5DMK3-bitrate ?

Started by jafa, October 20, 2014, 02:00:15 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

jafa

What is the bit rate coming out of the 5D3 with MLRAW enabled  is it 8 , 10 or something else or maybe it dose not apply to raw ?

Cheers
Jafa 

Walter Schulz

Sorry, please edit your question because - for me - it is completely incomprehensible. What does 8, 10 mean to you? Bit depth? SD-class? MByte/s?
If you have troubles because english isn't your native language: Feel free to add some text in your native tongue.

chmee

? Bitdepth is natively 14bit - the "datarate" you can calculate yourself.

width[px] * height[px] * [framespersecond] * 14 = datarate[bit/s] / (2^23) = datarate[mbyte/s]  (video only)
[size=2]phreekz * blog * twitter[/size]

jafa

Color Depth ! and dose this apply to raw.

chmee

as i said, the valuerange is nearly 14bit - so you can say simply say, the colordepth is also. because videoapps work usually with 8, 10, 12 or 16bit, most of the time its converted in another bitdepth. most of the converterapps try to hold a higher bitdepth instead of going to 8bit/channel.
[size=2]phreekz * blog * twitter[/size]

jafa

Thanks chmee ive been looking at other newer cameras such as the sony a7s and gh4 and was wondering just how much color depth difference there was between working with raw and the smaller 8-bit coming out of thoes cameras.

chmee

imho if you re spot on with your recordings (and kind of s- or c-log inside the body), 8bit/channel are fine. guys like andrew reid (eoshd, sony a7s) say, its raw-quality without coping with the converting-thing. me as well, i still use the simple h.264 output from 5diii, because of the simple workflow. quite one year of work is done on ml-raw, mlvfs is an approach getting faster workflow without losing the benefits and time.

so, my oppinion. if you've got a canon with magiclantern, go on. if you re thinking about spending money, wait or look on other bodies, yes.
[size=2]phreekz * blog * twitter[/size]

dmilligan

It's complicated to talk about the "color depth" of raw data because of the bayer matrix.

Usually, color depth refers to the total number of bits used to represent a color in a single pixel. In a typical RGB image, this might be 8 bits for each of three color channels giving a total of 24 bits => 24bit color or 8 bpc (bits per channel). So the total bits per pixel and the bits per channel are different. With raw data each pixel only has one color, so the bits per pixel and the bits per channel are the same.

Therefore, trying to compare the "color depth" of an RGB image to the "color depth" of a raw bayered image is like trying to compare apples and oranges. A "14 bit" (per channel) RGB image has 3 times more data (potentially) in it than does a "14 bit" raw bayered image. In this sense, you could think of the bayer matrix as being a sort of lossy compression in the analog domain. With raw, we have a possible representation of 3 * 2^14 different colors, but we also don't have complete information for each pixel, the other colors must be interpolated from neighboring pixels.

chmee

but after interpolating there are 14(high)bit RGB data available. no one works with pure bayer-data. nearly every film-sensor extracts his data from a bayer-sensor. so, @dmilligan its not that much apples and oranges, because they re all 2/3 synthetic..
[size=2]phreekz * blog * twitter[/size]

dmilligan

Quote from: chmee on October 20, 2014, 06:37:40 PM
14(high)bit RGB data
This merely proves my point. You can't really say how much exactly RGB data is truly there. This is because of the interpolation. Some of the data is "synthetic" as you say. So what percentage of that synthetic data is "true" and actually adds to the "color depth" (in an information theoretical sense) and what percentage is just false data/noise? It's impossible to precisely quantify, it depends on many things: the nature of the scene, the correlation between color channels, the type and quality of the demosaicing algorithm, etc. Hence: apples and oranges.

chmee

im with you about the nature of synthetics. but that applies to all footage you've seen, red or arri or ikegami or sony.. all of its data is 2/3 synthetic and not exactly. mathematically you can say, independently from debayer-algorithm, the calculated values wont be under or beyond the valuerange of the "real" values. i striked the 14bit, because only the dng-converter does that, only photoapps/acr and resolve can cope with it. videoapps are not interested in this bitdepth. but finally 14bits is, what comes from a canon-sensor and you can't map the native valuerange into 13 or 12 bit.

being "exactly" is not about being "right" in terms of imagequality.. 
[size=2]phreekz * blog * twitter[/size]