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 :)

Komodo 12.2 stops of dynamic range in CineD test.

T

Tomasz Nizielski

Member
Joined
Feb 3, 2020
Messages
21
Reaction score
0
Points
1
According to the CineD test, Komodo offers only 12.2 stops of dynamic range. I know that it's a global shutter sensor but I'm a bit disappointed, I was hoping for a better result. What's your opinion?

https://www.cined.com/red-komodo-6k-lab-test-dynamic-range-and-latitude-test/
 
More interesting than the Komodo DR measurement (and something I can anecdotally confirm/have noticed with all REDs), is the "highlight recovery algorithm" being decidedly monochromatic (Graeme has said as much about IPP2; reconstructs highlight DR out of the non-clipped channels). What I didn't realize was that it's two full highlight stops worth that are compromised. Not only is that a lot, but it also makes even over/under exposure almost impossible with r3ds (outside of maybe Monstro).

This just acutely demonstrates that "usable stops" (including recoverable stops that actually have decent info) are far more important than maximum measurable DR, especially since Xylas are only black/grey/white so you wouldn't see that massive colour shortcoming in the high end. Case in point, look at how many more colour consistent/recoverable high-end stops are available in ALEVIII (which is advertised as *only* "14+") or even BMD cameras. Heck, the overexposure section of that CineD article shows it clearly enough; only able to recover 1 stop over exposure on the skintone highlight.

The other question is why the prores file doesn't have the massive green colour shift in recovered under-exposure? I can understand why the prores is less noisy (it's a 6k>4k downscale for prores, right?), but don't understand why there's such a massive green tint? The only difference I can see is that they used the raw settings to bring up the r3d, but used the LGG for the prores file... Would the green colour cast still be there in the r3d if they left ISO at 800 and used LGG to boost it 5 stops instead?
 
Last edited:
Phil, it's that time of the year again!
 
Mike P. said:
This just acutely demonstrates that "usable stops" (including recoverable stops that actually have decent info) are far more important than maximum measurable DR
Click to expand...

As I have trying to say for a while.

However, I never read what IPP2 profile for R3D they used? Did they use that instead of a flat starting point? Because it looks like the graph bends far more like a curve that uses those end stops to curve off highlights and shadows.
 
This falls exactly in line with what I have seen from the footage thus far. It’s comparable to the A7S3 and behind the C70. But great footage has been captured with less. Motion cadence with the global shutter is better than the C70. That’s said the shutter of the A7S3 is fast enough that motion cadence is about the same for all intents and purpose.
 
Gregory Karydis said:
Phil, it's that time of the year again!
Click to expand...

It very much might be. This is not correct usage of Xyla, which is a ~3200K illuminate, and the chart seems to be set to 5600K which would taper off this way to the clipping point, which will also have an impact on their shadow stops due to how light works essentially. Red channel will clip first on tungsten for sure. But they are visibly off by more than 2 stops here within their own charts based on the small web graphics alone.

But yeah. I guess it might be that time of the year again.
 
Why would anyone use 3200k temperature on a camera sensor (cmos) that prefers 5600k light due to its blue starvation? If anything 5600k would improve the performance.... If they used 3200k light they would get blue noise in the shadows much sooner.
 
Jacek Zakowicz said:
Why would anyone use 3200k temperature on a camera sensor (cmos) that prefers 5600k light due to its blue starvation? If anything 5600k would improve the performance....
Click to expand...

Various emissive tests still utilize tungsten due to it's full spectrum. Not to mention it's still highly relevant in the industry in general. Utilizing corrective gels or optics can change this to whatever you want. All sensors in general have a different native color temperature and compounded by optics in the path that could also vary.

5600K would generally just improve blue channel noise, not really sensitivity ideally.

Also generally true is sensors do contain more luminance information than chroma, mostly commonly seen in really pushing still camera raw files, but what you are seeing here is a white balance error and not examining the clipping points.

I'll underline again there are slides included with the Xyla chart to block the patches that will inevitably eat up a stop or two in the shadows if you have the clipping patch and likely the first few highlight patches in frame mostly due to the flare creating lifted blacks/contamination.
 
I think the title is a bit misleading. The article clearly states that the 12.2 stops is for ProRes 422. RedCode Raw HQ is 10.8 stops of dynamic range.
 
Han Vogen said:
I think the title is a bit misleading. The article clearly states that the 12.2 stops is for ProRes 422. RedCode Raw HQ is 10.8 stops of dynamic range.
Click to expand...

This actually highlights a great deal of how and why IMATEST gets tricked so very often. Look at the two Xylas and compare those.

Want to have more fun? Trick it again via exporting out at ISO 1600 or 3200 and running NR of any sort.
 
As I wrote on another forum:

Myself said:
And it doesn't really matter anyway. There is a really simple formula to convert between DR formats: ADR * c = DDR, where ADR is Actual Dynamic Range, c is a 'constant' and DDR is Desired Dynamic Range. Using the appropriate constant, you can convert between numerous formats...

Example: ADR = 12.6 and DDR is 15.2, then your constant is c = 1.206. Just make sure you apply the same constant to all cameras under consideration, lest you end up with FDR (Forum Dynamic Range).
Click to expand...

Call it 8 stops, call it 20. As long as you compare camera tests done by the same individual/site you should be good. All is relative.

I personally much prefer results that land around 12 stops than those that reach 16+.
 
So, they did the test wrong?

