Images generated by RastaConverter

[drpeter]

22 minutes ago, ivop said:

Perhaps one can detect the specific loads and stores within a 6 cycle window

The latest version of RastaSlide produces annotated disassembly of RastaConverter .xex files, highlighting lines which will produce the known <=5 cycle write window bug. It may be that @Sheddy would consider updating RastaSlide so that it also highlights 6 cycle write windows...

Edited May 21, 2021 by drpeter

[ivop]

2 minutes ago, drpeter said:

The latest version of RastaSlide produces annotated disassembly of RastaConverter .xex files, highlighting lines which will produce the known <=5 cycle write window bug

Now that algorithm to detect bad lines has to be added to the evaluation function of RastaConverter, and then evaluate minus very bad

 

How does RastaSlide detect lines with the known write window bug?

 

[drpeter]

7 minutes ago, ivop said:

How does RastaSlide detect lines with the known write window bug?

It's a simple mechanical process of highlighting writes to a player horizontal position register that will complete <=5 colour clocks before the left edge of the player is due to be displayed...

 

See (very) detailed explanation here.

Edited May 21, 2021 by drpeter

[+Stephen]

I'll run this test on my PAL 1088XLD later this evening.  It uses all original Atari chips, but the motherboard is a custom one.  It will be interesting to see if that makes a difference (I don't expect it to).  I also have a 576NUC I can test with.  That will allow me to use the "freeze spray trick" on the GTIA and ANTIC to see if it makes a difference.

[a8isa1]

7 hours ago, drpeter said:

This is a rejig of drpeter_Artefact_1 that places Player 3 on the left of the screen.  I wonder if with this the leftmost bar jumps up to 6 after warm-up?

drpeter_Artefact_1a.xex 18.29 kB · 4 downloads

a8isa1's 800 shifts to bar 6 in 50 seconds from a cold start <-- click here to view

[+Philsan]

9 hours ago, drpeter said:

 

Fascinating.  So on this particular machine it only happens with Player 3, after about 8 minutes of warming up, and only on the far left side of the scanlines. Furthermore, there seems to be a consistent slight variation dependent on the vertical point on the screen, as the rectangles don't all disappear at an identical time/rate...

 

I wonder whether the horizontal variation is due to exactly that, or the fact that the leftmost rectangle on each line represents in-scanline reuse of the player (whose previous incarnation provides the leftmost vertical colour banding).

 

This is a rejig of drpeter_Artefact_1 that places Player 3 on the left of the screen.  I wonder if with this the leftmost bar jumps up to 6 after warm-up?

drpeter_Artefact_1a.xex 18.29 kB · 6 downloads

  

 

[drpeter]

Hmmm, looks like it becomes more widespread if you leave it long enough...  How long did you run drpeter_Artefact_3.1 for?

[drpeter]

I thought I'd have some fun with this and at the same time reward my valiant colleague testers with an Atari 8-bit first-

 

The Atari 8-bit motherboard temperature monitor utility ?

 

I can't test it on real hardware and by definition it won't work on Altirra (although the picture will display)

 

I would love feedback from those trying it out on genuine Ataris...

 

Tatra Mountains, Slovakia

 

57 colours.

 

 

drpeter_GTIA_Temp_Monitor.xex

 

N.B. No assurance as to the accuracy or usefulness of the GTIA temperature monitor is offered or implied ?

 

N.B.2  It will only work from a fully cold start, on machines which have demonstrated artefactual behaviour with drpeter_Artefact_3.1 (see recent posts)

Edited May 22, 2021 by drpeter

[+Philsan]

Who wants to know outside temperature, I want to know Atari temperature!

 

Here's the magic that happened after 23 minutes:

 

 

And after circa 40 minutes, I found that other changes happened!

Unfortunately I don't know whether 45 and 35 appeared at the same time or not, I didn't expect it.

I can repeat the test if necessary.

 

 

An hour has passed and nothing changed.

 

Edited May 22, 2021 by Philsan

[+CharlieChaplin]

Hmmm, could it be that you (re-)discovered DGF...?!?

 

 

Besides, one of my Atari 800XL computers (PAL, XL-OS Rev 2) showed your GTIA temp monitor pict. (from post #3258) and displayed 15 degrees when loaded, then after about ten minutes it displayed 35 degrees, thirty seconds later it displayed 25 degrees and one minute later 45 degrees...?!?

 

EDIT: Three minutes later 55 degrees appeared... and all other temperatures were still shown and all of them were stable after a few seconds (they started with flicker, but in less than ten seconds flicker disappeared). So the total running time was approx. 14 minutes 30 seconds.

