STE V1.5 32MHz BOOSTER - STATUS

[exxos]

Been doing more 40 MHz tests. It was a little unstable :( though I have found it is the ROM which is tripping up here. I'm not really surprised as the timings were very tight even for 32 MHz. I have added tiny wait states into the ROM part of the GAL. A couple tests run just 1% slower, and now seems a lot more stable.

I think realistically 40 MHz is what I found before being the max speed it can run at. Though this does not mean it will run at this speed on every machine. Only thing I could do is create a tiny clock board with 35 and 40 MHz oscillators on it. Basically just have a jumper to select the oscillator.

I don't really want to mash on a 40 MHz oscillator directly onto the booster as it may not work on every machine. Tolerances of every part will obviously add up and vary considerably across machines. So I think a small add-on board is the only way to go. Of course people can still use the motherboards 32 MHz line if they want to play safe anyway.

Though I am trying to move away from using the motherboards 32 MHz line is running the wire across the motherboard is causing some interference and some stability issues. Of course this has been known right from the start, but I would rather try and avoid the problem altogether.

[exxos]

Did you check how stable your GND is ? did you check it against the Mainboard GND ? This could also be an issue, we had alot of GND problems with our Extension. You can check it with scope against the Mainboard ground.... (just an idea)

Yeah the gnd isn't great anywhere on the STE :( Its as good as it can be either way. it doesn't help with it not being soldered into the motherboard either.

[exxos]

I put back the motherboards 32 MHz, and that works perfectly fine as expected. I then tried the 35 MHz crystal, and that was intermittently crashing the same as I was getting with the 40 MHz crystal.

So now I have added a wait state onto /DTACK.. ROM actually runs at 266% at 32MHz.. But when we ramp up to 40 MHz, the proper speed of 309% then returns as it is compensating for the slower /DTACK. So the delay on the /DTACK is probably about 25ns.

So far it seems happy, so I will leave it running for a while and see what happens.. And I will take out the WS and just use the GAL as a buffer, so the delay will be probably 7ns. though the good thing about running /DTACK through the GAL is that it will buffer and clean it up a little before it gets to CPU.

Current results below..

40-ws.jpg

UPDATE:

I tried a 7ns delay via the GAL and saw some graphics artifacts. So 7ns is to low. Even if I chained 2 IO's for 15ns.. its not far off 25ns anyway. Have to think of tolerances also. So a shame to lose a bit of speed over my first 40MHz test, but no choice really. Its still a fraction faster than the stock 32MHz.. but not much. Only test which shows any difference is int-div 355% to 430%.

I may see if I can do a DTACK bypass (speed up) for ROM access to see if that will gain a bit more speed....

[exxos]

New 40MHz results.. ROM is back up to full speed at 32MHz.. DTACK is full speed for ROM access, but delays 25ns for MB access..

So results are like 1% slower on display, but it seems a lot more stable now!

40WSv2.jpg

Only problem is..

IMG_2418.JPG

The GAL seemed to screw up with the extra code :( So only way to test was to hardwire a 2nd GAL ontop :roll:
I tried a few ways and pins to make it fit, but its not having it.. so a new problem to solve now... :roll: Might be possible with a tri-state nand gate or something..... to tired to care right now though :)

[Bobrane]

Hi Exxos, you are doing an amazing job.
I'm not able to help you in any way, but you can be sure that i'm following this thread with a huge interest.
Keep up your magic!

[exxos]

Hi Exxos, you are doing an amazing job.
I'm not able to help you in any way, but you can be sure that i'm following this thread with a huge interest.
Keep up your magic!

:thanksyellow:

[exxos]

Only just had chance to do some more testing today..

It seems for some reason the GAL does not like a second "output enable" used, even though the basic "gate" code works fine.

I thought I could use a simple OR gate to control DTACK, but then I thought I could just use a simple diode instead.

But it just dawned on me that actually do not need output enable at all on DTACK. Because it is isolated via the GAL, actually do not need to isolate it on the CPU itself, I can just drive it high or low . So really this is a big "d'oh" moment.. I guess it's the side-effect of trying to code things at nearly midnight :roll: :lol:

I have also found sometimes I have to press the reset button to get the machine to boot up. I think the reset line is a little unstable. Actually had this issue on the STFM with my V1.5 booster ages ago. But I think the reset line just needs a better value capacitor on it.. Will look into this another time.. Meanwhile I need to do a little editing PCB get some of this revised board ordered :)

[PhilC]

And some sleep by the sounds of things :P

Very interesting thread, have been checking it the last few days with interest.

[exxos]

And some sleep by the sounds of things :P

That would be nice yes!

Very interesting thread, have been checking it the last few days with interest.

Good that some people find my progress interesting. I never really know if anyone knows what the hell I am talking about half the time tho :lol: not sure I know myself sometimes :lol:

[PhilC]

I understand most of it. In reality I've been re-learning electronics by fixing computers.
Even found myself reading a Dram data sheet the other day to verify the waveform i was seeing... how sad is that :lol: