The holy grail (Canon Raw Rumor)

shaocaholica said:
Not quite. If you read the link and quote in my post, it describes the format as a "partial debayer". Hence, its not as heavy as a full RGB image but its also not as sharp. However, at 2.5K, downsizing to 1080p will sharpen it up. But thats not that important. The important fact is that it retains the original bit depth values without gamma correction so you still have the full latitude and bit depth compared to true RAW. You just loose resolution which you couldn't have used for motion delivery anyway.

Ultimate point is that there is an efficient middle ground between full sensor RAW and what we have now that doesn't require lineskipping or partial readout or RGB bloat. I would wager that the current Canon sRAW algorithm is much faster than a full debayer and resize.
Click to expand...


hmmm ... so if canon went sRAW with a CMOS dedicated to video – in order to purely rely on sRAW demosaicing instead of line skipping and pixel binning, what would the height and width of the sensor be in order to yield 1080p. Would it be 3840x2160?

If it were 3840x2160 and aps-c sized that brings us back to larger than 7D photosites
 
J Davis said:
hmmm ... so if canon went sRAW with a CMOS dedicated to video – in order to purely rely on sRAW demosaicing instead of line skipping and pixel binning, what would the height and width of the sensor be in order to yield 1080p. Would it be 3840x2160?

If it were 3840x2160 and aps-c sized that brings us back to larger than 7D photosites
Click to expand...

Current 7D sRAW is 2.5k which I think is perfect for 1080p delivery. I guess it could be a bit lower but thats unlikely to happen.
 
Y'know I've had my 7D coming up on 6 months now and I never use it for stills. Well rarely. This struck me as i was looking up dpreview to find out what res sRaw, mRaw and Raw my 7D shoots at! Lol.

Back on topic - I think you are right, canon may well go for sWar or even mRaw off their current batch of sensors - seems practical solution to avoid line skipping.
 
J Davis said:
hmmm ... so if canon went sRAW with a CMOS dedicated to video – in order to purely rely on sRAW demosaicing instead of line skipping and pixel binning, what would the height and width of the sensor be in order to yield 1080p. Would it be 3840x2160?
Click to expand...
I think you're confused on some things here. The demosaic to yield proper 1080p would require about 3 megapixels. But that has nothing to do with raw, sRAW, line skipping or pixel binning.

The line skipping/pixel binning is solely a function of the sensor design, and trying to read out the sensor fast enough to deliver video frame rates. The sensor is not designed for video, it's designed for stills, and it cannot be read out fast enough. It doesn't matter how they encode it (AVCHD, h.264, raw, sRAW, redcode raw, jpeg, whatever) it still has to skip pixels and bin pixels to get the speed necessary.

If they were to deliver a sensor that was purely aimed at video, I would hope it would have no more than about 3 megapixels. That should be able to be read out at video speeds, could be stored as raw, and demosaic'd to 1080p, and would be pretty much ideal for all aspects of video work.

But it'd be horrible for stills, and that's why you'll not see something like that put into a DSLR form factor.
 
Barry_Green said:
The line skipping/pixel binning is solely a function of the sensor design, and trying to read out the sensor fast enough to deliver video frame rates.
Click to expand...

We all get this but even if this issue is solved today with better sensors, we still have a mismatch of full sensor resolution and something more manageable and closer to delivery resolution. The math just doesn't add up for full sensor RAW from a 18mp sensor even if it were fast enough. However, sRAW would, in theory, allow DSLR makers to keep their stills MP counts while also being able to output a high fidelity lower resolution motion format.

Barry_Green said:
But it'd be horrible for stills, and that's why you'll not see something like that put into a DSLR form factor.
Click to expand...

The 3K, 6MP APS-C DSLR would still have a decent market IMO. At that spec, you could offer D3s level low light performance and just enough pixels for clean 2K/1080p. You'd be surprised how many photographers would jump on such a camera.
 
Last edited:
Barry_Green said:
I think you're confused on some things here. The demosaic to yield proper 1080p would require about 3 megapixels. But that has nothing to do with raw, sRAW, line skipping or pixel binning.

