No more FF35 Scarlet?

[toke lahti]

Not necessarily, sensor size has nothing to do with how Red designed the active pixel sites or circuitry in each sensor. Even though it too does not tell the whole story, pixel density is the first factor to start looking at. And even beyond that, within the pixel density, how much of that area is active and how much is supportive circuitry. If they used the same exact circuitry in all their sensors, they could be all exactly sensitive, but I'd argue that the larger the sensor, the more issues with voltage drop and on chip variation so you could lose some sensitivity across the entire array if you were stringent enough in your yield selection. Which is why it makes sense to me that the larger sensors are so much more expensive.

All comparison I have talked about is always assuming the same specs: pixel size, etc.
AFAIK, MX will have to types 3.2µm pixel size and 5.4µm.
I guess that these 2 are technically on same level so when the latter has 2.8 times larger pixels, it should also be 2.8 f-stops more sensitive.
They could also use exactly the same design for different sized sensors.
And adjust the pixel size or resolution...

There are of course always minor differences, but the big thing is that with Scarlet, we could use bigger sensor than what we have used for decades in eng, efp, event & documentary (2/3"), but S35 is just overkill and therefore waste of money.
But of course I'm trying to figure out the optimal Scarlet just because I'm poor.
Those who have money, just go and pick up FF Epic and shoot whatever with whatever lenses happily ever after.

 

[Rick Burnett]

One thing I have not seen described is the actual active pixel size per density and how that plays out on operation. Obviously the more area that is active the better, but not something that anyone really shares. And what I mean is this, if you have 100 pixels in 1000um of distance, then your pixels are 10um, however, that entire 10um is not all active pixel. A percentage of that area is for other circuitry to support the active circuit (row select, reset, noise, even some designs have multiple active circuits in the same pixel). So while 3.2um and 5.4um are different densities of pixels, it does not indicate the physical size only that you know it has to be smaller than some number lower than 3.2um and 5.4um.

And most definitely, you are correct they could use the same pixel size at different pitches, but I imagine, for the larger sensor, you'd make the pixels larger so you could get more light into them!

Honestly, even if I had more money to throw around, I still want a S35. I like that format the best in what I shoot. But, as many have thrown around here, the FF35 is not so much a video question but a stills questions and I can see that. I like shooting stills more on a 5D than a 7D, so I think it's valid product request!

 

[Gavin Greenwalt]

But aren't those sensors picking up red, green and blue independently? if so, how could you average them since they are not measuring the same thing?

Yes. But a Bayer pattern is *filtering* out Red Green and Blue separately on each photo-site anyway.

So the prism vs the filter should be similar light loss.

 

[David Rasberry]

But after all, I still think that there should be something between 2/3" and S35!
For example "widescreen 4/3" (18 x 9 mm) would be less than $5k and very nice with about 2kg s16 zoom lenses for motion and APS-C or 4/3" lenses for stills.

I agree having an intermediate size is a good idea, and a 4/3 based format is the logical choice. Wide angle super 16mm lenses won't cover it though. Super 16mm image width is 12.75mm vs 10.46 for 16mm and 9.6mm for HD 2/3".

 

[Gavin Greenwalt]

Luminance is only sampled in the green channel though and you have the optical losses through the prism to contend with too. There are more luminance samples with Scarlet's 2/3" sensor, not including the R and B filtered pixels, than on a 1080p camera. On Scarlet based on the 3.3micron pixel spec, out of a 4 pixel Bayer block, you have two luminance samples with a total area of about 21.8 sq. microns, vs a single luminance sample of about 25 sq. microns with a 3 chip 1080p camera. Little more than 10% difference in area. I think the underlying chip fab process and the inherent noise levels will have more impact than the minor difference in sample area. At low sample levels where read noise or quantization error predominates, the two samples of the Bayer chip may well be superior to the single sample 1080p chip.

Pretty sure both RED and Sony use more than just the green channel for luminance. Especially on a camera like the F900 which is a 4:4:4 camera.

 

[David Rasberry]

Pretty sure both RED and Sony use more than just the green channel for luminance. Especially on a camera like the F900 which is a 4:4:4 camera.

So does a Bayer sensor, but the sensitivities of each color channel on both types are mimics of the sensitivity of the human eye, and around 80-90% of your visual acuity and sensor luminance sensitivity are in the green channel. If you extracted a black and white image using only the green channel vs using all three channels, you would see almost no difference in the two images.

The reason a Bayer sensor works so well is that it is the closest mechanical analog to the natural distribution of luminance and color receptors in the human eye.

 

[Gavin Greenwalt]

A 3 Chip 2/3" has an entire sensor the size of a 2/3" Bayer that's ALL green. It doesn't have to sacrifice any luminance collecting space to red or blue.

So it still goes back to the 3 chip having a larger photon collection surface.

 

[David Rasberry]

A 3 Chip 2/3" has an entire sensor the size of a 2/3" Bayer that's ALL green. It doesn't have to sacrifice any luminance collecting space to red or blue.

So it still goes back to the 3 chip having a larger photon collection surface.

But fewer photons to collect. You lose about 1/3 of a stop thru the prism.

 

[Rick Burnett]

Yes. But a Bayer pattern is *filtering* out Red Green and Blue separately on each photo-site anyway.

So the prism vs the filter should be similar light loss.

No, it's not going to be similar loss because with the 3 sensor design you have both a prism AND a filter.

Here is a diagram I found online for a 3 CCD, but shows a generic configuration:

http://www.fluxdata.com/wp-content/uploads/fluxdata3ccdcamera.jpg

So not only are you splitting the beam to each sensor which means each sensor will receive less than a single sensor would receive, you are then filtering out the color as well.

But beyond that, like I said, what you gain in my opinion from the 3 sensor design is not more sensitivity, but more detail. Just because you have say 1920 pixels in a row for red compared to 960 doesn't mean those 1920 are more sensitive, it just means that you won't require interpolation in between those 960 to fill in the missing detail. You won't need to use a complicated algorithm to try and decide how to fill in that information.

I guess one could argue from a theoretical standpoint that the sensitivity of those missing pixels are what you gain and that would be true in my opinion as well, since those pixels are always going to be some combination of the 4 pixels near them in a bayer pattern which means they'd never be fully as sensitive in terms of the output that is calculated for them. Which of course goes back to detail. Again, however, on a whole, the sensor is no more sensitive.

 

[toke lahti]

Pretty sure both RED and Sony use more than just the green channel for luminance. Especially on a camera like the F900 which is a 4:4:4 camera.

When camera records RGB, it doesn't record separate luminance (Y).

 

[toke lahti]

I agree having an intermediate size is a good idea, and a 4/3 based format is the logical choice. Wide angle super 16mm lenses won't cover it though. Super 16mm image width is 12.75mm vs 10.46 for 16mm and 9.6mm for HD 2/3".

Yep,
and that's why you should crop/window the frame in the post or in the camera depending on lens, f-stop & focal length.
Having optically blurred corners is pretty nice effect to get certain feeling also.

Maybe gen2 Scarlet has more and economical choises.
Right now I wouldn't like to pay S35 sensor, since I need to use handheld zoom and there are none that light for S35.
So it's all about lenses.
Something like this converted for video work might work:
http://www.tamron.com/lenses/prod/18270_vc.asp
Although it would be very slow in the long end of the zoom.
But next in line is 35mm cine zoom, which weights over 5kg and costs 10x Scarlet.

Also DSMC idealogy is just not there, when it is much cheaper to use 5D & Scarlet 2/3" or S35, thanjust have one Scarlet FF.

So, RED, hear me out!
Make a 4/3" Scarlet! And conver this to video work:
http://www2.panasonic.com/consumer-.../model.L-RS014150.O_11002_7000000000000005702

There seems to be more and more interesting glass for 4/3":
http://noktor.com/products.php
DOF of 100mm/f2 FF lens!

 

[David Rasberry]

Right now I wouldn't like to pay S35 sensor, since I need to use handheld zoom and there are none that light for S35.
So it's all about lenses.

Also DSMC idealogy is just not there, when it is much cheaper to use 5D & Scarlet 2/3" or S35, thanjust have one Scarlet FF.

So, RED, hear me out!
Make a 4/3" Scarlet! And conver this to video work:

There seems to be more and more interesting glass for 4/3":

I am sticking with 2/3" for the same reasons. Compact broad range zooms are my weapon of choice. It is one reason 16mm film has maintained its popularity over the years. In fact I will be using 16mm film lenses on my Scarlet Cinema. 16mm used to be my favorite shooting format and I have longed for the day when I could afford to own the digital equivalent to my old Bolex. It is still not going to be cheap. I expect to wind up somewhere between $10k and $15k for a fully configured Scarlet Cinema including a decent selection of used lenses. I already have more than $3k in lenses now.

DSMC will come into its own. It will allow for styles of shooting that are not possible with any other type of camera. It is not just choosing to shoot either motion or stills. It is the ability to shoot motion frame rates for stills without compromising image quality.

4/3 is a very new format. I think it will come into its own in time. It should be the super 16mm of digital cinema.

 

[toke lahti]

4/3 is a very new format. I think it will come into its own in time. It should be the super 16mm of digital cinema.

Lets hope so.

All big camera manufacturers (except Canon) are bringing out 4/3" changeable lens EVIL systems. And since they are EVIL, all can shoot video and most will do fullHD. So after these companies push over 20 models out a year and over 10 million cameras sold, there will be lots of lenses available for them.
Companies like Tamron and Sigma will make lenses for them.
Question is, will any company make professional/prosumer lenses for motion work? Maybe Red could have this place in camera ecosystem?

Canon is also rumored to bring out new prosumer camera, because they already announced new video codec for that. Maybe that will have 4/3" sensor? Canon produces their own chips and lenses, so that gives them freedom to introduce new size class.
Canon is also rumored to bring our dslr with raw movie codec. That will take some quality advantage away from Red.

So in the long run Red will have very hard time to compete with these giants, when they have economics of scale in their side. If Canon announces 6D with APS-C and raw movie recording for $1499 next year, that would decrease interest for any Scarlet models quite a bit.
And if you want upgrade from S35 to FF, it's not $5000 more, it will be $1000 more to 5Dmk3.
But if Red has superior lenses to Scarlet, pros will keep on buying Scarlet.
So it's all about lenses.

Talking 'bout lenses brings me to this weird new small company again:
http://www.senkoadl.com/factory.php
From there: 1" 25mm f/0.95! 1" is very near of 16mm.
And: 2/3! 17mm f/0.95 for scarlet2/3"!
If just optical quality is good enough with these...

 

[David Rasberry]

Canon is also rumored to bring out new prosumer camera, because they already announced new video codec for that. Maybe that will have 4/3"

Based on size and form factor in the proto photos, the new prosumer video cam, at least the fixed lens version, is still a 1/3" 3 chip camera. It is stepping up to full 1920x1080p chips and solid state media with a new 4:2:2 pro level codec though. Should be a significant step forward over the current A1 series.

So in the long run Red will have very hard time to compete with these giants, when they have economics of scale in their side. If Canon announces 6D with APS-C and raw movie recording for $1499 next year, that would decrease interest for any Scarlet models quite a bit.

Nobody will be producing full motion raw recording cameras for $1500 for a few years yet. It takes too much processing power and processing power is cost. When they do it it won't fit in a DSLR form factor either, heat and power consumption alone will prevent that. So far Canon hasn't even demonstrated a sensor technology that can do full motion raw frame rates at all. Best they have done so far is 8fps. Even the rumored new prototypes are having moire and aliasing issues, so they are still doing a cheap shortcut down sample to some form of video codec.
When you understand that a $4000 8 core Mac pro doesn't have the processing power to even play 4k footage real time, much less record it, you can appreciate the amount of processing power that will have to be packed into such a camera. Its not cheap to do for anyone. 35mm motion raw for $1500 is wishful thinking. It is not a realistic expectation anytime soon, I don't care who is trying to do it or what their resources are.

 

[Gavin Greenwalt]

When you understand that a $4000 8 core Mac pro doesn't have the processing power to even play 4k footage real time, much less record it, you can appreciate the amount of processing power that will have to be packed into such a camera. Its not cheap to do for anyone. 35mm motion raw for $1500 is wishful thinking. It is not a realistic expectation anytime soon, I don't care who is trying to do it or what their resources are.

Yeah but an 8 core mac pro can barely play back 1080p h264 from a 5D. So ... that isn't really a good comparison.

It wasn't very long ago that 1080p processing was all but impossible but now I can buy a pocket cam which processes 1080p.

It's not a question of if Canon will release a camera which shoots RAW video. It's a question of when.

 

[Jim Retar]

Even the rumored new prototypes are having moire and aliasing issues, so they are still doing a cheap shortcut down sample to some form of video codec.

You seem to be making a rumor into a fact there. There is no real evidence they are making such a cam... therefore there is no evidence they are "doing a cheap shortcut"

 

[toke lahti]

Based on size and form factor in the proto photos, the new prosumer video cam, at least the fixed lens version, is still a 1/3" 3 chip camera. It is stepping up to full 1920x1080p chips and solid state media with a new 4:2:2 pro level codec though. Should be a significant step forward over the current A1 series.

Ok, I haven't seen the mockup.
Seems to be that canon, jvc and panny are still stuck on 1/3". How long?

Nobody will be producing full motion raw recording cameras for $1500 for a few years yet. It takes too much processing power and processing power is cost. When they do it it won't fit in a DSLR form factor either, heat and power consumption alone will prevent that.

You should know, that compressing that raw data to h.264 takes much more processing power than just saving the raw data.

 

[David Rasberry]

Ok, I haven't seen the mockup.
Seems to be that canon, jvc and panny are still stuck on 1/3". How long?

1/3" isn't going away. It is a major format with a very healthy market. Not all video shooters want large sensor cameras and especially they don't want to deal with shallow DOF or raw format work flows.

You should know, that compressing that raw data to h.264 takes much more processing power than just saving the raw data.

