A VFD question - if I may

[Richard Scales]

I acquired a bunch of Itron FG2013A1 VFD displays. These are 16 segments (plus DP/comma) 20 character displays.

I wanted to use them as a reason to try and grasp the nettle of multiplexing (which has eluded me for some time). I hooked the filament up to 4.5V and applied 12V to a grid and 12V to some segments - all good so far.

I had a bunch of HV5812 to hand and tried driving the display with these to turn on grids and segments - still good.

I made up a board with 2 x HV5812 drivers cascaded to give me a total of 18 segment drives and 20 grid drives. All I had to do then was to crack the multiplexing thing.

After some time and with assistance I got an interrupt routine firing every 1mS which would set the segments and grid I wanted to turn on and an SPI.transfer later - the display showed what I wanted. 1mS x 20 characters gives a refresh of 50Hz (if my math is right) which looks just fine.

.. however, when I want to show all 20 characters (in my test case 0-9 twice) I see some 'interesting' results.

The A/C filament drive is connected at the left hand end of filament, the other end to GND.

I am using 12V for VPP to the HV5812 drivers so I expect this to be what is going to each grid and segment that I want to activate.

The ISR just sets the segments for the current grid and turns them on in one hit (the SPI.Transfer is sending 40 bits - 20 for the grids, 18 for the segments and two are wasted)

Everything checks out using a logic analyser and there is plenty of time left to do other stuff - I am using an ESP-32 from AZ_Delivery for the testing and am using a CD40109 to convert the 3V3 logic to 5V.

I can see that it really is showing the right segments on each digit - it's just that there is a whole bunch more illumination of other unwanted segments which is far worse at the A/C supply end of the filament and completely gone ant the GND end.

I call upon those more experienced in these things to see if anyone can point me in the right direction?

I plan to try a DC filament drive next just to rule that out.

 - Richard

[Richard Scales]

This is supposed to show 0-9 twice - it is there but mostly masked by this 'effect' on the left.

 - Richard

[gregebert]

I think you want to drive the filament from a center-tapped transformer, and ground the center-tap. This minimizes the voltage gradient across the display, basically half of what it would be if single-ended.

With your current setup (single-ended), one side is grounded, and the other going positive/negative via AC, you are still setting up a voltage gradient across the display.

The other thing to consider is applying either a negative bias onto the grids to turn them off better, OR...if you follow the above suggestion with the center-tap transformer, apply a small positive DC bias to the center-tap. This will effectively make your grids negative relative to the filament (cathode) and improve the cutoff. You will also want to increase the segment voltage roughly by the same amount of the DC bias at the filament transformer center-tap, otherwise the brightness will be reduced.

Be careful about driving the grid positive relative to the filament; it will definitely increase the brightness as well as causing current thru the grid. Normally, grid-current is close to zero, because it's intent is to cause electrons to be repelled (blocked) from reaching the anode (segments). The datasheet should have details about this.

NIMO tubes behave similarly to VFDs, though the anode voltage is quite a bit higher.

BTW, I dont see any ghosting so that's a good sign.

[Richard Scales]

Many thanks indeed.

As part of my debugging/research, I tried using 5V DC for the filament and got this:

Here you can see that the brighter digits are at the end nearest GND and the less bright digits are at the end which is at 5V. I can understand that - at the 5V end there is only 12-5=7V between the filament and the grid/segments whilst at the GND end there is a full 12V between them - I am guessing that this is entirely relevant.

- Richard

[David Pye]

Does a short blanking interval fix the issue?

--
You received this message because you are subscribed to the Google Groups "neonixie-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neonixie-l+...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/neonixie-l/323902ee-1ed3-4235-8979-b38e7e63cd2bn%40googlegroups.com.

[Grahame Marsh]

I've always had success using a LM9022 VFD filament driver. It produces a square wave AC on two pins at 5V which you connect across filament. So the filament gets biased to about +2.5V.  Go look for the datasheet and you'll see how easy it is to use. You can also feed the outputs into a Dickson charge pump to generate 25V for the grids and anodes.

