IV-9 numitron lifespan

[theold...@gmail.com]

I seem to recall there was some discussion on the life expectancy of IV-9 numitrons some time ago.

 

In 2015 I built a number of clocks with these tubes, and they started failing last year. The clock in the picture is on my bench now, the three good tubes were replaced in November last year. So today all 6 tubes will be replaced. I have replaced all 6 tubes on a few other clocks also. One or more segments will no longer light up, and I do not see any blackening of the glass tube (The black you see in the picture is my permanent marker, s I will not replace the wrong tube when I disconnect the power).

 

The tubes are driven directly from a 4511 chip on a 5V power line.

 

There does not seem to be any order in which they fail, so I do not have the feeling that any of the tubes are significantly impacted by thermal stresses from flashing on and off. I would say based on my experience with them, the expected lifespan is about 7 years of continues use.

 

Bill v

[Chris]

Hi,

While i am not sure for the IV9 numitrons have a look at the DA2300 lifetime expectancy,



Basically reducing the voltage increases the life expectancy exponentially.

So i usually run my numitrons at 4.5V or lower and have had no issues so far.

[theold...@gmail.com]

Hi Chris,

 

The numitrons are driven by a 4511 decoder, and according to their data sheets the max. output at that voltage is 4.1 V, so that is well within specification for the tube. Yes, I agree, lowering the voltage even further should increase lifespan.

 

Bill v

 

From: neoni...@googlegroups.com <neoni...@googlegroups.com> On Behalf Of Chris
Sent: Saturday, March 04, 2023 8:19 AM
To: neonixie-l <neoni...@googlegroups.com>
Subject: [neonixie-l] Re: IV-9 numitron lifespan

 

Hi,

While i am not sure for the IV9 numitrons have a look at the DA2300 lifetime expectancy,

Basically reducing the voltage increases the life expectancy exponentially.

So i usually run my numitrons at 4.5V or lower and have had no issues so far.

On Friday, March 3, 2023 at 8:05:48 PM UTC+1 theold...@gmail.com wrote:

I seem to recall there was some discussion on the life expectancy of IV-9 numitrons some time ago.

 

In 2015 I built a number of clocks with these tubes, and they started failing last year. The clock in the picture is on my bench now, the three good tubes were replaced in November last year. So today all 6 tubes will be replaced. I have replaced all 6 tubes on a few other clocks also. One or more segments will no longer light up, and I do not see any blackening of the glass tube (The black you see in the picture is my permanent marker, s I will not replace the wrong tube when I disconnect the power).

 

The tubes are driven directly from a 4511 chip on a 5V power line.

 

There does not seem to be any order in which they fail, so I do not have the feeling that any of the tubes are significantly impacted by thermal stresses from flashing on and off. I would say based on my experience with them, the expected lifespan is about 7 years of continues use.

 

Bill v

--
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 on the web, visit https://groups.google.com/d/msgid/neonixie-l/81b5b7e6-d476-4b83-9a8e-c9d6cc468db9n%40googlegroups.com.

[gregebert]

The picture above confirms my suspicion that more-frequently-changing digits will die first. The dying hours tube is still a mystery to me; maybe it was previously used ? I learned about this as a kid when I would do stupid things like turning my room light on-and-off to annoy others. Yeah, my bulb always died pretty quickly...

There are 2 design enhancements you can do to prolong filament life. First, limit the inrush current by using a simple current regulator or drive from a higher voltage and add some series resistance. For example, if a filament typically runs at 40mA for a 5V supply, it's HOT resistance is 5/0.04 = 125 ohms. I've seen cold resistance 4-5 times lower, so such a bulb could easily see surge-currents of 200mA. Now, if you drive the same filament from a 12V supply, and add a series resistance of (12-5)/0.04= 175 ohms, the peak surge current has been reduced to 60mA. This will reduce thermal shock, but not thermal cycling.

For rapidly-changing digits, such as the seconds, you may want to have all segments slightly energized just below visibility to reduce thermal cycling. I would guess somewhere around 10-20% of normal operating current is a good place to start.

What about PWM ? Well, that's a good question and I suppose if the frequency is high enough, say hundreds of Hz, there wouldn't be any added thermal cycling.  But you would likely hear some audible noise. Faster still, say 20-30 kHz, wont be audible but might have some unwanted harmonics that cause RFI in the AM band.

