that are suitable.  If you do not like that then you can probably convert the output stages to PNP+NPN drivers.

It's entirely up to you, you can use four rusty nails if you want to and can get it to work

[David Hess]

The KSA1381/KSC3503 are my current go-to transistors for high voltage applications, but they are only 7 watts and 300 volts, limiting current to perhaps 20 milliamps in the real world.  Their high speed also makes them more likely to oscillate.

For a 500 volt and 100 milliamp design, that leads to maybe 2 or 3 in series and 6 or 8 in parallel, which is not quite as bad as it seems because parallel operation only requires adding emitter ballasting resistors.  That is about $30 or $40 worth of transistors, but still less expensive than the power MOSFET solution, and performance would be higher.

I wonder what high voltage high current PNP transistors used to be available.  The old 2N6214 could sustain 40 milliamps at 200 volts.

The modern MJE5852G ring emitter audio transistor can sustain 50 milliamps at 250 volts, so only two are needed in parallel.

[MrAl]

Hi,

You reminded me of the old 2N6250 transistor which is a high power device, although it is an NPN.
I used something like 5 of these in 'series' to test an old CRT television set where the flyback transistor died and I wanted to test it to see if it would work with a new transistor that I would have had to purchase at a premium cost.  I already had several 2N6250's so I rigged it up to test the TV.  It turned out it worked so I purchased the right transistor and fixed the thing.

We used these back in the 1980's to drive huge bipolar transistors for use in high power (1000Watts to 30kWatts) converters.  Those transistors were over $100 USD each back then and the big stud mount case was made of copper with a 1/2 inch bolt stud also copper.  They must have weighed in at least a pound each.  Pretty amazing transistors.  After that we went to the big rectangular type which were cheaper but also very high power (100 amps).  After that we started going to MOSFETs for their higher switching speeds.

I have a feeling not all the transistors still available out there can be found on regular outlet sites like Digikey and Mouser.

[langwadt]

The KSA1381/KSC3503 are my current go-to transistors for high voltage applications, but they are only 7 watts and 300 volts, limiting current to perhaps 20 milliamps in the real world.  Their high speed also makes them more likely to oscillate.

For a 500 volt and 100 milliamp design, that leads to maybe 2 or 3 in series and 6 or 8 in parallel, which is not quite as bad as it seems because parallel operation only requires adding emitter ballasting resistors.  That is about $30 or $40 worth of transistors, but still less expensive than the power MOSFET solution, and performance would be higher.

I wonder what high voltage high current PNP transistors used to be available.  The old 2N6214 could sustain 40 milliamps at 200 volts.

The modern MJE5852G ring emitter audio transistor can sustain 50 milliamps at 250 volts, so only two are needed in parallel.

FJPF5021 seems to be in that range too,

[David Hess]

The KSA1381/KSC3503 are my current go-to transistors for high voltage applications, but they are only 7 watts and 300 volts, limiting current to perhaps 20 milliamps in the real world.  Their high speed also makes them more likely to oscillate.

For a 500 volt and 100 milliamp design, that leads to maybe 2 or 3 in series and 6 or 8 in parallel, which is not quite as bad as it seems because parallel operation only requires adding emitter ballasting resistors.  That is about $30 or $40 worth of transistors, but still less expensive than the power MOSFET solution, and performance would be higher.

I wonder what high voltage high current PNP transistors used to be available.  The old 2N6214 could sustain 40 milliamps at 200 volts.

The modern MJE5852G ring emitter audio transistor can sustain 50 milliamps at 250 volts, so only two are needed in parallel.

FJPF5021 seems to be in that range too,

I was only looking for PNP candidates since they are less common.  There are plenty of NPN candidates and audio output transistors have complements.

[langwadt]

The KSA1381/KSC3503 are my current go-to transistors for high voltage applications, but they are only 7 watts and 300 volts, limiting current to perhaps 20 milliamps in the real world.  Their high speed also makes them more likely to oscillate.

