I've got a breadboard up and running with an adapter that uses four LTP-305G displays, each is 5x7, so this is 20 columns by 7 rows.
Using some dual logic level mosfets (DMN2004DMK-7) for the rows.
Using a 182 ohm resistor for each column.
I've targeted the current so that it is around 12.5mA pulse as 14% duty cycle which results in 1.785mA average pixel current. All pixels on should be around 250 mA.
On the AVR side, the datasheet says:
current in/out vdd/gnd pins - 200mA - the device has 3 sets of pins.
i/o sink/source - 40mA
max sink/source per group i/o pin group - 200mA @ 25C
the footnotes say that PA, PC, an PD are in independent groups. I've got 8 columns on A, 4 on C, 8 on D.
All off = 5mA.
All segments on 178-5=173mA, a far bit from 250mA. 5 columns on, 15 columns off (1 char all on) 64-5 = 59mA.
Looking at a pin that was left on while toggling the others off I saw a 0.35V difference in the pin voltage.
If I turn down the voltage from 5V to 4V or 3V, the problem of still present.
I finally measured the current starting with 1 column and turning on 1 more at a time - the first column is total current, the second is average.
| 14.9 | 14.9 |
| 26.5 | 13.3 |
| 37.8 | 12.6 |
| 48.8 | 12.2 |
| 59.2 | 11.8 |
| 69.4 | 11.6 |
| 78.2 | 11.2 |
| 86.5 | 10.8 |
| 95.9 | 10.7 |
| 104.2 | 10.4 |
| 111.8 | 10.2 |
| 119.2 | 9.9 |
| 126.8 | 9.8 |
| 135.5 | 9.7 |
| 142.2 | 9.5 |
| 149 | 9.3 |
| 156 | 9.2 |
| 162.1 | 9.0 |
| 168.3 | 8.9 |
| 174.3 | 8.7 |
So the question is what could be done to improve consistency?
I thought about doing 10 columns by 14 rows, but then I'd have to double the pulse current and I suspect that might be even worse than what I am doing now!
If I can't drive the columns from the AVR, it just won't be a workable project.
, you could "mount" sockets & plug the displays in....then you'd be able to tilt & wiggle the displays around slightly to keep the side-to-side distances exact. However, the tape should do the trick 
