How does the Denise chip (and Lisa) generate the display?

[TroyWilkins]

I hope this is in the right section, if not I apologise.

I'm trying to get a better understanding of how the Amiga works exactly so I can write it up on my web site, but I'm not entirely sure how Denise/Lisa generates the display.

What I mean is, does Denise/Lisa simply read from the appropriate place in Chip RAM as the electron beam scans across the display and generate suitable output to the display device, and thus Agnus/Lisa changes the data in the chip RAM if need be (say for a copper effect or display change), or is it done some other way?

[Daedalus]

Yeah, that's basically it alright, although Lisa & Denise produce digital output that needs to be converted to analogue to be displayed, so there is DAC circuitry between the custom chips and the actual video output you get from the RGB or composite ports.

There isn't a specific place in chip RAM that contains the display information, so the position in memory the information is read from can be changed easily. This is the basics of double buffering - one bitmap is being drawn while the other is being fed to the display, then they're swapped over so the previously drawn bitmap is now the one being fed to the display.

[pandy71]

Principle is relatively simple - there is group of counters that can be programmed (and some of them not), they count clock and based on outcome together with comparators trig particular strobes inside Denise and Lisa, there is group of latches where video bitplane data are load and latched, there are some comparators working together with counters, finally there are parallel to serial shifters and bunch of gates to detect collisions.

[Toni Wilen]

Denise/Lisa (and Paula) do not have direct access to chip RAM.

Agnus/Alice read from chip ram and then write the data words to Denise/Lisa BPLxDAT (if bitplane) and SPRxDAT (if sprite) registers.

[TroyWilkins]

Ahh, thank you very much. So one of the DMA channels in Agnus/Lisa is used to provide data to Denise/Lisa, and the same for Paula? With Paula, is it one DMA channel per sound channel? Sorry if these are annoying questions, but I've foudn that this information isn't easy to find using google.

[Toni Wilen]

HRM DMA allocation chart explains most of it, indirectly.

Paula has 5 channels, 4 for audio and 1 for disk (or 2, disk DMA can work in both directions). Works same way as Denise, Agnus/Alice writes to AUDxDATs and DSKDAT (or reads from DSKDATR).

Note that audio DMA writes do not cause any immediate sound change. DMA fetches are not fully hardwired to audio state, they are done early. When new sample word starts playing (first byte of word), next sample word DMA request goes to Agnus/Alice. (This is probably getting too technical but at the beginning of each scanline, Agnus/Alice gets serial DMA "mask" from Paula that tells which Paula's DMA slots require new data. This is not officially documented anywhere and explains smallest possible audio period without repeating samples.)

[TroyWilkins]

Not too technical, thank you very much for the detailed response Toni, very much appreciated! And thank you for the reminder about the HRM, just having a look at that now and it looks like it should prevent me from asking what must seem like basic questions.