Interlaced picture demo/test (160x384)

[Rybags]

Continuing on from My previous thread about real interlaced pictures

 

ifalcon1.zip Demo picture at 160 x 384 resolution (XEX file).

 

Real machine with 40K or more RAM required - just Load it from DOS or put onto a Binary loader menu disk.

 

This uses an updated version of my interlace system and is a bit more user-friendly to tune. Hopefully it will work on NTSC also.

 

Press [sPACE] to toggle between the settings screen and the picture. Other keys:

 

[cursor keys] change the Delay and Duration parameters to create the interlace.

[D] resets the Delay/Duration parameters to defaults.

turns the interlace on/off. Leave it on of course, unless the screen becomes too distorted to read properly.

[F] is Field Swap. If the diagonal line is out of step or the picture looks fuzzy, use this key.

 

Note that you can use the keys in either screen.

 

 

 

The program is now 100% Assembler. It's still a bit flaky, you might notice stuff that doesn't belong there around the top/bottom of screen but this should be fixed soon.

 

Could people try this out and post what Delay / Duration settings worked best?

Note that you should use the text screen, and 16-stepped diagonal line to verify that in fact you're getting an interlaced display. If you only see 8 steps in the line, keep trying.

[Cybernoid]

Awesome!

[+Stephen]

Continuing on from My previous thread about real interlaced pictures

 

ifalcon1.zip Demo picture at 160 x 384 resolution (XEX file).

 

Real machine with 40K or more RAM required - just Load it from DOS or put onto a Binary loader menu disk.

 

SNIP

 

Could people try this out and post what Delay / Duration settings worked best?

Note that you should use the text screen, and 16-stepped diagonal line to verify that in fact you're getting an interlaced display. If you only see 8 steps in the line, keep trying.

I'll try this as soon as I get home from work (NTSC machine)

 

Stephen Anderson

[_The Doctor__]

works on NTSC just fine, you have to adjust it a bit from the default but works great!

[+Stephen]

works on NTSC just fine, you have to adjust it a bit from the default but works great!

Looks good here too. I really have to try this on a proper TV or my 1084S monitor instead of this VGA convertor.

 

Stephen Anderson

[Bryan]

Looks good at

Delay 13 Duration 19-1E

Delay 12 Duration 1B-20

Delay 11 Duration 1B-20

Delay 10 Duration 1D-22

 

and many others combinations!

 

(NTSC)

Edited February 19, 2009 by Bryan

[Rybags]

Could people post the TV/monitor types also (LCD, tube etc).

 

Glad to see NTSC machines appear to like it.

 

Also, remember to use the stepped line on the text screen - it should have 1/16th character graduations.

 

I'll try and get a couple more pics done today... might even whip up a BMP convertor program so people can try their own.

[Bryan]

Mine's a professional CRT monitor made by JVC.

[Bryan]

I'm wondering if interlace might help with color enhancement in any way. I know the flicker from interlace isn't as bad as the flicker from most high-color tricks. I just don't have a clear idea of how to use them together yet...

 

Most PAL TV's will mix color line to line in a single field. NTSC doesn't do this and can't display those Mode 9/11 pictures, but maybe the overlap of fields might result in some apparent color mixing on NTSC displays.

[Rybags]

 

I've tried a couple more pics, although I'm not totally happy with them. More a case of the pictures I've generated (two alternating screens of data) not being so great than anything.

Note that those are capture card grabs so look a little different to what you'd see on the TV.

 

The Asteroid is in TIP mode. 2 pictures, first alternating palette with luma mode, second one luma with paletted.

The beach one (a FarCry screenshot) is simple APAC mode, although with 2 pictures alternating.

 

The reason I've not included executables... well, at the moment they actually look better without using interlace... so some work still needed.

 

I think I'll give the pics a rest for a bit ... I'm thinking I might do a couple of simple Editor enhancements:

 

. A "super small" text mode with half-sized characters (still 8x8 text though) - this would need a DLI each text line, to use the VScrol trick for shorter characters, 1K needed for character set data.

