Constant current source design

[Paul Andrews]

I wanted to start a discussion about constant current sources (as opposed to sinks, because I want to provide a constant current to the anode regardless of which cathode is pulled to ground). I've found many articles on the web. Some for Nixie constant current sinks, some for LED constant current sources, some more theoretical, some very simple, some very complex. But no constant current sources for Nixies - i.e. designs for a constant current source that include actual part numbers and component values. I wanted to start simple, to make sure I have at least some grasp of this topic, so here is my first stab. I would be grateful if anyone could let me know if it would work as is or if I have made some fundamental errors - ignoring improvements such as temperature stability for now (BTW R3 is Re in the equations).

Be gentle - this is all new to me!

[Tomasz Kowalczyk]

I would recommend using design like first one from link below: I haven't tested it myself, but it looks like it loses less power (doesn't have any resistors going from 200V to GND), and the mentioned minimum voltage drop on it will be not noticable in nixie application. It is not really a current source, but a current limiter, but again - it should do it's job. 

It uses one more transistor, but I think that with MMBTA42 or MPSA42 prices it shouldn't matter much.

http://electronicdesign.com/power/current-limiter-offers-circuit-protection-low-voltage-drop

[Laurence Wilkins]

Yes, I agree with Tomasz, the first circuit shown in the article he links to is the way to do it. Advantages: No "wasted" current draw, cheap and easy.  Ignore the comments in that article about the problems of voltage drop - that might be important when you've only got 5V to play with, but with 200V on tap... !!

Suggest RSense of 120 Ohms should regulate the current to 5mA, then make R1 anywhere between 100K and 220K. Not critical.  Use a high voltage MPSA42 transistor or similar for T1, T2 can be any old jelly bean type.

While this constant current source will be inherrently short-circuit proof (the most current that will ever flow will be 5mA), be aware that under those conditions a humble MPSA42 will be dropping 1 Watt and will fry pretty quickly! Something a little beefier in the power stakes might then be a better choise.

[gregebert]

That's basically what I use in my designs. I'll highlight the differences:

• I use a PMOS instead of PNP, mainly because it requires no drive-current.
• R1 & R2 are replaced with a pot to make the current adjustable.
• The above pot can driven from a small DC-DC converter (my preference), or between the HV supply & GND. There's essentially zero current for PMOS gate-drive, so high resistance values are fine. Not the case with PNP, though, due to finite base-currrent.
• A zener diode is added to clamp any spikes that may arise at the gate of the PMOS device. It's a paranoia item.
• A filter cap was added, in case there is unexpected noise from the DC DC converter, and also to suppress any very-short transient that may arise that are too fast for the zener to kick-in. (paranoia item).
• A large resistor across the PMOS to bleed any potential ESD. Without it, there is a remote possibility of charge-buildup. (paranoia item)

So, this circuit is replicated for each anode. When multiple anodes are driven, they all share the same gate-drive signal, which I call PDRV on the attached schematic.

[gregebert]

Some other things I forgot to mention...

1. Generally, you would use a current-regulator on the anode side for non-segmented tubes (0-9), where all cathodes use the same current.

2. Segmented displays (b7971) have different currents for various segments, so you will need cathode-side current-regulation. I chose to include an anode-side current-limiter as well because the Burroughs datasheet specifies a max total current. This max current is about half the sum of all the spec'd individual segment currents.

3. If you use MOS devices, be very careful not to exceed the max-voltage specs. Doing so will damage the device (impaired reliability, or outright destruction). I've done some research on MOS device-failures and they often fail-shorted, which is disastrous. Fortunately, those kinds of failures are mostly caused by overstress (voltage or temp).

[ZY]

I'd like some opinions on using something like a LM317 inline on the high side, as in the image attached. I've tested it and it seems to work although maybe there is a flaw that I'm not seeing somewhere.

[gregebert]

I dont see anything fundamentally wrong; I just have paranoia about running devices in an environment where this is potential for overvoltage. The LM317 is rated for ~35V, and the HV supply is around 180V, so you are relying on the voltage drop across the nixie tube. There are all kinds of unexpected things that happen during power-up/down; maybe there are scenarios that could damage the LM317.

[Tomasz Kowalczyk]

It all depends on R11 and desired current. Most nixies have a minimum voltage drop of around 120V (although they will maintain 130 for most of the time, they will just continue to work with lower voltage), so if voltage drop on R11 is more than (supply voltage - 120V - LM317_max_voltage), then it is safe - the only case where the nixie doesn't have that voltage drop is when it is internally shorted.

In this schematic, assuming that U15 is fully open, so voltage at R11 left side is just about 180V, everything will be okay - the LM317 is set for 8mA, so voltage drop over R11 would be around 34V and everythin would be safe. The problem is that if current is significantly lower that (nixie is heavily poisoned), then stuff can get bad: as far as I understand, R11 voltage drop has to be lower (Ohms law), so the voltage across LM317 will get higher, I think - it has to drop somewhere.

Long story short, you are right - LM317 can fail in non-standard situation. That's why I would go for a two-transistor design using MMBTA42 (or MPSA42), which will withstand almost anything. The only problem is power dissipation - 5mA with 50 volts drop across current limiting circuit is 0,25W. With such constant current design one should lower HV voltage to minimum (just above striking point of all nixies), normally the significantly higher voltage (200V) is used to make choosing limiting resistor easier.

[Luka C]

@greg

I'd like to ask something related to the B7971 since I'm designing a PCB for the clock at the moment. I have implemented the controlled current sinks for each cathode and will fine tune the current for each segment (already discussed in the HV control chips thread with the schematic). I'm wondering if you implement both the cathode sink drivers and the anode driver you proposed here:

1) If sum of all lit cathodes currents exceeds the maximum value in the datasheet (the current set on the PMOS controlled source) what will happen to the currents of the individual segments, will they proportionally go down for each activated cathode? If so, will this cause the tube to reduce the glow?

