ramapcsx2 / gbs-control

GNU General Public License v3.0
784 stars 110 forks source link

Add user controls for output LPF / interpolation filters #72

Closed tomudo closed 5 years ago

tomudo commented 5 years ago

I compare commit 523ca42a2084f25b09a514abf0017d63e54a1dd6 (Date: Mon Aug 20 16:16:42 2018) with latest versioin. I use 240p test patterns in picodrive (megadrive) emulator in PS2. When I test by test patterns -> color bleed check, the old version color is sharp but latest version is blur. I found white background and blue object has a little yellow breeding. Is it possible to look into to picture processing in the old version?

ramapcsx2 commented 5 years ago

I'm often changing the various filters when I discover new things. The current setup prioritizes jailbar and ghosting reduction over static test picture results. This approach works better when playing games, but it isn't set in stone yet.

Some day I can imagine deciding on possible options for this, and then I'll make them available as user options. It's a lot of parameters to consider, especially all the different hardware variations, so it's not simple.

ramapcsx2 commented 5 years ago

Small list of filters and other factors:

ramapcsx2 commented 5 years ago

@tomudo What Picodrive output mode did you test this in? I'm changing the delays for 480p once more. They are not correct.

tomudo commented 5 years ago

It tested with NTSC mode.

I just tested 480p but picture of bleeding test look similar.

ramapcsx2 commented 5 years ago

Okay, that's odd then. NTSC (or PAL) modes either interlaced or not look correct on my PS2. I even use Picodrive now :p

One more question: Are you use Component cables, or RGB?

tomudo commented 5 years ago

I use RGBS with citrus3000 modded and Component cable, however, both have the same result.

I just swap to Aug 20,2018 firmware to test again with all available mode (960p, 720p,480p). The vertical lines are sharper and have all same width in bleeding test. Then I upload latest Jul 25,2019 firmware, the vertical lines are not sharp as old firmware and line width is not equally width.

The blue lines have the most noticable problem, red and green have less noticable, white line has the least.

The pattern looks like these, sharp and normal width- blur and fat - blur and fat - sharp and normal width ...... (repeated)

ramapcsx2 commented 5 years ago

Hm, could you take a picture of the effect? I'm not sure we're looking at the same test.

tomudo commented 5 years ago

Picture from test of latest firmware: https://imgur.com/kumQ9Yj

Picture from test of old firmware: https://imgur.com/3jAgfEJ

ramapcsx2 commented 5 years ago

Okay, this is fine. It is missing the 3_24 VDS Tap6 filter. Enabling the filter will "fix" that issue, but comes with drawbacks such as pronounced jailbars and scrolling artefacts.

If you leave that test screen and run any regular game, the graphics will be good. You can toggle the filter as well, if you want (t3t24t3), but keep the drawbacks in mind.

tomudo commented 5 years ago

I tested filter and have result as you mention. Thank you for your help.

Should I close this issue?

ramapcsx2 commented 5 years ago

Alright :) The issue can stay open. It reminds me to add some form of user control for the filters.

ramapcsx2 commented 5 years ago

Update: I checked just the effect of this filter on 3 boards. Sometimes it tends to increase jailbars and shadows while scrolling, sometimes it doesn't. It really depends a lot on how well the sampling clock (thus, the preset) fits a particular source.

This filter definitely will be a user option soon! :)

ramapcsx2 commented 5 years ago

Option is included now. I've also defaulted the filter to enabled, as it was before.