 

[drpeter]

41 minutes ago, CharlieChaplin said:

Hmmm, could it be that you (re-)discovered DGF...?!?

Lots of interesting stuff here!

 

Although it seems to be thermally sensitive, this also appears to be a different temperature-sensitive timing effect to the one that produces DGF graphic modes.

[drpeter]

45 minutes ago, CharlieChaplin said:

displayed 15 degrees when loaded, then after about ten minutes it displayed 35 degrees, thirty seconds later it displayed 25 degrees and one minute later 45 degrees...?!?

 

EDIT: Three minutes later 55 degrees appeared... and all other temperatures were still shown and all of them were stable after a few seconds (they started with flicker, but in less than ten seconds flicker disappeared).

 

In the limited testing that's been done for this effect so far, it's clear that the dependency on horizontal and vertical screen position and the player involved varies somewhat from machine to machine.  GTIA temp monitor was programmed to follow the common dependencies seen so far, in an attempt to get the temperatures to appear in the correct sequence.  The flickering appearance followed by stabilisation is the expected behaviour.

 

I wonder if you could run drpeter_Artefact_3.1 from a cold start on your machine and post video (ideally) or describe in detail what happens?

[drpeter]

2 hours ago, Philsan said:

An hour has passed and nothing changed.

 

It works! ?

 

Seems your machine's GTIA may be resistant to that final change though...

 

I see that people have resorted to blowing hairdryers or even heat guns at their Atari in efforts to provoke other temperature-dependent GTIA timing 'features'!

 

Edited May 22, 2021 by drpeter

[+CharlieChaplin]

Alright,

 

did the Artifact 1 test (from post #3191) and all four colour stripes (nr. 1-4 from left to right) showed one black line at the start. After eight minutes stripe nr. 4 showed two black lines, after nine minutes stripe nr. 2 showed two black lines, after twelve minutes stripe nr 1 showed two black lines and after fifteen minutes stripe nr. 3 showed two black lines - so after 15 minutes all four stripes showed two black lines. (But I only took three pictures, after 9, 12 and 15 minutes.)

 

 

 

 

After that I switched off the A8 for thirty minutes to cool down, then I did Artifact Test 3_1. After three minutes the first stripes began to disappear and then every minute something happened and several stripes disappeared - after 17 minutes all stripes had disappeared. In this test the stripes of row nr. 1 disappeared first, then the stripes of rows nr. 2 and 4 disappeared and at the end the stripes of row nr. 3 disappeared. (I only took one picture, 17 minutes after the start, when all stripes had disappeared, maybe I should do a video?)

 

 

 

Will do Artifact test 1a and Artifact test 2 next...

 

[+Philsan]

6 hours ago, drpeter said:

Lots of interesting stuff here!

 

Although it seems to be thermally sensitive, this also appears to be a different temperature-sensitive timing effect to the one that produces DGF graphic modes.

In fact it seems similar but not the same.

I've run DGFINIT.XEX but after 80 minutes progress is always 0% (screenshot is from Altirra but on my 800XL it's the same, time elapsed on the right apart).

 

 

[drpeter]

53 minutes ago, Philsan said:

In fact it seems similar

DGF is situation (as I read it) whereby when the chips are hot the horizontal synchronisation of ANTIC and GTIA can be induced to drift out of alignment by up to one colour-clock. Absolute player/missile positioning is unaffected, its just the playfield that shifts 'underneath them' from which it I take it that the player/missile data that ANTIC reads from RAM at the beginning of each scanline is uploaded to GTIA graphic registers pretty much straightaway and is not waiting for any 'permission' from ANTIC relating to player positions.  If that is the case, from very early in the scanline GTIA is no longer dependent on ANTIC in any way for player display, as GTIA has its own horizontal colour-clock counter and presumably generates the trigger to prepare a player for display, including loading the shift register, internally and subject to no outside influence.

 

drpeter_Artefact_1 also demonstrates that there is no relative shift in playfield and player graphics display when this effect is active, so my feeling is that this is an independent effect governed solely by GTIA's internal pipelining relating to GTIA's internal horizontal trigger for player-display-preparations- but not actual subsequent display: player display positions don't shift after warm-up of affected GTIAs. The observable fact that the contents of the shift register can get overlain by new data mid-display if the player graphics data register is updated and the player retriggered demonstrates that the loading of the shift register only occurs after triggering, not earlier in the scanline. So it seems that the (primary-trigger->load-and-prepare-shift-register) sequence is lengthened from 5 to 6 colour clocks, beginning one colour clock earlier when affected GTIAs warm up, but the secondary trigger to then start shifting out bits for display still occurs at the appointed colour clock, so the horizontal position of the graphic image is not changed.

