Sign in with
Sign up | Sign in
Your question

Are we ignored regarding dynamic range?

Tags:
Last response: in Digital Camera
Share
Anonymous
December 4, 2004 11:33:19 PM

Archived from groups: rec.photo.digital (More info?)

We all became used to the Web sited with tests of equipment. They are
often much better and more thorough than what we can read in the
printed press. What we see on these pages influences our purchasing
decisions and as it is in business: meanwhile Manufacturers "sponsor"
these sits often to bias the tests just a bit...

While looking at the web sites with tests of digital cameras for
several years already I never stopped to wonder: Why they still
fail to include all essential characteristics of the sensor and
of the image processor. For me not just the megapixels count.
I would like to know what is the bit depth per pixel. Or I would
like to see a warning: "manufacturer does not disclose bit depth
of a pixel."

Just look at the dpreview review of Canon G6, what is considered
a prosumer camera. Page 2 (Specifications) lists RAW mode, but
does not specify the bit depth. It makes me angry. Same Steve
Digicams (he mentions however that the digic processor processes
12bit signal) and you can go on so along all review sites of rank.
Same with (say) Nikon 8800. It has NEF and RAW modes, it says.
And? How many bits per pixel does it deliver???

This feature is so fundamental, so decisive. And yet is being
mentioned only in DSLR or middle format digital back reviews.
Why? More bits per pixel provide the so important richness of
detail in highlight and in shadow, allows to manipulate image to
a far larger extend. Of course, whoever ignores RAW or tiff mode
and shoots *.jpeg, has always already lost and obtains 8bit per
color in pixel, regardless the used gear! But the raw data and the
tiffs support bigger depth per pixel and can deliver much richer
image detail. Just a reminder to all who do not deal with data
processing: Obvious calculation shows that 12bit can hold 4096
levels of luminance, 8bit merely 256, it's 16 times more!! Even
mere 10bit per pixel allows already for 4 times more levels of
luminance.

I would suggest to *always* calculate as an additional technical
spec "image data" in megapixels. As an example lets compare two
cameras. I took in both cases Pentax to escape the usual
Canon/Nikon bashing. I used factor of 1024 to calculate Kbyes
and Mbytes.
camera A has camera B has
6Mpix 7Mpix
sensor 3008 x 2008 3056 x 2296
depth 12bits per color. 8bit per color
raw data 69.12Mbytes 53.53MBytes
converted tiff 207.37MBytes 160.6Mbytes

And there is the extra megapixel gone... Why everybody puts
cameras in "megapixel categories," but never not in "image data"
categories?

Of course higher resolution has its merits, but I would always
weight it against more dynamic range! If the smaller resolution
is big enough to match the quality of my glass and to achieve
the largest size of prints which I can do, I will rather always
go for a camera which has better dynamic range and less pixels!
I just would like to know... what is it!!! If not provided,
I assume its mere 8bit per color.

Thomas

More about : dynamic range

Anonymous
December 5, 2004 12:22:43 AM

Archived from groups: rec.photo.digital (More info?)

ThomasH <henrymot@some.net> wrote in news:41B21F1A.EC07571C@comcast.net:

> whoever ignores RAW or tiff mode
> and shoots *.jpeg, has always already lost and obtains 8bit per
> color in pixel, regardless the used gear!

The JPEG picture uses 8 bit gamma 2.2.
That is just as much dynamic range as 12 bit linear.


/Roland
Anonymous
December 5, 2004 1:42:00 AM

Archived from groups: rec.photo.digital (More info?)

