Welcome to our community

Be a part of something great, join today!

Register 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

Thanks for the test,

Did you use 24p for the test and frame averaging only for the 100'000 iso test?

I don't understand why red did RG4 as its the hardest gamma curve that kills so much details in the greys. It's the opposite of the Alexa look which is now so trendy... But it doesn't help Red's reputation in being a better DR camera.

Pat
 
Patrick Tresch said:
Thanks for the test,

Did you use 24p for the test and frame averaging only for the 100'000 iso test?

I don't understand why red did RG4 as its the hardest gamma curve that kills so much details in the grays. It's the opposite of the Akexa look which is now so trendy... But it doesn't help Red's reputation in being a better DR camera.

Pat
Click to expand...

I can totally follow that thinking, as people tend to underexpose a lot, and then come back and complain about noise, and with STH it is more like giving people a bit of lowlight headroom (or maybe rather "footroom" for post...
 
  • Thread starter
  • Moderator
  • #43
Raven question. Can we assume that at 4.5 K 3:1 the Raven will have similar characteristics. Obviousl;y Weapon goes beyond, but 4.5K-4.5K these DR characteristics will translate?
 
Gunleik Groven said:
I can totally follow that thinking, as people tend to underexpose a lot, and then come back and complain about noise, and with STH it is more like giving people a bit of lowlight headroom (or maybe rather "footroom" for post...
Click to expand...

But you also give them less latitude to work with. They will play in something like 10 stops black to white on set.
 
Recoverable Highlight detail

Recoverable Highlight detail

Phil,

One thing I'm curious about still, is the top 3-4 chip in the highlights seem to have luminance values very close to each other. How much detail can be pulled out of these say in the highlights of a cloud? Are all 3-4 of these stops still usable even though to my eye they are all the same shade of white?
 
  • Thread starter
  • Moderator
  • #49
John Hafner said:
Phil,

One thing I'm curious about still, is the top 3-4 chip in the highlights seem to have luminance values very close to each other. How much detail can be pulled out of these say in the highlights of a cloud? Are all 3-4 of these stops still usable even though to my eye they are all the same shade of white?
Click to expand...

Great question and something I noticed as well immediately when hitting the chart with GIOSCOPE once exposed. The first 3 chips are indeed 1 stop apart in the real world and what we are seeing is where the highlights are rolling off as that's at the tip of the exposure. The value distance between them what's captured and displayed is seemingly small, but the stops are indeed captured. From what I can see it's how the curve is mapped for REDlogFilm. In Linear the separation is still slight, but also slightly more apparent. So I think what matters is that the camera captures that slight separation in this case, but this is what happens "up top" right at the moment before clip. If I made that first patch middle gray, it maps differently of course.
 
Patrick Tresch said:
But you also give them less latitude to work with. They will play in something like 10 stops black to white on set.
Click to expand...

Yup, but are less likely to mess up.
Not saying I agree with the philosophy, but judging from all the "my image has noise when I underexpose at 3200 ISO" threads... I think it may have some merit.
Nothing is stopping you from shooting with RG2 (or even view your image in RLF) if you want a bit more control, and know what to do with that.
 
Very useful stuff, as usual. Thanks Phil!

I was just having a convo this morning with somebody who was thinking about film school and I told him I learned everything from YouTube, Vimeo, blogs and forums.
 
Phil : Thankyou for creating such a fantastic resource for us.

If the Weapon 8k sensor (when it comes out) was used to finish in 4k - what are peoples thoughts on what a light-starved image would look like?
Would the image look Less saturated but with 1 or 2 stops more usable into the dark (when compare to the 1:1 image recorded)?

Going forward - do we think that a Signal-to-noise on each XLYA-21 checker board / or a mini Histogram above each box would be useful?

My guess is that there will be a consistent correlation between Grain/Texture and dB Signal-to-noise ... and that this sort of measure would help people get a feel for their own grain limit is.
It might be a fair metric to compare with other cameras : both their native resolution, and for 4K UHD finishing (using an unsophisticated averaging algorithm).

AJ
 
Antony Newman said:
If the Weapon 8k sensor (when it comes out) was used to finish in 4k - what are peoples thoughts on what a light-starved image would look like?
Would the image look Less saturated but with 1 or 2 stops more usable into the dark (when compare to the 1:1 image recorded)?
Click to expand...

My thoughts are that mathematically this is absolutely no different than using a 4K sensor to finish in 2K. Which means if you already have a DRAGON of any kind, you can do the experiment yourself and find out.

I also think that since we have seen nobody saying "Hey, I discovered that I can get an extra 1-2 stops more useable in the dark when I shoot in 4K and finish in 2K!" that there's no there there.
 
Well, more I'm looking at the test more I see 250 iso and RG4 as a risky bet, I would only use it in some special situations.
With this setting you get really clean blacks but you get "only" 12+ stops.

I would go with 1000 iso and RG3 to get close to 16 stops if I need more DR.
2000 iso has this beautiful texture in it that I love so much, a full debayer from a 6k r3d will make it shine IMHO.

Michael Tiemann said:
I also think that since we have seen nobody saying "Hey, I discovered that I can get an extra 1-2 stops more useable in the dark when I shoot in 4K and finish in 2K!" that there's no there there.
Click to expand...

But you can push the iso a bit further when you downscales from 6k to 2k instead of 4k (3200iso "looks like" 2000). Not only the denoiser will work better with a greater resolution but the noise looks also less present in the downscale process. I don't know if this is also due to the algorithm used?

Pat
 
  • Thread starter
  • Moderator
  • #56
Antony Newman said:
Phil : Thankyou for creating such a fantastic resource for us.

If the Weapon 8k sensor (when it comes out) was used to finish in 4k - what are peoples thoughts on what a light-starved image would look like?
Would the image look Less saturated but with 1 or 2 stops more usable into the dark (when compare to the 1:1 image recorded)?

Going forward - do we think that a Signal-to-noise on each XLYA-21 checker board / or a mini Histogram above each box would be useful?

My guess is that there will be a consistent correlation between Grain/Texture and dB Signal-to-noise ... and that this sort of measure would help people get a feel for their own grain limit is.
It might be a fair metric to compare with other cameras : both their native resolution, and for 4K UHD finishing (using an unsophisticated averaging algorithm).
Click to expand...

Oddly you mention 8K. Just did a consultation and talked about how 8K capture could be used for 4K delivery. In terms of what the sensor can see, if it is the same exact tech found in Weapon 6K today, it's going to "see" the same. However, an image scaled from a full 8K frame to a 4K frame will have less apparent visible noise then compared to a 100% crop of the 8K frame of course.

What I like about the Xyla-21 is it's a practical chart. It's no nonsense, a bunch of evenly separated steps, and something you can expose for just like you would for a given scene. A mini histogram? That I don't see the need for, maybe a dB ratio though, but also not so "practical" I think. Patches can be read and examined now. The changes I want are more geared towards the future. We'll see what comes of that.


Akin A said:
Very useful stuff, as usual. Thanks Phil!

I was just having a convo this morning with somebody who was thinking about film school and I told him I learned everything from YouTube, Vimeo, blogs and forums.
Click to expand...

Thanks Akin. Film School still serves an important role, a very important one if you have never shot motion picture "anything" before as well.

Information found online is a mixed bag. Some good, some bad, some biased, some from individuals who are more interested in being a personality than a resource, etc...... The tricky thing there is finding the good stuff.

However, I do think in person and hands on experience you can get from Film Schools (and for the lucky, actually working on shoots) is much more important. Spending time in person with experienced professionals and getting hands on experience with the technology is much more enriching than staring into our digital windows to the world. It's one of the things that REDucation sort of excels at. You get practical information about the camera and you can also get down to shooting quality motion pictures with those tools. For people who have no experience with the demands of the industry or using the cameras, that's pretty damn important.

