Boosting STFM: 16MHz bus, 64MHz Shifter

[troed]

Well it's at least closer, but will still break some demos when in "stock" mode. I'm a purist so I would really want it to be 100% compatible

On another note, I had some slight hopes that the image problems that appear in doubleST high (1280*400) after a few minutes would be down to power regulation. I had recapped the PSU but the 4700uf on the motherboard was still 30 years old. However, it made no difference after having switched it out.

The Shifter does get hot to the touch and it's the mode where the pixel clock really runs at 64MHz so next up to try is to add a heatsink .. but in worst case it could be either unsolvable or specific Shifter IC dependent.

1280*200*4, where the pixel clock is 32MHz, has no issues.

[exxos]

Poor shifter Have you tried the IMP ones ? one brand may work better than the other..

[troed]

Oh, and I made a first implementation of a program to stabilize the Shifter (wraparound, shifted bitplanes etc problems) at boot. It's not perfect, but putting it in AUTO increases the chances of having a correct display once reaching desktop ..

[troed]

Poor shifter Have you tried the IMP ones ? one brand may work better than the other..

Yeah I have a few different Shifters, but will need to socket it first.

[troed]

Ah, some good news! Put my Falcon (!) 45x45mm heatsink (the only one I had, it's low profile so average cooling capabilities I guess) with some standard cooling paste on top of the Shifter and the glitches in 64MHz pixel clock (1280x400 mono resolution) after a few minutes are now gone completely.

(this is what they looked like before, and got progressively worse)

64mhz_pixel_clock_glitches.jpg (99.73 KiB) Viewed 7464 times

There's still some instability with jail bars coming and going if the image displayed is "complex" (like, the striped GEM desktop) but they disappear if the opened program has large blank areas (empty text editor). This might be solvable, don't know yet.

The heatsink gets warm to the touch - I will actually add a fan as well and see if there's further effect to be gotten.

I still have spontaneous reboots, which seem to correlate with floppy or ACSI device access. Will now check to see if they're only in the doubleST mode or also the Stefan-mode (can't see why the higher clocked Shifter would cause reboots, I'd be more inclined to blame CPU/MMU). I briefly checked how hot the 68000P16 gets and it's indeed _very_ hot to the touch, much hotter than the Shifter ... might need to heatsink that as well.

[exxos]

You could move to a HC CPU.. those run cold.. you likely need to change the SIL pull ups on the CPU either way...

See bottom of this page... you probably seen it before...

http://www.exxosforum.co.uk/atari/last/V ... /index.htm

[troed]

Yeah all pullups are changed already on this STFM. I don't have any DIP to PLCC adapters laying around but I have a 68HC.

Will test with heatsink first, and also see if I can figure out what causes the reboots. Could still be power rail I guess. I measured 5V to 5.08, maybe I should increase it a bit ..

[exxos]

You can get DIP HC types also

It could simply be your "wiring" which is causing the issues. I would check the voltage between the PSU 0v and the GAL 0v and see what noise you have between 0v points.. make sure you not got "ground bounce" issues.. its one of the main issues I have with my boosters

[troed]

Regarding the resets I'm starting to doubt it's HW. It's always when some file access happens and it's always an immediate reset - and it happens in all three modes. Never bombs or more random behaviour. I re-use the same HDDriver 10.10 and Unicorn+Bigdos combo as I have on other computers and maybe there's an incompatibility with TOS 1.04 that I don't see on other computers (that mostly run 2.06).

Regarding the flakey Shifter output in 64MHz pixel clock (1280*400 mono) - it's not just having to use a heatsink on the Shifter. As we learnt way back when overclocking PCs it's also about the voltage Since I have the SR98 PSU in this machine I can easily change the output with the pot:

- 5.08V a bit flakey
- 5.12V (with floppy motor running) very little
-- floppy motor off almost nothing (5.13V)
- 5.16V with floppy motor on, nothing

(nothing = rock solid picture, no glitches)

Again all of this is with the one Shifter IC I've used so far so we cannot assume this holds for everyone. Also, it's of course only in this specific mode (1280*400) - but it's a really nice video mode so good to get stable

Also, the software fix for shifted bitplanes/wraparound etc is getting better. It mostly fixes everything by just putting it in AUTO, but if it doesn't it can just be launched one or more times from desktop manually until the image is good.

Also, I've inverted the MMUCLK in the Shifter expression for "Stefan-mode" in the GAL. This sort-of moves the other clocks falling edge to Shifter rise, like in the other modes. It seems to have lessened Shifter bitplane errors, at least it hasn't gotten worse.

Still to do: Solve the bootup issues. Sometimes I need to power cycle ten times to boot. Also, there's a "wakeup" in the boosted modes where floppy access results in empty disk - stable over resets. I believe this to be a specific CPU/GLUE wakeup and so should be fixed if I can get some sort of RESET-init state to work.

edit: GAL code now under source control at Github: https://github.com/troed/DoubleST

[exxos]

Interesting!

At some point I was running the shifter on its own 5V regulator (driven from the 12V line).. The shifter doesn't exactly get stable voltage, even if you add 4,700uF across it!

So it may be worth trying a 5V regulator on the shifter ( break its 5V feed of course) and see if its really voltage related or noise related.