ALEXA | F65 | DRAGON | HELIUM | Holiday shootout!

[Anton Esteban]

Hey Hrvoje I'd love to see you take a crack at matching the two cameras, even from this little test. You definitely talk a lot of game. Why don't you show us how it's done? Where are all of your camera tests?

 

[Alex Boothby]

Thank you Christoph!!!

 

[Nick Morrison]

Cameras are like tools. Learn to use and master them.

The master makes the art - not the tool.

 

[Daniel Stilling]

There is no subjective interpretation in the statement that the adequacy of final outcome depends on the proper post treatment as well. Post can be only as good as the shot material allows but proper shot doesn't remove the need for proper post.

Someone needs to create that "decent LUT" and that doesn't fall from the sky into the lap or come from exporting something from Resolve from a few test shots but from thorough analysis and testing, and more complex methodology reaching far wider and deeper.

Also, with that decent LUT at hand the final outcome still depends on quite a few factors, which depend on the person sitting in post using it. LUT can provide improved results and reduced offsets, not a perfect camera match. That part still requires - a competent person in post.

As I said, I understand the point which was suggested - that the motion imaging tools quality has improved and certain issues are reduced. But...let's not get ahead of ourselves here and expect a fool-proof one-click solution to all variables and troubles, thinking the story ends by pressing the red button.

Your methodology and theory is fine. But it's also just as subjective as anything regarding the art of cinematography and how to use a camera. To say that the only way to do so is the way you say, shows me you haven't had enough varied experiences to know that there are so many different ways to skin this cat.
One way is to do it the way you described, sure. But don't disregard everyone else experiences.
You would be surprised on how many productions I shot, where they barely had time to cut it, let alone do any color work. Serious stuff. Live concerts, national commercials. From camera to delivery basically. So my job, and why I can command my rate, is to listen to those needs, and get the best result possible within those parameters. And if I have to process the look right in camera, I have enough experience and knowledge to do so, and not depend on post production.
I don't know you, but I pride myself in being able to deliver excellent results no matter what happens down the pipeline, and I also like to think that I'm quick on my feet and able to roll with the punches to be able to do so. I trust that you'll eventually be able to come around and understand what I mean. Age and accumulated knowledge tends to soften absolute and unmovable opinions...

 

[Michael Tiemann]

Your methodology and theory is fine. But it's also just as subjective as anything regarding the art of cinematography and how to use a camera. To say that the only way to do so is the way you say, shows me you haven't had enough varied experiences to know that there are so many different ways to skin this cat.
One way is to do it the way you described, sure. But don't disregard everyone else experiences.
You would be surprised on how many productions I shot, where they barely had time to cut it, let alone do any color work. Serious stuff. Live concerts, national commercials. From camera to delivery basically. So my job, and why I can command my rate, is to listen to those needs, and get the best result possible within those parameters. And if I have to process the look right in camera, I have enough experience and knowledge to do so, and not depend on post production.
I don't know you, but I pride myself in being able to deliver excellent results no matter what happens down the pipeline, and I also like to think that I'm quick on my feet and able to roll with the punches to be able to do so. I trust that you'll eventually be able to come around and understand what I mean. Age and accumulated knowledge tends to soften absolute and unmovable opinions...

Very much agree that when doing live events, where the event is #1 and everything else is #2, the best tools are one's own experience and creativity/adaptability. I do lots of live events, and I post-mortem every one so that the next time I can be one step closer to doing everything right. It's a painful way to gain experience, but it has served me well. I'm now very confident of what I meter, what I shoot, and what range of options I'll have by the time the footage gets to post.

 

[Björn Benckert]

Proper tests are done with mid point as a reference because that way you can see how much usable DR you get above mid grey and how much below, which, if you have measured the scene, is much more informative about camera's dynamic range capability and provides insight to the DoP for required proper exposure. Which also depends on implementation of camera log, not just the sensor available DR.

One shot is not telling you the full story, which is why proper camera DR tests consist of a range of stops, from complete underexposure to complete overexposure.

You are probably right. But to me looking at the midpoint seam to lead to a lot of misunderstandings when comparing different cameras to each other. As the mid point is no point at all but more like a rather wide field where you pick a mid point depending on your noise acceptance. And acceptable noise is varying on quite a few factors. Things like light temperature and final resolution very much affect what we find as acceptable noise and it´s most of all a very personal taste kind of thing. So starting from a mid point makes it rather difficult to look at an image from one camera and say specifically how it behaves different from another. Then if you add camera log to the mix its a bit like trying to compare different film stocks but using different developing fluids while doing it. Simply you can not really know if its the image processing that gives one result or the other or if its the actual camera. So to me its more interesting to look at cameraRGB or such and compare from there. Camera log is made for a mid point kind of exposure but then it also applies the camera manufacturer taste and preferences for what will be used for dynamic rounding etc.

