English Amiga Board

English Amiga Board (http://eab.abime.net/index.php)
-   support.Hardware (http://eab.abime.net/forumdisplay.php?f=20)
-   -   What is the real Amiga Maximum overscan in PAL? (http://eab.abime.net/showthread.php?t=45691)

superturrican2 25 June 2009 18:16

What is the real Amiga Maximum overscan in PAL?

what is the maximum resolution that an Amiga *in Overscan* can get?

I am a little confused: :confused

The site of Fellow (an amiga emulator) says that the amiga has a chart up to 768x580 pixels, but while configuring WinFellow it allows me to set up 752x576 pixels.

Through the software of preference "SYS:Prefs\Overscan" allows me to use the workbench at most 724x566 pixels.

a demo that I found (http://kestra.exotica.org.uk/demo.php?id=17557) shows a screen at a resolution of 351x296 / 702x592 pixels.

The Amiga1200 manual says that I can use a maximum of 720x576 pixels in Overscan.

while Wikipedia (http://en.wikipedia.org/wiki/Amiga_500) says that the amiga Overscan has a 700 x 600 (PAL).

What "precisely" is the right resolution? :banghead
(or if anyone has a link to a proper documentation, this is much appreciated)


alexh 25 June 2009 20:17

702x576 is the defined maximum analog PAL resolution (but most CRT's only display 75% of this area).

720x576 is the defined maximum digital PAL (SD) resolution. It defined wider than analog because it has to encompass digitised analog video which is known to drift sideways slightly.

Dunno what the Amiga can do, but anything above those sizes are unlikely to be displayed on any PAL compatible display.


Originally Posted by superturrican2 (Post 564912)
a demo that I found (http://kestra.exotica.org.uk/demo.php?id=17557) shows a screen at a resolution of 351x296 / 702x592 pixels.

How did you come up with that number?

Cammy 25 June 2009 23:09

From using PAL Amigas for years, I've been under the impression that the maximum overscan was 720x576 too. Wikipedia proves itself an unreliable source for information once again, it seems like someone just made those numbers up (700x600).

Sometimes if people want to find out the overscan resolution a game or demo is running in, they'll set their emulator window to a size larger than the possible PAL overscan (800x600 would be fine), then take a screenshot, crop and measure it in an image processing program. This could be how superturican2 came up with the 351x296 resolution.

The Amiga chipset can be tweaked to produce new resolutions anyway, so it's possible demo coders can push the limits a bit more than what the Workbench preferences allow us to choose.

Calgor 26 June 2009 08:39

I think the maximum vertical resolution for OCS for PAL is something like 568/569 there are some relevant calculations in the RKRM Hardware Manual. Using the hardware manual one might also be able to calculate the theoretical horizontal maximum. Or maybe ask Toni :) [Disclaimer: I am not a demo coder or winuae coder so I have not tried anything practical, only theorectical]

superturrican2 26 June 2009 09:40


Originally Posted by alexh (Post 564937)
How did you come up with that number?

Just using StampScreen and PaintShopPro :P
I use WinUAE with a 800x600 window with aspectratio 1:1 *doubled* for Amiga-LowRes games/demo.
that's why I say the true resolution and the doubled.


Originally Posted by Cammy (Post 564937)
Sometimes if people want to find out the overscan resolution a game or demo is running in, they'll set their emulator window to a size larger than the possible PAL overscan (800x600 would be fine), then take a screenshot, crop and measure it in an image processing program. This could be how superturican2 came up with the 351x296 resolution.

yep, exactly! :D

I think this is the only way, or not?

can you recommend a better way (that can do everyone) to verify what's the true resolution of a demo / game?


Originally Posted by Cammy (Post 564937)
The Amiga chipset can be tweaked to produce new resolutions anyway, so it's possible demo coders can push the limits a bit more than what the Workbench preferences allow us to choose.

I found a game that uses a resolution of 352x272/704x544 pixels. The game is Super Gem Z by KAIKO.
It would be interesting to see what is the game and demo that uses the highest possible resolution (obviously in pure PAL - ECS/AGA).

