Driving a VFD tube with an ATX PSU

[CrazyHans]

I have a spare ATX supply and have been thinking about using it to drive 6 multiplexed 7-segment VFDs. My problem is that the grid needs 24V and the filaments need 5V, getting the 24V is easy by using the 12 and -12V lines from the PSU, but the filaments need to have 5V which I can't get because there's no -7V line coming out of the PSU. I was wondering whether the following circuit would be appropriate, or blow up in my face.
Should work right? idunno. pls help before I set my house on fire.

[amyk]

No, shorting the 5V and 12V outputs together will definitely NOT work.

[Whales]

Be wary that the negative rails will only be rated to a tiny amount of current.  Additionally many modern ATX PSUs simply don't include them.

[CrazyHans]

No, shorting the 5V and 12V outputs together will definitely NOT work.

They aren't? My diagram sucks. The 5V is running through the VFD filament, and the 12V is running through the VFD grid; then they run to common ground (or in this case some janky ass weird shit).

Be wary that the negative rails will only be rated to a tiny amount of current.

What about the current coming from the +12V rail?

[Whales]

If it's a modern PSU  the +12V rail (or rails) will happily do many, many amps, and the 5V should be able to do a few.  Your PSU should have a ratings sticker attached to the side that has all of the stats you need.  If your PSU is cheap or old and scavenged: derate what is claimed by a massive amount (eg half).

Your diagram does look odd -- it's not clear you are trying to make a virtual ground.  I presume that "filament 5V" means the opposite end of the filament is connected to +5V and similarly for the grids?.  Should one of the diodes be the other way around?  Why do their reverse breakdowns not add up to 24V?

I don't believe an ATX power supply is a good choice for 24V.  It will be difficult to find one that can deliver the current you need on the -12V rail, and if you do it will either be expensive or old.  Modern computers primarily use the +12V rail and a small amount of 5V and 3.3V with the negative voltage rails generally omitted as legacy requirements.

Side note: computer power supplies expect they have to only deliver current to the positive rails and only receive it on their negative rails.  If you hook something up between the 12V and 5V rails (net difference=7V) then you are taking current from the 12V rail and sending it into the 5V rail, which will charge up the capacitors on the 5V rail.  This will slowly raise the voltage on the 5V rail until the PSU supervisor/controller decides that there is a fault on the 5V rail and shuts down.

[Cliff Matthews]

I agree, the ATX psu is unsuited and overkill the purpose. I think you'd be better served by series joining two laptop adapters to get some more usable + and (-) rails. Maybe a cheap charge pump or resistor divider will then help wrap it up?

[Ian.M]

VFDs need AC drive to the filament, otherwise one gets an objectionable brightness variation from one side of the display to the other.  The traditional way to do this when a mains transformer windingfor the filament isn't available  is with a Royer, Armstrong or blocking oscillator with a center tapped secondary, the desired cathode DC bias voltage being applied at the center tap.   The bias voltage was often derived from rectifying another secondary on the same transformer.

However a capacitively coupled push-pull 50% PWM would also be suitable, with the bias applied at the center tap of a resistive divider across the filament.   The display multiplexing rate should be locked to the PWM period so N whole PWM cyles are used for each multiplexing time slot.   

Getting a 24V supply is a bit more awkward if you aren't using an oscillator for the filament.  As you only need a small current, an isolated DC-DC converter module is an option.

[CrazyHans]

I agree, the ATX psu is unsuited and overkill the purpose.

I already have an ATX supply which I got for free (practically brand new, 900W Antec PSU. Free. idek), and I've been thinking about just forking over the $20 for a 12-24V dc-dc converter (actually it's 24-12 but if my high school physics serves me it should work in reverse right?).

VFDs need AC drive to the filament, otherwise one gets an objectionable brightness variation from one side of the display to the other...
...an isolated DC-DC converter module is an option.

So all I need to do is get an oscillator for the 5V rail and a converter to take the 12V up to 24V?

Also damn you guys are helpful. <333

[Ian.M]

DC to DC converters DO NOT work backwards.   a 24V to 12V converter will be absolutely useless for this.  Also the price sounds about three times too high.  You would need a little PCB mount module rated at a watt or two e.g. http://export.rsdelivers.com/product/recom/ro-0524s/1w-isolated-dc-dc-converter-vin-45-55-v-dc-vout-24v-dc-1000v-dc/6728773.aspx
which runs from 5V and gives you approx 40mA at 24V out.  We'd need to know what VFD you have to tell you if 40mA is enough.

The whole circuit would run from 5V e.g. a 1A or 2A USB phonre charger PSU.

[Whales]

Yes, be wary that transformers are not the same as DC-DC converters. 

So all I need to do is get an oscillator for the 5V rail and a converter to take the 12V up to 24V?

Yes, this line of thought sounds good.

It's worth seeing how much current your filament consumes at 5V.  If it's decently low an SN754410NE motor driver would be one solution to turning the 5VDC into AC.  You would need to tell it to oscillate back and forth using eg a microcontroller (I presume your project has one) or a 555 timer.

