Now, I haven't done this for a long time, but ... IIRC ... the signals all have to be stabile, and only then can the E strobe be raised. The signals must never, Never, NEVER change during the posititive phase of the E strobe, and only after the E strobe is low once again, may the remaining signals and data change. This protocol was "stolen" from MOTOROLA, which used it with a number of its devices, notably the MC6845, where the timing was similar, and the protocol was the same.


The clearly idiotic practice of grounding the nWE signal in order to avoid using one pin has become entirely too widespread. It was commonly done when people were unable to change the direction of their PC printer port back in the '80's, when people were "playing" with LCD's on their PC printer port, and, for some inane reason has persisted. If you can't "afford" the one pin for the R/W signal, then go to 4-bit mode ... it's not slower, since the typical HD44780 instruction execution time was specified at 40 microseconds. As an alternative, you might consider memory mapping the LCD onto the MCU and operating it at a rate compatible with your LCD, since few MCU's can extend their external bus cycle timing.

RE