exxos's Atari, Amiga & retro forum

Think improving the clock signal at the DMA CLKIN is paramount and your buffer solves this.

While a stock Falcon's timing is fairly lazy, there could be a condition where the CPU is ending TS before the DMA asserts TA.
That's going to cause errors.
Think the delay mod is shifting the TS signal timing riding on the clock. Why the delay works on some machines, and not others.
Using the CMOS logic, we are kind of tied to built in delays, so the buffer is adding ~5ns, ~10ns, etc. Surely, they fit somewhere in there, just at what point would be optimum for each Falcon, and how to adjust it on the user end.
It's just a theory after reading up on some Amiga documentation.

When you test your new boards, maybe you could look at those signals with your analyzer, that's the only way to see it.

In the Amiga docs, it deals with those 2 signals with respect to fast ram and the 040 CPU. It details how to program a GAL to actually widen the TS signal where it doesn't end before the receiving device can acknowledge transfer.
Terriblefire here would be the one to discuss this with, if he has time. From his post, sounded like he was taking a break, and out of respect didn't chime in on any of his discussions.

Noticed too, on the Sparrow, the DMA has its own internal clock independent of the COMBEL. They are near identical design of the Falcon otherwise in that area. It requires no clock patch, SCSI, floppy and SDMA work perfectly. Too bad Atari went to save a buck with the Falcon design.

Came across the V3 prototype boards while trying to tidy up my workshop a bit. So before I lost them again I quickly hooked one up.


Results are as expected (thankfully)

DELAY OFF = 7ns default delay of the IC.
DELAY ON = 14ns (2 Gate delay).
EX-DELAY = 24ns total delay (14ns + 10ns extra) .

It gives different delay steps over the fixed on/off delay of V2. Though the total delay isn't much more I think currently. Its possible it could be tweaked to add 20ns extra delay instead of 10 by adding a 220pF 0805 ontop of the one near the bottom right of the PCB. That would give 14ns+20ns = 34ns delay. I can't really remember what the requirements were now.

If @Rustynutt or @Steve wants to try the extra delays then will sort something out in getting you a board.

As a note to anyone else reading this.. V3 is only a experimental board for a small number of people to experiment with extra delays with bus speeders. It is certainly not going to replace V2 anytime soon (if at all).

Yeah I wouldn't mind spending some time to beta test this. Just out of interest do you know what the original phantom clock patch delay was? Cheers

Steve wrote: 27 Jun 2022 22:14 Just out of interest do you know what the original phantom clock patch delay was? Cheers

IIRC the delay on the V2 was there comparable to what the Phantom was doing by the "Delay on" jumper. But it was calculated as I never had original hardware and not sure if anyone even tested the actual delay.

I am closing this thread as the V2 series development has come to a end and we are talking about V3 series now. V3 I have actually abandoned in favour for the V4 series which is currently in development and prototyping. V4 will have better control over more problematic / noisy Falcons and is a total re-design from the ground up. Also V4 will have a greater "delay range" for the SDMA end of the patch for bus speeders. I will open up a relevant topic about this in the future. V2 is still perfectly fine for 99% of users.

Currently Falcon clock patch V2 still available in my web shop.

https://www.exxosforum.co.uk/atari/store2/#0073