New Epic-W vs. Alexa Mini test

[Ben Scott]

Funny that Arri are now promoting this test on Twitter.

 

[Dan Kanes]

The only valid test is the one you do yourself, testing how anything works for you in the conditions and style you use the tools.

By the way, I now love my Helium based cameras, and Arri cameras both. People always love a "war" - we are of course in a time where if you know how to use a camera, and how to light or use the light of the world around you to your advantage, you can make a great image.

I'd also like to predict the world is in for some surprises cinematography wise between now and cinegear in June. Like, as in, no one saw these things coming.

 

[Phil Holland]

Funny that Arri are now promoting this test on Twitter.

They ramped through a bunch of different apertures. So just compare a T4 Alexa shot with a T5.6 Helium shot. Remember to ignore sharpness/color/vignetting though, because of inconsistent T stops.

This will in effect be the most damning thing about this test, because in effect whether aware of it or not, they did indeed test Helium at it's Base ISO/Balanced ISO and those results are not shown in the video which has gained traction on the internet. Which leads to bigger questions.

And in that sampling it shows truly how Helium stacks up against Alexa/Alev III, which tells a different tale than what is shown and written.

 

[Rakesh Malik]

Side note. REDWideGamutRGB/Log3G10 is being "buttoned up" and will be done soon-ish. Proper IPP2 implementation will reveal even more detail out of RED footage due to the new demosiac, which at the moment is supported only in beta REDCINE-X Pro. Helium is still young and things are changing rapidly. None of that is a defense really, but there's more to this sensor than what is seen in the test.

Are you serious? I'm already getting results from mine that are making me lean ever more toward retiring (and once I have the right lens complement for my Red to o this this, selling) my A7R because my Helium images look so much better already.

And there's a new de-Bayer implementation? Whoah. I can't run the beta RedCine on my machine (GPU limitations), so I'm still using the latest released version...

Exciting...

 

[Patrick Tresch]

IMHO Dragon LLO 1600 iso on a Weapon body performs better in DR than Helium and Alexa.

I see also better color rendering in the shadows.

More matured Dragon technology is still the king (until the Helium color science is finished).

Pat

 

[Nick Morrison]

I mean if they shot Helium at 800, then they've skewed the DR a stop. Helium's ISO is really 1600. That would explain the 1.5 stops diff in highlights.

I'm not saying the Mini isn't great, it is...for many reasons.

But they clearly started with Alexa Mini's 800 ISO base iso, and never took into consideration Helium's unique characteristics.

 

[Phil Holland]

Are you serious? I'm already getting results from mine that are making me lean ever more toward retiring (and once I have the right lens complement for my Red to o this this, selling) my A7R because my Helium images look so much better already.

And there's a new de-Bayer implementation? Whoah. I can't run the beta RedCine on my machine (GPU limitations), so I'm still using the latest released version...

Exciting...

Well more accurately, at this point the:
- Transform LUTs
- In Camera Menu Implementation
- REDCINE-X PRO w/ IPP2's New Demosiac
- and SDK

are being buttoned up.

REDWideGamutRGB/Log3G10 as a Color Space and Gamma Curve are locked as far as I know and productions have been using them as a basis for some time.


The New Demosiac is pretty great, it's certainly producing more detail straight away.

 

[Rakesh Malik]

REDWideGamutRGB/Log3G10 as a Color Space and Gamma Curve are locked as far as I know and productions have been using them as a basis for some time.

The results with the preview LUTs Graeme shared already look phenomenal.

The New Demosiac is pretty great, it's certainly producing more detail straight away.

Cool. Just more reason to consolidate my photography + cinematography kit into a single one based on my Epic-W. The only catch is battery life. My Sony A7R great on backpacking trips on that side; I can carry two tiny batteries, and have enough power for a week long backpacking trip. I hope I'll be able to charge batteries most nights on the Himalayas trip this summer... since I'll probably be taking along 4 HyperCore Slim/Slim Red batteries on that trip, and it'll be a bit over two weeks.

 

[Mike Finegan]

Well more accurately, at this point the:
- Transform LUTs
- In Camera Menu Implementation
- REDCINE-X PRO w/ IPP2's New Demosiac
- and SDK

are being buttoned up.

REDWideGamutRGB/Log3G10 as a Color Space and Gamma Curve are locked as far as I know and productions have been using them as a basis for some time.


The New Demosiac is pretty great, it's certainly producing more detail straight away.

Eager to test with this weekend!

 

[Chris Jordan]

Guys... JUST GO SHOOT. WHY ARE YOU ON THE FORUM ARGUING ABOUT... sorry gotta go.... on set actually living

 

[Sean Keefe]

Guys... JUST GO SHOOT. WHY ARE YOU ON THE FORUM ARGUING ABOUT... sorry gotta go.... on set actually living

I'm sure a lot of us will, when we get a camera!

 

[Stacey Spears]

Eager to test with this weekend!

