Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Off the chart...

Mohammed El Sharqawy said:
18 stops is exceeding what your eye can handle.. not what a display device can handle..
tone mapping is the thing everyone going to do to make use of these HDR images..
Click to expand...

Well, we don't really want a screen that shows the light with the intensity of the sun in the one end of the spectre, what we want is to be able to record as much as possible of that, to reproduce it through a non-painfull DR to view in theatres and TVs.

At least, that's what I want...
 
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...

Meryem Ersoz said:
The real question is: how am I going to process all the juicy data that this camera can deliver??

Engine room, more steam!
Click to expand...

If by process you mean in post, I take solace in Jim's posting above about "easy post solutions."

If you mean how are you gonna handle the amount of data this big ol' shark is gonna take in a bite, I think you're just going to need a bigger boat.
 
Pawel Achtel said:
If you do two exposures instead of one, you will gain at most 1 stop only and introduce motion artefacts: Hardly HDR.
Click to expand...

If you do an extra exposure at 1/1024 of the main exposure time, you gain 19 bits in the highlights. If you then use the clipped pixels of he main exposure as a mask to add the underexposed image, motion artifacts would probably not be a problem.
 
Stephen Smith said:
Is the motion hdri a Epic only thing?
Click to expand...

Eren Ozkural said:
No Stephen, it's been advertised for both Scarlet and Epic.
Click to expand...


Maybe I've missed a post or two, but earlier in the thread there was this from Graeme:


Victor Papalucas said:
Please tell us is it going to work just fine in motion ? :P
Click to expand...

Graeme Nattress said:
All my tests so far have been in motion.

Graeme
Click to expand...

Subhadip Sen said:
Any hints on what kind of tradeoffs we can expect?
Click to expand...

Graeme Nattress said:
Only tradeoff I can think of is that you'll have to buy an Epic.

Graeme
Click to expand...
 
Congratulations Red Team, this is stupendous news!

I've been mulling the possible technique in my head and just for fun, here's my theory given the things Jim et al have said it ISN'T using to get HDR:

The Mysterium-X sensor's read/reset rate is 5ms. Now everyone seems to think of time-slicing a single pixel read into time sections of varying length to give the HDR, but Red have said this isn't the technique (plus it would give temporal displacement, even if only slight).

Now we know that the M-X sensor is read top-to-bottom CMOS style. But if you think of it at the single line/pixel level, there has to be a pulse of electrical energy flowing through each pixel from top to bottom as it reads. Everyone thinks of digital sensors as having digital waveforms, but at the sensor itself (before the analog-to-digital converter part) it's still an analog waveform pulse of electrical energy.

Here's the clever part - that single pulse of energy (which returns a value at peak energy normally to be converted to a digital value) has a ramp-up and a ramp-down time, even if very small. Now if Red's new hardware is fast enough to accurately read and process multiple points on the waveform of that single read pulse, the lower voltage at the beginning of the waveform would return a different value than what it would at peak voltage and again at the tail end of the curve.

The great thing if this is possible, is that your exposure highlights can be based on the bottom of the read curve of a single pulse and the additional analog voltage can be used to effectively read further into the shadows without adding any noise compared to non-HDR mode. A second benefit is that it's still a single read pulse. No temporal displacement beyond the sensor's inherent 5ms skew on the entire sensor read per frame.

Essentially if this is possible, the limit of the sensor becomes the limit of pixel voltage resolution (i.e. how accurately one can measure between the points on the curve of a single pulse). If the delta between maximum and minimum voltage read is sufficient you could have an astonishing range of stops, limited only by the noise floor of the sensor.

Thoughts/corrections (I'm not a sensor engineer!)?

Cheers from Tokyo,

Paul
 
Maybe they've just taken interlaced down to a new (fractal) level.
 
Gunleik Groven said:
Well, we don't really want a screen that shows the light with the intensity of the sun in the one end of the spectre, what we want is to be able to record as much as possible of that, to reproduce it through a non-painfull DR to view in theatres and TVs.

At least, that's what I want...
Click to expand...

I'd love a device able to show the whole DR. It would be like a window.
 
