Technology depresses me. I'll just wait and look at the pictures.

[Eric Rodemann]

I red this article by John Galt (old news now I'm sure) at Creative Cow, the 5 years in business special edition.

"The great perpetrators of the mythology of what I call “marketing pixels”
have been RED and Dalsa. Their cameras with Bayer-pattern sensors
are basically subsampled chroma cameras. In other words, they have half
the number of red and blue color pixels as they do luminance, or green pixels. You have two green photo sites for
every red and blue."

Yadda yadda, etc., etc., but it goes on to explain why Red 4K is NOT 4k.

Sort of exactly how Jim likes to say 1080p DSLRs are NOT really 1080p.

The more I learn about the details of the technology the more confusing it all its.

I'm just going to wait to see the footage posted from the Scarlet, and if it looks good, that's that. Will it REALLY be 3k? I don't have any idea, honestly all the subsampling and bayer pattern stuff is over my head.

 

[Terry VerHaar]

My first impression is that there is some tortured logic in his discounting 4K and "dissing" RED but failing to come back around to discuss RED when he talks about what a camera resolves. I can't quite put my finger on it but it seems to to miss the relative arguments of what 4K is (and resolves) as compared with other digital cameras. I am still studying it but I'd love for Graeme or someone to weigh in about the article.

Also, it is worth noting that this gentlemen works for Panavision. :001_unsure:

 

[Eric Rodemann]

My first impression is that there is some tortured logic in his discounting 4K and "dissing" RED but failing to come back around to discuss RED when he talks about what a camera resolves. I can't quite put my finger on it but it seems to to miss the relative arguments of what 4K is (and resolves) as compared with other digital cameras. I am still studying it but I'd love for Graeme or someone to weigh in about the article.

Also, it is worth noting that this gentlemen works for Panavision. :001_unsure:

Well, my point, probably not really made in the original post, is the important thing is the subjective IQ.

In other words, what the screen looks like when I watch it. In the end, bayer patterns and 3K (is it REALLY 3k becaues of the bayer patter and low pass filters?) don't matter as long as it looks pretty on the big screen.

That's what I'll ultimately judge the Scarlet 2/3 by, what the footage looks like.

 

[Sam Eilertsen]

Graeme HAS weighed in on this, several times I think, if you search you will find it.

RED has consistently said that there Bayer sensors resolve about 80% of the output resolution. So the REDONE resolves 3.2K, the Epic will resolve 4K, and the Scarlet will probably resolve about 2.4K.

Whatever Galt says, these numbers can be confirmed by shooting a test chart.

 

[Jim Retar]

The consensus seems to be that the 3K Scarlet will resolve around 2.4k... so if you take the absolute best 1080p camera on the market... the scarlet will have more resolution.

Keep in mind that lenses will play a part in how much the camera can resolve.

Likewise, a 5K Epic will resolve at least 4k footage.

 

[Emyr R. E. Pugh]

http://reduser.net/forum/showthread.php?t=51550

 

[Julio Quintana]

In the end, bayer patterns and 3K (is it REALLY 3k becaues of the bayer patter and low pass filters?) don't matter as long as it looks pretty on the big screen.

The problem with lower resolutions is that they might look "pretty" most of the time, and then you film a brick wall, mini blinds, a chain link fence, or a tweed jacket, and your image falls apart. I get what your saying, but just because true resolution is tricky to determine doesn't mean it isn't important.

 

[Terry VerHaar]

Graeme HAS weighed in on this, several times I think, if you search you will find it.

RED has consistently said that there Bayer sensors resolve about 80% of the output resolution. So the REDONE resolves 3.2K, the Epic will resolve 4K, and the Scarlet will probably resolve about 2.4K.

Whatever Galt says, these numbers can be confirmed by shooting a test chart.

You are right that he has discussed what the sensors resolve, etc. I am very aware of that. But this article does seem to sort or "end run" that discussion with, as I put it, "tortured logic." It seems a little like an apples to oranges comparison - talking about how 4K resolution isn't 4K but that seems to be a different claim than whether 4K resolution actually resolves 4K. Of course, it doesn't but neither does the Genesis or any other digital camera. I am totally clear on the issue of what RED resolves, just not clear on how the Bayer pattern discussion doesn't seem quite right. :undecided:

 

[Terry VerHaar]

http://reduser.net/forum/showthread.php?t=51550

Thanks for the thread. Looks like a good discussion. Sorry for proliferating a rehash here. :embarassed:

 

[Graeme Nattress]

Best read through the earlier threads. My main comment is that the article would have been vastly superior if John had spent as much energy investigating the visually obvious issues with the RGB stripe sensor pattern the F35 and Panavision Genesis use as he did on Bayer pattern based sensor patterns.

In the end, the proof is in the puddling and the RED pudding tastes really nice.

