4MB Upgrade & 16MHz Booster progress

[stimpy]

Basically the round up is I have the CPU running ASYNC finally and had it running at 25MHz. I couldn't go any higher (didn't have anymore osc's to hand) but the CPU E-clock running faster causes the ACIA to screw up. After a lot of work, I gave up and decided to do a new test board using the 68SEC000.

Seem to remember when looking at the datasheets it was possible to ignore the 6800 bus and basically drive the 6800 signals in sync with the 68000 bus and assert DTACK as usual. It's also used for autovector interruptrs or something (VMA?) but I think you can do a similar thing by placing the vector level on the bus and handling it differently. Sounds like a few lines of code and must have been done to get 020+ CPUs to work as they don't support 6800 bus.

Maybe forget the fast ROM and just load ROM into fast ram on startup. Isn#t there auto boot prgs to do that?

Sounds like you need to move to SDRAM to keep up with the CPU...

[exxos]

Seem to remember when looking at the datasheets it was possible to ignore the 6800 bus and basically drive the 6800 signals in sync with the 68000 bus and assert DTACK as usual. It's also used for autovector interruptrs or something (VMA?) but I think you can do a similar thing by placing the vector level on the bus and handling it differently. Sounds like a few lines of code and must have been done to get 020+ CPUs to work as they don't support 6800 bus.

Autovector is the problem, the 020 has the AVEC pin to emulate the 68xx cycles, but its not so easy on the 68000. However the 68SEC000 does have the AVEC pin and is wired much like how the 020 is done. That emulates and uses DTACK as you mention. That is proven stuff on the 020 so shouldn't take to long to adapt it for the SEC CPU. That SEC CPU claims up to 80MHz speeds, I maxed out the HC at 38MHz, so I need to move over to that SEC CPU anyway. So I have abandoned the 68HC000 stuff now.

Maybe forget the fast ROM and just load ROM into fast ram on startup. Isn#t there auto boot prgs to do that?

ROM is running at 32MHz anyway, the ROM decoding logic in the STE is pretty fast and with a 55nS ROM, it *just* manages to keep up before the CPU reads DTACK on S4. So loading into fast-ram (32MHz) isn't going to be any faster.

Sounds like you need to move to SDRAM to keep up with the CPU...

SRAM seems to be about 45ns, I am hoping to ramp up to 64MHz with the SEC CPU, but Likely after somewhere around 50MHz will have to start using waitstates, unless faster SRAM can be found. Though currently I am working at 32MHz for fast-ram as its a sensible point to work to now the HC stuff is proven to work up to those speeds.

As a side note, I may prototype my SRAM board with a adapter to fit my STE V1 booster. Then at least people can add it to the current series of booster, however I am not sure on the height between the booster and the shielding , so if anyone can look into that it would help me out.

[exxos]

Small update, I have ordered the below PCB. It will adapt my dev-board to the SEC CPU. Will need some bodging to my PLD, but I thought its better to actually try this CPU's top speed before investing £100's in new booster PCB's.

The SEC has to run at 50MHz or better to make it worthwhile doing a new booster series based on it. I want to hit 64MHz, but if it runs faster then any speed can be used. Problem is at some point the ROM and SRAM speed will be maxed out. The CPU could still run faster with internal instructions.. But will have to do a lot of testing and research before making any decisions.

I want to max out the 68000 design as much as possible, and as said before, once finished, I will work on speeding up ST-RAM again. Likely will be in a years time as have so much work to do on various Atari related stuff. Though once ST-RAM and the CPU is boosted as much as possible, I think that will mark the final end of the 68000 series boosters.

While I would like to do a 020 and 030 board, I really don't want to spend the rest of my life only working on boosters. I am working with others in a 020 and 030 based design (possible even 040), so while I may take a step back from this work, it doesn't mark the end of boosters being created and produced.

SECADAPTER.jpg

[troed]

BTW - just to clarify this mod is working fine on my STFM now I think if the master 32mhz osc is replaced with a 64mhz one, might need buffering, then of course 16mhz CPU and 70ns RAM. Then downclock the MFP & GLUE back to stock speeds. Then this should allow the new video modes and 200% speed gains.

For those interested in the "Stefan 16MHz mod" and increased horisontal resolutions, I'm now trying to build this up in one of my 1040 STFMs as well. Documenting it here: https://www.exxoshost.co.uk/forum/viewt ... f=18&t=283

(Currently I have issues with the floppy)

[mrbombermillzy]

Thats great news!

Ive been hoping someone would pick up where Chris left the double pixel clock observations with his booster tests. I hope you get the floppy drive stable at some point.

[troed]

Thats great news!

Ive been hoping someone would pick up where Chris left the double pixel clock observations with his booster tests. I hope you get the floppy drive stable at some point.

