Getting to the bottom of the noise complaints - is it h.264 uncompression software?

No problem! Given I actually own and am in possession of an AF100, it's only in my best interest to get this right. Like most people, I want to spend my time on the creative aspects of my productions, and not spend a lot of time with technical issues (even though I am highly technical, as you can tell).

Ultimately, step-by-step will be perfect. That said, once I test this with newer clips, I'd really like to get feedback from others on the mac side. Also, I do have a snow leopard machine (mac pro) with FCP6 and CS5 on it so once I am done with the laptop tests, I will pop over there and test QT7 and QTX on there as well as the gamma tests. (Someone else will have to test FCP7 though).

Since I have a few other things I need to do first, I am going to go ahead and email clipwrap proactively because I'd rather be able to use that since I am often passing files over to another editor and it's MUCH easier to just batch convert for them. (he uses final cut)
 
I had a few minutes to wait on something so I went ahead and just shot an ISO1600 of some electronics equipment I had.

CLIPWRAP:

In PREMIERE the gamma shifting on play & scrub isn't happening on this clip.
In FCP6 all the same problems as before, no change (with just wrapper change).
Transcoding in CLIPWRAP to ProRes 422, clip acts like Log&Transfer clip inside Final cut.

PREMIERE

If you are using PREMIERE, it looks like the CLIPWRAP clips look okay now, however, since you can work with MTS directly, not sure the value. Just wanted to note the change in operation for completeness

FINAL CUT PRO 6

DO NOT USE CLIPWRAP to MOV with H.264 (wrapper change). Same problems as before (noise distortion, gamma shift problems, etc), HOWEVER, if you transcode to say ProRes422 in clipwrap, the clip works just like the log&transfer everywhere (and better in QT).

You still have the issue of the gamma going a little darker than what AVCCAM sees inside final cut, but at least if you would rather batch your ProRes conversions for FCP in clipwrap, it looks like it will work.

As a side note, ISO1600 looked better with 'editmans' settings than ISO400 on the preproduction camera in my opinion in terms of noise. Compared to the 7D, they were both just as almost usable (which is a good thing to see). As this is a 7D replacement for me, but that is neither here nor there.

A lot of people are fleshing out through the settings inside the camera, and most have come to the same setting conclusions about what looks the best. I still want to do a comparison of all the settings in the scene files, but that's a lot of work. What I would rather do is hook up a monitor to it, and adjust the settings while watching the monitor and see how they affect things with a scene with a long of range.

So Barry, I have emailed clipwrap and I'll see about getting to the bottom of the H.264 rewrap. I want this because the H.264 rewrap is nice to do a quick convert at the end of a shoot and cycle through them before the next day.

I will ponder more the Final Cut Pro issue of editing gamma, since I think it is hard to edit something darker OR lighter than it should be, even though the output file is right now if you use one of the two methods to get the content into FCP.

One thing to not is that if your Final Cut import looks wrong after export, make sure in the Preferences under the editing tab that Importing RGB video gamma is set to source (the default) which is why the export looked right. (although why this didn't affect the H.264 import from clipwrap I have no clue)

In my searching for more information on Final Cut Pro, and the gamma issue, it seems that Final Cut tries to account for the gamma difference between video displays and playback mechanisms. Given that normally OSX has a higher gamma (which I found a reason stated by Apple is to increase detail in the whites at the expense of blacks for print work) it looks like Final Cut is trying to compensate. And, I don't see any way to adjust it.

What I haven't done is taken something in Final Cut and then play it on a variety of devices to see how it looks OUTSIDE the computer screen. It might match more closely, but wow, is that frustrating.

So let me repeat DO NOT USE CLIPWRAP REWRAPPED H.264 IN FINAL CUT OR VIEW IN QT at this time. Both Final Cut and QT are showing something wrong, and in Final cut, it IS actually adding crazy noise and glitching in the clips. Use CLIPWRAP with PRORES transcode.
 
Remember these tests were on Leopard (10.5), so I have not tried anything with Snow Leopard yet (although the only thing I feel to test on there would be QTX which I don't like anyways) :)
 
Fotojorno said:
Dino,
Try your exact same settings, except change the Matrix to "Fluo" and see if that helps. I was using very similar settings to yours and saw a good deal of noise in the shadowy gray area's. I took Editman's suggestion of using the "Fluo" Matrix and I haven't seen any noise since. Not on my MBP and definitely not on my TV. You can check out the test I posted on my Vimeo page.
Click to expand...

Wow, i never actually thought the color matrix would have an affect on the lower exposure end noise. I put my marshall monitor on 1 to 1 pixel zoom, pointed it at my dark blue sofa towards the corners, and between the color matricies i noticed that they were in this order, from noisiest to cleanest:

1) Cine-like
2) Norm 1
3) Fluo
4) Norm 2

A lot of the mojo is from Cine-like because of the punchier colors. I guess it's a tradeoff? How does this affect footage with H.264 compared to the other users?
 
Papa
i'm super curious.
would you upload a screen grab from Norm 2?

would you say that "Cine-like" is HVX-type-of-noise?
 
