Epic and Scarlet release in 2010

Status
Not open for further replies.
Otis Grapsas said:
18stops is probably achieved with dual exposures. This method has major artifacts in motion but it can work when there is no motion in frame. Does anyone know if it's a single exposure?
Click to expand...

Jim stated whatever process they are using has "never been done before".

People have made many (i.e., dozens) of guesses as to how the tech is working, none of which have been confirmed as the technique being used.

Given that the RED team wants this to work for motion images as well as stills, and their commitment to really pushing the envelope in the past, I kind of doubt they would take an approach which would introduce severe temporal or spatial artifacts.

Which kind of leaves me with a big question mark over my head and really wanting to see some footage. But of course...how am I going to view something with 18+ stops? What monitor does that?
 
Multiple exposures is a negative. (http://reduser.net/forum/showpost.php?p=638278&postcount=134) I don't think dual exposure is a reasonable way to do HDR at all. Obviously RED have a trick or two up their sleeves.

There are several ways to map a 20 stop image onto a 10 stop display (as many modern LCDs are). From the still world, tone mapping seems to be the most popular. Aggressive tone mapping of 20+ stops images onto 8-bit video creates quite a surreal image that could find its way to cinema language. (Though I fear abuse/overuse) The idea of keeping a dark object in a dark room exposed without blowing out the exterior is very appealing to a cinematographer as well, regardless of how you output the final image. It could eliminate visual effects like sky replacement. We will have to see it work. We haven't had the privelege of such dynamic range in motion pictures before, so this is uncharted territory. Still photographers have used HDRI, of course, and some of the images are rather interesting. Exciting times.
 
Michael Olsen said:
Which kind of leaves me with a big question mark over my head and really wanting to see some footage. But of course...how am I going to view something with 18+ stops? What monitor does that?
Click to expand...

You can view HDR in any monitor, you just have to map it to the low dynamic range of a modern display. This is about 8 stops in consumer devices and 10stops in the best monitors. There is compression though, and many monitors that cannot really represent the last 2 stops (TN panels are an example) so you can't use the full range.

The real problem is how to map. I have spend years in building gamma curves, film models for negative and print, and all that, and I don't see room for more than 10 stops in a commercial delivery. 12 or 13 stops are an advantage but only in the shadow quality (better signal/noise in the shadows) not the actual dynamic range perception. If you want more than 10 stops, it will either look flat or very unnatural. 99% of the DOP out there will prefer the gamma curves that make the HDR camera look similar to the normal camera.

Think of dSLR cameras. These capture 10 to 15 stops. But all of them limit that to 8 or 9 visible stops with a gamma curve. Most photographers do not really care about the technical dynamic range, they just want good skin tones and good midrange contrast. That's what that gamma curve does. This is such a great requirement that none of the cameras that got a true dynamic range advantage on the sensor actually got ahead in the market (Fuji S5 etc).
 
To put it simply, in order to get a high dynamic range image that looks great, you need equally high dynamic range displays. Then you would have nature, reality, beauty. As long as you are mapping to a low dynamic range display you have to allow the image to map properly to the respective range of human vision (a small part of its full dynamic range) and that requires a curve with high midrange contrast. There is not much room on shadows and highlights with one of those.
 
Let us wait for it to come out and then we can analyze it for ourselves. I personally don't care how its achieve as long as the final product is to my liking. I'm a storyteller and not an engineer.

I've soured on Red in the last year but it wouldn't surprise me at all if Epic lives up to Red's predictions, which means good news for Scarlet. Only time will tell.
 
I see that Jim Jannard has mentioned that the Epic is using the same sensor with Red One MX. That's not 18 stops and its noise level looks normal for a sensor with that pixel size (quite good). The sensor cannot get 18 stops in a single exposure, so it has to be one of the conventional HDR techniques.

All it takes to demonstrate HDR is a single full resolution still with some fast motion in frame. I wish Red would do more of that. The Scarlet fixed lens video is 9 pixels reduced to 1 pixel (33% size or 1024x512 pixels for a 3K camera). A 3k still would be great.

It amazes me that Jim Jannard chooses to ignore one of the basic rules of marketing. Always promise less than you can deliver. Artifact-free 18stops from a conventional sensor we have already seen in practice does not look possible. People will not only expect that from Red, but from every single manufacturer!

