Dummies guide to building a 120watt 140 volt power supply to drive IN-9 Nixie tubes.

[spec]

I like the above circuit.
This is more in line with my idea of driving with OpAmp's.

This way I can apply the ideas in this sheet about offsets to zeroing the tube etc.

Edit: this pdf also drives with PWM in figure 2 ....

Trouble is the arduino's don't have enough PWM outputs.

This schematic from your posted pdf is exactly the way I was suggesting the current generator transistors should be driven with an opamp. This will allow you to control the ionizing current to the n th degree.

From what I have seen so far, you should be able to lead the way with absolute precision control of the CC tube current using a micro and the circuit would should be relatively simple too.

 

[fireant007]

Ok.... we are agreed.

Components.

Transformer.
Opamp.

Procedure.

Breadboard.
Test.
Fix stuff ups in design.
Test again.
Fix more stuffups ... probably in the code this time.
Test again.

Design/Build verro board version.

Find more errors .. coz I didn't think about some crosstalk problem or something else like a ground loop etc in the layout.

Stop drinking and building.... it is making it harder ...

Get an eagle board design out.

Throw them in the bin... cos I forgot to add some part or some part won't fit !

Pull my last 3 remaining hairs out.

Have a little rest.

Start again.

Victory !

Complete case.

Make bloody amazing viral youtube video and lie through my teeth to the world ... "this stuff is dead simple if you know what you are doing"

Make son and daughter proud !
They just want to see something of mine on youtube

P.S. .... I have just ordered this scope for this project.

Anyone think much of Rigol ? Looks ok ... I am only used to old analog ones ?
https://www.ebay.com/itm/Rigol-DS10...518700?hash=item3d0dcc9e2c:g:2m8AAOSwKtlWiuiX

 

[spec]

"The striking voltage is highly dependent of temperature too..

Really ? .. why ?

Another unknown fact is that any nixie requires a tiny bit of stray radiation (light, cosmic rays, etc.) to pre-ionize the neon gas, and strike.
I have noticed that a lot of tubes won't light straight away out of the box.

It is all to do with molecule excitement which is proportional to degrees Kelvin (0K= -273 C): the hotter the more active and thus the easier to start the ionization. As you say, electromagnetic radiation, including light, can help start ionization. I had a laptop where the cold cathode fluorescent lamp (CCFL) LCD backlight had aged and could only be started by heating the laptop screen with the wife's hair dryer (when she wasn't looking that is).

Yes, that is right about a new CC tube being difficult to start. They will often need a higher striking voltage to get going. After that the CCFL will then be to specification. I cant remember why this is though.

 

[fireant007]

Just one more little clue about what this thing is going to do.
1. Tell the time.
2. Measure barrometric pressure.
3. Be a high tech disco ball.
4. Use an ST-21 Geiger muller tube .... (decatron ?)
5. Have a couple of paddles .. for restarting your heart.

 

[fireant007]

Hi again fire,

here is a transformerless CC display PSU which can be switched for 110V and 240V mains supplies. It is non isolated so the control signals for the tube current driver transistors would need to be isolated or the whole system referenced to the voltage rails.
For 110 V operation only, the two BJTs and eight MOSFETs would not be required.

spec

View attachment 98348​

ERRATA
(1) R8 should be 4K7 not 22R
(2) For every 10 of CC display tubes (CCDTs) add another 100uF capacitor across the CCDT supply line, in parallel with C3 and C4 as appropriate.

NOTES

Hey Spec,
Was wondering about this circuit.
Why have you used 8 Mosfets ? and mixed N and P types ?
Can't you do it with just 2 big fat ones.

400volt P channel ones at 23watts seem to be rare ?

 

[spec]

Hey Spec,
Was wondering about this circuit.
Why have you used 8 Mosfets ?
Can't you do it with just 2 big fat ones.

400volt P channel ones at 23watts seem to be rare ?

