Text Application in MDA

[yyartsworld]

Hello
I found a new bug.
When i use text mode application like Edit right side is not shown completely.

MCEBlaster

86box

[Bs0Dd]

Same problem

The image is cut at the right when the correct pixel clock is set (my card has a 16MHz oscillator).

I can see the entire screen if I decrease the clock, but then, of course, I have a noisy image

Same situation with the Hercules Graphic mode

Tested on the Paradise Hi-Res Graphics Card (MDA/Hercules mode) and IBM EGA (MDA monitor mode).

[scrapcomputing]

Thanks for the bug report and apologies for the late reply.

Let me see if I can reproduce it locally to come up with a fix.

[Bs0Dd]

Also, not related to MDA, but EGA.

In the EGA mode, the screen also seems to be slightly cut on the right.

But, in the MDA mode, I can use Manual-TTL mode to move the image to the left, and then I'll see the rest of the screen (then the left side will be cut).

This does not work for the EGA mode

Again, I can change the pixel clock and see a full, but noisy screen.

Looks like this problem appears only in EGA mode 10h.

The card is an original IBM EGA with a 16.257 MHz oscillator.

[Bs0Dd]

P.S.: Since the EGA mode uses 640x350, but Blaster uses 640x400, there is some empty space at the bottom.
Maybe it's better to make the image centered with visible top and bottom borders.

[scrapcomputing]

Thanks again for the detailed bug report!

I have seen these border issues before. It has to do with the automatic offset calculation which can be off by a few pixels. The best solution would be to increase the buffer size horizontally by a few more pixels, but the problem is that the Pico 1 doesn't have enough memory to fit these few extra pixels. So the only alternative is setting the horizontal offset manually.
This should work in all modes, but there might as well be a bug in EGA, I would need to check.

Eventually this issue should go away once the Pico 2 variant supports a slightly larger buffer.

As for centering the 640x350 mode, this looks like a bug too. I thought I had taken care of it in a similar way as in MDA/Hercules.

I will look into both issues and will get back to you.
Thanks!

[Bs0Dd]

So the only alternative is setting the horizontal offset manually.

The problem seems to be not only with the offset, but with the width too.
Looks like the MDA screen is "wider" than it must be, so it can't be fitted completely. So by changing the offset, I can only make a picture cut on the left or the right, as seen in the pictures.

Also, as I see in the Wiki, MDA/Hercules uses 720x350 resolution, but the Blaster shows "MDA 752x350" string, same in the Manual-TTL mode.

[scrapcomputing]

Yeah, it looks like the width is missing 4 pixels or so. I will need to look into it.

As for the screen width being reported as 752, this is deliberately wider to mitigate for the inaccurate offset detection. So instead of showing 720 pixels horizontally we are showing 720 + 32 (=752) which includes some black borders. But this is obviously not working as expected given the missing width.

[scrapcomputing]

@yyartsworld Could you try the latest release? I increased the buffer size in MDA. I don't see any clipping on the right hand side.

@Bs0Dd I increased the buffer size for the Pico 2 firmware and it seems to fix the Checkit tests. Sadly I can't do the same fix for Pico 1 because it's running out of memory.

[yyartsworld]

I tried v0.3 beta 8 Pico 1.
the Clipping disappeared.
Thankyou very much!

[Bs0Dd]

Could you try the latest release? I increased the buffer size in MDA. I don't see any clipping on the right hand side.

Yep, tried now and finally it works well!

However, in the HGC Graphic mode, the image is shifted to the right, but I think it's a card "feature".

I also noticed that in Beta 7, MDA mode was broken when observing #28 (and wanted to write about this today); it looks like there were problems with HSYNC width, so it wasn't constant.

I increased the buffer size for the Pico 2 firmware and it seems to fix the Checkit tests. Sadly I can't do the same fix for Pico 1 because it's running out of memory.

Sad! I see that now EGA mode is centered (however, top-bottom borders invisible), but... a bit clipped right side... not so good.

[scrapcomputing]

Yeah I noticed that the MDA image is not centered horizontally too. I will need to investigate this further.

As for the EGA 640x350 clipping, it is really annoying I know. If I mange to save a few bytes here and there I may be able to fix it eventually :D

[scrapcomputing]

I think the centering issue may be related to the VGA timings for 800x600@56Hz. I am currently using:

	Value
H Back Porch	128
H Visible	800
H Front Porch	24
H Sync	72
V Back Porch	22
V Visible	600
V Front Porch	1
V Sync	2

But I don't know if these are correct.

I changed the values a bit and the image did move a bit to the left or to the right, but I could not get it perfectly centered. Perhaps these values are off and the monitor tries to guess the image placement, or perhaps it expects some ON pixels 800 wide and it does not see any so it gets confused and cannot center the image properly. And by the way my monitor reports 800x600@60Hz, which is not correct.

[scrapcomputing]

@Bs0Dd please try out the latest release (v0.3 beta 9), it should fix the horizontal clipping in EGA 640x350. I managed to increase the horizontal buffer size in the expense of the extra vertical buffer size, which is now 0. I also increased the number of offset values to 16. Let me know if you encounter any crashes/hangs (usually caused by out-of-memory errors) or perhaps any vertical image clipping.

[Bs0Dd]

Oops, looks like the situation is worse now.
The right side is still cut (so the fix doesn't work), but now the bottom part of the image is also cut, and there is a blinking line.

beta8/beta9

https://youtu.be/YGJCQlt3OtU

As I said earlier, I can see the full image if I change the pixel clock (in both directions), so maybe it's something related to this?

https://youtu.be/1nwrhIEix2Y

[scrapcomputing]

Have you tried different sampling offsets other than 0? Because in my case I could see the whole image with a sampling offset of 15 or so.

The extra line is strange, I wonder what's causing it.