2) If the sum of all lit cathodes is less than the maximum defined in the datasheet, it's supposed to stay that way because the PMOS source only limits the upper maximum current but does not enforce it constantly to flow trough the tube when it's activated?

3) How important do you think the maximum current per tube control is? Before deciding to go with the sink controller for each cathode, I browsed for other solutions to drive the B7971 and most of them use resistors, but these resistors have to limit the current to rather low values, are these enough to light the tube properly?

For example, http://tayloredge.com/storefront/1386_B7971SmartSocket/1386A.pdf

Segment 11 in the datasheet is defined for a max current of 5.5mA, but the above schematic would produce (170V - 140V) / 24 kOhm = 1.25mA. This seems rather low?

[gregebert]

Timely subject....my 7971 PC boards just cleared US customs and are on the plane to Oregon....

I have independent current-regulators on each cathode. 4 tubes * 15 cathodes = 60 current regulators. The driver transistor is a dual NPN in a surface-mount package to save area. The cathode (segment) currents are not adjustable, but there are SO many available resistor values it's easy to tweak to fine resolution. I purposely chose 1% tolerance SMT (0805) resistors. The NPNs are controlled by a TTL-level shift register (74HC595) running at 3.3V.

For the anodes, each tube has an overall current limit (PMOS), set to 22mA (datasheet max). Total of 4 per board.

The boards are end-to-end abuttable, so my first (and only) project will have 7 tubes to start with.

Of my seven b7971 tubes, I gathered I-V data for each segment. This gave me an overall idea how much current it takes to get decent illumination.

Next, I wrote a spreadsheet that calculates the total current for a variety of ASCII characters (0-9, A-Z, _ ^ - * ). By setting the individual segment currents in the spreadsheet so that most characters are at/below the tubes max-rated value, I get an idea where to start. A few characters go over the 22mA spec value, and that's where the anode regulator kicks-in. Hopefully it will result in segments fair-sharing their current. From there, it's easy to calculate the cathode resistor and select from available values.

[robin bussell]

> For
> example, http://tayloredge.com/storefront/1386_B7971SmartSocket/1386A.pdf
>
> Segment 11 in the datasheet is defined for a max current of 5.5mA, but
> the above schematic would produce (170V - 140V) / 24 kOhm = 1.25mA. This
> seems rather low?

I use those boards in this: https://hackaday.io/project/9155-nixiebot
and it seems plenty bright enough to my eyes. Occasionally when the sun
is full on the tubes through the window the pictures it produces can be
a bit less legible but that's likely the camera having to adjust the
sensitivity down due to bright reflections on the tubes.

Cheers,
Robin.

[gregebert]

To elaborate a bit more about what happens to the b7971 when the anode current-limiter kicks-in, I will need to study the I-V plots I generated from tube measurements. What will happen is that the anode current-limiter will effectively vary the voltage at the anode to maintain the total current at the limit.

If all segments had identical I-V characteristics, they would equally share the reduction in current imposed by the anode current-regulator.

I think that the segment with the highest voltage-drop will "suffer" the most, followed by the segment with the next-highest voltage-drop. This is why I tried to keep all displayable characters at/below the max current limit. The intent of the anode current-limiter is for protection only.

[Paul Andrews]

My parts arrived, so I finally got to try it. It actually works! Now on to trying variations - I plan to follow the course here.

[ZY]

I just want to add that for the low current scenario, the voltage should be consumed by the nixie. The voltage across the LM317 will continue to drop with reducing current, down to about 0.75V at 0.2mA under a test I did. Only in overcurrent would there be an issue, as the voltage across it will increase beyond it's limit. 

[gregebert]

A word of caution...the next article in the allaboutcircuits link in Paul's posting discusses current mirrors. You can use that technique for multiple anodes with a few caveats

1. If you use the R1//R2 voltage divider, you will need to account for base-current of additional regulators. This means smaller R1 & R2 values, hence more power dissipation in those resistors (heat).
2. You can also use a PMOS device instead of the PNP [base--> gate, emitter --> source, collector --> drain]; the gate current will be zero. Just be careful about ESD handling procedures, and be sure to have plenty of voltage margin (Vds > 200V for Paul's circuit)
3. You can omit R2, and replace R1 with a negative voltage supply; isolated DCDC converters are perfect for this. -5V for PNP, -12V for PMOS is a god choice. Remember: the gate or base must be negative with respect to the emitter.

Don't follow the current-mirror literally in the allaboutcircuits link, because there will be device-to-device variations in current for a given Vbe. This technique is fine for IC's when the mirrors are in close proximity on the same die and other symmetry rules are followed. Entirely different story when offchip. Having resistor R3 swamps-out the effect of Vbe variations between devices.

[Paul Andrews]

If I just look at all of the circuit diagrams for various current limiters, it begins to seem like any combination of a couple of transistors and a bunch of resistors works! Hence my need to do some 'practicals', it makes me go in to it all in enough depth to actually begin to understand.

BTW, how exactly would I drop 5V from 200V using a DCDC converter? I.e. which converter (part number?) and how would it be wired up (diagrams help, I'm a visual type of person).

Thanks in advance.

Maybe I should just do a masters in electronics.

[JohnK]

I suspect that he meant if you wanted the 1ve supply, then the devices he
mentioned give it easily directly from your low voltage supply - I don't
think he meant you to use up your precious High Voltage. Presumably you are
running the high voltage generator DCDC conv off a low voltage? You aren't
getting High Volts dangerously from the mains?

John K

--
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 an email to neoni...@googlegroups.com.
To view this discussion on the web, visit
https://groups.google.com/d/msgid/neonixie-l/c8b42ecf-2ec3-4b13-9bdf-60756b034537%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Paul Andrews]

As I see it, the base voltage needs to be -5V wrt to the emitter, so if the emitter voltage is 200, then the base voltage needs to be 195. Am I missing something?