ThomasH wrote:
> We all became used to the Web sited with tests of equipment. They are
> often much better and more thorough than what we can read in the
> printed press. What we see on these pages influences our purchasing
> decisions and as it is in business: meanwhile Manufacturers "sponsor"
> these sits often to bias the tests just a bit...
>
> While looking at the web sites with tests of digital cameras for
> several years already I never stopped to wonder: Why they still
> fail to include all essential characteristics of the sensor and
> of the image processor. For me not just the megapixels count.
> I would like to know what is the bit depth per pixel. Or I would
> like to see a warning: "manufacturer does not disclose bit depth
> of a pixel."
>
> Just look at the dpreview review of Canon G6, what is considered
> a prosumer camera. Page 2 (Specifications) lists RAW mode, but
> does not specify the bit depth. It makes me angry. Same Steve
> Digicams (he mentions however that the digic processor processes
> 12bit signal) and you can go on so along all review sites of rank.
> Same with (say) Nikon 8800. It has NEF and RAW modes, it says.
> And? How many bits per pixel does it deliver???
>
> This feature is so fundamental, so decisive. And yet is being
> mentioned only in DSLR or middle format digital back reviews.
> Why? More bits per pixel provide the so important richness of
> detail in highlight and in shadow, allows to manipulate image to
> a far larger extend. Of course, whoever ignores RAW or tiff mode
> and shoots *.jpeg, has always already lost and obtains 8bit per
> color in pixel, regardless the used gear! But the raw data and the
> tiffs support bigger depth per pixel and can deliver much richer
> image detail. Just a reminder to all who do not deal with data
> processing: Obvious calculation shows that 12bit can hold 4096
> levels of luminance, 8bit merely 256, it's 16 times more!! Even
> mere 10bit per pixel allows already for 4 times more levels of
> luminance.
>
> I would suggest to *always* calculate as an additional technical
> spec "image data" in megapixels. As an example lets compare two
> cameras. I took in both cases Pentax to escape the usual
> Canon/Nikon bashing. I used factor of 1024 to calculate Kbyes
> and Mbytes.
> camera A has camera B has
> 6Mpix 7Mpix
> sensor 3008 x 2008 3056 x 2296
> depth 12bits per color. 8bit per color
> raw data 69.12Mbytes 53.53MBytes
> converted tiff 207.37MBytes 160.6Mbytes
>
> And there is the extra megapixel gone... Why everybody puts
> cameras in "megapixel categories," but never not in "image data"
> categories?
>
> Of course higher resolution has its merits, but I would always
> weight it against more dynamic range! If the smaller resolution
> is big enough to match the quality of my glass and to achieve
> the largest size of prints which I can do, I will rather always
> go for a camera which has better dynamic range and less pixels!
> I just would like to know... what is it!!! If not provided,
> I assume its mere 8bit per color.
>
> Thomas
Good post Thomas

This info is gratefully received

Aerticeus

ps - have you given much thought the the optical zoom conversion?

58mm on my digicam looks purty darn close to 200mm on a 35mm yet its
called 380mm (35mm equiv) Let me know what you think on this one

A
Related resources
Anonymous
December 5, 2004 1:43:13 AM

Archived from groups: rec.photo.digital (More info?)

ThomasH wrote:
>...How many bits per pixel does it deliver???

Canon RAW is 12 bits per pixel.

Don't get angry. If you see some specifications that you don't like, buy some
other brand that you do like.

---Bob Gross---
Anonymous
December 5, 2004 2:29:05 AM

Archived from groups: rec.photo.digital (More info?)

In message <Xns95B5E3A5E5AFEklotjohan@130.133.1.4>,
Roland Karlsson <roland_dot_karlsson@bonetmail.com> wrote:

>The JPEG picture uses 8 bit gamma 2.2.
>That is just as much dynamic range as 12 bit linear.

Actually, an 8-bit gamma2.2--adjusted scale has more dynamic range than
12-bit linear. Of course, the 8-bit gamma data comes from the 12-bit
linear, so nothing is gained, only lost, in the 12-bit linear to 8-bit
gamma.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
December 5, 2004 2:33:12 AM

Archived from groups: rec.photo.digital (More info?)

