Bought the Canon 1DC

[FilmMan]

Here's  a few seconds from one shot from today (it is one of the less gross shots).   50 mm lens.  Neutral style.  Played with the colors then uploaded.

http://www.youtube.com/watch?v=73zdOAM-n7I&feature=youtu.be

[Sean Cunningham]

Hi richg101,

 

I use 32 bit workflow so I was surprised with the 1DC's footage not holding better. 

 

Without applying some kind of chroma filtering (some codecs do this, or have options to) you will retain banding in a 32bit workflow if it exists in the original footage.  32bit isn't going to create transitional data where there was none (not without intentionally applying another process expressly to do so, or by some byproduct of that process).  If the 8bit source bands a first step transcode to 32bit will band unless it's filtered or dithered.  Make sure you're also not introducing it with some discrepancy in your color management, just as a "crazy check".

 

8bit 4:4:4 can band, and does band, given the right material.  That's why all the better 601 gear quickly had to step it up to 10bits by at least the early 1990s at the latest.  Even with only 720x480 pixels you couldn't rely on 8bit uncompressed to not band, especially when transcoding.  

 

The fellows at Amblin found this out the hard way when they were doing the FX for the pilot episode of SeaQuest DSV and had to put in an 11th Hour call to the image format experts at ASDG who wrote them special drivers for their digital disc recorders to filter the incoming chroma, smoothing out the big, nasty bands that all of a sudden showed up in all of their foggy, blue underwater shots.

[FilmMan]

Hi BurnetRohoades,

 

There was no banding in the original footage.  I think my workflow is okay - although I'm prone to errors.   The footage banded in a very small area when I did a quick grade - on a very small portion of skin banding was apparent.    I tried the footage in another program and it doesn't band as bad when applying the same correction.  I'm trying to learn the weaknesses of the 1DC so I know what to look out for.  This camera is very sensitive to light.  I wish Canon would have elected upon 10 bit instead of 8 bit.  Canon may get lots of banding footage posted in the near future if this problem persists.  To be fair, I can't really make an accurate assessment yet, till I experiment more.  Cheers.

[hmcindie]

 

The fellows at Amblin found this out the hard way when they were doing the FX for the pilot episode of SeaQuest DSV and had to put in an 11th Hour call to the image format experts at ASDG who wrote them special drivers for their digital disc recorders to filter the incoming chroma, smoothing out the big, nasty bands that all of a sudden showed up in all of their foggy, blue underwater shots.

 

No, that was actually because of 4:2:0. That's why dvd's where there is a predominant color (blue/red), the chroma looks horrible. Has nothing to do with 8 bit.

 

One thing that takes away most of the banding in 8bit material nowadays is Neat Video. Great noise reduction but when using it in a 32-bit workflow, it really does miracles. You can just use it with almost no noise reduction at all and it will still smooth out the material into 32-bits really, really well. Surprising how well it works really. I've had shots where regular 32-bit grading showed a bit of banding that disappeared completely after a very slight adding of Neat Video.

[Sean Cunningham]

No, that was actually because of 4:2:0. That's why dvd's where there is a predominant color (blue/red), the chroma looks horrible. Has nothing to do with 8 bit.

 

Uh, no, this was being shot to a $60,000-80,000 Accom or Abekas digital disk recorder before being put to uncompressed digital tape for network broadcast.  DVDs were almost ten years away from existing at the consumer level.  It has everything to do with transcoding between YUV and RGB at 8bits.  You don't know what you're talking about here.

[richg101]

Without applying some kind of chroma filtering (some codecs do this, or have options to) you will retain banding in a 32bit workflow if it exists in the original footage.  32bit isn't going to create transitional data where there was none (not without intentionally applying another process expressly to do so, or by some byproduct of that process).  If the 8bit source bands a first step transcode to 32bit will band unless it's filtered or dithered.  Make sure you're also not introducing it with some discrepancy in your color management, just as a "crazy check".

Indeed.  I meant that since changing to using purely 32bit colour correction effects in premiere I have not been introducing any hugely noticeable banding when applying quite prominent grading.  I can certainly obtain cleaner gradients and smoother skintones/skys when I use just 32bit effects.  Re. capturing in 8bit colour, I am quite impressed with Sunset profile on my nex5n.  I hadn't used it before but since reading some articles on how it tries to, and quite successfully replicates 10bit colour in the way graduations are captured and processed.  EOSHD Andrew said something along the lines of that the profile is almost too good to have been included in the nex5n.   It would be interesting to work out exactly how Sony process the image in sunset profile to do what it does.  Must be some type of smoothing in the same way Neatvideo does it.

[chauffeurdevan]

Hi richg101,

 

I use 32 bit workflow so I was surprised with the 1DC's footage not holding better. 

 

Besides really heavy compositing, I don't care much about 32bit. 16bit should be enough for most grading. But that is another thread.

