DMA port "update" ideas..
Posted: 14 Oct 2018 22:59
I started a seperate thread for this...
For those not aware... https://www.exxosforum.co.uk/forum/viewt ... 500#p11552
We finally solved the "idea" behind the STE DMA buffers. Basically the 1772 does not "let go" of the bus until , up to 150ns later...
The Odd thing is the STFM really should have never worked with a hard drive attached. It is possible as the 1772 delay is from 20-150ns, that maybe some batches of 1772 worked, when others did not due to a higher delay...
If I remember rightly, Atari hard drives did not suffer from issues, but 3rd party drives sometimes did.. What we do know is the Atari hard drives had buffers on the inputs... and a fair amount of logic on the buffer enable line... of course on the STFM, if the buffers did not isolate the DMA, the bus would conflict.. So this could explain those 3rd party drives not working if they didn't add the "delay buffers".. All food for thought I guess....
Looking at the STE circuit...
The DMA port gets "delayed" by one clock cycle after FDCS goes high. This makes sense as the 1772 can hold the bus for that time. So if the DMA port was active, it could actually short out the 1772 bus to the DMA bus (same bus different DMA devices).
What I don't like is the "delay" on the DMA port which could attribute to DMA issues on the STE..
So my thought here is a patch board on the 1772 to add in a 74LS245 bus transceiver to isolate the 1772 once the DMA de-selects the 1772 via FDCS. This way, there is no bus conflict and no delay on the DMA port. It would also not matter the "delay" (20-150ns) of the 1772 as it would be isolated automatically.
As a side thought, it may still be a idea to add buffers on the DMA port itself (like the STE does but without delay).. Though I am still in 2 minds if its really needed or not yet... Really they would be there to protect the DMA chip more than anything.. cheaper to replace LS chips than the DMA... So I am leaning towards adding the buffers in later revisions...
For those not aware... https://www.exxosforum.co.uk/forum/viewt ... 500#p11552
We finally solved the "idea" behind the STE DMA buffers. Basically the 1772 does not "let go" of the bus until , up to 150ns later...
The Odd thing is the STFM really should have never worked with a hard drive attached. It is possible as the 1772 delay is from 20-150ns, that maybe some batches of 1772 worked, when others did not due to a higher delay...
If I remember rightly, Atari hard drives did not suffer from issues, but 3rd party drives sometimes did.. What we do know is the Atari hard drives had buffers on the inputs... and a fair amount of logic on the buffer enable line... of course on the STFM, if the buffers did not isolate the DMA, the bus would conflict.. So this could explain those 3rd party drives not working if they didn't add the "delay buffers".. All food for thought I guess....
Looking at the STE circuit...
The DMA port gets "delayed" by one clock cycle after FDCS goes high. This makes sense as the 1772 can hold the bus for that time. So if the DMA port was active, it could actually short out the 1772 bus to the DMA bus (same bus different DMA devices).
What I don't like is the "delay" on the DMA port which could attribute to DMA issues on the STE..
So my thought here is a patch board on the 1772 to add in a 74LS245 bus transceiver to isolate the 1772 once the DMA de-selects the 1772 via FDCS. This way, there is no bus conflict and no delay on the DMA port. It would also not matter the "delay" (20-150ns) of the 1772 as it would be isolated automatically.
As a side thought, it may still be a idea to add buffers on the DMA port itself (like the STE does but without delay).. Though I am still in 2 minds if its really needed or not yet... Really they would be there to protect the DMA chip more than anything.. cheaper to replace LS chips than the DMA... So I am leaning towards adding the buffers in later revisions...