The line skipping/pixel binning is solely a function of the sensor design, and trying to read out the sensor fast enough to deliver video frame rates. The sensor is not designed for video, it's designed for stills, and it cannot be read out fast enough. It doesn't matter how they encode it (AVCHD, h.264, raw, sRAW, redcode raw, jpeg, whatever) it still has to skip pixels and bin pixels to get the speed necessary.

If they were to deliver a sensor that was purely aimed at video, I would hope it would have no more than about 3 megapixels. That should be able to be read out at video speeds, could be stored as raw, and demosaic'd to 1080p, and would be pretty much ideal for all aspects of video work.

But it'd be horrible for stills, and that's why you'll not see something like that put into a DSLR form factor.
Click to expand...


You would have to read my earlier posts – we were talking about a hypothetical cmos dedicated to video not stills and a natural assumption of a video sensor is that it would be fast enough for 24 or 30 frames.

Shao suggested sRaw might be the way canon might go. Either way none of us are confused on the line skipping pixel binning issue - we all agree its bad.

I'm not sure how form factor came into it as it has nothing to do with how the sensor performs - I don't think any of us have doubts that a CMOS dedicated to video would not be in a DSLR body.
 
Last edited:
Actually the real crux is, to my knowledge, Canon has never developed a cmos fast enough for 24 or 30 frames without line skipping.

Can they, or are they willing to do that? Is it profitable for them?
 
J Davis said:
Actually the real crux is, to my knowledge, Canon has never developed a cmos fast enough for 24 or 30 frames without line skipping.

Can they, or are they willing to do that? Is it profitable for them?
Click to expand...

Yes. Because this would benefit the amount of still frames per second that their cameras can capture.
 
Yes. Because this would benefit the amount of still frames per second that their cameras can capture.
Click to expand...

But, there would likely have to be a tradeoff somewhere along the line... either in price or in other features/functionality.

later,
Jaosn
 
roxics said:
Yes. Because this would benefit the amount of still frames per second that their cameras can capture.
Click to expand...

I'm pretty confident in saying that any CMOS developed by canon that is dedicated to video would not also be aimed at the stills market. Not when their DSLR's are getting so inexpensive.
 
J Davis said:
I'm pretty confident in saying that any CMOS developed by canon that is dedicated to video would not also be aimed at the stills market. Not when their DSLR's are getting so inexpensive.
Click to expand...

That's kinda it to me... if there aren't megapixel or other tradeoffs then the price goes up. Then you are looking at a different market. One that RED aims to capitalize on with their upcoming DSMC line. One of high framerate, full sensor, raw recording. Not one of sell boat loads to photographers dabbling in video, and videographers seeking affordable DoF above all else.

One is more broad. one is more targeted and, ultimately, a more expensive product. If you are going to get involved in the latter, people won't settle near as much for compromises that they are more than willing to settle for with current dslr's.

Maybe they can find a middle ground in there somewhere, but the trend that is making these dslr's sell like hotcakes (other than their large target audience) is one of cheaper and cheaper prices. Once the price goes up... that target audience begins to get narrower... The narrower the potential customer base gets, the higher profit margin per unit they will need as well. The quality threshold will be higher among potential buyers. They start moving away from the DSLR market, and towards a specific one that RED is targeting, and hasn't yet proven to be one they (canon) need to go after to protect their profits.
 
J Davis said:
Actually the real crux is, to my knowledge, Canon has never developed a cmos fast enough for 24 or 30 frames without line skipping
Click to expand...

Has anyone? I mean, the Red One doesn't line skip but it still has a noticeable rolling shutter so it still isn't fast enough either.
 
maarek said:
Has anyone? I mean, the Red One doesn't line skip but it still has a noticeable rolling shutter so it still isn't fast enough either.
Click to expand...

That's an issue of CMOS sensors in general, at the moment. The new sensor (Mysterium-X) is much much better. The read-reset time has been cut down to 5 microseconds. Film is 4. Pretty close.
 
