Nixie tube clock ghosting issues

[Cristinello]

Hello guys,

I built a relatively simple nixie clock and I am encountering severe ghosting issues. To drive the mainboard I used an Arduino ( because I lack programming knowledge) . I will attach the schematic and the .ino file from the Arduino IDK . The schematic does not contain the nixie board beacuse it is a modular design  but it is straightforward. I am using 4028 BCD to decimal decoder to drive the nixies and the transistors are SMD versions of the FMMTA92 and 42. The power supply has also a strange behaviour , when left without any consumer , the output voltage is going down wih aprox 1Volt/30 seconds. If someone has any idea about what may be wrong I will be very very grateful .Thank you in advance!

[David Forbes]

Hi,

I think you have a timing issue. Your code turns on the anode and then
changes the cathode.

Try modifying each tubx() routine to turn on the cathode first, then
delay one millisecond, then turn on the anode.

--
David Forbes, Tucson AZ

[threeneurons]

Yes, alter your timing, as David suggested.

But also, those 100K resistors across the bases and emitters, of those A92s can be made smaller. Say 10K. The junction, will keep the voltage below 0.7V, so most of the current will still flow thru those junctions. Those resistors' primary purpose is to bleed off charge once the voltage drops below that 0.7V junction, and it starts looking like an open circuit.

Though, Its more likely a timing issue.

[Cristinello]

Thank you very much for your proposals but unfortunately none of them really work. I tried to modify my timings as Mr David suggested but it does not get rid of the ghosting. I modified also the 100k resistor and that does not really work either. Any other ideas? Thank you!

[Grahame Marsh]

Another technique is call "centre pull" where all anodes and cathodes are connected to a 90V supply (potential divider across 180V supply to ground) via high value resistors (say 470k).  Used in this design.

http://www.sgitheach.org.uk/nixie6.html

It means any residual charge is bleed off the electrode when you switch it off using the transistors and the unused electrodes float to the mid voltage point.

Not my invention!  But I always have found it effective with multiplex designs.

Grahame

--
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 post to this group, send email to neoni...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/70c37e3f-4cbc-4dbb-8368-9c3d4d8e643a%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[greg...@hotmail.com]

Leakage current ? I've seen past designs that use drivers with a breakdown-voltage-rating lower than the high-voltage supply. I always use drivers rated at or above the anode supply voltage, even though there are reasonable arguments for relying on the voltage-drop of the tube to reduce the voltage seen by the driver.

[David Forbes]

I wonder if the Arduino background program is causing the problem? It
does something to keep ocntrol of hte machine. Does anyone know what
that is? I don't.

The reason I ask is that you have no method of turning off the display
entirely. Your program changes one bit of the anode select at a time,
and one bit of the cathode select at a time. It is posible that the
clock spends some time displaying the wrong cathode or the wrong anode,
and this causes the ghost.

You could try to remove U1, the CD4028 anode decoder, and drive all six
anodes directly from the computer. That way, the code only changes one
bit to enable that digit's display.

Alternatively,you just have too much wire between the circuits and hte
tubes.
Please send a photo of the clock.

[Mark Moulding]

On Wednesday, May 21, 2014 8:33:17 AM UTC-7, nixiebunny wrote:

I wonder if the Arduino background program is causing the problem? It
does something to keep ocntrol of hte machine. Does anyone know what
that is? I don't.

Actually, there is no background process running on the Arduino; the Arduino is put back into program (boot-loader) mode by pulsing the DTR line from the host computer, which does a hardware reset of AVR. There is a timer interrupt at 1 kHz, but that uses very little CPU (less than 1%); the PWM outputs are also done using timers, but via hardware on the chip, so no instructions are executed.

[greg...@hotmail.com]

I looked at your driver schematics, and checked the transistor datasheet. The only possible culprit I see from a leakage perspective is the anode-driver; the NPN device has a 0.1uA Icbo (spec-max), and it could cause the PNP device to conduct several uA of current for an anode which is 'off'. The amount of current is hard to predict because its the product of NPN leakage, and the PNP's beta; both of these can vary significantly.