Not true, you can buy a 2MP HD webcam that compresses and encodes real time to H.264 with a $10 off the shelf chip. This is what Canon is doing in their DSLR's , reading 2 megapixels off of an 18 megapixel sensor and encoding that to AVCHD. They do not have sensors or data paths or processing power to do the A/D conversion for all 18 megapixels and pass it to either an internal or external storage medium faster than 8fps currently. To do so requires a dedicated A/D converter for each pixel and a data path and processor that can move that data faster than an 8 core Mac Pro can. Not happening. My Toshiba laptop can encode H.264 off of its internal web cam or play back 1080p H.264 at real time. The best it can do with a 4k 12 megapixel raw file from a red is 1-2 frames per second play back and 8-10 seconds per frame encode to another format. Data rates are exponentially higher. It is not easy and not cheap to do at full motion frame rates. For their sub 10k consumer markets I don't think Canon will even try at the current state of technology. They are at least two years away from having a sensor that can do only 24fps at the rate they are improving, and that's assuming they even think it is worth doing for their market. So far they don't.

 

[Stephen Gentle]

You should know, that compressing that raw data to h.264 takes much more processing power than just saving the raw data.

Well, that would be if you're just saving uncompressed RAW, but that's completly unmanageable in terms of storage speed and space required. RED's wavelet based REDCODE RAW compression would likely be far more taking than compressing H.264

 

[toke lahti]

1/3" isn't going away. It is a major format with a very healthy market. Not all video shooters want large sensor cameras and especially they don't want to deal with shallow DOF or raw format work flows.

I'll bet that 1/3" is going away from professional/better prosumer cameras. Sony has already moved to 1/2". Others will follow. Nobody wants 1/3" over 1/2" if they can choose. It's just economics. Everybody wants more sensivity and bigger sensor gives that. I haven't hears anybody ever complain about DOF of 1/2" beeing too shallow.

Not true, you can buy a 2MP HD webcam that compresses and encodes real time to H.264 with a $10 off the shelf chip. This is what Canon is doing in their DSLR's , reading 2 megapixels off of an 18 megapixel sensor and encoding that to AVCHD. They do not have sensors or data paths or processing power to do the A/D conversion for all 18 megapixels and pass it to either an internal or external storage medium faster than 8fps currently. To do so requires a dedicated A/D converter for each pixel and a data path and processor that can move that data faster than an 8 core Mac Pro can. Not happening. My Toshiba laptop can encode H.264 off of its internal web cam or play back 1080p H.264 at real time. The best it can do with a 4k 12 megapixel raw file from a red is 1-2 frames per second play back and 8-10 seconds per frame encode to another format. Data rates are exponentially higher. It is not easy and not cheap to do at full motion frame rates. For their sub 10k consumer markets I don't think Canon will even try at the current state of technology. They are at least two years away from having a sensor that can do only 24fps at the rate they are improving, and that's assuming they even think it is worth doing for their market. So far they don't.

It is true, what I said.
Bu of course it's not about if it's compressed or not, it's about how it's compressed.
And what computer's CPU/GPU can do doesn't matter, cameras have dedicated chips that are far more powerful with fraction of power used.

Do you really think that Canon couldn't do what Red does, if they wanted?
Canon knew everything that Red knew in the same day Canon got their hands to their first Red One. There might be some little secrets in the small details of the Red's encoding, but I guess that was also reverse engineered long time ago.
It's just they will do it when they can get good profits out of it and Red does it when it's technically possible without getting massive profits or losses.

Difference is that Scarlet (at least 2/3" & S35) will be mainly for motion and eg. 7D or 5Dmk2 mainly for stills. If Scarlet would have as much pixels than EOS's, it couldn't handle the raw compression. Same thing that if Canon would introduce, lets say, 10Mpx aps-c 8D they could also put raw compression in it. Quality wise it doesn't have to be so high end than Red's are, they could choose to do that cheaper.
They already have m-raw/s-raw, so they could use that for motion also. Or some kind of debayered rgb compression with more color depth than 8-bits. Not exactly raw, but much higher quality than h.264 from today's models.

I guess that 5Dmk2 & 7D have only 1080p recording only because when they were designing them, they didn't think that higher resolution or raw compression would be needed. Then they heard about Red and next they might start competing.

7D already can process 8 fps raw, which means 200MB/s from the sensor and it can also handle jpg compression of that, which means about 100Mpx of jpg compression per second.
Of course Scarlet can do 640Mpx/s with wavelet, which is propably 10x harder than jpg, but this shows that 7D is already quite close to what's needed. Ten times more processing power to 7D would be something like $1000 to chips and one quiet fan to keep it cool.
If Canon wants to compete with Scarlet they con easily have competitor on market next year.