@sohus Yes I have, and I've shot with it for a few days with no problem. I still haven't tried it at all at AQ4 though.

@bkmcwd Nice one, those graphs look good. I wish I could give it a go and shoot with it now, but can't really till I get all this work out of the way, or get a decent break. When I post here at the moment it's usually cause I have some rendering running in the background. I wouldn't worry about the result from the extreme test at all, it is unlikely to get near that sort of grief in real world shooting. In fact your moving through the trees footage is a lot more useful tbh.

My only concern is what looks like a flatline between frames 440 and 480, any ideas of the cause ? As for the T4 T1 thing @vitaliy said it would cure cadence problems, and thats pretty much all I know about it. But as I and @proaudio4 have noticed, on lower bitrates it can cause a large drop in frame sizes (and then continues in a stable fashion) a few seconds in. So, a few seconds in when used on lower bitrates it can drop its QP quite a lot, but with high bitrate patches 100M+ the T4 T1 shuffle doesn't seem to exhibit this drop. Have you tried your patch settings as they are now without it ?

Is there any way to combine these settings with say, CBrandin's or mpgxsvcd's stable settings on the "low" setting? That way we can retain spanning, autofocus, etc. and also have the option to use driftwood's settings when applicable.

Doing some tests right now with all the patches in my vault.

@Driftwood GOP3 AQ2 132M is not stable for me. It starts with 1.010.000 iframes and than drops to below 800.000 and stays there. iframes are the same size as b frames and not p frames in the stream.

@Driftwood GOP1 NOAQ 176M QUANT ME BABY. I noticed in my tests that 24H and 24L have the same effective bitrates and iframe sizes. 24L looks more stable to me as the extra bitrate in the H setting is apparently wasted.

@CBrandin 66M GOP12 AQ2 (and variations) is my favorite. Very stable and good frame sizes,stability and recording times.

@Sohus 132M there were two versions. I'll take a look. Can you try out the two 176M GOP3s - I'd like to know your reaction. Thanks

@Stray
Thank you for commenting, although you are busy!

"looks like a flatline between frames 440 and 480"

Although I do not understand a reason, I think that it is the same as the flat line which has come out in the highest stress test.
If it adjusts so that this may not come out, whenever it sets to the same level as AQ2 or it in 3GOP, condition that a write-in speed error will appear like my 176M AQ2 shown before.
I think that this phenomenon will be unavoidable if the same level as AQ2 or it is used in 3GOP.

Although I had, of course, also tried simultaneously the setting which removed T4, I did not find a difference. :-(

I was able to test this setting only a few in the town of night today.
About the degree of the frame size drop, if not bad, I feel.

Some more observations:

@Driftwood GOP1 Quant Me Baby = slightly larger than stock settings max video frame size. In other words, with this codec you achieve the highest single frame of the stock codec continously. This is very good performance wise. And Max = AVG = MIN so the quality is good throughout the 24 frames per second. It shows.

@cbrandin 66M & 66M MAX VARIATION. They are very close in my test. I do believe that the MAX VARIATION patch is slightly more efficient and in very complex scenes can direct the bandwith to the right properties. You can't go wrong with either of these. I think these give max. performance for min. bitrate (66M) and both are very stable.

I do have problems with the FSH setting in this patch. Was it made for 50i or 60i? Is it possible to use a 720P setting from another patch and put it in there without negative affects on the 1080p settings?

Nick, I can look at your other patches. Which ones do you mean? The last two posted here?

Test results (I didn't test all my patches):
https://docs.google.com/spreadsheet/ccc?key=0AuvH0y9LX_pkdFJHdnA2UnBnSEY2WlpSWVZjUGdVRHc&hl=en_GB
(sheet TESTING SOHUS)

No graphs but easy to compare the numbers. I look at cadence so all of them passed, only the @Driftwood 132M did not show even cadence (started high, dropped, stayed there).

@driftwood Yep I was just saying with a GOP3 for example that you should have much higher I-Frames and relatively smaller B-Frames to really take advantage of it being GOP 3, like the patch bkmcwd is working on. B-Frames can outstrip I-frames under heavy motion, but they really shouldn't on a deathchart test. The results of the GOP3 you've posted don't look good in comparison (I-frame size is too low), and the high B-frame size are worrying and wasteful (as in it may as well be a GOP1). I completely agree though that the GOP 1 is a massive achievement and I appreciate entirely its advantages when it comes to motion. But all of my tests lead me to conclude the sweet balance between image quality and motion rendering is going to be somewhere in a GOP 3 setup. I think bkmcwd is close to that at the moment, although it still has some stability problems. Maybe the stable frame sizes will fall some way between what you're achieving and his, though hopefully at a bitrate more around the 150M mark than the 170M.

Yeah, I cant check/test much right now. I'm stuck in a laborious loop of prepping footage interspersed with mind numbing manual creation of roto shapes in order to isolate elements within shots. Labour intensive not intellectually taxing work. You've altered the QP values since you've uploaded it ? Serioiusly, I'd rather be testing yours and bkmcwds patches and being on the tweak over doing what I'm doing now.

@bkmcwd Yeah, the T4 thing I think wouldn't solve the flatline, it was to solve the pulsing cadence issue IIRC. All I can think of is slightly pulling back the top/bottom settings a bit. I mean this 154M GOP3 is really, really close now.

The 66M is not as good as Chris's 44M for stability all round.

@Stray
Thanks! I got it.
Although it is a result at the time of lowlight that I am the most interested, since a result was not different in ON/OFF of T4, I do not care.

Since the value 154M was best balance with 3GOP in my having tried in the range out of which a card speed error does not come from 100M to 176M, I chose.
Since according to my feeling writing stopped when the Average Video Bitrate exceeded 150, I took care so that this might not be exceeded.

I forgot to attach the graph in my earlier post.

Have anyone got stability using Transcend Class 10 cards?

@stray ive had 1,108,000 sustained for around 30 seconds but it will fail eventually, hence the pullbacks for reliability.

The SD card test wont take place this weekend but sometime during the following week as Kingston and Delkin evaluation samples are still on the way to me.
Also I'll be doing some Elecard Video Quality Estimator comparison tests to demonstrate over a number of GOPs one to another - same scene/lighting/settings, with samples from different frame numbers (or GOP sequence numbers) to establish the pros and cons of short or long.