Graeme

 

[A. Clint Litton]

Best read through the earlier threads. My main comment is that the article would have been vastly superior if John had spent as much energy investigating the visually obvious issues with the RGB stripe sensor pattern the F35 and Panavision Genesis use as he did on Bayer pattern based sensor patterns.

In the end, the proof is in the puddling and the RED pudding tastes really nice.

Graeme

That is some tasty pudding.

 

[Paramon Ivanovich]

.
I'm just going to wait to see the footage posted from the Scarlet, and if it looks good, that's that. Will it REALLY be 3k? I don't have any idea, honestly all the subsampling and bayer pattern stuff is over my head.

I don't even care about bayer pattern, and I don't even need to see the footage from Scarlet, enough for me I see and work with R1 footage;
I'm buying one (at least) as soon as it'll be available in the Red store;
I know there are no miracles and you always get what you pay for, but if you can name me another 2/3" camera that shoots 120fps raw at 3K(well, whatever it is, to me it looks stunning) under $10K, I might change my mind and become more cautious about my choice.
and i am not being sarcastic or anything, I really mean it

 

[Terry VerHaar]

I don't even care about bayer pattern, and I don't even need to see the footage from Scarlet, enough for me I see and work with R1 footage;
I'm buying one (at least) as soon as it'll be available in the Red store;
I know there are no miracles and you always get what you pay for, but if you can name me another 2/3" camera that shoots 120fps raw at 3K(well, whatever it is, to me it looks stunning) under $10K, I might change my mind and become more cautious about my choice.
and i am not being sarcastic or anything, I really mean it

I am totally there with you. I think the RED ONE, and now the Epic footage, has pretty much proven the case. I think Jannard and Company have the highest standards and way more knowledge than I do. If I could buy it , sight unseen, this afternoon, I would! :thumbup:

 

[Pietro Impagliazzo]

I red this article by John Galt (old news now I'm sure) at Creative Cow, the 5 years in business special edition.

"The great perpetrators of the mythology of what I call “marketing pixels”
have been RED and Dalsa. Their cameras with Bayer-pattern sensors
are basically subsampled chroma cameras. In other words, they have half
the number of red and blue color pixels as they do luminance, or green pixels. You have two green photo sites for
every red and blue."

Yadda yadda, etc., etc., but it goes on to explain why Red 4K is NOT 4k.

Sort of exactly how Jim likes to say 1080p DSLRs are NOT really 1080p.

The more I learn about the details of the technology the more confusing it all its.

I'm just going to wait to see the footage posted from the Scarlet, and if it looks good, that's that. Will it REALLY be 3k? I don't have any idea, honestly all the subsampling and bayer pattern stuff is over my head.

Eric, well, it gets confusing... Just perseverate.

And this article is really old. Scarlet actual resolution will be something like 2.4K.

We have done these tests several times and posted them somewhere.

What I can remember offhand is:

4K scan of fast film- 2.8-2.9K
4K scan of slow film- 3.1-3.2K
RED 4K- 3.2K
RED 4.5K- 3.6K
RED 5K- 4K
Vistavision- 4K

F35 1080P- 1.7K (best 1080P we ever tested)
Alexa 2K- 1.6K (due primarily to low con filter on OLPF).
5D MKII 1080P- marginal 1.2K

Remember to use both height and width for calculations... 4K is 4 times the resolution of 2K at the same aspect ratio.

We are now using 11K zone plates for resolution testing.

Jim

Scarlet will have twice the measured resolution of a 5D. Not to mention REDCODE, 120fps, M-X DR, HDRx, etc.

 

[Lee Saxon]

Whatever Red One "really" resolves is still more than what 1080p cameras costing 12x as much resolve so....what are all these peoples' points, exactly? If you believe Red lies to you, fine, but Sony steals from you, so which is worse?

 

[Jeff Kilgroe]

Other things to keep it all in perspective...

Most all current SLR cameras use CMOS sensors with Bayer-pattern RGB filtration. If you're going to attack what RED, Dalsa, etc.. say about resolution based on Bayer pattern sensors, then you are also attacking Canon, Nikon, Panasonic, Sony, Fuji, Olympus, etc.. and all of their CMOS products. They may as well just throw their "megapixel" numbers for still images right out the window becuase they must be lying just like RED... Right? But that's really not how it works. A Bayer-pattern sensor still registers luma information for every single pixel on the sensor. The resolution or detail is all there for every pixel. But red, green and blue filtration is done in a checkerboard pattern, so chroma resolution is lower and must be interpolated or sampled between all the different RGB values recorded. It's actually a very effective process and is currently the most commonly used high-resolution sensor process for stills and video. In a good Bayer-sensor implementation, a resolution chart will show that the sensor can resolve about 80% to 88% of its actual recorded pixel resolution. Most of the loss of resolving power is due to the optical path and not the sensor itself. The primary effector in the optical path is the Optical Low Pass Filter (OLPF), which is a lensing system that sits in front of the sensor and helps to blend light between pixel edges on the sensor. It has a softening effect and helps to eliminate harsh aliasing and ringing. It is a trade-off. The OLPF could be removed and measured resolution would increase, but so would harsh aliasing.

