Page 1 of 1
Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 10:11 am
by Lisko
Hi, what is the difference and why both exists? What is the nearest to the real thing?
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 11:19 am
by LamerDeluxe
Lisko wrote: ↑Thu Feb 25, 2021 10:11 am
Hi, what is the difference and why both exists? What is the nearest to the real thing?
I find myself looking for the variation that is closest to the original one, just by going through all options and trying to gauge the difference to the original aspect ratio. I think an automatic mode, like how vertical integer scaling works now, would be best.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 11:50 am
by Yim
It depends on the native resolution and aspect ratio and the upscaling which is closer to the real thing. There’s no one definite answer. I’ve been fiddling with the code trying to implement an auto version that would go to whichever is closer, but I haven’t got it working yet. Not sure if I can write it well enough to be accepted even if I do get it working.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 1:16 pm
by meauxdal
I asked sorgelig about an Automatic HV-Integer option on one of his patreon posts, and from his reply it looks like he's not interested in adding it.
https://www.patreon.com/posts/minimig-4 ... d=54369989
I was checking the situations and i came to conclusion that Auto won't give much but more chaos of random options. Current algorithm tries to follow original aspect ratio. If you use 4:3 monitor then in most cases it will give only one horizontal size as wider version most likely will be wider than screen resolution.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 2:43 pm
by Insert Disk Two
Depends on the core.
PC Engine = wider
NeoGeo = narrow
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 2:55 pm
by redsteakraw
If you are on a wide screen TV wider fills more of the screen and looks better if you are on a 4:3 flat panel narrower would be better as it would fit the whole image on screen. One way to test is to simply hook up a CRT and Flat panel at the same time and mess with the settings until it looks the closest and best to your eyes. I can tell you for the NES I found wider to look better.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 3:53 pm
by Zorlac
On a 4K panel myself and so far wider seems to be the closest on the cores I have tested. An automatic mode would still be a better option though IMHO.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 4:46 pm
by Yim
Lisko wrote: ↑Thu Feb 25, 2021 10:11 am
Hi, what is the difference and why both exists? What is the nearest to the real thing?
Realised I didn’t actually answer the main question here. When you use integer scaling, you’re limited to resolutions that are integer multiples of the original resolution. The mister will use the largest vertical resolution that fits (either full screen or largest integer multiple depending on whether vertical integer scaling is on), and then generate the horizontal resolution by multiplying the vertical resolution by the aspect ratio. Sometimes the result will already be an integer multiple, often it won’t. Narrow scaling rounds the multiple down to the nearest integer, wide scaling rounds it up.
Say for example you’re playing a 320 pixel wide mega drive game on 1080p with vertical integer scaling on and 320 correction off. The vertical resolution will be 896 pixels and the default aspect ratio in the mister 64:49, producing an 1170 pixel wide image, 3.66x the original resolution. Narrow will round it down to 3x, 960 pixels wide, and wide will round up to 4x, 1280 pixels. Wide is closer to the intended aspect ratio in this case, but in other instances it might not be.
Re: Narrower or wider hv integer scale?
Posted: Thu Feb 25, 2021 6:45 pm
by LamerDeluxe
I don't know what is possible in a core, but if you add .5 to the multiplier and round down (floor) you always end up at the nearest integer (round).