> You received this message because you are subscribed to a topic in the Google Groups "neonixie-l" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/neonixie-l/gbJ0Beiu66I/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to neonixie-l+...@googlegroups.com.

> To post to this group, send an email to neoni...@googlegroups.com.

> To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/B1651375EFEC4E4096A7089BAAA54BC5%40compunet4f9da9.

[JohnK]

I think he was talking about a -5supply that you make from a low voltage AND
that both rails of teh 5V supply are insulated very well from earth/ground.
ie Floating.
Then you can wire each of its leads to a place that could be a voltage well
above [or below] earth/gnd. Let's wait for him to pipe up. I was only
replying early in case it put you on the right track.

Look into supplies that have "floating" outputs. Get that concept figured
out firmly. [I noticed that you said you are willing to learn.]

jk

https://groups.google.com/d/msgid/neonixie-l/6875DA44-C17F-4161-AAE8-AD4B877C89E9%40gmail.com.

[gregebert]

There are lots of inexpensive isolated DC-DC converters. I've been using products from Cui and Recom. Here's a 12V unit I use: http://www.digikey.com/product-detail/en/VIBLSD1-S5-S12-SIP/102-1432-ND/989887

They are available in all sorts of input and output voltages. Be careful that some require a minimum load, which can be for driving a power-on LED indicator. Also be careful filter caps on the output; they often have max-cap limits in the datasheet.

Anyways, they are easy to use. Connect the 2 input pins to your low-voltage supply observing correct polairty.

For the DC-DC converter outputs, connect the positive terminal to your anode supply; the negative output becomes the bias supply for your current regulator(s).

To make the anode current adjustable, connect a pot across the DCDC converter's output, then use the wiper to supply the bias voltage. I'd suggest 10-turn trimmers, rather than standard (cheap) 3/4 turn.

• If you are using PNPs, be mindful of the base current. It may force you to use lower-value potentiometers (ie,below 1K).
• If you use PMOS, there is no current so just use any pot you have; anything from 1K to 100K or more is fine. Best to add a zener diode to clamp any potential voltage spikes that could happen (startup/shutdown transients, ESD, etc) to protect the PMOS. The zener voltage needs to be greater than the DC-DC converters output voltage, AND less than the max Vgs specified in the PMOS datasheet.

[Paul Andrews]

Thanks Greg and John. That is very interesting about the isolated DCDC converters.

I've been trying to get the simulator in KiCAD 5.x nightlies to simulate some of these circuits, but no luck so far. I might give LTSpice a try.

I've tried out a few variations on the constant current source theme (and there are many more to go from the suggestions on this thread). The latest one I tried was a current source version of a temperature stabilized current sink as described here. I'm not entirely convinced that I translated it from a sink to a source correctly, however it works, except that it is no more stable than the basic constant current source in my first post! Perhaps because the first version is operating well within its parameters anyway, or the NPN isn't really much of a match for the PNP? Anyhow, I would be grateful if you could tell me if I at least got the principle right, or just messed up badly! (BTW, I'm very grateful for your input, but I also don't want to be a pain, but I assume that some other people might find this stuff useful too!).

I also tried this current mirror, which was a complete disaster - BY my calculations, the current should have been limited to about 0.5mA (200/390K), but it grew rapidly to about 30mA before I managed to break the connection, I don't really understand why. But now I do understand why there is no Re in those mirrors - Vb in the equations in my first post is equal to k (i.e. Vbe) so the value for Re is always zero?

Again, many thanks for all of your input. I'm going to try the current limiter that Tomasz suggested earlier next. Then on to PMOS.

[gregebert]

Dont use the current mirror without emitter resistors; it will likely be inaccurate with discrete devices.

The current mirror as shown uses the operating point of Q2 (which is very controllable and stable) to set the current through Q1. The problem is that in this configuration, Q1's collector current is exponentially sensitive to Vbe. Any mismatch in characteristics between Q1 & Q2 will result in substantial current mismatch between Q1 and Q2. I believe a 26mV mismatch at room temp can produce a 2:1 mismatch in collector current. Circuit wiring will also be a factor because it only takes milliohms to contribute to mismatch.

In order to match the currents between Q1 and Q2 independently of device variations, you need to add equal resistance at their emitters. If you are targerting, say 3mA, an emitter resistor of 1K will greatly improve matching. The reason why is that variations in Vbe are much smaller than the voltage drop across the emitter resistor.

[Paul Andrews]

Next installment. First I tried the current mirror with 2K resistors between +300V and the emitters. This works fine, but is also no more stable than the first solution. It presumably also wastes more power as R2 has to be higher - it is setting the current, so I chose 100K here to get 2mA:

Next I tried the current limiter as described in several places, but in particular here as suggested by Tomasz. I used an Rsense (R3) of 330 Ohms as I wanted 2mA. This was not successful. With test loads it delivers <2mA (or exactly 2mA for a short circuit!). So I tried a 6844A nixie (which is what started off this whole current source thing). It delivered approx 0.5mA, so the tube barely lit.

[Paul Andrews]

Of course, I meant +200V, not +300V!

[gregebert]

I'm not able to reason-out how the last circuit acts as a regulator. Basically, the path from the 200V supply to Rload (output) is 2 diode-drops (base-emitter junction) in series with 100K of resistance. Assuming the nixie tube takes about 150V to ionize, a current of 0.5mA thru R1 causes a 50V drop.

Adding-up the voltage drops around the circuit, everything is accounted for:

200V supply = 0.5mA * 100K + Vbe +Vbe + Nixie_tube. Solving, the nixie tube will have around 148.6V, assuming Vbe of 0.7V. This is in the ballpark for nixie tubes.

