@tcarretti please try the seta from the last zip upload in post above, it should work in your card. you can read more details in the txt file or in the yellow box in PTool (mouse over the patch letter).
another matrices option using 4 as dc number: (possible to do the 2000/8000 versions)
Matrix 1:
004 006 008 012
006 010 014 018
008 014 020 024
012 018 024 036
Matrix 2:
004 008 012 016
008 014 020 024
012 020 028 032
016 024 032 048
Matrix 3:
004 008 012 016
008 014 020 032
012 020 040 048
016 032 048 064
Matrix 4:
004 008 012 016
008 014 020 064
012 020 080 096
016 064 096 128
and last combo: (propbably isos around 1250/5000 or 1600/6400 versions)
Matrix 1A:
004 006 008 012
006 010 014 016
008 014 018 020
012 016 020 024
Matrix 1B:
004 006 008 010
006 010 012 012
008 012 014 016
010 012 016 020
Matrix 2:
004 006 008 012
006 010 014 018
008 014 020 024
012 018 024 036
Matrix 3:
004 006 008 012
006 010 014 024
008 014 030 036
012 024 036 048
Matrix 4:
004 006 008 012
006 010 014 048
008 014 060 072
012 048 072 096
Great @Apefos !!! I will try as soon as I can. Parabéns !!!
@Gallo22 thanks, hope you enjoy it on your tests.
@frullaccia reply to your comment about sete PAL vs. last uploaded I: the difference is sete does not have fallback to try to keep same texture all the time on image using the high datarate from PAL. The I has fallback so it is more safe for iso increase and heavy motion, but to avoid the camera activates fallback sete was created for PAL lands to get the best possible image under iso 2000 in hbr and 720p. differences can be minimal, more perceive if using 400% zoom in timeline frame by frame, or a big screen projection.
Versions with DC number 4 are finished.
Same kind of frequency matrix design as previous versions, but using numbers accordingly to dcn4.
This means lower quantization in all similar matrices compared to previous versions with dcn6, so more finer grain and lower iso limits.
Previous versions are here also, renamed to dcn6.
Generally speaking, I consider DCN6 as an improvement, a refinement of FlowMotion, and DCN4 as an improvement, a refinement of Nebula and Smooth Cinema. But all with a complete new matrix design.
Please compare and say which version (dcn4 or dcn6) are better for your image taste and why... (try to perceive differences in noise appearance, noise homogeneity, texture appearance, texture homogenity, light gradients quality, overall sharpness and edge sharpness (local contrast).
Deblocking Tables was adjusted, but not all recording modes and versions was fully tested about the diagonal rain pulsing pattern in the dcn4, if you see it, please report.
my choice: dcn6 2000s
I believe this is the last patch I build.
The idea behind the patch is only one patch work for everything, so I was not satisfied about changing patchs for iso 2000 and 8000, so I did a mix. Now one patch works for all isos below 8000.
Also I did an improvement in the fallback using ideas from the denoise patches.
"The End"
all isos below 8000
two versions, stable and high datarates
@apefos Dcn6 2000s is also very good. Do you think this is better? I still don't know. I will try again. Both versions are excellent!
@frullaccia the last upload is tweaked to work with all isos under 8000 in the same patch. I did careful look at the image and I found that the same patch can do everything. The last upload has the correct main matrix and fallback/deblocking to handle this. Thank you and all PV members for talking in the topics, you all helped to find this patch.
My final conclusions:
GH2 is a great camera and can survive long time, even in the 4k world. For paid professional work iso 800 is the maximum iso limit. Better to keep under iso 320. For situations in low light iso 1600 is the safe limit, 2000 maximum. More than this just for documentary work to get the shoot or for web upload when people will not see the textures so much. Professional work for big screen uses light when needed, or a better camera for higher isos. Even a denoise with neatvideo delivers amateur results no matter the settings in denoise if iso is higher than 2000. Denoise in 1600 iso and below can be useful for big screen with some film grain in post.
The last uploaded patch reaches the camera limit for stability and quality, there is nothing else to do.
Thank you all for talking in the topics and helping to find this settings.
@frullaccia the last upload is tweaked to work with all isos under 8000 in the same patch. I did careful look at the image and I found that the same patch can do everything. The last upload has the correct main matrix and fallback/deblocking to handle this. Thank you and all PV members for talking in the topics, you all helped to find this patch.
My final conclusions:
GH2 is a great camera and can survive long time, even in the 4k world. For paid professional work iso 800 is the maximum iso limit. Better to keep under iso 320. For situations in low light iso 1600 is the safe limit, 2000 maximum. More than this just for documentary work to get the shoot or for web upload when people will not see the textures so much. Professional work for big screen uses light when needed, or a better camera for higher isos. Even a denoise with neatvideo delivers amateur results no matter the settings in denoise if iso is higher than 2000. Denoise in 1600 iso and below can be useful for big screen with some film grain in post.
The last uploaded patch reaches the camera limit for stability and quality, there is nothing else to do.
Thank you all for talking in the topics and helping to find this settings.
@apefos can you please summarize this (final?) patch...i mean what bitrates are used in 1080p, hbr, 720p, for what is this patch good for and no good for, what matrix was used...maybe i´am not only person which is confused from all that stuff you put here...sorry for my english
there is no need to worry about what is inside the final uploaded patch. I did a carefull look in all matrix in timeline in 400% zoom to see the pixels in noise and texture, and the final patch has the best configuration. I also compared it with other patches from other developers. the final patches shows the best image for iso from 160 to 8000. About the datarate, you have two options, to be in the stable way or to use the highest possible, two patches. to be honest the stable datarate is good enough because any iso above 800 will need denoise and add film grain to look good no matter the patch you use. if you are willing to do a big screen projection the safe maximum iso is 800, more than this needs denoise and add film grain in post. also for web upload denoise is a must for iso 1000 and above. trust your eyes about what you see in the last upload showing to you.
I have done a comparison recording among some of the popular patches, including your work. I still have to sort through these video clips. It's just a preliminary testing to figure out which patches to do more testing on. Hopefully I will have some good examples and then post for others to see in a new topic.
I tested 7 of your patches. Some of them, including The Patch/End, truly might win over every other patch. So I thank you for your work @apefos.
Although I wish there was simple way to find bookmarks to learn how these patches and matrices work. Do you have any bookmarks saved for reading about patch work? I look through archive and it's a mess having to read thousands of comments to find just one good comment to learn from vitality or someone else. I am interested in messing around with patching for my own purpose to keep myself busy.
In 640iso, it was the same as a human can see F3.2-3.5 (because of the adaptor could not know exactly, it was not wide open WB 3200K lens in 16mm Tokina 16-50 f2.8
This is a screenshot of comparison I'm playing with in premiere pro. It's a 200% zoom and crop of the each patch listed. http://i.imgbox.com/LTol1WyE.png
Oh my God!!! It seems that my work failed, to my eyes "the patch end" is the worst!!!
@Manicd Just a tip for comparisons: in all patches the noise and texture are different in the I, P and B frames, so when comparing videos it is important to do a frame by frame looking to find the I or P or B frames. also comparing videos in playback and slowmotion is important to perceive differences and quality.
@RdC you can open the ini files using the notepad and read all informations there. also put mouse over the patch letter in ptool, also some information in the txt file inside the zips.
@haribabis thanks for the test
@haribabis Man, if I did that test with my gh3 I would have nasty exposure/gain/noise shift blotches and the underwater swimming noise blotchy effect starting to get noticeable past 1600 ISO. Also horizontal noise patterns. I only saw that barely starting at 6400 here (except no noise patterns). Does the gh2 just not do that? Looks pretty good.
@apefos Not to worry. The patch end is not the worst when viewing as video. I only upload screenshot to show what I am doing in premiere. I will soon have a video comparison finished. We will be surprised by the results ;)
i suppose the 6400 is the new 3200 !!!
@joethepro his video was at 720 and compressed. I can promise you there is plenty of noise in the gh2 above 1600 even with the hacks.
Edit: My browser is messed up, It wouldn't give me option for 1080 but it's fixed now, so never mind about this.
@Manicd i record in HBR (1920x1080) and export also in 1920x1080. The youtube player start with 720p, i think. Also, we have the poor youtube compression. in my opinion, with apefos hack we have more useable footage in high iso.
@Manicd I did another look to your test, very useful. In your test MoonT8 has something better, it has less chroma noise, there is something very different in it's chroma matrix which I was avoiding in my patches in an attempt to get better textures and a more vivid image. To my eyes The End is sharper, but MoonT8 has less chroma noise, but the colors are less vivid in MoonT8 and in some parts the blue becomes more grey on it. Probably there will be a balance between The Patch End chroma matrix and MoonT8 chroma matrix. The MoonT8 chroma Matrix has some similarities with the Denoise patches I developed. In test from @frullaccia in the Denoise topic the Denoise patch shows less chroma noise than The Patch, the chroma matrix in Denoise has some similarities with MoonT8 also. So there is room for an improvement in chroma.
It looks like you're new here. If you want to get involved, click one of these buttons!