Jason Ramsey said:
That's kinda it to me... if there aren't megapixel or other tradeoffs then the price goes up. Then you are looking at a different market. One that RED aims to capitalize on with their upcoming DSMC line. One of high framerate, full sensor, raw recording. Not one of sell boat loads to photographers dabbling in video, and videographers seeking affordable DoF above all else.

One is more broad. one is more targeted and, ultimately, a more expensive product. If you are going to get involved in the latter, people won't settle near as much for compromises that they are more than willing to settle for with current dslr's.

Maybe they can find a middle ground in there somewhere, but the trend that is making these dslr's sell like hotcakes (other than their large target audience) is one of cheaper and cheaper prices. Once the price goes up... that target audience begins to get narrower... The narrower the potential customer base gets, the higher profit margin per unit they will need as well. The quality threshold will be higher among potential buyers. They start moving away from the DSLR market, and towards a specific one that RED is targeting, and hasn't yet proven to be one they (canon) need to go after to protect their profits.
Click to expand...

Werd!


maarek said:
Has anyone? I mean, the Red One doesn't line skip but it still has a noticeable rolling shutter so it still isn't fast enough either.
Click to expand...

Rolling shutter and line skipping are two completely different things and the conversation was about like skipping and that branched off the thread topic of canon raw.

Several companies other than RED have developed CMOS sensors fast enough for video. SI-2K being one of them,
http://en.wikipedia.org/wiki/Digital_cinematography_cameras#SI-2K

There is also a really small company in scandanavian europe who developed one but for the life of me I can't remember their name. (anyone please? would love to be reminded)

Rolling shutter in physical film cameras is caused by the spinning mirror shutter, in digital its the read-reset time. The new breed of red sensors are supposed to be 1ms off film.
Quoting Graeme Nattress - Arri being 4ms, red's mx sensor 5ms (edit:michael beat me to it)
 
I think we can count out it being an APS-C sensor camera, looking at Canons product life cycle there aren't any more APS-C cameras in the pipeline this year. A DSLR with significantly improved video performance is going to be a pro camera. The most likely candidate is the 1DsMKIV which is due soon. As it's a full frame camera and the demand is for S35 I suspect it might capture video from say a 1.3-1.6x crop of the sensor to reduce the data rate. And yeah it would have to be something like a video sRaw with some debayering to keep the bit-rate under control to say 100Mbps.

If Canon is like any of the tech co's I've worked for then all product development is driven by the marketing department not engineering so when predicting what they are going to do in product development you have to consider what their market demand is. We're seeing pro photographers being asked to shoot video for their clients and it's them that this DSLR is aimed at with us being perhaps up to 10% of that market. I doubt it will be a match for the S35 scarlet but it could be "good enough". The only thing is at around $8000 who's going to want a 1DsMKIV when you can get a S35 scarlet with the minimum necessary options for around $10000? If the 5DMKIII shoots 100Mbps sRaw true 1080p 4:2:2 for under $3000 it's a different story, that would certainly be good enough for me to buy 2 instead of 1 scarlet S35.

In the meantime canon should release the MKII firmware instead of sitting on it until Nikon brings out a MKII competitor.
 
squig said:
I think we can count out it being an APS-C sensor camera, looking at Canons product life cycle there aren't any more APS-C cameras in the pipeline this year. A DSLR with significantly improved video performance is going to be a pro camera. The most likely candidate is the 1DsMKIV which is due soon. As it's a full frame camera and the demand is for S35 I suspect it might capture video from say a 1.3-1.6x crop of the sensor to reduce the data rate. And yeah it would have to be something like a video sRaw with some debayering to keep the bit-rate under control to say 100Mbps.

