remember when this is a new idea - glad someone who is better at these things than me (Nick Driftwood) stepped up to the plate and made it happen.

Was messing around with low GOP settings and rig's GOP1 would crash constantly on my Transcend 16gb Class 10. I loaded This patch and changed Gop to 6 and it worked much better, didn't crash once in similar shots. Set it to GOP4 and....so far so good.

All I did was change 1080/25p to GOP5, 24p to GOP4, and under AVCHD Research I dropped the audio buffer to 4....no idea why...and upped the video buffer to max, 84. Works so far, but only done very basic shots. But it recorded super over exposed and in pure darkness it recorded 3 times as long as the GOP1 patch.

EDIT: Just did a quick test to see how long it would record a all black scene and it went around 7-10 seconds when you start the scene, all black. It also did 7-10 recording a bright all white scene. However, when I started with a standard lit scene then moved to all black it went over 20 seconds before I got cold feet and moved the camera around. And by "This" patch, I mean LPowell's 100 mb Max Latitude patch.
http://www.personal-view.com/talks/discussion/162/gh1-100mbps-max-latitude-patch-v2/p1

sammy asked: "...do you think its best to stay at a gop where we get the best of both worlds ,say about gop of 6? "

Selection of GOP-size is a trade-off between static image quality versus dynamic image quality. With a limited average bitrate, shortening the GOP-size will reduce the bitrate available for encoding image details in each key frame, which will degrade static image quality in highly detailed scenes. On the other hand, a shorter GOP-length will improve the encoder's ability to track rapidly moving objects, which will increase dynamic image quality in highly agitated scenes.

At the 50-100Mbps bitrates that can be reliably used with fast Class 10 SD cards, I've found that extremely short GOP-sizes as low as 3-5 frames can be used on 1280x720 videos recorded at 25p or 30p. 1280x720 videos recorded at 50p or 60p work more reliably with a GOP-size of 15 frames or more. 1920x1080 videos also need moderately long GOP-sizes of 12-15 frames.

Reducing GOP-size to 1 causes the AVCHD encoder to encode each frame as an individual I-frame. As long the amount of data required to encode all the details in each frame does not exceed the available bitrate, an I-frame-only encoder will track all moving objects as quickly as the frame rate allows. However, this mode of operation forces the AVCHD encoder to work more like an MJPEG encoder, and sacrifices AVCHD's most efficient compression techniques. As a result, image quality will rapidly degrade in highly detailed scenes, since the encoder is forced to use excessive amounts of bitrate to redundantly encode the same details repeatedly in each individual frame.

For aesthetic reasons, full-HD 24-25p films are shot with only a limited amount of object motion or fast panning action. For this type of filming, GOP-sizes up to 15-frames can be used without noticeable motion-tracking artifacts - so long as the encoder's maximum bitrate is high enough to handle the amount of image detail required. With the excellent compression efficiency provided by P-frames (and when available, B-frames), average bitrates do not need to reach very high levels. With the most highly detailed scenes, maximizing the peak bitrate is what really counts.

Telkitatchki's GOP 1 opens up interesting possibilities although I can't get smooth video with my cards. The bit rate was high but the video stuttered. There were regularly spaced skipped frames (the upper limit was set at 48mbps, my class 10 cards cant reliably handle anything higher). With GOP15 I only got an average of 20mbps for the same scene.

Mere speculation here - but does anyone think 100mbps is possible on gh2?

> you'll find that I-frames are typically about 3-4 times larger than P-frames...
> I-frame compression is much less effective than P-frame compression...
> Thus, a video entirely out of I-frames would [be] 3-4 times [the bitrate]

Uncompressed P-frames contain less data than uncompressed I-frames, if I get it right, so I do not know if the compression itself plays a factor as well or if it is just the same for both.

The size increase for only I-frames would limit to 3-4 times, I guess, depending on what GOP you use: each GOP already has one "large sized" I-frame, of course...

its possible, but not true 24p, but it rather drop frames

isnt mjpeg at 24p possible with gh13/gh17?

what s the advantage of gop1 avchd over mjpeg?

Not a stupid question to me, I would like to know the answer as well, because shooting with the native 24p blackout powell patch averages in streamparser show a low average anyways, so if I used all I frames OR mostly all I frames I do not know the technical differences enough to know which of the variations would have the best visual output in comparason to the blackout powell native 24p patch. Perhaps someone with technical knowledge about I, P, etc frames and 24p/60p can give details about this idea from telkitachki and how it possibly could (at least in theory) produce better results, or if not now at least in the future when sdhc cards can run faster.

Here is a question, is it better to have lower bit rate with all iframes or higher bit rate iframes w pframes? Im sorry if its a stupid question ..

Well it looks correct in StreamParser, but in the NLE and in VLC and Splash Lite its obviously not correct. 458MB for 1:49 of footage as well

Yeah. If it worked correctly 35mb with just Iframes should be better quality.