[Moses]

The TI CD4511B datasheet lists the lists the *minimum* high level output (at 5v VDD) of 4.1v but typical is 4.55v, maybe others are different?

https://www.ti.com/lit/ds/symlink/cd4511b.pdf

At 4.55v the segment current on the IV-9 is going to be about 25ma. Typical IV-9 segment current is about 20ma I believe, so that would be a 25% overdrive.

I've put up a drive voltage vs segment current table on my site, maybe it's useful to someone: http://www.neonixie.com/IV-9-6D-RR/#tech

I settled with powering the tubes in my designs in one of two ways.. 5.0v direct with a recommended PWM of about 65% (the 165 (out of 255) value in the table) OR 5v through a regular silicon diode, dropping the voltage to about 4.2v at the tube and PWM that to 76%. Both result in a segment current of 20ma. Don't know if one or the other is better in regards to lifetime, I was not able to find any official sources of information. I experimented with driving them using a 3.3v regulator.. but at that voltage they were a bit dim.

NOTE: The IV-9 datasheet does mention to avoid running the tubes at a frequency between 105 and 1000 Hz, presumably to avoid mechanical resonance. I elected to run them at 1600 Hz.

I looked through some of my testing data.. at a segment current of 3mA the segments are visible, 2mA they are "barely visible" and at 1mA I noted them as not visible. Keeping the unlit segments warm with a slight current seems to make the most sense to me. It's all software on my clock so maybe an option for the next revision.

That's all I know so far.

Regards,

-Moses

[gregebert]

Very good info, Moses.

From that, I would target the  operating current well-below 20mA to prolong the life of the display. You could use an NPN current-limiter (driver) driven from 5V TTL logic. Let me know if you need circuit details, but it's a very simple design (1 NPN + 1 emitter resistor). As a starting point, a 270 ohm emitter resistor will limit current to 16mA.

Do you hear any noise with your PWM running at 1600Hz ?  If not, I would stick with PWM and not use the NPN driver, as it's not necessary. As long as the PWM cycle-time is shorter than the thermal time-constant of the filament (probably a few milliseconds), you should be fine. I think you can measure the thermal time-constant of the filament with a phototransistor and a scope. Starting at a low PWM frequency, the phototransistor will show fluctuations in the bulb's intensity. As the frequency is increased, you should see the amplitude decrease towards zero. I've never actually done this but I'm pretty sure it will work. You have to use a phototransistor (or photodiode); a CdS photosensor will have too-slow of a response time.

[theold...@gmail.com]

Using a current limiter sounds like a LOT of circuitry, 7 times the number of tubes? I simply use a forward biased diode to drop the voltage, about 0.7 for a standard silicon diode, and about 0.3 for a Schottky. That requires only one or perhaps two diodes for the whole set of tubes. Perhaps not as accurate, but definitely a whole lot simpler.

 

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/13ceb519-9b4d-469e-9e03-8d844e317239n%40googlegroups.com.

[Benoit Tourret]

Thats the case on Mose's clock: one (bypassable) diode driving all the "com" pins

[Dekatron42]

There are a few different datasheets at the bottom of this page: https://danyk.cz/avr_num_en.html and they all seem to mention the filament voltage starting at 3.15V - how much would this increase lifespan?

/Martin

[gregebert]

Under "normal" circumstances, I agree that individual current regulators add a lot of circuitry. The reality is that most tubes are no longer manufactured, so their numbers are dwindling as the price increases, essentially irreplaceable. You wouldn't believe the lengths I went towards protecting NIMO tubes in the clock I'm still working on.

I've debated about building a numitron clock, and the route I think I will go with uses LED 'filaments' that are used in retro light bulbs. Filaments scare the heck out of me because they will fail from thermal cycling, but there is no alternative for NIMO and VFD displays.

[martin martin]

How about this clock I built in the early 80s? 5314  clock chip.  I used a 1k pot for dimming.  Has been running for  many moons now!

~

mcve...@gmail.com

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/e87e8845-5e21-4b9d-9408-49ee93a3578dn%40googlegroups.com.

[theold...@gmail.com]

Martin,

 

How did you do that? (schematic)

 

Problem with a pot (or any resistor) is that you would need one for each element (7 per tube) since you cannot put one in the common line. Each filament draws about 22mA, so depending on the number displayed, the current would be anywhere between 44 and 154 mA. This would result in different voltage drops depending on the number displayed. It would result in uneven light output. Same reason why you could not use a single current regulator pre tube, but one per filament would be needed. A voltage regulator (such as a simple diode) does not have that disadvantage.

