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 :)

RED Weapon - Dynamic Range and Latitude Examined

Phil Holland

Phil Holland

Administrator
Staff member
Joined
Apr 19, 2007
Messages
13,392
Reaction score
796
Points
113
Location
Los Angeles
Website
www.phfx.com
Hello all. I've been going through a few hundred test images the last few days and feel this post might be a useful one to share.

Upon taking delivery of Weapon I got straight into shooting. However, I was beginning to feel something was a bit different. Then after further digging and getting intimidate with the camera I discovered there indeed was a difference in the Total Captured Dynamic Range between RED Epic Dragon and Weapon. It seems that RED has found a way to get a bit more out of the sensor, the most obvious changes are a slightly lower noise floor and a bit more highlight detail retention. Which essentially equates to more Dynamic Range.

That got me curious enough to test this. First and foremost I'd like to thank Stacey Spears and SpectraCAL for the use of their DSC Labs Xyla-21 Chart. This allowed me to compare my normal "patch reading" method of measuring Dynamic Range and the Xyla-21 Chart's individual steps for additional confidence in the results. I did a great deal of work here. Reading individual patches, masking off sides of the chart to eliminate glare, composting results, OLPF noise comparisons, gamma experiments, etc.... But there's certain information that I feel is very important towards the practical and technical ways of shooting that I want to share. And I think this not only effects Weapon, but likely other upcoming DSMC[SUP]2[/SUP] style RED cameras. The results found here are accurate within a tenth of a stop btw.

Let's dig in.


What Can Weapon See?

Well that was my first question and I wanted to see exactly what I could dig out of a Weapon image and the Xyla-21. Which led to a rather ludicrous ISO 102400 pushed test using Frame Averaging.



Remember this is a stress test, but yes, Weapon can indeed "see" all 21 stops amazingly. Are those "Usable Stops"? No, not exactly, but it does tell a bit of the story.

Something to remember through all of this, when you see these Xyla-21 frame grabs and graphics all of the images are exposed to not clip in Red, Green, or Blue Channels. Meaning that first cube isn't thrown away and is a "counted stop".


Recommended ISO Range

To date RED has recommended an ISO Range of 250-2000 on the Dragon sensor. Though the cameras do indeed go up to ISO 12800. While I personally wouldn't push it that far, after shooting this test footage (specifically the ISO Tests on the Color Chart at the end of the video), you can indeed go beyond the recommended ISO Range and get very usable results.

With Raven on the horizon I feel there's going to be a new audience who don't have a full grasp of how ISO and FLUT work on RED cameras. I don't want to re-tread too much here, but take a look at this graphic if this is new to you:
http://artbyphil.com/phfx/red/dragon/images/phfx_redDragon_ISOFlutBreakdown.jpg

Long and the short of it, RED cameras capture all the potential Dynamic Range possible for a given scene and exposure. RED doesn't apply any in camera noise reduction as they want to maintain the most possible Dynamic Range, Detail, and Color Information without losing it to aggressive lossy image processing techniques.


Total Captured Dynamic Range

RED has famously received a bit of criticism about the 16+ stops of Dynamic Range of the Dragon sensor, but it's something that I've been able to measure in the past:
http://www.reduser.net/forum/showthread.php?111400-Dragon-Flare&p=1322384&viewfull=1#post1322384
Stacey and Gavin also did the Xyla-21 test on Epic Dragon way back when:
http://www.reduser.net/forum/showthread.php?117540-Dragon-Dynamic-Range-Test-21-stop-chart

Weapon is stated to have 16.5+ stops of Dynamic Range.

Let's take a look at ISO 250, 800, 2000, and 6400 REDlogFilm Examples:













Remember that's REDlogFilm with No Curve applied. On the ISO 250 image on a calibrated display I can see separation down to 17th Captured Stop. ISO 800 down to the 18th, ISO 2000 down to the 20th, and ISO 6400 down to the 21st.

