Different SHIFTER versions

GFA, ASM, STOS, ...

Moderators: simonsunnyboy, Mug UK, Zorro 2, Moderator Team

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Different SHIFTER versions

Postby ijor » Thu Jul 05, 2018 5:00 pm

Breaking news ... :)

I still need to confirm with actually testing multiple chips. But at this time I am quite confident that there are some incompatibilities depending on the version of the Shifter chip. e.g., I believe older Shifter version won't display correctly the famous Alien 4-bit (or 4 pixel) sync scroll at the Punish Your Machine demo.

The older versions are C025914-20 or C025914-38. Newer versions are C025914-38A (note the extra A at the end) and C07013.

If you have one of the older Shifter chips, please test that demo.

Possibly other demos using non standard stabilization techniques might be affected as well.

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Thu Jul 05, 2018 6:57 pm

Interresting. It indeed does not work with the C025914-38 shifter (tested on an 260 ST). Never noticed that. Works fine on 38A though (obviously).

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Thu Jul 05, 2018 9:30 pm

Thanks a lot for the test!

It doesn't work or the video is corrupted? Can you describe in which way, or post a picture if that wouldn't be too much of a trouble. Thanks again.

User avatar
alien
Atari maniac
Atari maniac
Posts: 87
Joined: Sat May 01, 2004 4:01 am
Location: USA
Contact:

Re: Different SHIFTER versions

Postby alien » Thu Jul 05, 2018 11:05 pm

I'd like to see the effect too. Thanks!
Alien / ST-Connexion

User avatar
Smonson
Atari maniac
Atari maniac
Posts: 92
Joined: Sat Feb 20, 2016 9:45 am
Location: Canberra
Contact:

Re: Different SHIFTER versions

Postby Smonson » Fri Jul 06, 2018 12:18 am

It sounds like this could explain why I've never been able to see that sync scroll effect work on the FPGA model. Very interesting indeed!

User avatar
troed
Atari God
Atari God
Posts: 1380
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: Different SHIFTER versions

Postby troed » Fri Jul 06, 2018 6:16 am

tin - if that 260 has 1MB of RAM, does {Closure} look ok?

(it uses "non standard stabilization" ... )

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Fri Jul 06, 2018 6:36 am

ijor wrote:Thanks a lot for the test!

You‘re welcome - I have one MB rigged with a ZIF socket just for shifter tests like this :) Had to do this test on two differrent machines for now though, will retest with the same motherboard ASAP.

ijor wrote:It doesn't work or the video is corrupted? Can you describe in which way, or post a picture if that wouldn't be too much of a trouble. Thanks again.

See attached. This is with the same disk images that work fine on 38A.
Didn‘t have the time to do a proper direct source video capture, but will do that asap.
Could do LA traces too, if needed.

troed wrote:tin - if that 260 has 1MB of RAM, does {Closure} look ok?

(it uses "non standard stabilization" ... )


I know :) Yes, it has 1mb. Closure is next on my list, I‘m pretty curious, too.

Edit: gnnn, „Attachment too large“. Well, here it is: http://absencehq.de/atari/IMG_0540.JPG http://absencehq.de/atari/IMG_0541.JPG

czietz
Hardware Guru
Hardware Guru
Posts: 685
Joined: Tue May 24, 2016 6:47 pm

Re: Different SHIFTER versions

Postby czietz » Fri Jul 06, 2018 10:01 am

I'm afraid that particular sync scroll already glitches heavily on my test machine, 1040STfm with the IMP Shifter. (That's the computer I also had trouble with while running the SNDH intro: http://atari-forum.com/viewtopic.php?f=15&t=31742)

IMG_2938.JPG

IMG_2942.JPG

(Excuse the photographic screen captures, I don't have time to capture a video, right now.)

I'll have to postpone more tests (incl. with a -20 Shifter) until I have more time.
You do not have the required permissions to view the files attached to this post.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Fri Jul 06, 2018 1:59 pm

tin wrote:You‘re welcome - I have one MB rigged with a ZIF socket just for shifter tests like this :)


Good idea. Btw, it is possible to fit a ZIF socket into a regular socket, or it can only be soldered?

See attached. This is with the same disk images that work fine on 38A. Didn't have the time to do a proper direct source video capture, but will do that asap.


That would be nice if you can. What I expected is the "16-pixel background" bands. It might be what is happening but it is difficult to be sure with that picture.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Fri Jul 06, 2018 2:04 pm

czietz wrote:I'm afraid that particular sync scroll already glitches heavily on my test machine, 1040STfm with the IMP Shifter. (That's the computer I also had trouble with while running the SNDH ...