 

Bill v

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/CAKYv7NTaviwTXv09wHvzZ3mHt5PFK-5GRSRBArjLtmKp78GekQ%40mail.gmail.com.

[Nicholas Stock]

If I'm not mistaken, the clock kit that Martin is referring to is this one..

http://rrsignal.com/projects.htm

I have the schematic at home somewhere and will dig it out. There is a dimming mod using a transistor/pot that works quite nicely. Mine has been running for about 8 years now with no issues..... it uses the MM5314 clock chip, so no software to be seen and is mains referenced, but keeps remarkably good time over long periods (it sits next to a few GPS disciplined clocks).

Cheers,

Nick

On Tue, Mar 7, 2023 at 7:18 AM Batareyka <batar...@gmail.com> wrote:

There are two ways to solve this problem.
1. Software, but as the author of the topic, there is no software and the possibility of editing it (I
  I think so), then method number 2 remains.
2. The 4511 chip has a wide power supply range from 3-16 volts. Put the power regulator and turn it up to 4 volts, which will change the output power of the lamp filaments accordingly.
Checked more than once.

вторник, 7 марта 2023 г. в 17:02:08 UTC+2, theold...@gmail.com:

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/1d59be9a-50d3-4fa4-b783-f349b6be7d1dn%40googlegroups.com.

[Dekatron42]

A quick Googling of MM5314N and Numitron finds this webpage: http://www.tuberadios.com/numitron/ with schematic that is said to be compliments of Westdave.

/Martin

[Nicholas Stock]

That's the one. Thanks Martin. I've spoken to Richard about getting more PCBs made (I have some MM5314's that need a good home) but he can't locate the board files.

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/9d522a7e-0a80-4bf2-9f5b-854289216bccn%40googlegroups.com.

[theold...@gmail.com]

Not as easy as that sounds.

 

If you want to use software (I presume PWM) you cant use a 4511 anymore since their greatest advantage (besides the BCD to 7 segment decoding) is its latch function. No way to implement any software based dimming. This would require a very drastic hardware redesign.

 

Turning the 5V line down would work, but may cause issues with the rest of your circuit. A dedicated supply for the 4511’s could work, but depending on the final setting, there would be a discord in logic levels between the rest of the circuit and the 4511. Nasty, could result in unreliable communications. Level shifters could fix that, but wow, we are getting complicated again.

 

The diode is still the best solution in my mind. Working with silicon and Schottky diodes you can obtain steps of 0.3V, 0.6V, 0.9V, 1.0V, 1.2V etc. Some diodes have slightly different forward voltage drops, so with a bit of searching you can obtain a lot of very simple options. Using jumpers of DIP switches you can even make it very simple to adjust. Why all the complicated solutions?

 

Bill v

 

From: neoni...@googlegroups.com <neoni...@googlegroups.com> On Behalf Of Batareyka
Sent: Tuesday, March 07, 2023 10:19 AM
To: neonixie-l <neoni...@googlegroups.com>
Subject: Re: [neonixie-l] Re: IV-9 numitron lifespan

 

There are two ways to solve this problem.

1. Software, but as the author of the topic, there is no software and the possibility of editing it (I
  I think so), then method number 2 remains.
2. The 4511 chip has a wide power supply range from 3-16 volts. Put the power regulator and turn it up to 4 volts, which will change the output power of the lamp filaments accordingly.
Checked more than once.

вторник, 7 марта 2023 г. в 17:02:08 UTC+2, theold...@gmail.com:

Martin,

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/1d59be9a-50d3-4fa4-b783-f349b6be7d1dn%40googlegroups.com.

[theold...@gmail.com]

If you send me the schematic (with any modifications you may want) and preferred board size and shape, I can create board files for you.

 

Bill v

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/CAOX%2BRH%2BXP9qskehCnadpeQ%2Bws6z%3D8iHGHE%2BTvgzEiH6PVdTWhQ%40mail.gmail.com.

[Dekatron42]

This RCA datasheet shows their simple voltage regulator on page 6 for brightness control and it also shows the multiplex design with diodes.

https://www.one-electron.com/Archives/RCA/RCA-AppNotes/RCA%201971%20AN-4277%20Description%20and%20Application%20of%20NUMITRON%20Devices.pdf

/Martin

[Moses]

I agree with Bill. I like a low part count.. the lower the part count the greater the ultimate reliability.

