High voltage output amp(with NPN transistors)

[Gunders]

As mentioned in an older thread I have a final project at school where my group have to make a 3-phase generatorsimulator.
My part is to make an amplifier which have to give out +-800V.

I have now done some design with the amplifier, but I still have some problems and I hope somebody here can help me.

So far I have tried Marshall Leach's double barrelled amplifier, and added an extra cascode stage to protect the outputtransistors.
Even though I have added an extra cascode stage it isn't enough to protect the PNP transistors.
The max voltage the transistors see now is about 530V. For the NPN transistors this is okay as NPN types are easy available with voltages up to 1200-1500V.
For the PNP types it's an another story, here the maximum voltage for available transistors seems to be about 400-500V. If I could find 600V PNP types it would help a lot, but it's still a bit on the risky side so 700V PNP types would be great.

What I try to do now is to design an amplifier with NPN types only, but I haven't yet found the right way to do it.
It's here I hoped somebody could help me with an proper outputstage with NPN transistor only.


Here is the schematic for the amp so far. This design is only used to verify that this kind of design works properly and the voltage over the transistors is shared equally. The final design will of course use other transistors than those used in this schematic.

[Frank Berry]

How much output power?
Could you use an output transformer to step up the voltage?

[Gunders]

Originally posted by Frank Berry
How much output power?
Could you use an output transformer to step up the voltage?

Not very/extremely much output power, absolutely max outputcurrent is 100mA but 10-20mA is more realistic under normal conditions.

An output transformer have been considered, but the company I work for would like to see an amp without transformers.
They have a solution today which uses outputtransformers, but they would like to make it smaller and lighter.

[Christer]

I haven't tried to figure out how or if it would work with cascoding
for high-voltages, but have you considered something like a
quasi-complementary output? You will still need PNPs, of course,
but not for the output stage.

Another alternative, could a bridged amp be acceptable?

[Gunders]

Originally posted by Christer
I haven't tried to figure out how or if it would work with cascoding
for high-voltages, but have you considered something like a
quasi-complementary output? You will still need PNPs, of course,
but not for the output stage.

Another alternative, could a bridged amp be acceptable?

I don't know how to implement a brigded amp to a 3-phase system where you have the phases L1-L2-L3 and GND as a referance for the 3 phases. Suggestions are welcome.

I just took a look at the quasi-complementary output stage, if the PNP transistor in the outputdriver stage could be removed it would be great.
Wouldn't it be possible to drive a outputstage of this kind in a push-pull manner?
If I could use a lower voltage for the driverstage a quasi-complementary output may be a alternative, but I can't see how I can use a lower voltage in the driverstage.

[jcx]

cascading doesn't have to be symetrical, just add a few more stages to the pnp side

while you're at it you might want to look more closely into the divider chain ac load division and include device input Z in the calc


on a system level i would start with a nonisolated ac rect bridge dc supply and work with 400/600 V devices, then the 3 phase ouput transformer provides isolation and step up with a low system cost - control inputs may be isolated with signal xfmrs or opto isolators - or use wireless digital tech (irda for really cheap)

[Gunders]

well, here's my only working NPN outputstage so far.
The output waveform is pretty symetrical and with low distortion.
Even if it's a working design i don't know if I have biased it properly so it would be nice if smoebody could help me a bit.
The problem is the frequencyresponse, it looks rather awful.
The amp seems to have some zero's in the transferfunction.

[Gunders]

Here's the AC analysis.

[Gunders]

It is something wrong with the first schematic, here's a correct one.

[Workhorse]

Hey Man why are u stick around those Bipolars , Why dont u use IGBT s with rating upto 2KV . U can implement total n-type design with them very well.

With High Voltage Upto 100000000000000KV Regards
AAAAAAAAMMMMMMPPPMAAAAN