[CrazyHans]

DC to DC converters DO NOT work backwards.   a 24V to 12V converter will be absolutely useless for this.  Also the price sounds about three times too high.  You would need a little PCB mount module rated at a watt or two e.g. http://export.rsdelivers.com/product/recom/ro-0524s/1w-isolated-dc-dc-converter-vin-45-55-v-dc-vout-24v-dc-1000v-dc/6728773.aspx
which runs from 5V and gives you approx 40mA at 24V out.  We'd need to know what VFD you have to tell you if 40mA is enough.

The whole circuit would run from 5V e.g. a 1A or 2A USB phonre charger PSU.

http://www.tubeclockdb.com/vfd-tubes/85-large-vfd-tube-ilc1-17.html here are the specs on the tubes, 40mA isn't quite enough.
http://export.rsdelivers.com/product/recom/ro-1224s/1w-isolated-dc-dc-converter-vin-108-132-v-dc-vout-24v-dc-1000v-dc/6728791.aspx Would I be able to use one of these? Overkill though it may be, why not use the atx supply?

It's worth seeing how much current your filament consumes at 5V.  If it's decently low an SN754410NE motor driver would be one solution to turning the 5VDC into AC.  You would need to tell it to oscillate back and forth using eg a microcontroller (I presume your project has one) or a 555 timer.

Yeah I'll be running the whole thing using a MAX6920 and a knockoff arduino nano.

EDIT: seeing as the -12V rail is capable of 500mA, I was thinking maybe I could run the +3.3V rail through one of these http://export.rsdelivers.com/product/murata-power-solutions/mej1d1505sc/1w-isolated-dc-dc-converter-vin-135-165-v-vout-5v-52kv-dc/8294792.aspx + an oscillator to -12V. Would that work?

[Ian.M]

Those are *BIG*.   The total grid + anode current is 180mA. 

 A common way of driving VFDs is to apply the HV supply as a negative potential on the cathode, then use PNP transistors hanging off your MCU's Vdd rail to control anodes and grids.  As negative output DC-DC converters are somewhat rare I suggested an isolated one.   

However your MAX6920 takes a positive HV rail and does all the level conversion internally, so you don't need an isolated DC-DC converter, you just need a boost converter.  At approx 200mA out,  it would be unwise to boost from 5V as that would need about an amp for the boost converter alone and as your Arduino can run from a 9v or 12V supply, I suggest powering the boost converter from that.  It shouldn't need more than 600mA at 9V and less at 12V.

Why not use a cheap adjustable DC-DC boost converter module like this one on EBAY?
http://www.ebay.com/itm/DC-DC-Adjustable-Step-up-boost-Power-Converter-Module-XL6009-Replace-LM2577/371054648639

[CrazyHans]

Those are *BIG*.   The total grid + anode current is 180mA. 

 A common way of driving VFDs is to apply the HV supply as a negative potential on the cathode, then use PNP transistors hanging off your MCU's Vdd rail to control anodes and grids.  As negative output DC-DC converters are somewhat rare I suggested an isolated one.   

However your MAX6920 takes a positive HV rail and does all the level conversion internally, so you don't need an isolated DC-DC converter, you just need a boost converter.  At approx 200mA out,  it would be unwise to boost from 5V as that would need about an amp for the boost converter alone and as your Arduino can run from a 9v or 12V supply, I suggest powering the boost converter from that.  It shouldn't need more than 600mA at 9V and less at 12V.

Why not use a cheap adjustable DC-DC boost converter module like this one on EBAY?
http://www.ebay.com/itm/DC-DC-Adjustable-Step-up-boost-Power-Converter-Module-XL6009-Replace-LM2577/371054648639

I'm not entirely sure I'm following what you're saying, but if I am, you're saying I just have to put the 12V rail through a cheapo boost converter, then into the driver chip, then I use an oscillator for the 5V rail into the filament?

[Ian.M]

Exactly.
Your next task is to power up just the filament on its own at its rated voltage and measure the voltage and current so you  can work out how much filament power you need.  I'd start with an adjustable supply and work up from zero in case you've misunderstood the display's specs.  If the filament starts to glow enough to notice, you are probably overrunning it so should stop, switch off and double-chec the specs.

[CrazyHans]

5V on the filament drawing however much it wants from the PSU works great, and funnily enough just using 12V on the grid and segments is reasonably bright (enough to see under my room's lightbulb) I can't wait to see how bright these things get at 24V. Also I couldn't make out any difference in brightness for the upper and lower segments running the filament on DC, should I bother with the oscillator, or wait til I get all 6 tubes running and tested?

Thanks so much for your help btw.

EDIT: If I can get this thing running off a much smaller 12V supply then I should probably do that, I figure I get one of each of these
http://www.ebay.com.au/itm/DC-DC-CC-CV-Buck-Converter-Step-down-Power-Supply-Module-7-32V-to-0-8-28V-12A-GA-/181617142420?pt=LH_DefaultDomain_15&hash=item2a4939aa94
http://www.ebay.com.au/itm/10-32V-to-12-35V-6A-Step-Up-Voltage-Charger-Power-DC-DC-Boost-150W-Converter-/281282823819
To get the 5V for the filament and the 24V for the segments, picking these ones over the cheaper ones because heatsinks are good. (right?)
Found this baby in my drawer

Should fit the bill right?