The diode drop is a good idea and is what I used along with PWM of high power '595 shift registers. I thought about adding a separate DC-DC converter for just the tubes.. but that would have been another half dozen parts, and it's not 100% efficient, so it would not have saved too much power vs just a plain old diode burning off some voltage.

I'm going to definitely try to see if I can keep the segments warm in software, as everyone agrees the thermal shock of constantly changing segments is not ideal.

A side by side test of some overdriven tubes, one with warm segments and one without, may be an interesting test.

Regards,

-Moses

[Moses]

Gregebert,

I've not heard these tubes at all. I was initially running them at 100Hz, then changed to 400Hz for a while before settling at 1600Hz.

I've got half a dozen prototypes running for several months, so if they were going to make noise I would have heard them by now.

Neat idea on measuring the time-constant, I'll give it a try.

Regards,

-Moses

[gregebert]

Very interesting. Others who have multiplexed nixie tubes, which run at a lower current, have reported audible noise. I would expect more noise at higher currents due to magnetism effects, but that isn't happening here with your numitrons. I've heard it myself with incandescent bulbs; I havn't done any work with numitorns.

[theold...@gmail.com]

I just actually realized I was totally wrong with my earlier statement about the 4511. If you look at the truth table for this chip you see it has a blanking input (pin 4, ^BI). That would make it actually very easy to implement software dimming by simply tying all of them to a mpu pin coded with PWM. On my own clock this line is already connected to the PIC, so I should be all set to implement dimming. Just never thought of using it that way.

 

Amazing where a bit of discussion can lead you!

 

Bill v

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/000e1347-33db-4054-b90c-26411fa884can%40googlegroups.com.

[theold...@gmail.com]

I just actually realized I was totally wrong with my earlier statement about the 4511. If you look at the truth table for this chip you see it has a blanking input (pin 4, ^BI). That would make it actually very easy to implement software dimming by simply tying all of them to a mpu pin coded with PWM. On my own clock this line is already connected to the PIC, so I should be all set to implement dimming. Just never thought of using it that way.

 

Amazing where a bit of discussion can lead you!

 

Bill v

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/000e1347-33db-4054-b90c-26411fa884can%40googlegroups.com.

[Moses]

I got around to adding a "keep warm" function for unlit segments on my IV-9 clock kit. My experiences and a short video for those interested.

I thought this would be a quick and easy modification.. it was anything but! I added a hardware interrupt at 1600hz, and added code to keep unlit segments warm with about a 1/10 duty cycle. Worked well.. but seemingly random reboots of the ESP32 micro drove me nuts, I chased a non-existing software bug for more then an hour (much more). Later (much later!) I realized that the load imposed by largish changes in brightness, either on power up or a commanded brightness change, combined with the keep warm function was a pretty big load on the power supply. Yes, even higher then just turning all tubes and segments on all at once. I tried different power cables, different clocks, half a dozen different power supplies from boat anchor linears to modern switchers, I was able to trip up all of them. I even have a very low ESR polymer cap on the clock as well, but nope, the spike load was just too much.

My conclusion is the first few moments of the keep warm feature at 1/10 duty cycle, on 6 tubes and all segments, is a greater load then just turning on all segments normally. I'll have to hook up some proper equipment to quantify this soon. I could have brute force fixed this in hardware by throwing a bunch of caps at it.. but I went with the more elegant software solution. The brightness is no longer directly controlled, but ramped up and down (on bootup too) to match the requested brightness, this fixed the problem entirely.

With a normal lit segment being driven at 20mA, the 1/10th duty cycle gives the unlit segments approx 2.5mA current, which is barely visible. I provided a user adjustable setting for the duty cycle, but having no official manufacturer guidance to go on I think 2.5mA is as good a default value as any.

I hope this saves someone 4 hours of profanity ridden shaking of fists at a clock.

Side note, while I was in there.. I added cross fading on the digits, which I must admit looks better then I had imagined on a numitron clock. See the video and be well!

https://youtu.be/dkOwaZvVQrg

Regards,

-Moses

[gregebert]

Inductance can be a real pest; V=Ldi/dt , so switching a bunch of segments on or off at the same time (even if they are lit dimly) can cause problematic voltage-drops across a relatively small inductance. One trick I've done on IC's is to stagger the turn-on and turn-off of current-consuming circuits by a few microseconds; can also be done in software.