Loading ...
Sorry, an error occurred while loading the content.
 

Re: Designing a generic switchmode/flyback AVR controlled 180V power supply

Expand Messages
  • fnatmed
    ... jt - This is to drive a Sudoku game in Nixies, 81 IN-17 s to be exact. best candidate right now is a pair of 6000mAH LiIon batteries for the whole thing,
    Message 1 of 19 , Feb 14, 2007
      --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@...> wrote:
      >
      > >Since the NixieDoku will be battery operated, efficiency is a
      > >concern, and I hear that flyback is more efficient.
      >
      > Just how much power are you looking to deliver?
      >
      > The main drawback with the flyback conversion mode (Any converter
      > where the energy storage is in the voltage transformation element) is
      > that the entirety of the delivered power must be stored during the
      > primary switch on-time which tends to drive the I/R losses up when a
      > boost is performed. A buck flyback, such as used in a PC power
      > supply, starts with a high voltage so that the primary currents are
      > very low compared to the secondary currents.
      >
      > Another drawback of the flyback is the leakage inductance, which is
      > essentially lines of magnetic flux which do not contribute to the
      > power delivered to the load. This must either be dissipated as with a
      > snubber or returned to the primary supply via a reactive energy
      > transfer mechanism.

      jt -

      This is to drive a Sudoku game in Nixies, 81 IN-17's to be exact.
      best candidate right now is a pair of 6000mAH LiIon batteries for the
      whole thing, plus a dedicated charger from TI or Linear. It would
      also power the main micro controlling the whole thing.

      81 Nixies is a lot. So I'm thinking to break it down to 3 groups of
      27, multiplexed. At 2mA per tube, that's 54mA needed at 180V. Since
      we like to have headroom, I'm spec'ing the boost portion to be able to
      supply 80mA at 180V.

      I'm basing a lot on what I've read in Ronald Dekker's "Flyback for
      Dummies" treatise :)
      http://members.chello.nl/r.dekker49/flyback/flyback.html

      Given that we're battery-powered, I would prefer not to dissipate
      unecessarily. Can you describe a reactive energy transfer mechanism
      that would slot into a simple flyback topolgy ?

      Dean.
    • fnatmed
      ... Dave - Thanks for the pointer. I m messing with WeBench on the National site now. Looks like there s no way to use a standard transformer - it has to be
      Message 2 of 19 , Feb 15, 2007
        --- In NEONIXIE-L@yahoogroups.com, David Forbes <dforbes@...> wrote:

        > I have had good success with the LM2585 from National for use with a
        > flyback step-up transformer. It's easy to wind a transformer for it,
        > and the efficiency is reasonably high. The chip is a breeze to use -
        > a half-dozen parts and you're there. I recommend a 12V supply for
        > reasonable power output.

        Dave -

        Thanks for the pointer. I'm messing with WeBench on the National site
        now. Looks like there's no way to use a standard transformer - it has
        to be a custom one, which jibes with what you said about winding your
        own. Something I'm assuming you've done before ... I guess I need to
        find a source for the bobbins, cores etc. Any hints ?

        An example run using 12-16V Vin, 180V Vout and 0.08A Iout gives this a
        BOM :

        Ccomp Vishay-Vitramon VJ1206Y563KXAAT 0.056000 uF
        Cin Nichicon UPL1E561MPH NumCaps=1 560.0000 uF,0.0550 Ohms
        Cout1 Custom Custom NumCaps=1
        1.58 uF, 4.83 Ohm
        D1 Motorola MURS160T3 1.250000 V
        Dz Motorola SA60A 60.000000V
        IC National Semiconductor LM2587S-ADJ ADJ,Flyback
        Rcomp Vishay-Dale CRCW1206-2941FRT1 2940 Ohms
        Rfb1 Vishay-Dale CRCW1206-1211FRT1 1210 Ohms
        Rfb2 Vishay-Dale CRCW1206-1743FRT1 174000 Ohms
        Transformer Custom Custom 571.43 n, 28.57 u, 8.85 , 68.04 m, 10.32 , 1_P

        Hrm - I also note that the Cout cap is a custom : 1.58uF. Like I've
        got a few of those in my junkbox harhar. How critical are the values
        here ?