So how much Dynamic Range does Weapon have? For me if we are talking within the Recommended ISO Range of ISO 250-2000 up to 17-20 Captured Stops. Those stops nearing the noise floor aren't going to be the cleanest, but they are a part of the Shadow Roll-Off and reflect what's captured in camera. The safest "real answer" is indeed 16.5+ Stops of Total Captured Dynamic Range.

To show the effect of what happens once we toss a Gamma Curve on there I've also created samples for REDgamma4:
ISO 250
ISO 800
ISO 2000
ISO 6400


And to show the Dynamic Range properties more clearly, here's an image of the Xyla-21 with each ISO Rating shown in REDlogFilm:




And if you'd like to see it, a link to the REDgamma4 result:
http://www.artbyphil.com/phfx/cinematography/2015_REDWeaponCameraTest/bigs/phfx_REDWeaponTest2015_xyla21_ISOPatches_RG4.jpg


Differences Between the OLPFs

We currently have 3 Optical Low Pass Filter options for standard shooting. Skin Tone - Highlight, Standard, and Low Light Optimized. These 3 OLPFs are tuned in camera to accurately produce a consistent middle gray and fairly close color response.

There are slight differences between the OLPFs however.

All 3 OLPFs "see" about the same into the dark. However, the Low Light Optimized OLPF for instance is cleaner in the shadows when compared to say the Skin Tone - Highlight OPLF or Standard OLPF. The other big difference is how each OLPF's properties effect the Highlight Roll-Off. Also, possible Optical Artifacts are reduced the most with the Skin Tone - Highlight OLPF at the cost of a slightly noisier image.

Here's a few images that show those differences a bit more clearly:




Looking at the brushed steel you can see each of the OLPFs are clipping. Within the circled area you can see the different properties of the Highlight Roll-Off for each OLPF.




To show off the image texture noise it's easiest to go to an extreme ISO like 6400 and look at these Middle Gray through shadow tones for each OLPF. Theses are 100% crops of 6K material. The square next to 3 is Middle Gray. You can see that the STH is noisiest, the LLO is cleanest, and the STD is in the middle between them - though closer to the LLO in my opinion in many ways.
 
Last edited:
  • Thread starter
  • Moderator
  • #2
Brief Film Stock Chat

Film is relevant to my life, career, and to the industry. I've already done a few interesting tests regarding ISO and Film Grain here:
http://www.reduser.net/forum/showthread.php?117820-Red-Dragon-ISO-Texture-Versus-Film-Grain

However, what I can safely say is Epic Dragon and Weapon are cleaner at each equivalent ISO/ASA.

The conversation of Dynamic Range is also a bit different now. If we are talking about Kodak Vision3 we are looking at a total of 13 stops of potential Dynamic Range. Vision3 500T 15 stops max.

In terms of technical merit, we have surpassed motion picture film on Resolution, Color Accuracy, and Dynamic Range at this point. However, there's still a lot of intangible and tangible aesthetics to film and the chemical process that will continue to be attractive to some filmmakers. For me personally, especially since we have access to the improved Data Transfer Rates of Weapon (lower REDCODE RAW Compression Ratios), I think I've moved on for good at this point. The future of our industry likely falls in the 4K and 8K delivery realm and with things like HDR display technology on the rise, this color and Dynamic Range performance will become even more important for some.


Dynamic Range and Workable Latitude

So with all of this information there's something else I tested during this last week under a few conditions, and I know somebody (Martin!) will ask where the middle point between clipping and crushing is on Weapon. The answer to that question is "it depends". Truly what we need to to know is what the Workable Latitude is. For those of us that still use light meters this is really important.

So here's a breakdown of Workable Latitude from Middle Gray to Highlight Clip:




In terms of a practical example, if you are shooting, exposed properly for a scene at ISO 800 and find you have something that's 9 Stops above Middle Gray, there's a good chance one or all of your channels is clipping.

As for where your Shadow Latitude falls, that's going to depend on where you really want it to fall and how you want your footage to look.

My general advice is to work within the Recommend ISO Range of ISO 250-2000 with Weapon. Scaling from 6K, 5.5K, 5K, 4.5K to 4K or 2K makes a difference as well, so be aware of that if you are feeling like you're on the edge of what you're comfortable with. I've been working a lot at the Base ISO of Weapon so far at ISO 800, but I'm sure during the course of my Weapon shooting I'll be all over the place.