I never used an HVX in my life so i can't really comment, but i can tel you im getting very nice and clean images now with:

Detail: -6
VDetail: -7
Coring: +2 (still trying to figure out how this really works, just used it cause that seemed like the popular choice)
Chroma: +2
Gamma: CineD
Matrix: Norm2

Very happy :)

I don't have access to my editor right now because the project has currently open is a feature film that im working on tomorrow so i don't want anything to go wrong if i try reopening it, but you can take a look at the compressed video file i posted in another thread to compare noise ( to a certain degree )

I think it will be quite apparent, even with compression, that NORM 2 has the nicest setings.
 
PaPa said:
I never used an HVX in my life so i can't really comment, but i can tel you im getting very nice and clean images now with:

Detail: -6
VDetail: -7
Coring: +2 (still trying to figure out how this really works, just used it cause that seemed like the popular choice)
Chroma: +2
Gamma: CineD
Matrix: Norm2

Very happy :)

I don't have access to my editor right now because the project has currently open is a feature film that im working on tomorrow so i don't want anything to go wrong if i try reopening it, but you can take a look at the compressed video file i posted in another thread to compare noise ( to a certain degree )

I think it will be quite apparent, even with compression, that NORM 2 has the nicest setings.
Click to expand...

When detail is on the + side the coring kicks in !
 
Just not sure what exactly coring is lol.I know it has to do with reducing the apparent sharpness of noise? I think??
 
Update: I have sent clips to the clipwrap developer and they are looking at the issues I was seeing with the MOV files with QT and FinalCutPro.
 
I have been perceiving massive noise in most clips from the AF100 available on-line. There was a clip just posted to Vimeo that seems to tell me something. Go to: http://www.vimeo.com/18732208 to download "Af100 Dynamic Range Stretch". The original download is 443 Megs. Now watch the first part of the clip. It appears almost noise free on my system (iMac i7 calibrated using apples built in calibration but set to gamma of 1.8). Now, after Burmoco does a DYNAMIC RANGE STRETCH OF +2, 38 seconds in, the blacks on my monitor show extreme noise. I was interested to try out Barry's thoughts on the decoding of the codec being the possible culprit so I transcoded the downloaded clip to ProRes422HQ using MPEG Streamclip (as I do all my Canon 1D Mark IV footage for editing) and I perceive less noise. The image is also a little darker so I'm wondering if there was a gamma shift. I changed the gamma on my monitor to 2.2 but it did not help the noise on the original clip. I find it interesting that the clip from Vimeo looked so clean before the Dynamic range stretch of +2. Also, looking at my download of "A Verse Before Dying", freezing on some interiors with lots of shadows, the blacks look solid...no gradation or noise whatsoever. I'm wondering if the blacks were crushed or if a de-noiser was used on this footage? I would really like to get to the bottom of this before buying this camera.
 
Dynamic Range Stretch over 2 inherently has noise much higher than DRS:1. This not taking into account the actual scene being filmed. The DRS is intelligent so to speak, so, having dynamic range stretch when it isn't necessary causes noise.

Barry has stated that Dynamic Range Stretch with a setting of two or more should be only in "emergencies."
With that said, even Dynamic Range Stretch at 1 should only be used if needed.
 
Last edited:
What did Barry say they used for dynamic range stretch on the exterior shots of a verse before dying again?
 
CineMan1 said:
I have been perceiving massive noise in most clips from the AF100 available on-line. There was a clip just posted to Vimeo that seems to tell me something. Go to: http://www.vimeo.com/18732208 to download "Af100 Dynamic Range Stretch". The original download is 443 Megs. Now watch the first part of the clip. It appears almost noise free on my system (iMac i7 calibrated using apples built in calibration but set to gamma of 1.8). Now, after Burmoco does a DYNAMIC RANGE STRETCH OF +2, 38 seconds in, the blacks on my monitor show extreme noise. I was interested to try out Barry's thoughts on the decoding of the codec being the possible culprit so I transcoded the downloaded clip to ProRes422HQ using MPEG Streamclip (as I do all my Canon 1D Mark IV footage for editing) and I perceive less noise. The image is also a little darker so I'm wondering if there was a gamma shift. I changed the gamma on my monitor to 2.2 but it did not help the noise on the original clip. I find it interesting that the clip from Vimeo looked so clean before the Dynamic range stretch of +2. Also, looking at my download of "A Verse Before Dying", freezing on some interiors with lots of shadows, the blacks look solid...no gradation or noise whatsoever. I'm wondering if the blacks were crushed or if a de-noiser was used on this footage? I would really like to get to the bottom of this before buying this camera.
Click to expand...

All I can tell you is that I did pixel to pixel verification of the MTS files using AVCCAM, ProTools and Premiere Pro, using Photoshop to do subtractions between the footage. The H.264 decoders used in Final Cut, Premiere, QT and AVCCAM produced the same exact noise with different gammas. How did you use MpegStreamClip with an MTS file? Using clipwrap to change the wrapper to MOV introduces an issue. I am working with the clipwrap people to determine why. It has something to with colorspace they have told me.

Other than that, I cannot find a difference in the decompression, and I have spent A LOT of time testing things.
 
grimepoch - have you tried other methods to re-wrap (e.g. ffmpeg)? is it clipwrap only that introduces the issue ?
 
The version of ffmpeg on my machine won't read the MTS file. I thought that it should, but I checked all versions of everything I have installed and yet it still will not allow me to read the MTS.

ClipWrap is on the case, and they have sent me some stuff to try out to see if it helps. Will keep everyone posted.
 
You must log in or register to reply here.
Back
Top