        Dean.
      • David Forbes
        ... It can be difficult to track down the necessary parts in prototype quantities. I was lucky to have Magnetics Inc. send a few samples with a promise to buy
        Message 3 of 19 , Feb 15, 2007
          At 5:23 PM +0000 2/15/07, fnatmed wrote:
          >Dave -
          >
          >Thanks for the pointer. I'm messing with WeBench on the National site
          >now. Looks like there's no way to use a standard transformer - it has
          >to be a custom one, which jibes with what you said about winding your
          >own. Something I'm assuming you've done before ... I guess I need to
          >find a source for the bobbins, cores etc. Any hints ?

          It can be difficult to track down the necessary parts in prototype
          quantities. I was lucky to have Magnetics Inc. send a few samples
          with a promise to buy hundreds of parts later on. Little did they
          know that my transformer winding company used a different brand of
          parts!

          I have a good supply of various cores and bobbins from a big box of
          parts purchased at a local estate sale. I also got a Morris Coil
          Winder there. Very lucky find.

          I could wind a transformer for you to play with. The flyback topology
          is not critical with regard to the number of turns. Some
          experimentation will reveal the comfort zone of any given
          transformer, so a transformer to meet a different set of operating
          conditions can be extrapolated from the test unit.

          I would start with something I have lots of, like a Ferroxcube P2616
          pot core (26mm diameter, 16mm tall. The typical 100 kHz switching
          frequency works with the vintage 3C8 core material from Ferroxcube.
          Here's a recent catalog:
          http://www.elnamagnetics.com/library/catalogs/ferroxcube/HB2005.pdf

          My first guess for windings would be:
          Primary: 12 turns bifilar (parallel) wound, #22 gauge.
          Secondary: 180 turns #30 gauge.

          >An example run using 12-16V Vin, 180V Vout and 0.08A Iout gives this a
          >BOM :
          >
          >Ccomp Vishay-Vitramon VJ1206Y563KXAAT 0.056000 uF
          >Cin Nichicon UPL1E561MPH NumCaps=1 560.0000 uF,0.0550 Ohms
          >Cout1 Custom Custom NumCaps=1
          >1.58 uF, 4.83 Ohm
          >D1 Motorola MURS160T3 1.250000 V
          >Dz Motorola SA60A 60.000000V
          >IC National Semiconductor LM2587S-ADJ ADJ,Flyback
          >Rcomp Vishay-Dale CRCW1206-2941FRT1 2940 Ohms
          >Rfb1 Vishay-Dale CRCW1206-1211FRT1 1210 Ohms
          >Rfb2 Vishay-Dale CRCW1206-1743FRT1 174000 Ohms
          >Transformer Custom Custom 571.43 n, 28.57 u, 8.85 , 68.04 m, 10.32 , 1_P
          >
          >Hrm - I also note that the Cout cap is a custom : 1.58uF. Like I've
          >got a few of those in my junkbox harhar. How critical are the values
          >here ?
          >
          >Dean.

          The parts are mostly not critical. Of course, the feedback resistors
          determine the output voltage, so select them according to your needs
          for accuracy. The filter capacitors are OK within a factor of two.
          The compensation network is not too critical either, as far as my
          experience has shown. It will affect the stability under changing
          load. 20 percent variation in parts should not affect its behavior.

          You will want to use something like a UF1005 diode for the HV
          rectifier. It's ultra-fast to reduce switching loss. I have a few
          lying around.


          --

          --David Forbes, Tucson, AZ
          http://www.cathodecorner.com/
        • fnatmed
          ... Lucky :) ... There s one on ebay right now ... ... That would be awesome. Whatever you think would do the trick for pulling 80mA at 180V from 14.4V (from
          Message 4 of 19 , Feb 16, 2007
            --- In NEONIXIE-L@yahoogroups.com, David Forbes <dforbes@...> wrote:
            > It can be difficult to track down the necessary parts in prototype
            > quantities. I was lucky to have Magnetics Inc. send a few samples
            > with a promise to buy hundreds of parts later on. Little did they
            > know that my transformer winding company used a different brand of
            > parts!

            Lucky :)

            > I have a good supply of various cores and bobbins from a big box of
            > parts purchased at a local estate sale. I also got a Morris Coil
            > Winder there. Very lucky find.

            There's one on ebay right now ...

            > I could wind a transformer for you to play with. The flyback topology
            > is not critical with regard to the number of turns. Some
            > experimentation will reveal the comfort zone of any given
            > transformer, so a transformer to meet a different set of operating
            > conditions can be extrapolated from the test unit.
            >
            > I would start with something I have lots of, like a Ferroxcube P2616
            > pot core (26mm diameter, 16mm tall. The typical 100 kHz switching
            > frequency works with the vintage 3C8 core material from Ferroxcube.
            > Here's a recent catalog:
            > http://www.elnamagnetics.com/library/catalogs/ferroxcube/HB2005.pdf
            >
            > My first guess for windings would be:
            > Primary: 12 turns bifilar (parallel) wound, #22 gauge.
            > Secondary: 180 turns #30 gauge.

            That would be awesome. Whatever you think would do the trick for
            pulling 80mA at 180V from 14.4V (from LiIon 6000mAH batteries). I'll
            play with it, compare it to doing a basic inductor-based design. I'll
            email you my details.

            Plons from avrfreaks.net had a very slick method of doing a safe
            design ...

            http://www.aplomb.nl/TechStuff/NixNies/Switcher.html

            Basically he's using a 555 timer to send the pulses to the FET. It's
            a monostable timer that has to be triggered, and then reset to kill
            the pulse. Without the reset, it times out anyway. This way even if
            the driving AVR or switcher IC dies a horrible death, the timer will
            not trigger more than once, and will self-time-out.

            > The parts are mostly not critical. Of course, the feedback resistors
            > determine the output voltage, so select them according to your needs
            > for accuracy. The filter capacitors are OK within a factor of two.
            > The compensation network is not too critical either, as far as my
            > experience has shown. It will affect the stability under changing
            > load. 20 percent variation in parts should not affect its behavior.

            Good - makes part selection much simpler ... Will be easy enough to
            trim the feedback resistor network with a pot.

            > You will want to use something like a UF1005 diode for the HV
            > rectifier. It's ultra-fast to reduce switching loss. I have a few
            > lying around.

            I have a bunch of Mur160 diodes - pretty fast.

            Dean.
          • taylorjpt
            I was thinking of using IR transmitters and receivers in a 9x9 grid over the game as the user input for my Sudoku so there would just be nixies and nothing
            Message 5 of 19 , Feb 28, 2007
              I was thinking of using IR transmitters and receivers in a 9x9 grid
              over the game as the user input for my Sudoku so there would just be
              nixies and nothing else: Poke a tube and it cycles through 0-9 and
              then off. Maybe a switch on the master computer so a wrong value will
              blink in protest (Total beginners mode!) Has anyone done anything with
              IR in this manner? It would have to reach across the width of the
              game but that should be doable.

              I have a VB6 solver that works well which would be easy to port. Are
              you planning to have the game know the answer or just solve it in the
              background from the initial numbers in known puzzles?

              jt
            • taylorjpt
              ... A forward converter has an output reactor to average the square wave input to a DC value but the Iron needed to do that is far less than a flyback s
              Message 6 of 19 , Feb 28, 2007
                > Given that we're battery-powered, I would prefer not to dissipate
                > unecessarily. Can you describe a reactive energy transfer mechanism
                > that would slot into a simple flyback topolgy ?
                >
                > Dean.
                >

                A forward converter has an output reactor to average the square wave
                input to a DC value but the "Iron" needed to do that is far less than
                a flyback's inductor which has to store the entire cycles energy load
                in the switch on time. The transformer in a push-pull forward stores
                no energy so it's power handling capability is only limited by the
                current density of the windings.

                I've had really good results with the one I designed for my brothers
                clock and am getting better than 80% at 14 Watts out in under a half
                cubic inch: If you want to try it I'll send you one.
                http://www.tayloredge.com/storefront/1329_NixiePS/1329_graph1.jpg

                All the design calculations were done with this calculator program
                (freeware):
                http://www.tayloredge.com/utilities/vbapps/iCalculators.e_e

                jt
              • fnatmed
                ... That s an interesting approach. How about using 18 transmitters and receivers around the edge, just above the tubes, one set of 9 horizontal, one set
                Message 7 of 19 , Mar 2, 2007
                  --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@...> wrote:
                  >
                  > I was thinking of using IR transmitters and receivers in a 9x9 grid
                  > over the game as the user input for my Sudoku so there would just be
                  > nixies and nothing else: Poke a tube and it cycles through 0-9 and
                  > then off. Maybe a switch on the master computer so a wrong value will
                  > blink in protest (Total beginners mode!) Has anyone done anything with
                  > IR in this manner? It would have to reach across the width of the
                  > game but that should be doable.
                  >
                  > I have a VB6 solver that works well which would be easy to port. Are
                  > you planning to have the game know the answer or just solve it in the
                  > background from the initial numbers in known puzzles?

                  That's an interesting approach. How about using 18 transmitters and
                  receivers around the edge, just above the tubes, one set of 9
                  horizontal, one set vertical, intersections right over the tubes. As
                  your finger comes down over a tube, it blocks a unique pair.

                  Or if I understand you correctly, you'll have 81 pairs in small black
                  tubes pointing up. As your finger covers the tube, it will reflect IR
                  back to the receiver, so you know which Nixie is selected. It would
                  be nice if you could have the xmit/recv set under the Nixie, so you
                  can poke the Nixie as you say. I wonder if the neon glow gives off IR
                  - probably does, which would mess up your sensing.

                  If you have a chance, shoot me that VB solver. I have one in python,
                  and another one in C, both work well. You're welcome to them if you want.

                  Oddly enough, I never even considered just having completed puzzles :)
                  It was always going to be normal puzzles, and the system would do the
                  solving to compare/help the player. I think a lot of the art of
                  Sudoku is in the revealed numbers, so that would have to be
                  incorporated into the puzzle database.

                  I'm going to have two modes (so far). When you select an empty spot
                  and spin through the numbers, mode 1 would show every number
                  regardless, letting you put in a patently wrong entry. Mode 2 is
                  semi-guided, only displaying potentially valid numbers. For example,
                  if there's already a 5 in the row, the spin would go 1 2 3 4 6 7 8 9.

                  Part of the demo/standby mode will be animated auto-solving. It will
                  solve a puzzle slowly, illuminating the LEDs to show the progress as
                  it goes. Taking maybe a minute to solve a puzzle, so you have a
                  chance to see how it's doing it.

                  Dean.
                • fnatmed
                  ... 14W in under 1/2 cubic inch sounds awesome. I would definitely like to try one - that very nice of you to offer. My information is earlier in the post
                  Message 8 of 19 , Mar 2, 2007
                    --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@...> wrote:
                    >
                    > > Given that we're battery-powered, I would prefer not to dissipate
                    > > unecessarily. Can you describe a reactive energy transfer mechanism
                    > > that would slot into a simple flyback topolgy ?
                    > >
                    > > Dean.
                    > >
                    >
                    > A forward converter has an output reactor to average the square wave
                    > input to a DC value but the "Iron" needed to do that is far less than
                    > a flyback's inductor which has to store the entire cycles energy load
                    > in the switch on time. The transformer in a push-pull forward stores
                    > no energy so it's power handling capability is only limited by the
                    > current density of the windings.
                    >
                    > I've had really good results with the one I designed for my brothers
                    > clock and am getting better than 80% at 14 Watts out in under a half
                    > cubic inch: If you want to try it I'll send you one.
                    > http://www.tayloredge.com/storefront/1329_NixiePS/1329_graph1.jpg

                    14W in under 1/2 cubic inch sounds awesome. I would definitely like
                    to try one - that very nice of you to offer. My information is
                    earlier in the post train.

                    Do you have any pointers to more information - schematics etc ?

                    Thanks -

                    Dean.
                  • Nick de Smith
                    ... There s part of me, reading this, thats thinking about 9 IV18/21s in a row... t would have far fewer cables etc. Sending IR tuples, as in
                    Message 9 of 19 , Mar 2, 2007
                      --- In NEONIXIE-L@yahoogroups.com, "fnatmed" <halfwalker@...> wrote:
                      >
                      > --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@> wrote:
                      > >
                      > > I was thinking of using IR transmitters and receivers in a 9x9 grid
                      >...

                      There's part of me, reading this, thats thinking about 9 IV18/21s in a
                      row... t'would have far fewer cables etc.

                      Sending IR tuples, as in <row>,<col>,<value> would obviate the need for
                      too much complexity as well...

                      nicko
                    • fnatmed
                      ... Dammit - that s another potential project ! I can see those 9 IV18s floating in air, just connected at their wire-ends. I will NOT start thinking about
                      Message 10 of 19 , Mar 2, 2007
                        --- In NEONIXIE-L@yahoogroups.com, "Nick de Smith" <nick@...> wrote:
                        >
                        > --- In NEONIXIE-L@yahoogroups.com, "fnatmed" <halfwalker@> wrote:
                        > >
                        > > --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@> wrote:
                        > > >
                        > > > I was thinking of using IR transmitters and receivers in a 9x9 grid
                        > >...
                        >
                        > There's part of me, reading this, thats thinking about 9 IV18/21s in a
                        > row... t'would have far fewer cables etc.
                        >
                        > Sending IR tuples, as in <row>,<col>,<value> would obviate the need for
                        > too much complexity as well...

                        Dammit - that's another potential project ! I can see those 9 IV18s
                        floating in air, just connected at their wire-ends.

                        I will NOT start thinking about this.
                        I will NOT start thinking about this.
                        I will NOT start thinking about this.
                        I will NOT start thinking about this.

                        Dean.
                      • Jan Wuesten
                        To: NEONIXIE-L@yahoogroups.com From: fnatmed Date sent: Fri, 02 Mar 2007 17:14:19 -0000 Subject:
                        Message 11 of 19 , Mar 2, 2007
                          To: NEONIXIE-L@yahoogroups.com
                          From: "fnatmed" <halfwalker@...>
                          Date sent: Fri, 02 Mar 2007 17:14:19 -0000
                          Subject: [NEONIXIE-L] Re: IN12 Sudoku
                          Send reply to: NEONIXIE-L@yahoogroups.com

                          > --- In NEONIXIE-L@yahoogroups.com, "Nick de Smith" <nick@...> wrote:
                          > >
                          > > --- In NEONIXIE-L@yahoogroups.com, "fnatmed" <halfwalker@> wrote:
                          > > >
                          > > > --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@> wrote:
                          > > > >
                          > > > > I was thinking of using IR transmitters and receivers in a 9x9 grid
                          > > >...
                          > >
                          > > There's part of me, reading this, thats thinking about 9 IV18/21s in a
                          > > row... t'would have far fewer cables etc.
                          > >
                          > > Sending IR tuples, as in <row>,<col>,<value> would obviate the need for
                          > > too much complexity as well...
                          >
                          > Dammit - that's another potential project ! I can see those 9 IV18s
                          > floating in air, just connected at their wire-ends.
                          >
                          > I will NOT start thinking about this.
                          > I will NOT start thinking about this.
                          > I will NOT start thinking about this.
                          > I will NOT start thinking about this.
                          >
                          > Dean.
                          >
                          >

                          sure you will....not....or...

