AVCHD maximum image quality, series 2

Ralph_B

@bkmcwd
Focus isn't an issue. Look at the pictures at 200% in Photoshop and turn the AVCHD layer on and off. For Q=10, you can clearly see the noise go soft on the AVCHD (especially in the area of the orange building, center top)

driftwood

@olegkalyan A lot of 64gig cards are lousy some only working at 10-14mbs only. Jardly useful for any hibitrate settings and quite probably incompatible. Why use them without testing first?
Im still waiting for sdhc cRds to come thru for evaluation.

bkmcwd

@Ralph_B
A difference is visible also to me clearly.
If it is a result as you say supposing it is not an issue of a focus, I will also consider.
It is very interesting.
I will hear an opinion since my friend is also using my patch.
Thanks! :-)

cosimo_bullo

Our Transcend 16 GB cards have worked well with the older 132 gop 3 for the most part, but fail with 176 too. We've now moved on to Sandisc 30 MB/s and are having now problems (though we now have new problems with Henry's insane new patch).

olegkalyan

Will be looking for Sandisc cards.. thanks!

proaudio4

Ralph_B thanks again man! Another great test.

It's real close, both of them - that is to their HDMI RGB frame bottom layer.
There is a difference, and it's not just white balance. As you know of course..

I tried matching colors (which is almost impossible under these conditions where it's hard to define black, middle gray, and white)
When you zoom in and toggle layers, there's no doubt the blocks are more visible with the AVCHD. (Damn pixel peepers! ; )

Ralph, I'll need to hunt down your low light high ISO test on driftwood's 176M GOP1. I can't remember if you layered the 176M tests?
Also, did the 176M hold up better to it RGB comparison under this low light shot? I believe it did, but it's been awhile?

Everyone, remember to load into Photoshop and toggle on and off the AVCHD top layer, the backgroud layer is the HDMI RGB comparison.

You can of course zoom in for even more of a comparison. Of course, you don't need to zoom in to see the RGB frame has better shadow detail. Look at the back pine tree in front of house and tree to the left of it. The RGB does visually hold better shadow detail. Also, looking at the top front side of house you can see there is more detail in the RGB frame. Even the noise seen on the wall has more clarity. It's not a large difference, but it can be seen at normal frame size.

Having said that, this is an extreme example. We can thank Ralph for this!
The low light is an excellent test since we know detail in the shadows are the first to go. Also, remember that this is ISO6400, so it'a using bandwidth for the vector motion created by the high noise. Granted it's GOP3, it still gives up some i-frame size to share with the other two vector motion frames. Again this is due to noise is seen as motion.

That aside, I imagine under normal low ISO settings the 154M GOP3 i-frames "may" be large enough to be even closer to it's RGB frame comparision. Nice job bkmcwd!

bkmcwd, if is real close between your Q10 and Q12.
If Q12 seems better here (i'm not real 100% sure), it may be due to Q12 offering better motion rendering due to the excessive noise.

driftwood (Stray, or anyone else heavy into testing), could this explain this assuming there is a difference seen here between the Q10 and Q12 frames?

driftwood

@proaudio4 load up a copy of the demo or cheap lite version of streameye and check the MBs and QP - to see the difference.

Ralph_B

@proaudio4
I tested driftwood's 176M one page back in "Battle of the Titans".

bkmcwd

@proaudio4 or anyone else expert
Doesn't the still picture of a noise change with places to stop in except 1GOP?
Although it was my amateur's idea, when how depending on which the stopped place of a noise can be seen with the I frame or the B frame when I say from experience which has observed the noise until now had a difference, I was imagining.
@Ralph_B
How do you think?

Since frame size becomes large Q10 from Q12 after all under LOWLIGHT, I would like to use Q10.

BTW, my friend gave the name of "NINJA" to these patches from the character in which these patches "can be used anywhere and good at especially in darkness." :-)

EDIT:
@proaudio4
Thanks for your polite description!
Since my English is sometimes strange, please point out positively the place which is not known. 　;-)

Ralph_B

@bkmcwd
Yes, your English is a little difficult, but I think I understand your question. You're asking if the frame I used happened to be a B frame, then the noise would be smeared more than if it were an I frame. That's a good question, and to answer it, I went back to the original footage and grabbed three consecutive frames! Three from Q=10 and three from Q=12. Here they are in a zip file.

[url]http://www.sendspace.com/file/7qumoi[/url]

And you're right, Q=12 does indeed have some smeared frames. However, I think the smearing it a little worse in Q=10. But, it's close. You decide!

bkmcwd

@Ralph_B
Thank you for understanding my strange English. ;-)
And, thanks for showing a very clear sample.

I tested on ISO6400 over conditions, and I also felt that Q12 is clearer only very slightly than Q10.
However, since a difference was not able to view to me at the time of ISO3200 under, I think that I will use Q10.

sphipps

is it possible to use an i frame patch on 24h and a cbrandin patch on 24L as well as on 72060p?

Stray

@sphipps No as both 24H and 24L share too many parameters, including the GOP setting. You can however have completely independent 720p settings.

sphipps

@stray thank ya buddy. good to know. i need to hunt me down some 72060p settings.

Ralph_B

@sphipps
I've been working on 720 60P/50P and 1080 60i settings. I'll release them after a bit more testing.

sohus

What setting controls the i-frame size? My i-frames are now at 1,180k, I would like to reduce them to around 1,000k. How can I achieve that?

bkmcwd

@sohus
If "Frame Limit" is changed into a value smaller than 7864320?

driftwood

yes

AdamT

Bit of a newbie here and wonder if someone can explain.... I'm reading about a card speed error: is that an actual error message that the camera reports, or is it a typical sort of artifacting in playback?

I have two class 10 Transcend cards -- 32 and 16gb. With each one, I can't seem to get reliable performance above 42mb (mpgxsvcd settings). Cards too slow, or possibly just a less than stellar GH2?

tia

cosimo_bullo

I imagine it could be your cards. We've had good luck with our Transcend cards until we hit the wall with driftwood's 176 (then moved to SanDisk Extreme HD Video 16 GB SDHC Class 10 30MB/s).

bkmcwd

@proaudio4
Since the series 3 of the Low GOP thread does not stand yet, I stick here promptly.
Since former Q15 was slightly unstable, what was improved is this patch.
Please try this. :-)

Former version:
http://www.personal-view.com/talks/discussion/comment/22309#Comment_22309

proaudio4

@bkmcwd
Thank you!

I will check your new patch.

disneytoy

So, for slomo, is 720p 60 or 1080i 60 better. I realize 1080i is not 1920x1080 correct? But is their more pixels than a 1280x720 frame?

I have a project requiring 50% slow motion to regular speed and need the best option.

nomad

Well, you need to do your own tests. With the best de-interlacers the vertical resolution of 1080i will be just about as good as 720, but theoretically the horizontal resolution will be more. I doubt that you'll see a huge difference, since the whole de-interlacing process is smearing pixels together.

pingpongcomputer

I did a bunch of slow mo scenes with cbrandin's 66 settings last month, the 720 looks better and is easier to work with

edit: don't know if that holds true for bkmcwd settings, but the workflow is still easier with 720

Howdy, Stranger!

Categories

Tags in Topic

Top Posters