Falcon clock patch V4 series

[exxos]

@Rustynutt are you trying V4 on a stock machine first ?

@Steve reported issues when using phantom and went back to the V3. V4 may need tweaking for faster bus speeds but i'm not sure its worth the trouble anymore.

Though the V4 on a stock machine still remains somewhat untested. I have one running in my Falcon but not done much testing yet. Just never have time :(

[Rustynutt]

@Rustynutt are you trying V4 on a stock machine first ?

@Steve reported issues when using phantom and went back to the V3. V4 may need tweaking for faster bus speeds but i'm not sure its worth the trouble anymore.

Though the V4 on a stock machine still remains somewhat untested. I have one running in my Falcon but not done much testing yet. Just never have time :(

Time. And pain level :( I understand this, and you are far more committed to projects for sure!

Yes, will start with accelerated bus, no processor boards (CT, AB).

40MHz COMBEL will be the starting point.

End of simple reply :)

At 40MHz, we are still at a speed where the Falcon will boot without the TOS edit below.

Above 40MHz, on a stock Falcon with accelerated bus (no additional expansion processors, Phantom or Nemesis hardware), when TOS 4.04 blits the Atari LOGO to screen, results in a crash.

I use a modified TOS image in ROM.

The Falcon bus control register allows setting the Blitter to run 1/2 of the CPU clock.

In the test Falcon, the Blitter is set to run at 1/2 CPU clock. This is set before TOS blits the Atari LOGO. This is how the test Falcon can boot with upto a 62MHz COMBEL clock. At 1/2 CPU clock, that's 15.5MHz BliTTer, right where the standard 1:1 setting would be with a 32MHz COMBEL clock.

A really nice solution to the above is to burn EmuTos into ROM. EmuTos doesn't use the Blitter during boot to display it's logo, so on a stock Falcon with accelerated bus, EmuTos works very well.

The EmuTos team, from what I've read, utilized some of Didier's CT TOS, a part where he ripped out use of the Blitter and included alternate routines.

It also has the benefit of bypassing the Falcon RTC at boot.
********
Note: The CT60's boot with a stored CT TOS, not read from the Falcon TOS ROM.
So when the CT is enabled, the modified TOS ROM has no effect on the system.
The CT doesn't use the Blitter at boot anyway.

As far as the Afterburner and EmuTos goes, it is not fully supported. Given the handful of actual users, Roger said he'd have to actually have a unit to test what is going on with the AB. Same story with the Sparrow. He'd have to have a machine to include support.

So I'm basically "stuck" using TOS 4.04 with the edit for testing how the bus is accelerated on this machine considering the Blitter issue overclocked, and no full EmuTos support on the AB040.


*******

Looking at the results achieved with the buffers clock outputs you have, think it well worth the investigation on my part :)


Have another theory about the effect of CPU's on the expansion bus (Afterburner, CT6X's, and the DFB.

So think making sure the "stock" accelerated bus Falcon works with the V4 and V2 DMA board a good solid starting point.

I understand the first point on the V2 board, "0", is actually about 5ns, from reading your article.

Was following your post about purchasing a logic analyzer. I'm still in the market, and have my choice narrowed down. Just need the wife's final approval :)

That will help exchanging information, once I have one setup and understand the basics. My scope experience is pretty limited to old Textronic 1960-70's units.

[exxos]

@Rustynutt I don't think a LA will help unless you get a really expensive one. It seems like the SDMA clock is very sensitive to timing in the order of like 1ns. I doubt many LAs will have the bandwidth to keep up. Also a LA doesn't tell you the more analogue nature of those MHz clocks. IMHO you be better with a 4 channel scope.

When I was talking about a LA for booster work, I'm dealing with signals around 2mhz. I really need a 4 channel scope but I can't justify the cost for the amount it would get used.

[Rustynutt]

Thanks for that info.

[leonk]

For someone who has a Falcon with Atari's patch from back in the day and is considering going to a better clock patch (not planning on using accelerators or any add-on cards) is there a big difference between V2/3 and V4? I'd like to get something installed once and be done with it. I'm comfortable with soldering (do a lot of micro-soldering of FPGAs, DRAM, 0603 and smaller SMDs, etc under a microscope) so that's not a concern.