If you want 2mA, reduce R1 to around 25K. But I dont think it will act as a regulator when you change the supply; the current will also change.

[Tomasz Kowalczyk]

Sorry for late reply, I've destroyed my boost converter somehow and had to rebuild it from scratch. I've tried the current limiter, but without success - I don't know why, I followed the article and I think that this just won't work for such voltage drops or something. Funny thing is that I've chosen a small R1 for one try and with intention to blow things up I've connected it directly to converter, without any load. Of course, R1 blew up, but funny thing is that after burning the circuit was passing 1,2mA - and I was aiming at 1mA :D

Of course this was just a resistor burning from 1,2k to around 250k, not a working limiter.

I have another idea. What do you think about using optoisolator in active range? It would require testing each batch of components, as they tend to have varying current transfer ratio, but you'd be able to control the current on low voltage side (even from DAC).

I'm looking for a stable method of controlling it without any resistors going from HV to ground, as it would waste alot of power, at least with bipolar transistors - with a p-mosfet it could work with resistances in voltage divider like 4,7MOhm, but mosfets tend to change their characteristics with temperature.

[Tomasz Kowalczyk]

Update: success! I've found a way to limit current using few components and without any voltage dividers.

The key was using a N-JFET. I guess if one used a N-MOSFET which opens with negative gate voltage, the result would be simmilar. I wonder if a similar circuit with a P-MOSFET would work, but I'm not going to check it soon.

The schematic is ultra simple:

https://i.stack.imgur.com/sKtpJ.png

I used a BF256B (the only N-JFET I have laying around) and changed resistor to 1k. The current is limited to 2mA - I've tested it with voltages from 0 to 30V and as soon as I reach 2V across it, it starts blocking current. It changes slightly with voltage, on 3V across circuit it is exactly 2mA and with 30V it is 2,1mA, but I think it is close enough. A 270 resistor gives about 4,5mA, the difference between 5V and 30V is about 0,15mA - still very good,

Drawbacks of this method? JFETs are only low voltage, max 30V difference between any two pins, usually. So to use those with Nixies, you'd have to use a resistor + this limiter - so if you have 50V to drop on current limiter and you want it to work on 2mA, you have to use a 22k resistor + this limiter to drop 44V across resistor and 6V on limiter. And, of course, it will blow if Nixie gets shorted, so whole circuit will have to drop 200V instead of 50v.

I'll probably test this on nixies during next days. 

Idea and schematics taken from http://electronics.stackexchange.com/questions/105326/is-there-any-physical-device-called-current-source/105394

[Tomasz Kowalczyk]

Update 2 - I've found a 130,5V transil at work, so I've been able to substitute a nixie tube with a nice dummy load. Results are: 208V supply, 139V dropped on transil, 49V dropped on 3x33k resistors in parallel, rest dropped on my limiter. Current is 4,55mA. Everything seems to work just as expected! 

[Paul Andrews]

Here is a link to a paper about using a depletion mode MOSFET for the same purpose. It gives this equation for calculating the resistor value, which I assume would be the same for all depletion mode MOSFETS (!?):

VGS(th) is the gate threshold voltage of the MOSFET, IDSS is the on current at VGS = 0 V and ID is the required current.

There are plenty of these around with suitable voltage characteristics.

[Paul Andrews]

Although a mosfet current mirror would use fewer components as you only need one Rset (section three of this paper).

I guess I need to place yet another order with digikey to try some of these out.

[gregebert]

My concern with current limiters that rely heavily upon the datasheet specs (Vgs  for Depletion-mode regulator; Vbe for current-mirror)  is that variations due to process & temperature will have significant impact on the actual current. Using a slightly more complex+costly design will mitigate this; well-worth it in my opinion when you consider the value of the tubes you are protecting. 

[Paul Andrews]

Totally agree. This is one reason I like to actually try these things out.

On Apr 7, 2017, at 11:02 PM, gregebert <greg...@hotmail.com> wrote:

My concern with current limiters that rely heavily upon the datasheet specs (Vgs  for Depletion-mode regulator; Vbe for current-mirror)  is that variations due to process & temperature will have significant impact on the actual current. Using a slightly more complex+costly design will mitigate this; well-worth it in my opinion when you consider the value of the tubes you are protecting. 

--
You received this message because you are subscribed to a topic in the Google Groups "neonixie-l" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/neonixie-l/gbJ0Beiu66I/unsubscribe.
To unsubscribe from this group and all its topics, 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/5c9c7941-efba-4967-acee-24c8701874b4%40googlegroups.com.

[Tomasz Kowalczyk]

I'm afraid all active current limiters will change with temperature, just some will be affected more and some less. But small changes (0,1mA or less) won't be critical in a nixie clock, I think. So if a circuit will be stable enough to provide desired current +-0,1mA in temperature range of 0°C to about 70°C, with supply voltage changing 20V max - I'd call it good enough.

About variations of Vgs curves - within one production batch of transistors differences are usually minimal. So after ordering a batch of transistors you have to check just one of them and you can safely assume that all other will behave almost identically. Vbe differences in current mirrors are much more critical, as the difference gets amplified.

If those two factors are still too big, then you have to use a circuit using a current sensing resistor, an amplifier of Rsense voltage drop and a drive transistor. Multiply that by 15 segments and by number of tubes and you end up with a large and costly circuit. I think that that level of current stability is not needed here.

I'll test J-FET current limitter with a hot air gun - how it behaves in temperature range of 20-150°C, both with 5V and 30V dropping on it.

If I find some time, I'll do the same with a very similar circuit using LM317 (similar - both use only one resistor and one cheap active component).

[GastonP]

Yes... JFETs are extremely variable on their parameters so circuits using them need trimming of the external components if one needs a certain degree of precision. That, or negative feedback.

