Is this a hazardous design?

[Mr.CRC]

Hi:

I am designing a few vacuum fluorescent display clocks using Russian
single digit VFD tubes. One thing I spent some time thinking about how
to do in my own way was the filament drive. I wanted AC drive, but not
with magnetics. They are either too big, too expensive, too much
trouble to design to a cost optimum, or too time consuming to wind
myself in quantities of more than 1-2.

So I settled on a capacitively coupled half-bridge.

The only drawback, is that at present it is open-loop. It takes its
bridge power from either the +5V main logic supply, or an optional
alternate +12VDC bus. The PWM source is designed to be flexible. The
PCB can be configured (by not populating some parts, and setting
appropriate jumpers) to take alternating PWM pulses from a uC capable of
this. Or a single PWM signal from a uC, with pulse steering logic on
board. Or by using an on-board 555 oscillator with an inverter
available to allow configuring for nearly any duty cycle to feed to the
pulse steering.

The AC coupling, as would be true with a transformer, removes the risk
of blowing the filaments if the PWM locks up, which is a significant
risk if a uC is the PWM source.

So the remaining hazard is that the power supply voltage might rise too
high and fry the filaments.

Oh, a secondary hazard is what happens if one tube filament opens. Then
the voltage will rise on all the other tubes a little bit. Perhaps the
user would notice a blank tube and seek repair before this leads to
runaway, so this is probably a negligible concern. It's more of a
concern on a socketed tube design where one can easily pull out tubes.

This is a rather remote danger but it does disturb me a bit. If it is
just for my personal hobby clock, and I blew a set of tubes while having
a lot of spares, it wouldn't be a big deal.

But I plan to sell a few of these, probably when I retire. So I'd hate
to have one go poof where my reputation is involved (well, honoring a
warranty and being generous when out of warranty can go pretty far as
well). But the costs of going closed loop are not desirable.

I know many VFDs are driven open loop from AC lines (via transformers)
and the like. So am I being paranoid?

This is actually an interesting exercise for me because I don't normally
have to think about costs in my work. So having to deal with cost vs.
reliability trade-offs is an interesting new challenge.

 

[Rich Grise]

I am designing a few vacuum fluorescent display clocks using Russian
single digit VFD tubes. One thing I spent some time thinking about how
...
So I settled on a capacitively coupled half-bridge.
...
The only drawback, is that at present it is open-loop. ....
This is actually an interesting exercise for me because I don't normally
have to think about costs in my work. So having to deal with cost vs.
reliability trade-offs is an interesting new challenge.

Zener diodes?

Good Luck!
Rich

 

[Joerg]

Hi:

I am designing a few vacuum fluorescent display clocks using Russian
single digit VFD tubes. One thing I spent some time thinking about how
to do in my own way was the filament drive. I wanted AC drive, but not
with magnetics. They are either too big, too expensive, too much
trouble to design to a cost optimum, or too time consuming to wind
myself in quantities of more than 1-2.

So I settled on a capacitively coupled half-bridge.

Make sure that capacitor can handle the AC current over the long haul.

The only drawback, is that at present it is open-loop. It takes its
bridge power from either the +5V main logic supply, or an optional
alternate +12VDC bus. The PWM source is designed to be flexible. The
PCB can be configured (by not populating some parts, and setting
appropriate jumpers) to take alternating PWM pulses from a uC capable of
this. Or a single PWM signal from a uC, with pulse steering logic on
board. Or by using an on-board 555 oscillator with an inverter
available to allow configuring for nearly any duty cycle to feed to the
pulse steering.

The AC coupling, as would be true with a transformer, removes the risk
of blowing the filaments if the PWM locks up, which is a significant
risk if a uC is the PWM source.

So the remaining hazard is that the power supply voltage might rise too
high and fry the filaments.