DSLR video... Look at the 7D or 5D2, GH1 or GH2, etc.. They are all CMOS-based, Bayer pattern sensors. They are subject to the same pitfalls that RED's sensors are. But wait, they throw pixels away to create their HD video. They don't simply record the center window area because they try to get most of the usable sensor field of view into the recorded frame. They bin or combine pixels and/or skip lines altogether to down-size the image area into a raster size that the internal components can deal with. This line-skipping methodology introduces further resolution loss, increased aliasing and moire.

As pointed out already, that article is rather old. And the author did a rather poor job of presenting his case. He failed to mention all those other CMOS cameras in the world and failed to address shortcomings of the alternative technologies in use out there. 3-chip systems and RGB stripe systems have some glaringly serious faults of their own. The RGB stripe is a misfit cousin of the Bayer-RGB system, but instead of inter-woven checkerboard RGB patterns, it uses vertical solid stripes of alternating red, green and blue pixels. 3-chip cameras suffer light and resolution loss via their prism system to divert the proper wavelengths of light onto the appropriate sensor assigned to R, G or B reception.

 

[Tom.Wong]

I was just wondering. r3d is 12 bit, soon to be 16 bit from epic. the article mentioned in these threads argue that cameras like f35 and genesis are 10 bits per CHANNEL, for the RGB, thus making it 30 bits a pixel. For color accuracy this is pretty much the best, and avoiding calculation errors.

is r3d 12 bit, or 12 bit per channel? if it's 12 bits, doesn't that make it 12 bit for the entire pixel? meaning 4 bits a channel? I'm just trying to clearify some things, not pointing or prodding at anything. I grade r3d, it's amazing. but it would be nice to know these numbers as fact.

 

[Graeme Nattress]

R1 is 12bit raw, which means each pixel on the sensor is read through a 12bit A-to-D. We record that full bit depth in the R3D file. When we decode the raw data to make an RGB image we decode it to 16bit per channel.

Think on the F35 - it's a colour filter array sensor just like the RED, but in a stripe rather than Bayer pattern. I don't know what their A-To-D bit depth is off hand, but as you say, it gets recorded at 10bits per pixel as an RGB (rather than as a RAW) image.

RED is NOT 4bit per channel :-( Or if it is, our algorithms are a hell of a lot better than I thought they are! A 4bpc linear light image would have a max DR of 4 stops, which is significantly less than we achieve.

Graeme

 

[Eric Rodemann]

Eric, well, it gets confusing... Just perseverate.

And this article is really old. Scarlet actual resolution will be something like 2.4K.



Scarlet will have twice the measured resolution of a 5D. Not to mention REDCODE, 120fps, M-X DR, HDRx™™™, etc.

That was really useful information that I can process. 5D is 1.2K, Scarlet will be 2.4K, which is double. That's pretty easy to understand. I like it.

Oh, and I do appreciate the enthusiasm for the Scarlet, but I think I will look at some footage before I purchase one.
As for the article being old, please blame it on Creative Cow and their 5 year anniversary edition for including the old article. I'm not scouring the net to dredge up old stuff, I read it in the brand spanking new edition of Creative Cow, which does have some interesting articles.

The article brings up another point, which is 24p. We get 24p from film, and that was because of sound which wouldn't work correctly under 24 fps.

Then generations grow up watching 24 fps to the point we think that's what we gotta see in a movie, so we copy that in a digital format. We can shoot 24p, hurray!

Will audiences ever appreciate anything besides 24p, or have we now ingrained that in everyone to the point we're stuck with it?

 

[Michael Olsen]

Aside from the fact that RED ONE footage can look great, M-X footage, can look really great, and so far Epic-X footage seems to be considered spectacularly good...

The numbers game is ultimately fruitless. If anyone is lying, everyone is lying. If anyone is using "marketing pixels", everyone is using "marketing pixels".

What it comes down to is that no camera resolves 1:1 with each photosite. Which makes sense - nothing is 100% efficient. 4K bayer might resolve 3.5K. 5K bayer might resolve 4K. 6.5K striped RGB apparently resolves a bit less than 2K. To me, it has become a semantic argument, to which the only solution are test charts. Instead of specifying camera resolution as how many pixels/photosites they have on the sensor or how many they output in a given format, the only real meaningful data in determining how well they actually perform is...well, how well the actually perform. Best measured on charts under controlled conditions.

Other things to keep it all in perspective...

Most all current SLR cameras use CMOS sensors with Bayer-pattern RGB filtration.

Too, true. And they, like the RED(s), can produce some stunning images.