ThomasH wrote:
> We all became used to the Web sited with tests of equipment. They are
> often much better and more thorough than what we can read in the
> printed press. What we see on these pages influences our purchasing
> decisions and as it is in business: meanwhile Manufacturers "sponsor"
> these sits often to bias the tests just a bit...
>
> While looking at the web sites with tests of digital cameras for
> several years already I never stopped to wonder: Why they still
> fail to include all essential characteristics of the sensor and
> of the image processor. For me not just the megapixels count.
> I would like to know what is the bit depth per pixel. Or I would
> like to see a warning: "manufacturer does not disclose bit depth
> of a pixel."
>
> Just look at the dpreview review of Canon G6, what is considered
> a prosumer camera. Page 2 (Specifications) lists RAW mode, but
> does not specify the bit depth. It makes me angry. Same Steve
> Digicams (he mentions however that the digic processor processes
> 12bit signal) and you can go on so along all review sites of rank.
> Same with (say) Nikon 8800. It has NEF and RAW modes, it says.
> And? How many bits per pixel does it deliver???
>
> This feature is so fundamental, so decisive. And yet is being
> mentioned only in DSLR or middle format digital back reviews.
> Why? More bits per pixel provide the so important richness of
> detail in highlight and in shadow, allows to manipulate image to
> a far larger extend. Of course, whoever ignores RAW or tiff mode
> and shoots *.jpeg, has always already lost and obtains 8bit per
> color in pixel, regardless the used gear! But the raw data and the
> tiffs support bigger depth per pixel and can deliver much richer
> image detail. Just a reminder to all who do not deal with data
> processing: Obvious calculation shows that 12bit can hold 4096
> levels of luminance, 8bit merely 256, it's 16 times more!! Even
> mere 10bit per pixel allows already for 4 times more levels of
> luminance.
>
> I would suggest to *always* calculate as an additional technical
> spec "image data" in megapixels. As an example lets compare two
> cameras. I took in both cases Pentax to escape the usual
> Canon/Nikon bashing. I used factor of 1024 to calculate Kbyes
> and Mbytes.
> camera A has camera B has
> 6Mpix 7Mpix
> sensor 3008 x 2008 3056 x 2296
> depth 12bits per color. 8bit per color
> raw data 69.12Mbytes 53.53MBytes
> converted tiff 207.37MBytes 160.6Mbytes
>
> And there is the extra megapixel gone... Why everybody puts
> cameras in "megapixel categories," but never not in "image data"
> categories?
>
> Of course higher resolution has its merits, but I would always
> weight it against more dynamic range! If the smaller resolution
> is big enough to match the quality of my glass and to achieve
> the largest size of prints which I can do, I will rather always
> go for a camera which has better dynamic range and less pixels!
> I just would like to know... what is it!!! If not provided,
> I assume its mere 8bit per color.
>
> Thomas

Personally I find looking at actual results is a much better comparison
than stats.

--
Joseph Meehan

26 + 6 = 1 It's Irish Math
December 5, 2004 7:06:22 AM

Archived from groups: rec.photo.digital (More info?)

> I would like to know what is the bit depth per pixel. Or I would

> Of course higher resolution has its merits, but I would always
> weight it against more dynamic range!


Which is it you want? Dynamic range or color depth?

Perhaps I am mistaken, but it seems that you are running the two
together.

As I have come to understand it, color depth is one thing; but it is not
dynamic range as applied to photography and sensors. From what I have
gleaned, dynamic range is a measure of how many gradations (stops) of
luminance can be recorded sensor-wide, rather than how many gradations
of color value can be recorded at one photo site.
Anonymous
December 5, 2004 7:20:19 AM

Archived from groups: rec.photo.digital (More info?)

Just buy a fuji camera.
Anonymous
December 5, 2004 11:26:39 AM

Archived from groups: rec.photo.digital (More info?)

Aerticulean Effort wrote:
[]
> Aerticeus
>
> ps - have you given much thought the the optical zoom conversion?
>
> 58mm on my digicam looks purty darn close to 200mm on a 35mm yet its
> called 380mm (35mm equiv) Let me know what you think on this one
>
> A

Sounds like a bad measurement or a faulty zoom - I don't think the
specification would be that far out.

David
December 5, 2004 4:15:40 PM

Archived from groups: rec.photo.digital (More info?)

It would be nicer if they stated more tech info on products.

Jpeg is about all the printers can handle, sRGB is the standard for color
printers so you loose nothing other then options to correct errors with raw.

If they could get the cameras to see more stops of light and the printers to
match... then I will be happy.

