EEVblog #948 - Nixie Tube Display Project - Part 1

[Spuddevans]

#1 - One page mentioned something about AC power - can nixies be driven with an ac?

You can use half wave rectified AC, but I don't think you can use pure AC.

#2 - I read that is isn't wise to have it turned off - maybe this was a reference to older nixie type drivers though.  With a newer style higher voltage driver with zener's like has been talked about here, is it ok to have all segments off?

As long as your driving circuitry is fine with seeing the main HV voltage (170-200v), or you have a voltage clamp system to limit the higher voltage to one that is safe enough level to not exceed your driving circuit, then it's ok to have all segments unlit.

#3 - What is the best way to have an adjustable brightness for nixies?  Given you have a fixed resistor to limit current...

There may be a number of ways, more experienced Nixie builders will hopefully chip in, but while you have a fixed current limit resistor, it may be possible to design a means of varying the HV voltage, and thus varying the current drawn by the Nixies, and thus varying the brightness. (though I suspect that this method would have a limited range of brightness, nowhere near the range you would get using PWM and LED's)

Tim

[alank2]

Thanks Tim!

[Macbeth]

You can use half wave rectified AC, but I don't think you can use pure AC.

If I recall the IN-xx Russian tubes are indeed designed to run using simple un-filtered/un-smoothed half wave 240VAC, and this should lead to longer life than a 170VDC supply?  (Cathode poisoning?)

Direct rectified mains is of course quite insane for a hobbyist but as bigclivedotcom shows pretty much all the great unwashed who enjoy pound shops/dollar stores have plenty of devices around that can injure or kill just to shine a pretty LED. I haven't seen a rise in deaths from them, but they are often shit like a ripoff 2A Samsung knock-off charger I had went up in magic smoke.

Using proper design I would be happy to use half wave from the mains providing all the normal precautions are used X/Y capacitors, PTC, slots in PCB etc. The Nixies PCB should be isolated by opto-isolators too, though wifi and bluetooth modules are now so cheap why bother?

Low current so just use two 240/12 transformers back to back for safety during development. Switch to direct mains for final implementation. 

[b_force]

Why are you using TPIC6C595 (in your component library drawing video)? They have a breakdown voltage of only 33V instead of 50V for A and B variants and have a different pinout!

They have clamping zeners built in.
It will have to be the 50V B series though, otherwise the digits would bleed on with the 30V C series.

The TPIC6B596 or the TPIC6B595?
There is even a TPIC6A595, but that one is a bit more expensive.
Nice packages and easy all in one solution, but pretty big packages though, 96mm²
You will need 11 of them (for 88 connections), which is 1056mm²

Just pure on board space/total surface area;
44 dual BJT arrays (like the PMBTA42DS/MMDTA42) incl some resistors (which you practically can put anywhere you want) uses less board space for 88 connections.
The BJT arrays together are only 198mm², which is equivalent to two HV shift registers.

[alank2]

Those SmartNixie boards shown above use BSS131's for switching.

In terms of dimming, if you were NOT multiplexing, could you simply adjust the on time in milliseconds?  If you were doing 60 Hz, that would be 16.67 mS.  You could leave it lit for 1 mS out of 16...

Also, for cathode poison prevention, can you light the unused digits periodically for small run times so they are not on long enough to be visible.  Would that be sufficient?

[FxDev]

Why nobody thinks about convert high voltage with using microcontroller   

We have very good uCs today to convert voltage, so don't play with dimming issues, just decrease the voltage from 170V to for example 130V. Or you can convert it high value.

[SNGLinks]

I hate the look of the 5 digit.

This is much better.

[Towger]

Why nobody thinks about convert high voltage with using microcontroller   

We have very good uCs today to convert voltage, so don't play with dimming issues, just decrease the voltage from 170V to for example 130V. Or you can convert it high value.

That's what this design does, uses an MEGA328-P (arduino) to multiplex one 74141/K155ID1 to drive 6 tubes and at the same time control a boost converter for the high voltage.
http://www.open-rate.com/Downloads/NixieClockInstructionManualRev4V42.pdf
http://www.open-rate.com/Store.html

[robertbaruch]

I agree with some of the previous posters: The 50V clamp is very iffy. If the maintaining voltage is 120V-170V, then when a digit is on and you attempt to turn it off, having a 120V differential across the tube isn't guaranteed to turn it off.

This was my circuit. I use a MAX6922 VFD driver [datasheet] with its output set to 60V for "off". That gives the tube 110V across, which is low enough to guarantee turn-off. You don't need an output clamp: the  voltage at any output will be driven to either 0V or 60V.

Granted, it's a PLCC, which isn't exactly breadboard-friendly. The HV518 [datasheet] is equivalent in DIP package (and can go up to 90V, so you have an even better safety margin).

[EEVblog]

I agree with some of the previous posters: The 50V clamp is very iffy. If the maintaining voltage is 120V-170V, then when a digit is on and you attempt to turn it off, having a 120V differential across the tube isn't guaranteed to turn it off.

I'm using 160V in my design with 50V clamps for this reason. You'll see an update and quick test of this in my next video.

[bjrn]

Very nice project with a good roundup about the specific parts and details for this clock . I will definitely follow all of your videos and posts regarding this project.

