Page 1 of 1

Scaler issue, expected behavior?

Posted: Wed Aug 12, 2020 9:54 am
by Harrumph
Hi
I have an issue where v-sync =2 only works consistently in 720p mode, whereas it will "bug out" for some cores in 1080p.

It seems related to the refresh rate of the specific core. E.g. SNES works with v-sync=2 in both 720p and 1080p (60.1Hz). While NeoGeo works with v-sync =2 in 720p only (59.1 Hz ? I forget exactly) . In 1080p it looks deinterlaced somehow (vertical height stretches beyond the screen and looks similar to bob deinterlace flicker). In both cases, my TV itself reports the correct resolutions (1280x720 and 1920x1080 respectively).
It also happens on many arcade cores where refresh can be significantly below 60Hz.

From my (admittedly not fully exhaustive) testing, it seems to make no difference which scaling mode (i.e. integer or fill vertical) is used.

I know that v-sync=2 is not expected to have full compatibility, but I thought it weird that my TV is reporting a valid format (1920x1080), but still displaying this weird image.

Is this expected behavior? Anything to test to resolve this?
Cheers

Re: Scaler issue, expected behavior?

Posted: Thu Aug 13, 2020 10:23 am
by paulbnl
Maybe your TV does not like to go much lower than 148.5Mhz pixel clock in 1080p so what you can try is to increase the pixel clock slightly.

Code: Select all

video_mode=1920,122,44,148,1080,4,5,36,150795
This video mode has an increased front porch from 88 to 122 with a pixel clock of 150.795Mhz @ 60Hz. This will go back down to 148.5Mhz @ 59.1Hz.

Don't forget to disable video_mode_ntsc during testing. You can change the video mode per core: Adding Core-specific Settings

Re: Scaler issue, expected behavior?

Posted: Thu Aug 13, 2020 8:47 pm
by Harrumph
Thanks Paul, that was it, it now works perfectly!
:D

Re: Scaler issue, expected behavior?

Posted: Fri Aug 14, 2020 9:52 am
by paulbnl
Nice! Which TV is it?

Re: Scaler issue, expected behavior?

Posted: Sun Aug 16, 2020 5:13 pm
by Harrumph
It's a Philips 43" 1080p TV, 43PFT5301. It has been very tolerant with OSSC, so I was a bit surprised I ran into this problem actually.