CFHD- A Waste of Time and HD Space??? at DVinfo.net

Robert Young · October 8th, 2010, 07:40 PM

I could use a little intellectual firepower here.
I'm basically getting trashed on another forum for using CF for AVCHD sourced editing.
A rather elaborate rationale is being offered to support the notion that AVCHD, edited natively, is not even just as durable as CFHD.avi, but more durable.
It goes like this:
Part I: In a modern NLE (CS5), the AVCHD codec sitting on the timeline is decompressed one time only to uncompressed frames by the NLE. You can then apply effects, color correction, MBL, animated graphics, whatever, and it's all applied directly to these great big uncompressed frames- no harm done.
Part II: (This is where it begins to sound funny to me) The big, now edited, uncompressed frames just sit there in the NLE, never being recompressed to AVCHD unless you command the NLE specifically to do so.
When you render out to delivery formats, the NLE will do the render from the uncompressed frames, not from a compromised, recompressed AVCHD codec. Therefore, image quality is pristine.
Part III: (Really a stretch for me now) So, If you are converting from AVCHD and editing in CFHD, your final image quality actually will be poorer than a native AVCHD edit because you are decompressing the AVCHD only to then recompress it into CFHD.
This all certainly contradicts my years of experience with CF HD editing, but, not being a tech guy, I don't have a good response.
So, to the smart guys out there- True, or False?? Details...

Adam Gold · October 8th, 2010, 07:45 PM

Oops, never mind, you've already been active in the thread I originally pointed to... I was confused because your thread title refers to time and space but the post itself is about quality, and I think that the answer will depend upon which is more important to you.

Robert Young · October 8th, 2010, 08:43 PM

The title of the thread is a quote from my CF critics:
"Editing in CFHD is a waste of time and hard drive space"
I, myself, have no problem with either time, HD space, or quality :)
At issue is the position that, in an up to date NLE, editing AVCHD sourced footage in CFHD offers no image quality benefit to the final delivery product, compared to editing in native AVCHD.
I believe that this is untrue, but I do not understand the technology of exactly how the AVCHD codec is actually handled in the NLE well enough to argue the case. Is AVCHD truely only decompressed once, and never recompressed as you edit, render out, and etc., or not?

David Newman · October 8th, 2010, 09:02 PM

You can argue that 1 through 3 is technically correct, the falsehood comes from thinking the source is perfect/ideal, when it is anything but.

Some points missed.
a) The above argument requires that you stay within the NLE for all processing (not using outside tools), and that you never return to enhance, edit, modify an existing export. Also assumes none of the following it true. :)
b) While NLE could expand 4:2:0 to 4:2:2 to 4:4:4, just as we do, they don't do a very good job. Up-filtering helps in color correction and looks more natural.
c) There is actually a subject quality gain is storing an 8-bit source into a 10-bit wavelet (this is not true for DCT codecs like DNxHD and ProRes), the wavelet is a continuous image representation, removes some of the step discontinuities of the source, allowing you more flexibility in color correction. This is mathematically true, and it reported to me weekly how much better you can color correct a CineForm file than both native and other digital intermediates.
d) Many NLE filters are still RGB 8-bit, this can result is colorspace clipping (losing highlights and shadow detail) and 8-bit banding, which can all be avoiding using FirstLight to do your base corrections, even if downstream some 8-bit filters have to be used.

So the if source is sacred, and you want to reproduced it with no adjustments, believe the other argument (and if you stay within one video tool/NLE.) If you are more the creative type, and you want control over the look of the final image, you will do better sticking with CineForm.

Robert Young · October 8th, 2010, 09:13 PM

Thank you, David.
So, it's sort of like "True, True- Unrelated"
If you are so inclined, you can believe whatever you need to.
I'm noticing, as the NLEs and systems are getting better at handling native formats, I'm getting more flack about recommending CFHD.
I think it really revolves around what kind of editing these posters are doing- trim and stitch, vs. more complicated projects.

