If I may draw a parallel from the audio world, there are many microphones that have great specs, but only a few that are musical. The flow motion patch to me has a cinematic, visual appeal that transcends the specs. I'm not going to blow up frames and look at still images--no one is going to look at stills in the final product. Flowmotion is a great gift to videographers. We all have lots to choose from.

@Zaven13 Thanks for your praise and recommendations. What I want to emphasize is that rather than experimenting with the FSH/SH peak bitrate, it is better to use custom bitrate settings in FH and H modes. You can freely set the FH/H peak bitrate to anything below 95Mbps without affecting the 100Mbps bitrate of HBR mode. Aside from the peak bitrate, there is no inherent image quality drawback to using FH and H modes - they work exactly the same as FSH and SH modes.

There is also a technical reason why I can't guarantee the reliability of adjusting the FSH/SH Video Bitrate setting. In FM2, peak bitrates of 1080i and 720p modes are determined by balancing a combination of internal settings, rather than simply setting FSH/SH Video Bitrate to the maximum desired bitrate. For this reason, lowering FSH/SH to 88Mbps may produce suboptimal results, such as lowering the threshold of Fallback Mode. (One very notable testing hazard with 720p60 is that you can produce a significant, but spurious, improvement in apparent reliability by prematurely forcing the encoder into Fallback Mode.)

@Zaven13 Since releasing v2.02, I haven't encountered any reproduceable problems with 1080 HBR or FSH modes in either PAL or NTSC frame rates, nor have I seen any issues with PAL 720p50 modes. With 720p60, I improved the reliability of SH mode up to the point where in extreme conditions, the encoder just begins to show a mild and virtually imperceptible form of fallback behavior. (Macroblock QP range begins to expand beyond +/-2.) With this setting, 720p60 mode remains failsafe, and isolated instances of recording failures have been rare and difficult to reproduce.

On your patch suggestions, I'm not seeing the point in restricting the 100Mbps HBR, FSH, and SH modes to a bitrate of around 60Mbps. Since Flow Motion v2 already provides complete reliability at 60Mbps in FH and H video modes, these modes serve as secure alternatives in any situation where it's important to insure that recording will not halt, including cases where 4GB file-spanning is required. And in the case of HBR mode in Flow Motion v2.02, to date I have not seen any reliability issues at all.

At this point, I've hit a high enough bar with v2.02 that further refinement of the settings would require a great deal of time and effort. In terms of reliability, the problem is that I can no longer find test subjects that will reliably provoke the camera into a failure pattern that I can test and analyze. While it's inherently impractical to confirm that any patch is 100% reliable, it's nearly as challenging to determine whether proposed new settings are in fact any more reliable (or less!) than v2.02.

On the motion picture quality side, I've done enough pixel-peeping to confirm that FM2's Quantization Tables work as intended to significantly reduce the macroblock artifacts that plague low-bitrate AVCHD encoders. There are literally hundreds of tweakable adjustments that interact to produce these results, and it would have been wildly impractical to attempt to test all possible combinations. To bring this challenge under control, I used well-established psycho-visual principles to narrow down the possibilities to a working theory. In practice, my hunch turned out to be a good solution, not in general, but in the particular case of the GH2.

Improving motion picture quality further would require coordinating a comprehensive series of subjective viewing tests, with multiple patch variations, test subjects, and reviewers, followed by a statistical analysis of the results. This is how the perceptual quality of lossy audio encoders is certified in professional work, using suites of MUSHRA tests to generate quantitative data. To my knowledge, no comparable methodology has yet been developed for evaluation of video encoders. It would be a far more extensive project than I'm prepared to undertake.

@LPowell I can as soon as I get to a town Lee. My farm is 100% off grid and I get Clearwire using a cantenna. Bandwidth is about like dialup. I have the file ready and will send it up ASAP.

@GH2_fan Reducing Flow Motion v2 to a GOP-length of one will not only cause the videos to no longer play back in-camera. It also degrade the image quality and very likely undermine its reliability as well. FM2's 100Mbps bitrate is geared to produce optimal image quality with a GOP-length of 3. To maintain comparable quality levels with an all-Intra GOP-1 patch, you would need a peak bitrate of over 175Mbps.

High-bitrate patches require complex and often delicate adjustments to operate properly as intended across the wide range of GH2 video modes and options. With Flow Motion v2, the only optional AVCHD encoder customizations I recommend are the 24L Video Bitrate and FH/H Video Bitrate patches. These patches affect only the 60Mbps 24L, FH, and H video modes, and can be adjusted without disturbing FM2's 100Mbps video modes. If desired, you can adjust the 24L and FH/H Video Bitrate patches to any peak bitrate below 95Mbps.

Well I did the same bitrates when using Driftwood's patches and I was able to play them back incamera, and those were the ones that wasnt holding well in 1080i50 at ISO12800 on both High and Low bitrates, thus I find this quite strange.

Hi everyone, I thought i would give the latest Flow Motion a go with a couple of adjustments and it does seem to hold on quite well where I was having problems with Sedna Q20 before at Manual Movie Mode at ISO12800.

With this, I somewhat changed the GOP to 1 on Manual Movie, HBR and Cinema modes and it's bitrates to 100M (Highbitrate) and 50M (Lowbitrate). This as I said holds up quite well for a good few minutes at that ISO but my problem comes when trying to play them back (even after normally stopping record) I get an error message saying that this motion video cannot be played back but it is readable on the PC.

This is quite frustrating that I cannot even preview the video files that I have just recorded. Does anyone know why perhaps? Forgot to mention, this was all done on a 16GB 95Mbit/s Sandisk card. Hope this help! :)

Well...uploads aren't going well. I had a ten second clip that I uploaded to Exposure room but I got: We're Sorry, there was an Error Attempting this Operation Insufficient system storage. The server response was: 4.3.1 Unable to accept message because the server is out of disk space.

That upload took 6 hours so I compressed it and sent it to You Tube.

Anyway, it worked out really well on rough seas. This was still in harbor.

@jrd: Regarding: "Surely somebody is editing native Flowmotion 2 24p footage and can comment on time-line responsiveness?" - Well, I do, and I can comment, but I am not sure whether this is of any help to you, as you might be using a completely different hard- and software for editing.

I can say that using kdenlive and avidemux26 on a dual core i7 CPU with HD4000 GPU (mobile) or a Phenom X4 960 with a GeForce 8800 GPU (desktop) I am satisfied with the responsiveness of the timeline with GOP1, GOP3 and GOP9 material.

@LPowell , thank you, i lowered FH/H to 32, it works great when more card space is needed. I got 2h7min :-)

Lee your tests are NOT equal either. Checkout the sunlight on the top video against your shot. So much more highlights in the recorded picture on mine. Please redo your tests. In fact I will do my own comparison this wknd. Ill be taking 4 GH2s down with the same lenses.

@LPowell Is that the 12/15 GOP, 6 GOP or 3GOP variation of Cluster v4 in your images? Your results differ from my own tests! Also Im interested just to see if the classic GOP structure can handle thee 'scene' that Longjohnsilver is having problems with.