EDIT: Remember 3K for $3K? Many people thought reasonably priced cameras was incredibly expensive after that. That was good for nobody, even Red.
 
Last edited:
Otis, even if it's under-delivered, let's say 14 usable stops, it's going to be a killer.

What other camera will offer this? And if Scarlet has it, wow.

Anyway, on the original topic Jim asked Michael Cioni to check out for himself and he seems impressed.

Let us wait and see.
 
Pietro Impagliazzo said:
Otis, even if it's under-delivered, let's say 14 usable stops, it's going to be a killer.

What other camera will offer this? And if Scarlet has it, wow.

Anyway, on the original topic Jim asked Michael Cioni to check out for himself and he seems impressed.

Let us wait and see.
Click to expand...

Red say MX is 13 stops. It would be hard to sell an one stop improvement as HDR, it has to be something far better that actually does a few extra stops.
 
Otis, are you saying that 18 stops is simple "impossible"? Or are you open to the possibility that RED have invented a technique we (and you) aren't aware of? It is not a multiple exposure thing, that has been confirmed.

Also, it is not 18 stops, but at least 18 stops. They simply ran out of stops on the DSC test chart. Michael Cioni testifies seeing "more stops captured than my own eye. No exaggeration". Sounds like 20+ stops to me.
 
I believe we should all wait and see, the proof will be in the images when the working man gets his hands on these cams.
 
Subhadip said:
Otis, are you saying that 18 stops is simple "impossible"? Or are you open to the possibility that RED have invented a technique we (and you) aren't aware of? It is not a multiple exposure thing, that has been confirmed.

Also, it is not 18 stops, but at least 18 stops. They simply ran out of stops on the DSC test chart. Michael Cioni testifies seeing "more stops captured than my own eye. No exaggeration". Sounds like 20+ stops to me.
Click to expand...

Yes, it's impossible with a conventional sensor and a single exposure. The current state of the art in digital sensors is many years away from a goal of 18 or 20 stops. These techniques are not invented like that. Even 1 stops takes great research and development. A method like this needs to be implemented in the sensor, and in this case you can predict the performance on the design stage. If I had a method to do 18stops video without artifacts I could make $billions within a couple of years on the defense industry alone, and I would shake the industry in all high end imaging applications, I wouldn't put it on a digital cinema camera:) We have seen the very sensor in Red One MX anyway, it's a conventional design and unlike the older Red One sensor, the MX performs up to what you would expect for the pixel size, which is great. It's as good as the sensors on dSLRs from the great manufacturers.

If it actually did real HDR without artifacts in fast motion, that mode would be the only mode the camera would have. It wouldn't be a special mode.

You can't run out of stops on a chart. You simply change the aperture by a few stops, keep everything else the same, reshoot, and then you know for sure what dynamic range the sensor/HDR method gets.
 
Pietro Impagliazzo said:
RED may say this, but tests show between 11 and 12 of readily usable stops.

http://provideocoalition.com/index.php/aadams/story/next_stop_the_last_stop_red_mx_latitude_tests/
Click to expand...

Yes, just like Red one was not 11.3 stops. The sensor range is only a spec and not achieved in the real world. It's different between the three pixel types (RGB) because of the required white balance digital/analog gains. Testing methodology also affects the result. In this case the output is also compressed and compression has a given PSNR which reduces the usefulness of the lower stops due to artifacts combined with the noise. On CCD cameras there are other issues like the electrical environment in which the camera is operating. The lower stops are not very useful if the electrical interference manages to show itself.
 
Otis Grapsas said:
Yes, it's impossible with a conventional sensor and a single exposure. The current state of the art in digital sensors is many years away from a goal of 18 or 20 stops. These techniques are not invented like that. Even 1 stops takes great research and development. A method like this needs to be implemented in the sensor, and in this case you can predict the performance on the design stage. If I had a method to do 18stops video without artifacts I could make $billions within a couple of years on the defense industry alone, and I would shake the industry in all high end imaging applications, I wouldn't put it on a digital cinema camera:) We have seen the very sensor in Red One MX anyway, it's a conventional design and unlike the older Red One sensor, the MX performs up to what you would expect for the pixel size, which is great. It's as good as the sensors on dSLRs from the great manufacturers.

