rickb at bensene.com
Tue Mar 6 11:32:57 CST 2007
> It is not a good idea to multiplex Nixies. They don't always
> strike exactly when the voltage is applied. This means you
> can't go too fast. Also, if the current is too high ( not
> just the average ), it causes the gas to heat up too much and
> it changes the ionization.
> This causes a more distructive violet glow that will
> eventually cause the bulb to get contaminated by blasting the
> Unlike LED's, running them at too low a current doesn't just
> make them glow less brightly, it tends to make them glow only
> over part of the electrode.
> Most applications use one driver per Nixie.
Multiplexing Nixies was extremely common in all but the earliest
electronic calculators. The early machines
(Sharp Compet 10, 15, 20, 21, 30[and Facit/Addo equivalents]; early
Sumlock/ANITA machines; the Wang LOCI 1/2; and a few others) used
discrete drivers per Nixie tube, but the cost of such circuitry (11
transistors per tube [digits plus decimal point], along with diode
arrays to do BCD to decimal conversion for each tube) made the machines
quite expensive with very high component counts.
Multiplexing had two major advantages. The first was that the driver
circuitry was MUCH less expensive. One BCD to decimal converter, 11
transistors for driving the anodes (digit 0-9 & decimal point), and as
many transistors as there are tubes for switching the cathodes. Along
with the transistors, there was usually a series of RC networks, or
sometime small-signal transformers (Canon liked transformers) involved
in the cathode switching. The other major benefit of multiplexing was
that the architecture of most electronic calculators is bit-serial.
This was done because the speed of parallel math is not necessary on a
simple calculator, and that the ALU circuitry was much less complex.
Also, most of the machines had a memory that was bit serial, such as
shift registers or magnetostrictive delay lines (and in at least two
known cases, magnetic drums). Those that had magnetic core memory, or
stored charge capacitive memory were generally addressed in a bit-serial
fashion. The bit serial nature of the logic made multiplexing a whole
lot easier, because the bits are already circulating around in the
machine, and this architecture fit very will with multiplexing.
Typically, there would be a four-bit shift register that would
accumulate a BCD digit, and this register would be fed to the
BCD-to-Decimal converter, which would feed the anode drivers. A
separate shift register or counter would step through the number of
digits there are in the display, which would drive the cathode drivers.
Everything was synchronized by phases of the master clock circuitry.
The result was a much simpler design in terms of parts count (which
translated directly to lower cost and higher profits), and also smaller,
lighter, and more reliable machines.
In terms of failure rates, I have yet (after many years of collecting)
to find a multiplexed Nixie calculator in which a Nixie tube has failed.
It's relatively common to find failed anode or cathode driver transistor
failures, which result in either a single Nixie that refuses to light,
or a single numeric digit (or decimal point) won't display on any of the
tubes. In the case of a cathode driver failing, many might think the
tube is bad. In every single case I've had, replacing the driver
transistor(s) brings the tube back to life.
Remember, calculators were typically turned on in the morning, ran all
day long, and shut off when people went home for the day. So, the
number of runtime hours on the tubes by the time I've gotten my hands on
a machine is very long.
I do believe that Nixies made for direct drive have some different
characteristics than those made for multiplexed drive. I do not know
specifically, but recall reading somewhere that the gas mixture in
Nixies made for multiplexing is somewhat different, to allow for faster
switch-on/off times. I seem to recall that there was a trace amount of
mercury added to the Neon gas mixture. I also do know that the
switching times for mux'd Nixies were rather strict. Violate the duty
cycle and rise/fall times and current levels, and the tubes would indeed
suffer a reduced lifetime. Calculator designers took all of this into
account, and designed the drive circuitry very carefully to assure
maximum life for the Nixie tubes.
The common problem with old Nixie tubes (multiplexed or not) is long
warm-up time. Some of the old machines in the museum take about 30
seconds to a minute before all of the tubes light up. This is generally
caused by slight contamination of the gas in the tube by the anodes
slowly eroding, or by molecular-level leakage through the glass envelope
over very long periods of time, reducing the gas concentration in the
tube. But, once the tube "warms up", it runs fine. This is part of the
reason why I power up each calculator in the museum at least once a
month and run it for at least a half hour, partly to exercise the tubes,
and also to keep the power supply filter capacitors from malforming.
I generally do a few machines per day, and over the course of a month,
I've pretty much run through all of them.
I put a priority on the Nixie and CRT display machines, as they are the
most likely candidates for problems, then the VF tube calculators,
followed by LED's, then the calculators (like the early Canon 161 and
130S) with incandescent displays, and lastly, the early LCD calculators.
So, in summary, Nixies multiplex, and do so just fine. However, some
Nixies aren't designed for it, and multiplexing these tubes may shorten
their liftimes. But, the tubes that are designed for multiplexing can
have extraordinarily long lifetimes (some of the Nixies tubes in
machines in the museum are nearing 40 years old, and still work
The Old Calculator Web Museum
More information about the cctech