"ThomasH" <henrymot@some.net> wrote in message
news:41B21F1A.EC07571C@comcast.net...
>
> We all became used to the Web sited with tests of equipment. They are
> often much better and more thorough than what we can read in the
> printed press. What we see on these pages influences our purchasing
> decisions and as it is in business: meanwhile Manufacturers "sponsor"
> these sits often to bias the tests just a bit...
>
> While looking at the web sites with tests of digital cameras for
> several years already I never stopped to wonder: Why they still
> fail to include all essential characteristics of the sensor and
> of the image processor. For me not just the megapixels count.
> I would like to know what is the bit depth per pixel. Or I would
> like to see a warning: "manufacturer does not disclose bit depth
> of a pixel."
>
> Just look at the dpreview review of Canon G6, what is considered
> a prosumer camera. Page 2 (Specifications) lists RAW mode, but
> does not specify the bit depth. It makes me angry. Same Steve
> Digicams (he mentions however that the digic processor processes
> 12bit signal) and you can go on so along all review sites of rank.
> Same with (say) Nikon 8800. It has NEF and RAW modes, it says.
> And? How many bits per pixel does it deliver???
>
> This feature is so fundamental, so decisive. And yet is being
> mentioned only in DSLR or middle format digital back reviews.
> Why? More bits per pixel provide the so important richness of
> detail in highlight and in shadow, allows to manipulate image to
> a far larger extend. Of course, whoever ignores RAW or tiff mode
> and shoots *.jpeg, has always already lost and obtains 8bit per
> color in pixel, regardless the used gear! But the raw data and the
> tiffs support bigger depth per pixel and can deliver much richer
> image detail. Just a reminder to all who do not deal with data
> processing: Obvious calculation shows that 12bit can hold 4096
> levels of luminance, 8bit merely 256, it's 16 times more!! Even
> mere 10bit per pixel allows already for 4 times more levels of
> luminance.
>
> I would suggest to *always* calculate as an additional technical
> spec "image data" in megapixels. As an example lets compare two
> cameras. I took in both cases Pentax to escape the usual
> Canon/Nikon bashing. I used factor of 1024 to calculate Kbyes
> and Mbytes.
> camera A has camera B has
> 6Mpix 7Mpix
> sensor 3008 x 2008 3056 x 2296
> depth 12bits per color. 8bit per color
> raw data 69.12Mbytes 53.53MBytes
> converted tiff 207.37MBytes 160.6Mbytes
>
> And there is the extra megapixel gone... Why everybody puts
> cameras in "megapixel categories," but never not in "image data"
> categories?
>
> Of course higher resolution has its merits, but I would always
> weight it against more dynamic range! If the smaller resolution
> is big enough to match the quality of my glass and to achieve
> the largest size of prints which I can do, I will rather always
> go for a camera which has better dynamic range and less pixels!
> I just would like to know... what is it!!! If not provided,
> I assume its mere 8bit per color.
>
> Thomas
Anonymous
December 5, 2004 5:19:06 PM

Archived from groups: rec.photo.digital (More info?)

In message <41B2893A.9B3FBB7F@none.com>,
None <none@none.com> wrote:

>From what I have
>gleaned, dynamic range is a measure of how many gradations (stops) of
>luminance can be recorded sensor-wide, rather than how many gradations
>of color value can be recorded at one photo site.

If you specifically qualify, "dynamic range of the sensor", yes, and the
dynamic range of the camera's lowest ISO will be just slightly less, but
at any higher ISO, the dynamic range of the output is limited on the top
end by the cutoff point of analog to digital converter (4007 on the
12-bit Canons), and on the bottom end by the noise.

At the lowest ISOs, the shadow range is limited more by quantization
than by noise.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
December 5, 2004 5:45:50 PM

Archived from groups: rec.photo.digital (More info?)

"Sabineellen" <sabineellen@aol.com> wrote in message
news:20041204232019.22097.00001861@mb-m06.aol.com...
>
> Just buy a fuji camera.

Nah, for real depth buy a sigma.

I'm sure out of all cameras the sigmas have been buried the most
Anonymous
December 5, 2004 5:45:51 PM

Archived from groups: rec.photo.digital (More info?)

In message <Towsd.399$A6.20366@nnrp1.ozemail.com.au>,
"ed" <ed@ed.com> wrote:

>Nah, for real depth buy a sigma.

If you add the RAW data from the R, G, and B sensors from each pixel for
greyscale output, the output will have very good dynamic range. The
problem comes when you try to separate colors, giving a lot of hue
noise.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
December 5, 2004 9:24:40 PM

Archived from groups: rec.photo.digital (More info?)

YoYo wrote:
> It would be nicer if they stated more tech info on products.

That would complicate things for the average consumer at whom these
products are aimed.

> Jpeg is about all the printers can handle, sRGB is the standard for
> color printers so you loose nothing other then options to correct
> errors with raw.

No, printers only "handle" JPEG if they accept memory cards directly. I
will let others comment on printer gamut.

> If they could get the cameras to see more stops of light and the
> printers to match... then I will be happy.

Happy, perhaps, but broke!

