It is currently Wed Oct 18, 2017 2:23 am

All times are UTC - 7 hours





Post new topic Reply to topic  [ 150 posts ]  Go to page Previous  1, 2, 3, 4, 5 ... 10  Next
Author Message
 Post subject:
PostPosted: Sat Oct 15, 2011 3:56 pm 
Offline

Joined: Fri Aug 26, 2011 3:20 am
Posts: 41
I went a totally different route and simply reviewed over 100 games on the real hardware hooked up via composite to my display. It took several hours, but I eventually got every damn color to look exactly like the real NTSC hardware (and how I always remembered it looking). I've yet to see an emulator default come as close as mine does to looking like the US NES front-loader's color output.

This is my pal file:

http://www.firebrandx.com/downloads/fbx3pal.zip

I even made my own bsnes 83 build with my own palette edited in since byuu refused to consider it based on it being 'hand-coded'. My opinion is if the colors look the closest to what we all remember in the US, then it shouldn't matter how they were arrived at. I've seen games like Castlevania on a dozen different CRTs, and my palette makes it look exactly like the real one did.

I will state though, that I don't use any filters on my graphics. No scanlines, no filtering, nothing. Filters can of course alter the appearance of the palette, so keep in mind that mine is meant for unfiltered graphics.


Top
 Profile  
 
 Post subject: tool
PostPosted: Sat Oct 15, 2011 3:57 pm 
Offline
User avatar

Joined: Fri Oct 14, 2011 1:09 am
Posts: 248
http://bisqwit.iki.fi/utils/nespalette.php -- create and download tailor made NTSC palette with this tool.

EDIT: The closest match to mr. Firebrandx's palette that can be generated by my tool is by selecting hue -0.081, saturation level 1.53, contrast 0.94, brightness 1.07 and gamma 1.99. (Brute force searched.)
Which is still not much said; the average squared difference between the color channels in the two palettes is still about 172 (average absolute difference 9.53).


Last edited by Bisqwit on Sat Oct 15, 2011 4:50 pm, edited 1 time in total.

Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 4:49 pm 
Offline

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 19093
Location: NE Indiana, USA (NTSC)
Bisqwit wrote:
His seem to be: none, blue, blue+red, red, red+green, green, green+blue, all; which is not binary; it's gray code.

I bet this was on purpose, to make the hues in the emphasis channel follow the same color wheel pattern as the hues in the low-order bits.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:01 pm 
Offline
User avatar

Joined: Wed Apr 02, 2008 2:09 pm
Posts: 1019
Firebrandx wrote:
I even made my own bsnes 83 build with my own palette edited in since byuu refused to consider it based on it being 'hand-coded'. My opinion is if the colors look the closest to what we all remember in the US, then it shouldn't matter how they were arrived at.

I agree your palette is totally awesome. I used it on my pixel artist's guide to NES restrictions, and it beat out many others I found in my search. (Though it was pointed out that the last two columns should be full black. It's quite possible I used a bad version of your palette.)

But( as I understand it) the goal of BSNES is 100% perfect accuracy. It makes sense to me that your palette was rejected because there's no hard logic behind it. Your method would seem to go against everything I've read about the stated goals of BSNES. There are probably at least a few games that play closest to what we all remember in the US on other SNES emulators. Does that mean Byuu should simplify or compromise his code or goals? How good are you at recognizing color? Can you get a perfect score on this test?

Bisqwit: I am fairly impressed by how this looks, and this tool. It'll help me with picking the right palette colors for the pixel art in my game, for sure.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:19 pm 
Offline

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 19093
Location: NE Indiana, USA (NTSC)
I got a 23. Is that good?


Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:31 pm 
Offline
User avatar

Joined: Wed Apr 02, 2008 2:09 pm
Posts: 1019
I don't know what's good (you could have checked your results against others when you finished the test), but I got an 8. I know someone who has got a perfect score, and another who has got a 4.

A perfect score is all that matters for something like a definitive hand made palette for NES. Even if one could perfectly color match, it relies on a specific set of monitors. A logical solution is better than a brute forced one, and can be tweaked to emulate different CRT TVs much easier.

tl;dr: My point was I completely understand byuu's reasoning for rejecting firebrandx's palette.


Last edited by Kasumi on Sat Oct 15, 2011 5:32 pm, edited 1 time in total.

Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:32 pm 
Offline
Formerly 65024U

Joined: Sat Mar 27, 2010 12:57 pm
Posts: 2257
vNES moved to a "more accurate palette" and it now looks like shit.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:38 pm 
Offline
User avatar

Joined: Wed Apr 02, 2008 2:09 pm
Posts: 1019
The NES palette in general is pretty awful. There are a lot of NES palettes I've found over the years that looked great, and I wished were the real one. I like the one from YY-CHR. But that's not how it is. If your goal is accuracy, that's not the kind of thing you care about.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Oct 15, 2011 5:53 pm 
Offline
Formerly 65024U

Joined: Sat Mar 27, 2010 12:57 pm
Posts: 2257
But are the colors in the accuracy? I mean why can't it be an option? Default to right palette and then switch to the one that actually looks good. Colors on the screen are the last on the list for "accuracy" IMO.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 5:45 am 
Offline

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 19093
Location: NE Indiana, USA (NTSC)
Kasumi wrote:
I don't know what's good (you could have checked your results against others when you finished the test)

I looked, but it just gave minimum and maximum scores, not any sort of percentiles. I guess a lot of people click through without even rearranging the swatches.