Oh, a secondary hazard is what happens if one tube filament opens. Then
the voltage will rise on all the other tubes a little bit. Perhaps the
user would notice a blank tube and seek repair before this leads to
runaway, so this is probably a negligible concern. It's more of a
concern on a socketed tube design where one can easily pull out tubes.

Operating the thing with some tubes pulled would fall under the category
"user stupidity", or maybe even "daft". That's like roaring off in a car
with numerous lug nuts missing.

This is a rather remote danger but it does disturb me a bit. If it is
just for my personal hobby clock, and I blew a set of tubes while having
a lot of spares, it wouldn't be a big deal.

But I plan to sell a few of these, probably when I retire. So I'd hate
to have one go poof where my reputation is involved (well, honoring a
warranty and being generous when out of warranty can go pretty far as
well). But the costs of going closed loop are not desirable.

Cost? Huh? A few resistors and caps, that's all. If you want it simple
you could consider hysteretic operation which even a uC should be able
to do.

I know many VFDs are driven open loop from AC lines (via transformers)
and the like. So am I being paranoid?

This is actually an interesting exercise for me because I don't normally
have to think about costs in my work. So having to deal with cost vs.
reliability trade-offs is an interesting new challenge.

You have to get the datasheet for those tubes and check the filament
voltage tolerance. Usually quite high. I don't know about Russian tubes
but figure 10 in this Noritake document shows that the luminance remains
relatively flat from 60% to 100% filament voltage:

http://www.noritake-elec.com/vfd_operation.html

Often people "revive" tired VFDs by increasing the filament voltage to
where you can see the wires emit a faint red glow in the dark and they
still live on for years. If your 5V is reasonably stable the risks seem
to be a grossly wrong PWM ratio in the uC case or maybe a blown
capacitor. Doesn't sound too unusual or unreasonable to me. I mean,
everything is going to go kaputt some day, nothing lives forever. Even
you and I won't.

 

[Joerg]

You could drive the filaments pulsed, from +12 to ground, at a low
duty cycle. Pulse the filaments when the display segments are off,
which eliminates the filament voltage interaction. The filament driver
can be a p-fet, with capacitive coupling from a cmos (port pin?) level
up to the gate. That's fail-safe.

Until that P-channel fails and shorts out. 12V is a bit much for most
VFDs, I'd drive it from the 5V rail if available.

 

[Joerg]

Or the power supply catches fire. Or until a supernova goes off and
melts the Earth. Hey, life is risky.

Where you guys are I'd be more worried about this clock falling off the
shelf because the shelf fell over, because the wall behind the shelf
fell onto it after a major rumbler ...

I wonder what those Russian VF tubes are like.

Right now there is an amazing flurry of NOS tube stuff showing up. Like
the PL81, a flyback sweep tube from the early 50's. A guy on a German
forum had an interesting problem where he needs to swing over 4kV in
nanoseconds and I was musing aloud that, yeah, if this old PL81 was
still around it would be sweet. Then someone posted a link where you can
buy those off the shelf for a couple of bucks. Blew me away. It's
essentially like a 6KD6 but much skinnier and can do 7kV.

 

[Joerg]

I'm surprised there's not more mixed designs... tubes with
bipolars/MOS/OpAmps under the cathodes... making for an ideal HV
device.

They aren't that ideal. Even when flooring it (G2 at abs max and then
some, G1 in the white-knuckle range above 0V) you can barely make them
do a couple amps. To really shut the off good enough needs a swing of
easily negative 100V G1-cathode.

Probably the filament power?

That, and also the fact that new tubes are only made in Russia. The
longer term supply of HV-capable tubes such as the GP-5 is not really
guaranteed, only with audio stuff will there be an unshakeable group of
customers who'd never ever touch newfangled trainsistahs and to whom
class-D is the devil's work.

Speaking of green idiocy... has anyone seen "Cool It" yet?

Not yet. Doubt it'll be a box office hit though. The whole climate
debate on a "neighborhood pub level" seems to have ended with
Climategate, at least out here. That was sort of the last straw in many
people's minds.

