IN-9 drivers

[Mihai]

Hi all,
I know there are a lot of discussions on this topic and I believe I read almost everything. Not sure I understood everything so that's why I would like an extensive opinion on this matter. I recently bought some IN-9 Nixies from eBay for building a stereo 10-channel spectrum analyzer. I believe the schematic is very well known, Instructables - http://www.instructables.com/id/Nixie-Tube-Music-Visualizer/. Well, aside the modifications I made regarding the number of channels and the lin-to-log converter, I have an issue with how the Nixies are lit; of course, not all are lit from the bottom and their behavior are somehow erratic.
I boiled down what I read to the points below:
1. Burn-in - this was new to me - it seems that the tubes must be burned-in prior to using them. My tubes are lit all the way up, just not when they are supposed to. Do I still need to burn them?
2. Driver stage - I've put together all the op amp driver schematics I found, including the one in the original schematic (Instructables); there are basically 5 variants for which I do have some questions

a. "blind" op amp driver + limiting current resistor - this is in the Nixie documentation so it's the 1st building block
b. same as "a" + a pot for varying the current
c. same as "b" but interestingly, it has a resistor in the op amp's negative reaction loop. What is its purpose? Does is raise the amplification?
d. same as "b" but with a low pass filter. I understood this should fix the IN-9 issue regarding firing up from the bottom. Question is: do I have to have a low-pass filter tuned for each channel or is one for all? If it's one for all, what should be the cut-off frequency? I found different lox-pass Nixie filters and they were usually built with R=10k ad C=100n. If we go with the RC low pass formula, we get ~160Hz for the given values which I really don"t get what it's for
e. The Instructables schematic (see link above) mentioned a voltage divider of the input voltage but ...how? Is the voltage divider using the low-pass filter resistor for doing this?

Either way, I tried almost all variants in my schematic and none has a definitive and decisive result. I ended up with option "b" and the thing is that some work almost fine, and some are firing up from a third from the bottom. Sometimes they go in the middle but not always.

Any opinions on the above? The HV PSU is digital with MAX1771 not analog with transformer. For some reason, the voltage reading is not constant, it varies - 145-170V

PS: shall I give up on IN-9 and buy IN-13? I read though that it's not guaranteed they will work absolutely fine.

Thanks

[gregebert]

I gave up entirely on the linear indicator tubes after experimenting with them. If you dig around on this forum you'll find another posting I did.

Basically, I found that they were non-linear, non-repeatable, and did not light uniformly from bottom-to-top.

My original plan was to use 2 of them as the mechanical hands of a clock. Instead I used over 300 individual neon bulbs to simulate clock hands, and it works great.

Would a string of neon bulbs work for you ?

[Mihai]

Well, I really like the vintage look of the Nixies and this is part of a bigger project. Basically I would like to transform an old valves radio, changing only the insides and snapping a Nixie spectrum analyzer on it.

But I am interested in your schematic, it sounds interesting, could you please share it?

PS: any ideas on the initial email? I think I will have to switch to IN-13 either way because the consumed power (most probably) exceeds the limits of my power supply:
10 Nixies x 12mA each (average) x 180V = 21.6W and 10 Nixies x 20mA each (max) x 180V = 36W

...adding the rest of the circuits (active filters, lin-to-log) will be an additional 5-8W

[gregebert]

I attached a jpg of my neon-bulb driver, but it wont be much help because it's just a way to drive neon-bulbs from a CMOS/TTL level. Note that I use an opto-triac; this is for isolation as well as for AC operation. Operating a neon bulb with DC only illuminates 1 electrode (the cathode), rather than both. Ignore the diode; it's not applicable for your circuit. Funny thing is, I had considered using regular triacs (not isolated), and keeping the digital circuitry on the AC line, but the discrete triacs were larger and more expensive than the opto-triacs.

In your case if you want to drive 10 bulbs, you will probably need 10 opto-triacs. I'm sure there are clever ways to do a linear bargraph with less than 1 driver per bulb thru multiplexing; I have not experimented with multiplexing neon bulbs yet, but I have a project in the future that will. Dont feel bad about using lots of bulbs or drivers; my clock has 306 neon bulbs, 72 opto-triacs, etc. Someday I'll post a video on youtube.

Another option would be a plasma-panel from an arcade game; they are typically 128x32 pixels. You will still need to develop all the A/D circuitry, and the software to run the display, but it will give you maximum flexibility. However, it might look "too modern" . That's the main reason I've stayed away from them.

[Alex]

IN-9 is much brighter than IN-13, hence the higher current draw - it is also (IMO) better for display pieces for this reason. IN-13 is a bit better at starting at the bottom but still has its issues. If you want to make it 100x better at behaving, feed it full wave rectified DC but without any smoothing, A smooth DC voltage allows the ionisation to simply travel around on the wire - maybe this is due to contaminants building up with age I don't know... When I tried unsmoothed 100hz (50hz rectified) DC they worked so much better, though I guess this can make driving them a little more tricky. It does also help to let them sit for a few minutes at full power for a while, maybe even drive them 150% or so to clean up the internal structures. 

They are nice tubes, I can see the appeal over discrete bulbs. I have about 300 or so stashed away, I was going to make a spectrum analyser also at some point, so will keep tabs on your success!

- Alex

[Mihai]

I have multiple projects with Nixies, the spectrum analyzer is the first one and it will be followed by a PC monitor (CPU and MEM). Rectified DC sounds like a better option in terms of results but I think it's too bulky; I found a schematic with rectified DC + uplifting diode bridge but I didn't consider it yet. When I will build the PC monitor I will definitely need something small to fit on the back of a PC bay panel (the plastic bay on the front of the PC case).
I also built a spectrum analyzer schematic based on a MCU but that wasn't successful at all; I jumped too soon into building it and I don't think it was well thought, well designed or even built. I was enticed by the idea of having an MCU doing a FFT rather than separate active filters for each band. The beauty of the MCU schematic was that it was scalable so, in theory, you could attach 10-20 tubes with no issue. I can share the schematic if you want.
Can someone still tell me what should be the parameters of the low-pass filter on the driver stage?

[David Forbes]

Greg,

Your circuit works, but there may be a better way to get both electrodes
lit on the neon lamps, one that costs less and doesn't put AC mains
power in your circuit boards.

It's possible to use a standard NPN transistor with a pullup resistor to
+90VDC to turn the neon bulb on and off. It's also possible to put such
a circuit on both ends of the neon bulb and drive the two transistors
with combinations of binary 00, 01, or 10 to allow you to alternate the
glowing electrode at whatever rate you like. You can increase the duty
cycle to 100% using this method to get a brighter display.

The transistors can be TI SN75468 chips or similar to save wiring
effort. They also make SIP resistor networks to consolidate the pullup
resistors.

On 2/15/15 12:23 AM, gregebert wrote:
> I attached a jpg of my neon-bulb driver, but it wont be much help
> because it's just a way to drive neon-bulbs from a CMOS/TTL level. Note
> that I use an opto-triac; this is for isolation as well as for AC
> operation. Operating a neon bulb with DC only illuminates 1 electrode
> (the cathode), rather than both. Ignore the diode; it's not applicable
> for your circuit. Funny thing is, I had considered using regular triacs
> (not isolated), and keeping the digital circuitry on the AC line, but
> the discrete triacs were larger and more expensive than the opto-triacs.
>

--
David Forbes, Tucson AZ

[gregebert]

Funny I immediately dismissed this idea originally because I always think "H-driver" for DC-to-AC; using 4 switches. For my clock this would have been too costly. I didn't realize that replacing the upper 2 switches with resistors was perfectly fine and resulted in negligible wasted energy.

In the case of my big clock, the PCB layout & routing was too dense in 2 layers to use 2 resistors per bulb, even using resistor SIPs (though SMT would have been fine). WIth a high-enough DC supply, I could connect bulbs in-series (which saves on resistors).

[Mihai]

Any chance of going back to my initial question?

[Grahame Marsh]

On 16/02/2015 16:23, Mihai wrote:

Any chance of going back to my initial question?