. A normal sized text mode, except with a nice 8x16 font with serif characters - this would need no extra DLIs, but 2K worth of character data.

[+warerat]

Nice work, Rybags.

 

I was able to get this is to work on my NTSC systems on a Maganvox RGB monitor 80 with settings similar to what Bryan reported, both older and newer ANTICs.

[Rybags]

 

ijor was kind enough to do some oscilloscope captures for me.

 

There's some samples using various settings, showing the area around VSync (actual VSync is the 3 scanlines where the phase is inverted).

 

First one is from a normal situation, remainder are using my program with various parameters.

 

As you can see, it seems we are altering the spacing of the sync pulses, but it's still an imperfect and not well understood process.

Hopefully, more testing and/or analysis of the Antic and GTIA schematics might help to better understand the intricacies of the whole process.

[Heaven/TQA]

keep it going!

 

Maybe Fox or Eru can help in terms of good algorithm for the interlacing stuff?

 

in theory HIPi or TIPi modes would gain 160x384?

[Rybags]

The algorithms aren't really the problem. You can work that stuff out by simple paper drawings.

 

I've tried various combos... the problem is that modes like TIP traditionally rely on alternating the pixel type on the luma line.

 

With interlace, you have your luma line (every line in fact) alternating between one spot and another. So, naturally you either alternate it each frame or leave it the same for 2 frames, then change it.

 

I've tried various combos... having stuff that changes only every 25th of a second just becomes way too obvious.

 

I went back to a HIP pic... I would have uploaded an XEX but I made a couple of mistakes in the generation of the data itself. It's too "jumpy" ie, I've sampled from the master picture at the wrong intervals, and it's got a few pixels that alternate between black and something like Luma 4. Just not happy with it.

[FastRobPlus]

 

I've tried a couple more pics, although I'm not totally happy with them. More a case of the pictures I've generated (two alternating screens of data) not being so great than anything.

Note that those are capture card grabs so look a little different to what you'd see on the TV.

 

The Asteroid is in TIP mode. 2 pictures, first alternating palette with luma mode, second one luma with paletted.

The beach one (a FarCry screenshot) is simple APAC mode, although with 2 pictures alternating.

 

The reason I've not included executables... well, at the moment they actually look better without using interlace... so some work still needed.

 

I think I'll give the pics a rest for a bit ... I'm thinking I might do a couple of simple Editor enhancements:

 

. A "super small" text mode with half-sized characters (still 8x8 text though) - this would need a DLI each text line, to use the VScrol trick for shorter characters, 1K needed for character set data.

. A normal sized text mode, except with a nice 8x16 font with serif characters - this would need no extra DLIs, but 2K worth of character data.

 

Very nice! I actually recognized the second image as a FarCry screenshot before reading your post.

[Xebec's Demise]

These images are on an Atari 8-bit?!

 

Really nice!

 

[Rybags]

 

Update.

 

I went out and bought an oscilloscope the other day... the other one I have is practically useless.

 

Now I can get some nice low-down views on what's going on... made a modified version of the program that can trigger a sweep by outputting on the joystick port at strategic times.

 

That's a view of some HSync pulses with the default settings... obviously a bit out, so some work needed there. I was able to establish a near 50/50 spacing (frame/frame relative) pretty quickly, but we still have the unwanted diagonal interference at the top/bottom of the TV to try and get rid of.

[danwinslow]

Which scope did you get?

[atarian63]

Continuing on from My previous thread about real interlaced pictures

 

ifalcon1.zip Demo picture at 160 x 384 resolution (XEX file).

 

Real machine with 40K or more RAM required - just Load it from DOS or put onto a Binary loader menu disk.

 

This uses an updated version of my interlace system and is a bit more user-friendly to tune. Hopefully it will work on NTSC also.

 

Press [sPACE] to toggle between the settings screen and the picture. Other keys:

 

[cursor keys] change the Delay and Duration parameters to create the interlace.