Quote:
Even if one could perfectly color match, it relies on a specific set of monitors.

NES and PC connected to same LCD TV in side-by-side mode perhaps?

3gengames wrote:
But are the colors in the accuracy?

If you're trying to draw graphics and check on a PC that they look OK before evaluating them on an NES, yes.

But for now, Bisqwit's palette is probably the best we have.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 10:09 am 
Offline
User avatar

Joined: Fri Nov 19, 2004 7:35 pm
Posts: 3943
I think the Nintendulator palette looks pretty good. I might even make it the default palette in PocketNES.

_________________
Here come the fortune cookies! Here come the fortune cookies! They're wearing paper hats!


Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 10:34 am 
Offline
User avatar

Joined: Fri Oct 14, 2011 1:09 am
Posts: 248
Where accuracy is concerned, I am not entirely sure whether the overall hue for my palette is correct. I assumed that it begins from 0.000 degrees, but this assumption may be wrong. It just looks neat in code, and authentic on the screen as well. It may also depend on the particular TV/display device. At least it is very, very close.
As written earlier, at least mr. FireBrandX's hand-picked palette matched mine better (mathematically) when all the hues were shifted left by 2.43 degrees (i.e. a value of -0.081 half-cycles of the master clock was added to the "p" parameter in the sin/cos equation).


Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 12:47 pm 
Offline

Joined: Wed Mar 31, 2010 12:40 pm
Posts: 207
I wonder when someone will do the same kind of thing for PAL signals, and generate a PAL palette.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 1:12 pm 
Offline
User avatar

Joined: Fri Oct 14, 2011 1:09 am
Posts: 248
Yeah, I would like to, but I don't have the information about how either the PAL NES or the PAL circuitry works. Well, the latter I can find out.

http://wiki.nesdev.com/w/index.php/Overscan wrote:
The PAL color subcarrier is exactly 4,433,618.75 Hz. The PAL NES master clock is six times that, and the PPU generates one pixel for every 5 master clock cycles, or 5320342.5 Hz. This makes the width of a scanline 5320342.5*768/14750000 = 277 pixels, and the pixel aspect ratio 7375000/5320342.5 = (59*125000)/(165*64489/2) = 1.3862:1.

http://en.wikipedia.org/wiki/PAL wrote:
Both the PAL and the NTSC system use a quadrature amplitude modulated subcarrier carrying the chrominance information added to the luminance video signal to form a composite video baseband signal. The frequency of this subcarrier is 4.43361875 MHz for PAL, compared to 3.579545 MHz for NTSC.

This makes it deducable that the color generation on the PAL may be largely the same (12 cycles per pixel). On PAL, however, the phase reverses on each line. According to Wikipedia, this phase reversal system existing on PAL (and not existing in NTSC) is why PAL does not need the manual hue correction setting that NTSC users need.
But then, PAL uses YUV whereas NTSC uses YIQ.

Just for fun, I tried the simplest change, and changed the YIQ matrix into YUV, reversed the polarity of the neutron flow, err, the signal phase, and adjusted the offset by integer steps until it looked right. The attenuation/voltage levels were not changed at all. Here is the result for the PAL palette. Naturally, this is 99 % nonsense, only posted for curiosity reasons:
Image
I do have a PAL system (an NTDec, which may not actually do the same processing as the front-loading NES), but I do not have a color TV, so I cannot do comparative tests. Nor do I have an oscilloscope for analysing the generated signals, though I do know how to use it if I had one...

EDIT: Quite fascinatingly, it looks almost identical to the NTSC chart at http://bisqwit.iki.fi/kala/snap/nesemu_ntsccolor_sa.png. I guess the YIQ and YUV conversion matrices are just pure 60 degree rotations/mirrors of each others?
For those who want to repeat what I did, change the "color+p+8" into "color+12-p", and the I&Q coefficients into 0, 1.139800, -0.393800, -0.580500, 2.0279 and 0 respectively.
(Careful inspection reveals they are not exactly identical, though.)

EDIT: The same palette, offset 15 degrees to the left and 15 degrees to the right for the two possible options as per the post by tepples below. Cropped to 13 columns for better clarity in reduced resolution.
Image Image


Last edited by Bisqwit on Sun Oct 16, 2011 2:00 pm, edited 3 times in total.

Top
 Profile  
 
 Post subject:
PostPosted: Sun Oct 16, 2011 1:32 pm 
Offline

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 19093
Location: NE Indiana, USA (NTSC)
Bisqwit wrote:
But then, PAL uses YUV whereas NTSC uses YIQ.

I think the fact that the results are identical comes from the fact that YIQ is pretty much just a phase rotation of YUV. So the "PAL palette" is the same as the NTSC palette at one specific tint and color setting. However, this tint setting might be offset a half a hue unit from the NTSC "neutral" setting. Colorburst on PAL is aligned +/- 45 degrees from the main axis, putting the closest phases at -15 and +15 degrees, while NTSC has a colorburst directly on the main axis. The NTSC colorburst has the same hue as color $x8, but we'll need someone with a PAL NES and a scope to confirm exactly what hue the PAL colorburst is.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 150 posts ]  Go to page Previous  1, 2, 3, 4, 5 ... 10  Next

All times are UTC - 7 hours


Who is online

Users browsing this forum: No registered users and 6 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group