Edited May 22, 2021 by drpeter

[ivop]

@ijor 's reverse engineered schematic:

 

[drpeter]

12 minutes ago, ivop said:

@ijor 's reverse engineered schematic:

Fascinating reading (the thread that is -the schematic is all Greek to me!)

 

Perhaps @ijor, @Bryan or @Rybags have some thoughts on the cause/mechanics of this phenomenon, as I'm just speculating wildly...

[a8isa1]

12 hours ago, drpeter said:

I thought I'd have some fun with this and at the same time reward my valiant colleague testers with an Atari 8-bit first-

 

The Atari 8-bit motherboard temperature monitor utility ?

 

I can't test it on real hardware and by definition it won't work on Altirra (although the picture will display)

 

I would love feedback from those trying it out on genuine Ataris...

 

Tatra Mountains, Slovakia

 

57 colours.

 

 

drpeter_GTIA_Temp_Monitor.xex 21.95 kB · 10 downloads

 

N.B. No assurance as to the accuracy or usefulness of the GTIA temperature monitor is offered or implied ?

 

N.B.2  It will only work from a fully cold start, on machines which have demonstrated artefactual behaviour with drpeter_Artefact_3.1 (see recent posts)

My 800 records 55 degrees almost immediately after program loads via SIO2PC-USB.  

 

My 800XL never changes from 15 degrees.   Thanks to my RTC module, a DS3231, inside the 800XL I know the ambient temperature is 89 degrees F (31.67 C).

[+CharlieChaplin]

3 hours ago, Philsan said:

In fact it seems similar but not the same.

I've run DGFINIT.XEX but after 80 minutes progress is always 0% (screenshot is from Altirra but on my 800XL it's the same, time elapsed on the right apart).

 

 

 

Same here, after running 90 minutes on real Atari with AVG cart. the DGF init program still showed 0%, so I gave up. Maybe the program should run over night, until the next morning (approx. 12 hours) ?!?

 

[+CharlieChaplin]

Some more tests for drpeter...

 

1) Test 1a after 15 minutes:

 

 

2) Test 1a after 17 minutes:

 

 

3) Test 1a after 20 minutes:

 

 

4) Test 1a after 23 minutes:

 

 

So for Test 1a the two black lines appeared in this order: row 1, row 3, row 2, row 4.

After that I switched the A8 off for one hour, then I did run Artefact test 2. Result after 23 minutes looked like this:

 

 

Enough for tonight (1:30 am here in Germany). Now I am running the DGF init over night, let's wait and see what happens after 8, 10 or 12 hours...

 

[+CharlieChaplin]

And here is the last test, DGF Init after 12 hours, 9 minutes - progress 0%, so I give up here. Maybe this bug/feature only works with "cheap" XE/XEGS computers ?!?

 

[ijor]

23 hours ago, drpeter said:

Fascinating reading (the thread that is -the schematic is all Greek to me!)

 

Perhaps @ijor, @Bryan or @Rybags have some thoughts on the cause/mechanics of this phenomenon, as I'm just speculating wildly...

I can't check the whole thread right now and give a detailed answer. But we talked about this before with some experiments that, IIRC, Bryan did years ago. As I explained in that thread with the reverse engineered schematics, GTIA has some weak synchronization paths. That means that synchronization logic is at the edge and might break in some special conditions that depend on the specific chip, the temperature AND the voltage. On that thread I also described the one that is perhaps the most significant path that might break sync.

 

Exactly what happens here in this case I can't really comment without reading the thread and checking what the program exactly does, and I'm afraid I don't have the time right now.

Edited May 23, 2021 by ijor

[drpeter]

22 hours ago, a8isa1 said:

My 800XL never changes from 15 degrees.

I don't think you've reported the result of running drpeter_Artefact_3.1 on your 800XL. If there's no change to the image of drpeter_Artefact_3.1 after ~1hr of warm-up it's unlikely you'll see any temperatures >15 degrees appear on drpeter_GTIA_temp_monitor.

Edited May 23, 2021 by drpeter

[drpeter]

On 5/22/2021 at 5:28 PM, CharlieChaplin said:

after 17 minutes all stripes had disappeared. In this test the stripes of row nr. 1 disappeared first, then the stripes of rows nr. 2 and 4 disappeared and at the end the stripes of row nr. 3 disappeared. (I only took one picture, 17 minutes after the start, when all stripes had disappeared, maybe I should do a video?

Video would be great, if at all possible.?