[D] resets the Delay/Duration parameters to defaults.

turns the interlace on/off. Leave it on of course, unless the screen becomes too distorted to read properly.

[F] is Field Swap. If the diagonal line is out of step or the picture looks fuzzy, use this key.

 

Note that you can use the keys in either screen.

 

 

 

The program is now 100% Assembler. It's still a bit flaky, you might notice stuff that doesn't belong there around the top/bottom of screen but this should be fixed soon.

 

Could people try this out and post what Delay / Duration settings worked best?

Note that you should use the text screen, and 16-stepped diagonal line to verify that in fact you're getting an interlaced display. If you only see 8 steps in the line, keep trying.

Very Cool! You are amazing!

[Rybags]

The 'scope is branded "Digitech".

 

The manual is Chinglish and I can find next to no information on it or that brand. There is "Digitech" musical stuff like amps, pedals etc. but I'm 99% sure that it's in no way associated with this brand.

 

It's a 10 MHz job... bottom-end these days. I really would have liked one of the USB logger devices but the cheapest one is $299+ and it only does 200 KHz which is what I was getting on the sound card input anyway.

[candle]

funniest thing in those manuals is no matter what target language was, result is more or less still chineese, for polish manual it would be chilish i suppose

with 10mhz analog scope you should get pretty nice looking waveforms, but need to work on triggering

 

and for chineese manufacturer i wouldn't be so sure evevn if You would find they produce teddy bears, or diapers

[Rybags]

I've taken a bit of a step back to the older version... it gives me a near perfectly stable screen, so I'm doing some caps of that to see what it might be that I'm doing right.

 

 

 

I've modified the program so I can "Hold" on a particular field type.

 

First pic is of a normal everyday Atari VSync. Note that it's from composite video so some colourburst gets in on the action, but it doesn't interfere with what I'm trying to see.

 

The second pic is of a forced "interlace" field. I can hold the program on that type so that I can get a clean 'scope display.

 

Note the VSync pulse becomes somewhat fractured. Also there are a few scanlines where 2 HSync pulses are being sent instead of just one. The pics don't show it so well, but you get the second colourburst as well.

 

 

Another funny "trick" I discovered last night. With the right timings, you can actually make the VSync signal virtually disappear. Of course, all you get for your troubles then is a screen that constantly rolls.

Edited March 3, 2009 by Rybags

[Rybags]

I managed to modify the Plus/4 interlace program tonight.

 

Also got readings from it in standard mode.

 

The Plus/4 actually generates a more standard VBlank/VSync signal pattern.

 

Unlike the Atari, which just outputs a simple VSync of 3 scanlines of Sync Level with Blanking Level where the HSync pulses normally are, the Plus/4 actually does it by the book.

 

The 3 scanlines leading up to, and following VSync have the extra HSync pulse in the middle of the Scanline. Also, the HSync has the extra pulse (inverted) in the middle of the scanline.

 

The Plus/4 interlace system merely shortens every second VSync from 3 to 2.5 scanlines to achieve it's goal.

Edited March 4, 2009 by Rybags

[Laemeur]

Very cool demo.

 

I get a good result with delay around 2F, and duration around 14. That's on an NTSC XL hooked-up to a Commodore 1702. There are a lot of tunings where the vertical gradation is good, but the horizontal and vertical spaces in the line segments look best (so far) around that area.

[Rybags]

Thanks for that.

 

I've got to rework the whole thing... the scope help out there. Funnily enough, something that looks good on the TV usually looks good on the scope too.

 

I might go the overkill route and see if I can force the extra sync pulses before/after VSync - from Pic 2 in that previous post, you can see that it's almost there.

 

That particular setting from the INTV3.BAS program also gave best results on the 4 or 5 TV/monitors I've checked on myself.

 

 

I should add - people have on numerous occasions posted about having LCD TVs that just won't work properly with the Atari. In theory, a variation of this technique might be able to generate a stable standard display.

 

Although of course, that's of limited use since you can't really hack it into every bit of software you use.

Edited March 5, 2009 by Rybags