On Tue, 24 May 2005, der Mouse wrote: As logic elements that's the only thing gas tubes have going for them. A neon bulb will remain infinite resistance (off) until the striking voltage. Once struck, it's a quasi-constant voltage drop depending on the gas and other details. For neon, from flawed memory, strike-over is around 100V, and sustain around 80V. Look in a tube manual for miniatures with part numbers beginning with 0 (zero) for various gas mixtures used as voltage regulators (picture a zener in series with a resistor). 0A2, 0B2, etc. The glow is caused by quantum level changes in the electrons tricked into moving by the high voltage (gives off photons as they drop back down). The glow appears on the negative element where all the donor electrons are leaving the building. You could probably make a coincident-voltage memory from them... but they'd be slow I think, but pretty. Gas tubes are also light sensitive, somewhat. A lot of neon-type lamps contain a tiny bit of radioactive Krypton gas as a source of electrons (what you call beta particles when they are moving slowly) for more predictable strikeover. Like all quantum devices, it's a statistical thing. All sorts of gaseous logical/counting elements were available commercially for about 15 minutes in the late 1940's/early 1950's.

Yes-- and they were pig-slow. (I used PLATO in the late '70s, and even then it felt slow to me.) /~\ The ASCII der Mouse \ / Ribbon Campaign X Against HTML mouse at rodents.montreal.qc.ca / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

der> Yes-- and they were pig-slow. (I used PLATO in the late '70s, der> and even then it felt slow to me.) Don't confuse the capabilities of the display with that of the comm system. PLATO terminals used 1200 baud comm lines, though with efficient encoding they would do 180 characters per second (not 120 as with ASCII) and 60 vectors per second. The panels could do far more. The second generation PLATO terminal was an 8080 based device (the first generation was hardwired) and supported local execution of downloaded programs. Those would go dramatically faster. Even that show CPU would display text at 3000 characters per second, and that was limited by the 8 bit nature of the CPU (it would do block erase twice as fast, since there the CPU interface allowed 16 bits to be touched at a time -- and the panel was probably substantially faster than that). paul

On Wed, 25 May 2005, Eric Smith wrote: Actually, it might be easy enough to be fun. There aren't many storage method that are direct-readout of all cells! (It would make a display too). I think your idea is generally right, but this detail seems wrong: when X goes from +7.5V to 0V, the entire line will see 105V, striking the entire line. I think the approach is right though, but I think entangling address and data is needed: I'll call Ne tube turn-on strike; turn-off break. Lessee: Strike >= 100V. Break <= 80V. Select steady-state cell row (X) and column (Y) voltage to be 90V. Off cells remain off; on cells remain on (series R's per tube allow per-cell variation). Let's say you arbitrarily run X= 90V, Y=0V, for non-addressed cells. WRITING A 1: * Raise line X to 95V. * Drop line Y to -5V. - Line X cells see 95V; line Y cells see 85V; no state-change. - Cell X:Y sees 100V, strikes. WRITING A 0: * Drop line X to 85V. * Drop line Y to -5V. - Line X cells see 85V; line Y cells see 95V; no state-change. - Cell X:Y sees 80V, breaks. Now we know that these things are awful, and the strike/break voltages probably vary all over the place, and margins will be hard to determine and maintain, but likely with 1% or better regulation on each V and/or temperature compensation or whatever it'll work. Reading is problematic; the problem is that the X and Y lines can't be used for readout, as you have to hold them at constant V to maintain cell contents. But we know that a conducting cell maintains a constant voltage across it, so if you yank it's line to voltage N, the other end must be at voltage N+80V. You mentioned measuring current to sense cell contents; this implies destructve readout, but that's been done before. I now have to do work I'm paid for.... :-) but this is pretty interesting! As far as I can tell from reading literature, basically the metal electrodes develop non-conducting surfaces (eg. oxidation) from impurities in and outgassing from the glass and metal. Heat accellerates it (typical chemistry). Too much current makes them hot. 50 year old Nixies routinely work though so I dont see any short-term problems with them. You could easily measure strike time and off time; I never have. I suspect off is slower, you have to wait for the electron cloud to dissipate, and quantum processes distribute out in time quite a ways.

On Wed, 25 May 2005, Eric Smith wrote: I stand corrected! Ahh... the Nixie lit explicitly talks about element life. If one element is held lit for long term you get a limited life; if it varies as much as once a day life dramatially increases. I think for memory purposes we'd be safe... I think the numbers were "short" life 10K hours, "long" 100K hours. I bet it's also dependent on current. app note here: http://wps.com/archives/Burroughs/index.html I think paper N102.

Tom Jennings wrote: That is exactly the storage mechanism of the orange plasma display panels used by the Plato terminals. I'm sure the individual cells of a panel tracked each other quite well as compared to an assortment of a few hundred thousand neon lamps.

Tom Jennings wrote: I replied: [matrix drive ideas snipped] I've just looked at the specifications for Chicago Miniature Lamp neon bulbs, and it looks like they don't have anything suitable. The problem is that the specifications for the "maintain" and "breakdown" (start) voltage ranges are too wide, and often there is overlap. So unless you are willing to cherry-pick bulbs that meet tighter specs, 2-D matrix drive doesn't seem practical. For instance, the 4AD and 4AE bulbs do have a nice wide separation between the maximum maintain voltage and minimum breakdown voltage, but the ranges are too wide: rated maintain breakdown current voltage voltage ------- -------- --------- 4AD 0.50 mA 60-80V 105-135V 4AE 0.50 mA 60-80V 120-150V For symmetric drive you need to deal with three voltages (in opposite polarities for X and Y): A - below minimum maintain when applied to both X and Y, within maintain range when applied to just one B - nominal maintain when applied to both X and Y C - above maximum breakdown when applied to both X and Y, within maintain range when applied to just one This results in five inequalies to be satisfied. FOr the 4AD: 2*A < 60 - guarantee turn-off 2*B >= 80 - guarantee maintain 2*C >= 150 - guarantee turn-on A+B > 80 - don't turn off non-selected bulbs B+C < 120 - don't turn on non-selected bulbs Unfortuantely there is no set of numbers for A, B, and C that satisfy all five inequalities. For the asymmetric case, you have six variables, since there are different voltages used for each of three states on X and Y. A+D < 60 B+E >= 80 C+F >= 150 A+E > 80 B+D > 80 B+F < 120 C+E < 120 I haven't tried to find a solution to this. Is there any readily available software (preferrably free) that solves sets of inequalities like this? I also can't find a distributor that stocks the 4AD or 4AE bulbs. The C2A is avaialble for $0.186 each in quantity 5000 from Newark, but the characteristics are even more unsuitable. Eric

Eric wrote: This doesn't rule out the use of commodity neon lamps. It's always true that V_breakdown > V_maintain. The specs are just written loosely. I've built several neon lamp gadgets over the years (I think you saw my 5-neon-in-a-sequence-flasher in fact) and if you start with a bunch of bulbs from the same batch you'll end up with matching way better than you'd expect from just the spec. Whether or not they're good enough for a given logic circuit may drive you to hand-match etc. The same thing is done with semiconductors, you know :-). Tim.

On 26 May, 2005, at 00:34, Eric Smith wrote: Automatic testing and sorting should not be too difficult. As long as each row had about the same characteristics, the driver circuitry could compensate for differences. Many early portable computers used orange plasma displays which were actually such memories. They did not need any refresh, so the only thing missing was the readout circuitry. -- -bv