Been working on yet another Falcon clock patch because of issues in this thread
viewtopic.php?p=130589#p130589
Either has to be noise/ringing related, or timing issue. Though the 100pF cap which improves things actually does both !
I did a new version 6 patch but not really sure is worth even producing at this point. Without a machine to actually diagnose issues I am just guessing at the problems
V6 has inverting and non-inverting options. Plus a 10ns delay which can skew the SDMA clock in either inverting or non-inverting mode. I also think there is probably too much variation in tolerances and issues across Falcons to use the existing clock traces.
So I would advocate that the FPU gets driven from its own clock (or not at all if a FPU isn't needed). Then a new wire to the DMA much like the Phantom clock patch did. This way it bypasses the whole bad clock trace and fully isolates the SDMA from FPU clock. The SDMA will need the 10K pulldown much like the V3 had as its helps with DC-offset issues across the clock. So basically its a amalgamation of every other clock patch design all rolled into one !
- Capture.PNG (44.49 KiB) Viewed 1112 times
Like I said previously, it feels like we are just all "dancing around" ground bounces on various parts of the board or something like that. Figuring out issues on Falcons is almost impossible because of the sheer expense and lack of boards I have here. It's not like working on STFMs where I can try on 100 different motherboards !
The version 5 clock patch which had the phase locked loop has been working fine in my Falcon since whenever I created it months ago. But I don't really see the point in that series either because again it only gets tested in my own Falcon machines to accept pretty much anything I throw at it anyway.
I don't know if there were different PCB companies making the same board or not ? Maybe the layer stack order is different on some boards than others causing variations ?
I was asking around if someone could loan me a rev J board as they seem more problematic but its difficult to say if board revisions are a issue or not. The only other option is if someone has a board which simply fails the SDMA related tests, which doesn't work with any other general clock patch out there already, that could be loaned to me to run tests on and compare against my own falcon here.
I would like to get to the bottom of these weird issues once and for all, but without any amount of Falcon motherboards to test and diagnose, I simply don't have a large enough sample set to draw any sort of conclusions.
I mean how can my Falcon work fine with any clock patch I threw at it and yet other people's boards seem to fail miserably.. It makes no sense.. Obviously this has been an issue since day one because there are so many variants of the clock patch out there already. There must be some weird difference between batches of motherboards or revisions, or maybe there are different batch numbers or revisions of the chipsets causing problems. Who the hell knows...
The diagonal one stops the cross brace from springing off by itself as well.
