Unfortunately AC filament drive is a requirement for longer filament multi-digit VFDs.
Once you have more than a few digits, the voltage drop across a DC filament becomes significant compared to Vg-k so affects the brightness. With AC filament drive, either synchronous to the multiplexing frequency, or at a much higher frequency, the variations in Vg-k average out over the digit's on time so all digits are of similar brightness.
However the AC drive means you still need to provide a DC return path for the cathode current. Traditionally that is done with a center-tapped secondary on the filament drive transformer, so that if you hold that center tap at a DC level, that's the DC potential on the center of the filament. If the filament drive is capacitively coupled, for a virtual center tap, you can put a pair of resistors, typically of about five times the filament hot resistance, in series across the filament and apply the required DC potential to the junction of the two resistors. However that's equivalent to adding a series cathode resistor of half of one of the CT resistor's value, and if the voltage drop across the cathode resistor is significant compared to the static Vg-k, you are back to uneven brightness, this time varying with the number of segments illuminated in the digit.
Thoughts on how to provide transformerless AC drive, preferably from a supply as low as 5V would be appreciated. I know I can 'brute force' it using a low pin count MCU, with a PWM to alternating pins driving a small MOSFET H-bridge, capacitively coupled to the filament. By varying the ON time to dead time ratio, I can even control the effective RMS filament voltage. However it would be nice to have a single chip solution that doesn't require firmware, so I'm wondering if any of the commonly available push-pull SMPS controllers can be coerced into it, *IF* one can be found with a low enough minimum operating voltage and enough drive current capability.