In this case the bands are more clear, those are not just glitches. I believe however that the cause is very different on the IMP Shifter than on the -38 (or -20) one. It would be interesting to see if adding the diode at the VCC pin fixes the problem. I suspect it might.

czietz
Hardware Guru
Hardware Guru
Posts: 685
Joined: Tue May 24, 2016 6:47 pm

Re: Different SHIFTER versions

Postby czietz » Fri Jul 06, 2018 2:25 pm

ijor wrote:In this case the bands are more clear, those are not just glitches. I believe however that the cause is very different on the IMP Shifter than on the -38 (or -20) one. It would be interesting to see if adding the diode at the VCC pin fixes the problem. I suspect it might.


I will definitively test that; just not in the next few days...

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sat Jul 07, 2018 11:45 am

ijor wrote:Btw, it is possible to fit a ZIF socket into a regular socket, or it can only be soldered?

You can squeeze them in, but that probably damages the original sockets‘ contacts (the ZIFs pins are pretty wide). I did that on this particular test MB since I don‘t plan to remove the ZIF.

troed wrote:tin - if that 260 has 1MB of RAM, does {Closure} look ok?

(it uses "non standard stabilization" ... )


Did not have time to setup the full test rig, but was able to check the -38 on the ZIFfed MB.

Aliens PYM screen shows bands as expected and seen on Christians screenshots. Closure also shows bands in all fullscreens, but they are shifted each line. I‘ll capture some videos in the next few days.
The additional screen corruptions in the 260 are strange, have to retest that, too. Only noticable difference is an old MMU (old top border start), but that shouldn‘t have that effect.

Everything is fine with 38A on the same MB/setup.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sat Jul 07, 2018 1:19 pm

tin wrote:Did not have time to setup the full test rig, but was able to check the -38 on the ZIFfed MB.
Aliens PYM screen shows bands as expected and seen on Christians screenshots. Closure also shows bands in all fullscreens, but they are shifted each line.
...
Everything is fine with 38A on the same MB/setup.


Thanks a lot again! Yeah, that was what I expected. You were immensely helpful :)

I'll post a small resume of all the Shifter variants later.

The additional screen corruptions in the 260 are strange, have to retest that, too. Only noticable difference is an old MMU (old top border start), but that shouldn‘t have that effect.


You mean older GLUE, not MMU. The old top border issue was well known, so I expect it was tested or at least taken into consideration. But don't know for sure.

User avatar
troed
Atari God
Atari God
Posts: 1380
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: Different SHIFTER versions

Postby troed » Sat Jul 07, 2018 1:38 pm

Closure fullscreens should work with the old top border* - but the scrolling-dude screen might not when I think about it.

/Troed

*) side effect of them running at 60Hz at the end of all scanlines _except_ the ones which on 50Hz would cause vblank/end screen conditions.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sat Jul 07, 2018 1:54 pm

troed wrote:Closure fullscreens should work with the old top border* - but the scrolling-dude screen might not when I think about it.


I understand you tested Closure with an IMP Shifter, right? That's why I suspect czietz failure is due to his specific motherboard not being fully compatible with an IMP Shifter (as it seems some Megas are) and no diode fix.

You now need to acquire an old chipset and make Closure compatible with it :)

User avatar
troed
Atari God
Atari God
Posts: 1380
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: Different SHIFTER versions

Postby troed » Sat Jul 07, 2018 8:00 pm

ijor wrote:
troed wrote:Closure fullscreens should work with the old top border* - but the scrolling-dude screen might not when I think about it.


I understand you tested Closure with an IMP Shifter, right? That's why I suspect czietz failure is due to his specific motherboard not being fully compatible with an IMP Shifter (as it seems some Megas are) and no diode fix.

You now need to acquire an old chipset and make Closure compatible with it :)


Heh. Closure was tested on 9 different STs ... but yet will fail in some conditions. Now what I didn't know back then is that it's possible to detect whether it's an IMP chipset (GLUE, at least) or not ... don't know if that will help.

Otoh, a lot of modern demos don't work on all configs .. people are happy they get it working on Youtube it seems ;) Many demos don't care about Spec512 pixels is one thing I've noticed.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sat Jul 07, 2018 8:59 pm

troed wrote:Heh. Closure was tested on 9 different STs ... but yet will fail in some conditions. Now what I didn't know back then is that it's possible to detect whether it's an IMP chipset (GLUE, at least) or not ... don't know if that will help.