Someone advised me a demo/game where the maximum value of Overscan has been exceeded?

thank you so much for the infos! :bowdown

Calgor 26 June 2009 10:35

See my post here:


Ed Cruse 26 June 2009 19:44

In the include files there's four standard overscan modes listed.

#define OSCAN_TEXT /*entirely visible*/
#define OSCAN_STANDARD /*just past edges*/
#define OSCAN_MAX /*as much as possible*/
#define OSCAN_VIDEO /*even more than is possible*/

OSCAN_TEXT corrisponse to the 'text' setting in overscan prefs.
OSCAN_STANDARD corrisponse to the 'graphics' setting in overscan prefs.
OSCAN_MAX PAL=724x566 NTSC=724x482
OSCAN_VIDEO PAL=736x566 NTSC=736x482

These are the actual display sizes I got when I opened the screens using the above overscan modes. I suspect OSCAN_VIDEO is the biggest the chip set is capable of since it's bigger then what is possible, I could be wrong. :)

mr.vince 26 June 2009 21:38

The TV standard defines 625 lines for 625/50 (commonly called "PAL" because of the color encoding, resolution is identical to "SECAM"). 576 lines are visible, the rest is used for transport of additional data like videotext, closed captions, timecode. The horizontal maximum is 768 pixels (720 is the 601 digital representation with a pixel aspect ratio of 1.067:1, whereas the native resolution is square).

A frame border of 10% is considered action safe, 20% are title safe which means that titles should go into the inner 80%.

I am not sure if the Amiga really was able to output 768 pixels per line, but at least the background color should be able to fill the complete 768 pixels.

525/60 ("NTSC"):

525 lines, 480 visible, 640 pixels horizontal, 720 for 601 digital video with a pixel aspect ratio of 0.9.

Toni Wilen 26 June 2009 22:15

There are no pixels in analog TV signals, only lines. Remember superhires resolution :)

768 should be the maximum possible bitplane width (384 in lores). Easy to test but can't do it today :)

Probably never used because it is far too big for any real display device..

Note that for example AGA sprites and border color changes can be anywhere (of course they won't be visible inside vertical or horizontal blanking)

superturrican2 27 June 2009 13:31

so much information, thanks guys! :bowdown

now I have a better idea, even if this argument is not easy to arrive at a exact conclusion, since the amiga has to interface with a technology that was not formed for pixels matrix.

The same question can affect the C64, where it, through tricks, can use the screen edges (and beyond) for sprites or raster effects.

if I remember correctly, there are also some games for PlayStation 1 that used a high resolution through the Overscan.

@mr.vince tnx for the infos :agree

@Toni Wilen
I think for testing we should imagine a special TV with particolar settings regarding Overscan, or we should open own TV and put hands on the overscan knobs (I dont know how to traslate this to eng), so you can increase or decrease the viewing rectangle.


mr.vince 27 June 2009 14:07


Originally Posted by Toni Wilen (Post 565366)
There are no pixels in analog TV signals, only lines. Remember superhires resolution :)

You are right, Toni, 768 is still considered the maximum, which can not be surpassed due to bandwidth restrictions in the analog signal (speaking of a standard tv set here). I guess I could fetch my tv technicians handbook about tubes from the shelf and do some calculations, but I am pretty sure a standard monitor (in the range of a 1084 or similar) from "back then"will "fail" after trying more than 450-600 vertical (!) lines... That means getting some kind of grey surface where one would expect a black/white checkerboard pattern...

Anyway, with LCDs replacing the old tubes, we are constantly losing the ability to display all the special modes, until only the very obvious ones are left. I can only recommend keeping your tubes as long as they last...

@superturrican2: There is an amazing website I always recommend, unfortunately it's in German only:


Maybe someone can recommend something similar in English. It will for sure take months or even years to fully understand analog video with all the math involved, but it definitely is worth a read.

Ed Cruse 27 June 2009 23:06

Correct me if I'm wrong, but I think for PAL the rastor is always 625 lines, the center 576 lines are used for video, the beam is cut off for the other lines, I think by the blanking pulse. My guess is the Amiga always generates video for the center 576 lines which would include the boarder with the overscan rectangle inside of that. The overscan rectangle should be capable of using the full 576 lines.

