Official Low GOP topic, series 3

NickBen

Anybody know where the "Great Pumpkin" patch is today?

... old american humor ;-)

bkmcwd

@proaudio4
"The only question at this point is- can we see the difference between the decent exposure high detailed image between 176M GOP3 verses 176M GOP1? I'm not sure....yet. Do you have a test for this? "
I am planning to do the test of this very sake tomorrow.
Again, I love reAQuainted especially in about "noise". ;-)
I also think that your view about a noise is exact.

"these motion frames take away some of the bandwidth to reconstruct noise, even at GOP3"
I agree.
I think that that my friends care is how I-frame size affects the FEELING of resolving at the time of seeing not as a still picture but as a movie.
Although a result of a favorite problem is eventually likely to be brought, and again, it is likely to become a very interesting test also including a relation with wonderful 12GOP.
I think that 3GOP will be called "cream skimming setting" if I-frame size has much values.

Is the meaning transmitted correctly in my strange English? :-(

proaudio4

You bring up a good point bkmcwd. Although my tests were done to put a demand on the moderate to low under exposed detail, you do have a point about largest i-frames under high detail decent exposures.

The only question at this point is- can we see the difference between the decent exposure high detailed image between 176M GOP3 verses 176M GOP1? I'm not sure....yet. Do you have a test for this?

My tests are not the best for this, but they do show low exposures -2 stops to over -6 stops under exposed. "IMO" I believe what I'm seeing is, even though the 176M GOP3 has minimal motion vector frames (compared to GOP12), these motion frames take away some of the bandwidth to reconstruct noise, even at GOP3. I believe these low exposures are very important. Remember, white is set to100IRE and as we drop through the luminance to blacks, this is where a camera's signal-to-noise issue start to show. Now as these noise levels become more apparent, ideally we need the compression codec to represent this detail (noise) or we are left at the bottom levels displaying more visible blocks.

The GOP1 displays the noise more true to the raw output. With the GOP1 frame-to-frame images, I believe you can make a more accurate model of noise for your noise reduction program such as Neat Video, Pixel Farm, etc...

bkmcwd

@proaudio4
Thank you for testing! :-)
I am also going to borrow GH2 of my friend at this week, and to do a motion test on the same conditions again.
Although I love reAQuainted, there are people who care about whether less than 800k is really enough as frame size.
I think that this is a natural question.
It is because I-frame size is over 1M in some of other wonderful 12GOP.
Although it may eventually become the conclusion the "right man in the right place", I think that it becomes a very significant test.
I am looking forward to, also seeing the concrete result of your test, of course.

proaudio4

I did peform some testing on bkmcwd 176M GOP3 Q15 and driftwood 176M GOP1, but I need to evaluate all the data. I was mainly Stray's extreme dense chart shot at under exposures and a range of 160, 800, & 3200 ISO.

Both are really good. I will compile this data and post it.
I wanted to also get Strays latest 88M, but ran short due to the wife had other plans! LOL.
I will also shoot this test next week. If anything, it will let us know where the tradeoffs are bewteen higher bitrate verses moderately high.

This really was a one-for-one test.

First of all, anyone who thinks driftwood has noiser blue channel is seeing things, or something was different is the capture.
Also, if you believe this, please explain to me how the hack increased noise; afterall, noise is ACTUAL captured detail information represented with ones and zeros.....
My tests show it's actually better at producing finer noise. The finer blocks cause less smearing on any of the color channels!

drifwood's 176M GOP1 AQ3 does a beter job of low level signal representation.
My jaw hit the floor on this... You can actually see the noise is finer in my tests. Again, exact same settings and fixed lighting. reAQuainted has smaller blocks which allow the low level illumination to come through. This represent the actual fine noise closer to the raw signal. Since more light is not smeared from blocks, it seems to illuminate more in the low IRE... fucking amazing! I can cetainly see why driftwood said it sees in the dark!

I imported this into Vegas and analyzed RGB with the mid curve cranked up a bit to see even better into the noise floor.

I will post the images later.
The only issue with my test is my daylight lamps are fluorescent which causes the moving bands (shutter sync) issue. Nonetheless, they both were under the SAME conditions.

Both are VERY good, but after seeing this, it just "may" be worth the loss in drive space and use driftwoods reAQuainted.
Thank you sir! and of course Vitaliy!

bkmcwd

Although nobody may use this, I made Q14 version.
Since the maximum frame size must be made small even if it raises Q value more, this is the last. ;-)
Although not tried in detail, also in this patch, I think that you cannot use 30p(24p80%) in high ISO.

EDIT:
I stick Streamparser charts of the result of having done the quick test outdoor.
If 30p(24p80%) is also with about ISO800, it is used satisfactorily in normal situation. :-)
But, I think that 30p may be not stable.

proaudio4

3hats,
To use driftwood's latest reAQuainted patch.
Download the file here: http://personal-view.com/talks/discussion/953/gh2-patch-vault-most-popular-patches-in-one-place#Item_14
It's close to the last message.

