Official Low GOP topic, series 5

bkmcwd

Of course, you can combine the settings of MJPEG. :-)

leejb4

@bkmcwd So would you says users of the last GOP3ZILLA would find benefit to moving forward to the Natural setting? I know they are different and preference does mater, but in general terms, would you say natural has more "umph" towards quality and performance?

bkmcwd

@leejb4

The difference between natural and GOP3ZILLA will be a difference of Q18 and Q15, if I say simply. The merits using natural are that file size becomes somewhat small and that the scene which is highly minute and has a motion can also maintain high definition. The merit which chooses GOP3ZILLA is realizing high image quality also under lowlight also at a bright place.

On a scene which is shot while moving in green forest, natural is suitable. If it is a night view of a town, or indoor, the shot with GOP3ZILLA becomes the highest image quality.

driftwood

@marie Because there are a few variations of sizes for anamorphic lenses its best to size up yourself and input your own settings. This is why the majority of settings are now at base size levels to begin with.

leejb4

@bkmcwd Awesome! Well, I may be going to a pretty nice shooting area with a lot of trees and things so maybe I can get some test footage. Are your test from a few months back of Natural that you filmed about the same as if you would have filmed them today on the settings you released?

towi

@bkmcwd

"Silent users of my settings :-)"

From now on I will be a "loud" user of your settings - YOU ROCK :-)

Apart from Ralph_B's Sanity and my 24p25p-NINE your bkmcwd natural v1.11 patch is the only one that produces high IQ for PAL users at 50i (FSH) and 25p (HBR) even when the codec is stressed much.

The issue for us PAL users is the GOP settings for 24p and 50i are coupled. This is why many patches don't work as supposed to in PAL mode at HBR/FSH. Your settings do "work". However on my SanDisk 30MB/s cards the camera crashes on highly detailed scenes with motion (e.g. with slow pans on a "bush of death" or so...). So I've modified your settings for my SanDisk 30MB/s cards. I've lowered the bitrate for FSH from 80M to 72M and I've added the matrix of my 24p25p-NINE patch. By now these modified settings survive various codec torture tests. The footage shot in HBR mode is literally indescernible from shots at 24H. Though the bitrate was lowered a bit from 80M to 72M I-frames in HBR mode almost match the I-frame size of your undmodified patch (930K vs. 940K for the same scene). Streams analysed in Streamparser look healthy (ratio of I-frames and B-frames looks as supposed to with motion).

Have to do some more testing but things really look great with your settings. Many thanks for the great patch!

Attatched Streamparser grabs of a death chart shot in PAL mode at 24H, HBR (25p) and SH (50p) with my modified patch. The modified patch is also attached.

edit:

here's a 4'' comparision of 24H and HBR/25p of my modified patch.

ISO400, Smooth -2|-2|-1|-2, Zeiss 2.0/50, ETC mode.

(Foto JPEG QT movie:) http://www.file-upload.net/download-4177568/natural_modified_FOTOJPEG.zip.html

Stray

Argh why you all got to change bkmcwds lovely predictable high IQ settings with your matrices ;). Seriously though, as I always turn to this patch (GOP3ZILLA) for 'thee balance' of beautiful motion with ridiculous detail, I am looking forward to trying it with the new 'sedna' matrix. Those screenshots driftwood posted from it look very nicely sharp. Mind you @Towi, if thats going to create problems it'll be HBR where they'll be. Can you post us some screenshots for the HBR you've built on 'Natural' ? Edit : Sorry, just saw your download link.

bkmcwd

@leejb4

Thanks! :-)

I welcome your test footage. Since the weather is bad in my town every day, I cannot go to usual forest to test.

@towi

Thanks mate! :-)

I regret that I cannot talk with you loudly since I am poor at English.

I did not know the PAL user's situation well. I will be glad if helpful.

@driftwood @Stray

Thanks always! :-)

towi

@Stray

"why you all got to change bkmcwds lovely predictable high IQ settings with your matrices"

True ... just compared "bkmcwd natural v1.11" with and without my Matrix (and some other Matrixes from Driftwood). In my "24p25-NINE" patch I prefer my HQ matrix as the transition in the blacks is somewhat smoother. But in "bkmcwd natural v1.11" Panasonic's stock matrix does a better job.

Attached an updated version of my modified "natural" version for PAL users with SanDisk 30MB/s cards.

Kihlian

@bkmcwd At silent users.. I´m trying your settings too, wonderful!! I think I keep thys settings. I´m using the modified version from towi because I´m a 25p user too and I have the same SD card. The only think I find the reds a little bit pushed up, maybe I combine it with sedna looks great in the video.

bkmcwd

@towi Thanks for your modified version. :-)

I think that everyone should just combine a favorite matrix.

@Kihlian Thanks for using my setting. :-)

Please give me feedback of results.

bkmcwd

@ Silent users of my "settings" :-)

I release the newest "GOLGOP3-13_v4" setting. This is a TEST setting for using low qp with 3GOP.

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.(Maybe)

*Spanning in 24H, 24L, maybe in FSH, HBR, and SH with SanDisk 95MB/s 64GB SDXC Card.(In my death chart test,24p,25p,30p,50i,50p are spanning in ISO12800 repeatedly. In ISO3200, 60i and 60p are also spanning.)

Kihlian

Here a screenshot made with GOP3ZILLA featuring towi ;-) 25p HBR. The camera is moving around 33500 Kbt/s.

bkmcwd

@Kihlian

Looks good. Thanks! :-)

towi

"In my "24p25-NINE" patch I prefer my HQ matrix as the transition in the blacks is somewhat smoother. But in "bkmcwd natural v1.11" Panasonic's stock matrix does a better job."

Testing goes on...