I'm interested in knowing if the Amiga can do 768 horiz like Toni says, I think the max for PAL that the Amiga can do is 736x576 if you use custom overscan.

mr.vince 28 June 2009 09:58

Yes, 768 is the absolute maximum of the display device given the fact that you have 576 lines at 4:3... Some of this area is masked by the tv set housing, so yes, all display data goes inside this rectangle. TV sets usually have a much larger overscan area than monitors, which, in some cases, even do a bit of underscan (and show a black border around the picture).

Btw 576 is interlaced, so usually 1/4 of the original resolution is used for video consoles. While 384 x 288 would be the possible maximum (for 625/50, read: PAL and SECAM territories), the modes used are usually smaller.

Coming back to the original emulator question in the first post... one should be very safe setting up a window at 800 x 600.

Ed Cruse 28 June 2009 16:47

Here goes somemore confussion. :)

According to the Amiga hardware manual, because of the horiz and vert blanking of standard PAL and NTSC video signals, the max size is, PAL=736x566, NTSC=736x482. This is what you get using the OSCAN_VIDEO standard overscan listed in the include files.

The display window that is used by the chip set is 512x384 (1024x768), everything is specified in lowres/non-interlace. The hardware imposses a limit of no further to the left then 24 and no further to the right then 472, and there appears to be no limit impossed on the vertical. So, I think the chip set is fast enough to do 448x384 (896x768), this probably allows just enough time for the chip set to do the horiz blanking. Assuming this is correct, then if your monitor is fast enough and has enough adjustment range you can probably do 448x384 (896x768).

Keep in mind this is strictly speculation, the only things that are actually spelled out are the sizes listed in the first paragraph. This is also for OCS and I believe ECS, my book doesn't cover AGA at all, but it's probably the same as far as the impossed limitations.

Does anybody have any words of wisdom here, because I'm definitely not an expert.

Toni Wilen 29 June 2009 14:08

A1200 tests with FMODE=0:

I can see (*) exactly 344 pixels (336+8) when I set all display parameters to maximum possible. (ddfxxx and diwxxx) Total DMA words fetched per line: 25 which makes theoretical and useless max lores width of 400 pixels.

Hires was exactly doubled as expected (50 and 800)

AGA max fetch mode and lores: 7x64-bit DMA fetches = 448 pixels. hires: 13 DMA fetches = 832 pixels. This was totally useless test :)

*) 16:9 LCD TV so 1084 and similar displays may show more or less..

All tests done with single bitplane graphics.

Vertical size is exactly as PAL/NTSC spec says, visible data from end of vblank to end of field.

btw, it is easy to calculate maximum number of visible pixels if you know the length of visible horizontal line (which I don't remember..) and pixel clock (Amiga lores=140ns,hires=70,superhires=35)

Photon 29 June 2009 23:43

I had Asm-One open, so I squeezed the picture on my 1081 together vert and horiz, and made my current demo max overscan (OCS) \o/

Here are the slightly weird results.

I started with the hardware stop in Figure 6-9 in HRM, $d8. After that I tried to find the min start position for bitplane data fetch. The same figure says there's a hardware stop at $18, but no pixels are displayed below ~$28. When using $20 as DDFSTRT, I get a correct 384px wide display, but the first 14 pixels are outside the displayed picture bounds, and there's no way to shrink or move the picture to get them in view.

I made sure to first set DIWSTRT and DIWSTOP to display all ($00...$ff).

(Actually, I could go from $28...$d8 to $2a...$da without any change whatsoever, but that might be that bit 2 is simply ignored, so it could have side effects on other models or modes.)

I've no idea why I couldn't show pixels down to a DDFSTRT of $18; 400px should be possible on other models or CRTs, maybe.

I then tested the vertical limits. It was easy to find the vertical start, just set the background color and decrease DIWSTART until it lined up. The result was $1b. Maximum value was harder, since background color was displayed beyond the DMA fetch stop, but the maximum is $38.

This gives a 384x285 display, although in interlaced mode, probably line $138 only gets shown every other frame.

So, a max overscan OCS copperlist would look something like:


dc.w $8e,$1b00
dc.w $90,$37ff
dc.w $92,$20
dc.w $94,$d8

But in this 384 pixel wide display, only the last 370 pixels were displayed.

