MX vs Alexa

[Sanjin Jukic]

You can't be serious.

Sorry but just contrary,

I'm a serious man :laugh:

So... you want people to compare production stills from a 5D with ProRes from an ALEXA.

I just compared production stills I pulled off the web from David Fincher's "Seven" (shot with a Panasonic Lumix) with 35mm stills from Lawrence of Arabia. From that, I have determined that the 5D still looks better. :laugh:

Sanjin - you *really* don't like the ALEXA. Yet you have never shot with, never done post with it, and don't have any actual real-world experience with it. That, combined with ridiculous comments like this (really) make it hard to take your input seriously on this particular subject!

And you do have serious input, so... go out and do some of your own real-world tests and I bet you'd have some great results to post that would be very interesting for the community here!

With a smile,

Lucas

Lucas Wilson
------------
ASSIMILATE, inc.
LA, CA, USA

Lucas,

we don't need one more battle here about this topic.

Frankly I did comparison about the images that were available online and also that were directly related to the topic.

Nothing bad with it and only you could like it or not.

Also I know that you were grading those Alexa shots and can understand that of course you are emotionally involved with it.

If I want to replay on your joke will say something like this:

Oooooohh, sorry but I didn't know that David Fincher shot "Lawrence of Arabia" :laugh: and David Lean shot "Seven" :laugh:

just because I was confused and stupid to mix some of these production photos :laugh:

Then also totally it's not true that I don't like Alexa.

Think about that I'm one of the guys who posted a lot of relevant quotes, images and links about Alexa more than any other RED users.

If I think about Alexa with sort of criticism doesn't mean that I do not like it.

Criticism is a part of so called German Idealism, a philosophical thinking that started I. Kant, J. G. Fichte, F. W. J. Schelling, and G. W. F. Hegel,

then followed by Nietzsche and finishing with Phenomenology · Hermeneutics · Existentialism from E. Husserl and M. Heidegger.

Also still remains even today in traces of thinking from Peter Sloterdijk.

So that's a part of wider historical tradition and culture of thinking here in Europe.

Also don't forget France, UK and Italy with their critical involved philosophers during 20th Century.

Nothing bad with a little slice of criticism here just to encourage a healthy bit of skepticism over published and marketed specs.

 

[Dan Hudgins]

9-to-4 vs. OLPF

9-to-4 vs. OLPF

Though - as a slight comment to your CML post - I cannot really see evidence that there is applied heavy sharpening in the Alexa. So far.

If you can see the Berlin (?) same demo reel projected on a 1920x1080 projector you will be able to judge that better. In a fully un-sharpened film scan there is some blur of sharp edges as the film acts as a low pass filter.

What method the projector used to resize or how the reel itself was resized is unknown to me. However, less sharp source images might resize with less aliasing showing later. When RED ONE first came out some people were having aliasing problems making DVD because they did not blur the images before down sizing them.

With any digital image size reduction of an already somewhat "crisp" image can make the aliasing worse (become visable above the threshold of consciousness).

The Alexa seems to be using a 9-to-4 pixel reduction to get to 1920x1080 for the prores output.

The question is, does the OLPF have enough blur to avoid all aliasing after that reduction. Because I did not see it projected on a 1920x1080 projector, I can just guess from the absence of "fuzz" at the edges of lines that there is not a great deal of anti-aliasing blur left in the images that were shown, since I did not do the post or see the original prores full size on a 1920x1080 monitor, people with access to the camera can say more, and use a very sharp wide angle lens at prime stop to pan over buldings with many sharp lines to see if any runners develop on viewing of the prores files after contrast increase in grading, trans code for broadcast, etc. Such artifacts are f/ stop dependent, and as can be seen in some HDSLR footage, if you are wide open or stopped down you can maybe not see aliasing, if you are at prime stop you may under some high edge contrast conditions.

If you process the full 2880x1620 data, then the OLPF radius of blur can be seen directly, and any needed gauss blur can be added before additional reduction in size. So far as I know, the 2880x1620 data has not be published as a set of 48bpp TIF frames for people to look at in motion?

==

The question of how sharp ANY digital image should be made comes up, as for the most part anything over 0.707 of the pixel width or height would be in part more aliasing than true image detail at higher MTF.

To avoid aliasing the line witdth should be more like three pixels, so the "resolution" of a 1080 image should be about 360 lines per image height, which is about the radius of blur in a 4th gen 35mm print if you use the whole point spread as the diameter (radius varies with density and could be larger than midtone in the highlights and shadows through negative to positive to negative to postive printing).

Its good to use 5K post to filmout or 5K projection to avoid aliasing. If you push a 1080p image line width to just one scan line you can get an image that looks "as sharp" as a 5K image with 3 lines blur for the line width, but the 5K image will not show much aliasing and the 1080 image could because there is not enough edge blur to avoid it, that is why CGI images get anti-alias blur added, otherwise you can/may see the "runners" etc.

The question is now, does anyone care about "runners" or remember what "film-to-film" printed images looked like? Does aliasing matter any longer in the all digital image world?

==

As for letting people do their own de-Bayer on Epic and Alexa footage, there are some effects you can get with various filters that change the ratio of red to blue and red/blue to green.

With the "factory" de-Bayer you do not have enough control to re-balance all the de-Bayer settings to take advantage of using filters that would otherwise cause problems in the compressed images by putting the white blance off center.

When you have full true RAW you can do anything you like with the white balance (through the use of optical filters etc.) since it does not induce problems with the de-Bayer settings that are normally out of reach of the end user. These optical adjustments trade EI/ISO for S/N ratio and dynamic range, so are not within the "one-size-fits-all" scope of the OLPF+IR/UV cut filter picked by the camera maker.

Also for archive, its best to save the actual sensor data in an open format, later advances in de-Bayer can be used without licence issues or baked in artifacts. 100 or 1000 years from now an open published format like uncompressed Bayer DNG will still be an open published format and accessible to anyone who needs to work with the original camera files.

I don't see how it hurts the camera maker if someone comes up with a "better" result for their own use? People are not going to forced to use any of the other options that may come on the market or be free, for Red, there is no revenue loss from letting people develop alternative de-Bayer solutions since their de-Bayer solution is free anyway?

==

Anyway, with better and uncompressed data people can say more about what the camera (OLPF+SENSOR) can do and less about processing, compression, and re-sampling artifacts, so I am sure everyone is looking forward to seeing what comes out of the camera(s)...

 

[Sanjin Jukic]

Finally I've got some of Alexa ProRes 4444 short color chart tests from a friend that want to stay here anonymous.

Then I immediately compared with MX Light Digital color charts tests that are available here.

I drop Alexa PR4444 to FCP timeline and export tiff one frame file.

Took screen grab from videoscope RGB parade to show it here.

Also resized 4K HD RED PSD/tiff to HD 1080 tiff file.

Then import those two HD files in AE older version 3 and then clicked to blue channel to see what is going on there.

The results are below at the images.

Then make a conclusion for yourself.

Also this is a scientifically free test that only relies on the samples were found on the Internet and nothing more.

Alexa REC709 ProRes file blue channel 62% magnification (Don't get confused here because Alexa shot RED-DSC Lab color chart).

Alexa REC709 ProRes file blue channel 100% magnification (Don't get confused here because Alexa shot RED-DSC Lab color chart).

Alexa REC709 ProRes file videoscope parade in FCP (Don't get confused here because Alexa shot RED-DSC Lab color chart).

MX @ ISO 800, T5.6 from Light Iron Digital color charts tests MX vs. Sony F35, here is shown blue channel in AE3, 62% magnification.

MX @ ISO 800, T5.6 from Light Iron Digital color charts tests Mx vs. Sony F35, here is shown blue channel in AE3, 100% magnification.

MX @ ISO 800, T5.6 from Light Iron Digital color charts tests MX vs. Sony F35, videoscope parade in FCP.

