Official Low GOP topic, series 5

bkmcwd

Thanks! looks nice. :-)

It turns out that a cadence issue sometimes occurs in 60i and HBR of my settings. (I am not sure in PAL.) I am improving now.

Stray

@bkmcwd do you mean that there is a cadence issue with 60i just with the new GOLGOP3 settings youve posted ?

bkmcwd

@Stray "GOP3ZIILA" "natural" "GOLGOP3-13" all my settings. However, I have already improved GOP3ZILLA at first. I am just going to make preparations of the release now. :-)

bkmcwd

@ALL :-)

I release the newest "GOP3ZILLA_V2.0" setting. This is also a TEST setting for using low qp with 3GOP as same as GOLGOP3-13_V4.1. I think that you understand if you look at Streamparser what kind of character it is. When the result of the Stray's test chart was looked at by StreamEye, it turned out that the codec is not wavering. The codec was wavering in the previous version.

And the stability of FSH and HBR is raised a little.

Of course, this is also the setting focused on image quality with LARGE I-FRAME SIZE except 60p and 50p.

Please try it.

NOTE:

*24H mode is high quality 3GOP setting with mega size I-frame.

*24L mode is for 24p80%(30p).

*In camera playback is available.

*EX Tele works.(Not tested yet, but probably)

*Spanning in 24H, 24L, maybe in FSH, HBR, and SH with SanDisk 95MB/s 64GB SDXC Card.(Not tested yet)

*New natural and GOLGOP3-13 are coming soon.

duartix

@bkmcwd : I appreciate your work and please don't take it as a critic, I know this is a test chart and all, but the QP fluctuation on I-Frames seems just too much.

There are some settings developer monsters here that don't care about it as long as IQ is delivered, and that's just what I'm trying to investigate. I've seen some of those mega fluctuations and when I checked video for IQ everything still looked fine, so I can be chasing ghosts but can you please post a PNG screenshot of an I-Frame from 00000.MTS?

Thanks.

Stray

@duartix don't you mean 00001.mts ? 00000 is from the previous settings for GOP3ZILLA, which this file replaces.

duartix

Thanks @Stray ! I didn't notice 00000 was from the previous version. @bkmcwd : about the previous post, the new one looks a lot healthier on death chart (00001.mts) and about perfect on regular footage (00002.mts).

balazer

bkmcwd, I find that Gop3zilla 2.0 improves the stability of HBR mode somewhat compared to your previous version. What did you find the problem and the solution to be?

bkmcwd

@duartix I am sorry to have displayed the unclear result. ;-) Yes, second is latest version. In this result, the codec is not wavering.

@Stray Thanks mate! :-)

I am thankful to you who always supplement with my shortage of explanation.

@balazer Thanks for checking my settings, mate. :-)

A cadence issue arises, when the bit rate which a frame requires is insufficient. This was not able to be solved if FL and FB were not lowered until now. I do not know whether it is the method that this is right. However, I maintain balance by operating GOP_related and ES. By this operation, the bit rate which a frame requires can be lowered relatively. I think that it is stabilized as a result.

Was it transmitted with my strange English?

bkmcwd

@all

I release the newest "natural_V2.0" setting. This setting is also made as the same character as my latest settings. When the result of the Stray's test chart was looked at by StreamEye, it turned out that the codec is not wavering. The codec was wavering in the previous version sometimes.

And the stability of FSH and HBR is raised a little.

Of course, this is also the setting focused on image quality with LARGE I-FRAME SIZE except 60p and 50p.

Please try it.

NOTE:

*24H mode is high quality 3GOP setting with mega size I-frame.

*24L mode is for 24p80%(30p).

*In camera playback is available.

*EX Tele works.(Not tested yet, but probably)

*Spanning in 24H, 24L, maybe in FSH, HBR, and SH with SanDisk 95MB/s 64GB SDXC Card.(Not tested yet)

guzak

so which one is better? I mean, what is between

bkmcwd natural_v2.0

and

bkmcwd GOP3ZILLA_V2.0

?????

I am Europiam user, so my interest is HBR 25p

Thanks a lot for your great job anyway :-)

bkmcwd

@guzak

I do not understand the "better" meaning you say. And honestly, since I have not done strict comparison in 25p, I do not understand. Please let me know the compared result. :-)

guzak

I mean, which of them has less grain/noise in black areas for example when shooting in dark and higher ISO.... By question ,,what is different in between them,, I mean, there must be some difference, if you dvelop two versions :-)

bkmcwd

@guzak

OK, I got it.

"what is different in between them"

If I say simply, they will be the difference in image quality, and the difference in data size. The difference among both is a difference in qp. However, I have not checked yet how a difference appears in 25p. Although low qp realizes high image quality, as you know, it does not necessarily bring such a result depending on the object to shoot. Please try them. :-)

guzak

OK, I am going to make some night shooting with your new GOP3ZILLA_2.0 and Canon FD 50mm F1.8 ... I wish I would have two GH2s with two identical lenses side by side so I would not need to rewritw NATURAL_v2 and back again.... :-)

flaschus · vlcsnap-2012-03-14-11h39m14s69.png

@bkmcwd... can u point me somewhere to learn more about QP and Quantization and IQ?

gop3zilla 2.0 snapshot 20mm

duartix

@flaschus : http://www.slideshare.net/vcodex/introduction-to-h264-advanced-video-compression Slide #20.

Basically, lower QP = higher IQ = bigger size. QP = quantization parameter, think of it as a common divisor for every value of a block when encoding and a common multiplier when decoding.

flaschus

@duartix :

thankyou!..

any chance u have more information (links) on the topic?

duartix

@flaschus : Google is your friend. Search for H.264.

This one looks good: http://www.vcodex.com/h264.html

proaudio4

@flaschus Nice frame grab. Do you remember what ISO and f-stop? Wide open I imagine.

flaschus

umm, 160 , 5.6

bkmcwd

@flaschus Thanks, looks nice! :-)

@duartix Thank you for explaining instead of me. ;-)

flaschus

@ bkmcwd//@duartix.. so the lower the qp the more accurate color space we are seeing, per the rounded colour value?

like 134 can get coded to a qp 40 and come out encoded 120

but 134 into qp 10 gets decoded 130...

ok, so that makes sense, BUT how does attaching different matrixes affect this?

towi

@bkmcwd

"I release the newest "natural_V2.0" setting. This setting is also made as the same character as my latest settings. When the result of the Stray's test chart was looked at by StreamEye, it turned out that the codec is not wavering. The codec was wavering in the previous version sometimes."

Unfortunately with V2.0 the stream shows an uneven cadence in PAL mode (HBR/25p) on a static death chart. With V1.11 you'll notice it too, but much less.

PAL, HBR 25p:

Bonzai

GH2 Hack GOLGOP3-13_V4.1 Vimeo test video. Shot in an old unused stairwell of the 115 year old mill building where I live. Lotsa shadows and textures, difficult AF conditions and horrible lighting (the only CB setting which comes remotely close is manual 2.5K). However the conditions are always exactly the same - it's like nothing ever changes which is great from a video test perspective.

GOLGOP3-13_V4.1 32kISO Cine WB-2.5k -2 -2 +1 -2 SS-30 HBR.MTS Pany 14-140mm ISO3200 High Bit Rate F4.0 SS1/30 AF OIS:ON CINEMA [-2.-2.+1.-2] DL original 1080P/30Hz MTS file for analysis.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters