No, of course I don't expect you to do that!

It just seems as though you've occasionally ignored the salient part of my comments. I don't mean to give offense. It just seems that, in one of your previous remarks, at one point in a previous thread, you say,
though I'd already pointed out at that point that what might be faster and more effective than registered SIPO registers, is parallel bus interfaced '273-type latches driving ULN2803's which then should sink current through a resistor. It certainly would be less costly than the current-controlled cathode drivers, yet would, because only one anode in each column is driven at a time, allow the current to be limited by a single resistor in each column. That way, each darlington sinks current, through that resistor, from a single cathode. Each one is capable of up to 500 mA, but would have to sink only 100 mA.

I had pointed that out, I believe, on several occasions, including some in which I made misleading comments when I had not noticed that the LED array to which I pointed with that URL was not the one I thought it was, being, instead, a row/column-wired one. That was my mistake, certainly, and may have led to considerable confusion as to what I was "getting-at." I made remarks asserting that one should control the current with individual resistors at the anodes, though that was clearly not possible with that particular array.

Since there's a need for external memory, due to the choice of AT89S52, he has not as much memory on-chip as he needs. He wishes to use an array consisting of 80 columns x 16 rows. That would be at most 26 full (using 5x7 matrix) characters (using 5x7 matrix) on each of two lines. He wants to display a 50-character-long message which needs to be stored, in the message memory, and at 5x7 that would require 5 (column) bytes per character in a vertically oriented bitmap buffer from which the columns are driven. 50 characters would presumably require 250 bytes of recirculating bitmap buffer space needed to support scrolling.

I previously misunderstood what Aamir wanted to be 80 character messages, and that is not the case. Nonetheless, I believe that the parallel register and ULN2803 is the cheapest and most readily accessible path for Aamir to follow in building his sign. Using that hardware scheme, he can easily switch to an externally controlled SIPO scheme that uses X2 to clock his SIPO and to gate the clock if he wants the speed. It would, of course be easier with the parallel registers, since he then wouldn't have to concern himself with the generation and counting of clocks. He'll still have plenty else to do.

I actually like your 8:1 multiplexing scheme, and I see that it allows either the parallel or serial approach to signal routing to be used. I'd say it's his most promising option, particularly since it would probably work very well with parts he's already decided he can obtain.

RE