Cathode-driver looks solid; glad to see you used a high-voltage device.

Other than that, if you have a scope, I suggest you check the cs1-cs6, and k.0 thru k.9 signals very carefully. Your eyes will definitely see a faint glow on a cathode even if it's only energized for a few microseconds.

Can you describe the ghosting you see ? For example, If you illuminate a single '0', and all other digits are dark, do you see a faint '0' on the dark digits ?

Are the ghosts about the same brightness, or do they taper-off ? 

If you display all digits with the same numeral, are there any ghosts ?

If you display 123456, what sort of ghosting is seen at the 2 vs 3, etc ?

Those are important clues.

[David Forbes]

Thanks for that info. I was wondering because a previous Arduino project
I worked on had timing troubles.

This clock may just have a lot of wiring capacitance - I still haven't
seen it!

[David Forbes]

I don't see why that is necessary. Neither did National Semiconductor.

The 7441A driver chip contains 70V transistors; it relies on one output
always being "on" to ensure that the breakdown voltage isn't exceeded.
Once the neon ionizes, the voltage on the other cathodes is much lower.

That said, the TI SN74141 was redesigned with 55V Zener diodes on each
output to allow blanking. However, blanking requires that the anode
voltage be lowered, or an indistinct orange blob will appear in the
tube. But the anode transistors don't have to swing the entire anode
voltage either. They just have to make it low enough to not ionize -
about 100V.

I've sold a boatload of Nixie watches that use a Toshiba integrated 50V
Darlington driver. They work just fine. (I put a 50V Zener-reglated
supply on the commutator diode pin.)

[Cristinello]

marți, 20 mai 2014, 15:43:16 UTC+3, Cristinello a scris:

Hello guys,

I built a relatively simple nixie clock and I am encountering severe ghosting issues. To drive the mainboard I used an Arduino ( because I lack programming knowledge) . I will attach the schematic and the .ino file from the Arduino IDK . The schematic does not contain the nixie board beacuse it is a modular design  but it is straightforward. I am using 4028 BCD to decimal decoder to drive the nixies and the transistors are SMD versions of the FMMTA92 and 42. The power supply has also a strange behaviour , when left without any consumer , the output voltage is going down wih aprox 1Volt/30 seconds. If someone has any idea about what may be wrong I will be very very grateful .Thank you in advance!

Hello ,

I abandoned my design for a few months but last week I retried fixing the ghosting issues. I used an oscilloscope to try and identify the problems and I observed the following :

-> Due to the fact that my nixies are soldered on a separate board , the connections being only the anodes and cathodes , I observed that the wires connecting the boards are capacitively coupled and I could observe command impulses from one anode to the next ( lower amplitude of course ) . I fixed this adding one capactior in paralel with a resistor from each of the anode wires to GND ( on mainboard).

-> After fixing the clock timings , there was still ghosting on the tubes which I was not able to fix from software. A colleague of mine came up with the solution to add on each cathode an diode ( I added 1N4148 , 60 of them :) ) and the ghosting dissapeared . The problem with my design is that on the nixie board all the cathodes are tied together and the ghosting occurs like this : if I select tube A ( anode to VCC ) and pull the furthest cathode (from the anode ) to GND , all the cathodes inbetween will have a proportional charge , which in the tubes  B , C ,D ,E ,F translates as discharges between cathodes , aka ghosting.

I posted my findings so that it may help others in dealing with these issues .

Good day!  

[gregebert]

Interesting!! How exactly did you connect the 60 diodes ? It's not clear to me how they would solve the ghosting problem.

I've never measured anode-to-cathode capacitance of a nixie tube (but I will do that tonight....). I suppose if there's enough charge stored there, and if the multiplexing frequency is high enough, it seems possible that ghosting could occur from the brief period where stored charge illuminates the tube as it is dissipated.