--
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/f27b729a-6356-4aaf-b306-4893bddb5d88%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

I built a simple "7 segment" clock using IN13 tubes. Even though they have the primer, I found that initially the tubes struck in different places and were irratic when trying to sweep a bar across the tube. I ran them 50% over current for a few hours and then used the controller to continuously sweep them backwards and forwards. After a few days of this treatment they started to behave and have worked faultlessly since. I've no experience with IN9 tubes. If you look at the schematic you will see it is very crude as I do not use any opamps to offset the voltage drop but only run the DAC over a limited range.  Don't know if this helps or not.

http://www.sgitheach.org.uk/ss.html

Grahame

[Mihai]

Thanks for the tips.

I would just like to understand the low-pass filter's cut-off frequency. Do I have to get one for each band (frequency) or is it "one-size-fits-all"? From the values I found around, R=10k =1nF, the filter's frequency is 160Hz and I don't know to what extent this applies to my problem.

Could it be that the low-pass filter suppresses any high(er) frequency noise which might have a negative impact on the Nixies?

[Grahame Marsh]

I can't answer that question from knowledge. Gas filled tubes (being
pedantic these indicator tubes are not nixies) are relatively slow
devices as ionisation and deionisation are a slow processes. My answer
would be that HF noise would be "ignored"; now someone who does know can
knock me down :-)

But if you are building a spectrum analyser with the bandpass filters
done in a micro, why do you need LPFs?
Or have I missed somthing? If you are going to use external filters then
you need one per band and they must be band-pass not low pass?

> --

[Mihai]

Let me take a step back - the current schematic is not MCU based, let's say it's analog. The building blocks are:
input amp -> bandpass filter -> lin-to-log converter -> AC-to-DC converter -> driver

The LPF is between the AC-to-DC converter and the driver. At that point, the audio signal has been already filtered, so in theory I should have only the "DC" component of a specific channel/audio frequency. Does it make more sense now?

[Grahame Marsh]

AC-to-DC converter > LPF > driver
rectification > smoothing > DC to driver

Are in some way equivalent. So decide on the degree of smoothing required and you are home. But I think we have come full circle back to your LPF requirement

I can suggest two ways - build a channel and experiment increasing the amount of "smoothing" until you get the look right - this is not going to be a precision instrument. Or try modelling the circuit. I use the open-source Qucs as it is free but you have to learn how to use it.

Grahame

--
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/3c052adc-f786-4679-9e02-83978eea8a33%40googlegroups.com.

[Nick]

Just a quick thought. With a lin-> log converter after the bandpass filter you will need one per channel.

With the lin-> log converter before the set of bandpass filters, you need only one, saving a whole bunch of components...

Nick

[Mihai]

The whole schematic is built on 2 boards: logic and drivers. The logic board is manufactured because it was a bit too complex for me to do it (~200 vias) so I cannot modify the electric connections. I paid around 100$ for 5 pcs so I would like to use it as-is. The LPF + drivers + HV board is made by me so I can redo it; I have to redo it either way because I used the wrong opamps when I designed it.

@Grahame
You're close to what I also tested on a breadboard; what I found out is that the lower the frequency the lazier the tubes. However, the behavior is the same from a certain frequency above, including not having the LPF in there. This is connected to my initial question(s): what frequency should I use, why and is it the same for all drivers?
Simulating the schematic, or parts of it, in a simulator will not do. The simulation programs are not always tell the entire story. For example, when I simulated the wave gen + comparator (lin-to-log) the response I got in LTSpice was looking very good. However, in practice it didn't work because the output of the wave gen was higher than the audio signal therefore driving the comparator into low-state.

@Nick
I am using one wave gen for all channels and a comparator for each channel. I was thinking to build the lin-to-log before the filters but the owner of the schematic (Instructables) confirmed that it is working this way. As said above, since I already have the board, I cannot modify it.

[Mihai]

PS: I forgot the attachment

[Nick]

Hi - this doesn't look much like the original Instructable - I'm trying to understand your lin/log converter - the 555 as shown should produce about a 4.6kHz spike with an exponential decay curve - do you have a reference for how this is supposed to work, end-to-end?

Regarding the PSU - apologies for not replying sooner - if its the one from my web site, you really need to use the component values I recommend and to read the instructions carefully, especially those relating to layout - the output capacitor in particular needs to be SMALL and low ESR. A big capacitor, e.g. 100uF, will cause problems...

