Jez,

The reference design looks to be a good VGA character (or tiled cell) display and with additional memory, the basis for an equally suitable graphic display.

Did you implement the "super-pixel" character display or did you add memory for bit-map capability?

My recent experience seems to indicate the definition of the desired utility of the display rapidly expands the memory requirement where as timing is generated rather more easily.

Impacting this memory definition as well as expected utility is how tightly coupled the display sub-system is to it's host system.

In regards to "TV" targeted video, and Atmel/CCellar/Zx-80 video references. I'm finding that area of information not published deals with the displayed geometry of the resulting video.

The ability to adjust this geometry is more important as the television device becomes less expensive. App note demonstrator projects can avoid this concern (but does seem to dictate the published video images). Anyone intending a respectable display on a TV device might find it harder to ignore.

Rock solid timing does not, by itself, guarentee square field display. Seems inexpensive TV's depend upon overscan to mask edge and initial line display problems.

Having the ability to adjust timing by row, addresses this pretty neatly and so far seems to not be "dynamic" but instead tailored to the target device.

regards,
p