David Newman · October 8th, 2010, 09:22 PM

Yes, CineForm is optimized toward independent film crowd than "trim and stitch" as you put it.

Alex Raskin · October 8th, 2010, 10:01 PM

In my view, CFHD produces controlled and known results. With internal NLE processing, on contrast, you are relying on the processes that change from release to release, and from NLE to NLE.

FirstLight, which comes with CF, is fantastic in quick color correction that comes without any rendering time.

If you capture in CFHD natively, like I do with SI-2K camera, then you definitely want to stay in the CF realm in post.

Also... With vDSLRs like Canon 7D, native footage's timecode is not correctly recognized in Premiere CS5. However, if you transcode the native footage into CFHD with HDLink (from the original media with THM files present), then HDLink apparently re-stamps the footage correctly with the timecode recognizable by the Premiere (plus THM gives it 1/100s of a second timecode resolution.) Could be just my setup, but it seems to work for me that way.

Robert Young · October 9th, 2010, 12:45 AM

Quote:

Originally Posted by Alex Raskin

In my view, CFHD produces controlled and known results. With internal NLE processing, on contrast, you are relying on the processes that change from release to release, and from NLE to NLE.

FirstLight, which comes with CF, is fantastic in quick color correction that comes without any rendering time.

Amen.
Exactly my experience as well (since PPro 1.5). CF neutralizes all the quirks and weirdness that infest this entire process. Doesn't matter which version of NLE, what the source codec, etc., etc.
If I want to do a project, simple or not, and deliver top quality DVD (or whatever) images, I know before I even start that my workflow will consistantly provide the expected result. No surprises, no last minute disappointments, hand wringing, racing the clock... all the drama.
And First Light... priceless

October 8th, 2010, 07:40 PM	#1
Robert Young Trustee Join Date: Mar 2005 Location: Coronado Island Posts: 1,472	CFHD- A Waste of Time and HD Space??? I could use a little intellectual firepower here. I'm basically getting trashed on another forum for using CF for AVCHD sourced editing. A rather elaborate rationale is being offered to support the notion that AVCHD, edited natively, is not even just as durable as CFHD.avi, but more durable. It goes like this: Part I: In a modern NLE (CS5), the AVCHD codec sitting on the timeline is decompressed one time only to uncompressed frames by the NLE. You can then apply effects, color correction, MBL, animated graphics, whatever, and it's all applied directly to these great big uncompressed frames- no harm done. Part II: (This is where it begins to sound funny to me) The big, now edited, uncompressed frames just sit there in the NLE, never being recompressed to AVCHD unless you command the NLE specifically to do so. When you render out to delivery formats, the NLE will do the render from the uncompressed frames, not from a compromised, recompressed AVCHD codec. Therefore, image quality is pristine. Part III: (Really a stretch for me now) So, If you are converting from AVCHD and editing in CFHD, your final image quality actually will be poorer than a native AVCHD edit because you are decompressing the AVCHD only to then recompress it into CFHD. This all certainly contradicts my years of experience with CF HD editing, but, not being a tech guy, I don't have a good response. So, to the smart guys out there- True, or False?? Details... __________________ Bob Last edited by Robert Young; October 8th, 2010 at 08:47 PM.

October 8th, 2010, 07:45 PM	#2
Adam Gold Inner Circle Join Date: Jan 2007 Location: Woodinville, WA USA Posts: 3,467	Oops, never mind, you've already been active in the thread I originally pointed to... I was confused because your thread title refers to time and space but the post itself is about quality, and I think that the answer will depend upon which is more important to you. __________________ "It can only be attributable to human error... This sort of thing has cropped up before, and it has always been due to human error."

