New Epic-W vs. Alexa Mini test

[John Schick]

But your original point of just moving an image into a different colourspace can be a transform too, especially from a wide space to a narrow display space but there are an infinite number of ways to go from LogC to 709.

So below is a comparison of a technical 709 transform compared to that Alexa one, whilst you can see the same broad movement, the Alexa shows more desaturation in the mids and highlights. They are closely bunched together compared to a simple 709.

https://s3-eu-west-1.amazonaws.com/inventome-media/709.jpg

Hey Paul, that’s an interesting way to view the LUTs.

Just to clarify my position, I'm using ARRI’s colorspace documentation found here:
http://www.centron.sk/Storage/Document/Product/3453-2012-08-02_Alexa_LogC_Curve-Usage_in_VFX.pdf

I’ve rebuilt the LogC2Rec709 transform myself, and all it needs is a color matrix (in linear light) and then an s-curve applied to the LogC data. It doesn’t require any special highlight desaturation, although for my version I did include some gamut-mapping to retain clipped colors.

ARRI also has a tone-mapped version of the transform, which includes some additional luminance compression (possibly what you're seeing on the charts?) and a slightly different color matrix. This is the default version you will find on ARRI’s LUT Generator, etc.

RED’s IPP2 implementation on the other hand, has much more going on under the hood - involving some advanced out-of-gamut correction / saturation adjustments.

Great that this thread is getting to the core of it all. Comparing cameras that shoot raw without separating the image processing from the discussion is like comparing lenses but shoot some of them netted and some with stacks of filters in front and then talk about their characters as lenses. Arri has some creative luts that is liked by a large group of people and they also do pretty good cameras. But when saying that one camera has more dynamic rounding then the other then the discussion is already in the ditch.

Regarding the general discussion, how the images are processed makes a huge difference. I think using something like an ACES pipeline would be a good way to compare cameras moving forward, since each manufacturer’s curves are removed and all the footage is processed equally in scene-referred linear light.

 

[Hrvoje Simic]

Regarding the general discussion, how the images are processed makes a huge difference. I think using something like an ACES pipeline would be a good way to compare cameras moving forward, since each manufacturer’s curves are removed and all the footage is processed equally in scene-referred linear light.

In which case camera judgements would be affected by principles and limitations od IDTs and CTL.

It is not processed equally, IDTs differ for cameras.


Every time you judge a raw recording camera performance on a final developed image you are judging the whole package required to get to that image.


https://en.m.wikipedia.org/wiki/Blind_men_and_an_elephant

 

[paulcurtis]

I’ve rebuilt the LogC2Rec709 transform myself, and all it needs is a color matrix (in linear light) and then an s-curve applied to the LogC data. It doesn’t require any special highlight desaturation, although for my version I did include some gamut-mapping to retain clipped colors.

Hi John,

Yes, you can do all this with Math, but i don't think that's what the Arri 709 LUTs from Arri's LUT builder are doing though. I *think* they are doing more than that for a 'look'. I confess to not bothering rebuilding the same although Nuke has the LogC and WideGamut conversions built in so never really needed to. In their LUT builder online you can choose different types of output.

I could be wrong though but there looks to be some non linear behaviour in that particular transform.

RED’s IPP2 implementation on the other hand, has much more going on under the hood - involving some advanced out-of-gamut correction / saturation adjustments.

I suspect so does Arri but they hide that in-camera. ArriRAW isn't really true RAW, it's processed in camera and LogC is a result of that too - all that gamut mapping is happening in there. Some of it is exposed externally like WideGamut to 709 but that initial sensor RGB to WideGamut i'd assume is fairly similar to RWG mapping.

Regarding the general discussion, how the images are processed makes a huge difference. I think using something like an ACES pipeline would be a good way to compare cameras moving forward, since each manufacturer’s curves are removed and all the footage is processed equally in scene-referred linear light.