This chip is now unobtanium but the LM4871 (about £1 from Mouser) is identical although not listed as a VFD filament supply.

Hope this helps 

Graham

On 02/04/2026 12:03, David Pye wrote:

Does a short blanking interval fix the issue?

On Thu, 2 Apr 2026, 07:20 'Richard Scales' via neonixie-l, <neoni...@googlegroups.com> wrote:

Many thanks indeed.

As part of my debugging/research, I tried using 5V DC for the filament and got this:

To view this discussion, visit https://groups.google.com/d/msgid/neonixie-l/CAOQ6x0GwffP7q6gUZkY5peRvRW47_Ah4oXoaG7TPT%3DPJr9_NwA%40mail.gmail.com.

[Dekatron42]

I've looked a lot at the designs used by Kerry D. Wong. where he used a small ferrite toroid core: http://www.kerrywong.com/2013/06/05/vfd-filament-driver-using-555/

There's also this design in Nuts & Volts QA-section: www.nutsvolts.com/index.php?/magazine/article/october2011_QA - this design has a voltage adjustment so you can vary the output voltage!

/Martin

[Tony Adams]

The common Chinese mono amplifier IC usually known as an '8002' and available in lots of 10 for pennies, appears to be a pin-compatible clone of the LM4871 and works in the above circuit. I currently have one driving the filaments of 8x IV11 tubes. It's very efficient and doesn't warm up at alll.

To view this discussion, visit https://groups.google.com/d/msgid/neonixie-l/d4a94a84-95d0-4f37-b55e-4dae6189a418%40googlemail.com.

[Tomasz Kowalczyk]

" The A/C filament drive is connected at the left hand end of filament, the other end to GND."

I need to make sure: how exactly was this applied? The result is in line with what I would expect if you hooked up a transformer tap with one end to the left side of the display, other end to the right side, and grounded the right side. Which, for half of the cycle, would mean the cathode is negatively biased compared to the grid and anodes, letting some current through.

As Gregebert said, if your filament transformer has a center tap, use it. I think you should hook it up to a positive voltage of at least half the filament peak-to-peak voltage to prevent any part of the filament ever going below ground, so if your transformer produces 4,5 V, the center tap must be at least 2,25 V. Slightly more might be needed to get rid of off digit glow completely, it's display dependent. 
These displays are basically triodes, after all, and to shut off a triode completely you need to pull the grid below the cathode voltage. 

For chopped 5 V, I think dedicated filament drivers are not strictly necessary, any H bridge capable of operating at the filament current should be OK. Check what frequency the dedicated ICs produce as a reference. The filament is a thin, long wire which probably can resonate mechanically at some frequency, and I have absolutely zero idea what the range might be. 

[Tomasz Kowalczyk]

I am also curious: with the multiplex scan being relatively slow and close to mains frequency, do you observe any instability of brightness, like a beat frequency? 

[Richard Scales]

Very many thanks for the tips - a short blanking interval made no difference but the switch to DC showed that the multiplexing was working well.

I need to look at the setup of my current A/C filament drive.

As for the grid/segment voltages, I am currently using 12V - perhaps I need to change that?

- Richard

[Richard Scales]

@ Tomasz: With the filament on DC, no visible beat frequency at all.

There is a lot about this which is outside the scope of my current knowledge - though I am learning all the time thanks to the good people here.

I would like to crack it as I have a number of these displays that I would like to put to use.

I also plan to do the same trick with a large panaplex display that I have (16 digits of 16 segments) - at least that should be easier to drive (famous last words!).

- Richard

[gregebert]

Panaplex is easier to drive than VFD.

I recently brought-up a BR13404 (13 character / 9-segment) Panaplex. It's important to follow the datasheet very carefully, and bias the anodes and cathodes when they are *OFF* to prevent ghosting. Proper blanking time is important. I got mine to work on the first try, and didn't need to tweak any timing. Ping me offline for details.