Here are some new measurements of key outputs:
|AC Mains||120.7 VRMS @ 60 Hz|
|Transformer output||7.38 VRMS @ 60 Hz|
|Rectifier output||7.85 VDC|
Note that the rectifier output is lower than the 10-11 VDC I recalled. At this voltage I saw no evidence of discontinuous operation of the voltage regulator, which makes determining the filament power input much easier.
Here's the measurement across the filament on the second P170-DH, with the positive probe on VF1 and negative on VF2. Since the blocking oscillator only drives the transformer one direction rather than a push-pull arrangement, the output isn't symmetrical and has a bit of an offset. I've adjusted the zero-volts point to take that out of the picture.
Ignoring the brief spikes on the negative-going side, this looks a lot like 4.6 Vpeak at 115 KHz.
Lowering the DC input to the first calculator board so the rectifier output is 7.85 VDC produces this across the 30 ohm resistor substituting for the VFD filament:
It's not as spikey on the negative side but otherwise looks a lot like the second calculator board. It's shifted a bit more towards VF1 and the frequency is a tad lower, but even flipping back and forth between the two in an image viewer doesn't show a lot of difference.
Let's compare these with the output drive from my MAX253 circuit using the 4:3 ratio transformer I selected.
The first thing that jumps out is that the waveform is much cleaner and far more symmetric with no offset. This is not a surprise given that the MAX253 was designed to drive a transformer symmetrically. The second thing to note is that the frequency is higher: 400 KHz if I'm counting correctly. Again not surprising, as I have the MAX253 strapped for high-frequency operation and that's in line with the datasheet graphs (Vcc is 6.1 V). The important point, though, is that the voltage across the resistor looks pretty close to what the original board supplies.
While supplying 4.5 V across the 30 ohm load (675 mW), the MAX253 circuit is drawing about 207 mA at 3.3V (683 mW). Changing the strapping on the MAX253 to select low-frequency operation produces this output:
Looks like 238 KHz to me. That's about double the frequency of the original design, but the circuit takes only a fraction of the space.
Unfortunately it looks like I'm asking too much of the MAX253 and its Halo transformer. The MAX253 datasheet contains these limits:
Elsewhere in the datasheet is the statement, "The average current in each MAX253 switch must still be limited to less than 200mA..." This is confusing to me, as each switch draws current only 50% of the time, which I would think means I can draw 400 mA per switch. Yet the datasheet also says it can provide power levels "up to 1W with a 5V supply", which is 200 mA, and "600mW with a 3.3V supply", which is 189 mA. Maybe this limitation is because the current through an inductor is a ramp rather than constant as it would be through a resistor; regardless, at an input current of 207 mA it looks like I'm over the documented limits.
The MAX253 isn't even getting warm (+3C over ambient). The transformer seems a bit warm (+8C over ambient), but the resistor is soldered directly to its terminals and the resistor is hot (+20C over ambient) so I expect some of that heat is thermal conduction from the resistor. But I'm not interested in designing flawed circuitry.
As an experment I reduced the input voltage until there was a noticeable change in segment brightness, and it looks like the original circuitry may be overdriving the filament. To reduce the current demands I replaced the 4:3 transformer with one having a 2:1 ratio that I bought for just this situation. The modified circuit draws only 120 mA from the 3.3V supply, which is well within the datasheet limits and doesn't appear to change the brightness noticeably.
I'll keep this configuration until I see what things look like with the digits illuminated at a 1:13 duty cycle. If the display looks good I'll use it as-is; if not, I'll switch to a MAX256 or SN6505 and a transformer that is rated for the power.