Lcd Colour Correction Test Programs

[A_SN]

[cut]This thread is dedicated to compiling the results from GP2X MKII screen colour test program (as a follow up to this year-old thread) I'm hereby releasing. [/cut]The LCD colour correction test consists in a series of simple tests designed to find the precise colour characteristics of GP2X screens and to correct them. The test comes in two parts, one SDL program you run on your PC, and one program you run on your GP2X.

Both programs display the same colours, the goal is to make the colour in the center of the screen on the GP2X match as closely as possible to the one on the PC screen. [cut]It assumes that your PC screen has a decent colour calibration. One thing that is essential to remember is that your eyes shgould be perpedicular to the GP2X's screen, which means you should see the reflection of your eyes right in the middle of the screen. Additionally you may want to turn all the lights around you off.

So here's how it goes :

-Download the PC program (Windows and Linux versions as well as the source that is compiled with the command 'gcc pc_colour_test.c -o pc_colour_test `sdl-config --cflags --libs` -lpng')
-Download the GP2X program, copy on SD card, launch
-The screen fills with a red gradient. Move the joystick from left to right to move the vertical black bar to the point at which the gradient stops on its right hand (and brightest) side and becomes uniform bright red. Then press A and do the same for Green, Blue and White gradients.
-Launch the PC program. Both the PC program and the GP2X programs now say "Number of points: 4). We'll leave it to that, as we want to all use the same number of points in the curve. Press Return on the PC and button A on the GP2X.

-Now you see both screens divide in three rectangles. A black rectangle, a dark red rectangle, and a bright red rectangle. The goal is for you to make the middle rectangle on the GP2X more or less bright by moving the joystick left or right, and press A when you're satisfied with the resemblance between the colours on both screens. Please remeber that the middle rectangle is not always a 50% mix of the two other rectangles.[/cut]

[cut]-There's 4 such tests for each colour channel, which means a total of 16 tests. Make sure that you're always on the same test on both programs. On the PC you can go back and forth using the arrow keys or the Return and Backspace keys.

-When you're done, the PC program will simply quit, and the GP2X program will display its results, which look like this :

CODE
Red Green Blue White
204 200 196 200
127 126 102 120
101 87 77 90
167 162 138 157
185 183 164 178


-Write them down and post them here. If you don't "feel" the White test, feel free to not include its column when you report the results.

If you wish you can run the test many times and average each value. Just make sure to mention how many times you run the test so that I can appropriatedly weight your results in the final average (I won't make it proportional but the square root of it, as running a noisy test N times only makes it sqrt(N) times less noisy).[/cut]

Note that this test program will not actually attempt to do any sort of colour correction. It will be the object of the later release of colourd (the colour correcting daemon, see 2nd post of this thread) and will be made part of the final release of the Open2X firmware as a menu option.

[cut]As a result of this thread, once I'll have compiled all the results into a weighted average, we'll get a very precise colour correction.

Important note : make sure to report the model of your GP2X. MKI's don't have the same LCD as MKIIs (I don't know what F200's have).[/cut]

 

[A_SN]

And here is the actual colour correction daemon, colourd.

Simply copy the folder over to the SD card, and launch the script colourd.gpe. It uses the settings (in the file named 'colourd') I posted above (for a F100 MKII screen), calculated by giving a weight of 1 to Red and Blue, 2 to Green, and 4 to White. Note that the numbers aren't in the same order. The first number is the number of points (here 4), and the other values are sorted ascendantly.

EDIT : It nows uses the final figures obtained thanks to this thread

 

[Alex.]

F-100 MK2,
CODE
Red Green Blue White
204 212 208 208
108 102 102 102
54 51 38 51
153 153 153 153
178 170 167 177

I only found the 4th test of every colour channel to require significant calibration. Hope this helps, and hope more people will submit their numbers. Thanks for doing this

Edit: fixed first line.

