Is there a way to use NPN transistor for high side of H bridge at around 400 volts?

Question

I have tried using 2 PNP and 2 NPN transistors for a full H bridge but as high voltage PNP transistors aren't easily available, I want to use all 4 NPN transistors for high voltage.

Answer 1

Yes, it’s possible. You need to provide base drive above your supply voltage in order to not make an emitter follower and at this point you might as well use a MOSFET instead.

If your supply voltage is high enough, an emitter follower might be acceptable from losses point of view.

Answer 2

The old way was to use 4 power NPN Darlingtons with 4N35 opto drivers. If your current is great, the heat loss can be greater than modern FETS.

Quad Nch FETs are popular as ESC H bridge drivers because PWM on the low side is used to with a diode cap to a charge pump Vboost for the high side Vgs. Cross-conduction is eliminated with careful time-constant choice deadtime during commutation to determine if snubbers are needed.

Answer 3

I don't think H-bridges specifically were used very much, back in the day (before MOSFETs displaced BJTs for SMPS application), but they can be driven in the identical way to the half bridges popular in older ATX computer supplies: use a base driver transformer, complementary secondaries to drive diagonal pairs during positive or negative phase, plus a short loop through the output to obtain positive feedback and thus constant-hFE drive. Control is shorting-mode, usually from a TL494/KA7500 plus a pair of drive transistors which act to short out the drive transformer, ensuring turn-off.

If you need DC drive, I wouldn't recommend it. You might be better off with individual drive transformers, output rectified to supply base current -- if you don't mind a dreadfully slow turn-off as there's only self-discharge acting on such an arrangement. This would be good enough for an AC inverter ("modified sine wave") or motor switching, if probably not motor control or PWM.

For faster switching, probably leave the transformer on CW as a power supply (DC-DC isolator), and add an optoisolator and drive circuit to kick the base into and out of conduction.

Bootstrap drive is also possible, which I suppose isn't too bad with a 5V-capable gate driver -- they might not be suitable for continuous output current though, being made only for brief pulses of current. A discrete solution may be needed. Obviously, bootstrap only works well when the peak negative-going voltage is limited; this is easy for MOSFETs (body diode), but with BJTs we may need to include clamp diodes.

Some Linear Technology (now Analog Devices) switching regulators are bipolar in design, and use a bootstrap at fairly low voltage (2-3V). They tend to work at surprisingly low supply voltages, while having good efficiency (the bootstrap allows them to achieve Vce(sat) properly, in contrast to e.g. LM2592 which uses a Sziklai style arrangement, losing a bit over 1*Vbe voltage drop).

Big industrial units -- often using "trilington" modules -- I believe used the latter, isolated base drive circuits. The triple Darlington arrangement needs relatively low forward base current, but turn-off is quite rough: they usually provide E-B diodes to allow you to shunt base charge from the whole stack of NPNs, but you'll need to sink many amperes peak to be worthwhile.

Answer 4

This is almost one of those "if you have ask how you don't want to try" questions.

Yes, you can drive NPNs on the high side. Basically, most all of the methods that one might use to drive the far more common N-channel MOSFET circuit topology will work; you just need to account for the fact that you need to supply current to the base all of the time that the transistor is on.

Note that if you are going to try this, you want to look for transistors that are newer than the old standbys of the 1970's (i.e., the 2N3055 and the TIP-whatever are out). For a while in the 1980's and 1990's Siliconix (bought by Diodes, Inc.) made it their mission to design & manufacture bipolar transistors that would compete with MOSFETs in the switched-power arena. They may have made some transistors with \$V_{CEO}\$ ratings high enough to make this work, and there may even be some in current production.

For the most part, you'll suffer much less pain if you just design your circuit using the latest N-channel MOSFETs -- these are under continuous improvement, and will just plain work better than any bipolar transistors you can find.