PulkoMandy

For afew days I was thinking about ways to interface the EF9367 video chip with 4464 RAM (it is designed for 4116 originally). I thought I could use "page mode" to do two successive accesses to the RAM and avoid having to use 8 RAM chips. But it's not going to happen, the EF9367 is way too slow for that. I hadn't realized how tight its timings are. This could be fixed with some extra buffers, but then the circuit becomes way too complicated.

Well, one useless project out of my mind :)

Apr 12 at 7:39pmWeb
Show threadAGMSApr 13

Looks up EF9367. Ooo, ahh, it can do 512 pixel wide scan lines. That means 80 column text!

My current project #NthPongEars for the #NabuPC uses a 4 year older TMS9918A video chip that just does 256 pixels per line.

Show threadPulkoMandyApr 13
@AGMS00 it can even go up to 1024 very narrow pixels, that the previous generation EF9365 and 66 couldn't do. But in all cases, they need quite a not of supporting logic (external shift registers, sequencer made of a counter and a lookup ROM, multiplexers, ...). And all write accesses are done during the vertical and horizontal borders only, so drawing performance is slow unless you turn off the display. And also, it has line and text drawing commands but no direct write access to the RAM
Show threadPulkoMandyApr 13
@AGMS00 so, no scrolling, no bitmaps, etc unless you add extra support logic to achieve that. And then that's so much logic, you may as well do your own design from scratch?
Show threadAGMSApr 13
That actually makes the TMS9818A sound better! Its API is mostly about writing to VRAM: text, font, sprite table. Though it has a similar problem of missing writes outside vertical blank. But it does include all the video signal generation hardware on chip (has a composite video output pin), and DRAM support logic.
Show threadAGMSApr 13
#NthPongWars, not autocorrect Ears. That would be a bouncy rabbit game or something weirdly audible.