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 :)

New Epic-W vs. Alexa Mini test

I think that these types of "stress" tests are not very useful. What is instructive about seeing +5 stops on a face for either system? It's like saying, look how well (or badly) this system performs when you abuse its useful parameters. It's just not all that interesting.
 
A guy in the NFS comments mentions Graeme explaining why this test is flawed:
"Feel free to go on Epic-W's Facebook group and read the thread. The head engineer Graeme at RED explains why this test was flawed."


I can't find the FB group he's talking about. All I see is public posts from this https://www.facebook.com/REDEPICw/

Can anyone point me in the right direction or copy/paste here?
 
Hugh Scully said:
I think that these types of "stress" tests are not very useful. What is instructive about seeing +5 stops on a face for either system?
Click to expand...

Because that way the cinematographer gets to know:

1) How much usable dynamic range he has per particular setup when shooting the most important storytelling subject
2) When each camera hits its limitations
3) What happens when that happens.
 
paulcurtis said:
To me calling it a shootout is asking for trouble because immediately there's a winner and loser.
Click to expand...

Agree 100%. If they want a real shootout, then they need to compare absolutely EVERYTHING. Resolution, sharpness, noise performance, native ISOs, same ISOs, dynamic range, sharpness/detail, ergonomics, workflow, data rates/compression, framerates, etc. etc. etc.

Of course, it'd be subjective in the end because the "winner" would be someone's interpretation of the results. But even that would be much more helpful to DP's trying to pick a camera. Measuring a single metric while crippling one of the cameras is not a good "test."

Alexa probably will still win in the DR department, but not by a stop and a half up top. And I'm okay with that! There's trade-offs for everything and I think it's pretty incredible considering Helium is pushing out almost 5 times the number of pixels.
 
gabrielB said:
A guy in the NFS comments mentions Graeme explaining why this test is flawed:
"Feel free to go on Epic-W's Facebook group and read the thread. The head engineer Graeme at RED explains why this test was flawed."


I can't find the FB group he's talking about. All I see is public posts from this https://www.facebook.com/REDEPICw/

Can anyone point me in the right direction or copy/paste here?
Click to expand...

It's a closed group. Join us :)
https://www.facebook.com/groups/1816671215236474/
 
Hugh Scully said:
I think that these types of "stress" tests are not very useful. What is instructive about seeing +5 stops on a face for either system? It's like saying, look how well (or badly) this system performs when you abuse its useful parameters. It's just not all that interesting.
Click to expand...

Then i guess you haven't graded much of the material, where "cinematographer" completely blows his/her exposure and then blaming DIT as in "he said it was fine". Unfortunately, badly exposed material happens much more often than you think and that is when you're happy to have some extra maneuvering room...
 
Hrvoje Simic said:
Because that way the cinematographer gets to know:

1) How much usable dynamic range he has per particular setup when shooting the most important storytelling subject
2) When each camera hits its limitations
3) What happens when that happens.
Click to expand...

Right, as you say, "usable dynamic range." Once you reach the boundary of what is useful (blown plastic face), then what possible use is it to show results well past that boundary and certainly well past what a competent cinematographer would shoot?
 
jake blackstone said:
Then i guess you haven't graded much of the material, where "cinematographer" completely blows his/her exposure and then blaming DIT as in "he said it was fine". Unfortunately, badly exposed material happens much more often than you think and that is when you're happy to have some extra maneuvering room...
Click to expand...

This.

This test can help cinematographers understand where their dynamic range falls. And shows how each camera breaks up in extreme conditions.
 
Just a quick comment.
The shot where they show clipping in the face.That is never how I would shot. Clearly when they shot that they must have had the goal stops showing that clipping was happening. If I see that on the screen I will be sure to less less ligh hitting the sensor. For this to be a useful test for me I would like to see it properly exposed for the face so no clipping occures and then you bring up the dark areas. That way you can compare both the highlights and the dark areas.

Just my 2 cents.
/Andreas
 
Really strange to not use Helium at it's native ISO.

and this is not just an issue with camera tests... even within peer reviewed literature, it's very frustrating to see critical variables ignored or dismissed.
 
