F55: Dragon Sensor 20 stops

Considering Jim's prior pronouncements on prior RED cameras there may be a significant difference between claimed DR and actual usable DR. The Epic and Scarlet are just not good low light performers. When you add more photo sights the physics pushes towards less light gathering ability and HDR just causes ghosting with motion.

To me that just opens a big healthy dose of 'wait and see', especially as he apparently sees the need to push back against the F55.

Doesn't this make all those recently purchased Epic's obsolete?

Would make a an awesome STILL camera, however
 
...The bigger the body, the bigger the heat sink....You're eye just can't even seem that to begin with.
This is why I call them computer geek DP's

Interesting. Actually, heatsinking (for heat transfer/management) is related to surface area so it's not about a bigger box but a box with more surface area. You get that with a finned sink + fan
Human eye does indeed 'scan' but it easily accommodates a 20 stop DR, The more we push the technology, the more we learn about the human visual system in it's totality.That and 'Leveling'.
RED Digital is rather technology driven and their value added has always been 'pushing the envelope' . I jokingly refer to RED as Sony/Arri secret R&D enterprise.
 
The human eye uses all sorts of tricks to generate a perceived 20 stops of DR, but take those away and it's actually more like 7 stops at any instance in time. What this means is that although we can't push TVs to produce more than 7 stops of DR (otherwise our eyes will struggle to take in the highlight and shadow detail at once), we can still display an image made to represent more stops, which will look natural to us.

Would make a an awesome STILL camera, however

The Epic? Poor color performance, low-level softness, no advanced control of shutter and flash sync, native ISO800, no autofocus, no automatic control of aperture and whatnot, all of this against even a modest DSLR that costs 10x less. You can make REDs workable for photography, but it's a tremendous waste if that's all you intend to do. Cine cameras have gone a long way in DR compared to even the best still cameras, however; we have a bunch of s35 cameras that get up to 14 stops DR now, whereas the only still camera that does that is the D800.

P.S. In case someone doesn't believe the elephant photo, here is the original Raw: https://www.dropbox.com/s/oz3s9wxg8ajsw1k/_MG_0069.CR2
 
Last edited:
Another problem is the software. The resolving power of a camera is actually quite tied to how the data is processed. Moving over to photography, I've recently upgraded my favorite Raw converter, Capture One, to ver. 7 and now it quite literally resolves details a pixel in width from my 5D2 stills.
.

Looking at your elephant photo, the difference is quite amazing. Have you compared Capture One V7 to the latest version of Photoshop? Just curious if one can resolve more detail or dynamic range than the other?
 
The elephant picture is an 8 bit JPEG representation of the RAW image capture. There is no detail to recover in the overexposed picture as shown. This same issue came up when I photographed under controlled conditions Sekonic test targets to calibrate their light meter. It turns the overexposed images into JPEGS and uses that to interpret the dynamic range of the camera, which it interprets as only 7-8 stops. But if you take the RAW version of those pictures where all the lighter squares comes out totally white, it is possible in Aperture (or Lightroom) to bring back many more stops of what is seemingly blown out over- exposure. The RAW data file has more information than can be shown in the initial represenation of that image. This is the whole benefit of RAW capture, the hidden information (at least that is my understanding). Typically in still cameras, the RAW data is 12-14 bits as opposed to JPEGS and most video players which are only 8 bit.
 
I think I will throw away my 18% card and just carry an elephant around. That is a brilliantly simple observation Liam.....hilarious.
Oh yeah, I accidentally switched my camera mode dial to where I had different settings, the 5D is notorious for having an easy to nudge mode dial.

Looking at your elephant photo, the difference is quite amazing. Have you compared Capture One V7 to the latest version of Photoshop? Just curious if one can resolve more detail or dynamic range than the other?
I compared it to Camera Raw 7.2 in both Photoshop and Lightroom 4, which gave me less stellar results, the image was more washed out and lost red chroma, making everything yellow.

The elephant picture is an 8 bit JPEG representation of the RAW image capture.
This sounds plausible... But I compared v6 and v7 with the Raw set to linear gamma, and v7 still produces a stop more highlights than v6. You can push highlight recovery even higher, whereas in earlier versions, the recovery sliders were basically useless.

What I'm saying is, even if there is some sort of hidden data, the software still has a very big impact on how the actual full range of data is handled. If you have a Canikon or a Phase, C1 is currently the converter to beat in absolute image quality.
 