 

[Gunleik Groven]

Gee, I wish I had one of those DSC RED books, they look pretty usefull.

But I think I can tell that bluechanell noise is not a significant issue on either of these cams, from this comparison.

Haven't gotten around to that yet. Thanks.

 

[Jarred Land]

lets try and stay on topic of the test and avoid speculation of recent events. Moving todays segment to offtopic.

 

[Adam S]

Gunleik you know there is a difference between ArriRaw and Log C, right?

I was under the impression that Arri hasn't finalized the Alexa flavor of ArriRaw, let alone enabling it in camera, even for VIP's. Are you a Super VIP that has access to an ArriRaw enabled body?

 

[Gunleik Groven]

I know there is a difference...

 

[Gunleik Groven]

OK guys.

There's gonna be a delay... -
Before I release images and analysis...

And here's why:

These two cameras are in each their ways the two most sexy pieces of digital cinema camera equipment generally available in numbers these days.

The test is done to find out how they work and how one can apply both or one of them well for different situations and shooting conditions.

The first day was intended to "break" the images. See what happens when they go "bad".
That is usesfull for me and my employers, as it tells us more about what works well on them and why.

First of all I got some surprises, that I need to re-check and reshoot to verify in addition to discussing them off list with others. And the surprises made me change what I wanted to check. I need additional info before it is a point to release them.

Secondly, these images just show off my capacity to break cams! -
The destruction of beauty!

To me, that is informative. To release these images without context would only generate a big splash and possibly a silly, heated debate, where the primary issue would be lost: These are todays top digital camera choices for most of us. This is the context I want the images to be analyzed from, NOT from: "Gee, look how ugly!"

So, I cannot release the test bit by bit as I had initially planned, and I think I might add a third day of shooting, where I try to apply the knowledge I get from the first two days. Together with DP's and post people. Just to see how beautifull they can be. Because they are.

My initial tests were not intended to show off that, and they don't...

But I see pretty clearly that releasing them out of context will be counterproductive. Even though I always want to pixelpeek myself on the source materials for tests, and that is the main idea as to the concept of the test. And rest asured, it still is... -

To be clear, this is NOT due to bias towards one of the other of the cameras. I made both look ugly! (What an achievement! Having all this loveably image-making-kit in my hands, just to create uglyness is a sacrilege of sorts. LOL).

Just to clearify...

Cheers!

 

[David Rasberry]

Be interesting to know where the limits are for sure Gunleik. When you post the pretty stuff how about posting some of the uglies too with exposure data so we can learn what not to do.

 

[Gunleik Groven]

Be interesting to know where the limits are for sure Gunleik. When you post the pretty stuff how about posting some of the uglies too with exposure data so we can learn what not to do.

I'll def post both...

 

[Subhadip Sen]

Indeed, it is the ugly footage which could be more revealing. Look forward to the test results.

 

[Adam S]

Everyone can now officially record to ProRes for commercial work. Arri released the firmware update to enable that feature yesterday! Now we can all do our own tests and draw our own conclusions, if we can get a hold of an Alexa....

 

[Jeff Kilgroe]

if we can get a hold of an Alexa....

That's the key right there... The shoot I'm starting tomorrow was supposed to have an Alexa available to it. The director decided to bite the bullet so we could all play with it (still shoot RED MX as well), but that fell through. Turns out he was lied to and can't really get the camera. Bummer because it would have been perfect timing with SUP 2.0 and onboard ProRes...

I'll shoot with an Alexa sooner or later. I'm starting to wonder if I should take bets on what happens first; I get my first Epic-X or I shoot with Alexa. Hmmm... :biggrin5:

 

[Gunleik Groven]

At the moment Alexas are more plentifull than Epics, at least here... -

 

[Daniel García]

Has anybody tested the rolling shutter artifacts on Alexa? Does it have the half-frame strobe problem?
I checked a quick Google search and found Arri confirming it has an 8ms readout rolling shutter.

 

[Gunleik Groven]

There are rolling shutter issues on the Alexa. Not only with strobing lights.