Can maybe Red share a proper test like this because this shit is going viral as a negative post about Komodo and it’s gonna be the go to argument for many trying to discred the camera.

Personally, I would just like to have one proper test for once. There are so many people doing faulty tests and so many people arguing about what is what while Red just slap the usual 16+ stops onto the specs. No it’s not 16 stops, but I would rather see a proper analysis of the DR than the continuation of people not doing things right.
 
Christoffer Glans said:
So, they did the test wrong?
Click to expand...

At the day of publishing (yesterday) they originally posted with results stemming from not using full quality settings in Resolve. It's been updated now.

It doesn't change the overall performance (just brought the raw up to ProRes), but there was an issue with weak looking .r3d files and a strong green cast. They have also tested with WB at 3200 with identical results.

They summarise it as having better DR (marginally) than the P6K, but that's not what I'm seeing in the charts in the post. And that is when comparing Komodo WITH highlight reconstruction and P6K without.

If the end of 2020 and beginning 2021 taught us anything, it's that there is no truth anymore. Everyone just defines their own world and true north. And this goes all the way to the top, as we've seen.

Bob gives sage advice in post #2. Maybe people should compare less, just look at the images and decided for themselves if they are any good?
 
AndreeMarkefors said:
At the day of publishing (yesterday) they originally posted with results stemming from not using full quality settings in Resolve. It's been updated now.

It doesn't change the overall performance (just brought the raw up to ProRes), but there was an issue with weak looking .r3d files and a strong green cast. They have also tested with WB at 3200 with identical results.

They summarise it as having better DR (marginally) than the P6K, but that's not what I'm seeing in the charts in the post. And that is when comparing Komodo WITH highlight reconstruction and P6K without.

If the end of 2020 and beginning 2021 taught us anything, it's that there is no truth anymore. Everyone just defines their own world and true north. And this goes all the way to the top, as we've seen.

Bob gives sage advice in post #2. Maybe people should compare less, just look at the images and decided for themselves if they are any good?
Click to expand...

3200 brings RGB to a balanced point and doing that would add a stop to their count since they basically removed one stop because one channel dipped below it... which it does because they had the wrong white balance.
And there's no word on processing the files. How's the post IPP2 implementation. How does the DR handle after supersampling and higher quality NR like neat video? This test is so flawed that I'm surprised it comes from them. I was waiting for this test because I previously held them to a higher standard compared to everyone else. But people just can't seem to handle the Red infrastructure correctly at all. It's all over the place.

It's definitely possible to test this in a scientific way. If people had the knowledge required to do so. But all there is, is either people who do tests without having any knowledge of the system they use or compare between, while others do tests with so much bias that it becomes a parody.

If anything 2020 taught us, it's that people give up on truth in favor of just letting the bullshitters roam. Why can't someone do actual tests on this camera? Really dig deep and get to the true specs. I'm tired of marketing BS, I'm tired of self-proclaimed experts on youtube with their click bait thumbnails with texts like "BM destroys!" "This is insaaaaane!".

People seem to have lost all methods of rationality in favor of being clickbait crackheads eating bite-sized snacks of online bs before marching out to do yet another pointless vlog video about essentially nothing or the same as every other crackhead.

I just want to see a proper test... a comparison against the previous DSMC2 lineup. Pros and cons, charts, DR measurements etc. done by someone who's actually knowledgable with camera systems and post-processing of RAW footage. It would be nice if Red did this themselves, I trust Graeme over most people in the industry, he's in my opinion one of the backbone people at Red who really knows his stuff. I would like him and others like him to do some real tests giving real data. Not marketing, not badly conducted tests, but the real deal.
 
Christoffer Glans said:
3200 brings RGB to a balanced point and doing that would add a stop to their count since they basically removed one stop because one channel dipped below it... which it does because they had the wrong white balance.
Click to expand...

As I wrote above, they did test at 3200 too. The results were identical. I don't remember now if Gunther wrote it in the post or in the comments below as a reply to someone. And I'm assuming he's talking about the numbers Imatest spits out—those are the numbers they 'post' at the end of the day at various SNR levels.

I agree with a lot of what you say, but they can't change their method around every now and then. Super-sampling and NR have not been part of their standardised test (I think they do mention it sometimes in the text though), so I wouldn't expect them to include it all of a sudden.

The main win is that all cameras they test use a similar setup.
 
Seems like the test was updated due to Highest Quality not being chosen in Resolve, which of course has an impact on IMATEST due to how the R3D is rendered. Now they updated and state:

"IMATEST calculates 12.5 stops at SNR = 2 and 13.6 stops of dynamic range at SNR = 1"

Now if we can suggest to them to properly use the Xyla slides you'll gain a bit more in the shadows. And they aren't counting 2 of the stops of captured information in the highlights. Essentially getting you to what everybody else is getting.

No offense because mistakes happen, but the damage is already done based on how that wildfire spreads on the net. Thankful they updated most of the issue.
 
I do think the issue of whether to include recovered highlights or not is a complicated one. Is Komodo the only camera currently to have highlight recovery built in and visible on the monitor during recording? I would say it is a bit complicated since conceivably such highlight recovery is available to other raw cameras but not visible until processed in post. In your opinion how should the xyla chart be exposed with respect to RGB channel clipping? Should it the top chip be counted as in as long as at least one color channel is not yet clipped?
 
You must log in or register to reply here.
Back
Top