Cathode-poisoning prevention: IN-18

[gregebert]

For those of you who did your own poison-prevention algorithms, are you trying to run all cathodes for the same amount of time, or do you just activate the 'dead' ones for a few minutes each day when nobody is watching the clock ?

Thanks to the 'Tomorrowland' movie, I'm making my first IN-18 clock because my wife likes the larger digits. All of my previous clocks have been Burroughs or Sperry devices. I plan to build a 10-tube clock with hours:minutes:seconds month:day. Swapping date & time doesn't solve the poisoning risk for us 'backward-date' folks in the USA where "03/07" means March 7, rather than July 3. Going to 12 or 14 tubes so I can display the year doesn't seem necessary to me. Then again, this *is* a nixie clock, so by definition the whole contraption isn't "necessary"......

The tens-hours & and tens-month are probably the biggest problems (especially in 12-hour time format); I'm not worried about the slight imbalance in runtime that units-hours, units-days, and units-months will have. 

I'll be using direct-drive, at 5mA per tube (spec is 4-8mA). I plan to make the drive-current adjustable, so when I de-poison digits they could run at a higher current but I doubt I will go more than 10mA.

I'm looking into driving dark cathodes from the anode supply (basically a push-pull driver for each cathode), but that all depends upon how much PCB area and money I'm willing to spend on it. Putting pullup resistors on all cathodes, then driving one low, is a big waste of energy and generates unwanted heat.

Other suggestions I can use to maximize tube-life ?

BTW, the first batch of IN-18's I bought are 1991 date-code, and they cost quite a bit more. I've seen cheaper IN-18's for sale from the early 1980's that I've read have reliability issues.

[marta_kson]

If You are not going to shuffle around Your tubes, then de-poisoning will not help prolong tube life as the heavily used numbers still wear out first. If You shuffle them they will have some mechanical stress every time that is done and maybe cause a leak. Having moveable tube mounts and switch their allocations in software when moved is one solution, but will it really be done over time? People are lazy by nature...

The calendar tubes may benefit from de-poisoning as they shift a little bit too infrequent to keep their cathodes fresh. Just step thru the used cathodes at some hour when You are likely to be asleep. There are no need for fancy drivers, just let the cathodes float when not lit. The never used ones are just never used. No need to waste driver outputs on them. This is not a rocket science.

In my opinion the best way to prolong tube life is to use lowest recommended current and use PWM to reduce brightness. In the well known Weston book he states that tube life is inverse proportional to I^2.5, that means a lot of wear for just a little bit more current. He also states that the mean current when pulsed (don't remember dutu cycle) just had an impact of I^1.5. He had no definite explanation for that, just empirical data from sputtering experiments.

[Matt]

I concur. Direct drive with PWM, as opposed to multiplexing. I suggest adding a light sensor to lower the brightness automatically
via PWM. This way, as the sun changes angles, or lights get turned off and on, you are not subjecting them to any more current then
they need and they remain bright enough for the ambient light.

PWM may also prevent metal whiskers from forming by denying a long lived and continuous magnetic field that they like to form in.
Tubes with mercury, such as the IN-18, have less of a problem with this. So even if you operate them in full brightness (which I
like to do because I like brightly lit rooms during the day), they should spend some amount of time in PWM. This is probably a
bigger risk for digits that rarely change.

I have seen a lot of clocks run their de-poisoning cycle at regular intervals throughout the day, such as 3 seconds every 15
minutes. I am not a huge fan of this for cosmetic reasons. I have thought about adding a motion sensor to hold the de-poisoning
cycle until no motion has been detected for an hour or so, and then after another hour of inactivity, shut off the display
entirely. A de-poisoning cycle can be lengthened to make up for the amount of time that it has been put on hold.

It makes no sense to run through digits that go through 0-9 all day, but cosmetically, the poison-prevention sequence looks better
if all of the digits are showing the same value. But it does make more sense to cycle though the ones that stay off the most, in
proportion to how long they stay off in relation to the other digits. And if you are changing the brightness/current throughout the
day, then it might make sense to account for that too in that ratio.

Even if the clocks only display 1 or 2 in the left hours digit, usually all of the other digits are wired up to support menu
functionality or make the poison-prevention sequence look better (by displaying the same digit on every tube). Your design may not
need this. One reason to keep all of the digits functional is if you want to re-purpose those tubes for another clock in the
future. It is possible that the tubes might outlive the clock or your desire for that clock design.

