I S O settings on MX

[Ivan Kovax]

@ Ivan: Shoot 800 covers you for most situations - thus easy.
Easy = good.
I like

There's a small catch, though - you won't enjoy your extra hughlight latitude without reclaiming overexposure in post. Thus going redcolor @ 800 in post ONLY means you've lost 1.4 stops of sensor latitude. Then you'd probably been better off shooting 320 to start with... I like the 800 on set-approach, though, as it has enough headroom for the tungsten balancing of the sensor.
(If you don't understand why I am saing this, just shoot 800... - )

Gunelik. Correct me if I am wrong but MX @ 800 ISO = 13.5 stops of dynamic range?

And the sensor itself is rated at 800, right?

 

[Gunleik Groven]

Gunelik. Correct me if I am wrong but MX @ 800 ISO = 13.5 stops of dynamic range?

And the sensor itself is rated at 800, right?

I find it a risky business to be conclusive about total stops as that is a matter of tolerance, but I'd be surprised if you found 13,5 @ 800, more likely to find more stops @ 320.

Remember that RED has showed ca 13 stops from the Epic, which has more bits to record to and oversampling in camera than R1 MX - thus higher precission...

Find the stops you like, but it is often not good to throw away the highlight ones...

Remember that your post and on-set priorities might differ with RAW...

@ 800, you have thrown out 1,4 stops in the top if you have filled the sensor to its max.
the sensors total latitude is allways fixed, independently of where you set the iso, so if you throw out 1,4 stops - you don't "reclaim" them somewhere else...

 

[Gunleik Groven]

I can do this test all the up till the "bring the 320 up with a curve" part. I'll have to have a friend help me with that part I'm not exactly sure what I will be looking for in the blue channel, but I'm terribly interested now. Thanks Gunleik!

You'll be looking at how and if the noise is different from the 320 and the 3200 development, especially in the blue channel, as that's the one your stressing the most with this test.

I actually had to do it twice to confirm what I saw...

 

[Ivan Kovax]

I find it a risky business to be conclusive about total stops as that is a matter of tolerance, but I'd be surprised if you found 13,5 @ 800, more likely to find more stops @ 320.

Remember that RED has showed ca 13 stops from the Epic, which has more bits to record to and oversampling in camera than R1 MX - thus higher precission...

Find the stops you like, but it is often not good to throw away the highlight ones...

Remember that your post and on-set priorities might differ with RAW...

Agreed regarding on set and post priorities.

But just out of curiosity, what is the reason that you say more stops are likely to be squeezed out at 320, being the old rating of the mysterium sensor, as opposed to 800 being the new rating of MX?

Or is this going to be a very long and drawn out conversation if we get into this?

 

[Gunleik Groven]

Not long at all...

Digging stuff out of the bottom is quite simple.
Clipping your highlights is quite simple.
The sensor range is allways the same as it has some physical limits in the top and in the bottom.

Do this.
Expose an image (preferably a greyscale chart) untill (the highest chip) it blows in the highlights @ 320.

Adjust to 800.

You will see that you lose more highlight precission.
Again I am willing to do the beer-crate bet -

If you post @ 800 with redlogFilm you@ll get almost all back.
If you go redgamma, it will be gone.

If you go back to 320 for post, you'll have it all back both with redlogFilm and redgamma, but it will be hard to grade without clipping with the redgamma image, while quite easy with the redlogFilm thingy.

As a sidenote:
If you do it all as a primarygrade from RAW in RC-X, the gamma setting doesn't matter that much... (That takes a bit more of explanation, but the short version is:
What looks good is good...)

 

[Ivan Kovax]

@ Gunleik

Great info! Thanks!

I will do the test myself in the coming week and let you know the results of who owes who beer

 

[Gunleik Groven]

Cool

Mees likes beers... -

 

[Ivan Kovax]

So as an extension, one would expect Epic to behave in much the same way in regards to it's physical lower and upper limits of the sensor - with the major difference coming from its change in bitrate?

 

[Graeme Nattress]

The setting of ISO in camera is irrelevant to the DR of the recorded images.

From B30 onwards, the math always bends potentially clipping data away from clipping, putting it through a special curve that preserves highlights. If I shoot at chart where each step is a stop, and the top chip is clipping, even at ISO6400, you can still just see the tonal difference between the clipped chip and the one just below. At ISO800 that difference is very clear. Dropping to ISO320 you're increasing the visibility of the difference between those top two steps, but decreasing you ability to see into the shadows. There's a strict limit on how far a single curve will bend to give you an image with nicely viewable contrast and show every stop of DR clearly.