 

[senquack]

f100 MK2 test:

r g b w
228 212 220 224
142 141 107 138
92 90 69 93
178 168 151 157
193 192 187 190

f200 MK1 test (unit purchased Feb 2008, says ( B ) on back but shipped w/ firmware 4.0.0):
r g b w
240 236 236 240
136 131 107 128
94 76 68 60
168 157 153 166
199 200 193 202

 

[A_SN]

F-100 MK2,
CODE
Red Green Blue White
4 0 0 4
108 102 102 102
54 51 38 51
153 153 153 153
178 170 167 177

I only found the 4th test of every colour channel to require significant calibration. Hope this helps, and hope more people will submit their numbers. Thanks for doing this

Thanks a lot! However there's a problem with the first line, which matches to the very first test! All values should be around 200, yours are around 0. Crap I guess it wasn't clear that I was talking about the bright end of the gradients, on the right, not the dark end! I'll edit my post to make it clearer.

 

[Alex.]

Thanks a lot! However there's a problem with the first line, which matches to the very first test! All values should be around 200, yours are around 0. Crap I guess it wasn't clear that I was talking about the bright end of the gradients, on the right, not the dark end! I'll edit my post to make it clearer.

I assumed since the vertical bar was black, I was to see where it started to blend in... Sorry about that, fixed.

 

[Metalbrain]

F-100 MK2,
CODE
Red Green Blue White
208 192 204 208
142 128 115 126
 88 75 66 66
172 158 157 158
185 188 185 188


The white tests were the hardest to match because I saw a blueish gray in the gp2x, compared to the pure grey of my PC screen.

 

[fusion_power]

And here is the actual colour correction daemon, colourd.



Any improvements compared to the last release?
I always use your little tool for various Emulators, great program. It brings better colours onto my F100 MK2 especially in combination with GPSP or Final Burn Alpha etc. Picodrive has this Gamma Curve built in, very useful.

 

[A_SN]

And here is the actual colour correction daemon, colourd.



Any improvements compared to the last release?
I always use your little tool for various Emulators, great program. It brings better colours onto my F100 MK2 especially in combination with GPSP or Final Burn Alpha etc. Picodrive has this Gamma Curve built in, very useful.

Tremendous improvements! Two improvements really. First one was the smoothing function. Back then it was a gaussian interpolation, made the curve overly smooth. Now I use cubic interpolation, much better. The other thing is, back then, it didn't worry about burnt colours, so it wouldn't fix the problem. Strangely enough the first problem made up for the second, but well you still had burnt colours. Not anymore, provided you correctly chose the burn limit (the first series of test).

 

[Lithosphere]

F-100 MK2,
CODE
Red Green Blue White
208 208 200 208
121 102 104 103
68 64 51 52
153 146 152 152
192 179 195 205

 

[fusion_power]

Tremendous improvements! Two improvements really. First one was the smoothing function. Back then it was a gaussian interpolation, made the curve overly smooth. Now I use cubic interpolation, much better. The other thing is, back then, it didn't worry about burnt colours, so it wouldn't fix the problem. Strangely enough the first problem made up for the second, but well you still had burnt colours. Not anymore, provided you correctly chose the burn limit (the first series of test).

I've tested the new version and the Difference is big!
Gmenu2X is more washed out than with the old version but then I tested GPSP and Advance Wars2. Colours are not that bright and shiny but I think the current colourd is the most realistic compared to the original Game appearence. In AW2 now you can see every little colour gradient even nearly invisible stuff like the background lines/graphics/letters in the info windows:

Without the tool, all this BG graphic elements are nearly invisible because of the strong saturation of the F100 MK2 screen, you only see a plain BG colour.

Good Work.

 

[A_SN]

Tremendous improvements! Two improvements really. First one was the smoothing function. Back then it was a gaussian interpolation, made the curve overly smooth. Now I use cubic interpolation, much better. The other thing is, back then, it didn't worry about burnt colours, so it wouldn't fix the problem. Strangely enough the first problem made up for the second, but well you still had burnt colours. Not anymore, provided you correctly chose the burn limit (the first series of test).

