Voting resources, early voting, and poll worker information - VOTE. ... Adafruit is open and shipping.
0

Ice Tube Clock IV-18 To-Spec Hack
Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Fri Oct 18, 2013 2:48 pm

LM9022.png
LM9022.png (46.51 KiB) Viewed 2765 times
I know this goes beyond what you want to use on your board, but there is a chip designed for this. The example circuit may give some design ideas. I'm going a different power supply route, using a linear transformer, similar to the earlier posting PhilD13 had made HERE

EDIT: You of course would not need the voltage boost part of the circuit for the anode and grid, via the diodes and capacitors. Ice tube's inductor boost circuit takes care of that.


Are there any good reasons to provide cleaner power to the filament?


Raid's in grid IV-18 He claimed this was not so important, in his opinion. I don't personally feel that a square wave is inferior here, in the Noritake guide it is referenced as an alternative. Much of what I have done deals with alternating current, hence my interest in driving the tube this way, I like linear transformers. By today's standards, a linear transformer is a little old school for making DC. The 9V switch mode transformer used in the kit is a great little power supply. I don't know if you are aware, but a switch mode transformer boosts the frequency to 25 KHZ or higher. By doing this the transformer can be much smaller and supply a great amount of current, plus at that frequency the filter capacitors can be quite small, to filter DC ripple at that high frequency. As opposed to a linear transformer working at 60 HZ, which needs much more capacitance to filter ripple. But for a linear transformer in this circuit, most of the current is for the AC filament, and only a small amount of current needs to be filtered DC for the micro controller, boost and anode connections.


EDIT: Try driving your MOSFETs with a 9 volt gate potential and see if that helps, 4.7 volts from the micro controller is marginal. Or try boosting voltage from LM317 a little, with variable resistor. Adjust variable resistor while reading filament output from oscilloscope, see if you can cheat voltage up a little. Using some capacitance should clean up your wave.

LM 317.gif
LM 317.gif (5.58 KiB) Viewed 2743 times
Last edited by Russell 27 on Sat Oct 19, 2013 9:01 am, edited 4 times in total.
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by phild13 on Sat Oct 19, 2013 6:52 am

The LM9022 is no longer available. You can take a look at the MAX628, also known as the TSC428 to generate the filament drive.

http://www.maximintegrated.com/datashee ... vp/id/1378
http://datasheets.maximintegrated.com/e ... TSC428.pdf

phild13
 
Posts: 247
Joined: Mon Sep 10, 2012 1:05 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Sat Oct 19, 2013 5:02 pm

PhilD13 wrote:The LM9022 is no longer available. You can take a look at the MAX628, also known as the TSC428 to generate the filament drive.

Performance of the related MAX627 (on my breadboard) is comparable to the FET push-pulls. But the FETs are cheaper and require less board space than the DIP version of the MAX627, so I'm thinking that I want to stick with the FETs.

Russell 27 wrote:Raid's in grid IV-18 He claimed [filament drive] was not so important, in his opinion.

Raid was driving the IV-18 filament well below the 5 volts required by the tube specifications--around 3.5 volts DC. At that voltage display gradient is barely noticeable.

As I understand it, the predominant failure mode for vacuum tube filaments is cathode poisoning, which is greatly accelerated by running the filament outside of the specifications. If voltage is too low, the cathode is too cold and emits fewer electrons. The large positive charge at the anode and grid ionizes trace amounts of gas in the tube, forming positive ions. These positive ions are attracted to the cathode, but are usually neutralized by electrons. If the cathode is emitting fewer electrons, a larger proportion of the positive ions collide with the filament, damaging the oxide layer and resulting in reduced electron emission. Driving the filament at too great a voltage accelerates sublimation of the oxide layer--the other common cause of reduced electron emission--but running a filament too hot is not nearly as damaging as running it too cold.

The effect of cathode poisoning is readily apparent in the first post of this thread. Note how the time displays nothing on the 3rd and 6th digits. When the time is displayed and the cathode is too cold, the protective electron cloud surrounding the cathode moves towards the more positive regions--the lit digits--and that depletes the protective cloud surrounding the filament near the blank digits. Positive ions collide with the filament's oxide layer near the blank digits, resulting in reduced electron emission. In the clock menus, the 3rd and 6th digits are significantly dimmer than the other digits because the cathode has been disproportionally poisoned in those regions. Running the filament to spec--5 volts--should fix that.

Unfortunately, running the filament with 5 volts DC causes a slight gradient across the display, but that gradient becomes quite noticeable if brightness is lowered by reducing the anode and grid voltage. EDIT: And DC is associated with filament notching, although I don't think notching would be a problem given the relatively low current density in VFD filaments.

