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 :)

RED ONE MX and SONY F35

Michael Cioni said:
I'm reluctant to state exactly what ISO I think this sensor really is. However, a conclusion that we did come to today (thanks Deanan, Ted and Rob) is that ISO rating might not be the ideal or perhaps at least the only unit of measurement to best expose your MX footage ... All that is to say it is my opinion that unlike the RED One Mysterium, the best images from the MX require the constant adjustment of ISO rating and relative T-stop settings. This means it is about to become more important that production and post both have a perfect handshake between which meta-data fields to maintain and which to reset.
Click to expand...

Michael,

Could you clarify that thought more? What do you mean by a constant changing of the ISO and T-Stop? Are you talking about within a scene? Between scenes? What kind of changes are you finding most helpful / would you recommend for various scenarios? I'm really interested to hear your thoughts on this, especially since the "handshake" between production and post is so important.
 
The same could be said for film. You have to decide where you want 18% Grey to fall in any given shot,
while balancing that tonality over a scene so that no shot seems out of place (barring relights).
Thats the difference between shooting charts and real life. Obviously, you cant always peg your actors skin at .65ire
or whatever, and roll.

While its hard to compare apples to oranges, the "fruit" of this side by side comparo is pretty impressive.
It would be nice to see these charts after a more thorough grading. Still, its nice to see a pre-graded
image look quite decent post RedcineX. Great job.

With regards to ISO. I know this has come up frequently in the past and is a touchy subject. It would be nice to know
where the "relative ISO" lies even if the sensor behaves best around ISO 320. It feels like the lens length/field of view argument (50isa50isa50) but behaves like a (insert ISO here).
 
David Mullen ASC said:
I know it's hard for some people here to believe, but the competing products from Sony, ARRI, Canon, etc. are all made by decent and talented folks as well, and they all deserve a bit of respect for providing us with nice tools to choose from. There is a level of machismo around here that I find a bit distasteful. I feel like I'm in a college dorm as a bunch of male students watch the Superbowl. If you are not careful, you guys are going to turn Sony and ARRI into the "underdogs" that people will start rooting for to beat the "big guy", i.e. Red.
Click to expand...

amen David... it would be great if the noise was as objective as the tests.

But I also can't blame RED for wanting the M-X to be compared to companies with established product lines. Especially after the past 2 years of (mostly) unfavorable comparisons.
 
ISO/ASA utility

ISO/ASA utility

In the world of the MX, new color science and FLUT™ you have to alter the way you treat nominal ASA ratings. I really don't think its as simple as counting stops over/under 18% grey. The noise floor is so different in character and there is so much more real information in the shadows.

One metric I do respect is signal to noise ratio. IIRC Graeme touted the Mysterium at 66db S/N. My non-scientific intuitive sense is the MX is more than 2 stops "cleaner" to the eye, I am not sure how to measure that in db of S/N.

The dynamic nature of how the FLUT™ control affects shadows, mids and highlights resists a simple declination or exponent conversion to existing scales. The great news is the monitoring is more accurate so shooting by eye in a closer to WSIWYG environment is a more comfortable place.

By the way, build 30 rocks! You can now set up a user button to rotate through several color assists - OFF - EXPOSURE (this is a new false color display option that shows the clip and the noise floor only) - VIDEO (the old false color palette) - FOCUS (blue edge highlight) - OFF

More good news, several of the exposure meters are now based on a true RAW path. Love it.

Cheers - #19
 
Blair S. Paulsen said:
One metric I do respect is signal to noise ratio. IIRC Graeme touted the Mysterium at 66db S/N. My non-scientific intuitive sense is the MX is more than 2 stops "cleaner" to the eye, I am not sure how to measure that in db of S/N.
Click to expand...

Noise is a subjective quality to it so S/N can be misleading as to what's usable vs objectionable.
 
Deanan said:
Noise is a subjective quality to it so S/N can be misleading as to what's usable vs objectionable.
Click to expand...

If you peg the noise level to a common Std. Dev., like Graeme suggested, then, for the purpose of a comparison of DR of two cameras, it is not subjective. I haven't done the maths, but to my eye it looks like the DR on MX is a stop or two higher than on the F35.
 
Pawel Achtel said:
If you peg the noise level to a common Std. Dev., like Graeme suggested, then, for the purpose of a comparison of DR of two cameras, it is not subjective. I haven't done the maths, but to my eye it looks like the DR on MX is a stop or two higher than on the F35.
Click to expand...