                          Anyway, happy and nice weekend to all Nixieans all around the globe

                          Jan



                          Mit freundlichen Gruessen---best regards----cordialement

                          Dipl.-Ing. Jan Philipp Wuesten

                          Frag'Jan zuerst - Ask Jan First GmbH & Co. KG
                          Preiler Ring 10 ; D- 25774 Lehe, Germany
                          Phone: +49-4882-6054551 Fax : +49-4882-6054552
                          <http://www.die-wuestens.de> (Deutsch, English, Francais)
                          mail to <FJZ@...> or <FJZ@...>
                          HR A4788 PI, Vertretung: Wüsten Verw.GmbH, HR B 6295 PI
                          Registergericht Pinneberg. GF: Jan Wüsten
                          **************************************************
                          Aktuelle Info: Abonnieren Sie unseren Newsletter
                          Stay updated: Subscribe to our newsletter:
                          Restez en ligne: Abonnez-vous à notre courriel de nouvelles:

                          Info: <http://www.wuesten.net/news.htm>
                          **************************************************
                        • Alan J. Franzman
                          ... How about 9 ZM1204s (10-digit Pandicon)? I think it would be a bit difficult since I don t recall ever seeng *one* of those; only the 14- digit ZM1200 and
                          Message 12 of 19 , Mar 3, 2007
                            --- In NEONIXIE-L@yahoogroups.com, "Nick de Smith" <nick@...> wrote:
                            > --- In NEONIXIE-L@yahoogroups.com, "fnatmed" <halfwalker@> wrote:
                            > > --- In NEONIXIE-L@yahoogroups.com, "taylorjpt" <jpt@> wrote:
                            > > > I was thinking of using IR transmitters and receivers in a 9x9
                            > > > grid...
                            >
                            > There's part of me, reading this, thats thinking about 9 IV18/21s in
                            > a row... t'would have far fewer cables etc.

                            How about 9 ZM1204s (10-digit Pandicon)? I think it would be a bit
                            difficult since I don't recall ever seeng *one* of those; only the 14-
                            digit ZM1200 and 8-digit ZM1206 ever seem to show up. Maybe using 11
                            or 13 positions of the ZM1200 would work (3 digits, gap, 3 digits,
                            gap, 3 digits -- gaps could be 1 or 2 digits wide).

                            Does anyone own 9 of *any* model of Pandicon?

                            A.J.
                          • Nick de Smith
                            ... Best I can do is 1206 x 6, and they are only 8 digit. IV-18 are only 8 digit, and IV-27s are 13. So, IV-27s may not be nixies, but are cheap and plentiful.
                            Message 13 of 19 , Mar 3, 2007
                              --- In NEONIXIE-L@yahoogroups.com, "Alan J. Franzman"
                              <a.j.franzman@...> wrote:

                              > Does anyone own 9 of *any* model of Pandicon?

                              Best I can do is 1206 x 6, and they are only 8 digit.

                              IV-18 are only 8 digit, and IV-27s are 13.

                              So, IV-27s may not be nixies, but are cheap and plentiful. The extra
                              digits could be used for other purposes, or masked off...

                              Nickolovich
                            • taylorjpt
                              ... This is the approach I m going for, where multiple taps scroll through the digits etc. Measuring the speed of the taps and delays, this will also be my
                              Message 14 of 19 , Mar 3, 2007
                                > That's an interesting approach. How about using 18 transmitters and
                                > receivers around the edge, just above the tubes, one set of 9
                                > horizontal, one set vertical, intersections right over the tubes. As
                                > your finger comes down over a tube, it blocks a unique pair.