Hy fire,

Because you obviously know about electronics, the circuits I have posted are only engineering outlines and have not been optimized.

The eight MOSFETs are just what I think can do the job without me having to do any thinking (I haven't even checked for suitable types of MOSFETs).

Sure, if suitable single MOSFETs can be found to dissipate 92W at 160V and withstand 400V that would work fine and would simplify the circuit. In fact, bipolar junction transistors or, more likely, IGBTs may be more suitable.

A couple of things to consider are:
(1) secondary breakdown and safe operating area (SOA) is often the limiting factor with power transistors rather than the data sheet max power rating which is a fairly meaningless parameter and is only really there for marketing purposes.
(2) The other limiting factor is maximum junction temperature: same comments as (1) above about transistor power rating.
(3) 92W is a lot to dissipate with one device and would require a huge heat sink with a low thermal resistance to air. So it would really be a trade-off between more transistors or a bigger heat sink. The more transistor approach is liable to be smaller and cheaper, but I haven't checked.

It would be possible to alter the circuit to use a single sex power transistor- NMOSFET for example.

Incidentally, the transistors would only dissipate 92W if one bank of 50 tubes were all off and the other bank of 50 tubes were all fully turned on with 12mA each.

spec

 

[spec]

Just one more little clue about what this thing is going to do.
1. Tell the time.
2. Measure barrometric pressure.
3. Be a high tech disco ball.
4. Use an ST-21 Geiger muller tube .... (decatron ?)
5. Have a couple of paddles .. for restarting your heart.

Quite a comprehensive list of functions. My advice would be to produce a universal control interface that allows you to control each tube current individually with accurately and and with high resolution. Then all the fancy stuff can be done in software.

Processing speed is liable to be critical - have you investigated the complete range of processors: Raspberry Pi and Beagle Bone Black, or even a dedicated microcontroller with a multiplexed fast digital to analog output.

 

[spec]

About the tube PSU.

A switch mode power supply with the following specification would be a simple design exercise for an experienced SMPS engineer, but would require more work than the mains transformer approach. Also, you would probably have to wind/modify a transformer.

(1) Input/output isolation: yes
(2) Input voltage: 100V to 250V
(3) Input voltage frequency: DC to 500Hz
(4) Output voltage 165V DC
(5) Output current 0A to 1.5A

spec

 

[spec]

About a suitable opamp;

The LM324 quad opamp is probably the cheapest for the current to voltage converter function.
https://www.ti.com/lit/ds/symlink/lm124-n.pdf

spec

 

[spec]

Anyone think much of Rigol ? Looks ok ... I am only used to old analog ones ?
https://www.ebay.com/itm/Rigol-DS10...518700?hash=item3d0dcc9e2c:g:2m8AAOSwKtlWiuiX

Rigol stuff is nice.

My Son has a Rigol DS105- 42 for home electronics work (he is a software designer) and is very pleased with it. I had a little play with it too, when he brought it around about a year ago.

I'm old fashioned and have got a Tektronix 2235A

 

[fireant007]

Hy fire,

Because you obviously know about electronics, the circuits I have posted are only engineering outlines and have not been optimized.

The eight MOSFETs are just what I think can do the job without me having to do any thinking (I haven't even checked for suitable types of MOSFETs).

Sure, if suitable single MOSFETs can be found to dissipate 92W at 160V and withstand 400V that would work fine and would simplify the circuit. In fact, bipolar junction transistors or, more likely, IGBTs may be more suitable.

A couple of things to consider are:
(1) secondary breakdown and safe operating area (SOA) is often the limiting factor with power transistors rather than the data sheet max power rating which is a fairly meaningless parameter and is only really there for marketing purposes.
(2) The other limiting factor is maximum junction temperature: same comments as (1) above about transistor power rating.
(3) 92W is a lot to dissipate with one device and would require a huge heat sink with a low thermal resistance to air. So it would really be a trade-off between more transistors or a bigger heat sink. The more transistor approach is liable to be smaller and cheaper, but I haven't checked.