Yeah, but they might have to hand out sunscreen at desert island-based features. "I think that film gave me a tan."

Stephen
 
Okay, here's my take... and I think I'm right or Arri should use my method if Red isn't, LOL! :smilewinkgrin:.

I speculate that the Mysterium sensor allows for selective pixel reads based on how full of photons each pixel is. If all pixels were read faster than 1/48th, then you'd starve the pixels that are nearly empty and add a proportionally large amount of noise to them w/ each read.

If on the other hand, you could selectively read pixels that are close to overloading, then you'd empty them and allow them to capture more light. And the added noise of another read would be negligible.

So how to do this? Let's say for 1/48th shutter you actually do four selective reads of 1/192nd. The first three reads use some type of a gate that only offloads pixels filled more than a certain value. The results of each of the these reads are accumulated in a buffer. The final read is a "give me all you got" read that does not employ a gate. The final read is added to the buffer, and the cumulative value in the buffer is sent to RedCode encoding.

Which means that RedCode needs to be changed to either logarithmic or more bits. I don't see how you can store such a wide DR 12-bit linear.


P.S. A variation of this would be if the close to full pixels could offload themselves to the buffer when they are approaching full.
 
Jannard said:
We have said all along that we love film, but at some point it is headed to the retirement home. Our fear was that digital replacements would not be good enough. We believe that the primary areas needed for digital to "measure up" are:

1. Resolution

2. Dynamic range

3. "Feel"

Resolution has never been an issue with RED. And people seem very happy with the "hand and feel" of RED footage, primarily because we don't have to sharpen our footage due to lack of resolution.

Tonight was a big night for us. We enabled our EPIC variable HDR mode and the results are hard to believe. We were so blown away that we called Micheal Cioni over from LightIron to witness it... otherwise we risked being called a scam (again). I think there is a new dynamic range leader in the camera world. We will post a chart shortly.

It was a good night over here at the RED Studios Hollywood... :-)

Jim
Click to expand...

When I think about all of the technological marvels that have taken place during my lifetime it's amazing to ponder what has occured over the past 50 years. But, the Red team is making history right in front of our eyes...transparently...and letting us all be a part of it. In that regard, it's a new world of technological achievements that I never thought I'd live long enough to witness. Let alone, be a small part of. You can feel the fundamental shifting as it's happening. Too often our world focuses on negative aspects of humanity. This is an example of mankind at their best. Innovation, inspiration, passion and creation. Congratulations Jim and the entire Red team.
 
Peter Moretti said:
Which means that RedCode needs to changed to being either logarithmic or have more bits. I don't see how you can store such a wide DR 12-bit linear.
Click to expand...

What about processing @ 16, storing @ 14, like in Epic?
 
G Lap said:
Maybe I've missed a post or two, but earlier in the thread there was this from Graeme:
Click to expand...

Jannard said:
Graeme's response was tongue in cheek... the over-analyzing of posts here is quite amusing.

Jim
Click to expand...

HDRx should still be available on the Scarlet. To what degree? I don't know. Since we're not sure about the method behind HDRx there is no reason to think that it's exclusive to the Epic.

Scarlet's cheaper because "high frame rates cost real money" which means faster processor, bigger bandwidth, more expensive cooling solutions. Jim's already shot down the assumption that HDRx is temporal based (i.e camera records at 120 fps with various exposures and "stuffs" it into a 24fps .R3D)

Someone at RED's also mentioned that the RED 1 MX couldn't get HDRx due to the new REDCODEs.

Will the Scarlet's HDRx be less impressive than the Epic's? Probably.

When I asked Jim if the Scarlet 2/3" DR would be significantly worse than the S35/Epic's due to sensor size he intimated that it wouldn't. Smart pixel design carries a lot of weight apparently. And since HDRx seems to be tied to the new cameras smarter architecture and improved soft/firmware then who knows what they're capable of implementing in the "little dog" :smile5:

Pure conjecture: If the Epic's current DR is 13.5-18 stops (w.o/w HDRx) I'm guessing that Scarlet 2/3" will be 11.5-14 stops. Why these numbers? I dunno. A hunch. :willy_nilly:
 
You must log in or register to reply here.
Back
Top