 

Try applying some noise (not grain) at a really low level (+/- 0,5% - but should be 100/256 = 0,39%), this will dither the image and reduce banding.

 

Try it before, and try it after grading effects.

 

Dithering will be even more important when you'ill reduce from 4k to 1080p

[FilmMan]

Thanks Hmcindie and Chauffeurdevan. 

 

Everyone should read this link.  It is done professionally (not by me of course) and shows what occurs. 

http://www.hdslr.es/foro/index.php?topic=1711.0

 

If you scroll down on the website, there is some info in English (from the manual) and from Hurlbut's blog.  I remember reading Hurlbut's quote and thought Canon would have fixed any issues.  Well, they still persist. 

[FilmMan]

I put in a call to Canon.  I'll see if I get a return call today.  I will bring up this 'banding' issue and see what their response will be.   

[HurtinMinorKey]

are you grading with a 32bit workflow?  so all your effects applied to the image are 32bit?  Your banding should disappear all together with a 32bit workflow, even with 8bit original capture

 

This suggests that the banding does exist in the 8bit pregrade. In this case the 32-bit grade won't make a difference. 

[Bruno]

Have you tried using 5DtoRGB? Not sure if it has the settings for the 1DC compression type yet, but it does a great job with 5D/7D footage. It gets rid of a lot of compression artifacts and does a good job at transcoding to prores at a higher bit depth. It doesn't do any miracles, but the improvements to the footage are quite noticeable, especially if you're gonna grade.

Denoising software is usually also effective when trying to improve the bit depth, it's at least better than just directly transcoding to a higher bit depth codec.

[FilmMan]

Hi Bruno,

 

Tried Denoiser, its effective depending upon the footage.  I've been testing Canon Log more.  Using more range seems to work better.  Still premature.  Been getting better results now.  It seems how the footage is shot is very important.   I took a shot from the inside of the house looking out into the brightness.  Got good results  Thanks for the response.

[Andrew Reid]

It looks like the codec is allocating less bitrate to smooth gradated areas with little detail, like a painted wall for instance.

 

In the shadows it will allocate even less bitrate so compression and macro blocking will be worse.

 

As for 8bit, not much of a fix for that. There's only 255 luma levels so subtle gradations aren't possible. There will be big bands to get from one shade to another. It won't be smooth. May even be dithered, which has a rough texture and more noise in it.

[FilmMan]

Andrew,

You're right.  It seems that way.  I have contacted Canon and forwarded info to a rep.  I'm waiting to see what they will respond.  Hopefully a firmware update - wishful thinking.   

 

Article for people interested on color bit depth.  The picture of the banding (8 bit gradient) is showing up at times (with Canon Log after grading).

http://provideocoalition.com/pvcexclusive/story/color_depth

[Andrew Reid]

It is unlikely to be resolved.

 

My advice is to crush the blacks and avoid Canon LOG, shooting with a standard picture profile and exposing just spot on every shot.

 

This isn't the camera to do heavy grading on.

[Bruno]

I have to say that it's highly unfair to judge that footage based on the youtube version, can you find a way to share the original footage?

 

I watched the master projected in 70mm and it looked stunning, but if I was judging it from a youtube clip like the following one, I probably wouldn't have thought much of it, it doesn't look any better than DSLR footage with the crappy compression.

 

http://www.youtube.com/watch?v=QzqJMIN3Yco

[FilmMan]

It is unlikely to be resolved.

 

My advice is to crush the blacks and avoid Canon LOG, shooting with a standard picture profile and exposing just spot on every shot.

 

This isn't the camera to do heavy grading on.

Thanks Andrew.    I was thinking of shooting with neutral and not use log.  However, the log does work in certain conditions.  I want to test it more on shooting outside in good light and see how it holds.  

The grade can be pushed a fair bit too. I wish Canon would give a little more fire power to this camera (10 bit color). 

 Images are very detailed and quite amazing - sometimes not as amazing when I mess up.    Cheers.

[FilmMan]

Just looked at that file of my son.  In Canon Log there was banding on his neck before I color graded!  I looked at the footage again before any grading.  I remember someone posting that it was probably on the recorded log file.  They were right. 

[Leang]

Canon 1DC & 8mm, H. Chretien Baby Hypergonar anamorphic lens

 

http://www.youtube.com/watch?v=jDnM-HBb4E8

[FilmMan]

Thanks for posting Leang.   I still have to try some night time shooting with the 1DC.  I've been playing around with Log and Neutral under different lighting.  Also playing with codecs in post.  The 1DC is better than most realize but a draw back (in my opinion) is the 8 bit.  Andrew has posted a jpg of the 8 bit(1DC) vs 10 bit (Sony F3).  It is very true.  However under good lighting it is less of an issue (from my limited testing).  It is still a pain as it restrict.  Denoiser does help.   If Canon would give 10 bit, I think it would be a major benefit.   Cheers.