If it actually did real HDR without artifacts in fast motion, that mode would be the only mode the camera would have. It wouldn't be a special mode.

You can't run out of stops on a chart. You simply change the aperture by a few stops, keep everything else the same, reshoot, and then you know for sure what dynamic range the sensor/HDR method gets.
Click to expand...

Thanks for your views, Otis, much appreciated. However, something's gotta give here. RED and a third party, a witness, Michael Cioni, vouch for this new technology. It takes a lot to impress someone of Michael's status, but to leave him "speechless" is something else. On the other hand, you say it is impossible. Jim has also clearly stated that it is not even close to a multiple exposure/bracketing technique that might leave artifacts. If I were to believe that is impossible, it would be accusing RED and a respected person like Michael Cioni of misinformation and lies. Nope, Occam's Razor suggests that in this case, it is the miracle that is most likely. With all due respect, Otis, I am going to have to conclude that this is a technology that has simply never occurred to you. And as an engineer and a competitor, it is natural that you are skeptical.

Having said all that, neither you, nor me, know what HDRx is all about. So we will have to wait and see how it turns out. However, personally, I am never, never going to dismiss it straight away without seeing it.
 
Subhadip said:
Otis, I am going to have to conclude that this is a technology that has simply never occurred to you. And as an engineer and a competitor, it is natural that you are skeptical.
Click to expand...

I'm not a competitor to Red, especially for EPIC. It's just that if this technology is artifact free (as good as a single exposure, taking a 11-12 stops sensor to 18stops) it is worth billions. I'm surprised people actually accept figures like 18stops so easily while major manufacturers are strugling for an one stop improvement in dynamic range. Technology like this is not something I would expect to be announced in this fashion on a website like reduser.

If I ever run across such a revolutionary method I would stop the release of any product just to figure out what I can do with it. The entire digital cinema market is not important compared to that of such an achievement in digital imaging. I wouldn't have one camera out of the door to avoid attempts at reverse engineering the method.
 
Jim on their HDRx

"I'll give you some clues on HDRx©...

1. What we are doing has never been done before.

2. We have different (variable) HDRx© modes.

3. The test Michael Cioni saw was 18 stops.

4. We have a lot of work to do to characterize and implement variable HDRx© in EPIC.

5. We need a new chart. :-)

More details when we have the faintest idea of what we are doing... :-)

Jim"

http://www.reduser.net/forum/showthread.php?t=48494&page=13

from Gavin Greenwalt in reply to people trying to figure out HDRx

"Jim has already stated that it's not how they're doing it.

The ways we know they aren't doing it:

1) Multi-sample ala motion blur and stacked. (Aka shoot @ max burst frame rate for 1/48th of a second and combine the frames.)

2) Progressive exposure. (Aka each pixel checks to see if it's 'about' to clip and stops exposing at its max saturation point. Along with "how fast" it exposed". Delta(t) exposure. )

3) Multi shutter conventional HDR.

How they are doing it... well that's the mystery. :D"

http://www.reduser.net/forum/showthread.php?t=48494&page=11

they are not using tone mapping and claim there are no trade offs at this point but we'll find out eventually (well most likely they wont tell us how they did it but we will find out what trade offs there are if any other than file size).
 
Jannard said:
Over the long haul, we have a hard time believing anyone can stay with us in resolution, dynamic range and "feel". Now add easy post solutions, size and price.

Jim
Click to expand...

I know I can sound a bit fanboy-ish, but RED delivered with R1 and delivered with M-X.

I mean c'mon! R1 M-X is comparable to an Alexa that costs more than $50k!

I bet all my chips on RED with no regret.

Some people may dislike the attitude, the business model and what not (god knows why, prudes), but I'm fine with all that as long as they give me the cameras no other company can.

Well, now I don't sound like a fanboy... I actually AM a fanboy!

:evil:
 
I could add another 20 methods to the list, but all have artifacts. The methods that do work involve either huge multisensor setups (limited for resolution by allignment), pixels with different sensitivities (fuji), and sensor implementations inherently capable of high dynamic range (multi slope linear, currently problematic).
 
Status
Not open for further replies.
Back
Top