Aaron Lochert said:
Of course, it'd be subjective in the end because the "winner" would be someone's interpretation of the results. But even that would be much more helpful to DP's trying to pick a camera. Measuring a single metric while crippling one of the cameras is not a good "test."

Alexa probably will still win in the DR department, but not by a stop and a half up top. And I'm okay with that! There's trade-offs for everything and I think it's pretty incredible considering Helium is pushing out almost 5 times the number of pixels.
Click to expand...

Hence if it was shot like for like and compared in linear, with no faffing of post pipelines or 'grading' then it would be fairly quantitive.

What is pretty subjective is the rest of the camera and workflow. I'd like to see the Alexa doing HDRx for example. Or framing a 6K crop of an 8K source. Or doing internal wireless focus via an app on an iPhone. Or timelapse. And so on.

YMMV

cheers
Paul
 
paulcurtis said:
To me calling it a shootout is asking for trouble because immediately there's a winner and loser.

The fundamental error of rating the Helium wrong smacks of a lack of understanding.

Then the post pipeline is a bit dubious as well.

If they wanted to do a shootout - both cameras at native ISO. If one camera is more sensitive naturally than the other then stop it down or increase shutter speed but don't mess with ISO.

Then ALL tests should be done in linear space, with no faffing with LUTs and god knows what. Because those LUTs can affect the results hugely.

A shoot out should be quantative and this isn't.

But if you want to test them side by side like this, then it isn't a shootout because you're comparing peoples approaches, techniques, knowledge of their cameras and pipelines of which enormous numbers of mistakes can be made.

I've been through plenty of RAW Alexa footage, i'd still take my Helium over that IMHO

cheers
Paul
Click to expand...

I've never seen Red win in a shootout against Alexa.

Ok, resolution. Maybe.
 
It's just like the movies where the police car always catches the Lamborghini or Ducati LOL
 
Helium in linear
ykiLaI2.jpg


Arri in Log-C (they did not upload arriraw, only DPX export. Log-C seems to be just beginning to blow out, perhaps not all 3 channels are blown yet. Maybe a stop and a third more retained than helium.)
qAryPaq.jpg


Helium, 1280, Low contrast rolloff 4, full debayer log3g10 redwidegamut etc
7ephvGM.jpg

ADc01AZ.png


Arri Log-C -> Rec709 in resolve
s7pKbEo.jpg


Helium with single node grade to lift shadows like Arri
4l685ai.jpg

kgc7riY.png


Helium Eyes
bfsL2m6.png


Arri eyes
uIggexR.png


Clips are A001C0001_161229_R and B001_C018_1229R4
 
Last edited:
Would be great if an experienced Helium Operator was on the test. This also means the Red Team still has a way to go in terms of informing users how to properly use their cameras. If this is the result people are seeing and its the result of operator error, its easy to Blame the operator, But the truth is, with such a new camera and sensor, people without proper guidance will naturally make mistakes, or assumptions based on previous models. Dragon and Epic MX were rated as Iso 800.
 
If you want to rate the camera differently you can just download their raw files. They have a file for every stop or so, compare what you think is fair. Arri without a doubt has a stop or two more range, not sure it wins on any other battleground.
 
Arri, assuming 4000k 800iso like the others. You can't tell but the blownout lights in Log-C and Log3G10 are almost identical, i.e. the following two images have matching blow out points. Check how the shadows fall.
u7GiqsR.jpg


Helium, lowcon rolloff 4, no other grade.
UmFtE0n.jpg

rIFbK8E.png


Correcting tint seems OK to me, lifting shadows should only harm an image so I'd rather they match. Easy to crunch the arri down, harder to lift helium up. I wonder if resolve's SDK is as up to date as RCX?

By the way, all my images are 4k if you open them in another tab.

edit:

Using this shot, where the two lightbulbs look to be blown equally, I lifted the helium to 6400 ISO and then matched the arri using the offset wheel in resolve.

CElveGl.png

Arri left, helium right.
 
Last edited:
You must log in or register to reply here.
Back
Top