Sorry, didn't mean to imply you didn't test Closure well enough. I know how much time and dedication you invested in this. It is just that it is still not clear to me if the IMP Shifter has some kind of incompatibility. I suspect that at least as long as it has the diode fix, it should be functionally equivalent to Shifter -38A. But currently it is not much more than an educated guess, hence my question.

I'm not sure you can detect an IMP GLUE. I know you can detect an IMP MMU. The old top border can be detected, of course, but the new top border is already present in the GLUE -38A (newer Ricoh). It is true that detecting the old top border might be a hint that the computer has an old Shifter as well.

czietz
Hardware Guru
Hardware Guru
Posts: 685
Joined: Tue May 24, 2016 6:47 pm

Re: Different SHIFTER versions

Postby czietz » Sat Jul 07, 2018 9:04 pm

ijor wrote:I'm not sure you can detect an IMP GLUE.


While hunting down some bug in EmuTOS, I recently learned that IMP and Ricoh chipsets have different bus error behavior. See my tests summarized in this post on hatari-devel: https://listengine.tuxfamily.org/lists. ... 00104.html

Since afaik the address decoding is handled by GLUE, I think the difference mentioned above will specifically allow you to tell IMP and Ricoh GLUEs apart.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sat Jul 07, 2018 9:38 pm

czietz wrote:While hunting down some bug in EmuTOS, I recently learned that IMP and Ricoh chipsets have different bus error behavior. See my tests summarized in this post on hatari-devel: https://listengine.tuxfamily.org/lists. ... 00104.html

Since afaik the address decoding is handled by GLUE, I think the difference mentioned above will specifically allow you to tell IMP and Ricoh GLUEs apart.


I'm not so sure. I didn't actually test a mixed MMU and GLUE on an IMP/RICOH combination, but I think the different behavior is due to MMU, and not to GLUE.

When accessing MMU or SHIFTER, GLUE performs the coarse address decoding only . It is up to MMU to perform further fine address decoding and to decide if it asserts DTACK or not, and if so exactly when. If MMU doesn't assert DTACK, a bus error would be triggered.

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sun Jul 08, 2018 1:15 pm

Here are a few quick video captures (50Hz, no sound, direct ST RGB feed). All done on same MB with -38 and re-verified to run OK with -38A. Unfortunately I didn't log the respective wakestates.
PYM:



Closure Intro (two different power cycles, didn't check Wakestate):



Closure other parts:




Closure failure that happens only sometimes on different runs if left alone, but not always (demo resets itself, no user interaction):



Also checked with an IMP shifter on the board used above, but that one doesn't behave nice with a non-IMP chipset in general, it seems. Also got the banded effect with the 4-bit scroll with full IMP chipset on different MB (no recording, sry) - probably a wakestate issue, since this didn't happen on retry. But IMP is probably another story.

Last edited by tin on Sun Jul 08, 2018 1:38 pm, edited 1 time in total.

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sun Jul 08, 2018 1:36 pm

tin wrote: The additional screen corruptions in the 260 are strange, have to retest that, too.

Screen artefacts and bands are the same between that 260 and the captured MB. The camera I took the screenshots with did blend 2 frames, that‘s why the bands aren‘t clearly visible.

ijor wrote:Btw, it is possible to fit a ZIF socket into a regular socket, or it can only be soldered?

Forgot that I also plugged a stack of two sockets underneath the ZIF to raise it above the shifter shielding - so the original socket might not be damaged by adding a ZIF after all. That‘s probably also the reason for the faint colors in above captures.

User avatar
troed
Atari God
Atari God
Posts: 1380
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: Different SHIFTER versions

Postby troed » Sun Jul 08, 2018 2:35 pm

Thanks for the captures! The two different effects on the "big SYNC logo dist" with a power cycle in between are interesting. In the first video there's a bitplane error, and the cause is that my "stabilizer-free scanline" (which stabilize at the left border) has not clear the intended Shifter word and thus there's a bitplane shift.

The second video is at the correct bitplane, at least.

Closure runs different code paths for different wakestates so I'm guessing that's the reason. So, with that Shifter there's a timing difference that affects my code in at least one wakestate that way.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sun Jul 08, 2018 6:28 pm

tin wrote:Here are a few quick video captures (50Hz, no sound, direct ST RGB feed). All done on same MB with -38 and re-verified to run OK with -38A.


Wow. Thanks a lot for the videos, really very nice!