 

[Joerg]

[snip]
Right now there is an amazing flurry of NOS tube stuff showing up. Like
the PL81, a flyback sweep tube from the early 50's. A guy on a German
forum had an interesting problem where he needs to swing over 4kV in
nanoseconds and I was musing aloud that, yeah, if this old PL81 was
still around it would be sweet. Then someone posted a link where you can
buy those off the shelf for a couple of bucks. Blew me away. It's
essentially like a 6KD6 but much skinnier and can do 7kV.
I'm surprised there's not more mixed designs... tubes with
bipolars/MOS/OpAmps under the cathodes... making for an ideal HV
device.

They aren't that ideal. Even when flooring it (G2 at abs max and then
some, G1 in the white-knuckle range above 0V) you can barely make them
do a couple amps. To really shut the off good enough needs a swing of
easily negative 100V G1-cathode.

I easily trade 100V for 7KV capability

That's not quite how it works, no free lunches there

You can easily control 1kV with 10V gate swing. But not with a tube.
Pretty soon you can control up to 4kV with a new IXYS device but that
isn't in the showrooms yet AFAIK.

Problem is, most laser guys just use a beanstalk final and let that
avalanche. Naturally, this jitters like crazy because the effect has
about the same "order" as a real avalanche on a mountain slope.

It's supposed to be a rational look at the economics of global
"farts".

I think there's also a book. May be interesting.

 

[Mr.CRC]

Make sure that capacitor can handle the AC current over the long haul.

I'll have 0.6A RMS current with either of two tube configurations, one
with 1.5VAC, the other at 2.4VAC. My PCB has two locations for parallel
caps. Pana FM series, either a 150uF 35V or 220uF 25V can handle 0.95A
with 0.056ohms ESR. These should suffice.

Operating the thing with some tubes pulled would fall under the category
"user stupidity", or maybe even "daft". That's like roaring off in a car
with numerous lug nuts missing.

Alternatively, "within the reach of children." I certainly used to
abuse the TV set and other household electronics when I was a kid in
this manner ;-)

Cost? Huh? A few resistors and caps, that's all. If you want it simple
you could consider hysteretic operation which even a uC should be able
to do.

What is "hysteretic operation ?"

You have to get the datasheet for those tubes and check the filament
voltage tolerance. Usually quite high. I don't know about Russian tubes
but figure 10 in this Noritake document shows that the luminance remains
relatively flat from 60% to 100% filament voltage:

The IV-12 (7-seg 21mm high digit with DP) seems to allow a range of
90-110mA fil. current, at a nominal 1.5V . The IV-17 (16-segment 16mm
character) has poorer translations, with roughly 50mA current. We can
assume a similar +-10% tolerance.

http://www.noritake-elec.com/vfd_operation.html

Often people "revive" tired VFDs by increasing the filament voltage to
where you can see the wires emit a faint red glow in the dark and they
still live on for years. If your 5V is reasonably stable the risks seem
to be a grossly wrong PWM ratio in the uC case or maybe a blown
capacitor. Doesn't sound too unusual or unreasonable to me. I mean,
everything is going to go kaputt some day, nothing lives forever. Even
you and I won't.

With the 2.4V tubes, it will actually be barely possible to reach this
with 5V supply and accounting for losses. So no way to burn these out.

The 1.5V tubes are possible to fry with a 5V supply.


I agree it's not much worry.

 

[Joerg]

Mr.CRC wrote:

[...]

Cost? Huh? A few resistors and caps, that's all. If you want it simple
you could consider hysteretic operation which even a uC should be able
to do.

What is "hysteretic operation ?"