Unzip it in the SAME directory you put ptool in. Load in original base firmware. (Vitaliy's link for instructions are below) Now, in ptool-look at the bottom,
you should see the "a" highlighted. Double-click on it to load driftwoods settings. Save it.

Vitaliy's instructions are here:
http://personal-view.com/talks/discussion/666/ptool-v3.62d-topic/p1

Always use a fully charged battery, AND MOST important, DO NOT USE a power supply!

Also please donate to Vitaliy. He has allowed the GH2 to become something really amazing.

LOL!
I see Macalincag just mentioned the samething....

Macalincag

@3hats

http://personal-view.com/talks/discussion/comment/22150#Comment_22150

http://personal-view.com/talks/discussion/666/ptool-v3.62d-topic/p1

3hats

Macalincag-Thanks
I did find this info, however, I am working on a Mac with wine bottler to view these files. For whatever reason, I can't simply load these presets into ptools to update the settings. User error for sure. So I am just inputting the data manually. It looks like there are 3 settings to deal with for 24P. Bitrate, AQ and GOP. I have set those to 176, AQ3 and GOP1. Is that right or are there other settings? Regardless, my Transcend 32GB class 10 card chokes after a few seconds. I am gathering that Transcend is not the card of choice. At 88mb the transcend is doing fine, though. Sorry for the rookie talk.
thanks

bkmcwd

@HillTop1
OK! I have understood the thing which you want to know. ;-)

As written also above, I have shot on another day.
About a time zone, maybe... possibly it was different for about 2 hours...?

And, it seems that motion blur by vibration has also appeared with the sample of GOP3ZILLA as there was a difference of ON and OFF of OIS considerably.

It is difficult to arrange conditions correctly in move shooting with handheld. :-(

Macalincag

@3hats

http://personal-view.com/talks/discussion/953/gh2-patch-vault-most-popular-patches-in-one-place#Item_14

3hats

I am new here and a little lost. I am trying to find the actual settings needed for the "ReAQuainted Patch"
thanks
Bill

gameb

@HillTop1 cheers

proaudio4

gameb...
Do not make judgement on recompressed vimeo preview video files...
Of course they look bad.

You need the original files for comparison,

HillTop1

@bkmcwd, just that I noticed a change in lighting and was wondering what was the interval from shooting with reAQuainted to yours.
@gameb, click on the "vimeo" marker on the bottom right of the video and it will take you to vimeo where you can download the original mts files.

gameb

@HillTop1 where can I see the original files?

bkmcwd

@HillTop1
I recorded several takes for each among about 20 minutes on another day, and I chose the good takes from them.
The feature of each patch has appeared?

bkmcwd

@HillTop1
Thanks! :-)

@gameb
Thanks mate!
I am using my DIY steadicam displayed in my vimeo profile thumbnail. ;-)

"some quiet strong strange artefacts around the trees" "more evident in the first than in the second film"
With the first film, OIS of the lens was set to ON with the second film in OFF.
I think that it will be the cause if there is a difference in both.
If I also play this raw data as it is with PC, I also noticed that an occasionally strange motion might be carried out.
However, since it stopped having been worrisome after editing, I thought that it was because high load is only applied to PC because the bit rate is high.
How is an actual cause?

HillTop1

@gameb, download the original mts files, not such artifacts
@bkmcwd, about how many minutes apart did you shoot with Driftwood reAQuainted and your Q15 "GOP3ZILLA"?
They both look pretty good, reAQuainted looks good in low light.

gameb

@bkmcwd, the steady-movement looks amazing, cheers mate. What steadycam do you use? But there are some quiet strong strange artefacts around the trees, more preciser on the outline of the tree-trunk, when you move around the single tree - whats that? I think this effect is more evident in the first than in the second film.

HillTop1

@bkmcwd, I like these kinds of comparisons. Looks great so far, will try it out when I get home.

bkmcwd

Although I love driftwood's 1GOP, since my friends are careful in frame size, I am making my original patch.
This compares 1GOP with 3GOP on the almost same conditions.
I think that the feature has come out well.

EDIT:
Since the name "GOP3ZILLA" was got from user, I will name to this Q15 patch.

reAQuainted

My latest 3GOP Q15 "GOP3ZILLA"

apix

Good stuff TrackZillas, As for Davinci resolve, the lite version will give you only 2 nodes of color correction on your comps.

TrackZillas

@315Films thanks for the info gonna try out the free version of davinci I think it only gives you 2 nods. MB Colorista very user friendly that's what I use now with Looks.

nomad

The 4GB limit is coming from a pretty old PC formatting scheme, called FAT32. Unfortunately it's used by most cameras until today, and there is no chance to put away with this limit, since you'd need to change the whole operating system of the camera. So, in standard operation the camera has to begin a new file at every 4 GB limit. But it needs buffering for that. This buffer's size (and speed) is the problem we are running up against.

I'm afraid we might never find high bitrate settings that always survive spanning.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters