If you was following my efforts to improve the GH2 quality/stability/reliability, mainly in 720p 60p and 50p, please go to this topic:

http://personal-view.com/talks/discussion/11079/workhorse

to download all best patches after all tweakings, please go to:

http://personal-view.com/talks/discussion/11069/apefos-top-settings-6-patches-for-gh2-hack

These are the two best patches I found. The two are modified FlowMotion 2.02 Both are stable/reliable for recording with the GH2, Both cleans ok with NeatVideo and are great for high iso.

The setj patch settings (modified FM202) uses GOP3 for 1080p 24p and GOP9 for all others. It does not show diagonal pattern (rain) in 1080p 24p, but in 720p the problem exists. The problem is not recorded in the video file, it is a decoder problem in the player/editing software, so you need to use 5DtoRGB free software to convert the mts files to another format before editing, this solves the diagonal pattern (rain) problem. This is for people who wants the highest possible bitrate in 720p and shortest GOP in 720p

The seta patch settings (modified FM202) uses GOP3 for 1080p 24p and the default Panasonic GOP for all others. This avoids the diagonal pattern (rain) in all modes and there is no need to use 5DtoRGB, you can play/edit the original mts files in all softwares without the diagonal pattern (rain). This is for people who wants great quality in 720p but no transcoding before edit/play.

I am keeping the seta patch in my GH2 because it is great quality and no diagonal pattern (rain) in all play/edit softwares, no need to transcode before edit/play.

Try out Spizz T7. Very very good 50p and 60p modes high bitrate throuhout but is very smooth.

Ditto HBR modes.

In my experience analysing these 6 GOP / 3 GOP streams: too high a buffer (eg 95000 on bottom bitrates) although produces better Quantisation but at the expense of poorer i frame sizes, yet, too low a buffer (eg 38000 on bottom bitrates) which works equally great gives larger i frames at the expense of poorer quantisation of the picture. Try it!

Spizz T7 is tried and tested to sit in the middle: 55000 on bottom bitrates (which should be named buffer in ptools!) for HBR/50/60i and 62000 avg for 720p modes along with a measured scaling table rendering out artefacts. Subsequently bitrates stay high even for less detailed pictures and iron out the any mosquito noise/artefact shit. Death charts or very high detailed pictures will take it to the extreme high tops of the allocated bitrate. Other settings Ive looked at just dont produce the balance.

I tested apefos' modified version of Flow Motion v2.02 over the weekend and analyzed the results. I primarily tested 720p60 SH mode and 1080p24 24H mode, but verified that the other video modes functioned as intended.

Apefos stated his main purpose was to produce a more reliable 720p video mode, using FM2's 100Mbps GOP6 SH mode as a starting point. The version I tested uses a GOP9 format with a reduced maximum bitrate. FM2 was designed and tested for use with Class 10 SD cards rated at 30MB/sec or higher, for which it supports 4GB file-spanning in all 60Mbps video modes. Apefos did not claim file-spanning support for his modified version, and I didn't run tests on that feature. My tests all used a Sandisk 16GB 30MB/sec SD card, with a Lumix 14-50mm f2.8-3.5 zoom, set at f8, 1/125 sec, and ISO 200. Shooting conditions were very similar to those I used to develop Flow Motion v2 originally. Here's an example video, shot in FM2 100Mbps 720p SH mode:

I conducted separate reliability and image quality tests of both FM v2.02 and apefos' revised version. In all cases but one, both patches passed all reliability tests. Apefos' version failed to complete recording in one 720p60 SH trial out of a dozen shots of an intentionally overexposed, highly-detailed subject. I had to pop the battery to restart the GH2, but nothing else was amiss. While this may have been a fluke (since I was unable to reproduce the failure), it's pretty typical of what I've experienced with high-bitrate 720p SH mode patches in the past. Consequently, I was unable to confirm whether apefos' GOP9 720p patch is verifiably more (or less) reliable than FM v2.02.

To evaluate image quality, I used cbrandon's StreamParser and Elecard's StreamEye to evaluate average bitrates and encoded quality factors used in I, P, and B-frames. Here is a StreamParser plot of the FM2 720p SH video above:

Vital statistics - FM2 720p60 SH mode, GOP: 6-frames IBP, max bitrate: 105Mbps, avg bitrate: 75Mbps, avg I-frame: 450K, avg P-frame: 156K, B-frame: 80K

This video exhibits FM2 working in 720p mode as intended: GOP6, max peak bitrate, high avg bitrate, large I-frame size, compact P and B-frame sizes. Compare now to the same scene shot in apefos' modified patch:

Vital statistics - Apefos 720p60 SH mode, GOP: 9-frames IBP, max bitrate: 77Mbps, avg bitrate: 54Mbps, avg I-frame: 433K, avg P-frame: 115K, B-frame: 57K