David
December 5, 2004 9:24:41 PM

Archived from groups: rec.photo.digital (More info?)

"David J Taylor" <david-taylor@invalid.com> wrote in news:31h238F3aohf3U1
@individual.net:

> No, printers only "handle" JPEG if they accept memory cards directly. I
> will let others comment on printer gamut.
>

I think (modern) Postscript printers can actually print .jpg images
natively. An unrelated, but interesting tidbit: HPs newest color lasers
have added a new mode. They can now print PCL, PS, and .pdf. I think the
images in a pdf file can be stored as .jpg.

Bob

--
Delete the inverse SPAM to reply
Anonymous
December 5, 2004 10:47:36 PM

Archived from groups: rec.photo.digital (More info?)

JPS@no.komm wrote:
>
> In message <41B2893A.9B3FBB7F@none.com>,
> None <none@none.com> wrote:
>
> > Which is it you want? Dynamic range or color depth?
> >
> > Perhaps I am mistaken, but it seems that you are running the two
> > together.

It is irrelevant which I prefer personally, your choice might
be different. The point I am trying to make is to make aware
of the importance of the dynamic range and to make it a habit
to disclose this information so that everybody can make a choice.

The majority of users, especially if not involved in data
processing, seem to look after getting "more mega pix" and this
is not the only factor which we should take into consideration.
For *me* it is not even the most important feature to consider!
Your mileage may vary.

>
> >From what I have
> >gleaned, dynamic range is a measure of how many gradations (stops) of
> >luminance can be recorded sensor-wide, rather than how many gradations
> >of color value can be recorded at one photo site.
>
> If you specifically qualify, "dynamic range of the sensor", yes, and the
> dynamic range of the camera's lowest ISO will be just slightly less, but
> at any higher ISO, the dynamic range of the output is limited on the top
> end by the cutoff point of analog to digital converter (4007 on the
> 12-bit Canons), and on the bottom end by the noise.
>
> At the lowest ISOs, the shadow range is limited more by quantization
> than by noise.

Of course, calculating 2**n numbers provide only theoretical
upper boundary numbers. In reality noise will render some of
these possible values unusable. May I ask what is the source
of the cutoff values declared by Canon? I like though that
they disclose such detail.

Thomas


> --
>
> <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
> John P Sheehy <JPS@no.komm>
> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
December 5, 2004 11:38:31 PM

Archived from groups: rec.photo.digital (More info?)

In message <41B365E4.920FEE09@comcast.net>,
ThomasH <henrymot@some.net> wrote:

>Of course, calculating 2**n numbers provide only theoretical
>upper boundary numbers. In reality noise will render some of
>these possible values unusable.

>May I ask what is the source
>of the cutoff values declared by Canon? I like though that
>they disclose such detail.

They don't. That's the RAW value you get when you over-expose a blank
white wall by 10 stops (except for the dead pixels). I am writing
applets that give stats on the RAW data, using Adobe's uncompressed .DNG
files as a data source, reporting low values, and a full histogram for
the entire image and each color channel.

You can load the uncompressed .dng files into photoshop "as" .raw.

For the 10D, the dimensions are 3088*2056, and for the 20D they are
3522*2348.

1 channel, 16-bit, IBMPC, click on header number box, and click "guess",
then load.

These are not the pure RAW values, though, because they are posterized
by Adobe's "15-bit + 32768" system. As an additional pain, there is a
bug in the info tool that reports 16-bit values incorrectly, posterizing
the "info" numbers, but not the actual data. You have to convert to RGB
for the "info" tool to work correctly.

There are other tools that load generic 16-bit ".raw" data and correctly
display pixel values.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
December 5, 2004 11:40:35 PM

Archived from groups: rec.photo.digital (More info?)

In message <Xns95B69B4FEE7E7bobatcarolnet@216.196.97.142>,
bob <usenetMAPS@2fiddles.com> wrote:

>I think (modern) Postscript printers can actually print .jpg images
>natively.

Are there any printers that have rotation built into their drivers?

I've always thought that it was stupid to rotate and perspective scale
things in a raster which would be better done at print time, with the
higher-resolution output medium.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
December 6, 2004 12:17:07 AM

Archived from groups: rec.photo.digital (More info?)

JPS@no.komm wrote in news:ebs6r0tks6u19f9kb5j1fon1f8p2hianv2@4ax.com:


> Are there any printers that have rotation built into their drivers?

