Question about changing (bus) pullup resistors (Mega STE)

[antorsae]

Hello,

I plan to do some maintenance on my recently acquired MegaSTE. As discussed [anchor]viewtopic.php?t=1585[/anchor] Im going change the bus pull ups since I'd also like to use a 68HC000 processor (less heat).

I am not sure which of the SIL resistor are the ones that need to be changed and with what values:

1 - M9-1-122J
2 - M9-1-103J
3 - M9-1-472J
4 - M9-1-103J
5 - M9-1-103J
6 - M9-1-103J
7 - M9-1-472J
8 - M9-1-472J
9 - M9-1-472J
10 - M9-1-472J
11 - M9-1-102J
12 - M9-1-102J

Thank you!

20240715_212027-SIL.jpg (628.43 KiB) Viewed 1128 times

[exxos]

Bus resistor changes on the MSTE haven't been suggested or tried. I'm sure people have used to HC CPU without any ill effects though.

[atari030]

HC CPU just drops in and goes. Shed loads cooler than the old 12mhz ST CPU's.

[Steve]

I wouldn't recommend changing the CPU based on heat alone ... you could introduce unanticipated problems. ie; I have read that introducing a HC CPU can break certain timing critical demos. People tend to change to the HC CPU when they're experiencing some other problems. I would say, if you're not experiencing problems, stick a little heatsink on the CPU if it makes you sleep better at night, and be happy

Also another warning: Changing SIL resistors might sound straight forward, but more often than not people accidentally break motherboard traces pulling them out. So please beware. Definitely not to be done without a good reason.

[exxos]

I think tests would need to be done with both CPUs to see if it broke anything. I know some people on other forums claimed the HC CPU broke some stuff with demos on the STE, but I never was able to replicate that. Such problems were likely due to other, or , other related factors such as a unstable bus or bad connections or fake / bad CPU etc.

[sporniket]

Also another warning: Changing SIL resistors might sound straight forward, but more often than not people accidentally break motherboard traces pulling them out. So please beware. Definitely not to be done without a good reason.

I second that, I had to change a SIL for installing a mod to my STE, in order to be able to solder wires on them. I lifted 2 pads in the process. Fortunately they were pad not linked to a trace, the vital pad was on the other side.

[atari030]

I've replaced a couple of CPU's with HC's. No difference I could see. On my wonky MegaSTE the CPU was rooted so the HC was a pretty good upgrade. The other I did was an STE, also no discernible change.

Personally, I think a HC in a MegaSTE is a good idea. The Mega alone breaks demo's so no biggie. Oh, the SIL's not so much a good idea.

[ijor]

I think tests would need to be done with both CPUs to see if it broke anything. I know some people on other forums claimed the HC CPU broke some stuff with demos on the STE, but I never was able to replicate that. Such problems were likely due to other, or , other related factors such as a unstable bus or bad connections or fake / bad CPU etc.

I doubt that a CMOS 68K has any functional difference with the older NMOS part. I checked a couple of undocumented issues and the CMOS CPU behaves exactly the same. It is obvious to me that Motorola was very careful to avoid any possible incompatibility.

Now, the analog timing might be slightly different. CMOS parts are obviously faster and will tend to have faster transition times. We already discussed how this could provoke timing differences in some edge cases using Blitter, depending also on the exact Blitter version.

There was some claim about possible incompatibilities in atari-forum. IIRC it was about demos using Spectrum-512 effects. But this is known to be a "dangerous" technique. @troed can probably elaborate about this, but if you attempt to change the palette at the very last possible cycle, it might not always work. So this is not just an issue about CMOS vs NMOS cpu.

[exxos]

I doubt that a CMOS 68K has any functional difference with the older NMOS part. I checked a couple of undocumented issues and the CMOS CPU behaves exactly the same. It is obvious to me that Motorola was very careful to avoid any possible incompatibility.

Yeah. There was that wiki article which claimed some difference a while ago, but turned out the wiki was wrong anyway.

Now, the analog timing might be slightly different. CMOS parts are obviously faster and will tend to have faster transition times. We already discussed how this could provoke timing differences in some edge cases using Blitter, depending also on the exact Blitter version.

Yes, and that pullup resistor was in the mix as well. In fact those slight differences are why the HC fixes the DMA issues. So the HC *can* indeed change timings. I measured up to 22ns while doing the DMA research. I am not aware of any side "bad" effects though. Possible some rare edge-case exists, but like you say, different blitter versions cause issues, even TOS versions and everything else "random" which can exists across various machines also.

There was some claim about possible incompatibilities in atari-forum. IIRC it was about demos using Spectrum-512 effects. But this is known to be a "dangerous" technique. @troed can probably elaborate about this, but if you attempt to change the palette at the very last possible cycle, it might not always work. So this is not just an issue about CMOS vs NMOS cpu.

Another Problem is the amount of variations on the STE itself. I mean Atari changed various pullups on later machines. Later machines may "tolerate" the HC CPU and work fine, earlier machines may not. But like I said years ago, I already documented 10K wasn't good enough. People could blame the HC CPU for "not working" but that's not taking into account Atari's own fixes. The HC could simply show up stability issues on the bus for example, Which Atari fixed when they changed to 4.7K. Of course, this doesn't mean the HC CPU "doesn't work". Things are not that simple. Unfortunately a lot of people never understood that.

I mean you could have 2 identical motherboards where spectrum 512k might work on one, but not on the other. Thats just a side-effect of pushing the timings to the limit. If only there was some new motherboard which attempts to standardise everything...