Rumour: Canon 5D Mark 3 to feature new MPEG variant 4-2-2 at 50Mbit

If they can figure out a way for the camera to properly create a 1920x1080 image without line skipping, and include a clean HDMI output from that...I'de pick up one of these in a heartbeat. Would give me a good image, but also allow me some great monitoring options that currently are a pain on the MarkII. Of course, others are thinking HDMI and external recorder if the HDMI is clean, but I don't want to be picky! :)
 
Has anyone ever kept a scorecard of how often these rumor sites get it right? I would think that would be a quite illuminating exercise.
 
don't really understand what 4:2:2 color sampling has to do with smooth gradients...
 
EOSHD is the National Enquirer of HDSLR blogs. The guy writes great headlines and questionable editorials -- occasionally very right and often not.

He's just now written two back-to-back articles on what he HOPES the 5Dm3 will be (probably what we all hope). Unfortunately, he's got no real sources and his hopes are his hopes.

The real question is -- when was the last time Canon delivered something for which we all hoped? Answer: never.

What we should really be hoping is that they don't screw up the 5Dm3. You know, lower the bitrate in some effort to boost camcorder sales, etc.
 
He probably got it from our thread in the 5d forum titled "speculating on the 5dmark3 (check out the first post)....this was my idea first!
 
This may be BS, but it's probably one of the more realistic rumors. However I don't so much care about the codec if what they're doing before the video hits the codec hasn't changed.
 
The rumor itself is feasible. It might happen. The camera already does bitrates around 50 mbps (at times) and MPEG2 should actually be less CPU intensive than the H264 it already is doing. But Canon might decide that H264 plays better with consumer hardware and stick with that.

But I agree the source of the rumor isn't really reliable. The first clue is how the rumor tries to justify itself with an explanation: they're picking this codec so they don't have to pay licensing fees to Panasonic and Sony.

Real leaks usually don't have any explanations. They're just short and to the point. They don't know why it was done that way...only the top guys in Japan know. Also, H264 is licensed by MPEG LA, while AVCHD is licensed by Panasonic and Sony. AVCHD uses H264, but Canon doesn't necessarily have to sign up for AVCHD to do H264 (I think).

The real question is -- when was the last time Canon delivered something for which we all hoped? Answer: never.

Actually, Canon has delivered on a few occassions. Canon managed to come through with the 5D firmware upgrade with manual control (they never added a major feature like that in a free firmware upgrade before AFAIK). The 7D stunned everyone with manual control and a full range of frame rates. Then the T2i continued that trend at almost half the price! Then the 60D added an articulated screen.
 
don't really understand what 4:2:2 color sampling has to do with smooth gradients...
4:2:2, by itself, would have nothing to do with it. The things that improve gradients in a codec are 1) 10-bits, and 2) adequate bandwidth. A much higher bandwidth codec (or a much more efficient codec!) could reduce gradient banding substantially. I do see less banding in the NanoFlash recordings than in the native AF100 recordings, for example. If you've got more horsepower to throw at the situation, you don't have to take as many shortcuts (and gradient banding is one way a codec takes shortcuts).

10 bits would obviously greatly increase the resolution of the shades of gray (or shades of color) that the codec can represent, but would be wasted without enough bandwidth to take advantage of all that.

I don't know that 50mbps 4:2:2 would result in that much of a reduction in gradient banding. I would think you'd probably need more on the order of 100mbps of MPEG-2 to get the job done, or 50mbps of (fully implemented IBP) h.264.
 
they're picking this codec so they don't have to pay licensing fees to Panasonic and Sony.
Also, H264 is licensed by MPEG LA, while AVCHD is licensed by Panasonic and Sony. AVCHD uses H264, but Canon doesn't necessarily have to sign up for AVCHD to do H264 (I think).
You are correct. Canon is already paying license fees to MPEG LA to use h.264. Panasonic and Sony own AVCHD, which is a specific implementation of h.264, and anyone using it will have to pay a fee, although I don't know if that goes to Pan/Sony or if it goes straight to MPEG-LA.

Funny thing is -- MPEG-2 is *also* licensed by MPEG-LA, and those who use it have to pay royalties on it. So if they go to MPEG-2, they'll still be paying royalties...
 
Yeah I definitely agree...the XF codec is just such a logical step for Canon, I really think this will happen (not because of a rumor, I posted about it 2 days ago ) but because of where the arrows seem to be pointing. Now for that Digic V that makes it possible to get a relatively alias/moire free GH2 like 1080p signal from the sensor! oh and the picture styles including different gamma curves.
 