I agree about minimizing the current, but to use PWM on top of that. But I am not sure if the lowest usable strike(ionization)
and/or sustain voltage will be usable as the tubes age. I wonder if per tube and/or per cathode adjustment might be needed or
desirable. They use the same current regardless, so supply voltage and dropping resistance have to be modified in unison. It
sounds like it can get very complicated very quickly, and is probably why people just design to the spec and not add any feature to
adjust this. There is some discussion about constant current power supplies. One thing that I hate are clocks that change
brightness as the digits change. While more of a problem for multi-segment digits, a most significant digit turning on and off can
cause this in a single segment digit display.

The tubes might outlive the high voltage components. I wonder if there is any risk to the tubes from one of the switch mode power
supply's failure modes. This is slightly off topic, but this discussion is about keeping the tubes running for decades.

> --
> 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
> <mailto:neonixie-l+...@googlegroups.com>.
> To post to this group, send email to neoni...@googlegroups.com <mailto:neoni...@googlegroups.com>.
> To view this discussion on the web, visit
> https://groups.google.com/d/msgid/neonixie-l/f78b0dbd-30fa-4ce9-8f3f-d11b50d11713%40googlegroups.com
> <https://groups.google.com/d/msgid/neonixie-l/f78b0dbd-30fa-4ce9-8f3f-d11b50d11713%40googlegroups.com?utm_medium=email&utm_source=footer>.
> For more options, visit https://groups.google.com/d/optout.

[gregebert]

Thanks for the info.

I'm definitely using per-tube current-regulators on the cathodes; already did this on some of my other designs and I'm very happy with it. My HV supply is +200V (linear, not switching), so I'm hoping it will give enough margin as the tubes age. I have a jumper option to boost another +30VDC if necessary via aging. The tradeoff is wasted energy (heat) at higher anode voltages. I'll have a better idea about striking & sustaining voltages when my first batch of tubes arrives next week.

Regarding PWM, since this is an FPGA design I can probably do this well above 10kHz if needed, to avoid mechanical resonance in the tubes. It then becomes a tradeoff with switching losses on the drivers, which I think are much lower than reduced power consumption of the tubes, so the net product is less energy.

[Ian Sparkes]

I'm mulling over this statement, and because I don't have a copy of Weston I'm struggling to interpret it:

In the well known Weston book he states that tube life is inverse proportional to I^2.5, that means a lot of wear for just a little bit more current. He also states that the mean current when pulsed (don't remember dutu cycle) just had an impact of I^1.5. He had no definite explanation for that, just empirical data from sputtering experiments.

I'm interested if the "I" in the two equations is the same "I" (I suppose meaning the nominal, non-PWM adjusted, applied drive current), and over what range of duty cycle this empirical rule holds, or is the second "I" (in "I^1.5") somehow the first "I" adjusted for the duty cycle?

I'm running a series of experiments to do with tube life on IN-1s (there is a separate thread on this elsewhere), and one of the observations I have is that PWM (in the form of dimming or multiplexing) does indeed increase tube life significantly. I am also experimenting with dark cathode bias, but have not got any results for that so far.

Can anyone point me to an electronic copy of the book, (or let me see a scan of the relevant section)?

[gregebert]

Perhaps there is a thermal component hiding in the underlying physics of tube wearout, such as exp(aKT), where K is Boltzmann's constant, and T is absolute temperature. Tubes will run hotter at higher current. Until now it never even occurred to me the cathode surface temperature of a nixie could be significantly higher than the glass envelope.

I'm not sure how the heat generated by a nixie tube is produced, and more importantly, where it's produced. The cathode itself isn't actually glowing; it's the ionized gas surrounding the cathode that produces the glow. Most likely the ionized gas is the hottest part of the tube, but it's close proximity to the cathode will cause it to heat.

[A. Nonamus]

As I understand it, for current to flow in a "cold-cathode" tube, the gas atoms must impact the cathode to pick up electrons, and the attraction due to the potential difference means they hit it fairly hard. So the only reason the cathodes are called "cold" is that they are not significantly directly heated by I^2R of current flowing along them, nor indirectly by infrared radiation from a heater filament. Despite the lack of those heating mechanisms, "cold" cathodes actually get quite hot from those impacts. In some nixies, when looking from the side you can see cathodes distort slightly from thermal expansion.