Just wondering what's the recommended solution for me assuming staying with what Atari put it might not be the best option.

[exxos]

V2/3 will likely be fine for the vast number of users. But there are falcons with even worse problems which is why the V4 was designed.It addresses a lot more problems which the previous gen patches do not address. As to what falcons have what problems exactly, there's no way to tell.

[leonk]

Thanks for the prompt reply. Is it safe to go directly to V4 or do you suggest going to V2/V3 and only if I notice issues, consider V4?

I need to see the upgrade instructions; want to make sure that "undoing" Atari's fix is pretty straight forward.

[Steve]

Thanks for the prompt reply. Is it safe to go directly to V4 or do you suggest going to V2/V3 and only if I notice issues, consider V4?

I need to see the upgrade instructions; want to make sure that "undoing" Atari's fix is pretty straight forward.

Leonk if you have a stock falcon, ie; no bus overclocking, the V4 is the best you can get IMO. When I did oscilloscope tests on the V4 the clock signal it produced was extremely clean. But it did not like my bus overclock. V2/3 has a delay option which allows bus overclock support.

So in my opinion: for standard Falcons: V4 is the 'best' ie; provides the most perfect clock signal. For bus overclocked Falcons: V2/3. (FYI please read my last comment on this, as I probably shouldn't jump to any conclusions)

Here are screenshots of my measurements for V2/3 on the SDMA clock:

v3sdma.jpg

And here are my measurements for the V4 on the SDMA clock, as you can see the waveform is much cleaner:

V4sdma.jpg

One thing I want to mention is not every Falcon is the same, different Falcons can produce different quality clock signals. So my results might look different to others. But this is my experience testing both. Also, does having a nicer clock signal mean a more stable machine? Technically it should, but like Exxos said, it hasn't had much testing with users so far. So results may vary.

[leonk]

The Falcon clock patch V4 came in today and I proceeded to install it. I don't have an FPU in this Falcon, and it's pretty much stock. Here's the before (measured at R221)

IMG_7626.JPG

Here's the install (I also did the ceramic cap under the FPU and cap/resistor at the SDMA)

IMG_7631.jpg

Lastly, here's the same clock freq after the mod at R221. Overall, it seems to be pretty much identical (unless I'm measuring the wrong thing) I suspect I just got lucky and the clock fix from Atari was "good enough".

IMG_7633.JPG

[jwd]

I think the other ends of the run would need to be probed to compare properly, for example (as documented elsewhere) the SDMA signal once it gets to the SDMA IC is/can be very different from the signal that leaves the clock splitter circuit/clock patch.

That was my main issue I had an original atari clock patch + a simple CPU speeder (powerup2 which immensely cleaned the clock patch to the CPU), no FPU so didn't care about that and that setup got me through Uni, I would see some crashes in Falcon games (Steel Talons) and demos but just put it down to buggy software at the time. Fast forward a few decades more knowledge and Exxos's excellent coverage (+mikros clock patch web page) things are clearer and especially one demo (JOY) if that fails/crashes then even if you have a clock patch you need a better one.

In my case (seen with Osc) my CPU clock was nice and clean (through powerup2 speeder) but my SDMA signal was shocking even with original clock patch. So I got a phantom clock patch off @Steve at the original CyberLegends event. This massively cleaned up the signal to the SDMA and as final test JOY demo worked fine again and again.

What I will say is that even though the phantom cleaned it up (SDMA) nicely there was still some 'artefacts' present in the signal (compared to the super clean output from the powerup2) and if I wanted to put a further speeder (DBFX or Nemesis etc.) then I would probably need a cleaner signal to SDMA as it would all be put under tighter tolerances and therefore I will probably need (in my case) the V4 patch.

Saying that some falcons with no clock patch whatsoever can run speeders/accelerators and the JOY demo absolutely fine, each Falcon must be considered unique and what works for one Falcon might not work for another. Another example is the capacitor in the video circuit that is the wrong way round, voltages need to be measured and there are variances in the circuit on the board that aren't in the schematics, mine is correctly orientated from factory but vast majority aren't again need to measure and check your own bird.