DMA port "update" ideas..

[Icky]

If you can format a floppy fully, it will test the DMA out at the higher speeds better...

Assume gigafile doesn't like the faster port speed

I'll give that a go shortly - need food first

Formatting floppy under 16MHz works

[exxos]

Formatting floppy under 16MHz works

Thats good to know.

[rpineau]

I'm not sure it's going to speedup transfer to memory as the DMA is not the one addressing the memory, so might terminate cycle to FDC and DMA devices faster but not transfer to RAM faster.
You would need an accelerated MMU,GLUE and RAM for this.
In any case it's good to know it can take 16MHz.

[Icky]

I had a thought last night. Yes before anyone comments it did hurt

With all the talk around the DMA and buffering it too and the asks to add in an internal header could we not have yet another piggyback board for the STF on the DMA to buffer and provide the header out.

Haven’t had chance to check the schematics yet to see if it would be a simple plug in board.

And if so would this not be a good addition to the original STFM as well?

[exxos]

I haven't given it much thought about the DMA buffer yet. It's not really critical for the STF. I have half wondered about doing a buffer kit for the STFM etc but the problem is it also me getting sidetracked into other projects again Not saying its a bad idea, probably a good idea to have patches for the STFM, and may do that at some point in the future..

[troed]

And if so would this not be a good addition to the original STFM as well?

This is what I've done for internal floppy on the 520ST(M) - thanks to it having a nice row of holes for filters (or zero-ohm links) that can be used. Not all motherboard revisions have that.

floppy_schematic.png (213.72 KiB) Viewed 4062 times

(I've had three of these made)

[Icky]

And if so would this not be a good addition to the original STFM as well?

This is what I've done for internal floppy on the 520ST(M) - thanks to it having a nice row of holes for filters (or zero-ohm links) that can be used. Not all motherboard revisions have that.

(I've had three of these made)

Clever idea to use the filter holes.

[exxos]

Just taking a look the STE DMA circuit again...

Basically we have the LS245 data bus isolation.. And the LS244 or some other signal buffering..

As mentioned before, the LS74 would not be used in this case.. So basically we just have 2 chips to deal with.

I will see if I make a start with the schematic for this...

dma.jpg (42.88 KiB) Viewed 4022 times

[exxos]

I've started the schematic..

It is actually going to be rather problematic because of the 244 buffer... What we basically end up doing is breaking some connections between the DMA & 1772, which would also be buffered by the LS244 , where it should not actually be.. This may or may not cause issues, but I think it is better not to risk it... So if this board went ahead, I think leaving off the 244 buffer for now is a better idea.. Otherwise to link all this properly is going to involve a lot of track cutting and rewiring.. I don't think it's worth the trouble those signals..

Also for example, we don't have the reset signal from the DMA pins, so that signal would have to be cut to the DMA port connector and routed via the 244 buffer.. I have not checked the motherboard layout, but likely would end up in more patchwork to reroute the signal.

Of course this would all be done properly on the next revision motherboard, but all we can do on the current design is add the DMA dataline buffers.. That is mostly the main "upgrade" anyway..

Am not really sure it is even worth doing a patch board for the current motherboard.. As we have the 1772 patch board on the way which solves bus conflicts... The only potential issue is as said before, if the hard drive ( likely only "brick" drives) were slow to release the databus... It could cause a bus conflict...Of course this is only basically speculation at this point. Likely ICD "the link" interfaces probably be okay because it is not using all sorts of separate logic chips.

So at this point, I say it is not really worth the trouble , especially considering it would need a huge hack job to actually adding the 244 buffer.. A lot more trouble than it's worth basically, at least on the current motherboard.

dmap.jpg (70.02 KiB) Viewed 4017 times

[rpineau]

I agree that for this revision the 244 is not "needed".
The main fix is really the buffering of the data line to avoid conflict on the bus.