"Real" aspect ratio?

[akeley]

I am taking some photos-cum-screenshots from my TV for a project of mine. When I measured them later I can see the aspect ratio is not really 1.33 (4:3) but somewhere around 1.43.

A scr$ from an emulator measures at perfect 1.33 (256/192). Meanwhile my real ZX connected to a CRT TV produces the 1.43. How to explain this? (apart from the most likely answer: that I'm a moron and have missed something obvious or my "math" is backwards)

[Guesser]

When did you last calibrate the deflection amps in your telly?

[akeley]

Never. I hate calibrating, it's a rabbit hole, and the image quality/size hasn't really bothered me before.

So, are you saying it should definitely be also 1.33 on my TV? I could try to do some tweaks using the settings menu (no way I'm messing with CRT innards ) and the 240p suite for Genesis PAL.

[Guesser]

I've no idea, just that comparing an emulator against one TV isn't very scientific

Pixel aspect ratio might be mentioned in the ULA book

[Joefish]

I always found Spectrum pixels (particularly ZX81 ones; no colour crawl) lined up very well with the phosphor dots on a CRT telly. Certainly the rows of dots always lined up with the scan lines, and generally the pixels lined up with the phosphor dots too. I don't know if the RGB triple-dots on a TV are perfectly square in ratio or not, but I assumed they were, and it was all calculated to work.

It bugged me that using my ST with a CRT telly, the 320 pixels across the screen didn't line up with the phosphor dots. A checkerboard stipple would have vertical bands where the pixels drifted in and out of alignment.

[Ast A. Moore]

I normally adjust the TV’s geometry using the Spectrum. I draw a circle in BASIC with a single point in the center, and then measure from the center to the edge of the circle doing my best to make sure it’s a prefect one.

According to Chris Smith’s The ZX Spectrum ULA, the Spectrum’s pixels are perfectly square.

[zx81]

According to Chris Smith’s The ZX Spectrum ULA, the Spectrum’s pixels are perfectly square.

I guess that isn't possible. Chris explains that a timing difference between logic gates produces a non-equal pixel when the ULA do a transition from 0 to 1 and 1 to 0. The Woody's demo Brightminer shows an effect not replicated in any emulator, except on Woody's emulator Specemu.

[Joefish]

According to Chris Smith’s The ZX Spectrum ULA, the Spectrum’s pixels are perfectly square.

I guess that isn't possible. Chris explains that a timing difference between logic gates produces a non-equal pixel when the ULA do a transition from 0 to 1 and 1 to 0. The Woody's demo Brightminer shows an effect not replicated in any emulator, except on Woody's emulator Specemu.

No, if that were true, the lines of the display would end up different lengths depending on the pixels.
What is happening is 0-->1 and 1->0 transitions are momentarily early or late, but overall the pixels are still at the same even spacing.

[1024MAK]

Analogue systems like CRT TVs are set up by humans both picture width and picture height). There is also overscan to be taken into account (where the picture is deliberately set up so that at no point does the customer have any part of the tube not displaying a image, because some of the picture information is beyond the edge of the screen). So it’s perfectly possible that the aspect ratio will be slightly different between different TVs of the same make, let alone to other TVs or to a emulator output.

Mark

[4thRock]

For all practical purposes the pixels are square considering a 4:3 display ( 4/3 = 1.3333 so 192*1.3333 = 256 ).
The CIRCLE command in Basic considers square pixels, and all game graphics do the same.

If you do maths based on signal generation, you'll find that they are off by a few percent, but it's too small to notice.
So just set your CRT to display a perfect circle and you are OK.

For the sake of discussion you find rectangular pixels on the QL (notice the game board https://spectrumcomputing.co.uk/zxdb/si ... -run-2.png ) and the ATM (notice the horizontal stretch on the characters https://spectrumcomputing.co.uk/zxdb/si ... -run-1.png), but again the user would correct the CRT aspect ratio as needed.

[akeley]

Analogue systems like CRT TVs are set up by humans both picture width and picture height). There is also overscan to be taken into account (where the picture is deliberately set up so that at no point does the customer have any part of the tube not displaying a image, because some of the picture information is beyond the edge of the screen). So it’s perfectly possible that the aspect ratio will be slightly different between different TVs of the same make, let alone to other TVs or to a emulator output.

Yes, that's pretty much what I though (even if these measurements were nearly identical for 2 different Trinitrons I've tried).

I got in the Service Menu yesterday and chiseled some ~1.33 ratio using 240p suite and ZX measurements. Interestingly, the image was a bit off for PAL, but not NTSC (even though these TVs are from PAL zone)

Aaaand...of course it promptly broke the ratio on my Amiga's Workbench (perhaps otehr machines too). So I will probably just go back to the old status quo. Ah, the joys of CRT tweaking