Different SHIFTER versions

GFA, ASM, STOS, ...

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

MarkP
Atarian
Atarian
Posts: 5
Joined: Tue Jul 31, 2018 10:22 pm

Re: Different SHIFTER versions

Postby MarkP » Tue Jul 31, 2018 11:59 pm

Potentially silly question, but what type of screen are you recording from? CRT or LCD?

Whilst some of these look to definitely be shifter-internal glitches, others including the "interlaced" artefacts put me somewhat in mind of a particular old PC demo that messed directly with the VGA's CRTC register values to produce a code-economic vertical sine wibbler by shifting the sync positions instead of doing anything to the graphics data itself. Which doesn't replay properly on most LCDs, and those which do handle it still often claim a loss of signal here and there.

So if these pictures are from an LCD, I wonder if those particular shifters might be responding to the code in a way that alters the effective sync timings more wildly than later revisions and it's enough to upset the screen on the lines where those tricks are used?

(if it's actually on a CRT, then this is probably nonsense ... I've been trying to read and decipher all the research on these subjects the last few days, trying to pick out particular details for my own purposes and getting very confused ;) )

User avatar
Smonson
Obsessive compulsive Atari behavior
Obsessive compulsive Atari behavior
Posts: 125
Joined: Sat Feb 20, 2016 9:45 am
Location: Canberra
Contact:

Re: Different SHIFTER versions

Postby Smonson » Sat Aug 04, 2018 9:18 am

I have a small request: could someone confirm whether Spectrum 512 works perfectly on the -20 or -38 shifter versions in 60Hz?

Using my verilog model (which is similar to the -38), it's perfect at 50Hz but totally broken at 60Hz.

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sat Aug 04, 2018 3:07 pm

I finally came around to fix my IMP board (I fried the Vsync diode somehow).

So here's revision 2 of the LA capture package:

http://absencehq.de/atari/shiftercaptures_r2_c025914-38_38A_c070713-002.zip

Changes:
-added traces for C070713-002 8833 low/mid 50Hz using a full IMP chipset board
-fixed typo regarding the C070713-002 Shifter (hopefully throughout)

Please let me know if anything else is bodged.

For doing different wakestate traces on IMP we have to have cooler weather :)

I might add again that I really, really, really miss the AF wiki for storing and sorting findings like this and would (like probably many others) do alot to get it back up in it's full AF glory.

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sat Aug 04, 2018 3:12 pm

MarkP wrote:Potentially silly question, but what type of screen are you recording from? CRT or LCD?

I can only speak for the video recordings: there was no screen involved, it's a (more or less) direct RGB capture. The capture card proved quite stable in the past (which will ofc change, now that I mentioned it :) ), and I didn't see anything in the recording a real CRT didn't show.

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sat Aug 04, 2018 3:20 pm

alien wrote:Thanks for the video tin. That's really interesting (and a bit of a shame).

You're welcome :) Nothing to worry about - as you said, your screen worked on most STs people used at the time it was released. Thanks alot for your fabulous work, you made many of us scratch our heads back then. Unfortunately most of us outside of france only got their hands on your articles when it was too late ;-)

ijor
Hardware Guru
Hardware Guru
Posts: 3627
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sun Aug 05, 2018 3:06 am

Smonson wrote:I have a small request: could someone confirm whether Spectrum 512 works perfectly on the -20 or -38 shifter versions in 60Hz?
Using my verilog model (which is similar to the -38), it's perfect at 50Hz but totally broken at 60Hz.


I doubt it is related to the Shifter version, and it is actually the other way around. Spectrum 512, or at least some versions of it, only run correctly at 60 Hz. I guess what's happening is the following. Some Spectrum 512 distribution disks and images include a frequency switcher program in the auto folder, whenever it runs it toggles between 50 Hz and 60 Hz. It was included for being able to run Spectrum 512 correctly on European computers. If it runs on a computer that is already running at 60 Hz (such as booting with US TOS), then it will enable 50 Hz and it will have the opposite effect, Spectrum 512 won't display correctly. Probably you are booting such an image ...

If that's not what's happening to you then please provide more info: Did you test with a real Shifter (whatever version)? What exactly do you mean by totally broken, can you post a picture?

User avatar
Smonson
Obsessive compulsive Atari behavior
Obsessive compulsive Atari behavior
Posts: 125
Joined: Sat Feb 20, 2016 9:45 am
Location: Canberra
Contact:

Re: Different SHIFTER versions

Postby Smonson » Sun Aug 05, 2018 8:26 am

Ah, yeah, I forgot that I can use my real shifter to test this (I lost it for a long time and recently found it again). I confirmed that the copy of Spectrum 512 that I have cannot work in 60Hz, only 50Hz. I will post pics on the FPGA shifter thread. Thanks Ijor :D

ijor
Hardware Guru
Hardware Guru
Posts: 3627
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Sun Aug 05, 2018 2:39 pm

tin wrote:I finally came around to fix my IMP board (I fried the Vsync diode somehow).
-added traces for C070713-002 8833 low/mid 50Hz using a full IMP chipset board


Many thanks once again. But something is odd with this new trace on low rez (any chance you mixed up and it is actually on med rez?). Can you make one or two more. Only with the full IMP chipset, only on LOW rez. If you can with two different powerup cycles. And please make sure that the screen is solid. That is, the start of the B2 (or whatever RGB signal you trace) should be exactly at the left border. Thanks a lot!