Here is an animated GIF showing the difference between the old debayer and the new. Be sure to open in a new window (Right mouse click over image and select open image in a new tab) so you see it full resolution. Notice how the tip of the wedge on the far right is cleaner. You will also notice a reduction in cross color or chroma moire artifacts.

Because it is an animated GIF, there is dither due to limited color depth of animated GIFs, but you can clearly see how much more detail the new debayer produces.

Here is another example from an old CML image. Big thing here is the reduction in color artifacts on the B&W wedge.

 

[Karim D. Ghantous]

People always love a "war"

No, not really. We certainly do not like bullshit, though. I've heard much bullshit about a lot of aspects of film making and photography over the years. Some of it is actually quite funny.

 

[Aharon Rothschild]

So this isn't going to go over well but Erich Roland posted some very interesting .r3ds of a xyla chart for both Weapon and Helium.
http://www.reduser.net/forum/showthread.php?149166-Weapon-cf-vs-Helium-Xyla-chart
You can read stops by counting every stop down from clipping until the noise floor. Helium showing less dynamic range than Weapon and in line with the conclusions about dynamic range from Oren Soffer's test of Alexa Mini vs Helium.
Erich Roland posted the .r3ds so he can answer further questions but there are .r3d's posted to the thread and they speak for themselves.
First Helium then Dragon

 

[Phil Holland]

I was getting too many messages, but here's sort of the straight comparison if we are taking into account the approximate Base ISO of RED Helium and Arri ALEV3:

Straight Log3G10 to Rec.709 and Log-C to Rec.709, no grading.




And Log because it's a bit clearer:


*edit and because the conversation is continuing online for me, ISO 1600 and 3200 in Log3G10 as well as Log-C, which will indeed look different because they are different specs for Gamma Curves.




Long and the short of it:

ISO 1600 on Helium is already showing the highlight protection. Interestingly if looking into noise analysis you can likely get away with the Standard OLPF up to ISO 2500-3200 if you're playing at the same level of the ALEV3 at ISO 800-1000.

That's apparent if you look at the files in motion in particular, but these are just stills and you can get a feel for that.

Also, I'm comparing this at UHD 4K resolution (Helium downsample, Alexa Upsampled). If we were considering 2K the differences and range could likely be stretched further.

So this isn't going to go over well

The same as I posted. Helium on a fundamental level clips one stop earlier than Dragon.

Also it's important to use the slides on the Xyla as the chips themselves begin to contaminate results.

 

[Aharon Rothschild]

Phil, any chance you could post the .r3ds of your xyla chart? Would be great to have a look as well.

 

[Dan Kanes]

No, not really. We certainly do not like bullshit, though. I've heard much bullshit about a lot of aspects of film making and photography over the years. Some of it is actually quite funny.

For instance ?

I mean, what I'm saying here is that people actually do love bullshitting.

Personally, none of the tests I've seen from anyone really reflect how I use a camera.

Someone in favor of something will skew something one way, someone in favor of another thing will try to skew it back the other way.

Most of this Bs is thriving on a conflict based ideology.

Rather than "be smart, get the shot, make money, move on."

 

[Karim D. Ghantous]

For instance ?

Maybe you're right, at least in part, that people like bullshitting. But they don't know they're doing it, which is worse. A few instances: the idea that an APS-C sensor resolves more detail than a 645 negative (proof was provided, but it didn't accord with reality); the idea that shallow DOF is somehow 'cinematic'; the tradition that beginner photographers ought to start with b&w film (this was before digital); the idea that focal length influences 'compression'; the idea that reflex viewfinders (motion and still) show you exactly what the lens sees (surprisingly untrue).

 

[Nick Vera]

If we're going to be comparing the Helium with the Alexa at 1600 ISO, shouldn't both cameras be at 1600 ISO since one camera is being shifted towards the highlights, while the other one isn't?

 

[Nick Vera]

I was getting too many messages, but here's sort of the straight comparison if we are taking into account the approximate Base ISO of RED Helium and Arri ALEV3:

Straight Log3G10 to Rec.709 and Log-C to Rec.709, no grading.




And Log because it's a bit clearer:


*edit and because the conversation is continuing online for me, ISO 1600 and 3200 in Log3G10 as well as Log-C, which will indeed look different because they are different specs for Gamma Curves.




Long and the short of it:

ISO 1600 on Helium is already showing the highlight protection. Interestingly if looking into noise analysis you can likely get away with the Standard OLPF up to ISO 2500-3200 if you're playing at the same level of the ALEV3 at ISO 800-1000.

That's apparent if you look at the files in motion in particular, but these are just stills and you can get a feel for that.

Also, I'm comparing this at UHD 4K resolution (Helium downsample, Alexa Upsampled). If we were considering 2K the differences and range could likely be stretched further.




The same as I posted. Helium on a fundamental level clips one stop earlier than Dragon.

Also it's important to use the slides on the Xyla as the chips themselves begin to contaminate results.

That's still why I believe the 8k VV dragon sensor was just the better sensor to have. Better highlight performance, had proper 4:3 mode for anamorphics, and you can use full frame lenses properly.