Nick

[John Rehwinkel]

> Just a quick thought. With a lin-> log converter after the bandpass filter you will need one per channel.
>
> With the lin-> log converter before the set of bandpass filters, you need only one, saving a whole bunch of components...

Since it's a sampling lin->log converter, it would work anyway, the large bands would swamp the small ones. Even if it were a flash (analogue) lin->log converter,
interaction between bands would be a problem, so you'd still want per-channel ones. As it happens, it's a non-issue anyway as Mihai already has the boards
made up.

- John

[Mihai]

The simulation of the 555 circuit is in the attachment; I got this schematic/option from the Instructables author himself after I told him that 2 Quad comparators per channel is overkill. The schematic came with 2 issues which I didn't pay attention to check:
1. the type of the 555 circuit itself - I can't quite figure out what it is - looks like an astable but it's not quite there
2. the transistor's base resistor - in his original schematic this resistor was only 10K. Well, it seems that even LTSpice is having its bad days because, taken into consideration the other surrounding parameters, the schematic doesn't work in practice even though Spice says it's possible. If the base resistor is under 68-100k then the voltage of the function generator is greater than the audio input (as I said in one of my previous messages). If you look at the schematic you will see that the output of the wave gen is connected to the inverting input of the comparator; well, considering the audio voltage ~3V and the wave gen ~ 8V then the result is quite clear - no signal at the comparator's output. I had to increase the base resistor to over 68k so I can get the desired result.

PS: the PSU has a 4.7uF/250 at the HV output; shall I lower it or even remove it?

[Nick]

In your ASC file, C1, C2 & C3 all have invalid characters.... removing those changes the behaviour dramically - they should just be "0.1u" or "100n"...

I've included what I believe should be the astable circuit from your schematic, done in LTspice - seems to be ok...

Nick

[Mihai]

It's the same thing, at least for me; it's probably a character set thing.

[Mihai]

Well, in the mean time I received the IN-13s and they work absolutely fine. Moreover, for one channel I put the LPF in front of the driver and surprisingly enough it has no effect what-so-ever. I am working on the PCB now and I hope I can do it one side only and a few straps on the top side.
PS: it seems there is a huge difference between LM358 and NE5532 dual op amps. With the 1st it works fine but with the latter, though rated as newer and better, the behavior is a bit hectic, the most notable is that with no audio input the tubes stay lit all the way to the top.

[GastonP]

That behaviour in a NE5532 is not normal at all. If you have an oscilloscope available you should check what is happening, either HF oscillation, a defective solder injecting signal from a nearby transmitter, or even a PSU regulator that is oscillating in HF too and passing through the bypass capacitors. Also check your layout and distance of the decoupling caps to the IC. You should check too for (lack of) ground loops.

[Peace Frog]

Hi.
New member here, I have a question regarding IN9 and thought better to ask in the thread rather than starting a new one.

Read that IN9 should be driven by DC, would the device suffer damage from driving it with AC, or DC with an AC element?

Thanks.

[Mihai]

Hi,
I don't know if the Nixie will be damaged "per se" but it will surely not work as designed.

Do you have something in mind, like a project, or did you bump into such a schematic?

[Peace Frog]

It is something I noticed from experimenting with IN9.
The problem is that I cannot quantify the drive conditions just yet, waiting for scope delivery.
I am not keen to post what I might be mistaken about, but as long as I am not shorting the life of the tube in the process!

Shouldn't be too long,new year is over, and I keep chanting Rigol three times in the mirror each night!

[Mihai]

Well, powering the tube by other means than the specs may very well shorten its life.

[Peace Frog]

On Sunday, 22 February 2015 21:33:32 UTC, Mihai wrote:

Well, powering the tube by other means than the specs may very well shorten its life.

Had a look for specs and cannot fnd anything other than what appears to be written by enthusiasts.
Another question of Nixie builds, are the current limiting resistor best being flame proof/high voltage?

[Mihai]

The limiting resistors surely have to be there; to a certain extent you can thing of these devices as LEDs. The are probably sturdier than LEDs but still ...