 

[Sabovic Adis]

Then again, adjusting the lens, in order to hit the max input level of the sensor, leeds to all kinds of new variables such as
softer or sharper image depending on set f/T stop, chromatic aberrations or diffraction limit, DOF/FOV, rendering the test itself invalid. And not all lenses are the same...

It's only good for finding out the best conditions under which the particular camera gives its best, not for A/B comparison.

 

[Hrvoje Simic]

Happy new path around the Sun in a new Nature's cycle, to all the creative Souls across the globe, using any tools or means of expression...

 

[Björn Benckert]

Then again, adjusting the lens, in order to hit the max input level of the sensor, leeds to all kinds of new variables such as
softer or sharper image depending on set f/T stop, chromatic aberrations or diffraction limit, DOF/FOV, rendering the test itself invalid. And not all lenses are the same...

It's only good for finding out the best conditions under which the particular camera gives its best, not for A/B comparison.

That is why I sugest to use time/shutter to set the exposure. As it does not come with those variables. As sure changing stop on the lens changes pretty much everything. But sure it does not work for motion blur test but everything else stays quite intact as the adjustment is also rather slim, we are taking about shutter adjustments equal to less then 2 stops or such.

 

[Sabovic Adis]

That is why I sugest to use time/shutter to set the exposure. As it does not come with those variables. As sure changing stop on the lens changes pretty much everything. But sure it does not work for motion blur test but everything else stays quite intact as the adjustment is also rather slim, we are taking about shutter adjustments equal to less then 2 stops or such.

Not sure that exposure time (seconds) can be related to or translated to illumination (lux), or strength of the light current (Lumens) but I get Your point, Björn. :biggrin:

 

[Stacey Spears]

There is no subjective interpretation in the statement that the adequacy of final outcome depends on the proper post treatment as well. Post can be only as good as the shot material allows but proper shot doesn't remove the need for proper post.

I agree with that. I think Phil nicely demonstrated the importance of white balance.

Just because two cameras have an ISO called 800 or a color temp called 5600 does not mean that each one still needs to be properly adjusted in post to get things on equal footing. Otherwise they are not really comparing the same fruit.

I am sure one could argue that all cameras set to ISO 800 should match, and perhaps that is true in theory, but not in practice, not so much. Shoot your test chart and do proper post and then compare.

With that said, I think Christoph has done a nice job overall. This seems like the type of test where ACES should get them on more equal footing. At least that is what ACES should do in theory vs. practice.

Comparing multiple anything is never easy. Be it cameras, displays, cars, etc... You also need some thick skin. Comparing two or more of anything is always going to get put under a microscope and get lots of feedback both good and bad. Hence the thick skin. You really need to understand how each product works to get the most out of it. And then when a fanboy gets involved things can go sideways because of confirmation bias. I enjoy looking at the tests though. I learn something new from every new test done. Maybe a new way to look at something or stuff to avoid.

For some image processing stuff I am working on, I have been adjusting the light so that when I hit auto white balance it changes very little vs. letting it makes large corrections. Not sure if it is worth the effort to adjust the lights, but interesting data none-the-less. What really gets me is how far off these LED lights are in color temp from what their display reads. Then when you add a softbox, egg crate, or anything in front of the light it is even further off. And now I am just rambling.

 

[Christoph Tilley]

Absolutely agree with all of you that each and every camera has to be treated separately. But to show differences between these high end image acquisition devices we chose to shoot from a common midpoint which bases on standards: ISO800, 5600K and 0 Tint. Of course this test does not reveal which camera is better than the other in particular, but I'm not sure if any test is doing this.

That being said I tend to shoot the Dragon and Helium with Tint -10 applied. That gets rid of the odd green cast which is quite annoying for me to be honest. We shot this test with only one 800W HMI source, so CRI should not be an issue.

But way back when Dalsa revealed their first CMOS sensors and Medium Format camera manufacturers like creo/leaf started to use this technology in their products, we had the same problem in developing the imagery: either magenta or green cast.

 

[Joel Arvidsson]

Happy new path around the Sun in a new Nature's cycle, to all the creative Souls across the globe, using any tools or means of expression...

So what does each image represent?

 

[Hrvoje Simic]

So what does each image represent?

A shot from Christoph's test, exposed for the face, each from a different camera.

 

[Christoffer Glans]

I shot the Dragon on DC2/RG4 because afaik with the Epic DSMC it is not possible to use IPP2 in camera. But you can download the R3Ds from Dropbox and apply the look from there. Upload should be done by now.