If Canon is like any of the tech co's I've worked for then all product development is driven by the marketing department not engineering so when predicting what they are going to do in product development you have to consider what their market demand is. We're seeing pro photographers being asked to shoot video for their clients and it's them that this DSLR is aimed at with us being perhaps up to 10% of that market. I doubt it will be a match for the S35 scarlet but it could be "good enough". The only thing is at around $8000 who's going to want a 1DsMKIV when you can get a S35 scarlet with the minimum necessary options for around $10000? If the 5DMKIII shoots 100Mbps sRaw true 1080p 4:2:2 for under $3000 it's a different story, that would certainly be good enough for me to buy 2 instead of 1 scarlet S35.

In the meantime canon should release the MKII firmware instead of sitting on it until Nikon brings out a MKII competitor.
Click to expand...

I agree with all the above comments. Whilst I don't yet know much about video, I know how the DSLR market works. It is almost certain that the DSLR these rumours refer to is the 1Ds mkIV as it is overdue. This is likely to be very expensive (look at the introductory price of the 1Ds mkIII, then add on the extra due to the different exchange rates). I think the good news is that some of these features are likely to feature in a future 5D mkIII - but I don't expect this until late this year at the very earliest, and possibly later still. There is an outside chance of another APS-C DSLR because there is a gap with the xxD Canon DSLR range as the 7D was not a direct replacement to the 50D, which has no video.

I also agree about the marketing department driven bit. There is no doubt that Canon see the video as giving them a competitive advantage, and that there is an increasing call for video from commercial and pro stills photographers.

DSLRs will always be able to do something RED cannot due to the scale of economy. DSLRs sales are massive and very profitable, hence why Sony bought Minolta out etc. If DSLRs did not sell so profitably in such huge numbers neither the technological development or the prices would be what they are. Less than 2 1/2 years ago the Canon 1Ds mkII was the highest resolution DSLR on the market by a long way. The T2i/550D now exceeds it's resolution and many of its specs, but obviously not the build quality. Therefore the DSLR manufacturers can invest in R&D for sensor development in a way not possible for that used in products with much smaller sales. So if the DSLR manufacturers decide it is important to make sensors with better video read out they are more likely to achieve this than anyone else. They are also in a far better position to respond to other competitive products than are smaller companies. As far I can see the main difficulties with DSLRs as reqards video is that the trend for stills photography is towards much higher resolution, which obviously creates a bit of a problem with processing power when it comes to high speed down-scaling of the resolution for video.
 
J Davis said:
There is also a really small company in scandanavian europe who developed one but for the life of me I can't remember their name. (anyone please? would love to be reminded)
Click to expand...

I assume you mean the long delayed Ikonoskop A-Cam dII but they use a Kodak 2/3" CCD, not CMOS.
 
Most marketing departments try to have several different options in the development pipeline so they can respond rapidly to changing market conditions. Canon proved that they could do this when they responded to the D90.

And that is the wild card in all of this - after Nikon started the race they have done nothing new. Do you think they are sitting still? They have no pro division to protect. There were rumors that Nikon was working on RAW video for 2010 introduction. Whatever their response is it will bring a quick response from Canon.

Then there is the possibility that Panasonic will do something later this year with the GH1 upgrade. Canon knows enough to be preparing for that. It could be argued that the 550/Ti2 is the GH1 killer, and that Canon will defend that market niche with a quick upgrade. With the 550 Canon showed that it would rather own the mass market with advanced tech rather than reserving the advanced tech for high end. You can thank Panasonic for that. And in the future the moves of Panasonic into mass market high tech video cameras will drive Canon to move more and more tech to the low end.

Now assume that that happens, what will differentiate high and low end DSLR video? Could be audio controls, could be RAW, could be the dynamic range war just started by RED. In any case there has to be some reason to justify a pro model.

I think that we are entering a time when the pace of development well be accelerated due to quick responses to competition. In such an environment it is nearly impossible to predict what will flow. I think the days of the video capablities being roughly equal across product lines is pretty much over, but at the same time the video capabilities of the low end will be the prime battle ground.

And do not forget that the consumer camcorders are just beginning to see a tech battle, witness the new Panasonic camcorders with high bit rate AVCHD. We live in a small segment of the big world and our hopes and our wishes will be secondary to these market forces.
 
Last edited:
You must log in or register to reply here.
Back
Top