Apefos' patch is tuned to record at bitrates around 25Mbps lower than FM v2.02. This does not necessarily produce lower video quality, since it uses a GOP9 format and comparably large I-frames as FM2. Its statistics reminded me of FM2's 60Mbps 720p H mode, and I shot the same scene in H video mode for comparison:

Vital statistics - FM2 720p60 H mode, GOP: 6-frames IBP, max bitrate: 59Mbps, avg bitrate: 52Mbps, avg I-frame: 394K, avg P-frame: 89K, B-frame: 42K

Both apefos SH mode and FM2 H mode produce similar average bitrates of around 50Mbps. FM2 H mode, however, was designed to insure 4GB file-spanning for long-take static shots that don't require as much bitrate as 100Mbps SH mode videos. In FM2 H mode, P and B-frame sizes are tuned relatively smaller than its I-frames, preserving static I-frame image quality.

To compare the quality of these three shots in finer detail, I analyzed their I, P, and B-frames in Elecard Streameye. The key quality factor in H.264 encoding is the Q-level used to encode macroblocks, with lower Q's preserving higher image quality. In FM v2.02, macroblock Q-level ranges from 17 upwards, with varying Q-levels used in I-frames, and a single Q-level used in each P and B-frame. Here are typical Q-level statistics I gathered from these videos:

FM2 720p SH mode: I-frame: 17-23, P-frame: 19, B -frame: 21

Apefos 720p SH mode: I-frame: 20-24, P-frame: 22, B -frame: 24

FM2 720p H mode: I-frame: 22-24, P-frame: 24, B -frame: 26

From these statistics, we can see how FM v2.02 makes use of its 100Mbps bitrate. Most notable is its ability to encode I-frames at macroblock quality as high as Q17. FM2 uses Q17-Q19 primarily to encode shadow details that would be obscured at lower quality settings. It encodes P and B-frames at relatively lower Q levels, since they contain only corrective details rather than complete images.

While Apefos' patch uses measurably lower Q-factors than FM v2.02, it does a good job allocating its 75Mbps bitrate among I, P, and B-frames. Its performance places it roughly midway between FM2's SH and H modes. All three patches maintain worst-case I-frame quality at high levels, with FM2 H mode sacrificing P and B-frame quality in order to keep its max bitrate below 60Mbps to insure file-spanning reliability.

Yes, I did tests with the 14-42mm kit lens, zoom, auto focus, manual focus.

If spanning is not important use the setj from previous post. If spanning is important try this seti. In this seti I just changed GOP to 9 for 720p 60p/50p and also for 1080i 60i and changed the SH bitrate to 72000000. All other settings are from original FlowMotion.

this ini file setc is the same of previous setb but it changes the 720p and 1080i top/bottom limits to be in same range of the SH and H bitrate limits. To be honest I do not know if these changes will improve stability or affect spanning comparing to the previous setb file. I will not test it because I am using the seth file which is in a previous post here.

@apefos

The modified settings that I worked on is only for HBR30p mode. This was the main concern that I had with all of the patches. I have used Moon T-5/7, Cake V2.3, Sanity, etc etc etc. After extensive testing I found this patch to be the best patch for shooting in 30p mode with a modified Moon T-4 patch as a close second. Attached is my modified setting that will pass the screen of death video in 30p-24p. I have not tested 720p but maybe you could mesh the 2 together then we would have the best of both worlds ;-)

@apefos

I agree Flowmotion is an awesome patch I have my own modified version of that patch on 3=GH2's and I see no reason to use anything else. The bitrates in 24p is moderate and not excessively high LPowell did a remarkable job with this patch.

I found that the shortest the GOP the better the image, but there is a limit before camera stops recording. There is only one patch I did. In the patch GOP 3 is for 1080p 24p and GOP 9 is for 720p 60p. GOP 3 is stable for 1080p 24p and GOP 9 is the shortest GOP stable for 720p. I did not test 60i, 30p or 25p because it is not important for me. All the kudos goes to @LPowell because he did the great job in the matrices and tables, I just tweaked the stability in 720p changing from GOP6 to GOP9 and adjusting the bitrate and quantization.

Hello @TrackZillas I am doing tests before uploading something. I did some shoots under sunlight with everything in focus in iso 160. I used the 14-42mm kit lens. The image is ok. The GH2 720p is as good as Canon 1080p.

I can't believe! The modified Flow Motion makes the GH2 720p 60p usable at iso 12800 and just great at iso 6400. It upscales pretty good to 1080p. I just did a good denoise and add some film grain. There is no reason for me to upgrade for GH3 or GH4 at this moment!