[Tomasz Kowalczyk]

I ran the temperature test with 120°C hot air aimed directly at the JFET. Conclusion - the higher the temperature of JFET, the lower the current - so it is self stopping rather than increasing current, which is good for tubes (slight undercurrenting shouldn't damage them like overcurrenting would). Because of nature of the circuit (10k resistor + JFET limiter + Z567M as load), lower current means higher voltage drop on JFET, so power dissipation increases a bit, because voltage drop on the resistor decreases. I'm looking for a transistor with similar parameters as BF256B, but with higher breakdown voltage (at least 50V), so I could run the test without additional resistor. If I find one and it will pass the tests, I think that it would be it - a small circuit limiting anode current, allowing to conserve power by using HV just above striking voltage, while still having current under control. Next step would be some kind of a constant current step up converter.

If this is too unstable due to differences in Vg curves between transistors, then only an active current sensing and driving circuit is needed. I have an idea to use an isolated 5V supply, which would allow to use some simple circuits near HV voltage - an opamp amplifying voltage on current sensing resistor and driving a transistor, all connected somehow to HV. I'm afraid it is only a loose idea - I am not willing to test it myself, for me the JFET circuit is good enough, if I ever decide to use current limiting instead of good old resistors.

[Dekatron42]

Have you tried to ask for instance Supertex (nowadays Microchip) about their LR8, the IXCP10M45S (10M90S) or the DN2540 regarding CCS designs?

It seems like some audio people use the DN2540/IXCP10M45S in combination with either a low dropout voltage regulator like LD1085 or the LM317 to either make a high voltage regulated power supply or to make a CCS (see for instance the GlassWare PS-1 PSU: http://glass-ware.stores.yahoo.net/ps1kit.html and here: http://tubecad.com/2008/11/blog0151.htm). The LR8 (or the LR12) can be used as a CCS with just one resistor and it works up to 450V with a few milliamperes. Using the DN2540 with an LM317 is probably the easiest design. A DN2540 & LM317 CCS can be found here: https://vwws.wordpress.com/2011/06/21/dn2540-lm317-cascode-ccs/ .

/Martin

[Tomasz Kowalczyk]

I've ran some tests with BSS139 + a potentiometer as a variable resistor.

Conclusions:

 - this design offers cheap and stable current limiting

 - it varies with voltage dropped over the circuit, so it should be tunable - the difference is about 14% change between 5V dropped and 105V dropped across it (4,37mA vs 4,91mA).

 - it is thermally stable - at 100V dropped and 4,93mA cold current I am breaking the allowed power dissipation (allowed 0,36W, here - 0,5W), it heats up to a bit over 50°C (many probe cables connected close to the transistor act like a radiator), the current has changed by 40µA 

 - while testing it I found out that striking voltage of tubes is a max value - I've tested one Z567M and one LC-631, they both strike with voltages lower than their normal maintaing voltage! While powered correctly they sit at 140V across the tube, and I was able to lower power supply voltage to 135V, disconnect tube, connect it again and it worked. I suspected that due to large resistors and a cap in the feedback I have high pulses over 170V in my supply, but it is not true - at 135V the spikes measured with an oscilloscope were less than 5V. Of course, with voltage lower than 140V + minimum drop over BSS39 and resistor, the current was lowered.

 - due to that it is possible to create a clock wasting very low power on anode limiting - classic operation uses 170V, 180V or 200V supply + resistor, which wastes quite a lot of power on those resistor. here we can tune the power supply just above striking voltages of all tubes and be sure that we don't have to recalculate the resistors. This will have a very positive effect in clocks which run on batteries or USB supplies - with both lower voltage required and lower power consumption.

 - after tuning to desired current it is very stable - drift is max 20µA 

I wonder if this low striking voltage is common among different tubes or does the striking voltage change with temperature.

In my opinion the only drawback of this design is that it is changing the current with voltage drop across the limiter. This change in normal operation shouldn't be noticable, because normally power supply doesn't change much - maybe 10V under bad circumstances. With such voltage change the current will be pretty much the same, but a resistor calculated once isn't universal for every design. 

Everything else (simplicity, price) says that this is it, I'll be using this instead of resistors.

I'm attaching photos of setup - left multimeter measures voltage across the tube, middle measured power supply voltage, and the right one measures current (in mA). Voltage across limiter is of course the difference between middle and left.

[John Rehwinkel]

 - while testing it I found out that striking voltage of tubes is a max value - I've tested one Z567M and one LC-631, they both strike with voltages lower than their normal maintaing voltage!

Yes, it's a maximum value, so people can design circuits that are guaranteed to strike even with a worst-case tube, under worst-case conditions (see below).

I wonder if this low striking voltage is common among different tubes or does the striking voltage change with temperature.

Temperature has a minimal effect on striking voltage.  The big factor is something to start the ionization cascade.  If the tube is exposed to light, photons will do the trick.  Radiation of other forms will as well.  Worst case is in absolute darkness.  For some designs, striking speed also matters: the higher the voltage, the faster the tube will strike.  For some designs this can matter.

One workaround is to have a "primer" electrode, to provide a source of ions to start the tube.  While nixies don't normally come with primer electrodes, you can use a decimal point as a primer, just hook it up via a very large resistance.  This will reduce the striking voltage and time significantly in the dark.

- John

[gregebert]

My wristwatch uses a 'boost' approach to ionize the display above 180V for 25msec, then throttles back between 140 to 160V after the display is stable. The saved energy is significant. It's 3-1/2 digits, direct-drive, and uses NPN current-regulators for each segment (24 total).

My bench prototype has been running for over 2 years now on the original charge to the battery (3.7V Li-ion, 1050mA-hr). I dont display the time more than a few times per week, but the fact it's still operating is amusing. BTW, the battery was not new, either. It was used for a few years in my cellphone so it's capacity is diminished.