That's precisely what is misleading, take the std deviation that is the same between different sensors and the quality of the noise can differ dramatically from harsh to organic to patterned to chromatic to fin to broad to how it distributes over luminance, etc.

In going from Mysterium to Mysterium-X, tuning the pixel and the sensor for S/N is only a small part of getting the noise performance improvement we wanted. Sometimes you have to put the technical specs aside :)
 
Last edited:
Noise tolerance is subjective and different noise characteristics impact particular people's awareness idiosyncratically. Agreed. What I want to be clear on in my description of the noise floor on the MX sensor (as currently tuned) is that for me it offers 2 full stops more shadow latitude than the original Mysterium did. Will everyone who looks at the footage agree with me, of course not.

To my sensibilities the character of the noise in the MX is more of a pleasing texture than a liability. IMHO, 1,000 ASA on MX looks cleaner than 320 on the Mysterium due to the smoother look of the noise. :party:

Cheers - #19
 
I find this test a little misleading. If I had to guess looking at the histogram I would say that the RED footage got a nice toe curve in the shadows to crush all the noise out.

Here is my 1080p PSD match-graded so that the histograms look relatively similar:
10801280matchedcopy.jpg

http://www.sendspace.com/file/1omc3b

and mind you this was done in 8bit. And RED gets a little extra noise reduction from being run through wavelet compression.



The mx is definitely better. But not nearly as much as an unprocessed image would you to believe. Of course you're going to sacrifice a little DnR as a consequence but it seems like the MX natively does it by design. I'm not entirely sure how people are making comparison DnR conclusions from this test. I don't see it clipping anywhere.

You really need two linear images to make an empirical decision. There's a lot of apples and oranges. It's useful. And I think it's informative. But I also think people are perhaps drawing too many conclusions.
 
Can't download from your link, so can't see it - but did you also match saturation as that's another noise-maker.

Graeme
 
(Link replaced)

No I just did a quick curves adjustment on the grayscale and tried to approximately white balance the two more accurately as part of the curves. I just have PS here at home so I was trying to avoid as many operations as possible on the image since each one would be in 8bit space.

The color in the F35 is more muted in my quick comparison. So it's true that it gains a bit of an advantage from that.

My comparison is equally as misleading as the unmodified. I don't know what sort of gamma and rolloff is happening in the RED processing nor the slog exactly so I just sort of did a quick match by eye in the gradient.
 
Yes, thanks for the update link. Not a bad match at all - had to boost saturation a tad on the F35 to bring it up though. Is it just me or are the greens way off on the F35 though? Red and blue seem reasonable, but green is just bonkers.

Graeme
 
Thanks for your work here Gavin. Helps a lot.
 
Graeme Nattress said:
Is it just me or are the greens way off on the F35 though? Red and blue seem reasonable, but green is just bonkers.

Graeme
Click to expand...

Yeah don't know what's going on in the greens. Not sure if it's in the footage or something I did in the curves/levels.
 
Gavin Greenwalt said:
I find this test a little misleading. If I had to guess looking at the histogram I would say that the RED footage got a nice toe curve in the shadows to crush all the noise out.

Here is my 1080p PSD match-graded so that the histograms look relatively similar:
10801280matchedcopy.jpg

http://www.sendspace.com/file/1omc3b

and mind you this was done in 8bit. And RED gets a little extra noise reduction from being run through wavelet compression.
Click to expand...

Thanks for posting that, Gavin. I thought the same thing. I dragged the files into Photoshop and applied a simple curve using the eyedroppers. With a curve applied the F35 looked a lot better, but still seemed to have more noise than the MX.

-Fran
 
Would a good comparison be use (in 16-bit RGB mode in PS) the black, gray and white droppers in the Levels adjustment tool, on the respective areas of the chart, then also match the saturation (using the Hue/Saturation adjustment tool)?

That is what I did to the 1280/1250ISO 1080p files. It didn't seem to set the WB, or at least I see a yellow/green hue in the F35 grays, but it did fill out the histogram. Also, I suppose there may be variations due to noise in the images if the pickers only select a single pixel. For saturation I may have been to crude, with +40 on the F35 and -40 on the M-X. I think I should have gone a little lower, then again some of the colors do match closely.

If this is a decent method to compare then here are 100% TIFFs:

M-X ISO1280 (~RED saturation)
F35 ISO1250 (~RED saturation)

M-X ISO1280 (~F35 saturation)
F35 ISO1250 (~F35 saturation)

Gavin, what ISO was your comparison based on?
 
You must log in or register to reply here.
Back
Top