That's down to the IDTs and some of them if you open then are very simple and can't possibly be that accurate. It's down to the camera manufacturers to do a good job with that.

Again it's just calling that test a shoot out which irks and the fall out is that A is better than B, when the truth is always more complex than that.

I don't think it's that fair on Red because it propagates myths and of course we don't help with that, looking overly defensive either

cheers
Paul

 

[Rakesh Malik]

Again it's just calling that test a shoot out which irks and the fall out is that A is better than B, when the truth is always more complex than that.

The only thing that they really showed is why Arri it's taken so long for everyone else to catch up to Arri.

 

[Patrick Tresch]

The only thing that they really showed is why Arri it's taken so long for everyone else to catch up to Arri.

Prores in camera? Yes I'ts good that Weapon has it now.

 

[Rakesh Malik]

Prores in camera? Yes I'ts good that Weapon has it now.

That's been a big winner for me also, but I was referring to other factors, like the dynamic range and color and the new color space and gamma curve, and so on.

 

[Steven Abrams]

AFAIK Black Magic also uses a dual gain architecture in its cinema cameras.

Just like you Trump, you have no idea what you're talking about. The bmcc, pocket, and 4.6K sensor all use a dual gain readout from the sensor. The other cameras they have do not.

 

[Kemalettin Sert]

Just like you Trump, you have no idea what you're talking about. The bmcc, pocket, and 4.6K sensor all use a dual gain readout from the sensor. The other cameras they have do not.

your source?

 

[Patrick McGowan]

It would be awesome to see a test comparing the Scarlet-W against the Alexa Mini, just to see how close or far the image is considering the price difference.

RED has consistently been the best choice, at least for me, on many productions. Thank you for that!

One question for the more technically knowledgable though: I've read that the Alexa and the Panavision DXL with the Dragon sensor use a subtractive color space (though I admittedly don't know much about that). Does anyone find this is a discrepancy from an additive color space? Any chance of implementing this on other RED cameras or is there something already available that simulates this?

Thanks

 

[Anton Shavlik]

The subtractive / additive thing is misinformation. All these sensors use RGB pixels. Arri does seem to mix those channels, and favors magenta and teal cyan over pure red and blue. You can see that in these very tests.

Nothing that you couldn't do with a node or two in resolve.

 

[Patrick McGowan]

The subtractive / additive thing is misinformation. All these sensors use RGB pixels. Arri does seem to mix those channels, and favors magenta and teal cyan over pure red and blue. You can see that in these very tests.

Nothing that you couldn't do with a node or two in resolve.

Ok cool, I'll check that out. Thanks!

 

[Steven Abrams]

your source?

I know some people, but it is also well known these are the sensors (or variations of) that bm are using:

"The sensor has two ADC channels per column with one optimized for low light levels and the other optimized for high light levels, enabling high dynamic range data collection in a single image."

"ADC resolution 22 bits (2 x 11-bit)"

bmcc : http://www.fairchildimaging.com/files/cis2521_pb_feb_2015_vf.pdf

pocket : http://www.fairchildimaging.com/files/data_sheet_cis_1910f_0.pdf

4.6K : http://www.fairchildimaging.com/files/ltn4625a_pb_feb_2015_vf.pdf

It's how the bm dynamic range is so good and got compared to Alexa from the start. John Brawley also says its dual gain with 22bit and he's the main beta tester.

 

[Hrvoje Simic]

All these sensors use RGB pixels. Arri does seem to mix those channels, and favors magenta and teal cyan over pure red and blue.

It doesn't .
It only seems that way here.

You can see that in these very tests.

These tests are useless for judging actual color performance and if taken as referent misleading.

Nothing that you couldn't do with a node or two in resolve.

That is also incorrect.

"ADC resolution 22 bits (2 x 11-bit)"

bmcc : http://www.fairchildimaging.com/files/cis2521_pb_feb_2015_vf.pdf


It's how the bm dynamic range is so good and got compared to Alexa from the start. John Brawley also says its dual gain with 22bit and he's the main beta tester.

