Kai, could you explain, please? I have looked once again and your schematic still looks wrong for me. I am confused now...

Hallo Oleg,

as we have seen from the discussion up to now, it's not trivial to drive a 40x7 LED matrix by multiplexing: In order to prevent flickering a sufficient high refresh rate is needed. And if shift registers are used for the communication between micro and LED matrix, then micro wastes a lot of time with feeding these shift registers.

For this reason I would recommend to introduce some parallel communication between micro and LED matrix.

But how to multiplex? By rows, or by columns?
Multiplexing by columns has the disadvantage, that LEDs must be driven by a current which is 40 times the average current. And although most LEDs can tolerate this for a short period of time, they are in danger to become destroyed, if micro isn't able to maintain the proper multiplexing, means if there are situations, where columns are driven for much longer duration than intended. This can happen, when code still has some bugs, or with certain brown-out situations.

Here, multiplexing by rows is better, because momentary current through LEDs is only 7 times the average current, and a simple monoflop can provide a safe protection against most failures.
A big disadvantage of multiplexing by rows is, that the use of simple Darlington drivers will result in relevant fluctuations of intensity of LEDs, depending on how many LEDs of actual row have to be turned-on. It's obvious, that Uce of Darlington relevantly differs, if only one LED has to be driven or 40! Multiplexing by columns would derive benefit from much smaller range of load change, namely only 7:1 instead of 40:1. It can easily be calculated, that multiplexing by rows here can result in fluctuations of intensity of more than 20%, if simple Darlington drivers are used!

In the following schematic (which I attach here again) I have choosen a different approach, which would eliminate these fluctuations: The use of a 'strobed' LM317 for each row.








In order to prevent flickering, I would turn-on each row for 1msec every 7msec. 1msec is pretty enough time to be able to turn-on and turn-off the LM317. To prevent over- and undershots when switching the LM317, a 220nF capacitor is connected in parallel to 500Ohm trimmer. This capacitor will heavily slow down turn-on and turn-off times and by this reduce not only over- and undershots of LM317, but also the produce of EMI.

How are the LM317 controlled?
LM317 fabricates a very stable voltage of 1.25V between 'output' pin and 'adjust' pin. A resistor, here 240Ohm, placed across these two pins forces a current to flow, here 5.2mA, which also flows through 500Ohm trimmer and causes a voltage drop. If trimmer is adjusted to 375Ohm, then output voltage of LM317 is simply the sum of 1.25V plus voltage drop of 5.2mA x 375Ohm = 1.95V, yielding 3.2V totally. This is the way LM317 works as voltage regulator.
If 500Ohm trimmer is exchanged by a short circuit to ground, then according voltage drop is 0V and output voltage falls to 1.25V.

Assume now, that 'ROW 1' shall be turned-on. To achive this, 74HC138 is programmed by the micro in such a way, that '/Q0' output of 74HC138 is logically low. All other outputs of 74HC138 are logically high then. BC337 of 'ROW 1' is turned-off, which causes all the current flowing through 240Ohm resistor also to flow through 500Ohm trimmer. As consequence output voltage of LM317 rises to 3.2V.
The BC337 of all other rows are turned-on, and by this, shunting the current flowing through according 240Ohm resistors directly to ground. As consequence output voltage of these LM317 is about 1.35V, a voltage which is too little to be able to let the LEDs of according rows shine.

As already mentioned above, a CD4538 monoflop is used to provide a protection against excessive overload conditions of LEDs: Monoflop is strobed by the same port pin, that strobes the watchdog of MAX1232. Strobing pulses retrigger the monoflop all the time, so that '/Q' output of CD4538 is permanently emitting 0V. This signal is used to enable the 74HC138.
If, for any reason, strobe pulses disappear, monoflop is no longer triggered, '/Q' output of CD4538 goes high and all the LED drivers are disabled.
10msec was choosen for the monoflop pulse period and monoflop should be strobed every 7msec.

Further protection is provided by feeding the 'reset' signal to another enable input of 74HC138: Whenever the micro is reseted (brown-out situation, etc.) LED drivers again are disabled.


Kai