I hope that explains why I'd like to run the filament at 5 volts with some form of AC.

Russell 27 wrote:Try driving your MOSFETs with a 9 volt gate potential and see if that helps, 4.7 volts from the micro controller is marginal. Or try boosting voltage from LM317 a little, with variable resistor. Adjust variable resistor while reading filament output from oscilloscope, see if you can cheat voltage up a little. Using some capacitance should clean up your wave.

Surprisingly, those MOSFETs are already in saturation at 4.7 volts, according to the datasheet! The scale on voltage for the oscilloscope pics is 2 volts, so the wave is going -5 volts to +5 volts--just where I want it to provide 5 volts of square wave AC.

Russell 27 wrote:Using some capacitance should clean up your wave.

Yes, it does. But board space is tight, and I'd rather not squeeze in another cap, and if I add a capacitor, I'd also want to add a protection diode. My thinking is that if sine wave and square wave AC are both okay for the filament, then a slightly noisy square wave should be fine--as long as the magnitude of the wave never exceeds 1.4x the RMS voltage.

So here is my question: Do I need to clean up the wave? The filament should be insensitive to the exact shape of the wave, correct?

EDIT: For reference I'm attaching a draft of the xmas-icetube Rev. D schematic and board and some images.


xmas-d.zip
Eagle files for the draft versions of the xmas-icetube Rev. D board.
(177.71 KiB) Downloaded 82 times


filament-driver.png
The 5v slightly-distorted square-wave AC filament power supply.
filament-driver.png (10.02 KiB) Viewed 2734 times


xmas-c-front.png
The already packed xmas-icetube draft board.
xmas-c-front.png (213.07 KiB) Viewed 2734 times
--John <www.jarchie.com/email>

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by phild13 on Sun Oct 20, 2013 6:59 am

I was actually just mentioning that the LM9022 was not available but for anyone reading that wanted to try a different part then the MAX628 would likely be a decent alternative.

While there are many ways to generate the desired waveform, I think the FETs creating the AC is just fine for what is needed and the filament is not going to care about the noise. The noise is possibly of greater concern to the processors on the board and other stuff in the room where a clock is/may be located? I know there is not a lot of room, but you could try either a couple 0.1uF bypass caps to ground on the AC side (AC1, AC2) or maybe a 100uF cap between supply and ground. Since you have a scope it should be easy to tell if there is any difference.

phild13
 
Posts: 247
Joined: Mon Sep 10, 2012 1:05 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Sun Oct 20, 2013 2:56 pm

Thank you for the suggestions, although at this point, I'm leaning toward omitting the caps.

PhilD13 wrote:try either a couple 0.1uF bypass caps to ground on the AC side (AC1, AC2)

This combination significantly distorts the square wave. The rise time and overshoot take much longer, but have the similar magnitude.

PhilD13 wrote:or maybe a 100uF cap between supply and ground.

A 220 uF electrolytic cap reduces overshoot slightly, and a 0.1 uF ceramic cap has an identical effect.


Neither setup seems to effect the 9v output of D1 or the 5v power supply to the microcontroller--both seem quite well-regulated regardless of the presence of capacitors in the filament driver circuit.

EDIT: I do see some noise on the LM317 output without a cap, but adding a 0.1 uF cap completely eliminates that noise. So the push-pull output overshoot seems to be (partially) due to the FET switching behavior. Adding the 0.1 uF cap has no noticeable effect on the LM317 input, which seems pretty clean regardless. To me this indicates that the noise across the filament should not affect other circuits and should not affect nearby electronics. But if I'm missing something, please let me know!
--John <www.jarchie.com/email>

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Sun Oct 20, 2013 7:20 pm

One other method,While outside of your original idea, shows another way to achieve the same result, using a self wound torroid to make a square wave driver. This method is kind of between what I'm doing and your MOSFETS. Kerry Wong's blog can be viewed HERE. If you ever want to experiment with transformers this would be a great place to start. Instead of using the astable 555 timer, the driver transistor and torroid would be directly connected to the micro controller. By changing the windings, it could be totally fine tuned.

vfdfilamentdriver.png
vfdfilamentdriver.png (444.1 KiB) Viewed 2697 times



100 Nf capacitors are generally used to filter high frequency noise or pulses. The more logic circuits you encounter, you will see how important these are to filtering or absorbing unwanted logic changes.
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by phild13 on Mon Oct 21, 2013 8:57 am

Most info I see connect the center tap to ground or to VDD in order to prevent ghosting and not to the supply.
Take a look at the info in the link on filament power supply and filament bias voltage
http://www.noritake-itron.com/SubPages/ ... perapn.htm
Another connected center tap to grid
http://ferretronix.com/tech/vfd/mm5316_alarm_clock.jpg

phild13
 
Posts: 247
Joined: Mon Sep 10, 2012 1:05 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Mon Oct 21, 2013 5:38 pm

Thank you both for the interesting comments. The idea of using a center-tapped pulse transformer or a standard center-tapped transformer has crossed my mind several times, and I do like those ideas better than push-pulls. But I thought that soldering a few transistors would be easier and appeal to a larger segment of the hacker community.