Yeah I definitely agree...the XF codec is just such a logical step for Canon, I really think this will happen (not because of a rumor, I posted about it 2 days ago ) but because of where the arrows seem to be pointing. Now for that Digic V that makes it possible to get a relatively alias/moire free GH2 like 1080p signal from the sensor! oh and the picture styles including different gamma curves.

Sorry ...I was replying to a comment (gabrobot's) 5 posts ago...I need to type faster.
 
The rumor itself is feasible. It might happen. The camera already does bitrates around 50 mbps (at times) and MPEG2 should actually be less CPU intensive than the H264 it already is doing. But Canon might decide that H264 plays better with consumer hardware and stick with that.

Well, I'm thinking 50Mbps mpeg-2 is actually a significant step backwards from 44Mbps mpeg-4.

Having spent the past two years working with various forms of mpeg-4, I could swear I can see the difference between it and mpeg-2.

That being said -- I can't imagine Canon sticking their shiny new "broadcast" codec in a DSLR.
 
Well, I'm thinking 50Mbps mpeg-2 is actually a significant step backwards from 44Mbps mpeg-4.

Having spent the past two years working with various forms of mpeg-4, I could swear I can see the difference between it and mpeg-2.

But this is 4:2:2, with twice the color data of 4:2:0 MPEG 4 implementations. The specs are basically the same as Sony's XDCAM HD422 codec. Other advantages are broadcast approvals, NLE compatibility, easier editing and industry standard MXF container. Also, having 4:2:2 color space is an easy way to distinguish yourself from the other cameras that are recording in 4:2:0 onboard. It's an easy talking point for the sales rep: "Hey, we've got twice the color of those other guys and you don't even need a separate offboard recorder!"

The downside is a consumer might not know what do to with an "MXF" file. This could cause confusion that Canon doesn't want. So the decision whether to use this codec or not might be more of a marketing one, rather than a technical issue. I bet the camera can handle the 422 codec. I personally would prefer it.
 
But this is 4:2:2, with twice the color data of 4:2:0 MPEG 4 implementations. The specs are basically the same as Sony's XDCAM HD422 codec. Other advantages are broadcast approvals, NLE compatibility, easier editing and industry standard MXF container. Also, having 4:2:2 color space is an easy way to distinguish yourself from the other cameras that are recording in 4:2:0 onboard. It's an easy talking point for the sales rep: "Hey, we've got twice the color of those other guys and you don't even need a separate offboard recorder!"

The downside is a consumer might not know what do to with an "MXF" file. This could cause confusion that Canon doesn't want. So the decision whether to use this codec or not might be more of a marketing one, rather than a technical issue. I bet the camera can handle the 422 codec. I personally would prefer it.

I agree with the majority of your post, except I think most consumers willing to spend at least $2500 ($3000/4000 + with a lense) on a camera and want to use the video feature will be a well informed demographic. Putting it on the entry level models might be a different story (even in that case though I think in general people stepping up to DLSRs usually do some research).

Case in point
The cost of this camera will be essentially more than the XF100.
 
50mbps mpeg-2 should look cleaner than mpeg-4 at those bitrates, especially if it's 4:2:2. The differences between mpeg-4 and 2 are huge with low bitrates but negligible with high.

Also I don't get this clamoring for b-frames for the mpeg-4 variant in Canon DSLR's. When the bitrate is high enough, the b-frames only damage the picture. That's why the higher the bitrate, the less complex encoding should take place. I-frame is over 100mbps for a reason and it has nothing to do with "efficiency".
 
How long has this EOSHD animosity been going on now? Come on - almost 2 years now. I call a truce and I'd be happy to meet some of the DVX guys like Barry at NAB as well, if it means smoothing over the misunderstandings, although it is a long way to go to get my head smashed in (joke).

My site is not just about rumours. Once in a while something makes sense. I do have sources. The DSLR scene is a community of real people don't forget. Very occasionally employees of Canon and Panasonic - not at the Japanese HQ, mostly in marketing and training, come forward with information absolutely unprovoked. They only do so on the condition that I don't name them - not even a hint. The stuff I've been told about the MPEG codec goes along with what Tim Smith said to Jarad Abrams in 2010 as well, about Canon going their own way and producing their own codecs, like they did with the XF codec. It all checks out just fine. I have no need to do National Enquirer style headlines, it doesn't make the hits go up and more than they already have.

Peace!
 
Last edited:
Back
Top