If the footage isn't processed in post and this is just a test of in-camera perfromance that's bypassing a lot about Red infrastructure. I'm sorry, but I don't see the point of comparing just in-camera footage from a RAW-workflow based camera to cameras that does most of it's processing in-camera like the Alexa. The whole point of shooting Red is to process the material as a one-light first, be it white balance or adding sharpness. A true proper test of the new color science should be to do a super basic neutral one-light of all cameras with a 4K output to compare. Then all basic things like tint, white balance, sharpness, noise etc. would be a matched up comparison showing the real differences between the different sensors.

It would be more interesting to see a all white balances neutral between the cameras and both the Dragon and Helium put through the new color science, otherwise it's only representing the client monitor on set, not the final output capabilities of the cameras.

 

[Hrvoje Simic]

If the footage isn't processed in post and this is just a test of in-camera perfromance that's bypassing a lot about Red infrastructure. I'm sorry, but I don't see the point of comparing just in-camera footage from a RAW-workflow based camera

Which is why Christoph generously shared 60 GB of raw material.

 

[Björn Benckert]

A shot from Christoph's test, exposed for the face, each from a different camera.

Im just guessing from looking at you 3 graded jpgs. And I can not say I see any huge differences so it´s very much a guess:

top dragon.
middle alexa.
bottom helium.

But then where is the F65, is it in there?

Not sure that exposure time (seconds) can be related to or translated to illumination (lux), or strength of the light current (Lumens) but I get Your point, Björn. :biggrin:

I´m not really talking seconds here but fractions of a 25th of a second. If you shoot 360 deg shutter on one camera and 90 on the other you get a exposure difference equal two 2 stops. if you ignore the motion blur difference, black shading and a few other things you get as i see it a very clean comparison without altering lens settings or adding Neutral density filters or such. Sure, extremely short or long exposure times are not optimal but to even out these cameras it´s only small adjustments. So doing just that, shooting helium on 90 deg and alexa on 360 deg gives you a base to start from (or I actually think helium at 120deg would be more accurate). Doing so will show that alexa and red have distributed their DR´s differently simply the red "800 iso" midpoint is closer to clip than what is done for the alexa.

 

[Kemalettin Sert]

Im just guessing from looking at you 3 graded jpgs. And I can not say I see any huge differences so it´s very much a guess:

top dragon.
middle alexa.
bottom helium.

But then where is the F65, is it in there?

its on the top
F65
Alexa
Dragon

 

[Anton Esteban]

Great job on matching the shots Hrvoje! I'm about to post my own results soon. I've been able to get a good correction going that seems to translate well between various shots and clips. The problem is I've only been able to make the corrections on the LOG DPX clips and not the RAW files. I don't have access to high speed internet at the moment and I want to refine what I'm doing on the actual RAW clips of this test which I'll hopefully be able to download soon. The good news is, from this initial stab at matching the clips, I feel I've improved my original HELIUM to ARRI look. I've been testing the correction on a lot of other Helium shots and feel that I've been able to improve on color contrast, nuance in tonality, and reproduce more subtle skintones.

The next step for me is how to best translate this correction into a 3D LUT. There are plenty of limitations of using a 3D LUT, which have all been covered in great detail on this post as well as the ARRI Christmas LUT post. That being said, does anyone on this board have any key insights on what is the best way to translate corrections from a color program like Da Vinci Resolve into a 3D LUT? I'm sure there are plenty of Do's and Don'ts on how to do this, especially in regards to preventing banding issues. Of course we all now know that there will NEVER be this silver bullet LUT but I just want to know ways we can get close and just improve on what we have as another tool in our arsenal to make the best images we can.

I personally don't use LUTs for Color Correction at all. I've been much happier with the results I've had when I reverse engineer LUTS, to my liking, in order to gain maximum control of the image in post. I made my ARRI Christmas LUT mainly as a shooting tool to help me see how much I can dig out of the shadows of the Helium sensor in order to protect the highlights when I'm shooting. Below are links to tutorials (unfortunately form a paid site) that have helped me reverse engineer LUTS and make my own.


Matching Shots with LUTS:
https://mixinglight.com/portfolio/matching-shots-with-luts/

Reverse Engineering LUTS:
https://mixinglight.com/portfolio/reverse-engineering-luts/

Building your own Look Up Table:
https://mixinglight.com/portfolio/rolling-your-own-luts/

 

[Alex Boothby]

Can we all take a minute and acknowledge the near motion-control perfection of this anonymous model. This lady certainly knows how to hit her marks and an nail a repeatable expression. Pixel-peepers and skin-tone analyzers rejoice. BRAVO!