I want to make a driver board for my flipdot display. It is 28 columns and 16 rows. The circuit is pretty simple.

To flip a dot I need +6v on a high pin or -6v on a low pin and 0v on the column. I realize that this is the same as +12v high, 0v low and +6v on the column. But either way I am not aware of how to produce a middle reference except for a voltage divider which obviously is a bad idea.

It takes about 2 amps to flip the dot. Because of this I don't think it is going to be possible to flip an entire column at once but instead do them one by one. Luckily it is only supposed to take 1ms to flip them.

So my two questions are: How to I make a halfway reference between 12 and 0 volts that can handle 2 amps? And what sort of circuit should I be looking for that I can multiplex only one row at a time as well as one column at a time? My goal is to control the circuit with an atmega2560, is there a way for my row pins to be high, low or just open and reliably keep whatever is switching the rows neither high or low until actually set high or low?

 

You may want to look at an H-bridge driver such as L298.
I would be inclined to tie ROW1H and ROW1L lines together and let the L298 do the current steering at both ends of the coil, i.e. column and row.

 

I think I get what you're saying. Would I still be able to leave rows open with this chip so that I can control only one dot at a time? Can that H-bridge be left open?

I am also worried that it seems like each l298 would only handle two rows or columns each. That is going to take up a lot of real estate and money. The only other H-bridge that I am familiar with is the ULN2003 but that only goes up to .5A.

 

You may want to look at an H-bridge driver such as L298.
I would be inclined to tie ROW1H and ROW1L lines together and let the L298 do the current steering at both ends of the coil, i.e. column and row.

I'm not quite sure I follow.
How many L298's will that require?

 

There are four half-bridges per one L298 package. Unfortunately they are not independent, i.e. it was meant for driving two stepping motors.
You can use two independent half-bridges and ignore the other two.

You need 44 half-bridges in total, 28 columns + 16 rows.

Total of 22 L298 packages.

 

So here's my understanding of how the L298 would work.
The column and H & L rows-tied-together will have three states, 6V, open, or ground.

Thus to energize a coil positive on right, the row is set to 6V and the desired column is grounded.
To energize a coil positive on the left, the row is grounded and the column is set to 6V.
Unenergized rows and columns are set to open (L298 input disabled).

This does require a 3 state signal for each column but this can be derived from the row signals and a column enable signal.

1. If the column is enabled and the row signal is enabled high, the column is set to ground.
2. If the column is enabled and the row signal is enabled low, the column is set to 6V.
3. Columns not enabled are set to open, regardless of the row signal.

Note that the L298 has about a 2V drop through both transistors so you would likely need a supply of around 8V to get 6v across the coil.

 

So just to clarify, the L298 will act open when not enabled? Because it is going to be important that whichever column I am on, I am still only affecting one row (one dot) at a time to stay at 2A. So is it correct that the idea would be that only one row and column are enabled at a time? I just want to be sure that I can rely on it to act like an open circuit when not enabled.


Is there another option? This is looking like over $100 in L298's Are there other half-bridge packages that can do 2A that are less expensive per channel? What about a bunch of MOSFETs? Even if I needed two per row and column (16 + 28)*2 that still seems like it would be considerably cheaper.

 

Could you use 6VAC on the columns and connect each row as needed?

 

There are other approaches to this problem.

You can still use half-bridges on the columns and then use logic-level low side MOSFET driver (N-channel) on the Row_xL and logic-level high side MOSFET driver (P-channel) on the Row_xH.

 

Could you use 6VAC on the columns and connect each row as needed?

AC?

 

Triacs could be used for the row and column switches and run it off AC volts as each coil will run half wave rectified via the diodes.
This is not to say the best way to run it, but just another idea to muddy the waters

 

Triacs could be used for the row and column switches and run it off AC volts as each coil will run half wave rectified via the diodes.

Okay, that makes sense.
You could use one TRIAC for each column and one SCR or TRIAC for each L and H row signal.

So to flip a dot you would energize the appropriate column with its TRIAC and the SCR/TRIAC on the appropriate row.