It trips off the conduction phase when the voltage has reached target
but won't come back on until it has dropped a few ten millivolts below
target:

http://electronicdesign.com/article/power/switch-mode-ics-promote-efficient-power-management.aspx
http://archive.electronicdesign.com/cluster/planetee/files/29/20245/1217sdpart1fig6_web.gif

The comparator has a hysteresis and thus isn't as "nervous" as a regular
comparator. This does cause ripple on the output but for a filament
drive that shouldn't matter and those converters are simple, IME very
stable.

Think of it as the accelerator pedal of a Ferrari Testarossa on ice
versus that of a John Deere tractor in a muddy field

The IV-12 (7-seg 21mm high digit with DP) seems to allow a range of
90-110mA fil. current, at a nominal 1.5V . The IV-17 (16-segment 16mm
character) has poorer translations, with roughly 50mA current. We can
assume a similar +-10% tolerance.

For a VFD that would be a very narrow tolerance.

With the 2.4V tubes, it will actually be barely possible to reach this
with 5V supply and accounting for losses. So no way to burn these out.

The 1.5V tubes are possible to fry with a 5V supply.


I agree it's not much worry.

You might want to place two 1.5V tubes each in series.

 

[Mr.CRC]

You could drive the filaments pulsed, from +12 to ground, at a low
duty cycle. Pulse the filaments when the display segments are off,
which eliminates the filament voltage interaction. The filament driver
can be a p-fet, with capacitive coupling from a cmos (port pin?) level
up to the gate. That's fail-safe.

John

Presumably with the other end of the filament biased at +3 to +5V with a
zener diode so that anode and grid current keeps the cathode positive
w/respect to the grid, right?

I would have to do some experiments to see what the time constant is for
the decay of temperature. Those wires are awfully thin, so I have my
doubts.

Oh, come to think of it I am planning to have the 7-seg. tubes on direct
drive, for maximum brightness, rather than multiplexed. So this doesn't
really work with continuously on tubes.

The 16-seg. tubes OTOH, will be a set of 12 multiplexed as six pairs.
This might work there. But I am trying to create a universal driver
board with filament drive, boost voltage for anodes+grids, and up to 60
bits of serial-to-parallel HV output drivers. So whatever scheme I use
has to work in both situations.

I'll probably stick with the 1/2 bridge, since it's all laid out already.


Thanks for the input.

 

[Mr.CRC]

Until that P-channel fails and shorts out. 12V is a bit much for most
VFDs, I'd drive it from the 5V rail if available.

Hmm. This is an interesting idea, since it is way simpler than my
bridge/2, and thus cheaper. But it does create a new risk.

 

[Mr.CRC]

Right now there is an amazing flurry of NOS tube stuff showing up. Like
the PL81, a flyback sweep tube from the early 50's. A guy on a German
forum had an interesting problem where he needs to swing over 4kV in
nanoseconds and I was musing aloud that, yeah, if this old PL81 was
still around it would be sweet. Then someone posted a link where you can
buy those off the shelf for a couple of bucks. Blew me away. It's
essentially like a 6KD6 but much skinnier and can do 7kV.

Here's some examples of the VFDs:

http://www.tubeclockdb.com/vfd-tubes/1-iv-12-b-12.html

http://www.tubeclockdb.com/nixie-kits/59-iv-17-smartsockets.html


I have a bunch of Russian 40mm digit Nixies, as well as German 30mm, and
US and German 15mm. The smaller Russian tubes are not to my liking, do
to crude construction or silly upside down '2' characters for the '5' digit.

The US (National NL840) and Telefunken ZM1210 15mm tubes have very nice
aesthetics, so I have bought up much of the available surplus supply.
Once the Russian tubes run down in 10-15 years, I will retire and bring
out the nicer looking western made ones, which at that point should be
extremely rare.

 

[Mr.CRC]

Or MOVs. Anything to clamp transients that caps like a whole lot less
than transformers.

There are lots of LED nightlight and other low power products that use
cap dividers of various topologies right off the line.

Problem: a clock is a rather complex device, with lots of other voltage
sensitive goodies in there. If the VFD supply jumps one way due to a
power line spike but the uP stays put on another supply, what gives up
first?