[Ian.M]

The Jentec power brick should be OK.  Even it is probably over-kill.

Why use an expensive more bulky boost module?  Even if you run the individual VFD digit tubes non-multiplexed at full brightness, brightness, that's less than 0.75A for four so the tiny heatsinks wont be an issue.

Brightness variation across a VFD is far more of a problem with multi-digit ones with horizontal filaments.   You should be able to use DC for single digit VFDs without significant issues, however I would advise reversing it occasionally (swap filament connections) to even out the wear on the filament's high emission coating.   However, if you are building this as a general purpose VFD driver board and ever want to hook up multi-digit VFDs, the AC filament drive will be essential.   You don't actually need to AC couple the filament with the chip you are using, just use a DC coupled full bridge (i.e. a motor driver chip) fed using an adjustable buck converter module to set the filament voltage, and power that off the 12V rail.  To compensate for the DC coupling, simply increase the HV rail by half the filament voltage. 

[CrazyHans]

Cool, idk if you noticed but the brick has a 5V rail too so the step down won't be necessary (right?). I think I will put in an oscillator just because it's best practice and the display will be running pretty much non-stop.

Just to be clear, I'll be setting this up with the 12V rail running through a boost converter, through the 6920, to the tubes. And the 5V rail will run to the microcontroller, the 6920, and (via an oscillator) to the filaments.

Thanks again, a lot.

[Ian.M]

A step-down (buck) converter lets you tweak the filament voltage.  If you run it on fixed 5V you don't get that option.   Start at the lower limit of the normal filament voltage range and tweak it up a bit as the display ages to compensate.  NB. DO NOT underrun the filament and attempt to compensate by boosting the HV to the grids and anodes as that can cause rapid wear to the oxide coating.

The phosphor on the anodes wears out as well, faster at high brightness.  Again its best to underrun them initially and turn up the HV a bit as they age. 

You will want to be able to shut off filament power as leaving the filament running with the display blanked leads to cathode poisoning (anothe form of damage that loweres the emission).  Adding a PIR motion detector and shutting off the whole display after half an hour of no motion detected by turning off all segments and putting the the motor driver chip that is running the filiment into its stop mode will do a lot to extend the tube life.   I'd also add a photo cell so I could dim it to follow the ambient light intensity by PWMing the MAX6920 blanking pins or by reducing the duty cycle if I'm already using them for multiplexing.  That also will help extend the life by reducing wear on the phosphor.

 

[CrazyHans]

Well then since the current rating on the 5V rail is way overkill, couldn't I just run it through a potentiometer instead?

Regarding the duty cycle of this thing, it's going in an office so will probably need to be running about 10 hours a day, 5 days a week.

[Whales]

Potentiometers that can survive the heat of being used in series with a device pulling power tend to be very large and expensive.  You are much better off using a cheap potentiometer to control a tiny amount of current that tells a voltage regulator what to do rather than shoving all of the current through a massive pot.

[Ian.M]

Well then since the current rating on the 5V rail is way overkill, couldn't I just run it through a potentiometer instead?

An adjustable filament voltage is a refinement you may be able to skip completely, but if not, you could add a wirewound pot wired as a variable resistor in series with the filament.  However that is only any good if the desired filament voltage is less than 5V minus the voltage drop in the full bridge (motor driver) chip.   Depending on the full bridge chip you use the voltage drop could be significant.  Also a suitable wirewound pot is likely to be far more expensive than a buck converter off EBAY.   

Regarding the duty cycle of this thing, it's going in an office so will probably need to be running about 10 hours a day, 5 days a week.

That's 2500 hours a year - so it could be significantly faded in as little as four years if you start with the brightness too high.

[CrazyHans]

okey dokey, so I also get an LDR and a motor driver chip and that should do it. What kind of driver should I buy?

[Ian.M]

We'd need to know the filament current @5V and also its cold resistance to recommend anything specific.  I assume you need to drive the filaments for all four digits in parallel?

[CrazyHans]

Six digits, it's not a clock it's a counter that displays the number of clients the business has. I'm also pretty sure my multimeter is dodgy (not surprising, $10 digitech piece of crap) because it won't measure the amps being drawn by the filament no matter what I try. But seeing as the spec sheet has been dead on about everything else so far, I see no harm in saying the filament is probably drawing ~150mA.

[Ian.M]

So you are going to need a total of 900mA of filament drive, but significantly more when initially starting from cold.  I'd probably use a L298N quad H-bridge, paralleling its sections in pairs for higher current capability and lower saturation drop though it might be worth considering TI's DRV8833, DRV8835 or DRV8838 MOSFET H-bridges (Polulu do DIL formfactor breakout boards for them).

If you use a L298N you wont be able to run it from the 5V rail and get 5V Pk-Pk out due to its internal saturation drop, so will need the adjustable buck converter module so you can set  it for 5V across the filament.