@driftwood
It DOES have to do with the cards.... Sorry after such great news to pee on everyones parade...
After filling the 2 different cards I was sure this was it.... Then tried overwriting... Format the full card and fill it again... Write errors :(
If you need your files to span, I am afraid you will have to follow my instructions from the card formatting thread. Also @grunf has pointed to the tool I also use now, the official SDHC/XC format tool.
My cards were immaculate, so that is why it spanned, if you just format in camera, the files are still there, but marked as zero. As the write happens, it has to unmark and delete before writing a new file, so no luck. I still find it amazing thet 132 will span on virgin cards (or any cards).

@allenswrench- well I could spand for at least 14 mins last night. I am going to test 720p, 1080i, 24p on spanning and ETC modes and see if it spans all the time.

I post results.

P.S. Its pissing down at the mo so the camea goe some work cut out.

For some reason its not working now :-( and I have put the memory card in my laptop.

I am trying to see if you have to keep the memory card away from the laptop and just use the usb cable to transfer the video files. 720p ETC just refuses to write to the card it comes with an error saying "cannot write to card because the limitation of the card".

So I am going to completely format the card like it was new and then I am going to try and see if the spanning works. I may edit the firmare for 720p and just try to up lessen the limit so it does write all to deail.

We will get there in the end bloody thing.

@chrimsbroome Good stuff. That's interesting. How many records did you try before you reformatted?

Analysing the 44M / 32M AQ4 (12 GOP default) settings reports big drops on the buffer across the board on death chart & trees of death - it never crashes but the bitrate/bandwidth seems dire?!!! Ive tried 800 iso to 1600 iso test using 60 to 400 shutter speeds and seen rapid decline on the charts (close to zero) beyond 10 seconds. I am now going to analyse chris's tuned 44M/32M settings.

@chrimsbroome the 55M? This is ratioed to work with 132M/110M on 24p. It should be stable.

@driftwood Why run reverse telecine on 24p source ? Edit : er.. okay, I've read your explanation. What do you mean by 'reverse speed' it, and its already interpreted at 24p right ? And where does pulldown enter into this ?

When it comes to tuning I've discovered that framebuffers and the framelimit have a hard limit that causes the camera to not record at all and require the battery to be removed. Dunno exactly where it is but it's somewhere between 4x and 6x, just an FYI. You know, I think these high bitrates, excluding the GOP1 which is a special case, aren't really worth it at all. They seem to be resulting in bloating P and B frames and not actually improving the image quality at all. Ralph Bs comparisions with HDMI output show that we're pretty much there at 132M GOP 3. As @cosimo_bullo has also noted other strange things seem to occur, such as bitrate actually drops more than running at lower bitrates. I've seen similar things with changing AQ2 to AQ3 with certain settings. Such that the AQ3 version produces streams with lower bitrates on the same shot, but then at AQ4 the streams bitrates are higher again. Odd. Theres something else going on, either wrong or a compensation (throttling) is being done when some conditions are being hit.

@driftwood Thanks for your fast response. Sorry for asking again. What do I gain with GOP1 over GOP3 or GOP6. All settings are high bit rates and looking amazing. Just to understand what the difference are. I have done some shooting at the seaside and the files are 175Mbits after conversion to Proress 422HQ. I have some banding in the sky. This is the 220M GOP6 AQ1 setting. If I change to AQ2 will this look better or can the GOP resolve this? again thanks!

@daimon The motion rendering is better (to some of us) with lower GOPs as the frequency of I-Frames is increased. Also, arguably, image quality increases, but I'm not convinced it does as the codec maintains image quality very well (and more efficiently) with it's B-Frames and P-frames with a GOP of 12 anyway.

If you were to change the 220M GOP 6 to AQ2 it may improve the image quality, but may not cure the banding issue which is mostly down to the 8-bit colour space the codec is working in. Increasing the AQ will also probably affect stability, though pointing the 220M GOP1 at a highly detailed scene may do that anyway in that unless your card is very fast its going to give you write errors. If you check the streamparser output of Driftwoods test with 220M GOP1 in the lowlit room you'll see it only averages about 110M, upping AQ will increase this average. However, an average of 110M is so good you're better leaving the AQ alone, the shadow detail is damn fine already. In short, such a bloody large average bitrate kind of indicates it isn't going to get much better tbh.

Edit : Another effect of a lower GOP is that noise seems to get less smeared and therefore easier to clean up.

driftwood,

I tried the 88/66 AQ2 (Video buffer 24p=0x3600000), it worked with Pappas Death chart /Static and motion / ISO3200 / (1/1600 shutter speed) without crashing.

BUT, on a static scene using pappas death chart, streamparser chart developed a cadence issue after about one minute.

I then tried adding the T4 to T1. This seemed to fix the cadence issue, but shooting static on pappas chart, the i-frames start out large, but for some reason something happens and the I-frames drop down a bit. very strange?
So the T4 to T1 seems to help the cadence issue (viewed in streamparser), but this either causes the I-frame drop after a bit, or there's something else going on.

driftwood,

As much as I like the overall quality of 44M and 66M, I believe when these are set to AQ4, motion suffers a bit and there's visual loss. Raplh, HDMI vs 66 AQ4 motion tests shows this issue clearly. AQ2 seems to be our best bet at these bitrates.

Yes, I'd REALLY like to get (24H 88 / 24L 66) AQ2 settings optimized.

I wish I had a better understanding of how the buffers and limits are used.

So far Chris's 66 AQ2 posted on 8/29 seem to hold up fine on all tests (e.g. i-frames size stay during static shot; No cadence issue (streamparser); AQ2 holds up fairly well on motion).

But, I'd like to see 88M AQ2 optimized! This may be the "good enough" not hogging too much SDHC disk space and also offering real strong static and motion.

@driftwood

The purpose of my tuning parameters was to be able to increase I frame size while avoiding the cadence issue. Stability was not the primary goal - the I frame sizes was. Without the patch to increase I frame size, the other tuning parameters aren't necessary. It doesn't surprise me that stability would be the same without these patches.