EOS20D CR2 format and camera settings

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Jim Garrison - 11 Dec 2005 06:01 GMT

I understand that 'raw' format is supposed be just that:
48bpp of the raw levels recorded by the camera sensor,
totally unprocessed.

However, the ISO setting obviously affects the captured
data, so *some* processing is going on, even with
raw files.

What I'm looking for (if it exists) is a chart showing
which camera settings actually affect the raw numbers in
some way, and which are stored as metadata.

My guess is that the ISO setting results in different levels
of amplitude scaling, but all other image settings (WB, color
space, contrast, sharpness, saturation and color tone) are
merely recorded as metadata in a .CR2 file and don't change
the recorded pixel values.

Correct?

Signature

Jim Garrison (jhg@acm.org)
PGP Keys at http://www.jhmg.net RSA 0x04B73B7F DH 0x70738D88

Reply to this Message

Scott W - 11 Dec 2005 06:10 GMT

> I understand that 'raw' format is supposed be just that:
> 48bpp of the raw levels recorded by the camera sensor,
[quoted text clipped - 15 lines]
>
> Correct?

The camera changes its analog gain stages when you change the iso. I
believe that is the only setting that will affect the raw data, other
then thing like sutter speed.

Scott

Reply to this Message

Stephen M. Dunn - 11 Dec 2005 18:06 GMT

$I understand that 'raw' format is supposed be just that:
$48bpp of the raw levels recorded by the camera sensor,
$totally unprocessed.

36, actually; the sensor works at 12 bits.

$However, the ISO setting obviously affects the captured
$data, so *some* processing is going on, even with
$raw files.

ISO is analog, controlled by the gain level of the amplifiers on
the chip. Some sources claim that 3200 is actually the same as
1600 but with digital modification; Canon, to the best of my
knowledge, has never released a statement about whether or not
this is the case.

Signature

Stephen M. Dunn <stephen@stevedunn.ca>

>>>----------------> http://www.stevedunn.ca/ <----------------<<<

------------------------------------------------------------------
Say hi to my cat -- http://www.stevedunn.ca/photos/toby/

Reply to this Message

JPS@no.komm - 11 Dec 2005 22:39 GMT

>ISO is analog, controlled by the gain level of the amplifiers on
>the chip. Some sources claim that 3200 is actually the same as
>1600 but with digital modification; Canon, to the best of my
>knowledge, has never released a statement about whether or not
>this is the case.

It doesn't matter what Canon says; the RAW data has no unique
information for the least significant bit in the top 1 or 2 ISOs on the
Canon cameras. The 20D has all even numbers at ISO 3200 in the RAW
data, except for a few hundred bad pixels that are odd when averaged
from their neighbors. The 10D has patterns of 1s and 0s in the least
significant bit that are not possibly the real digitized data, at ISOs
1600 and 3200:

http://www.pbase.com/jps_photo/image/38841732/original

Signature

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>

><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

Reply to this Message

JPS@no.komm - 11 Dec 2005 22:34 GMT

>I understand that 'raw' format is supposed be just that:
>48bpp of the raw levels recorded by the camera sensor,
>totally unprocessed.

Unfortunately, the RAW data is actually only 12-bit for most DSLRs. It
is a 12-bit value that more or less represents the amount of charge in
the cells, in a linear proportion. Not totally true, however, as the
data is offset on many cameras (black is much higher than 0 in the RAW
data).

>However, the ISO setting obviously affects the captured
>data, so *some* processing is going on, even with
>raw files.

For most ISOs, the number is proportional to the amplification used
before converting the signals to numbers. Cameras may use
multiplication of under-exposed data instead of purely amplification for
their highest ISO(s). The Canon 20D's ISO 3200 is really 1600 with the
RAW numbers doubled, and a stop of under-exposure; not very useful if
you shoot RAW, as setting the camera to ISO 1600 and -1 EC yourself will
leave you a stop of headroom that is clipped when the camera is set to
ISO 3200.

>What I'm looking for (if it exists) is a chart showing
>which camera settings actually affect the raw numbers in
[quoted text clipped - 5 lines]
>merely recorded as metadata in a .CR2 file and don't change
>the recorded pixel values.

>Correct?

Pretty much. At any given ISO, the only factors that affect the RAW
data directly are the shutter speed and the f-stop, and the amount of
light coming through the lens. All those other things are just
metadata. There are possibly some exceptions, however; at least one
DSLR (Nikon D2X) purports to use the WB setting so that the red and blue
channels have independent amplification, and therefore, virtually
independent ISO.

Signature

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>

><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

Reply to this Message

JPS@no.komm - 11 Dec 2005 22:49 GMT

>Unfortunately, the RAW data is actually only 12-bit for most DSLRs. It
>is a 12-bit value that more or less represents the amount of charge in
>the cells, in a linear proportion. Not totally true, however, as the
>data is offset on many cameras (black is much higher than 0 in the RAW
>data).

I forgot to mention why it is only 12-bits. Each pixel only measures
one color band, red, green, or blue, depending on the filter over it.
The filters are over the pixels in a pattern like this:

R G R G R G

G B G B G B

R G R G R G

G B G B G B

R G R G R G

G B G B G B

The full RGB pixels are generally derived from the RAW data by a type of
interpolation called "demosaicing".

Signature

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>

><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

Reply to this Message

Jim Garrison - 11 Dec 2005 23:31 GMT

Thanks! That was more than I asked for, but just what
I needed to know.

>>Unfortunately, the RAW data is actually only 12-bit for most DSLRs. It

[snip]

Signature

Jim Garrison (jhg@acm.org)
PGP Keys at http://www.jhmg.net RSA 0x04B73B7F DH 0x70738D88

Reply to this Message

Paul Furman - 12 Dec 2005 05:54 GMT

>>My guess is that the ISO setting results in different levels
>>of amplitude scaling, but all other image settings (WB, color
[quoted text clipped - 11 lines]
> channels have independent amplification, and therefore, virtually
> independent ISO.

Ah, that's the same WB setting that caused the encryption controversy,
thanks for clarifying.

On a similar note, color filters (or polarizer, etc.) on the front of
the lens are another way to change the raw data.

Reply to this Message