Scarlet Resolutions

[David Mullen ASC]

exactly what i've been wondering, that and where the stills and sample clips from the 2/3" have been. I'd love to see the 2/3" with some mini primes sample work . . .

This is the only sample clip that was posted by Red of the new 2/3" sensor, with a RPP lens on it:

http://redgrabs.com/izzy/

 

[Steven Caesare]

But surely the low light capability of the 2/3" Mysterium-X isn't going to be as good as the s35 Mysterium-X sensor, is it?

The reason I say this is because it has a sizable difference in surface area. If I am mistaken, please let me know, for it would be GREAT news to know if shooting with natural light is a feasible....

For a given sensor size, more pixels are just as good (if not better), than larger pixels, when it comes to noise characteristics. (See Daniel Brown's excellents posts on this subject).

Given that that sensor design/chemistry is the same (ie- it's the MX design for both sensor sizes), you should see similar sensitivity and noise characteristics for them.

Of course, blowing up the smaller sensor image to the same size as the larger will magnify everything there, including noise.

But for DOWNREZZING to HD it should look fantastic. Of course the 5K Epic may even look more fantasticer when downrezzed, but we may be looking a tthe law of diminshing returns by that time consideirng the proce difference between the cams.

-sc

 

[David Mullen ASC]

But if there are no tradeoffs with cramming more and smaller photosites into the same size area, then why isn't MX sensor 9K if the 2/3" sensor is 3K? And why even bother with a 6K Monstro sensor, which could be 11K with those pixel dimensions?

 

[Jimmy Odom]

WHAT?:001_huh:

But if there are no tradeoffs with cramming more and smaller photosites into the same size area, then why isn't MX sensor 9K if the 2/3" sensor is 3K? And why even bother with a 6K Monstro sensor, which could be 11K with those pixel dimensions?

 

[David Mullen ASC]

If they are cramming 3K worth of pixels into a 9.5mm width, more or less, and the MX sensor is about 3X the width of the 2/3" sensor, then using the same sensor design, it could be 9K. And the Monstro FF35 sensor is about 3.7X wider than the 2/3" sensor, which is a potential 11K using the same sensor design.

But clearly there are some advantages to using larger photosites when you have the room for them.

 

[Steven Caesare]

But if there are no tradeoffs with cramming more and smaller photosites into the same size area, then why isn't MX sensor 9K if the 2/3" sensor is 3K? And why even bother with a 6K Monstro sensor, which could be 11K with those pixel dimensions?

I'd suspect that form factor, resolution, cost, processing and storage were just as significant factors in the design.

Smaller feature sizes may have different fabrication cost. A 9K sensor is 48megapixels. That's over 3 times a 5K sensor. That translates to more CPU horsepower, storage, bandwidth, etc... Additionally there may impact to overcranking ability, etc...

Also it's pretty clear RED has some very specific targets for resolution in each format size. They want at least 4K of resolvable detail for S35 format, hence a 5K sensor. They want at least 2K of resolvable detail for a 2/3" format, hence a 3K sensor (which whould actually resolve ~2400).

For all we know, the 2/3rds sensor may be more expensive per unit area to make becasue of the smaller feature sizes, but RED refuses to descend below a quality threshold.

That's not to say there may not be other sensor tradeoffs, but generally noise isn't one of them. While smaller photosites may be generally noisier per unit area, when there are more of them for a given size, then the overall net is about the same.

-sc

 

[David Mullen ASC]

That's all true, I'm just saying that the 2/3" sensor is not the same MX sensor just cut down to size from 30mm to around 9.5mm. And since it's not the same sensor, by logic, it's a different sensor. It may be part of the "MX line", but it's not the same sensor.

 

[David Rasberry]

But if there are no tradeoffs with cramming more and smaller photosites into the same size area, then why isn't MX sensor 9K if the 2/3" sensor is 3K? And why even bother with a 6K Monstro sensor, which could be 11K with those pixel dimensions?

Dan Hudgins indicated there is no significant noise/sensitivity difference for pixel sizes between about 3 and 7 microns, everthing else being equal. But higher pixel densities still create larger file sizes and require more processing power. This likely won't show any real advantage in actual resolution as the optical resolution will be limited more by the OLPF and the performance of the lenses than by pixel density. At f4 for instance, the minimum COC of an ideal lens would cover about four 3 micron pixels due to diffraction alone.

 

[David Mullen ASC]

So you are saying that the MX sensor is not 9K because there are no design advantages to using larger photosites, but simply because the EPIC processor, lenses, and OPLF can't handle or resolve 9K? (Yet it can handle 5K at higher frame rates?)

