[lmyllari]

Thanks, I will try with software filtering and check out the results... based on what I've reported, what do you think my best shot is?

Set full range with xrandr, 16-235 in XBMC, VAAPI and software filtering on. Check that you have all the picture settings in XBMC neutral. This should work because your TV assumes limited range input even when it is told otherwise (like my Panasonic). You should get BtB/WtW and no banding - just like with software decoding.

Once you have the levels correct, check your TV overscan and picture size setting and verify with the resolution patterns that there is no scaling happening. If you see 1-2 pixel scaling (which shows up as some brightness variation and blurriness in the patterns), check XBMC picture aspect setting.


I think the avshd709 patterns are a wonderful free resource and they have a nice PDF guide explaining them. Highly recommended to anyone following this!

[furii]

@lmyllari

i had never bothered with the individual resolution test patterns as i had purposefully set overscan off on my tv and the sharpness & overscan.mp4 showed everything was correct. however, i just now gave it a shot and i am definitely experiencing some squishing on the horizontal. in settings > video > video calibration my corners are set to 0,0 and pixel ratio comes out to an even 1.000. however when i pull up the same settings while the test pattern is playing i'm getting a pixel ratio of 1.001. any idea how/if this can be correct? is this an xbmc bug or something in the intel drivers?

[lmyllari]

i had never bothered with the individual resolution test patterns as i had purposefully set overscan off on my tv and the sharpness & overscan.mp4 showed everything was correct. however, i just now gave it a shot and i am definitely experiencing some squishing on the horizontal. in settings > video > video calibration my corners are set to 0,0 and pixel ratio comes out to an even 1.000. however when i pull up the same settings while the test pattern is playing i'm getting a pixel ratio of 1.001. any idea how/if this can be correct? is this an xbmc bug or something in the intel drivers?

It's an xbmc bug. Monitor size from EDID is used to calculate the pixel ratio and the precision is not high enough to be pixel perfect at 1920x1080. I think EDID uses integer millimeters.

I used to patch my builds for this, but you can also set the ratio to 1.000 using the on screen calibration or directly in the settings file. I think it's guisettings.xml. The ratio is specified separately for each display mode, so fixing it in 1080p60 menus won't fix it for 1080p24 playback.

I'll probably start making builds again once I get the dithering and 3dLUT code ready for my own use, and should include all this stuff too. Everything should go upstream of course, but I may need help cleaning it up and porting to other systems.

[voip-ninja]

Thanks, I will try with software filtering and check out the results... based on what I've reported, what do you think my best shot is?

Set full range with xrandr, 16-235 in XBMC, VAAPI and software filtering on. Check that you have all the picture settings in XBMC neutral. This should work because your TV assumes limited range input even when it is told otherwise (like my Panasonic). You should get BtB/WtW and no banding - just like with software decoding.

Once you have the levels correct, check your TV overscan and picture size setting and verify with the resolution patterns that there is no scaling happening. If you see 1-2 pixel scaling (which shows up as some brightness variation and blurriness in the patterns), check XBMC picture aspect setting.


I think the avshd709 patterns are a wonderful free resource and they have a nice PDF guide explaining them. Highly recommended to anyone following this!

I want to personally thank you yet again as you are an invaluable resource for troubleshooting these really irritating problems that are troublesome to those of us after best PQ we can get.

The recommendation above seems to have been the "jackpot" combination.. the only oddity I observed is that I seemed to get WTW in the test patterns but not BTB... PQ during playback (of at least H264) however looks great and I finally think I'm where I need to be with this.

The only thing to fret about now is that possibly during VC1 playback things will be a bit off since we are forced to use software decode for that currently.

Thanks again a million times over. I think that we need a "Panasonic Plasma" patch that corrects all of this stuff for NUC users who have a Panny TV!

[lmyllari]

The recommendation above seems to have been the "jackpot" combination.. the only oddity I observed is that I seemed to get WTW in the test patterns but not BTB... PQ during playback (of at least H264) however looks great and I finally think I'm where I need to be with this.

Did you try adjusting your TV brightness? You should now be getting BtB and WtW, so one step up on brightness should make the 16 bar light up just barely (which would mean that your brightness control was exactly where it needs to be - 16 is supposed to be black).

A lot of calibrators leave some headroom at the top, so if your TV can reach the target brightness at signal level 235 it can continue getting brighter up until where it runs out of power.

The only thing to fret about now is that possibly during VC1 playback things will be a bit off since we are forced to use software decode for that currently.

Don't fret about it. If you have enough CPU power to decode it, the results should look identical. Both h264 and vc1 are specified so that every decoder should produce bit exact output.

Thanks again a million times over. I think that we need a "Panasonic Plasma" patch that corrects all of this stuff for NUC users who have a Panny TV!

No problem. I think we just need a guide which explains pretty much the last page.

Also, the default settings (gpu auto range, xbmc full range) are already pretty universal and should work in your setup too even if the result is not optimal (some banding, BtB and WtW clipping).

[furii]

i had never bothered with the individual resolution test patterns as i had purposefully set overscan off on my tv and the sharpness & overscan.mp4 showed everything was correct. however, i just now gave it a shot and i am definitely experiencing some squishing on the horizontal. in settings > video > video calibration my corners are set to 0,0 and pixel ratio comes out to an even 1.000. however when i pull up the same settings while the test pattern is playing i'm getting a pixel ratio of 1.001. any idea how/if this can be correct? is this an xbmc bug or something in the intel drivers?

