Is there anything stopping Canon from making a 4mp Video DSRL thing?

[Tim Polster]

Hello,

I have been watching this market evolve as I use Canon still cameras for still work an other "video" cameras for video work. Just recently have I started experimenting with the DSLR's for video applications.

When I step back, I am a bit puzzled at the marketplace. On one hand, there is the RED company which has a $17,500 body which has a 11.5mp sensor that was designed to function in a motion picture environment. I have read that this cameras has had heat issues through it's development. Even though this camera has a 11.5mp sensor (which would give it relatively large photosites), it's ISO is not off of the charts.

Then we have the 5DMKII which sells for $3000, has 21mp and was designed for still use. This camera is limited to 12min of recording due to memory limitations but is that a heat limitation as well? Compared to the RED, the 5D seems to be more flexible with ISO and light gathering even though it has way smaller photosites.

So I wonder, if Canon could make a new sensor the size of the 5DMII sensor but at 4mp for video only, the ISO would be off of the charts. What is stopping them? Is it technical? Is it a relative low demand or market size?

RED and its future products was supposed to turn the industry upside down, but from my view, it looks like Canon has that ability through much better sensor development and user install base.

Will they do it? And how does RED fit into this landscape? Scarlet does not seem as groundbreaking anymore...

Your thoughts for discussion?

[Chris Hurd]

Back in January, I took my best shot at answering this question in as much detail as I could muster.

Read it here: APS-C or Full Frame CMOS Sensors? Not Happening in 2010

Then please report back to discuss...

[Tim Polster]

I agree with your entire article. The different style lenses and the established broadcast video market are huge.

But as I watch this niche develop, I am seeing a market where these cameras are mainly meant to be used with still lenses. Not so much in a traditional video role but as a specialty video role.

For example, I have resisted even testing my cameras for video until a month or two ago because I was thinking, "it can't be all that" and "what would I use it for?"

Well after testing I realized that this camera (if it had better audio) could be used quite lot in certain circumstances given its price and my ownership of Canon still lenses. It would never replace video cameras, but it could augment my video cameras and ease the burden off of them to be everything at all times.

To be precise, in the video camera market, you pay big for semi-shallow depth of field - 2/3". Well by owning a VDSLR or two, you would not need to spend so big to get this look and could happily use 1/3" ro 1/2" cameras for normal video work and use the VDLRs for interviews, cine applications etc...

Which brings me to my point. How expensive or difficult is it for a company like Canon to take an existing chip and decrease the megapixels?

The cameras I am seeing do not need all of the still resolution nor all of the still technology inside. Rather then re-invent the video market, I think it would be wise for Canon or any other maker to define a niche. but quick action is important as this space will fill up quickly due to the relatively low resolution requirements.

This approach uses existing Canon lenses, does not need any R&D for new lenses, and needs a much lower mp sensor in a more video friendly housing. Does not seem that far of a stretch to me...am I off base?

[Henry Olonga]

I thought RED was 4K - almost 12 and a half megapixels. Official specs say full array is 4900 (h) x 2580 (v)

I agree with you however that an ultra high megapixel count for video is not needed for 1080p delivery. Higher than that and there are advantages to oversampling e.g One of the Scarlets is 3k for a clean 2k or 1080p delivery.

Quote Tim

Which brings me to my point. How expensive or difficult is it for a company like Canon to take an existing chip and decrease the megapixels?

......... wouldn't that lead to line skipping or binning to decrease resolution - the bane of the current crop of DSLRs? Or at least a very high quality downsample would be needed - which Canon do not do well at all on their current cameras as evidenced by the aliasing.

[Tim Polster]

Specific wise, I don't know the details of how much line skipping (if any) can be used without negative effects.

I should have used 2.1 megapixels as that is 1080p but figured some would think that as limiting. But this sized chip would be even less demanding from a still-photo perspective.