The "22 bit ADC (2 x11bit ADC) " part = marketing BS

It is 2 x 11bit ADC signals. Which is not the same as 22 bit ADC.
Real 22 bit A/D makes zero sense with these S/N figures as much of it would quantize noise.

By that logic Arri would have 28 bit. It doesn't.
It has two 14 bit quantized signals merged into one with 16 bit processing.

 

[Ugo Del]

It would be awesome to see a test comparing the Scarlet-W against the Alexa Mini, just to see how close or far the image is considering the price difference.

RED has consistently been the best choice, at least for me, on many productions. Thank you for that!

One question for the more technically knowledgable though: I've read that the Alexa and the Panavision DXL with the Dragon sensor use a subtractive color space (though I admittedly don't know much about that). Does anyone find this is a discrepancy from an additive color space? Any chance of implementing this on other RED cameras or is there something already available that simulates this?

Thanks

An interesting test would be the Varicam vs Red (I just found one so far http://www.thebriannguyen.com/2016/03/red-scarlet-w-vs-panasonic-varicam-lt-complete-dynamic-range-comparison/). As a Red owner I would be more worried about the Varicam than the Alexa these days. Cinema had been won by Arri, Red doing great with TV due to strict resolution rules. Though, slowly but surely, more and more shows are shot on Panasonic, some big ones even switching from Red or Arri. The Varicam offers the best of both world plus crazy lowlight. People were unimpressed at first because it doesn't get crazy gadgets but many start to understand how impressive it really is.

 

[Gregory Dillard]

12 pages of which camera manufacturer does what? This always seems like I'm back in high school and EVERYONE wants to be in that group. Take your RED or ARRI cameras and go make beautiful images filled with amazing stories. We're in an era of some amazing technological advancements that are afforded to us inside of these cameras. Enjoy your craft, explore your skills and keep it moving! "ARE YOU NOT ENTERTAINED!" Gladiator

 

[Hrvoje Simic]

12 pages of which camera manufacturer does what?

You can always suggest or start more productive content on a camera forum in a camera comparison thread.

 

[rand thompson]

The Thread was going dead so I thought I would bring it back with fun .GIF I call "Forever Flashing Lights"

[/url]output_0jX51p by rand thompson, on Flickr[/IMG]

 

[Joel Arvidsson]

The Thread was going dead so I thought I would bring it back with fun .GIF I call "Forever Flashing Lights"

[/url]output_0jX51p by rand thompson, on Flickr[/IMG]

I have not watched the video yet but the image your posted looks like Helium had a red light that blinked at this frame. You see that the highlights its is totally different. So I don't know why they want to compare those two images.

Actually there is a better image posted from that test.

 

[rand thompson]

Joel,

Sorry for the confusion. The Gif above is comprised of two different grades of the same Red file listed below:

B002_C012_1229Z5.0000213F.R3D

From
https://www.dropbox.com/sh/b8vqyep7783oe8a/AAAvKWfpFQhGs3PTiHj_ty4pa?dl=0

The frame without the blue and red lights Is the one I graded to remove the blue and red lights just to see how the image would look without them, and it was graded from RedwidegamutRGB/Log3g10 plus IPP2 Low Contrast Rolloff 4.

The Frame with the blue and Red lights present is the RedwidegamutRGB/Log3g10 plus IPP2 Low Contrast Rolloff 4 only and is of the original image . I created the Gif file strictly for fun after seeing the two different grades back to back in Resolve and as basically a "Placeholder" to help bring back the thread, but it was probably dead anyway. Sorry again for the confusion.

 

[John Brawley]

Maybe you've been reading the FAKE NEWS, but since 2012, with their very first camera they have done dual gain architecture.

The 2.5K, the Pocket, the micro and the UM4.6k UM4.6K Pro all have dual gain sensors.

They've had many issues to be sure, but at least get the information right.

JB