I'm not sure there's much to worry about with regard to differentials
between the VFDs and the other electronics. There will probably be a
regulated 12VDC power supply from the line, a wall-wart with UL rating
so the safety issue is outsourced.

5V for logic would be derived from that. The VFD filament driver can be
set up to work off the 5V or the 12V. The worst thing that can happen
is the 12V supply goes bonkers. So perhaps something to protect against
that is sufficient. Some MOV or TVS, and a fuse.

 

[Joerg]

Here's some examples of the VFDs:

http://www.tubeclockdb.com/vfd-tubes/1-iv-12-b-12.html

http://www.tubeclockdb.com/nixie-kits/59-iv-17-smartsockets.html

Nice. Looks like the filaments aren't driven to the point where they
glow so there should be quite some reserves.

I have a bunch of Russian 40mm digit Nixies, as well as German 30mm, and
US and German 15mm. The smaller Russian tubes are not to my liking, do
to crude construction or silly upside down '2' characters for the '5' digit.

The US (National NL840) and Telefunken ZM1210 15mm tubes have very nice
aesthetics, so I have bought up much of the available surplus supply.
Once the Russian tubes run down in 10-15 years, I will retire and bring
out the nicer looking western made ones, which at that point should be
extremely rare.

Let's hope they'll keep their vacuum

 

[Joerg]

Hmm. This is an interesting idea, since it is way simpler than my
bridge/2, and thus cheaper. But it does create a new risk.

Well, if the tubes are super expensive and need to be babied a lot, why
not place a fuse in series with the filament? Or a Polyfuse, but make
sure that can't cause anything to catch fire.

 

[Joerg]

I'm not sure there's much to worry about with regard to differentials
between the VFDs and the other electronics. There will probably be a
regulated 12VDC power supply from the line, a wall-wart with UL rating
so the safety issue is outsourced.

5V for logic would be derived from that. The VFD filament driver can be
set up to work off the 5V or the 12V. The worst thing that can happen
is the 12V supply goes bonkers. So perhaps something to protect against
that is sufficient. Some MOV or TVS, and a fuse.

Since I assume there's some electronics hanging off of the 12V that
might go *POOF* before the tubes do I'd consider a crowbar. MOV and TVS
have a large tolerance. If you build it using a TL431 plus SCR it'll be
very precise and still cheap. In the end it all comes down to the angst
factor and how expensive the stuff is that could blow, and whether it
can cause consequential damage.

 

[Mr.CRC]

Let's hope they'll keep their vacuum

The NL840s have presumably all been tested. I double check them,
testing so far about 10-12% of the bunch. I've yet to find any bad
ones. Since they were probably made <=1970, there's a good shot they'll
last for another few decades.

At least the Nixies don't require a hard vacuum. There are dekatrons
that still seem to work (though maybe not at spec counting speed) for
visual purposes even though they have oxidized getters.

 

[Mr.CRC]

Well, if the tubes are super expensive and need to be babied a lot, why
not place a fuse in series with the filament? Or a Polyfuse, but make
sure that can't cause anything to catch fire.

Fortunately, the Russian VFDs are actually quite inexpensive. About
$1.50 per tube.

 

[Joerg]

Fortunately, the Russian VFDs are actually quite inexpensive. About
$1.50 per tube.

That is a good deal. Although, for a Russian this is a lot of money.

 

[Warren]

Joerg expounded in

Where you guys are I'd be more worried about this clock falling off
the shelf because the shelf fell over, because the wall behind the
shelf fell onto it after a major rumbler ...

Yes- don't forget cats. Our one cat likes to knock things off the dresser
until someone finally gets out of bed to feed him and give attention.

Scolding doesn't work. As soon as you climb back into bed he's back up
there again. We almost had the clock radio go sliding off one morning..
it's just a matter of time.

Warren