Floppy stable and my STFM can now do 1280*400 in monochrome and 640*200 in 16 colors

/Troed

[joska]

Documenting it here: https://www.exxoshost.co.uk/forum/viewt ... f=18&t=283

Would you mind posting a summary here when this is finished?

[troed]

Documenting it here: https://www.exxoshost.co.uk/forum/viewt ... f=18&t=283

Would you mind posting a summary here when this is finished?

I've now published a long and prosaic writeup here: https://blog.troed.se/projects/atari-st ... deo-modes/

/Troed

[joska]

Thanks

[TheNameOfTheGame]

Nice work and writeup!

[mrbombermillzy]

A good writeup there.

Troed: "Could Atari possibly have enabled these video modes in the STE, back in 1989? "
Well, Ive been studying the TT documentation lately and somewhere along the way, I remember seeing some sort of information that basically stated that the TT shifter in some form, was originally planned for the STE including a 320x200 x8bpp mode (320x400 ended up in the TT shifter), but this part was changed out at a later date.

[troed]

Thanks all. I've gotten RAM speed up to 200% as it should be now, post updated a bit. Also indicating which traces I cut etc if anyone wants to have a go at replicating this on the exact same motherboard version.

[joska]

This mod is very similar to the one I did to one of my Stacys almost 20 years ago. That was simpler though, as the LCD controller's screen output did not depend on clock speed I could just speed up everything except for the YM, WD and ACIA baud clock. The Stacy already had fast enough RAM and ROM so with very small modifications it could have ran at double speed from factory. A shame Atari didn't do that.

[joska]

Btw I wonder how this will work with a MonSTer card... Also, LaceScan would extend the resolutions even more. 800++ horisontal resolution would be possible in 16 colour.

[mrbombermillzy]

Im salivating at the thought of this new res coupled with a HW overscan mod, the twin shifter mod for 12bpp palette and beg Doug to do a custom copy of Photochrome/AGT for it...mmm

The above mods could be done relatively simply/inexpensively and it would turn the STFM into an interesting STE alternative. Maybe Exxos could sell modded boards or as a barebones kit.

Thankyou Troed! Yet another reason to stay Atari!

[arf]

Btw I wonder how this will work with a MonSTer card... Also, LaceScan would extend the resolutions even more. 800++ horisontal resolution would be possible in 16 colour.

Indeed, did anyone of the more-hardware-skill-enabled forum colleagues already try that? Would be much interested in both whether the MonSTer would work with 16 MHz board freq., and whether LaceScan could further improv the resolution above 640x200/16c.

[troed]

LaceScan will surely work. Regarding MonSTer I actually have one in my Mega ST and since I socketed the 68000 it should be simple to test.

"Simple" could still mean months ... I'm way behind on all the things I should do with the little retro time I get ...

[arf]

LaceScan will surely work.

Wow. Would not have thought that the ST’s graphic system is that stretchable.

Regarding MonSTer I actually have one in my Mega ST and since I socketed the 68000 it should be simple to test.

"Simple" could still mean months ... I'm way behind on all the things I should do with the little retro time I get ...

I know what you’re talking about …

[Maeke]

Btw I wonder how this will work with a MonSTer card... Also, LaceScan would extend the resolutions even more. 800++ horisontal resolution would be possible in 16 colour.

Indeed, did anyone of the more-hardware-skill-enabled forum colleagues already try that? Would be much interested in both whether the MonSTer would work with 16 MHz board freq., and whether LaceScan could further improv the resolution above 640x200/16c.

We already know it won't, at least the ide part. Sadly atm none of the ide adapters we have works at 16Mhz. Exxos is willing to try and develop one but he doesn't have enough time, so if someone else has the knowledge to try....

[arf]

Btw I wonder how this will work with a MonSTer card... Also, LaceScan would extend the resolutions even more. 800++ horisontal resolution would be possible in 16 colour.

Indeed, did anyone of the more-hardware-skill-enabled forum colleagues already try that? Would be much interested in both whether the MonSTer would work with 16 MHz board freq., and whether LaceScan could further improv the resolution above 640x200/16c.

We already know it won't, at least the ide part. Sadly atm none of the ide adapters we have works at 16Mhz. Exxos is willing to try and develop one but he doesn't have enough time, so if someone else has the knowledge to try....

Ah, what a pity. It has been my understanding that the MegaSTE version of the MonSTEr _does_ run with 16 MHz base frequency, though, including the IDE part?

[Maeke]

mmh dunno, but i can guaranty for having talked with exxos that it didn't work with his boosters, timing issue.

[troed]

Update on the doubleST overclock of the ST bus, MMU and Shifter: I've moved from using a 74F161 binary counter for clock generation to a GAL, which has allowed me to make it all switchable between doubleST, the original "Stefan"-mod and stock. It's not perfectly stable yet, but it's in good enough shape that I've posted the CUPL code for the GAL and logic traces on what it looks like atm.

Everything is documented in this thread: https://www.exxoshost.co.uk/forum/viewt ... 4595#p4594