IN-1 reliability

[Ian Sparkes]

Hi All,

first post here, so go easy on me... ;)

I read here that people have been having problems with IN-1 reliability, but in my (admittedly not huge) experience, I have not so far seen any IN-1 tube fail, even if I have a reasonable number of them in use (6 on a "SixNix" clock from PV Electronics, 6 each on two home brew Arduino based clocks, 4 on another home brew, 2 in two separate single digit home brew clocks). The longest running of these has been going full time for about 2 years (about 17000 hours). All clocks (except the single digit ones) run ACP every 10 minutes.

I have never had a failure, and I didn't even consider the possibility of a tube failure up until now, so I'm a little bit disturbed by the idea that others have had to replace tubes after 15 days.

All of the multi-digit clocks are 1 x n multiplexed, and the latest clocks I have built use 1 x 6 multiplexing, with weekend blanking and ambient light dimming. At night, each tube is on 10% (dimming factor) of 16% (multiplexing factor) of the total time. The oldest clock "SixNix" just has static night time dimming.

I have seen grid discolouration on only the single digit clocks which are running practically in a "direct" (non-multiplexed) mode, but even this is not clear. I have one with a tube from 1984, and one with a tube from 1987. The 1984 tube appears to be degrading, but the 1987 one does not. Degrading here means only that the grid has lost it's shine.

I'd like to try to understand better why I have not had any failures while others have. Is multiplexing simply extending the failure point, or are there perhaps other factors that influence the life time? I'm using K155s with 2k7 Anode resistors and a more or less regulated 180V supply (more or less because the regulation is done in software, so there is a bit of ripple about).

Did I just get very lucky, or is there something that I am unwittingly doing right? Is there any variation by year that anyone is aware of? Any idea when the Hg was added?

Thanks for any input or ideas.

Ian

[gregebert]

Perhaps your tubes were from a different manufacturer, and probably a different manufacturing process. Since they are from 1984 to 1987, that's even more likely.

My IN-1 tubes are from the Anod plant, and have date code of 1992. Each tube was displaying a single numeral 24/7, with no cycling, dimming, multiplexing, etc. I had failures every few days. Each failure was a micro-filament that developed between adjacent cathodes, causing an electrical short between them.

After switching to Burroughs tubes, I've had only 1 failure with 15 tubes after 2 years. That failure appears to be a broken internal spot-weld.

[Ian Sparkes]

Thanks for the info, it sounds like a slightly different use case, but a valid one. Given the age of the oldest tubes I have (early 70's) I can't imagine that multiplexing was the normal mode of operation, and direct drive was more likely. On the other hand, I find it hard to believe that these tubes would have had a 20+ year prouction run if they had been so terribly unreliable.

One thing that comes to mind is that perhaps the current was too high in direct mode. Do you have a sketch of the circuit? Perhaps a cycling of digits is particularly important for these tubes? Why should a bridge form from one cathode to the other? Was it the same digit pair each time? Is there perhaps a bias voltage that should be applied to undriven cathodes?

I'd love to understand the failure mechanism better, and perhaps I'll manage to reproduce it when I have some sacrificial tubes.

In the mean time, I'll report back if any of the 20 something IN-1s I have in action play up.

Ian

[gregebert]

My driver circuit is a direct-drive constant-current source, and it was set to the IN-1 spec current of 3.0mA. I was only using one numeral in each tube, so all undriven cathodes were floating. The clock uses 15 tubes to display 1 thru 12 on the clock face (it's a large neon-equivalent of a regular mechanical clock) so the numeral is displayed constantly.

What I found is that after a few days there would be 2 glowing cathodes: The desired numeral, and the adjacent cathode closest to the front of the tube. I confirmed a low-resistance short of a few ohms was present. I kept a lot of notes, but I would need to dig around for them. I also found that the short was caused by a tiny filament that formed between the cathodes by passing about 100mA which caused it to glow. It usually took 300 to 500mA to zap the short; it acted like a fuse. Afterwards the tube functioned as expected.

As I ran my clock longer, I noticed more tubes failing in the same manner, and each time I was able to zap them back to life. It got ridiculous having to zap another tube almost every day, so I replaced them with Burroughs tubes and never had any trouble.

I suspect the filament was growing in the direction of the electric field between the cathode and anode, and grew until it shorted the adjacent cathode. I dont have the equipment to analyze the chemical makeup of the cathodes or the filament that formed. I do know it was very small diameter; not visible until it was made to glow.

NASA has done considerable research on tin whiskers, which is probably a similar mechanism to what was failing in my IN-1 tubes. They form in electric fields.

I've heard that IN-1 tubes do not contain mercury, whereas Burroughs (and many other brands) do contain mercury. Perhaps there is some odd role mercury plays in preventing the formation of these filaments.

I suspect that after the USSR broke into separate nations that different materials or processes were used in IN-1 manufacturing; there may have been relaxed quality standards as well. My tubes are all date-coded 9205.

I have an A101 dekatron spinning 24/7 on the same clock, and it's been running flawlessly for more than 2 years. Despite a much higher operating voltage (hence higher electric field), it's fine.

[Ian Sparkes]

Thanks for the update, that's a clear description, and now I understand the reason that each digit is illuminated constantly.

I think you might be right in your suspicion that it is a materials problem: 92 is in the depths of the economic chaos in Russia, and people had been living without wages in many cases for months. I'll have a look in my collection if I have dates that have such a late date, and if so, perhaps I will be able to reproduce and understand what the issue is.

Thanks for the help, I'll report back any findings I can come up with.

[Jan Rychter]

Just took a quick look at my stash, and I also have a bunch of IN-1s coded 9205. I'm not sure if it's worth it to design something around them, then — if they fail so quickly and easily.

