C100: AVCHD better than Prores?

L

lexicon

Well-known member
Björn Kurtenbach shared a comparison he shot on a Canon C100 that shows off the difference between highly compressed internal AVCHD (420, 8-bit, 24 Mbps), external ProRes 422 (422, 8-bit, 122 Mbps), and ProRes HQ (422, 10-bit, 184 Mbps), recording to an Atomos Ninja Star external recorder via the uncompressed HDMI output.

conclusions:
pretty good implementation of AVCHD on the C100: internal AVCHD is ok for most of the jobs; differences are insignificant with ProRes 422, more noticeable with ProRes HQ (specially on motion and color).

 
Thank you. That was a great video. My conclusions:

  1. If you're going to shoot ProRes, shoot HQ.
  2. ProRes 422 HQ is a good codec for demanding scenes. It has the static detail of AVCHD but none of the artifacts in complicated motion.
  3. The differences between all of these codecs is visible only to pixel peepers.

Decisions should be a balance of all pros and cons, and I think ProRes 422 HQ is the best balance of file size, editing ease, and image quality, if you're shooting a wide variety of things and can afford the media --- what is it, 80 GB an hour at 24 fps?
 
Thanks for sharing that excellent video. I found it very informative particularly the codec behavior for static and motion.
 
AVCHD is not as NLE or FX friendly to edit with, especially if you need to modify what comes from the camera very much. Visual quality is quite good for playback though. All i-frame codecs with bit rates higher than 50Mbps are much better for editing. 4:2:2 10 bit codecs are better for color grading and FX work in general.
 
I'd be curious to see the difference between the three after some very light grading. I loathe the idea of using an external recorder but recently have been asked to deliver broadcast spec (50Mbps). If they're worried about degradation during the broadcast chain, why not shoot AVCHD, then edit and export as prores?
 
If you don't spend much time with your footage in post, shooting internal AVCHD footage is all you need. Recording externally to a 10 bit dnxhd container makes a MASSIVE difference in push/pull ability in post. When I'm not shooting 10 bit externally, I'm recording avchd internally.

Do some tests:
- undere/overexposing
- keying, rotoscoping
- lifting shadows
- skin tones
- noise quality
- recovering highlights
- camera motion
- green screen

A while back I posted a link to an article that detailed how the c100 was outputting more than 8 bits of data when recording to the 220x option on the ninja blade(dnxhd 10 bit).
 
It's great video, but I feel he should have used the working "ProRes implementation".

Obviously ProRes 422 in itself has no problem with foliage. Unfortunately, the Atomos implementation in this specific recorder isn't quite there yet.

Still, it will be tiresome reading posts referring back to this video that say that ProRes 422 sucks. Oh, well.
 
xocomil said:
If they're worried about degradation during the broadcast chain, why not shoot AVCHD, then edit and export as prores?
Click to expand...

It's all about various compression artefacts that you often can't see at the mastering stage. Once it gets mauled by the broadcast MPEG encoders, those artefacts start chewing up the available bandwidth in the multiplex signal.

So, here in the UK for example, you might have a programme on BBC3 - shot on H264/AVCHD - suddenly demanding more bandwidth than Channel 4, resulting in the Channel 4 programme (and the commercials in between) looking more blocky and pixellated - or dropping out completely.

They don't like that, so the 50 Meg standard was agreed by all the broadcasters (who make up the EBU) and is rigidly enforced by the bitrate police.

Ben.
 
Ben - so the compression artifacts are from the source footage, regardless of whether the video has been transcoded to prores for editing and delivery? And these artifacts demand extra bandwidth when the program is broadcast, taking it from shows being broadcast simultaneously on other channels? That's wild!
 
xocomil said:
Ben - so the compression artifacts are from the source footage, regardless of whether the video has been transcoded to prores for editing and delivery? And these artifacts demand extra bandwidth when the program is broadcast, taking it from shows being broadcast simultaneously on other channels? That's wild!
Click to expand...

Yes - bear in mind that, to get to ProRes, the transcoding process involves decompressing the Long GOP Inter frame AVCHD to baseband/uncompressed and then recompressing it to Intra Frame ProRes. In the process, new artefacts get introduced. The broadcast business often involves moving media between different facilities, so further compression/decompression is expected along the chain, before it then gets squeezed down to a Transport Stream for playout.

I'd wager that an AVCHD-ProRes transcode would actually be fine - it's just that the standard had already been agreed.

And that's the principle reason the C300 costs so much more than a C100 - cos it can.

Ben.
 
Thanks Ben. So would a camera like the XF100 (Mpeg 50 Mbps 4:2:2) have inherently less artifacts recording the same scene as a C100 (in spite of higher noise in low light, possibly poorer optics, smaller sensor, etc)?
 
xocomil said:
Thanks Ben. So would a camera like the XF100 (Mpeg 50 Mbps 4:2:2) have inherently less artifacts recording the same scene as a C100 (in spite of higher noise in low light, possibly poorer optics, smaller sensor, etc)?
Click to expand...

I don't know - I'm only referencing broadcast standards that were put in place ages ago. If you're publishing to the web, it's academic.

Ben.
 
In the video he mentions so many relevant things and pixel peeping is the only way to see the issues in ProRes 422. The viewer wouldn't see it. If you don't go in 400% you wont see it but color reproduction is something you will see without pixel peeping and that is a pretty solid trade off for a larger 422 codec file.

AndreeOnline

"It's great video, but I feel he should have used the working "ProRes implementation".

Obviously ProRes 422 in itself has no problem with foliage. Unfortunately, the Atomos implementation in this specific recorder isn't quite there yet.

Still, it will be tiresome reading posts referring back to this video that say that ProRes 422 sucks. Oh, well."

I never heard this before. Are you saying Atomos changed the way the Ninja Star or other models record the ProRes codec? If this is the case then is the Convergent Design products or other recorders better at ProRes implementation?

I think the presenter was very clear and not making a statement that ProRes sucks and AVCHD is better. Weighed the pros and cons fairley and even made fun of pixel peeping to see the issues.
 
Last edited:
I don't know if its just me but the Prores HQ seemed noticeably sharper too, not just at 400% but viewing the full clip. And yes, color looks much better. Obviously as many have said color grading will be the real test, I still think there's lots of benefit to using an external record, especially how easy and accessible it is these days with the ninja star.
 
Erik Naso said:
I never heard this before. Are you saying Atomos changed the way the Ninja Star or other models record the ProRes codec?
Click to expand...

Hi Erik,

I am not trying to slander anyone, and I don't own Atomos or Convergent products so I have no stake in it.

What I am saying is: if you gave me the uncompressed HDMI stream as a file on my computer (and we are assuming it is at least as good as the AVCHD internal—meaning the camera isn't doing anything funky) and I let FCPX spit it out as a ProRes 422, it would be visually lossless without any artefacts whatsoever.

That means I have to conclude that something with the Atomos implementation is off. Maybe they just need to know about it? Some kind of bug? Maybe a performance issue with these small recorders?
 
Prores is intended as a robust intermediate codec specifically for editing. AVC in long GOP non I-frame versions is not and NLE's do not always handle long GOP formats well. Best to transcode to Prores, edit, then export in whatever codec is required for final delivery. There will be less degradation of the original than editing AVC if you do much more than simple cuts only doc style production. Exporting an edited long GOP AVC to some other long GOP delivery codec , typically H.264, after editing and FX can pick up significant artifacts.
 
You must log in or register to reply here.
Back
Top