Falcon clock patch V4 series

[kohli79]

I don't know if it's helpful. I've attached a picture of my CT63 Falcon. It has an Atari/factory clock patch installed. I can't remember which one exactly. In any case, the Falcon runs in 030/060 mode without any problems. I hope that everything will be OK again soon!

Regards, Uwe

CT63.png (737.94 KiB) Viewed 862 times

[exxos]

Thanks. Looks like it's just the "classic" "inverter style variants. That is basically what most generic patches use, including my V2 and V3 series. Though the V3 uses better methods for driving the SDMA clock and has optional SDMA delays.

Without the termination resistors, signals will be a lot worse, but likely slightly more in phase. But even so, the inverter will skew the signals a lot anyway, that I can't see resistors mattering in that case.

But there lies the problem that I have used various clock patches on my Falcon and my CT60e has been fine anyway.

If @Swissfalcon tries the V3, He can basically have the same by bridging the series resistors. It might alter timings enough to work. But even so, its a inverter which czuba says to use none. So as to why more people haven't come across this odd problem is really strange.

[Swissfalcon]

Hi, I'm back after some days off with new information.

As a reminder, here's where I was the last time I wrote:

This is the signal on the SDMA before any patch applied:

witouth patch (Atari V0 mod).jpeg (588.72 KiB) Viewed 770 times

With this configuration. Joy demo crashed a lot of time (maybe 50%), and Falcamp hanged eventually after some minutes of playtime but it was sometimes possible to play music for more than 1 hour sometime. MX player worked forever (more than 48h). Everything else that I tested was stable.


Then, I installed the V4.
This is the signal on the Expansion bus, where the CT60e take the clock, with the V4 installed:

without CT60 on the bus my falcon.jpeg (669.82 KiB) Viewed 770 times

And this is the signal on the SDMA (after the resistor) with the V4 installed:

patch V4 after resistor.jpeg (224.37 KiB) Viewed 770 times

And this is both signal superposed with the V4 installed, we see that they aren't in synch:

désynchronisation V4 après résistance.jpeg (647.27 KiB) Viewed 770 times

With This configuration, Joy worked 100%, but Falcamp crashed almost immediately, and MXplayer crashed after maximum 30 minutes. The system overall was much less stable than before.

Then, I installed the V3 patch. here the results:
Here the signal on the CT60e bus clock pin with V3 installed

V3 CT60e.jpeg (605.58 KiB) Viewed 770 times

Here the signal on the SDMA with the V3 installed:

V3 SDMA.jpeg (237.35 KiB) Viewed 770 times

And here both signal:

V3 both.jpeg (661.33 KiB) Viewed 770 times

Not perfect, but much better in synch than the V4.

In this configuration, Joy demo worked 100%, Falcamp player for hours and MX player too. The system was more stable overall. But I had a very strange new problem: MINT did 2 times a reboot by itself. without me doing something. It simply decided to reboot.

Anyway, the overall stability is much better, so it's look that the main problem in my specific case is the desynchronising between the expansion bus signal and the SDMA signal.

Here a summary of all measurement:

V4 vs V3.jpeg (332.58 KiB) Viewed 770 times

IMG_7138.jpeg (431.84 KiB) Viewed 770 times

IMG_7137.jpeg (328.79 KiB) Viewed 770 times

[exxos]

Great research well done!

What I can tell then, something as small as 1ns might be enough to upset the CT60e Though I still don't fully understand why my CT60e doesn't exhibit those problems.

Still not sure if it may be worth you lending me your CT60 to see if it works in my Falcon or not. It may possibly help identify if the problems are more with the CT60 itself, or your Falcon. Also if the CT60 fails, I can maybe experiment with the clocks before doing anything else.

The majority of the parts for the V5 patch arrived in stock yesterday. I'll try and look at it again next week. It will likely take a few weeks to get the board manufactured and for me to test them and evaluate signals are in sync better. Once I am happy with it, I can send you one to try in your Falcon.