I've tested the new version and the Difference is big!
Gmenu2X is more washed out than with the old version but then I tested GPSP and Advance Wars2. Colours are not that bright and shiny but I think the current colourd is the most realistic compared to the original Game appearence. In AW2 now you can see every little colour gradient even nearly invisible stuff like the background lines/graphics/letters in the info windows:

Without the tool, all this BG graphic elements are nearly invisible because of the strong saturation of the F100 MK2 screen, you only see a plain BG colour.

Good Work.

Yeah the problem is going to be with programs that were designed to look good on MKII screens. And the screen isn't over saturated but more like each channel is burnt out around a value of 200, so gradient variations are destroyed. There's nothing that can be done about saturation, only about contrast and luminosity or even gamma.

 

[A_SN]

I have to say I expected a lot more participation than this.. I'm not sure what can be done with the few results we've got.. that's fairly disappointing.

 

[paulobecker]

Here are my results: (F100 MK2)

CODE

Red Green Blue White
208 204 200 208
133 141 148 156
87 98 117 115
172 173 180 186
194 193 193 198


Could the fact that I did this using a laptop LCD screen have any impact on the result?

(edited: first line was wrong because I'm stupid and didn't read the instructions properly)

 

[A_SN]

Here are my results: (F100 MK2)

CODE

Red Green Blue White
208 204 200 208
133 141 148 156
87 98 117 115
172 173 180 186
194 193 193 198

Could the fact that I did this using a laptop LCD screen have any impact on the result?

(edited: first line was wrong because I'm stupid and didn't read the instructions properly)

Your results look correct. Don't worry, most likely anyone here uses an LCD screen that's far from calibrated anyways. I know I tried to calibrate mine but couldn't even satisfy the test's pre-requirements regarding luminosity and contrast.

 

[Alex.]

Only 5 people bothered to spend 2 minutes for the cause of making a great piece of software even better? Come on!

 

[senquack]

Only 5 people bothered to spend 2 minutes for the cause of making a great piece of software even better? Come on!

Everyone is busy mindlessly obsessing over electronics that aren't available yet.

 

[hessiess]

Here are my results: (F100 MK2)

CODE

Red Green Blue White
208 204 200 208
133 141 148 156
87 98 117 115
172 173 180 186
194 193 193 198

Could the fact that I did this using a laptop LCD screen have any impact on the result?

(edited: first line was wrong because I'm stupid and didn't read the instructions properly)

Your results look correct. Don't worry, most likely anyone here uses an LCD screen that's far from calibrated anyways. I know I tried to calibrate mine but couldn't even satisfy the test's pre-requirements regarding luminosity and contrast.


My monitor is as accurate as I can get it by eye, though I don't have a GP2X. Monitor calibration isn't difficult, it can just take a long time, and a bit of practice. Just make sure you adjust the monitor and not the drivers where ever possible and you should be OK. This guide explains the process very well: http://www.normankoren.com/makingfineprints1A.html

also:

• Glossy screens tend to under saturate reds.
• most new LCD displays come with all the settings at max. This is almost always too bright, especially with 'ultra bright' displays. brightness should be set relative to your environment, so max brightness would be OK in a conservetoy or outside for example. as a referance, I had to set my monitor to around 10% brightness to get it displaying remotely accurately.
• LCD's have trouble displaying dark blacks, this is especially true if your monitor is too bright.

 

[Adventus]

F200 Mk1 FW4.0.0 (same as Senor_Quack's):

CODE
Red Green Blue White
216 204 200 208
145 136 157 164
97 86 112 111
166 169 186 190
196 195 198 207

Edit: It should be noted that i did this test in moderately high ambient light.

 

[A_SN]

Thank you all, here are the results I obtained (this is the config file, the values you put in file 'colourd' in the colourd folder) :

CODE
4
80.36
126.74
162.90
188.05
207.05



And that's also the values that will be used by default in Open2X when colourd will be turned on (turned off by default)

EDIT : I just tried these values and it looks pretty good. Good job people, I think we nailed it within 1 or 2 pixel value units on each point of the curve!

Oh and I've just updated colourd.zip with these values too.