Hopefully that is/was useful. It was fairly revealing to me and made me understand some of the "more" we are getting with Weapon.
 
Last edited:
  • Thread starter
  • Moderator
  • #5
Thanks guys. I had a lot of "ah hah" and "oh wow" moments during this test. It was time consuming to do everything, some tests replicated up to 5 times just to make sure. Fascinating stuff really.
 
Very interesting Phil. Thanks for looking into this!
I just did my first test comparing the colors between the MX sensor with the Dragon (yes a bit late to the party). Quite a challenge to make sure all variables were as close as possible I must say and this test look a lot harder.
/Andreas
 
  • Thread starter
  • Moderator
  • #9
AndreasOberg said:
Very interesting Phil. Thanks for looking into this!
I just did my first test comparing the colors between the MX sensor with the Dragon (yes a bit late to the party). Quite a challenge to make sure all variables were as close as possible I must say and this test look a lot harder.
/Andreas
Click to expand...

I won't say which test it was, but I did sprain my ankle in the dark doing one of these :)
 
Great!! Thanks for your time my friend ;)
 
"It seems that RED has found a way to get a bit more out of the sensor, the most obvious changes are a slightly lower noise floor and a bit more highlight detail retention. Which essentially equates to more Dynamic Range."

Phil,

I understand at the extremes it can be hard to quantify exactly but how much more dynamic range do you believe the Weapon has over the Epic weapon? Would you agree on the 1/2 stop in the shadows from Jarred and would there be another 1/2 stop in the highlights? Again thanks for the hard work.
 
Wow Phil!


Amazing/// thanks for the time and sharing... the house in going down.

This is sure beauty in form of graphics!
 
Joel Voelker said:
Phil, I really appreciate all your tests but they're going to drive me insane waiting for my Weapon upgrade!
Click to expand...

Agree at the moment I work on possibly the coolest stuff I touched in this industry... And sadly I tomorrow have to go rent a dragon to shoot some plates for it. As our dragon got stolen and I thought that the woven would come out this summer, if I was smart (I'm not by the way) I would have ordered the forged straight away. the only shit thing with red upgrade paths is that no matter what, it has always been moth beneficial to just order the shit expansive stuff as soon as it´s out the gate. If I would have ordred a forged right at nab, then those extra coins would have been long gone paid by now and I would actually have a camera in my hands, it hurts to think of, now I go grab another beer and put another render. You guys like David Bowie?
 
  • Thread starter
  • Moderator
  • #19
Tom Gleeson said:
"It seems that RED has found a way to get a bit more out of the sensor, the most obvious changes are a slightly lower noise floor and a bit more highlight detail retention. Which essentially equates to more Dynamic Range."

Phil,

I understand at the extremes it can be hard to quantify exactly but how much more dynamic range do you believe the Weapon has over the Epic weapon? Would you agree on the 1/2 stop in the shadows from Jarred and would there be another 1/2 stop in the highlights? Again thanks for the hard work.
Click to expand...

It is exactly that. Difficult to quantify/describe.

Here's what I can say. The images are slightly cleaner, so the noise floor is a smidge lower. There appears to be more highlight retention for sure as well. That lamp image shows a clear 1 stop advantage at the top end. So it feels like things have been tuned differently across the board. What's really telling is the ISO 2000+ samples. There's a lot of DR in there. For me in practical terms it's around a 1.5 stop "difference" overall perhaps. I rarely shoot beyond ISO 1280 for my typical shoots, but there's more meat now and where the meat is what is actually important.

Slight difference for some, noticeable more when you start taking the implications of things like high speed shooting into account. There's also the CMOS Smear improvement, that can't be overlooked. It's a big difference.

Right now I'm very, very happy with where the camera is, it's already revealed that in the tests and the handful of shoots so far. Weapon is just doing better slightly in pretty much every way. Feels a little bit like a huge body upgrade with a slight sensor upgrade on top, if you want to classify it that way.
 
You must log in or register to reply here.
Back
Top