HP DesignJets have rotation in the drivers, and scaling too. I don't know
for sure that Postscript supports rotation, but I think it does, and it
does support scaling.

You cannot, as far as I know, send a .jpg to a Postscript printer from the
command line, but that would be interesting.

Bob

--
Delete the inverse SPAM to reply
Anonymous
December 6, 2004 1:40:14 PM

Archived from groups: rec.photo.digital (More info?)

bob wrote:
> "David J Taylor" <david-taylor@invalid.com> wrote in
> news:31h238F3aohf3U1 @individual.net:
>
>> No, printers only "handle" JPEG if they accept memory cards
>> directly. I will let others comment on printer gamut.
>>
>
> I think (modern) Postscript printers can actually print .jpg images
> natively. An unrelated, but interesting tidbit: HPs newest color
> lasers have added a new mode. They can now print PCL, PS, and .pdf. I
> think the images in a pdf file can be stored as .jpg.
>
> Bob

Fascinating. I didn't know that. It certainly makes a lot of sense
rather than sending huge bitmaps over the line. How do they handle colour
spaces defined in the JPEG?

Cheers,
David
December 7, 2004 12:24:05 AM

Archived from groups: rec.photo.digital (More info?)

"David J Taylor" <david-taylor@invalid.com> wrote in
news:31ir8fF3b5083U1@individual.net:

>> They can now print PCL, PS, and .pdf. I
>> think the images in a pdf file can be stored as .jpg.
>>
>> Bob
>
> Fascinating. I didn't know that. It certainly makes a lot of sense
> rather than sending huge bitmaps over the line. How do they handle
> colour spaces defined in the JPEG?
>

In Acrobat, there are settings for color management, just like in
Photoshop. The default jpeg policy is to "preserve embeded profiles."

There are individual settings for the document as a whole, and for the
various image types (jpeg, tiff, etc.)

The settings for images include setting resolution limits and type of
compression. For color and greyscale images, compression options are jpeg
or zip.

I have found experimentally that regardless of the settings I use in
Acrobat, there always seems to be a difference in quality between the
original .jpg and the .jpg in the .pdf file.

Bob


--
Delete the inverse SPAM to reply
Anonymous
December 7, 2004 11:51:06 AM

Archived from groups: rec.photo.digital (More info?)

robertwgross@cs.com (Robertwgross) writes:

>Canon RAW is 12 bits per pixel.

Except in the case of the G2, where it is 10 bits. The G1 might have
been 10 bits as well. I think Canon switched to 12 bits with the G3.

Dave
Anonymous
December 24, 2004 12:48:23 AM

Archived from groups: rec.photo.digital (More info?)

In message <cp3qtq$451$5@mughi.cs.ubc.ca>,
davem@cs.ubc.ca (Dave Martindale) wrote:

>robertwgross@cs.com (Robertwgross) writes:
>
>>Canon RAW is 12 bits per pixel.
>
>Except in the case of the G2, where it is 10 bits. The G1 might have
>been 10 bits as well. I think Canon switched to 12 bits with the G3.

Canon doesn't seem to use the full 12 bits, either, at least on some
cameras. I have run the RAW data from over-exposed images through some
simple data analysis, and the value 4007 seems to be the maximum. The
lowest blackpoint bias I've seen is 124 at room temperature, so only up
to 3884 values can actuall be used.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
January 2, 2005 3:32:32 AM

Archived from groups: rec.photo.digital (More info?)

<JPS@no.komm> wrote in message
news:uuems05mqu8jtnenhhiepfhps5u7t55j3d@4ax.com...
SNIP
> Canon doesn't seem to use the full 12 bits, either, at least on
> some cameras. I have run the RAW data from over-exposed
> images through some simple data analysis, and the value 4007
> seems to be the maximum. The lowest blackpoint bias I've
> seen is 124 at room temperature, so only up to 3884 values can
> actuall be used.

Different Raw decoders may produce different numbers.

I don't know what happens in DNG conversions, but in one Astro program
(IRIS) I got a 120-4095 range (so 11.96 stops) from a 20D Raw CFA
image conversion (it was a night exposure outdoors, low temperatures
and with overexposed streetlights). In another Astro converter
(ImagesPlus) I got from that same image a CFA data range from 0-15873,
which after division by 4 had an 11.95 stops range.

Bart
!