Unfortunately I didn't log the respective wakestates.


No so sure GLUE wakestates are the main issue here. SHIFTER wakestates are critical here. GLUE wakestates might be relevant on Closure because as Troed is saying, he uses different code paths depending on the detected wakestate. But on Alien's scroll, it is almost certain that what matters is the SHIFTER wakestate.

Also checked with an IMP shifter on the board used above, but that one doesn't behave nice with a non-IMP chipset in general, it seems. Also got the banded effect with the 4-bit scroll with full IMP chipset on different MB (no recording, sry) - probably a wakestate issue, since this didn't happen on retry. But IMP is probably another story.


Interesting. Please give us more details :) What exactly happens when trying the IMP Shifter with the old chipset? Does the computer fails? Video is always corrupted, even on regular desktop?

And the other MB with full IMP chipset, is it a Mega or an ST? Does it have the diode fix connected at the Shifter VCC pin?

Would you mind doing some LA traces with the IMP? Yeah, I know, I am asking a bit too much :)

tin
Atari maniac
Atari maniac
Posts: 82
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sun Jul 08, 2018 10:13 pm

ijor wrote:Wow. Thanks a lot for the videos, really very nice!

You‘re welcome :)
ijor wrote: GLUE wakestates might be relevant on Closure because as Troed is saying, he uses different code paths depending on the detected wakestate. But on Alien's scroll, it is almost certain that what matters is the SHIFTER wakestate.

I know, but if the behaviour of a syncline changes between power cycles, I normally log the wakestates because that’s usually ssth suspicious. I built a special log chain (via ROM cart and CosmosEx) just for that.

ijor wrote:Interesting. Please give us more details :) What exactly happens when trying the IMP Shifter with the old chipset? Does the computer fails? Video is always corrupted, even on regular desktop?

And the other MB with full IMP chipset, is it a Mega or an ST? Does it have the diode fix connected at the Shifter VCC pin?
o
Would you mind doing some LA traces with the IMP? Yeah, I know, I am asking a bit too much :)

IMP in the board used above (a 1040) produces a stable, bitplane-shifted picture (e.g. also shifted on desktop), works fine otherwise. Interrestingly a reset can change the shift, sometimes even showing no shift at all, but that‘s pretty rare. The full IMP is also a 1040. I don’t think it has a diode fix on the shifter. Do you happen to have a picture how that (factory) patch looks like?
Should have an IMP Mega ST somewhere, too.

I won‘t mind doing LA traces, just point me in the right direction (e.g. which lines exactly, on full or mixed IMP setup, which screen etc.). Might take some days, though.

ijor
Hardware Guru
Hardware Guru
Posts: 3337
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Mon Jul 09, 2018 5:27 am

tin wrote:The full IMP is also a 1040. I don’t think it has a diode fix on the shifter. Do you happen to have a picture how that (factory) patch looks like?


Sorry, I don't. I understand it is a diode added manually, not on the motherboard. The SHIFTER VCC pin is disconnected from the motherboard and is connected instead to 5V through a diode. It is supposed to reduce the voltage to make it run slower.

Christian told me that some later Mega motherboard revisions were supposed to include some kind of built-in fix. Not sure if they include a diode already or has something different. And it is also possible that the same modification is integrated in the last 1040 revisions.

You can check if the VCC pin is connected directly to the 5V power supply or not. You can also measure the active voltage on the pin when the computer is on.

I wont mind doing LA traces, just point me in the right direction (e.g. which lines exactly, on full or mixed IMP setup, which screen etc.). Might take some days, though.


You need a screen showing a solid, neat left margin. We need to measure the point where video starts. Trace the following signals, all are present on SHIFTER:

- 16 MHZ clock
- DE
- LOAD
- One of the RGB signal (could be anyone that changes at the left border)

Trigger on the raising edge of DE to capture the left border. Capture frequency should be 64 MHz or higher. We need only a few samples. As long as the raising edge of DE and the initial edge of the RGB is seen, then that is good enough. There is no need to trace the 32 MHz clock. The 8 MHz signal might be useful but it is not strictly necessary. What we need is the exact time from DE and LOAD to the start of RGB video.

Make one capture on medium rez and if possible one on low rez as well (the one on med rez is more important). And yes, both on the full and on the mixed IMP setup :) Of course, we understand it could take some days. And nobody could blame you if for whatever reason you can't take the captures at all.


Social Media

     

Return to “Coding”

Who is online

Users browsing this forum: No registered users and 2 guests