DFB1X with Nemesis / clockpatch

[krupkaj]

I have just got the new DFB1X. Fortunately I have catched the last posting to EU :)

I have 2 Falcons. One with Nemesis + its clockpatch and second with Phantom + its clockpatch. First I tried the DFB1X with Nemesised Falcon and it did not work well. It just displayed Atari logo and that was all. So I installed the DFB1X to the second one and it looks like it is running OK. My question is did anybody tried it with Nemesis clockpatch? Or the problem is somewhere else? Both Falcons are running fine without the new booster.

[Steve]

I beta tested the DFB1 with Badwolf when it was being developed. We could not get it working with my Phantom bus overclock, even with different clock patches. It would work in standard 16mhz bus, but as soon as overclock was enabled it would crash. It is interesting that it works on your Nemesis ... they're basically the same thing, but different kinds of clock patch.

[krupkaj]

good to know. Maybe I wrote it wrong. I originally wanted to try it with accelerated bus, but it didn't happen. I tried just my 2 Falcons with different clockpatch. It does not work with Nemesis clockpatch at all, but it runs with Phantom clock patch.

Another question is, does it work with any other bus accelerator?

[Steve]

There are only Nemesis and Phantom. As far as I know they both operate in the same way (the overclocking part) the only difference is the nemesis has a different clock patch to the phantom. The phantom clock patch was famous for being the one which worked even with CT60. So far I have tested DFB1 with Phantom, it wouldn't work in overclock mode. You have a nemesis .. so between us we have tested them all. Just to clarify, are you saying you tested overclock mode on both or neither?

[krupkaj]

Not yet, just standard unaccelerated board. The accelerators are installed, but the DFB1X did not work with the Nemesised Falcon at all, I have just install it to the Falcon with Phantom.

I want to show DFB1X to Ctirad (Phantom author) and discuss it after the holidays.

[exxos]

The DFB1/X assumes a stock bus speed. Faster bus speeds is not something myself or BW have even considered, let alone tried. It would likely need a overhaul of the firmware to work at different bus speeds. But considering how problematical bus speeders are, I doubt either of us are going to give ourselves the headaches in all that stuff.

As for clock patches, My V2,V3,V4 patches work fine with the DFB1X.

Other Patches may work, but used "at your own risk" and if you try to use them, then you are on your own. Other patches may yield unpredictable results and have certainly not been tested.

[Badwolf]

(I originally responded to this on another thread by mistake)

I'm intrigued as to what the problem is. Not having got around to repairing my second Falcon motherboard, I've yet to experiment with an accelerated system clock (I'm not brave enough to do this on my original bird).

Conceptually I can't think of an obvious reason it shouldn't' work, but then I don't know the details of these bus speeders.

The DFB1(X) clock switching logic has problems when the clock is too slow* (special handling goes into the DFB1X firmware to look for the 8<-->16 MHz clock speed switch and disable acceleration when it's at 8, but it's not perfect).

Also DFB1(X) take their clock from the expansion branch of the clock fan-out. Do these patches desynchronise these somehow?

If the booster doesn't work in no-acceleration mode (one of the jumpers set on DFB1, I think this is carried through to DFB1X), then it's possibly not logic, but physical. The clock signal going into DFB1 is extremely sensitive to being overloaded.

This would probably be quite an interesting investigation for anyone so minded and suitably equipped (Steve?), but not one I have any time for, I'm afraid.

BW

* I abuse a related clock switching circuit for two completely unrelated clocks. This is because in general it switches faster than the proper circuit and is 'good enough' in most cases, but this does introduce up to one half cycle which is at half the speed of the slowest clock. At 16MHz, this means there's potentially one 8MHz half cycle in there. At 8MHz, this is up to one 4MHz half cycle. I theorise my problems with 8MHz mode is that this 4MHz half cycle is too slow for the 030's dynamic core. I could be barking up the wrong tree, but it fits the observed behaviour.