It's an xbmc bug. Monitor size from EDID is used to calculate the pixel ratio and the precision is not high enough to be pixel perfect at 1920x1080. I think EDID uses integer millimeters.

I used to patch my builds for this, but you can also set the ratio to 1.000 using the on screen calibration or directly in the settings file. I think it's guisettings.xml. The ratio is specified separately for each display mode, so fixing it in 1080p60 menus won't fix it for 1080p24 playback.

I'll probably start making builds again once I get the dithering and 3dLUT code ready for my own use, and should include all this stuff too. Everything should go upstream of course, but I may need help cleaning it up and porting to other systems.

that's what i thought as i vaguely remember you mentioning something in a post a while back.. sadly editing the guisettings.xml is not a viable solution as it seems xbmc repopulates the info on system start and video playback. as locking that down is likely to cause more problems than its worth i guess this is a bug we'll have to deal with for the time being. thanks again for all your help.

[voip-ninja]

The recommendation above seems to have been the "jackpot" combination.. the only oddity I observed is that I seemed to get WTW in the test patterns but not BTB... PQ during playback (of at least H264) however looks great and I finally think I'm where I need to be with this.

Did you try adjusting your TV brightness? You should now be getting BtB and WtW, so one step up on brightness should make the 16 bar light up just barely (which would mean that your brightness control was exactly where it needs to be - 16 is supposed to be black).

A lot of calibrators leave some headroom at the top, so if your TV can reach the target brightness at signal level 235 it can continue getting brighter up until where it runs out of power.

Unfortunately the ISF presets are locked (I at one time knew how to go in and tweak things but that was quite a while ago). Even if I change to a different video preset and crank up the brightness it has no affect, so maybe a bug with the VT30 model where with this settings combination BTB is not getting passed.

Picture is good, just would be nice to get that last little bit of range at the bottom. Nothing I can't live with.

Thanks again for all of the help. I might concatenate all of this info over the past few posts into some kind of FAQ but would probably have to get a mod to sticky it or agree it should be added to a Wiki page.

[lmyllari]

that's what i thought as i vaguely remember you mentioning something in a post a while back.. sadly editing the guisettings.xml is not a viable solution as it seems xbmc repopulates the info on system start and video playback. as locking that down is likely to cause more problems than its worth i guess this is a bug we'll have to deal with for the time being. thanks again for all your help.

I was able to fix it by editing the settings. I think I manually shut down xbmc, edited the file and then restarted xbmc.

You can also make the change through the menus - you just have to do it for all the refresh rates you use.

[nickr]

Yes if you want to manually change something in guisettings.xml you need to exit xbmc first. When XBMC shuts down it seems to write the current gui based settings (ie the settings showing in XBMC itself) to guisettings.xml, overwriting any manual changes you made.

[boredandlazy]

Hi Everyone,

I'm considering purchasing an Intel BOXD54250WYKH1 to use with XBMC. I'll be running Windows 7 Ultimate off of a 64GB SSD with 8GB of RAM streaming media via GB ethernet from a QNAP NAS. It'll be plugged into a 1080p Sony Bravia W900 TV.
This is a very long thread so are the any major issues I'll need to consider? Are the issues with 24p on intel GPU's solved on the Intel 5000 graphics?
Thanks.

[-DDD-]

The 24p Issues are solved on all HSW Graphics, also on 4400, 4600 and HD Graphics

[boredandlazy]

The 24p Issues are solved on all HSW Graphics, also on 4400, 4600 and HD Graphics

Oh, it's fixed on the 4400 series as well? My current I3 desktop cpu is a Haswell with the 4400 GPU. I thought it was only fixed on 4600 and above... Maybe I'll try the tiny antec case instead. Although I do like the smaller build of the NUC since it's about half the size.

[furii]

that's what i thought as i vaguely remember you mentioning something in a post a while back.. sadly editing the guisettings.xml is not a viable solution as it seems xbmc repopulates the info on system start and video playback. as locking that down is likely to cause more problems than its worth i guess this is a bug we'll have to deal with for the time being. thanks again for all your help.

I was able to fix it by editing the settings. I think I manually shut down xbmc, edited the file and then restarted xbmc.

You can also make the change through the menus - you just have to do it for all the refresh rates you use.

Yes if you want to manually change something in guisettings.xml you need to exit xbmc first. When XBMC shuts down it seems to write the current gui based settings (ie the settings showing in XBMC itself) to guisettings.xml, overwriting any manual changes you made.

thanks, that did it. i had to do it in a hacky way since i couldn't figure out how to not have xbmc automatically restart on oe. the wiki says to touch a lock file but that doesn't seem to be respected in the current builds. so i had to edit a copy of the xml and then added lines to my autostart to swap them before xbmc started up. like i said, hacky, but it works going through the gui didn't work for me because xbmc would go from 1.001 to 0.999 and up to 1.003.

edit: apparently it's systemctl stop xbmc just in case anyone else was curious.

[voip-ninja]

The 24p Issues are solved on all HSW Graphics, also on 4400, 4600 and HD Graphics

Oh, it's fixed on the 4400 series as well? My current I3 desktop cpu is a Haswell with the 4400 GPU. I thought it was only fixed on 4600 and above... Maybe I'll try the tiny antec case instead. Although I do like the smaller build of the NUC since it's about half the size.

More info in this thread (and specifically this post as far as 4400 goes)

http://forum.xbmc.org/showthread.php?tid...pid1705547

[elim]

can anyone tell me if the i3 nuc is quiet?

i will be sitting around 2m from it and it will be under my tv

will be used for flash streams, xbmc, and web browsing thanks