I gave some serious thought to using the original transformer and duplicating the original power supply circuit to power my redesign of the P170-DH calculator's guts, but I decided it involved too many compromises. I don't need a +5V supply; the FPGA needs +3.3V and +1.2V. Since the filament and VFD drive supplies are derived from taps on the same transformer, adjusting the output voltage down from 5V to 3.3V would also reduce those outputs proportionately, and I have no idea what current it could supply. I gave some thought to using it solely to power the VFD while implementing separate +3.3V and +1.2V supplies, but that just seemed silly.
My new approach starts with +3.3V and +1.2V step-down switching regulators, possibly based on the TI TPS62160. These can handle the unregulated rectified output from the mains transformer and could provide up to 1A each (far more than the calculator's transformer can supply) with efficiencies approaching 90%. This will more than meet the FPGA's needs and requires only a single inductor, a couple of small filter capacitors, and a resistor divider to set the output voltage.
The filament is best driven with AC from an isolated winding of a center-tapped transformer, so I spent some time last week looking for a suitable transformer and driver. I found some ICs capable of both driving an isolation transformer and regulating the output which would allow them to run from the unregulated supply, but they're intended for rectified DC output on the other side of the transformer and I'd like something with a fairly symmetric AC waveform.
My initial selection is the Maxim MAX253 transformer driver, driven from the regulated 3.3V supply, and a Halo TGM-230NSLF transformer. Here's a schematic from the MAX253's datasheet that shows the general idea, aside from the transformer ratio and the rectifiers on the secondary side.
With the primary fed with 3.3V at its center tap, the voltage across the whole primary winding is twice the input voltage. Ignoring losses, a 1:1 ratio transformer would thus give me 6.6V across the secondary, and I've already determined that 5V is too high. Since this is the real world I started with 4:3 ratio transformer to make up for losses I expected to see. As a hedge I decided to also get a TGM-210NSLF 2:1 ratio transformer, which would give me something under 3.3V AC out.
One of the interesting aspects of the recommended circuit for use of the MAX253 at
3.3V is the pair of diodes between the transformer primary windings and
the Vcc pin. While the MAX253's output transistors are happy with a 5V Vcc, they're a bit anemic with only 3.3V Vcc.
Since the transformer is fed through its center tap, pulling the D1 pin
to ground causes the D2 pin to rise to twice the supply
voltage (and vice versa). These two diodes use this to boost the Vcc to about 6V,
allowing the output transistors to get the gate drive they need to
operate efficiently. However, this limits the input voltage to 3.5V to avoid exceeding the Vcc maximum of 7V.
Early Monday morning I placed an order with Mouser for the parts to breadboard this circuit. The parts arrived in the mail this afternoon, and this evening (Thursday) I wired it up:
I loaded the secondary winding with the 30 ohm resistor visible on the left side of the photo to simulate the filament without risking damage to the VFD. With the 'scope's ground connected to the center-tap of the secondary, here's what I see across the resistor:
That looks like about 3.8 VAC, nice and symmetric, at a bit over 312 KHz. That's about a volt lower than the no-loss ideal. Later I'll take a look at the primary side of the transformer, but I suspect much of the loss is the resistance of the output switches in the MAX253. They have a typical resistance of 1.5 ohms at 100 mA, and I'm pushing closer to their max current of 200 mA.
After some preliminary testing with the 30 ohm resistor I connected the VFD filament. I don't have a good test jig yet, but it seems quite happy with this driver.
Edit: After posting this last night I realized that this 'scope trace shown above was taken with the filament wired in parallel with
the 30 ohm resistor, thus doubling the required output current and
pulling the output voltage down. With only the 30 ohm resistor the
output I still see that odd dip in the tops of the waveforms but the low point is above 4 volts. This may be higher than the filament should get, since this circuit isn't bursty like the original circuit. It may be time to crack open my other P170-DH and try to figure out the average power actually being supplied to the filament -- I may get use out of that 2:1 ratio transformer after all.