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sun Aug 12, 2018 2:51 pm

ijor wrote:But something is odd with this new trace on low rez (any chance you mixed up and it is actually on med rez?).
Can you make one or two more. Only with the full IMP chipset, only on LOW rez. If you can with two different powerup cycles. And please make sure that the screen is solid. That is, the start of the B2 (or whatever RGB signal you trace) should be exactly at the left border. Thanks a lot!

It's absolutely possible there's been a naming mixup or sth like that.
Here are some new traces (full IMP) with 20ns - unfortunately there was no time to do proper PNGs. If these are good traces, I'll add them to the set and add the PNGs.

http://absencehq.de/atari/shifter_tests.zip

The little testbed PRG used simply shows a red background color with a blue foreground in low/med/high (well high obviously not red/blue) switched via keys 1,2,3 (unsynchronized, not even to VBL). While switching I produced also some banded and shifted left states, all stable. DL4 and DL5 (meaning "DE to LOAD cycles") are included as well.

ijor
Hardware Guru
Hardware Guru
Posts: 3627
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Mon Aug 13, 2018 4:13 am

Hi,

The 3 "normal" low rez traces are now ok. So it sounds indeed that there was some kind of mix up with the previous one on low rez.

I didn't check the other two, banded and shifted. Not sure what you mean by this effect?

Thanks once more again :)

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Mon Aug 13, 2018 7:41 am

You‘re welcome :)

I‘ll add those traces to the big packgage including PNGs asap.

Regarding „banded“ and „shifted“:
„banded“ shows every 2nd column (16 pixel block)one column background color although it should show gfx data. In this case it might be a plane effect, have to check palette and plane data - but either way this is either a true „banded“ display or a stable (aka one-shot, one switch, never switched agin) plane shift, e.g. 2 bytes or sth like that. So it‘s interresting to document in any way.
„Leftshifted“ has the DE=>BG2 delay (aka left border) which is around 22/23 8Mhz clocks on normal display shifted left to around 9 clocks. Also stable with just one rez switch (manually) and then never switched again.

ijor
Hardware Guru
Hardware Guru
Posts: 3627
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Mon Aug 13, 2018 1:09 pm

tin wrote:Regarding „banded“ and „shifted“ ...


I suspected something like that. But the trace bandwidth (50 MHz) is a bit too low and then it's difficult to analyze unless you know beforehand what you are looking for.

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Mon Aug 13, 2018 4:14 pm

ijor wrote:
tin wrote:Regarding „banded“ and „shifted“ ...


I suspected something like that. But the trace bandwidth (50 MHz) is a bit too low and then it's difficult to analyze unless you know beforehand what you are looking for.

I'll do proper 8ns ones later, these effects are pretty easy to recreate on this IMP set, it seems.

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Mon Aug 13, 2018 7:03 pm

Here's the previous IMP set in 8ns and with dl3, dl6 and two different lowrez "shiftedleft" cases added.

http://absencehq.de/atari/shifter_tests_v2.zip

Also: rechecked that the banded case is a "true" banded case, not just a plane shift - my testbed had all planes fully set and only bg color was set to red. Have to refine that method, though, since that way I can't recognize stable plane shifts visually, since all gfx data is always displayed blue, regardless of a possible plane offset.

I'll do some automation to generate the PNGs from the CVSs automatically, as they take most of the time required to do a full package.

ijor
Hardware Guru
Hardware Guru
Posts: 3627
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: Different SHIFTER versions

Postby ijor » Tue Aug 14, 2018 7:30 pm

tin wrote:Here's the previous IMP set in 8ns and with dl3, dl6 and two different lowrez "shiftedleft" cases added.


Thanks once more again. But seems to me again that some of the traces might have the wrong resolution.

This looks like low rez, and not med rez: C070713-002IMP_med_50hz_dl3.vcd

tin
Atari maniac
Atari maniac
Posts: 99
Joined: Mon Jul 23, 2012 7:59 am
Contact:

Re: Different SHIFTER versions

Postby tin » Sun Sep 02, 2018 7:36 pm

ijor wrote:
tin wrote:Here's the previous IMP set in 8ns and with dl3, dl6 and two different lowrez "shiftedleft" cases added.


Thanks once more again. But seems to me again that some of the traces might have the wrong resolution.

This looks like low rez, and not med rez: C070713-002IMP_med_50hz_dl3.vcd


Thanks for spotting that. Redid that trace and merged all traces including their respective PNGs into a new release: http://www.absencehq.de/atari/shiftercaptures_r3_c025914-38_38A_c070713-002.7z

All traces are now @8ns.

(There's also a little gnuplot script in there for plotting the start of these traces).


Social Media

     

Return to “Coding”

Who is online

Users browsing this forum: No registered users and 2 guests