And I am wrong about RED, it is listed about 11.5 megapixels. I thought I had read it really only shot physically at 2k. I appologize for mis-stating the RED info.
(I have updated my original post)

[Evan Donn]

2.1Mp wouldn't be enough because you need to divide the photosites up between red green and blue. So to get a 1:1 sampling (4:4:4) for all three colors at 1920x1080 you'd need 5760x3240 or 18.7mp. Or you use a bayer pattern for 4:2:2 and you could go with an 8.3mp sensor. To allow for up to 2K resolution plus a little oversampling you'll want something in the 10-12mp range... kind of like Red. Or you need to go to a three chip system - which I imagine would result in a heavy, bulky lens block to house the sensors and beam splitter.

So Canon could produce a DSLR that shot better video - it would have significantly improved sensitivity and none of the current aliasing problems - if they went to something like a FF 12mp sensor. They've shown willingness to go this direction with the G11 vs. the G10. Unfortunately I don't think the DSLR market would bear it as it would probably severely limit the photographic market for the camera, which is significantly bigger than users buying these for video only.

[Tony Tibbetts]

I've always wondered why they can't get the best of both worlds. I could be wrong here, but why not make an XL style camera with a full (or APS-C) size sensor that takes EOS lenses and include a lens mount adapter to accept XL lenses. Once the adapter is attached the camera will only scan the center 1/3rd of the chip.

Again I'm not a big tech guy, but this sounds feasible to me. Maybe the flange distance would be an issue.

Is it feasible? I know there are guys on these boards with a lot of knowledge here that could weigh in on this.

Heck, maybe this has already been discussed.

[Tim Polster]

Originally Posted by Evan Donn

2.1Mp wouldn't be enough because you need to divide the photosites up between red green and blue.

Evan, it is clear you know more about this than I do, but I am wondering what the 5DMKII does now when it shoots video?

Does it take the 21mp chip and go down to the ~12mp range you are speaking of?

Or does it line skip all the way down to just get 1920x1080 out of the chip?

I do not think we will get a product from Canon if it is too far in the video R&D camp. If they can re-tool and existing design/chip then I think it would be less of a financial and market risk for them.

Thanks for your reply

[Evan Donn]

I don't know the exact details of the 5D's implementation, but my guess is that they line skip down to every third vertical line and then pull RGB from adjacent areas on each line - although I'm sure the details are much more complicated than that.

Originally Posted by Tony Tibbetts

I've always wondered why they can't get the best of both worlds. I could be wrong here, but why not make an XL style camera with a full (or APS-C) size sensor that takes EOS lenses and include a lens mount adapter to accept XL lenses. Once the adapter is attached the camera will only scan the center 1/3rd of the chip.

The difference between chip sizes is just too large for this to be feasible. XL lenses are designed for 1/3 inch sensors, which are about 1/7 the width and height of the FF sensor. So XL lenses would only be useable as super-telephoto. Plus you'd need the overall pixel density to be much higher in order to have enough resolution in the windowed area that an XL lens would use.

Video capable DSLRs are getting cheaper and more common every day. There are a ton of great options in 1/3" cameras. The solution isn't to try and combine them into one - it's simply to have both and use whichever is appropriate for the shot at hand. The great thing about that approach is that there's no need to wait for Canon or anyone else to do anything - you can have the best of both worlds right now.

[Tim Polster]

Thanks for your reply.

I know this is tough to armchair quarterback camera design. The cost of a new camera might not be what everybody is expecting either given the much smaller video market and the non-still friendly design.

Maybe an easy approach would be to release a 5DMKIIv that is a little more expensive but has an XLR or two and some video features. This camera could be labeled for the extra taxes and would not require much of a change to the existing 5DMKII?

[David Chapman]

This is all very interesting. I know it's easy to think of a combo camera in an XL2 body or something, but to me the deal isn't as much of "how do we get this concept to work in a video camera body with video features?" but more of a "how do we get the footage into a sRAW-type codec for larger than 4:2:0 8bit compression?"

The video side of Canon is really excited about their 4:2:2 HDV codec on their new 1/3" video camera, but you would think it would be easier to create some device that could attach to the 5D/7D via hdmi (after firmware updates resolves downscaling video and red record dot) to record that HDV 50Mbps. This device would be a lot cheaper to produce for our "niche" market that we are all in and would be an instant sell catering to us and not worrying about creating a hybrid camera that photographers are scared of.

