RAW and bit depth...a couple of questions.

[Matthew Scott]

Hey guys,

I've read some confusing things lately about best kelvin's to shoot at, best ISO's to shoot at, etc. If these factors of my recoding are merely meta data, then why would it be better to shoot one ISO over another? Or why would it be better to shoot one colour temp over another? (other than obvious exposure reasons of course). Have I missed something here? Can someone please explain it to me? Or better still, rather than posting "ISO and WB are simply metadata, you can change in post so don't worry about it"...I'd rather someone clarify what I'm confused about. I swear I'm not making it up...I've read on here numerous times about how you should shoot at 5000k because that's the sensor's native WB....or.....you should shoot at 800ISO because that's the sensor's native ISO. But WHY should I, if I can change it in post? I'm not annoyed, even though I might sound it....just curious and willing to learn.

Secondly, regarding bit depth. 16 bits of colour and light information is ridiculously awesome and I actually can't get over how awesome it is. But one thing that gets me thinking is this.....If I shoot a white wall with a graduating light over it (something the 8bit world of cams can't do..ahem......c300), grade it to my liking in a 32bit colour space program, and then export it to an 8-bit h264 or DVD or Bluray......is there any advantage to having the 16-bits in the firstplace? I'm talking about banding here....or posterization. Sure, I will be able to pull better keys and have the obvious benefits of 16bits as I grade...but how different will the final 8bit encode be compared to if I shot it in 8bit to begin with? I plan on doing some tests when I get my Scarlet, but this also fascinates me.

Appreciate you reading this long winded couple of questions....

Matt

 

[Dylan Macleod CSC]

Hi Matt
To address the first part of your post - not 100% sure what your question is. I think what you are saying is - why bother shooting anything other than native (daylight balance at 800ISO)? Because you want to craft the best image possible for the look you are going after. If you are shooting in tungsten light - you don't want to shoot at 5000K because obviously your image will be very warm. And for ISO - I use different ISO's as insurance. If I'm concerned about highlights, I'll shoot at 800ISO. If I am concerned about noise in the shadows - I'll rate at 320. But I am always aware that the sensor is native 800 - so I will constantly check RAW. Although you have the flexibility to rate the camera as you like - you might make certain decisions that WILL affect your RAW. Extreme example - you rate the camera at 2000ISO - you adjust your aperture because the monitored image looks really bright - you've now baked in a darker exposure by adjusting the aperture. So although adjusting the ISO does not affect the recorded RAW image - the decisions you make by viewing that image will certainly affect your RAW. Hope that makes sense?

 

[Hrvoje Simic]

Hey Matt,
Adressing the second part of your post - regarding 16 bit advantage. First off, banding your refer to with a white gradient is just one example, the advantages go way beyond that. Second, grading in 32 bit space is not a guarantee you will be able to grade the image to your liking, if it has been recorded in 8 bits. 32 bit based grading tool provides more precision in altering the captured image, but if the image has been captured with 8 bit precison, it does not come with a lot of data to work with.

Imagine having a screen with a very low resolution, on which you want to show a circle. Now imagine how small you want those pixels to accurately represent that circle. The smaller the pixels get and the more of them there is, the more precise the shape of the cirlcle will be, because each pixel will represent a more accurate value, a dot of that circle.

The same principle works in bits representing colour/luma values. More bits per pixel allow more precise steps. More precise steps allow more accurate digital representation of what was in front of the lens. With that, post work tends to "stretch" those quantization steps and with fewer "steps" (less bits used) image falls apart sooner.

Theoretically, 8 bits allow 16.7 M steps (256 per channel), 10 bits allow 1 B steps (1024), 16 bits 281.4 Trillion steps (65536). Provided it's a full 16 bit pipeline from sensor A/D to grading suite, this is 16.7 Million times the 16.7 M colour gradations used in 8 bit precision.

For an end user or someone not fully involved with post or interested in numbers, this means more data to work with (larger palette to paint with), more natural looking imagery, and less loss in compression of the final master.

 

[Rob Ruffo]

When they say "shoot at" x=ISO what they mean is use a light meter and set it at some ISo to judge light/iris/ND needed. (and then the camera too, only so that monitoring does not look weird. But it's just a preview of what the post approach can be ). You can, of course, do whatever you like in post, and yes ISO is just metadata (which means it's the same to change it in camera when you shoot as it is to change it afterwards via RedRaw settings in whatever software you use for post.) Same goes for white balance. You'd want to set it somewhere more or less right on set, otherwise it would be hard to see the color-mix of you light set-up (for example, is there some blue light leak from a window fighting a mostly tungsten scene? Are LEDs looking too green, etc.)

 

[Michael Lindsay]

Secondly, regarding bit depth. 16 bits of colour and light information is ridiculously awesome and I actually can't get over how awesome it is. But one thing that gets me thinking is this.....If I shoot a white wall with a graduating light over it (something the 8bit world of cams can't do..ahem......c300), grade it to my liking in a 32bit colour space program, and then export it to an 8-bit h264 or DVD or Bluray......is there any advantage to having the 16-bits in the firstplace? I'm talking about banding here....or posterization. Sure, I will be able to pull better keys and have the obvious benefits of 16bits as I grade...but how different will the final 8bit encode be compared to if I shot it in 8bit to begin with? I plan on doing some tests when I get my Scarlet, but this also fascinates me.

There are huge advantages to starting in 10bits+ for 8bit delivery. I can't believe Canon don't allow 10bits out of there c300.

Ocassionally a 8bit delivery pictures will band in a annoying way. If you have started at a higher bit depth then you have the opertunity to dither down to 8bit which is a slicker way of diminishing banding in a final image (rather than just adding noise).

regards

M