On another note, I think I figured out why the noise from the push-pulls doesn't seem to affect the other clock circuits. The relationship to charge (coulomb), current (ampere), and voltage is C = A*s and F = C/V where C = coulombs, A = amperes, s = seconds, and F = farads. The push-pull noise amounts to an overshoot of ~1 volt for ~25 ns. Since the resistance of the filament is ~50 ohms, the overshoot increases current by around 1v/50ohms = 20 mA. The charge required to compensate for that would be 20mA * 25ns = 0.5 nC, but the charge held in the 9v input capacitors (C1 and C2) is 47uF * 9v = 423000 nC. Although those calculations are extremely rough, I think they do show that noise is many orders of magnitude smaller than the charge held in the input capacitors, so those capacitors should be more than sufficient to buffer other circuits.

PhilD13 wrote:Most info I see connect the center tap to ground or to VDD in order to prevent ghosting and not to the supply.

I know that you're right, and there must be a good reason for connecting the center tap to ground. I would be curious what that reason is, because naively, connecting the center tap to the supply makes more sense to me...

In newer VFDs it's sometimes necessary to generate a negative shut-off voltage for the grid/segments to prevent ghosting. I think the idea of connecting the center tap to the supply is to make ground negative with respect to the filament, eliminating the need for a negative shut-off voltage.

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Tue Oct 22, 2013 8:50 am

I posted much of this to show how many ways things can be done, others take on a particular solution. I have studied much of this on my own, including the Noritake, Itron VFD guide. This is the basis of my drive circuit. Have not had time to test it, but a start concept idea.

POWER_ Filament driver_1.PNG
POWER_ Filament driver_1.PNG (39.4 KiB) Viewed 2639 times
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Tue Oct 22, 2013 3:25 pm

Russell 27 wrote:I posted much of this to show how many ways things can be done, others take on a particular solution. I have studied much of this on my own, including the Noritake, Itron VFD guide. This is the basis of my drive circuit.

Thank you for sharing!

It looks like the voltage across the filament will be ~5v sine wave AC, so the magnitude of that wave will have a minimum of -7v. From the IV-18 specifications, it looks like segments will glow slightly if the anode/grid are at 2.5-3.0v above the filament voltage. So it's possible that you'll need to generate negative shut-off voltage to prevent ghosting; the MAXIM driver chip does provide a Vss pin for the negative shut-off voltage.

But I'm not sure if I'm interpreting that part of the datasheet correctly, so I'd be quite curious if you notice any ghosting.
--John <www.jarchie.com/email>

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Wed Oct 23, 2013 8:57 am

Actually, that's what the virtual center tap is supposed to help with.
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Wed Oct 23, 2013 3:37 pm

Russell 27 wrote:Actually, that's what the virtual center tap is supposed to help with.

Ah, yes. Of course! My mistake. So the wave will only reach a minimum of -7/2 = -3.5v relative to ground, correct?
--John <www.jarchie.com/email>

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Thu Oct 24, 2013 2:40 pm

Ah, yes. Of course! My mistake. So the wave will only reach a minimum of -7/2 = -3.5v relative to ground, correct?


Actually, I don't look at it from that perspective. As an explanation, instead of 7 volts peak I would say to look at RMS, 5 volts. I look at it as more of a balance point. I'm sure you have seen Noritake's Itron VFD operation guide. Below I have posted quoted information from that guide, pertaining to my circuit concept. There is also detailed information in the guide, about how you have driven the filament and also how the Ice Tube Kit does.

Fig.11_ Transformer with Center-Tap.gif
Fig.11_ Transformer with Center-Tap.gif (2.26 KiB) Viewed 2579 times

Generally, the transformer is the most popular device utilized to supply the filament voltage (Ef) with a 60(or 50)Hz sine wave which also has a center-tap for cathode bias as shown in Fig.11. The center - tap technique is used to prevent luminance slant i.e. difference in brightness from one side of the display to the other.