[John Rehwinkel]

My wristwatch uses a 'boost' approach to ionize the display above 180V for 25msec, then throttles back between 140 to 160V after the display is stable. The saved energy is significant. It's 3-1/2 digits, direct-drive, and uses NPN current-regulators for each segment (24 total).

Is that done in software, or what?  It just occurred to me that you could add a parasitic multiplier "starting circuit" similar to the ones used in helium-neon laser power supplies to do this automatically*.  However, another multiplier segment just might produce enough voltage to endanger other components (depends on what the AC drive voltage is, and the margins on the other components).

* such lasers generally run on 1700V or so, but can require upwards of 10kV to start

My bench prototype has been running for over 2 years now on the original charge to the battery (3.7V Li-ion, 1050mA-hr). I dont display the time more than a few times per week, but the fact it's still operating is amusing. BTW, the battery was not new, either. It was used for a few years in my cellphone so it's capacity is diminished.

That is both amusing and cool.

- John

[gregebert]

It's done in FPGA code (verilog). I created a crude A-to-D converter using a resistor tree into 4 FPGA pins. The resistors are ratioed to give the FPGA indication when the anode voltage is 140,160, 180, or 200V. From there, the FPGA adjusts the duty-cycle of the DC-DC converter to change the voltage. I also use a soft-start on the HV DC to minimize Ldi/dt effects on the wiring, give the HV filter cap (0.3uF) a chance to charge-up, and minimize current-spiking the battery (my naive assumption is chemical reactions in a battery are not instantaneous).

[Tomasz Kowalczyk]

Wow. I didn't think much about how the ionization starts. I was quite surprised as after reading this I turned off all lights in my room and with 150V the same LC-631 didn't start - but as soon as I put some light on it, it indeed started glowing.

Unfortunately I can't use any decimal point as a ignition starter for the simple reason - almost none of B13B socket tubes have a decimal point :) and I own mostly those tubes (ZM1040, Z566M, LC-631, Z560M).

Thank you for sharing this information and making details of how nixies work more clear to me. Also thank you for sharing the idea of programming the boost converter to have a startup routine - this is so simple and yet I didn't think about it. 

Do you know if there is an effect of lowered striking voltage for some time after the tube is turned off? I'm curious if it is possible to add PWM dimming or even multiplexing with 145V power supply with 180V starting routine. As I tested my LC-631 it seems to light up properly in darkness after it was lighted once with my desk lamp - after that I can disconnect it, wait few seconds and reconnect and it works immidiately. I don't know if it is a rule or just a coincidence.

[JohnK]

I told this story here to the previous generation of readers...

In a facility where I worked for a few years back in the mid-1970s there was also a two-rack-cabinet trigger tube monster.

Its task was to check the teletype data that was arriving over an HF radio network. When [parity] errors were detected the device issued a request for a re-send.

The story goes that one Friday there was a fault and it was very intermittent. The guy working on it was tearing his hair out and went home late that night. It ran fine for the rest of the weekend.

Monday morning  came and the equipment room staff tidied up and closed the cabinet doors. Within hours the intermittent was back.

It took a while but eventually the penny dropped. The room was lit with strip lighting [fluorescent tubes] banks that were seperately switched and also a couple of PAR 38 flood lamps had been mounted on the ceiling pointing into these cabinets because of the 'strange' construction and difficulty of working on it. The room temperature did vary quite a lot too surprisingly. Desert conditions outside with temp differentials of twenty-odd [degrees] C.

They eventually noticed that the fault wasn't present when the cabinets were open and the bright lights were on.

The particular trigger tubes became hard to get and the particular guys working on the equipment didn't manage to identify which tubes were the problem and wanted to use the 'shotgun' approach with the fault. [Longer story there  :-))   ]

Eventually they took the doors off and made sure that the internals were brightly illuminated!

 

John K 

 

 

 

--
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/aa645b68-5fa0-4212-8b9d-b75f69e6794f%40googlegroups.com.

[Paul Andrews]

You need a few ions around in the Neon for the cascade to start. This can happen for a variety of reasons including cosmic rays, background radiation or doping of the gas mixture with something radioactive. There is at least one thread on here about radioactive nixies (which sent me off on a side quest that hasn't quite finished yet!). Apparently 5092's are dosed with Kr-85 and have a little radioactivity symbol on them.

[M W]

That story reminds me of one from work. Not Nixie-related but sorta funny :)

Years ago, work had an IBM System/360 ot /370 of some age; It started having faults at random times, so they called in the tech. He opened the system covers, tied in all the diagnostics tools - logic analyzer, etc, but, zero faults happened over a couple days of his running those; So they pulled them out & closed the system up. Next day a few faults happened, so he was called back in, wired it up and ran it another week, no faults. He pulled the test gear & closed it up, but it crashed before he even made it to his car; He was called back in and of course once wired up, zero errors...

The light finally lit up on someone's brain in there (reports varied on whose!) - That the system cover doors, when OPEN, would preclude any errors, but when they were closed, errors would occur; So they looked at the wiring harnesses and found the harness that was flexing when the doors were closed, which had a nasty intermittent in there, that was only going open circuit VERY rarely, if quite enough to be horribly ANNOYING.

Systems Engineers types and so on, HATE intermittents, they're the bane of their existence :P LOL And they're quite annoying to debug, sometimes you can't figure out where the darned things are hiding at all.

  Mark

[Terry S]

There is a very similar story about an early super computer developed at Control Data in the 70's. 

An AC outlet on the side of the machine was intended for engineering for debug instruments.

But nightly, when the cleaning lady plugged in her vacuum cleaner, the machine would crash. It took days to figure it out.