Last edited:
Ha ha. No the cover was not shot on RED at all. This issue had the first RED ad in the middle somewhere. I have the issue somewhere. Jarred posted this to show what the cover looked like.


http://www.reduser.net/forum/showthread.php?75055-VOGUE-RED


The cover was photographed by Mert Alas and Marcus Piggott
http://www.artpartner.com/artists/image/mert-alas-marcus-piggott/

This behind the scenes video was shot around the same time as the J Lo video which shows them using a medium format camera.
http://www.youtube.com/watch?v=7AgzbSGn3yc


http://media.vogue.com/files/filech...57358248.jpg_article_gallery_slideshow_v2.jpg
As I recall. This is allegedly a still grab from Red footage. I can;t remember if it was Scarlet or Epic. I studied the cover under a loupe. Good DR but crap resolution.
Still, not bad.
 
Last edited:
The elephant picture is an 8 bit JPEG representation of the RAW image capture. There is no detail to recover in the overexposed picture as shown. This same issue came up when I photographed under controlled conditions Sekonic test targets to calibrate their light meter. It turns the overexposed images into JPEGS and uses that to interpret the dynamic range of the camera, which it interprets as only 7-8 stops. But if you take the RAW version of those pictures where all the lighter squares comes out totally white, it is possible in Aperture (or Lightroom) to bring back many more stops of what is seemingly blown out over- exposure. The RAW data file has more information than can be shown in the initial represenation of that image. This is the whole benefit of RAW capture, the hidden information (at least that is my understanding). Typically in still cameras, the RAW data is 12-14 bits as opposed to JPEGS and most video players which are only 8 bit.

"RAW" gets a lot of credibility for its ability to contain this extra information but it has much more to do with the fact that when shooting RAW photos with DSLR's or RAW video with camera like RED, the signals are stored using a linear gamma. i.e. data directly off the sensor which is linear based. Later in post processing once gamma correction is applied to match how we perceive light (more logarithmically) the data which represents the entire tonal range becomes skewed and biased towards the bright tones. So "more information" in this case is really only available in the bright tones and not so much the middle or dark tones.

Probably everyone has heard someone say when using a histogram keep the hill to the right. With linear based encoding there is good reason for that since you will use more of the data bits available to represent the overall image. Example:

Original scene:
smooth_gradient.jpg


What the sensor sees and how a linear signal looks using 5-bit representation (32-steps):
gradient_5bit.jpg


Gamma correction applied to same linear signal so we can view it:
gamma_gradient.jpg

(note the increased levels in the bright tones)

Not all RAWs are equal either. ARRIRAW shoots RAW but uses LOG to encode the signal. BMC is also doing something similar to LOG, and the new F5/F55 is not very clear to me yet but judging from Ben Allen's latest article which indicated log is being applied in post processing, it leads me to believe Sony RAW is also perhaps using linear gamma.

So what is RAW really? A format that contains unaltered Bayer sensor information - but how it is encoded (ie. log or linear gamma) makes all the difference.
 
All those covers on that link are indeed shot on the RED. That said, I do remember when Jarred teased about the Vogue issue people did assume that it was the cover shot. It was cleared up that it was just the ad inside. In any case resolution isn't everything. I love some of the stuff Greg Williams gets from his RED.
http://www.emptykingdom.com/main/featured/greg-williams-red-filmstills/


Well. when the issue was released, it was promoted as a still grab on the RED forums. So someone had to have lied. Anyway, the resolution looked like 3 year old 8MP digicam which is about consistent with a still grab.
Anyway, here's a link to RED's Citations for Still Grabs.
 
All those covers on that link are indeed shot on the RED. That said, I do remember when Jarred teased about the Vogue issue people did assume that it was the cover shot. It was cleared up that it was just the ad inside. In any case resolution isn't everything. I love some of the stuff Greg Williams gets from his RED.
http://www.emptykingdom.com/main/featured/greg-williams-red-filmstills/

The content is beautiful but from an image quality POV... You could dupe that look with a grungy Mamiyaflex and Tri-X Pan ;)... for about $100
 
The Red One is actually a better designed camera in my opinion - it doesn't overheat as easily and has some good weight and ergonomics to it. The menu system is very disorganized but its built a little more solid. Oh well. I'm hoping for the F55 to be amazing when you take off the Sony Matrix and really go for it on your own.
 
Back
Top