For a 500 volt and 100 milliamp design, that leads to maybe 2 or 3 in series and 6 or 8 in parallel, which is not quite as bad as it seems because parallel operation only requires adding emitter ballasting resistors.  That is about $30 or $40 worth of transistors, but still less expensive than the power MOSFET solution, and performance would be higher.

I wonder what high voltage high current PNP transistors used to be available.  The old 2N6214 could sustain 40 milliamps at 200 volts.

The modern MJE5852G ring emitter audio transistor can sustain 50 milliamps at 250 volts, so only two are needed in parallel.

FJPF5021 seems to be in that range too,

I was only looking for PNP candidates since they are less common.  There are plenty of NPN candidates and audio output transistors have complements.

maybe something like this from when PNP power transistors were less common, https://sound-au.com/articles/dcp-f6.gif

[David Hess]

maybe something like this from when PNP power transistors were less common, https://sound-au.com/articles/dcp-f6.gif

There is nothing wrong with the quasi-complementary pair, however the PNP still has to sustain the same high voltage across its collector.  The advantage is that the PNP is lower power.  It would be more suitable if a higher current or higher power PNP was not available.

Did IXYS make linear p-channel MOSFETs?  I cannot find any reference to them.  Anyway, their standard 600 volt IXTH10P60 is suppose to support 350 milliamps at 500 volts if you believe their datasheet, and can find it in stock.  Pair it with a IXTH20N60 I guess, but of course a high voltage driver will still be needed to drive their input capacitance.

[magic]

Driving the output FETs really becomes simple in the "earthed source and floating HV supplies" configuration, where all drive electronics work at ground-referenced low voltage and high current is no issue.

Otherwise, one might consider a "rail to rail" common emitter output stage with FET cascodes. The problematic, high voltage Cgd is now driven by the power FET itself and Cgs is driven by the common emitter/source transistor but it only sees low voltage so it can be any standard part with low power rating. The common emitter transistors should also be controllable by means of fairly low current, so the preceding stage can operate at low power and use simpler and cheaper transistors. And they only need to handle one rail worth of collector voltage.

[amyk]

Am I the only one here who thinks a vacuum tube design would easily be able to handle this sort of voltage and current?

[David Hess]

Am I the only one here who thinks a vacuum tube design would easily be able to handle this sort of voltage and current?

A vacuum tube design is very feasible, but without a PNP or p-channel equivalent, I think it is limited to class-a operation.  (1) If you went this route, then the same topology will also work with only high voltage NPN or n-channel transistors and there are lots of options there.

A vacuum tube design will also require floating heater voltages that can sustain 500 volt swings, so some custom transformers.

How fast was that optically coupled design?  Did it use all n-channel or NPN output devices?

(1) A complementary class-ab design with p-channel vacuum tubes made of antimatter is left as an exercise for the reader.

[magic]

Couldn't this output stage work in class AB with tubes? You may not even need I5 thanks to R7 and negative grid bias on Q7. Connect R8 to a stacked negative rail whose value is twice the desired idle grid voltage and you have 100% freedom in selecting standing current, while both outputs can be driven up to zero grid voltage.



edit
To stay more on topic, this is also a possible way of avoiding P-ch power MOSFETs

[David Hess]

Couldn't this output stage work in class AB with tubes?

Yes, but see below.

To stay more on topic, this is also a possible way of avoiding P-ch power MOSFETs

The input stage is depending on 4 p-channel JFETs.

[GN89]

Thats a lot of input from you guys. Tank you. I continues working on this so I show the results so far. What I changed is to get rid of the biasing circuit and add a new stage for driving the output transistors which will also make quiescent current setting more easy (Q18-Q24).

Slew rate is still below the limit I want to make sure 20kHz works with full swing. The stacking of the VAS current source is not ideal but there are no BJTs with such SOA even Mosfets need stacking.