While the transitions in the deepest blacks are a bit smoother the stock matrix also shows a blue tint in the blacks. With my HQ matrix the blacks are consistent with the white balance of the entire image. Too, overall my HQ matrix will produce a slightly sharper image than the stock matrix… but without beeing harsh. I've also noticed my HQ-Matrix will produce larger B-frames (both with static scenes and motion) whereas I-frames are a little bit smaller (minor difference, though)… hard to compare but that is what I have seen in my tests so far. However, I am talking about fine nuances between the stock matrix and my HQ matrix… not about earth-shattering differences. It's up to ones personal taste...

Stray

@towi Are these the settings you posted above or are you still on the tweak ?

No more blue tint in the blacks has sold it for me. Not having a blue suppression step to do would be very nice, my processing scripts are getting a tad cumbersome. I guess its going to be impossible to say if the larger B Frames are an unnecessary inflation thats occuring. I have some shooting for a personal project tomorrow so I'll give it a whirl if the settings are the ones above. Are you testing 24p too or just focussing on HBR ? oh, and have you tried your matrix on GOP3ZILLA or just natural.

towi

@Stray

"Are these the settings you posted above or are you still on the tweak"

I was talking about the modified "natural v1.11" patch (for PAL & SanDisk 30MB/s card users) with and without my HQ matrix ... no further adjustments.

bkmcwd natural v1.11 + HQ matrix posted above in Comment_47102: http://www.personal-view.com/talks/uploads/FileUpload/8f/a951a7feb67b226cfe30b0afd0f5ae.zip

bkmcwd natural v1.11 + stock matrix in Comment_47173: http://www.personal-view.com/talks/uploads/FileUpload/23/b70b1d40ef33a44998e52bfc7646c7.zip

"Are you testing 24p too or just focussing on HBR"

My findings posted above are based on tests in HBR/25p mode shot under lowlight tungsten. Most likely color balance related issues apply to all modes... but I can't confirm for sure right now.

"oh, and have you tried your matrix on GOP3ZILLA or just natural"

only on "natural"

You can easily test it on GOP3ZILLA ... just insert the values directly in the ini file (with a text editor / notepad) after "Quantizer" and before "GOP tables" (see attachment). I've attached the matrix (with and without fallback mode ... I use the one without fallback mode).

Kihlian

@towi the sharpness in your matrix is incredible! look at the basket on my screen shot above. I have a beginner question, will not a Intra setting solve all issues of b and p frames?

Stray

@towi cheers, I'll give the HQ matrix version a good test tomorrow in natural light.

towi

@Kihlian

"will not a Intra setting solve all issues of b and p frames?"

Maybe... but unless someone makes an Intra patch that produces I frames at around 800K this is theory. Driftwood's patches are really great - also in HBR-mode. But in HBR mode only whith scenes that do not stress the codec much. As soon as you shoot highly detailed scenes footage will look heavily compressed and will show mosquito noise. (At least this goes for PAL mode). I am confident Driftwood will make a great HBR patch sooner or later. But he's focussing on 24p and he is an NTSC user... so it's sure very well understandable that HBR in PAL mode is not on top of his priority list ...

@Stray

"I'll give the HQ matrix version a good test tomorrow in natural light"

Cool - please let us know your findings.

Kihlian

@towi Thanks for the answer, 'll always stay to give a look to your settings. The last Gop3zilla with your matrix have a sharpness that takes my breath away!

towi

@bkmcwd

may I ask why you've set the bitrate for 24H that high? I'd like to lower the bitrate to somewhere around 100M for 24H. Any advise on a good value to start with?

many thanks!

Stray

@towi I believe the bitrate was set at that level specifically to get as close to the possible maximum I-Frame size (there is an in camera upper limit on I-Frame size IIRC, but I can't remember it) for a GOP of 3 (the longer the gop, the lower your bitrate needs to be to maintain equivalent I-Frame size). Mind you all these vague memories of mine relate to earlier versions of Ptool.

towi

@Stray

"the longer the gop, the lower your bitrate needs to be to maintain equivalent I-Frame size"

yes, I am aware of that. However, the bitrate is almost as high as in Driftwood's Orion, which is I-frame only, so GOP=1. I wonder if 110M, 99M or even 88M won't be enough. Of course I could just try... but bkmcwd as the creator of this patch maybe could give a well founded starting point.

Stray

Indeed, but the I-Frames in the Driftwoods GOP 1s, while consistent, cannot reach near the I-Frame sizes of this GOP 3 patch if you're shooting a well light/well detailed scene. 1,180,000 I think was the I-Frame max. But yeah, its a question of if taking a drop in I-Frame size (as you lower its bitrate) is perceptible.

In post, to me anyroad, it is definitely noticeable. Just push your levels really high and check the sizes of the visible blocking, particularly in the chroma. I did a lot of tests related to this with varying GOP lengths, and I always tested a B-Frame (or P-Frame) from the middle of the long GOP against a GOP 1 I-Frame, to see if it was maintaining the quality. And it did, quite impressively, I was expecting it not too (even with a GOP 12 of around 66M). Now the GOP1 settings have come on a long way I know, but I don't think this would have changed, larger I-Frames is still going to mean better overall detail. I guess its just one of the things that makes me love this camera, the ridiculous, and probably completely unnecessary detail it can achieve and maintain at a longer GOP than 1. The bkmcwd GOP3ZILLA is as far as I'm concerned the perfect balance between detail and motion, but true enough dropping the bitrate may not be visibly noticeable. It would take a while to get it properly stable at a lower rate I think (it used to be 10M higher than it is now too IIRC).

edit: Oh and on the whole from recent shooting, the GOP1 CBR styles are the better performer in very lowlight, which is to be expected. But I'm expecting that your matrix may close that gap.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters