Versions with mid constants and limit setting - experimental

constant for 1080p = 1572864

constant for 720p = 786432

limit setting = 1048576

hello @konjow @eyenorth @jleo @Aria @Futurity

please, can you give some feedback about the patches in first post of this topic? they are the best tweaked patches until now, no need to try all of them, just one... thanks.

these patches can give you stable recording in all modes, not just 24p.

I did all carefully using research about previous GH1 patches and previous knowledge from GH2. So I believe they are correct. But to know this for sure the patches needs tests. Better thing would be GH1 owners do tests, we are a community and basic idea in a community is cooperation.

@apefos, I will try to do some tests later this week. I have a job that I've got to do 1st.

Please use the last upload called ip8x_abcde, there was typo error in the matrix and now it is correct.

After more research I think the patches below can be stable for all recording modes, ntsc and pal.

all patches are: FHD/SH=32Mbps, H=28Mbps, L=24Mbps, overall 34Mbps, video buffer 40Mbps,

for 24p and 25p all the five patches are the same, wrapped in interlaced, default gops, default constant

for 720p the differences are:

in setA gops for 720p are default and constant for 720p was increased to 460800

in setB gops for 720p was lowered to 15 and 13 and constant for 720p was increased to 460800

in setC gops for 720p was lowered to 15 and 13 and constant for 720p is default

in setD gops for 720p was lowered to 12 and 10 and constant for 720p was increased to 460800

in setE gops for 720p was lowered to 12 and 10 and constant for 720p is default

These differences are for pursue the best stable 720p. When doing maths for 1080p considering the default constant value and default gop length I perceived that the I frames can be 8x bigger than P frames. To get same thing for 720p the five options was created to try. These patches are an attempt to keep the frames size under the stable limits.

I do not have the GH1 to do tests, but considering the research about previous patches development, the GH2 knlowledge and the math calculations, I believe that the setE can be the best one. I believe it can keep the frames size under the default constant values, so maybe it will be stable for both 1080p and 720p. Lowering the 720p gop length is the trick I am using to lower the I frames size, in 1080p the maths shows that the default gop is good for this.

Please use the last upload called ip8x_abcde, there was typo error in the matrix and now it is correct.

the patches and user file below are revised and correct. if they do not work, so I will need the GH1 body to do tests...

here are four patches, variants from ipx8.

they are an attempt to increase the datarate and lower the gop to try to keep the frames size under the constant values for both 1080 and 720. lowering gop too much can make the I frames small and bad quality, so they needs tests in death chart and in low light iso 1600 to perceive the frames size, image quality and stability.

setf and setg are 40Mbps with overall 42 and video buffer 50.

the setf is gop 10 - 8 - 8 - 7

the setg is gop 6 - 5 - 6 - 5

seth and seti are 43.5Mbps with overall 45.5 and video buffer 50

the seth is gop 10 - 8 - 8 - 7

the seti is gop 6 - 5 - 6 - 5

The lower one is stable. Normal settings with iso 1600 spanning. iso 3200 with high shorter do not. On camera screen image looks good but did not review on bigger screen yet.

@knojow thanks for the test, I hope it can give you good images

@yskunto the original panasonic matrix is already good. it does a good share of data between the I and P frames keeping a good size for both. the gop 10 for wrapped 24p and gop 20 for 60p improves the low light a little bit removing some posterization in dark areas. there is no way to lower the gop for pal, it must be 13 for wrapped 25p and 26 for 50p. there is no need to change the constant values, limit, qp and matrix, so the user file is not needed. the last two uploads I did are ok. using it with the exposure trick from dvxuser forum can give decent low light performance up to iso 1600. iso 3200 is too much for the gh1.

I believe I found the almost perfect patch for the GH1. It keeps the datarate high in low light!!! There is no diagonal rain problem in 1080, and works for pal and ntsc.

It is a very careful different tweaked matrix for 1080 and 720 and best gop configuration. I am calling it "advanced plus matrix"

There is no empty frames in death chart or low light in 1080 and 720 for pal and ntsc.

24p is wrapped in 60i, 25p is wrapped in 50i, this is for best stability, needs pulldown removal in post, easy to do.

the setG is a little lower datarate compared to setH, try setH first.

Enjoy!!!

edit post: in next post below a better version is uploaded, go for it.

The maximum iso for the gh1 is iso 1000. it cleans ok with neatvideo and looks like iso 100 after denoise. from iso 1250 and above the sensor starts to show ugly horizontal pattern which does not clean ok and also the color starts to desaturate. to avoid using neatvideo the iso needs to be 640 and below, better 400 and below.

using iso 1000 and below the creative movie mode works good, there is no need to use the exposure trick from dvxuser forum to avoid horizontal pulsing banding.

the vertical stripes can be minimized with a good white balance and using a fast lens with focal reducer, more light reaching the sensor helps minimize the vertical lines, also it is important to avoid underexposure.

the diagonal rain problem is minimized by the patches above in last upload, but not completely avoided, some very small amount almost unnoticeable is present. so converting the videos with 5dtorgb is welcome before editing.

the pulldown removal is easy to do for 24p, all clips shows the same kind of pulldown, no variation in the order.

I believe there is no way to improve the matrix more because the image looks pretty good with good data share between i and p frames keeping same datarate high in low light and death chart test is ok. the quantization and the inter/intra design in the matrix was carefully adjusted. also to completely remove the diagonal rain problem the deblocking tables must be tweaked and gh1 firmware does not allow to do this.