--J.

> --
> 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/53216fa6-0cd4-4b97-a0e1-7b0daccf73e6%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[gregebert]

The 9205 IN-1's are perfect for experimenting. They are inexpensive, easy-to-find, and have similar-enough characteristics of better tubes that you can use them for some prototyping. Just be aware that IN-1's use a bit more current (3mA) so you will need to adjust your anode resistor (or current-limit if you go that route) when you change tubes.

They might work OK in a clock that is not displaying 24/7, or perhaps a digital multimeter (yep, it's on my drawing board...) I only know from my experience that displaying a single static numeral on an IN-1 with  that date-code is going to cause tube failures pretty quickly.

BTW, dont waste time/money on sockets for the IN-1 (or A101 dekatron for that matter). I use crimp-connectors (TE Connectivity AMP Connectors 350415-1 , available for cheap from DigiKey) soldered to a wire, then covered with heat-shrink tubing. It works perfectly.

I have about 20 of these boogers, and I was thinking of arranging them in a 4x5 matrix into an "idiot box" where they all display numbers randomly. I've come close to putting a near-dead one in a microwave oven and documenting on youtube, but I have too much admiration for nixies to do that.....even THESE nixies.

[Ian Sparkes]

Hmmmm... Interestingly I think I have found a confirmation of a micro filament between two cathodes in the IN-1.

It's a long story, but for those who are impatient, skip to the "Round up of current observations" near the end.


In a previous version of a direct drive clock I made, digits 4 and 8 came on together, and then after a couple of weeks 1 and 6. Looking at the order of the cathodes it is (from the back): 1, 6, 0, 2, 5, 9, 7, 4, 8, 3. So, 1 and 6 are adjacent, as are 4 and 8, therefore a micro-filament is a reasonable explanation.

I previously put this down to a cacked Atmega 328, because I measured the pin voltage at the Atmega output pin, and read it as "just on" at about 0.7V. I made this single digit clock as a way of using up an Atmega that had blown a few of the pins due to me having two left hands, and thought no more of it, as I had already damaged the output driver of the microcontroller.

However an hour or so ago, now the "neat" version of the clock I made has started showing 1 and 6 together, although the Atmega328 is new and undamaged!

Previously I observed that giving the tube a firm but not violent tap (like you used to do with malfunctioning TVs) resolved the issue. But It would soon again re-appear. I put this down to a bad solder joint, but it always left me puzzled as to exactly which solder joint could cause 4 and 8 to appear. No single binary pin could cause that.

Now something starts to make sense. In my direct drive clock, I am not being at all fastidious about the driving voltage: the regulator software is just taken from another clock where I am a lot more careful about the timing and blanking, and where I turned off the HV generation between digits to avoid over-voltage.

(Anyone wants to see the code for the original clock, it is here https://github.com/isparkes/ArdunixNix6 and open source)

In the clocks that show the problem, I'm just doing the bare minumum to stop the worst over voltage effects (digits ghosting on even when they should be off), but I do know that I am driving the tube to desperation, because I'm getting a fair amount of "flare" on certain cathodes (see the picture - The strange glow in the background is emanating from the lead that goes from the pin on the rear of the tube to the cathode, and that is not normal).

A round up of current observations:
1) In my IN-1 clocks which are multiplexed and driven according to the voltage specification (i.e. I was careful about it), I do NOT see micro-filaments, even after around 2 years of continuous use.
2) In the IN-1 clocks which are direct driven, probably with out of spec voltages (TBV), micro-filaments appear after 2-4 weeks
3) My tubes in the offending clocks are from 84 and 87 (i.e. NOT due to poor materials after Soviet breakup after 87)
4) The offending clocks do not do any form of brightness control or ACP, whereas the "good" clocks do.

So, problem confirmed. Cause as yet unknown.

Next steps will be to monitor the voltages in the "bad" clocks, and make software changes to ensure the driving voltage remains "in spec".

I'll keep you all updated.

[gregebert]

OK, so there is definitely some goodness for the IN-1 when it's multiplexed.

With direct-drive, the tube has a constant internal electric field so there will always be favorable conditions for a metallic filament to form. What I dont understand is why a tube that is 4:1 multiplexed doesn't have any filament growths after a longer time (roughly 4X) versus direct-drive.

There is obviously something else in-play. Maybe scope traces of the multiplexed tube will reveal a possible cause. My guess is there is some kind of activation-level required to cause filaments; it will be a function of temperature and electric-field strength. But there must be something else, such as a minimum on-time.

[Ian Sparkes]

A quick update to draw a line under this topic: since the last post some time ago, I've had no more failures, even on the non-multiplexed clocks. Here are my observations:

In general, IN-1 are more sensitive to misuse, and will bridge over between digits (1 and 6, 4 and 8) if not driven within their parameters. Presumably this is because of the lack of mercury doping, which also gives them their warm glow

If run in multiplexed mode, there does not appear to be a large problem. You should still respect the current limits, but it is possible to moderately over-drive the tubes with no ill effects. My oldest multiplexed IN-1 clock is still looking great after a couple of years of continuous use and no special precautions.

If run in direct mode, you have to be more careful with the current limit. I had repeated failures until I really measured the current and adjusted it to be in-spec (3mA). Even with direct drive I still pulse the current, but that is because my design easily allows for it, and I use ambient light dimming.

IN-1s do need more care than other tubes, but if you are careful with the current or use multiplexing, you should be OK.

[gregebert]

I'm curious what date-code your IN-1's are; in an earlier post you mentioned 1984 and 1987.

[Ian Sparkes]

I've got examples from 77 to 91 in use, we did suppose at one point that it might be a quality issue due to the Soviet break up, so I used the earliest and latest I could find. It didn't make any difference...