Terry

[Nick Sargeant]

Oh, I could tell you some stories. When we did some early ships of advanced workstations to universities, I had a bunch of complaints from Cambridge University that their optical mice were failing randomly. I phoned the lab tec to find out what was going on .. these mice used two colours of LED, and a reflective pad with a grid of lines on. So, he told me that one in particular was random, jerky, and sometimes the mouse pointer wouldn't move at all. I asked him for the serial number of the failing mouse (thinking was there a bad batch?) .. he said Hang on, just have to shut the blinds so I can read the number - sunlight is too strong ... (a long pause) ... oh, wait, it's working now.

And that was a leading academic institution. Speaks to the difference between intelligence and common sense, I feel.

I had another complaint from a customer somewhere in NY state that one of my workstations had a CRT screen where the image wobbled. Just the one. .. on a hunch I asked him where it was in the room. Next to a wall, he said. Can you move it away from the wall? Oh, look, it has stopped wobbling now. What's the other side of that wall? A huge transformer serving power to the building apparently. *sigh*

[Paul Andrews]

I finally got around to trying this using a 200V supply and one of these. I tried it first for a 2mA load current. I had to determine the Vgs experimentally. The datasheet looks like it should be about -2.5V. I got about -1.9V. I gave it a 5K load. The mosfet heats up quite a lot while operating, so the current across the load gradually climbs, though it wasn't too bad - the end result was about 2.05mA

I then tried it for a 4mA load current. The mosfet got really hot, and the current climbed up to about 4.5mA before I disconnected it.

In short, I won't be using this method to limit current.

BTW I tried two different pieces of the same mosfet - the Vgs was the same for both. I ordered them at the same time so they could well be from the same batch. So far, if I wanted to use a constant current source, I would go with the very first option.

On Friday, April 7, 2017 at 5:34:49 PM UTC-4, Paul Andrews wrote:

[gregebert]

A 5K-ohm load with 2mA thru it only drops 10V; with a 200V supply, you would have 190V across the mosfet. At 2mA load current, that translates to 380mW dissipation in the mosfet. From the datasheet, the max Theta ja I saw listed was 170 C/W. That would translate to a 65C rise, which is fine for room-temp even in a hot climate.

At 4mA, though, you will get into trouble as you noted.

Typically you wont see anywhere near that much voltage across a nixie current-limiter. Assuming 150V typical nixie voltage, and 10V for the resistor, you will get 40V across the transistor. That translates to 80mW, and a temp-rise of about 15C, which is plenty of margin.

A current limiter wont reduce the overall power dissipation, but it will reduce the variation in nixie current over supply-voltage variations and nixie aging.

[Paul Andrews]

Well that makes me feel dumb (which is fine, I will learn)! I should have had the sense to calculate a suitable load value and I should read the data sheets more carefully! Hopefully my trial and error will be instructive to others as well as myself. 

I'll do it again with calculated load values, if only to ram the lesson home.

I will also get around to trying out the mosfet version of the pmos current source at the start of this discussion.

Thanks for all your input into this. 

--
You received this message because you are subscribed to a topic in the Google Groups "neonixie-l" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/neonixie-l/gbJ0Beiu66I/unsubscribe.
To unsubscribe from this group and all its topics, 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/820f82b0-8b70-4c70-b589-d224f70db2f3%40googlegroups.com.

[gregebert]

Nobody in this forum is dumb; some of us just havn't been bitten as hard as others........I found out the hard way you have to read and understand every spec item in the datasheet, or you will get into trouble. I've refused to use quite a few parts because a spec value wasn't provided; other cases (like the HV5530), I was forced to design-around a missing spec item, like min prop delay.

[Paul Andrews]

Hi (Greg?),

Using the datasheet for the PMOS transistor you use in your circuit, how do you calculate the value of R116 (the resistor between +200V and the source of your MOSFET)?

On Saturday, March 25, 2017 at 12:30:53 PM UTC-4, gregebert wrote:

That's basically what I use in my designs. I'll highlight the differences:

• I use a PMOS instead of PNP, mainly because it requires no drive-current.
• R1 & R2 are replaced with a pot to make the current adjustable.
• The above pot can driven from a small DC-DC converter (my preference), or between the HV supply & GND. There's essentially zero current for PMOS gate-drive, so high resistance values are fine. Not the case with PNP, though, due to finite base-currrent.
• A zener diode is added to clamp any spikes that may arise at the gate of the PMOS device. It's a paranoia item.
• A filter cap was added, in case there is unexpected noise from the DC DC converter, and also to suppress any very-short transient that may arise that are too fast for the zener to kick-in. (paranoia item).
• A large resistor across the PMOS to bleed any potential ESD. Without it, there is a remote possibility of charge-buildup. (paranoia item)

So, this circuit is replicated for each anode. When multiple anodes are driven, they all share the same gate-drive signal, which I call PDRV on the attached schematic.

[gregebert]

Assuming you have high-enough HV power-supply headroom for the nixies (180V or more), the choice for the resistor begins with the isolated power supply; I chose 12V because there are a lot of small, efficient DC-DC converters that provide 12V. You can use other values, but be careful to stay below the rated Vgs(max).

Next, you want some adjustment range for the anode current. I set the nominal gate voltage at 10V. That gives +/- 2V of range to adjust the anode current without changing resistors.

Next, from the datasheet, you need to find Vgs(on) for the MOSFET. It's usually plotted on a curve. For the FQD7P20TM that I use, it's about 4V (see the transfer characteristics curve).

Next, you need the target anode current. For a Burroughs 5092, it's 2.2mA.

The voltage-drop across the resistor is 10V - Vgs(on); in this case, it's 6V. Now you can solve for the resistor value:
   R=V/I    = 6/0.0022   = 2727 ohms. The nearest standard value = 2.7K.

Sanity-check the power in the resistor: P=I^2R; in this case,  P=13mW, so even with tiny surface-mount devices you are fine.

