DMA port "update" ideas..

[exxos]

Will look at this today. Is this worth a jumper on the Rev.H board to be able to select between 8MHz and 16MHz?

Yes, if its likely to work that is...

[exxos]

You are one hell of a multi-tasker.. how do keep every strait?

I don't, I'm quite insane you know

[Icky]

16MHz works on DMA and can see floppy with the new WD1772 buffer board. Reading, writing and formatting all work.

However as before if I plug in Gigafile as well it doesn't see HDD or Floppy.

[exxos]

Odd. Good that the floppy works.. Could be the DMA thinks gigafile is ready when its not. Could just be a timing issue in GF.. Not sure why floppy not working with GF, sounds like maybe data conflict .

What I may do is loan you my ultasastan to try out after Xmas.

There is another thought that we could run the DMA at stock speed while its getting data from a device but double clock it when its copying to ram. On a stock machine its not going to show any speed boost as ram isn't running any faster.

In any case. It does hint like the DMA is OK running at double speed if you can format a floppy OK and read from it OK..

[rpineau]

I'm not sure that running the DMA at 16MHz will accelerate any transfer as the GLUE/MMU are the one generating the memory address for the transfer and not the DMA itself (it only has A1 from the address bus). So running it at 16 MHz would only affect the signals generated by the DMA and it might end up out of sync with what the GLUE/MMU are doing doing.

[ijor]

I'm not sure that running the DMA at 16MHz will accelerate any transfer as the GLUE/MMU are the one generating the memory address for the transfer and not the DMA itself (it only has A1 from the address bus). So running it at 16 MHz would only affect the signals generated by the DMA and it might end up out of sync with what the GLUE/MMU are doing doing.

Right. These DMA cycles (as opposed to the ones generated by Blitter) are a coordinated effort among the 3 chips. The only thing that DMA does is to pulse the RDY signal. GLUE is the bus master and asserts all the bus control signals at its own pace. The addressing is generated internally by MMU.

It's interesting and even surprising that it works at all at 16 MHz. But not so much surprising when you consider that the interface with DMA is not fully synchronous. The exact behavior might depend on the specific version of the chipset, GLUE and DMA.

[rpineau]

Yes, I think that the fact it works is mostly because of the GLUE/MMU timing not being changed.

[Icky]

I guess it would be interesting to try this on a STE which is having DMA issues... *writes on 2020 to do list*

Well I did not get as far as my STE but was in the process of sorting out my STFM DEV Motherboard to allow easier access to some of the chips.

Today I socketed the 1772 so thought I would try out the 1772 Buffer Board from the STF Remake.

All running well and functioning correctly. I placed a Gotek along side to show that the board fits nicely with it and doesn't restrict the cables.

IMG_3638.jpg (110.31 KiB) Viewed 1207 times

[exxos]