I'm planning a project that will involve driving four 7-segment flip-digit displays. These are electromechanical devices where each segment is either displayed or hidden depending on the direction of current flow through a coil. It looks like the best way to drive each segment is via an H bridge. Even though these are cheap and easy to use, and you can get two in a 16-pin package, that's still 14 packages, plus five packages of inverters, so I started thinking about multiplexing them.

In the attached drawing I show four coils (well, motors since that's what was used in the drawing I edited, but pretend they're just coils). If each coil corresponds to the same segment in each of the four digits, in principle I could get away with only four H bridge packages and two packages of inverters, if I could selectively interrupt the current at each of the red Xs, of which there would be 28 in the full circuit. Ironically, this approach would increase the number of required IO lines (the opposite of what multiplexing is usually used for), but drastically reduce the package count. The Xs could be relays, but that's a lot of relays, and I'd prefer something solid state for all the usual reasons. They can't be transistors since current has to flow both ways. Would triacs work? I've never used them before, but they ARE intended for bidirectional current flow, even if it's typically AC. I realize they stay latched until current stops flowing, but I can disable the H bridge to induce that. Any other possibilities?

 

I'm not sure how you would multiplex these without incurring unwelcome complexity but if you use a split rail supply for driving the segments you could halve the number of driver chips, eliminate the inverters and simplify the wiring because one end of each segment motor would connect to 0V.

 

Unfortunately, the coils aren't center-tapped, so there are only to contacts, one at each end. Don't know how to connect a split-rail supply to that.

 

How much current is required per coil to flip the disk?
74126 can source about about 5mA.
74LVT126 can source about 30mA.

 

Each segment requires 350 mA @ 19V for 1 ms. to flip on or off.

 

Unfortunately, the coils aren't center-tapped, so there are only to contacts, one at each end. Don't know how to connect a split-rail supply to that.

I know they aren't center-tapped and they don't need to be. Imagine a power supply with 3 outputs; +V 0V -V, if you connect one end of the coil to 0V you can connect the other end to either +V or -V and thereby reverse the current by switching only one end of the coil.

 

Do you have a data sheet on the display?

 

I know they aren't center-tapped and they don't need to be...

Aah, of course. Thanks for pointing that out. Thinking it through, though, I'd have to use 56 transistors and at least that many resistors. Plus, I've only been able to find one switching regulator on Digikey with the necessary specs to produce -19V (a Linear Technology LT1614CS8). Conceptually this is a simpler approach, but I'm not sure ti would work out that way in practise.

 

Do you have a data sheet on the display?

Here you go.

 

I think you are in for some real work. Unless you can find a driver chip for that display. They have to be using something.
You would have to build a driver/decoder for each display.

The bi-directional current is a bummer.

Have you looked for an alternative?

 

Yeah, it's a challenge, but this is the one I'd have to use if I go forward with this project. For me, the biggest feature is they only draw power when changing state; nothing when stable. Since my application (a sculpture) will be battery-operated, that's a real plus. Also, since I need about 50 digits this size, an LCD is out of the question. There used to be a company that made large, individual LCD digits, but not anymore, and I can't justify custom ones. I also really like the "electromechanicalness" of these flip things.

I'm thinking of building a 4-digit prototype to see if I can get them to work (probably with H bridges). Then I'd build up the actual piece out of multiples of the prototype.

The digit vendor makes a driver board, but it's $13/digit and talks RS485. I'll be evaluating it, but I wanted to explore my own driver circuit in case I could do it more cheaply and simply

 

What were you planning to use to control the display? Are you controlling other parameters?

Can you buy one and copy it for less than $13? Is it worth your time?

See what kind of discount you can get for 50 units.

 

What were you planning to use to control the display?

An AVR plus the previously mentioned H bridges etc.

Are you controlling other parameters?

Like what?

Can you buy one and copy it for less than $13? Is it worth your time? See what kind of discount you can get for 50 units.

That IS the 50 piece price. One piece is $23, plus $50 for shipping from Poland (of course, I could ship a lot of pieces for that $50). I'm not sure I could copy it because it sounds like (their English isn't great) they use a custom decoder chip which is no longer available and they won't sell to me from their declining stock. Plus, RS485 adds a layer of complexity I'd prefer to avoid. As I said, I'm planning on ordering a couple to evaluate. Maybe they aren't using the custom chip, and I could leave out the RS485 from my version.

 

Whatever you do, your insistence on using those displays will take money and time.

I would research for an alternative display. Both mechanical and electronic.

Good luck on your project.

 

For the suggestion in post #6, you can use ULN2803 octal NPN darlingtons to reduce the complexity of the "pull down" direction. But there is no equivalent pnp part, so you are correct that there would be a bunch of small PNPs and resistors. Thinking this through, the +/-19V approach has some interface voltage issues.

First thought, the system has a a 3.3 V or 5 V supply for the uC, a +19 V supply for the common coil connection, and a +38 V supply. One bank of 2803's pulls the other end of each coil down to GND. Another bank of 2803's acts as level translators to drive the individual PNP transistors that pull the coils up to +38. which is only 19 V difference from the common coil connections. Complex, messy, lotsa parts... The H drivers are looking better...

ak