The H-Bridge I built was originally too slow. I perfected it using smaller pull up/down resistors and tuning the transistor choice further, and it’s now giving very good performances. On this matter I highly recommend the Art of Electronics’ chapter on bipolar transistors under “2.2.2 Switching circuit examples” which explains properly the sizing of the resistors used in this H-Bridge design.
Now, the problem I have is that my H-Bridge is still too slow to operate at an alternating 20+Khz. There are a few nanoseconds window where the bridge is fully off when switching from one half-bridge to the other. This in itself is not a big problem: you could technically operate at a slightly higher voltage to compensate for the RMS voltage lost in the switch. The voltage spikes would probably be detrimental to the VFD over a long term though, so I chose not to go down that path.
To get better switching performances out of a BJT H-Bridge I would have to diode clamp it and further complexify the design. So I gave up. I bought a couple of commercial h-bridges suitable for low voltage, one of them being the DMOS LV8405V to see if it would perform better:
… Yep, and here it is at 78Khz without batting an eye.
Test circuit built:
Of course it’s not perfect, there’s a big voltage drop as you increase the load that is typical of mosfet bridges; but because the LV8405V operates two full bridges into a single chip, you can balance out the load.
Conclusion on building a VFD filament power supply:
So my final solution on building a VFD power supply:
- An alternating H-Bridge is the correct solution to avoid luminosity issues (i.e. one side of the VFD brighter than the other).
- The square wave needed to drive the H-Bridge can be done through an inexpensive Attiny13 –here a Atmega328p for simplicity.
- You can build your own H-Bridge but for these low voltage russian IV-25 tubes; two commercial products are suitable:
- OnSemi LV8405V
- TI DRV8837 / DRV8838
- You will lose a lot of power on this H-Bridge because the voltage drop is big in comparison to the driving “motor” voltage.
I found that Texas Instruments make a great low voltage H-Bridge (DRV8837 / DRV8838); but as they are only available in a WSON package I could not prototype them.