All I am saying is that the 2/3" sensor can't possibly be exactly the same sensor as the S35 MX sensor just cut down to size. That doesn't mean it can't perform well, I'm just saying that it's a different sensor with the same name. I don't see why this is so controversial or even debatable.

If photosite size makes absolutely no difference in how a sensor performs, then why doesn't the industry all just use the same size?

 

[Steven Caesare]

That's all true, I'm just saying that the 2/3" sensor is not the same MX sensor just cut down to size from 30mm to around 9.5mm. And since it's not the same sensor, by logic, it's a different sensor. It may be part of the "MX line", but it's not the same sensor.

Well, I think there may some difference in how the terms may be being used.

Apparently uses the names for a sensor to apply to a specific architecture/fab-process (and therefore likely major performance charateristics such as DR/noise). Jim has already chimed in and stated that the 3K Scarlet should have similar noise perfomance as the Red ONE M-X sensor.

For example notice that the both the Scarlet and Epic S35 are refered to as "Mysterium-X", despite apparently drasticly differing pixel sizes (if my math is right, the 2K M-X sensor will have almost 100,000 pixels per mm^2, whereas Epic 5KS35 will only have about 1/4 that at about 25,000 !!):

http://redrentalsmexico.com/home/wp-content/uploads/2008/11/epic_scarlet_brochure_large-1.jpg

I'd suspect that overall image characteristics such as noise perfomance, color response, light senstivity, etc.. will be about the same for all sensors in a family (Mysterium, M-X, Monsto, etc..), but that their particualar resolution "slice" will primarily affect detail reproduction.

-sc

 

[Steven Caesare]

So you are saying that the MX sensor is not 9K because there are no design advantages to using larger photosites, but simply because the EPIC processor, lenses, and OPLF can't handle or resolve 9K? (Yet it can handle 5K at higher frame rates?)

All I am saying is that the 2/3" sensor can't possibly be exactly the same sensor as the S35 MX sensor just cut down to size. That doesn't mean it can't perform well, I'm just saying that it's a different sensor with the same name. I don't see why this is so controversial or even debatable.

If photosite size makes absolutely no difference in how a sensor performs, then why doesn't the industry all just use the same size?

I was answering Joseph's question which was:

But surely the low light capability of the 2/3" Mysterium-X isn't going to be as good as the s35 Mysterium-X sensor, is it?

The reason I say this is because it has a sizable difference in surface area...

It is a different sized sensor with the same architecture and chemistry/fab-process.

In the context of Joseph's question, the 2/3" sensor will likely have very similar low-light characteristics.

I'm not saying there aren't lots of other factors, including cost, overall pipeline/storage requirements, vendor product stratification, etc....

And obviously detail/resolution factor significantly in how large you can blow up the print (and hence magnifiy any noise present).

But I suspect a 3K RGB shot from Scarlet and a 5K RGB shot from EPIC will both look about the same from sensitivity/DR/CR standpoint. One will just be able to capture more detail.

 

[David Mullen ASC]

I just think some people here are too prone to extremes of emotion -- from first worrying too much about the 2/3" sensor performance (I'm sure it will be just fine) to then thinking that there will be no difference in image quality between two different sensors of different sizes, just because they have the same name. I don't understand that sort of mentality, that's all, this all-or-nothing thinking. "It's going to suck horribly" or "it's going to be the greatest sensor of all time".... the possibility of some state in between doesn't occur to them.

 

[Steven Caesare]

I just think some people here are too prone to extremes of emotion -- from first worrying too much about the 2/3" sensor performance (I'm sure it will be just fine) to then thinking that there will be no difference in image quality between two different sensors of different sizes, just because they have the same name. I don't understand that sort of mentality, that's all, this all-or-nothing thinking. "It's going to suck horribly" or "it's going to be the greatest sensor of all time".... the possibility of some state in between doesn't occur to them.

I'm not sure where you are getting that people said they were the exact same thing.

I was specific about what I was speaking about:

For a given sensor size, more pixels are just as good (if not better), than larger pixels, when it comes to noise characteristics.

I then gave the caveat later in the very same post that there will obviously be a tradeoff if you blow it up to the same resolution.

Jim also stated that: "Mysterium-X refers to the pixel design/technology that leads to dramatically reduced noise levels. This pixel design/technology is used in both the 2/3" 3K RAW Mysterium-X sensor as well as the Mysterium-X S35 5K RAW sensor... "

-sc