[Evan Donn]

Problem with that solution is that I don't think the necessary changes to the camera itself are as simple as a firmware update. The camera downscales the way it does for a reason - probably a combination of limited processing power plus battery life and overheating issues. I wouldn't be surprised if it drops the external signal resolution for a similar reason. I just don't think it's within the current hardware capabilities to do a clean downscale of the entire sensor and send it out as full res over HDMI.

I'm pretty sure we will see a more video oriented camera built around these sensors at some point, but my guess is it won't be for another couple of years. Canon's never been one to rush into a new video market, and the fact that they've emerged as a leader in video for DSLRs seems to be something they didn't fully anticipate. And the facts remain - they currently have little to no competition in this particular market, they're selling these DSLRs in their current configuration about as fast as they can make them, and the largest market for these cameras is actually to people who aren't interested in them primarily for their video capabilities. They'll get around to it when it makes sense to do so from a market perspective, and they'll have several years of experience and feedback from us crazy early adopters to ensure that what they do finally ship is still pretty competitive when it hits the market.

[Jon Fairhurst]

CAMERA SHOPPER

"Should I buy the Nikon or the Canon? What do you think?"

DVINFO MEMBER

"Will you shoot video?"

CAMERA SHOPPER

"It's really for photos, but I might shoot some video here and there"

DVINFO MEMBER

"Canon is the only choice for video. Nikon is way behind on video. For photos, both make great products."

CAMERA SHOPPER

"Okay. I'll get the Canon then..."

ME

I think video matters. ;)

[William Japhet]

Hi,

as you said : Canon knows how to produce HD Video only DSLR with small sensor. But...
Never forget that Canon is also a lens manufacturer, for broadcast world, for Sony, Panasonic, Grass Valley. Some very expensive lenses... And expensive Cinema lenses. Not a little photography prime. A big market.

They could not create this HD camera without annoying his broadcast world. They have already changes the market with the 5D and 7D. They can't make a crisis in the cinema market now...

RED had made a little revolution, but they were unknown before. Canon is Canon...

William

[Antony Newman]

HD output. I think this is a very interesting question.

I think it depends on:

Q1) If the kernel (ie Dryos) stealing all the power of the ARM (eg to make the GUI responsive)?

Q2) Is there a limitation preventing output 1080p via HDMI (either memory bandwidth, or HDMI hardware limitation)?

Q3) Is the ARM already able to access a the raw HD data already, or is it being passed the h.264 compressed image?

Q4) What the heat is the issue is caused by (assuming the issue is heat)?


My thoughts so far:

A1) What I have seen of the Dryos code is not optimised Assembly code. (which is good for us if there are simple inefficient block_moves that can be rewritten)

A2) I am hoping this is a software limit. I am guessing most recent HDMI hardware devices that can negotiate at 1080i can probably do 1080p.

A3) Don't know.

A4) The sensor can concurrently scan at 24Hz x 1080p, LCD at 480i, processing sound at 48kz (with more noise - doh!), compress the 1080p image into .264 and writing the compressed video to the Compact flash.

It can (alternatively) concurrently scan 24Hz at 1080p, LCD at 1080i (not sure if the screen image is compressed), also and process 48Khz sound.

And what I am wondering is, if all the power used for .264 compression is ASIC intensive, and displaying the LCD is intensive, and moving data to the LCD the 480i (1080i) to the LCD, and writing to the CCF was switched off and instead used to write 60% more data to HDMI than when it was doing 1080i x 30hz ... would we be onto a winner? What if we switched off all the sound processing too?

-
Now there is lot of What ifs. If I was canon, I wouldn't try and deliver these features in a 5d mk ii. I'd put this on the cherry of my next camera; Those with a serious interest in video will upgrade when the time comes. If they go all out to make the 5d deliver this (HD) now .. what would be the big cherry that would make you want to buy another body in two years time?

If we want this now - its going to be a collaborative effort outside Canon. There is a lot of code to understand - an awful lot of it. If 100 people started analysing the code - we may just get somewhere!

Antony