Official Low GOP topic, series 3

driftwood

@bkmcwd Yes, the mere fact its INTRA and every frame is sustaining hi quality, large picture size, quality mbs and Q and 'binning' less I feel its better, and looks better to me. The longer GOP patches can also look good and require less space but are cosmetically harder to edit and prone to more problems if the GOP gets corrupted. They actually take longer to decode and manipulate in the NLE. You can't 'cut' on a predictive frame, unlike INTRA.
Your own GOP3 AQ findings are interesting. But a different range of matrices will come into effect between the AQ settings and you need to analyse both macroblock sizes and Quantisation. Your perfect AQ quest depends on the detail/scenery/hi&low frequencies/movement as to what you prefer works for the subject matter.
For 720p I prefer AQ1 with a good bitrate/buffer ratio - a good balance for which is, after all, about slow motion quality.

bkmcwd

@driftwood

Thanks mate! :-)
And your new patch is amazing!
I also tried immediately, without any problems could span at 24H with SanDisk 95MB/s 64GB.
You are the man!

BTW, I have some questions.
As you said, a few weeks, with the streameye, I have observed macroblocks and qp.
"Setting the frame size is about 800k and larger max mb size" and, "setting the frame size is about 1M max mb size slight smaller" I've compared.
(Here, ignoring the difference between the GOP, just think about the I-frame.)
Is the idea "other mbs will be small size relatively if max mb size is large in the limited frame size" right?
After that, That is, since the one where frame size is larger becomes average as a whole and mb size becomes large, image quality is good?

Attached streameye is the result of my shooting the scene in the same room using your Quantum 6 Beta, and AQ4 and AQ2 of me, respectively.
What should just look at this result how, or I do not understand.
Do I hear that after all, AQ2 with the largest frame size is the best as for image quality?

*These are all shot panning in the same room.
Quantum 6 Beta
http://www.raitank.jp/wp-content/uploads/2012/01/juk38-QT6_room_800_sp.JPG
AQ4
http://www.raitank.jp/wp-content/uploads/2012/01/sm689-AQ4_room_800_sp.JPG
AQ2
http://www.raitank.jp/wp-content/uploads/2012/01/91887-AQ2_room_800_sp.JPG

proaudio4

@driftwood

Quantum v6 BETA !

Thank you!

proaudio4

@csync
Excellent work here!
Also, I was wondered the same thing regarding white balancing.

I've noticed while using the "daylight" WB setting the RED channel is noiser than the BLUE. Also, the RED can clip early and makes REDs look like shit and highlights look fucking nasty.

Zaven13

@B3Guy
Since you are such a nice guy calling people imbeciles, here is how you can download videos from YouTube. By the way, I actually put the link to the video in my post here but somehow it got replaced with the video thumbnail. Regardless, if you go to YouTube.come and search for example on video I posted (GH2 test foggy day), it is the first search result you see. Click on the video and the link shows up in the URL area. Right click and copy the URL. There is a free program called YouTube Downloader. Download and install that program. After you open the program, copy the URL into the program URL area and click download button. You get an MP4 copy of the video on your hard drive.
Make a new year's resolution to be more respectful in the future.

driftwood

@strangeways shoud be fine for all

Strangways

I've been switching between Aquamotion v2 and SeAQuake, using Sandisk 35MB/s cards. Will the 24H and 24L settings of Quantum v6 replace the need to switch between them for me? I know 24H won't span on my cards, but will it work for 4 or so minutes like SeAQuake?

driftwood

Its Quantum v2 with higher frame limit, tuned bitrate ;-)

Quantum v6 1080p24H is bitrate=147M (7M less than Q2 but you won't notice)
However, buffer=148846 (ie much higher buffer)!

The L setting is just about 100M and primed - I want to see how this gets on with ISO tests in particularly.
The 1080i and 720p 60 settings are 100M :-)

I have been analysing AVC INTRA 100. You may have guessed.

driftwood

@paglez Thanks for the intial feedback.

paglez

B3Guy
In Spanish your comment about camera test by imbeciles, could consider thay you are un gran gilipollas o "tontolculo".

danyyyel

Swiss_Boy very nice.

B3Guy

@csync Try Vimeo if you want to actually share the .mts (it will upload to their site and you can allow viewers to download the original file. It may not actually play back in Vimeo-s web player, though.) Vimeo+ is not horribly expensive, and IMO is worth it for the quality improvement over YouTube (as well as the absence of links to "camera tests" by imbeciles).

*edit* alternatively, dropbox.

paglez

@Driftwood
Thank you very much once again for your great dedication and hard work. This is the results of my first test about your Quantum v6 Beta:
1080p24 H Mode: Spanning correctly (Sandisk 95 MB/s), similar quality than previous versions.
24 L Mode: Spanning (not checked yet) similar quality than previous versions
1080i50: Similar quality than previous releases
720p50: "" "" ""
108024p (variable 80%): Works fine in L Mode (checked about 3 minutes)
108024p (variable 80%). Break in H mode (at 8/9 secs)

This results are subjectives, due are based in night shots (in very low light conditions)

danyyyel

@csync you can perhaps share some uncompressed screen grabs in png for a start.

driftwood

*** NEW *** Quantum v6 BETA
New changes new theory. Tuned buffers now set by 'Bottom' Bitrate settings.
1080p24H Quantum v2 style with slight bitrate/ buffer tuned.
1080p24L Great allrounder and amazing QP. AQuamtion v2 tuned. Also 80% mode and ETC mode should work, plus spanning dare I say?
1080i60AQ4/720p60AQ1 for movement, GOP3 and GOP 15 on a new bitrate/buffer ratio.
1080i50/720p50 as before apart from AQ1 for movement.

Try before you cry.

driftwood

Excellent report @csync.

csync

I redid my tests with Quantum 2, 5, Aquamotion, and the Panasonic 1.1 firmware, and it does look like the noise is there throughout all of the firmwares loaded but is slightly more noticeable in Quantum 5 when compared to Quantum 2. Quantum 2/5 both have finer grain and more detail in the shadows than Aquamotion, and the Panasonic firmware seems to have the least grain, but also the least detail in the shadows.

I have the raw MTS files for all four firmwares at:

ISO
160
320
400
640
800
2000

if anyone would like them.

I also tried different white balance profiles, to see if that changed the noise significantly, and white balancing to a card proved to have the least amount of noise.

The clip on Youtube was brought into FCPX using 422HQ from the MTS files, and then compressed to H.264 at 27Mbps and uploaded (484MB). The compression noise from this as well as Youtube's own compression knocks out a lot of the grain, and I'm not sure how to better share these MTS files for comparison. If anyone has better testing protocols or advice, I'm eager for it! I set the ISOs always from the ISO directly below it (on the grid, so, from a higher ISO), as well, so they should all be safe.

driftwood

Hi mate, @bkmcwd theyre all part of Quantum testing, Total safety on the limits for GOP1 is reallly around the 6008080 mark - see other Driftwood patches. A lot of my patches have 6108080 (or thereabouts in), the 6208888 version was to see how it got on with other people's high end sd cards.

bkmcwd

@driftwood

I made two kinds of AQ4 patches which can span if SanDisk 98MB/s 64GB card is used.
Since my work is sloppy unlike you, in a frame limit, these seem to be limit for me. :-)

BTW, when a frame limit is set to 6208888 and I try on your Quantum-Beta5 of you, recording has stopped in 36 seconds. :-(
Is there any mistake in me in anything?
Of course, I used SanDisk 98MB/s 64GB...