Fig.12_Transformer without Center-Tap.gif
Fig.12_Transformer without Center-Tap.gif (2.31 KiB) Viewed 2579 times

Using a transformer without this center-tap can not only cause luminance slant but also ghost illumination due to exceeding the amplitude of the filament voltage in excess of the specified cut-off bias voltage rating


A relatively simple task is actually quite complex when you really look at it. I read that Guitar amp guys, kinda cheat a virtual center tap when their transformer does not have a center tap, to alleviate hum. I imagine this is a rough correlation of what can equate to ghosting in the tube. They generally use 100 ohm resistors, one connected to one side of secondary line to ground; one connected to the other secondary line to ground, basically like my drawing. I chose 470 ohm resistors, in the hope to achieve desired effect with less wasted, dissipated current.

Now this is the way I understand it, there's actually a lot of complexity here for such a simple electron maker.
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Re: Ice Tube Clock IV-18 To-Spec Hack

by jarchie on Sun Oct 27, 2013 7:29 pm

Thanks for the explanation, although I guess I was thinking that voltage on part of the filament will dip below ground at certain times, so there might be a possibility of slight ghosting on some tubes.


On another note, I soldered the AC generation circuit below on a scrap of Adafruit prototyping board, and it generates almost exactly 5v AC with the 1% resistors below. This is my (hopefully final) push-pull circuit for this to-spec hack. The LM317 is a TO-92 package to save space, which works fine with a switching-type 9v power supply as included with the Adafruit kit. If the AC adapter is providing significantly more than 9v, a TO-220 package might be necessary.

AC.png
AC generation circuit
AC.png (18.85 KiB) Viewed 2559 times

Although the push-pulls do not provide sine-wave AC, they do provide a kind of voltage symmetry with respect to ground that should be similar to using a center tap. And because the voltage drop across Q5 and Q7 is around half a volt, ground is always negative with respect to the filament. As far as the tube is concerned, it's like using 5 volt, center-tapped, square-wave AC with ground as a negative cut-off voltage (around -3 volts) to prevent ghosting.

It might be tempting to add capacitors to the LM317 output to improve transient response, but the LM317 needs a 10 mA load to effectively regulate current. So when the display is off, the capacitor might carry a significantly larger charge than required to provide 5 volts across the filament. When the display is turned on, the initial voltage across the filament might be well above specification.

If anyone would like to use capacitors on the LM317 output, I also suggest (1) adding a 500 ohm resistor between between the LM317 output and ground (to ensure a 10mA load), (2) adding a safety diode between the LM317 input and output to allow current flow from the output to input (protects from a short-circuit on the input), and (3) using a TO-220 package for the LM317 instead of the TO-92 package that I used (the TO-220 can safely handle the additional 10 mA current draw, although a TO-92 might still work). But the filament should be indifferent to the presence of capacitors, and C1 and C2 on the Ice Tube Clock are more than sufficient to buffer other circuits from the slight electrical noise introduced by a lack of capacitors on the LM317 output.
--John <www.jarchie.com/email>

jarchie
 
Posts: 595
Joined: Sun Jun 24, 2012 2:16 pm
Location: Santa Cruz, California, United States

Re: Ice Tube Clock IV-18 To-Spec Hack

by Russell 27 on Mon Oct 28, 2013 11:11 am

The LM317 is a TO-92 package to save space, which works fine with a switching-type 9v power supply as included with the Adafruit kit. If the AC adapter is providing significantly more than 9v, a TO-220 package might be necessary.


It might be tempting to add capacitors to the LM317 output to improve transient response, but the LM317 needs a 10 mA load to effectively regulate current. So when the display is off, the capacitor might carry a significantly larger charge than required to provide 5 volts across the filament. When the display is turned on, the initial voltage across the filament might be well above specification.


Input voltage is not a direct correlation to output current. The regulator regulates voltage, not current. The capacitor will not charge above output voltage, it will just stabilize at full charge capacity @ 5.7 volts (according to your drawing resistors) with no load. The filament itself regulates current. This is unlike the inductor circuit, which induces an unregulated pulse voltage that is much higher than 9 volts, to charge the capacitor; sometimes called a charge pump or voltage pump; that voltage is only regulated by the ZENER for max voltage. Since there is only so much current there in the first place, stored in the capacitor, the voltage drops when the load of the anode is there. No current no voltage. The TO-92 is only good for about 200 MA max, so it may get a little warm. Until the push pull circuit starts to oscillate, there would be no real connection there.
Russell 27
 
Posts: 240
Joined: Thu Sep 12, 2013 3:59 pm

Please be positive and constructive with your questions and comments.