That would require a total of 32 SCR's and 28 TRIAC's, or 60 TRIAC's.

You could use a small TRIAC such as this, which is only $18.70 for a hundred.
It's rated for 1A continuous but can carry 2A for up to 200 cycles.
It should only take 1 full cycle to flip the dot.

Edit: Just realized you could simplify to one TRIAC per column and one TRIAC per pair of row lines, if you phase triggered one of the TRIACs on one half of the phase only, positive or negative, depending upon which way you want to flip the dot.
This would reduce the part count to 44 TRIACs.
Here's a concept schematic using the TRIAC half-wave trigger approach:

 

I am not crazy about the idea of using an AC power source. Maybe I am understanding it wrong but I would think that timing would be an issue. I would have to sense and wait for the current to be going the right way before doing a flip. I am also worried that generating an AC power source will add a lot of complexity.

What about using a triac at the column but then just using mosfets at the rows and going back to the + and - 6v idea so that I can do the same thing but instead of tying high and low together I just go back to controlling them separately?

(edit: or instead of mosfets, would SCRs be more appropriate? I am actually pretty unfamiliar with triacs and SCRs until now. Seems like we are building the same thing but I feel like having +/- 6v would be more simple than using AC)

Another issue is that I am playing with the l298, and while I can get it to work, it is taking something like 15 or more volts to actually sync and source 2amp. Apparently the dropout voltage increases a lot with higher loads. So I am not crazy about these either.

 

You cannot use Triacs or SCRs on DC, well, not easily. They can be turned on but rely on the current flow stopping to turn off.
If this unit is to be mains powered then you can use an appropriate transformer to supply the AC. And timing will not be an issue as you will be running this from a processor of some type. A"zero crossing" detector will be used, ( http://www.bristolwatch.com/ele2/zero_crossing.htm ) and it can also have an input to detect the phase. And as this happens 50 or 60 times a second depending where you live, it will be plenty fast enough.

 

as this happens 50 or 60 times a second depending where you live, it will be plenty fast enough.

There are 448 that can theoretically be flipped in 1ms each. So that's 7.5 seconds vs 0.5 seconds.

What about triacs on the columns and mosfets on the rows? Then I can turn off a column by turning off all of the mosfets.

 

You can flip more than one at a time, but I'd be careful about switching too many as then the power supply will need to be a lot bigger.
I've not built this circuit and probably will not anyway so I think if you want to try it, build a couple to see how it works.
No, triacs all around as FETs are DC only.

Ebay has some cheap triacs...
http://www.ebay.com.au/itm/2Pcs-Bta...873274?hash=item41c4566fba:g:iJkAAOSwB09YHbsF

and some opto triac drivers for the row triggering...
http://www.ebay.com.au/itm/5-10-203...hash=item46560a25e5:m:mWQAGi5mY7SuuNX544nsIQg

 

You can flip more than one at a time

To match the speed of just not using AC would require flipping 15 at once which would require 30 amps. This idea seems to just be making things more and more complicated.

 

To match the speed of just not using AC would require flipping 15 at once which would require 30 amps. This idea seems to just be making things more and more complicated.

You only need to supply that current pulse for a few mSecs so that is not a great problem. It is not running continually.
I think going the AC way will be a lot cheaper than having to make a DC system. For a start, no big DC supply is needed, and it will be easier to interface.
Whatever way you decide on, please post your results here so we can see how you went.

EDIT...
Have a look at using something like these SSRs..
http://www.ebay.com.au/itm/G3MB-202...779688?hash=item361aea0a68:g:mecAAOSwvjdZSN8X

There are boards with 8 on them available too, and if you run them from a 74AC138 1 of 8 decoder, 74AC139 dual 1 of 4, or some other gating, wired so you will not be able to turn 2 sides on together, you will protect the circuit a bit.

 

generating an AC power source will add a lot of complexity.

Generating AC just requires a transformer.
Generating high current DC is more complex.

 

I understand but I am more worried about having this thing update very slowly than finding a high amperage power adapter.