You should also sanity-check the PMOS transistor worst-case power dissipation. Assume a ridiculously low nixie voltage of 125V. The voltage across the PMOS is Supply_voltage - Nixie_voltage - Resistor_voltage = 200-125-6.

PMOS power = Max voltage * typical current = 69*0.0022 = 150mW.

This device has a Theta j-a of 110C/W, so the temp-rise = 110*0.15 = 17 degrees. Plenty of margin here.

[Luka C]

I'm glad you've done the test with BSS139 and it turned out to work fine. I've used the same transistor for the cathode current control on my B7971 clock, completed routing the PCB and currently sparing some money to send it to the fab house :)

[gregebert]

Luka - What segment currents are you using for the 7971 ?

If you turn all segments on at the rated datasheet current, there are several characters, such as 8 and Q, that will cause the total current to significantly exceed the max rating of 22mA. I analyzed current-draw for most characters, and even the average (~26mA) exceeds spec.

I scaled-back the current in my design so I never exceed the max 22mA.

7971's are really expensive & rare, so I'm not taking any chances.

[Luka C]

@greg, I have taken a look at this: http://tayloredge.com/storefront/1386_B7971SmartSocket/1386.pdf . When using 175V as supply voltage and assuming 140V voltage drop on the tube, the currents should be: 

S15, S4, S1 = 1.59mA

S14, S10 = 1.06mA

S13, S11, S9, S7 = 1.46mA

S12, S8, S6, S5, S3, S2 = 1.30mA

When you add them all together, you get 20.53mA which is just right under the specs of the datasheet so I adjusted the currents according to it. I know this will result in less brightness than the maximum possible when all segments are not lit up because plenty of margin will be left then, but I see no other way, other than implementing the anode current limiter as well as you have described in earlier posts.

[Nick]

I like the general idea of what your attempting, but it's really not suited to segmented tubes like the B7971.

The Burroughs data sheets for the valve clearly state the currents for each segment and give examples for the differing cathode resistors to use to achieve this on a per-cathode basis.

There is no "right" single anode current for a B7971 - they are also pretty bullet-proof - I've never known one fail and have several hundred out there. Bearing in mind that all known B7971 are ex-equipment, I would stick with the Burroughs recommendations on those.

For single-cathode/digit nixies, a current limiter would be fine.

Cheers

Nick

[gregebert]

The problem is that if the segment currents stated in the Burroughs datasheet are summed-up for various characters, many characters exceed the max-spec for anode current. However, the average character-current is on-par with the max anode-current spec. Unfortunately, we'll never know what Burroughs intended.

I took a different approach and gathered a lot of data on my 7971's, including a guesstimate of the current that gives good visual illumination for each segment.

When I scaled-back the spec-value for segment currents to get most characters within the 22mA max anode-current spec, I found that the 'visually good' current was on-par with the scaled current. What I ended up with is segment currents that are visually good, and within the max-anode-current spec limit (slightly over for 2 characters).

Hopefully I will have time this weekend to finish my board and try it out.

[Luka C]

@Nick - I know that cathode current control per segment might seem a bit of an overkill, but as @greg said, considering the price of the tubes, I think we should do our best to meet all of the datasheet specs if possible. And if you carefully look at the datasheet, consider the case you want to display a character like "*" (asterisk), then add together the currents of each segment from the datasheet and it will exceed the maximum anode current defined in the same datasheet, this is why each cathode current should be adjusted so that if you add them all together, the sum will be lower than the maximum anode current (of course, how much this really impacts the life of the tube will remain a mystery I guess, but we should still be careful)

This could be achieved by adjusting the resistor value in each of the segment lines but this carries a side effect. These tubes will presumably run in our clocks for years and during the that time, their characteristics will probably change. The voltage drop on the tube will probably change and if the supply voltage is fixed this will also result in changes to the segment current (Is = (Ua - Um) / Rs). If Um rises over time, this would result in lower voltage drop across the resistor and in turn lower the segment current (perhaps to the point where brightness might not be what we would like it to be) and this is why a constant current sink might come handy to make sure current in each segment remains fixed over the years of operation.

[gregebert]

In addition to current-control per-segment, I also have an anode-current limiter. It's probably overkill but it does cover the case where I have a few characters that exceed to max total current.

I'm trying to fire this up for the first time this weekend. I've never used a RasPi before, so I have a bit of software to develop.

[Tomasz Kowalczyk]

I'm happy to read that.

I'm curious what would happen in a circuit with both anode 22mA limiter and cathode limiters after the anode limiter kicks in - will the current of each lit cathode get lowered by same percentage or will it equalize to same value on each cathode and then lower further with the same value? 

Anode limiter could possibly lead to uneven illumination of segments, if first option was the real one. Unfortunately, B7971s are totally out of my league and I can't test it myself.

[gregebert]

I hope to find that out this weekend.  My hunch is that segments with the highest voltage-drop will be impacted the most. I dont have my IV curves handy at the moment, so I cant say which particular segments would be affected. I recall seeing 3 clusters of IV curves for each tube.

[Nick]

On Friday, 12 May 2017 17:42:58 UTC+4, Luka C wrote:

@Nick - I know that cathode current control per segment might seem a bit of an overkill, but as @greg said, considering the price of the tubes, I think we should do our best to meet all of the datasheet specs if possible. And if you carefully look at the datasheet, consider the case you want to display a character like "*" (asterisk), then add together the currents of each segment from the datasheet and it will exceed the maximum anode current defined in the same datasheet, this is why each cathode current should be adjusted so that if you add them all together, the sum will be lower than the maximum anode current (of course, how much this really impacts the life of the tube will remain a mystery I guess, but we should still be careful)

Are you sure you're not looking at the maximum rather than the normal current ratings for each element ?

Nick