                                This is the approach I'm going for, where multiple taps "scroll"
                                through the digits etc. Measuring the speed of the taps and delays,
                                this will also be my user interface so there will be nothing but
                                nixies showing. To run off batteries, I can power down the nixies
                                when no activity is sensed and then use only the lower row sensor to
                                periodically sense for a power up signal. I should be able to get a
                                couple years of standby from any type of pack this way.

                                http://www.tayloredge.com/storefront/1328_SmartNixie/Sudoku.pdf

                                jt
                              • taylorjpt
                                ... I ll send you one on Monday. jt
                                Message 15 of 19 , Mar 3, 2007
                                  > 14W in under 1/2 cubic inch sounds awesome. I would definitely like
                                  > to try one - that very nice of you to offer. My information is
                                  > earlier in the post train.
                                  >
                                  > Do you have any pointers to more information - schematics etc ?
                                  >
                                  > Thanks -
                                  >
                                  > Dean.

                                  I'll send you one on Monday.

                                  jt
                                • David Forbes
                                  Sudoku gadgets have some unique non-clockish requirements that need to be taken into account. The first is legibility. I received a $5 Sudoku game as a
                                  Message 16 of 19 , Mar 3, 2007
                                    Sudoku gadgets have some unique non-clockish requirements that need
                                    to be taken into account.

                                    The first is legibility. I received a $5 Sudoku game as a Christmas
                                    present. It uses a 7 segment LCD display with lines printed on an
                                    overlay to define the 3x3 groups. This gadget is rather hard to read,
                                    because a jumble of 7 segment numbers is hard to make out, especially
                                    on a passive LCD display in poor lighting. They didn't provide extra
                                    space between the groups, so it's hard to see which group a central
                                    row lies in.

                                    The Russian VFD calculator displays would be rather poor for Sudoku,
                                    as would the Pandicons, because the horizontal spacing would be 10
                                    times as tight as the vertical spacing.

                                    The ideal Sudoku display would be an array of round or oval nixie
                                    tubes, spaced uniformly in 3x3 groups with an extra smidgeon of
                                    spacing between the groups.

                                    The control system is also worth consideration. I see that some folks
                                    are considering using optical means of finger detection to select
                                    cells and numbers. That is a good idea, if it works.

                                    The $5 Sudoku game I have uses a touch screen with a built-in plastic
                                    stylus. There is a row of numbers along the bottom of the screen that
                                    is used to select what a cell will be set to, then a cell is touched
                                    to set it to the selected value. A "C" cell at the top of the display
                                    puts the unit into clear mode, where touching a cell will render it
                                    blank. The cells themselves have a tic mark in the lower right corner
                                    that indicates a pre-filled cell. The tic mark in the most recently
                                    changed cell blinks slowly.

                                    I was considering a capacitive touch system using a loop of wire
                                    wrapped around the top of the nixie tube to sense finger capacitance.
                                    It would take a bit of experimentation, but it could work well. The
                                    coils could be wound separately as formless self-supporting coils
                                    (like the ones in disk drive voice coils), then placed over the tubes
                                    and glued into place.

                                    Another possibility is a pushbutton switch or two next to each tube.
                                    They make very small right-angle tact switches that could be mounted
                                    on a PC board placed edge-up between the tube rows. You'd need nine
                                    of these narrow, long PC boards.

                                    The indication of pre-filled cells etc. can be done easily by
                                    brightness control via pulse-width modulation. I'm assuming that
                                    multiplexing is called for in the design to reduce the number of
                                    driver chips. With a gadget that's only activated occasionally (you
                                    have to sleep a few hours a night!), tube lifetime is not too big a
                                    concern.
                                    --

                                    --David Forbes, Tucson, AZ
                                    http://www.cathodecorner.com/
                                  Your message has been successfully submitted and would be delivered to recipients shortly.