It would be possible to alter the circuit to use a single sex power transistor- NMOSFET for example.

Incidentally, the transistors would only dissipate 92W if one bank of 50 tubes were all off and the other bank of 50 tubes were all fully turned on with 12mA each.

spec

Errr.. I know how to fix stuff.... I would not say I "Know" a lot !
Think you are right about the heatsink ... one point source of heat coming from a fat transistor would be harder to dissipate over a large heat sink using multiple devices.

Would love your input on a single sex.. N type Mosfet circuit.
I assume you would just chop out D3 and get rid of all the stuff at the bottom making the emitter of Q10 the common negative ? C1 Negative would also connect to this rail ?
Q10 biasing would be the same ? ... maybe ?

I think I have to go this way as the transformer is 2.2kg's .. and too dang big. .. and expensive.

Thanks for all the tips on picking a device .. mosfet of otherwise. No degree here. First I had ever heard of IGBT's ... looking at them now ... and they look GOOD !

 

[fireant007]

Quite a comprehensive list of functions. My advice would be to produce a universal control interface that allows you to control each tube current individually with accurately and and with high resolution. Then all the fancy stuff can be done in software.

Processing speed is liable to be critical - have you investigated the complete range of processors: Raspberry Pi and Beagle Bone Black, or even a dedicated microcontroller with a multiplexed fast digital to analog output.

Universal control interface is just coming straight out of the Arduino ... (it's just what I know) .... lots of pins on the MEGA and fancy stuff must be done in software ... that's a given.
Multiplexed DAC's probably don't have idiot proof programmers .. or a great open source community and I also know multiplexing regular Nixies has it's on fair share of problems.
Processing speed of the mega is 16Mhz .... and interrupts can always be written to keep an eye on the vitals !
If it does not work with a Arduino mega . I might go down the Pi way ... though I know naff all about it.

 

[spec]

Think you are right about the heatsink ... one point source of heat coming from a fat transistor would be harder to dissipate over a large heat sink using multiple devices.

You are on the right grounds but it is a bit more involved than that. With multiple devices you could run the heatsink at a higher temperature without exceeding the maximum junction temperature of typically 175 deg C which is the controlling factor.

I did a couple of calculations and a single devices (one NMOSFET and one PMOSFET) would need a single heat sink of around 0.75 degC/W. That would be a big expensive heat sink. You could probably get away with a single device on safe operating area grounds though. On the other hand, with four devices, a single heat sink of 4 * 0.75 degC/W = 3 deg C/W would probably do. This would be smaller and cheaper.
spec

 

[fireant007]

Rigol stuff is nice.

My Son has a Rigol DS105- 42 for home electronics work (he is a software designer) and is very pleased with it. I had a little play with it too, when he brought it around about a year ago.

I'm old fashioned and have got a Tektronix 2235A

Wow .. I think that scope is a 1Mhz model I used to fix TV's in the very early 1980's not storage from memory ?
Would still like to have a good old fashioned analog ... I think they show the real picture.

Yes . I bought the 4 input Rigol for looking at logic lines for software development too.
A dedicated rigol signal analyser with scope didn't get good reviews .. and you have to use either the scope or the analyser .. not both together. 4 channels is enough to look at for debugging methinks.

 

[spec]

Universal control interface is just coming straight out of the Arduino ... (it's just what I know) .... lots of pins on the MEGA and fancy stuff must be done in software ... that's a given.
Multiplexed DAC's probably don't have idiot proof programmers .. or a great open source community and I also know multiplexing regular Nixies has it's on fair share of problems.
Processing speed of the mega is 16Mhz .... and interrupts can always be written to keep an eye on the vitals !
If it does not work with a Arduino mega . I might go down the Pi way ... though I know naff all about it.