Tested in lores, 5 bitplanes normal and HAM, with shift=0 ($dff102)

A normal display line is about 63us. TVs show less than CRTs (i.e., CRTs have potentiometers for adjusting the picture ;)). How much or little LCDs connected to real Amigas show, or how adjustable they are to overcome it, is left to someone more interested in that. :) Well, maybe I will test my PS2 screen...

OK, so I will now make it a point to use the whole 384x285 px in my demo and check what WinUAE hides!! Hehe :P

There are some choices beside the max. You could go 285 in height and use a ratio of 4:3 to get a 380x285 display. That will hide 10px (on a 1081). 368 width gives 276 height, and loses 2px. On the other hand, you have 16px less screen area to clear, and the margin should be safe enough to work on the other chipsets. I'm gonna go with max for now, maybe I change my mind :)

superturrican2 02 July 2009 15:34

good work Photon! :bowdown

soon as possible, can you upload your demo-test? :great

are you the only coder who made a demo to test maximum overscan for 1084, LCD TV, CRT TV and WinUAE/WinFellow ? :spin

TCD 02 July 2009 15:40

/me wonders for what project superturrican2 will use the info... ;)

superturrican2 02 July 2009 22:47

it's not ready as a project for the near future :p
anyway, dont worry, I'll give you the credits :great

on the other hand seems to me that this thread has been a certain curiosity =)

ImmortalA1000 13 June 2018 14:23


If you check out the fifth image here for Digiview 4.0/Gold NTSC advert you can see Newtek allows the selection of 768x480 for an NTSC Amiga.


For PAL it is 768x580 according to NewTek in the French PAL advert for maximum overscan supported. I had the resolution 768x576 in my head for some reason even though I had read that advert many times (late 90s...home computer mags in the bog to make use of wasted time sitting doing not much etc lol!)

Here is an interesting thing, almost all RGB SCART capable portable PAL TVs from a reputable brand will have controls to shrink the screen both horizontally and vertically inside, the older the TV the more likely. So you can actually shrink the displayed image to show more of the overscan just by turning a couple of variable resistors on the motherboard. I think Atari did this with the SM124 monitors at the factory to zoom out the stupidly large C64 style borders on the screen for the ST to make it look less shit than a Mac at the computer show stands.

Also what you have to remember is that although technically some people say that 768 is the maximum horizontal resolution of each scan line due to bandwidth in the signal surely the dot pitch of the phosphor coating and the masking mesh grill dot pitch is also part of the actual resolution to. Weren't they all paired up to give standard 625line PAL TV broadcast standard in use in the 80s/90s.

You surely have to also consider just how good the electronics driving the display on the TV are as well. An Amstrad CD player will never relay as much of the 16 bit data to your speakers as a decent Kenwood or Pioneer etc due to higher accuracy of the electronics.

I had a Sony 14" portable TV at the time of my Amiga 1200, which I had bought together with a shiny new £400 Pioneer top of the range hi-fi in 1987 for my Amiga 1000 because I had got my first job and my friend worked in a high street electrical store and got them on staff discount! The later Sony portable TVs by the way had problems with convergence towards the edge of the tubes but on those early models with the PVM monitor quality tubes you could use workbench in hires lace and not have any convergence issues anywhere on the screen.

Of course using super hires interlaced mode was one of the first things I did and there was a clear difference displaying pictures in those two modes on that TV, perhaps a very subtle difference if I used my Commodore 1084SD monitor from 1989 (Goldstar tube with anti-reflective coating if I remember rightly).

So despite what people say....if you could write a 50hz game in super hires lace using Asteroids style vector graphic type visuals (anybody want to rewrite Cabaret Asteroids!) it would look better on a CRT compared to the same wireframe game engine in just hires lace.

(Of course if you have some rubbish gold fish bowl cheap and chearful portable TV with god awful convergence it defeats the object of it as far as bitmap photo-realistic images displayed in super hires lace via Dpaint IV AGA etc but the vectors would be smoother.)

All times are GMT +2. The time now is 18:27.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.

Page generated in 0.04697 seconds with 11 queries