Mike, I
think I'm right on this. Let's just consider the red OR the green for now. The display is 5 columns wide by 7 rows tall. The rows are the anodes (positive side) of the LEDs and the columns are the cathodes (negative side) of the LEDs. The anodes are tied together across all the rows and the cathodes are tied together down each column.
J
View attachment 67790
The above picture is from the data sheet.
The anodes are driven by a port pin via a resistor. A total of seven port pins and resistors are used. Each port pin can supply a maximum of 25 mA.
Each of the 5 columns (considering just the red or the green) are connected to ground via a ULN2803 (or ULN2003) low side driver chip. Each of these gates can sink 300+ mA.
If 1 - 7 bits of the port are HIGH, and EXACTLY ONE of the ULN2803 gates is active (I.e., conducting to ground), 1 - 7 LEDs will be illuminated, each port pin connected to an anode will be supplying < 25 mA and the ULN2803 gate will be happily sinking up to 175 mA depending on how many LEDs are on in a column.
Using both colors just doubles the columns to 10. Still, only one column is active at a time.
All of the above presumes Vf of the LEDs is less than 4 volts. Since MrDEB was able to get the LEDs to illuminate, this must be the case.
If the drawing MrDEB did above showing the current measurements is to be believed and I'm interpreting it correctly, the hardware configuration should match what I've described. He's just using it incorrectly by turning on more than one cathode at a time. When this is done, more than one LED is supplied by a single resistor, leading to variable brightness and unexpected results.
I hope this has clarified my thinking.