I also build myself a nixie clock based on the IN12b for my bachelor thesis over a year ago. Never had the time to do a little wrap up about the build. I also went with the ESP8266, which connects to my wireless network and gets the time via NTP, as a main controller, some 595 shift registers and discrete MMBTA42 transistors and a simple switchable high voltage step up power supply with a cheap MC34063. It is okay for only four tubes. Your solution seems to be much better for a higher number of nixie tubes. As the eagle layout designer was used as freeware, I was limited to the board size and I also wanted to "manufacture" the PCBs with the toner transfer method on my own at home. So I decided to make a one sided, modular design. The power supply or controller could be swapped in future, if I want to upgrade it or make it more efficient.

The PCB quaility of the main display module is not as good as the one for the ESP. A laminator does the work much better, than a normal iron with uneven pressure. I hope I will have some time to improve the layout and make some new PCBs again, but it works and the glow of the tubes looks so nice in contrast with modern computers and stuff on my desk.

[EEVblog]

PART 3:

[NivagSwerdna]

Keeping the elements OFF once they have been ON seems to be the challenge (since their sustain voltage is much lower than the initial conduction voltage) so the first test probably wasn't that relevant but the subsequent test of turning the 7 on and off seems to prove that at 160V with 50V clamp is do-able.  Might be worth repeating that second test with all digits in turn.

[twam]

Dave, if you use D5 of the ESP8266 module for your CLK and D7 for the data you could use its build-in HW SPI the get data out. As you're not using the pins for anything else, there's nothing to loose

[AustinTxBob]

It looks like all of the Nixie drivers are getting the same data fed to them.  How is the appropriate data being pushed to each one?

[bktemp]

It looks like all of the Nixie drivers are getting the same data fed to them.  How is the appropriate data being pushed to each one?

They are connected in series: All data get shifted into the first ic. After 8 clocks, the first bit appears at the output of the first ic and enters the second ic.
So for the 11 ics 88bits need to be clocked out. The first 8 bits end at the last ic. The last 8bits at the first ic.

[AustinTxBob]

It looks like all of the Nixie drivers are getting the same data fed to them.  How is the appropriate data being pushed to each one?

They are connected in series: All data get shifted into the first ic. After 8 clocks, the first bit appears at the output of the first ic and enters the second ic.
So for the 11 ics 88bits need to be clocked out. The first 8 bits end at the last ic. The last 8bits at the first ic.

So 88 clock pulses (DCLK?) and then the other clock (RCLK) is pulsed to push all of the data out to the tubes?

[bktemp]

So 88 clock pulses (DCLK?) and then the other clock (RCLK) is pulsed to push all of the data out to the tubes?

Exactly.
Because all the data get shifted through all shift registers, having the additional latch/register is really useful, otherwise wrong data would appear on the outputs while shifting.

[robertbaruch]

With regard to the voltages... well, if it works, it works. I do like that Dave validated that 120V wasn't sufficient to turn off the digits, and that 160V was sufficient to turn them on. Still, I'd rather see good engineering practice and conform to both the turn-off voltage (<120V) and turn-on voltage (>=170V). But again, like I said, if it works, it works.

[robertbaruch]

BTW, I really enjoy Dave go through the design process. I think this should be a regular feature: "Dave builds X".

[Pentium100]

This part got me a bit confused. With the two zener diodes arranged like this and a sufficiently high voltage across the output, wouldn't the gate of the transistor be at 20V? Well, if the gate is at 20V then the transistor starts to conduct and the voltage would drop, but wouldn't this result in the transistor being partially open?
I am most likely missing something, just cannot figure out what since obviously the chip has to work.

[bktemp]

This part got me a bit confused. With the two zener diodes arranged like this and a sufficiently high voltage across the output, wouldn't the gate of the transistor be at 20V? Well, if the gate is at 20V then the transistor starts to conduct and the voltage would drop, but wouldn't this result in the transistor being partially open?

The mosfet will start conducting at a much lower voltage than 20V. Normally you wouldn't need the 20V Z-diode because the mosfet will conduct and limit the drain voltage whenever it goes above gate threshold voltage (probably somewhere between 1V and 4V) + z-diode voltage. The 20V z-diode protecting the gate is only needed for fast transients being applied to the drain.
So yes, the mosfet will be conducing slightly, but only as much as needed to keep the drain voltage below 33V. As soon as the voltage goes below 33V, it stops conducting.
Basically the mosfet acts as a power amplifier for the z-diode increasing its power handling capabilities.

[alank2]

Really enjoying these videos - good job Dave!!!

[Pentium100]

So yes, the mosfet will be conducing slightly, but only as much as needed to keep the drain voltage below 33V. As soon as the voltage goes below 33V, it stops conducting.
Basically the mosfet acts as a power amplifier for the z-diode increasing its power handling capabilities.

That's interesting. But doesn't this mean that the clamp voltage is around 33V (the value of the "top" zener) and not 50V (both zener diodes combined)?

[niekvs]

The SN75468 indeed seems like a much better choice (together with a 90V clamping zener). That way, the design will have a lot bigger margin and not depend on these specific types of Nixie tubes. E.g., say someone wants to use this same design but use different nixies, the 50V clamp may well fall short. And since the SN75468 is also easily available, I don't see a reason to use the part with the built-in zeners. You can then also just use the datasheet strike voltage of 170V or 180V without a problem, rather than lowering it to 160V "because it works with these specific nixies I have in hand". It's better to have a comfortable margin.