I was very, very fortunate coming up through the feature film side of the industry. A weird and fascinating combination of pre-production, production, and post production on the job experiences. It was a different time in the industry when I was 17, smaller really. When I got my break around the age of 19 I was able to learn from some of the top working professionals on a nearly daily basis and on the job, which has assisted greatly in my own creative problem solving abilities. The on the job experience and mentor-ish driven learning environment accelerated things greatly for me. Still to this day really. I've now mentored at NYFA, instructed at Gnomon, and have done various workshops. At the same time I don't have a degree, but I do have a lot of experience. I think that's what it comes down to for film schools or not. They are right for some, not right for others. But if you are starting from square one it's important to start gathering quality experience. For some an educational institution is the right system. For others, self study and exploration is valid. However, at some point, I do feel it's important to reach out and learn from those you respect if possible. There's so damn much to be learned from working professionals as well as working on productions. And much of that isn't found online, but you get tastes of it here and there for sure.
 
I understand the desire to have one DR test to rule them all, but how do you account for noise tolerance? For some shooters, Evin and Phil are good examples, they want a certain amount of "texture" in the image when "clean" might undermine the creative tone. Of course they, like most of us, have some situations where kitchen clean is exactly what is desired. So there are two major variables in determining which is the last "usable" stop in the shadows - personal taste/style and type of look desired for that particular shot/scene.

As Stacey notes, Geoff Boyle has done a fine job of comparing DR on the top cameras, but any practical testing methodology is limited to certain stipulated reference points that may favor one capture tech over another. Moreover, the amount of digital signal processing (DSP) done upstream from recording can vary widely. As I have noted many times, R3Ds are more "raw" than what you see from other manufacturers outputs. RED has always tried to give us as much image data as possible so we can characterize the image more precisely during raw development. Influencing how the Bayer CFA data set is constructed into an RGB image is a powerful tool, but, as the saying goes, with great power...

Cheers - #19
 
Phil Holland said:
I guess I would need to test that to be precise. It will also likely depend on what your rated ISO was as well if you think about it.
Click to expand...

Thanks Phill.

Looking forward to your Weapon colour test.
 
I should elaborate on the matter of increased ISO or no. Frame averaging can increase signal to noise ratio (and hence dynamic range) because signal is correlated and noise is not. When you average frames, the correlated signal averages to its correct level whereas the noise averages toward zero. There are many references showing that every time you double the number of frames averaged, you get another stop of DR. 2 frames is +1 stop. 4 frames is +2 stops.

But this math doesn't work so well for averaging four pixels into one, such as reducing 8k to 4k or 4k to 2k. The reason is because the different pixels are seeing different signals, as well as having different noise. When 4 pixels are averaged as one, the signals don't accumulate vs. the noise. They average vs. the noise. Now, if you have a perfectly uniform object in shadow, such as the 21st chip of a Xyla chart, the signals are coincidentally correlated, so you could see deeper. But if the object is a garden-variety object that has color, texture, differentiation of some sort, and it's in deep shadow, a 4 frame average will see 2 stops deeper into the noise floor, whereas a four pixel spatial average will not.

So 8k->4k or 4k->2k may chart better, but in the real world it buys you something between 2 stops (for uniform textureless surfaces) and nothing at all (for realistic, complex objects).
 
Last edited:
John Hafner said:
Phil,

One thing I'm curious about still, is the top 3-4 chip in the highlights seem to have luminance values very close to each other. How much detail can be pulled out of these say in the highlights of a cloud? Are all 3-4 of these stops still usable even though to my eye they are all the same shade of white?
Click to expand...

It also seems like the final ISO is a bigger jump in exposure than expected.
 
You must log in or register to reply here.
Back
Top