October 8th, 2010, 08:43 PM	#3
Robert Young Trustee Join Date: Mar 2005 Location: Coronado Island Posts: 1,472	The title of the thread is a quote from my CF critics: "Editing in CFHD is a waste of time and hard drive space" I, myself, have no problem with either time, HD space, or quality :) At issue is the position that, in an up to date NLE, editing AVCHD sourced footage in CFHD offers no image quality benefit to the final delivery product, compared to editing in native AVCHD. I believe that this is untrue, but I do not understand the technology of exactly how the AVCHD codec is actually handled in the NLE well enough to argue the case. Is AVCHD truely only decompressed once, and never recompressed as you edit, render out, and etc., or not? __________________ Bob

October 8th, 2010, 09:02 PM	#4
David Newman CTO, CineForm Inc. Join Date: Jul 2003 Location: Cardiff-by-the-Sea, California Posts: 8,095	You can argue that 1 through 3 is technically correct, the falsehood comes from thinking the source is perfect/ideal, when it is anything but. Some points missed. a) The above argument requires that you stay within the NLE for all processing (not using outside tools), and that you never return to enhance, edit, modify an existing export. Also assumes none of the following it true. :) b) While NLE could expand 4:2:0 to 4:2:2 to 4:4:4, just as we do, they don't do a very good job. Up-filtering helps in color correction and looks more natural. c) There is actually a subject quality gain is storing an 8-bit source into a 10-bit wavelet (this is not true for DCT codecs like DNxHD and ProRes), the wavelet is a continuous image representation, removes some of the step discontinuities of the source, allowing you more flexibility in color correction. This is mathematically true, and it reported to me weekly how much better you can color correct a CineForm file than both native and other digital intermediates. d) Many NLE filters are still RGB 8-bit, this can result is colorspace clipping (losing highlights and shadow detail) and 8-bit banding, which can all be avoiding using FirstLight to do your base corrections, even if downstream some 8-bit filters have to be used. So the if source is sacred, and you want to reproduced it with no adjustments, believe the other argument (and if you stay within one video tool/NLE.) If you are more the creative type, and you want control over the look of the final image, you will do better sticking with CineForm. __________________ David Newman -- web: www.gopro.com blog: cineform.blogspot.com -- twitter: twitter.com/David_Newman

October 8th, 2010, 09:13 PM	#5
Robert Young Trustee Join Date: Mar 2005 Location: Coronado Island Posts: 1,472	Thank you, David. So, it's sort of like "True, True- Unrelated" If you are so inclined, you can believe whatever you need to. I'm noticing, as the NLEs and systems are getting better at handling native formats, I'm getting more flack about recommending CFHD. I think it really revolves around what kind of editing these posters are doing- trim and stitch, vs. more complicated projects. __________________ Bob

October 8th, 2010, 09:22 PM	#6
David Newman CTO, CineForm Inc. Join Date: Jul 2003 Location: Cardiff-by-the-Sea, California Posts: 8,095	Yes, CineForm is optimized toward independent film crowd than "trim and stitch" as you put it. __________________ David Newman -- web: www.gopro.com blog: cineform.blogspot.com -- twitter: twitter.com/David_Newman

October 8th, 2010, 10:01 PM	#7
Alex Raskin Trustee Join Date: Sep 2003 Posts: 1,435	In my view, CFHD produces controlled and known results. With internal NLE processing, on contrast, you are relying on the processes that change from release to release, and from NLE to NLE. FirstLight, which comes with CF, is fantastic in quick color correction that comes without any rendering time. If you capture in CFHD natively, like I do with SI-2K camera, then you definitely want to stay in the CF realm in post. Also... With vDSLRs like Canon 7D, native footage's timecode is not correctly recognized in Premiere CS5. However, if you transcode the native footage into CFHD with HDLink (from the original media with THM files present), then HDLink apparently re-stamps the footage correctly with the timecode recognizable by the Premiere (plus THM gives it 1/100s of a second timecode resolution.) Could be just my setup, but it seems to work for me that way. Last edited by Alex Raskin; October 9th, 2010 at 02:56 AM.