REDLogFilm will show you each recordable step clearly, but the image has low contrast and is not what I'd call nicely viewable. It's great for follow-up grading, but not for looking at as-is. But grading is it's primary use, so that's fine. It's a curve for a purpose. You can select ISO320 to ISO6400 and still clearly see each step on the chart.

Graeme

 

[Gunleik Groven]

There is a difference as to gradeability with the different choices and how much of the highlight gets blown.

I won't claim that the info isn't there in the highlights, and this example is extreme, but:

http://reduser.net/forum/showpost.php?p=656660&postcount=89

I see the same tendency in 320 vs 800

Vith redGamma the highlights leave me @ 320, while I have controll over them @ 320 and below.

It is not very difficult tu pull up the lowlights up with a curve.

The truly amazing thing is how ISO now works vs noise.

Curious if this is an MX only thing...

 

[Graeme Nattress]

Remember, there is no one perfect curve that does the most perfect job with every image! All we can do is "keep it raw" and give you the options to develop as your eye sees best.

Graeme

 

[Pablo Villegas S.]

Okay, thanks Pablo. What if I transcode to Pro Res though? I won't be grading a R3D file, right? How can Color be looking at an R3D file if I'm finishing in Pro Res? Again, this isn't the normal workflow...just a time/money sensitive workflow. Am I missing something?

I don't quite understand the entire process, but I guess that is due to my lack of knowledge in the software mentioned. I'm never involved in the post side of things, but I get a lot of questions from folks here in Turkey that don't understand this stuff either. Just trying to wrap my head around it. I think I need to play with the software more in order to fully grasp these details.

Thanks for all the time you are spending responding to my silly questions.

Well, in that case, yeah, you won't be grading r3d files.

But there is a workflow that uses proxies in FCP, and when you send it to color, it automatically links to the r3d files. That's the one I use, then when you render in color you end up with a ProRess file, and sen it back to FCP.

 

[Pablo Villegas S.]

Remember, there is no one perfect curve that does the most perfect job with every image! All we can do is "keep it raw" and give you the options to develop as your eye sees best.

Graeme

Yeah, that's pretty much it.

If the information is there, you can bring it back with Redgamma as well, I've done it, you don't have to use Redlog to do it. I find it easier to work with Redgamma or even Rec709 if it's going to end in HD for TV broadcast, and can get back all the information in the highlights that show in Linear mode or Redlog mode.

 

[Clint Lealos]

Well, in that case, yeah, you won't be grading r3d files.

But there is a workflow that uses proxies in FCP, and when you send it to color, it automatically links to the r3d files. That's the one I use, then when you render in color you end up with a ProRess file, and sen it back to FCP.

Just the type of info I need! Makes perfect sense. Thanks Pablo.

 

[Gunleik Groven]

Remember, there is no one perfect curve that does the most perfect job with every image! All we can do is "keep it raw" and give you the options to develop as your eye sees best.

Graeme

I know... -

Thanks


G

 

[Stephen Strangways]

For projects with less money, time, and personal would benefit from a workflow like: shoot ISO 800, transcode using Redcolor (ISO 800), edit and color in FCP.

This could just as easily be "shoot ISO whatever works for the shot, transcode using Redcolor (ISO whatever works for the shot), edit and color in FCP.

What if I transcode to Pro Res though? I won't be grading a R3D file, right?

It adds a bit of complication, but I do this all the time due to editors working off-site:

1. Transcode your footage to ProRes. I use ProRes Proxy to keep files size small.
2. Edit in FCP using those files.
3. When you're ready to grade, export an XML file from FCP.
4. Use Clipfinder or TP2 to replace all the references to the ProRes files in the XML with references to the proxy mov files from the camera, the ones that are alongside the R3D files.
5. Run the "fix Color looping bug" while you are at it.
6. Save that as a new XML and open it in Color.
7. Where Color sees the poxy mov files in your new XML it automatically links to the original R3Ds instead, and you can do your grading from those.
8. When you're done in Color, render and send to FCP.

Once you figure out the exact steps in that workflow, it only takes a minute or two to do it each time, and you're now grading with the original R3Ds.

 

[Uli Plank]

Regarding ISO, there's a very detailed explanation over at provideocoalition for the Alexa – the same principles apply to any electronic sensor. The ISO you choose is scene-dependent, only clipping is a fixed value.