[exxos]

Progress..

V5.jpg (60.43 KiB) Viewed 653 times

[leonk]

In the span of 2 months, this thread jumped 10 pages!

Finally got some free time to play around with my Falcon; install software, and test things out. Unfortunately, I think I ran into some issues.

System: Stock NTSC 68030 with no accelerator. Using MeanWell replacement PSU, Exxos NVRAM, Exxos 16MB RAM upgrade, and Atari's clock patch has been replaced by Exxos 4.2A patch. No FPU in this system. Games work and sound great, regular TOS as well as NeoDesk/Geneva work great (I use XBOOT to swap between the 2) I've posted pics of the clean clock output I see at the SDMA chip somewhere around page 3 here.

ISSUE: 2019's Signals music demo has crackling in the output. Aniplay and FalcAMP have issues playing 128kbps MP3 files (either crackling sound or crash with DSP error). DMA2DSP.PRG shows a green bar across the screen, with red background and small red spikes on the bottom. SDMATEST at 50 KHz with 4 tracks also has crackling in the output.

I think it's clear that the SDMA chip is not a happy camper. Before I rip the 4.2A clock patch and put Atari's patch back for baseline testing, any suggestions on what else I can try beyond the install docs as posted here?

[exxos]

@leonk I cannot assertain if you took a scope reading on X10 on the sdma pin after removing the 47pf cap, there only seems to be ones with it ?

But as your logic low is around 0.3v it may be a indication of the DC offset issue I fixed on the previous gen patch.. but on the V4, as the voltage swings were higher, there didn't seem to be any issues in doing things differently on the V4.

If your still not seeing a proper logic low, then you will have to remove the resistor on the sdma chip end, replace with a 100nf cap ( I'd need to check the actual value when I get home) then have a 10k resistor as shown in this image..
https://exxosforum.co.uk/atari/last/fal ... .htm#V2MOD
But note it doesn't show the cap on that image...

The signals on the original patch are likely worse anyway. So we are back to the clock sync problem again. There is something odd going on as some falcons seems to tolerate almost anything and others seem ultra picky. Though the CT60 adds another level of issues which seems to be clock sync related.

I was trying to avoid manufacturing multiple clock patches but if every falcon is behaving differently, then there may simpley be no way to solve all the issues with a universal patch

[leonk]

Interesting @exxos . So what you're suggesting is to undo the SDMA V4 patch (64/68 ohm resistor on the "bootleg" track and the 47pF capacitor to ground) and apply the V2/V3 patch in this area as well?

What about the 100pF ceramic cap under the FPU? I don't have an FPU installed and don't plan on adding one but that cap has been installed as well.

[exxos]

The 62R on the bootleg track gets changed for the capacitor. Then you have the 10k resistor on the right side of the cap to 0v.

Please make sure you measure the high and low voltages on the sdma pin before and after the mod an repeat your tests. Whereas you saw around 300mV low voltage, after the mod you *should* see a better low voltage..

Leave the 100pf in place for now as it's not good to change many things at once. We need to rule things out one at a time ..

[leonk]

Here's how the signal looks like at the SDMA chip after the standard V4 install: (using 10X probe settings)

IMG_8036.JPG (325.52 KiB) Viewed 517 times

I changed the install at the SDMA to have a 100nF capacitor inline and a 10K resistor going to 0V:

IMG_8038.jpg (380.96 KiB) Viewed 517 times

And now this is how it looks like now: (note - this pretty much looks identical with or without the 10K resistor!)

IMG_8039.JPG (366.03 KiB) Viewed 517 times

SDMATEST now passes, and I get a solid green bar with white background on DMA2DSP. I think the thing that did it is the 100nF inline cap and not the 10K resistor. MP3 files play now on Aniplay and FalcAMP. The only complaint I have is that the audio output is still noticeably scratchy on the MP3 playback as well as in demos (both on the SCART/RGB output on the TV as well as when I plug in headphones directly into the computer in the back). I don't have this noise in game music.