I know about processors etc, but have no specific knowledge of any modern micro systems like the Arduino, Pi etc. It is something I have intended to do for the last 20 years or so. One of the secrets with micro control is to do as much of the fast low level stuff in hardware and present the micro with a rationalised control inter face. In the case of the tube software all the programmer would want to see is which tube out of the 100 available and how much current say between 0mA and 12mA in 12/256 (8 bits) increments.

That way the programer would be relived of all the low level stuff and can concentrate on the artistic side of displaying various patterns on the tubes.

The other side that the programer needs to handle is the input signal which will define which out of a set of tube routines is required to execute.

There are heaps of user sites on the net for the various micro systems and there are experts on ETO too.

spec

 

[fireant007]

I know about processors etc, but have no specific knowledge of any modern micro systems like the Arduino, Pi etc. It is something I have intended to do for the last 20 years or so. One of the secrets with micro control is to do as much of the fast low level stuff in hardware and present the micro with a rationalised control inter face. In the case of the tube software all the programmer would want to see is which tube out of the 100 available and how much current say between 0mA and 12mA in 12/256 (8 bits) increments.

That way the programer would be relived of all the low level stuff and can concentrate on the artistic side of displaying various patterns on the tubes.

The other side that the programer needs to handle is the input signal which will define which out of a set of tube routines is required to execute.

There are heaps of user sites on the net for the various micro systems and there are experts on ETO too.

spec

Yup.. agreed.
it's called software layer abstraction !
You should look at the arduino .. it's really very easy for simple stuff.
If you know a little C++ is't a doddle.

Edit: https://www.arduino.cc/en/Main/ArduinoBoardMega2560

 

[spec]

Wow .. I think that scope is a 1Mhz model I used to fix TV's in the very early 1980's not storage from memory ?
Would still like to have a good old fashioned analog ... I think they show the real picture.

The Tek 2235A is a 1991 model. It is a dual beam real time scope- not storage- with a bandwidth of 100MHz and a maximum sensitivity of 2mV per division. We had loads of them at work and I always wanted one. So about a year ago I bought mine from a chap on Ebay.com in the States.

I love real time scopes, but digital/storage types are so useful for much work these days. Maybe I will save up and get a Rigol too.

spec

 

[fireant007]

About the tube PSU.

A switch mode power supply with the following specification would be a simple design exercise for an experienced SMPS engineer, but would require more work than the mains transformer approach. Also, you would probably have to wind/modify a transformer.

(1) Input/output isolation: yes
(2) Input voltage: 100V to 250V
(3) Input voltage frequency: DC to 500Hz
(4) Output voltage 165V DC
(5) Output current 0A to 1.5A

spec

I have been playing around with TI Webbench design.
Auto SMPS design ... I think you still have to tweek the layout.

 

[spec]

Yup.. agreed.
it's called software layer abstraction !
You should look at the arduino .. it's really very easy for simple stuff.
If you know a little C++ is't a doddle.

Edit: https://www.arduino.cc/en/Main/ArduinoBoardMega2560

I am preaching to the wise again.
I do know the various flavors of C, but focus on C# mostly now and some time will find out what Python is all about.
I also read all the books about Object Oriented Software as you seem to have done.
My experience with software/hardware systems is that the two development teams never communicate well, and you find the softies going through all sorts of antics to match a particular hardware interface when the hardies could reduce much of the code with a simple hardware redesign here and there. And the converse is also true.

Yeah, I would love to do some microcontroller experiments... its the time though.
spec

 

[fireant007]

About the tube PSU.

A switch mode power supply with the following specification would be a simple design exercise for an experienced SMPS engineer, but would require more work than the mains transformer approach. Also, you would probably have to wind/modify a transformer.

(1) Input/output isolation: yes
(2) Input voltage: 100V to 250V
(3) Input voltage frequency: DC to 500Hz
(4) Output voltage 165V DC
(5) Output current 0A to 1.5A

spec

Check this site out ....

https://www.ti.com/lsds/ti/analog/w...DCMP=hpa_design_center&HQS=analogdesigncenter

It's almost magic for SMPS design....