Sigma cameras try to appeal to real shooters

In article ,
RichA wrote:

https://www.dpreview.com/opinion/997...shoots-dng-raw

dng can't support foveon unless sigma releases their secret sauce, at
which point people will realize just how much they've been scammed.

In article ,
nospam wrote:

In article ,
RichA wrote:

https://www.dpreview.com/opinion/997...shoots-dng-raw

dng can't support foveon unless sigma releases their secret sauce, at
which point people will realize just how much they've been scammed.

These files are not RAW in any conventional sense. From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

150MB files from 25MP 8bit captures...
--
teleportation kills

In article ,
nospam wrote:

In article ,
RichA wrote:

https://www.dpreview.com/opinion/997...shoots-dng-raw

dng can't support foveon unless sigma releases their secret sauce, at
which point people will realize just how much they've been scammed.

These files are not RAW in any conventional sense. From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

150MB files from 25MP 12bit captures...
--
teleportation kills

In article , android
wrote:


https://www.dpreview.com/opinion/997...shoots-dng-raw

dng can't support foveon unless sigma releases their secret sauce, at
which point people will realize just how much they've been scammed.

These files are not RAW in any conventional sense.

foveon cameras do not produce true raw. a foveon raw has already been
partially processed by the camera. it's not possible to get the true
sensor data.

From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

translated: it's a tiff.

any 'foveon advantage' is entirely gone (not that there ever was one
but that's another discussion).

150MB files from 25MP 8bit captures...

that's because it's basically a tiff file.

and writing all that data on a camera that is already slow is not going
to be a good user experience.

In article , android
says...
These files are not RAW in any conventional sense. From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

150MB files from 25MP 12bit captures...

Well, the only difference to a RAW file from a Bayer camera is that a
Bayer RAW lacks two thirds of the colour information. The Sigma RAW
files instead have the full colour information and probably are so large
because they have many bits, allowing a good deal of image processing.

BTW, it's not 25MP 12bit captures, it's 25MP 12 bit x 3.
--
Alfred Molon

Olympus E-series DSLRs and micro 4/3 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site

In article ,
Alfred Molon wrote:

In article , android
says...
These files are not RAW in any conventional sense. From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

150MB files from 25MP 12bit captures...

Well, the only difference to a RAW file from a Bayer camera is that a
Bayer RAW lacks two thirds of the colour information. The Sigma RAW
files instead have the full colour information and probably are so large
because they have many bits, allowing a good deal of image processing.

BTW, it's not 25MP 12bit captures, it's 25MP 12 bit x 3.

No it's a constructed DNG. According to The Imaging Resource:

"The top layer is 25.6 megapixels and an additional pair of
6.4-megapixel layers below"

http://www.imaging-resource.com/PROD...igma-sd-quattr
o-hA.HTM

That would put the real resolution between 6.5 and 25.4MP...
--
teleportation kills

In article , Alfred
Molon wrote:

These files are not RAW in any conventional sense. From the linked
article:

"The Quattro H performs the necessary deconvolution and interpolation
process required to derive red, green and blue information for each
pixel, so that the Raw processing software doesn't need to."

150MB files from 25MP 12bit captures...

Well, the only difference to a RAW file from a Bayer camera is that a
Bayer RAW lacks two thirds of the colour information.

completely wrong, but even if that was true, human vision can only
resolve 1/10th of spatial resolution so bayer is still well ahead of
what you can actually see.

The Sigma RAW
files instead have the full colour information and probably are so large
because they have many bits, allowing a good deal of image processing.

the files are large because there's redundant (and inaccurate)
information in them.

BTW, it's not 25MP 12bit captures, it's 25MP 12 bit x 3.

it's not even 25 mp, let alone x3.

the bottom two layers are 6mp, making it a lowly 6 mp camera with a
blue layer that is higher resolution.

the true resolution is slightly higher than 6mp, except for pure blue
subjects of which there are none.

sigma is lying, as is anyone else who parrots the same bull****.

In article , android
wrote:

BTW, it's not 25MP 12bit captures, it's 25MP 12 bit x 3.

No it's a constructed DNG. According to The Imaging Resource:

"The top layer is 25.6 megapixels and an additional pair of
6.4-megapixel layers below"

which makes it a glorified 6.4 megapixel camera.

http://www.imaging-resource.com/PROD...igma-sd-quattr
o-hA.HTM

That would put the real resolution between 6.5 and 25.4MP...

closer to 6.5.

what's worse is that the resolution varies depending on the colour of
the subject.

only pure blue subjects matching the spectral range of the top layer
will get the full 25mp. *everything else* will be limited to 6.4 mp,
with the rest of the data being 'guessed' (with lots of artifacts).

ironically, 'guessing' is the very thing that the ignorant sigma fans
say about bayer. except it's *much* worse on sigma. not only is it
guessing colour but it's guessing spatial resolution.

In article , android
says...
That would put the real resolution between 6.5 and 25.4MP.

Well no, it's not that simple.

The real resolution of a Bayer sensor is probably around 50%-70% of the
pixel count, depending on the scene. This Foveon sensor, if it doesn't
have three complete colour layers as you claim, will have a somewhat
higher real resolution - perhaps 70% to 90% of the pixel count.
--
Alfred Molon

Olympus E-series DSLRs and micro 4/3 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site

In article , Alfred
Molon wrote:

That would put the real resolution between 6.5 and 25.4MP.

Well no, it's not that simple.

it's not simple but it's correct.

the actual resolution of that sigma camera varies depending on the
colour of the subject, something which is *never* desirable.

because the bottom two layers are 6 mp, it can only resolve slightly
better than an ordinary 6 mp camera, unless the subject is mostly blue,
in which case it will be a bit better because the top layer, which
responds to blue, has higher pixel count and can resolve more detail.

real world subjects are *not* mostly blue nor are they a spectral match
for the top layer.

in other words, it's a 6 mp camera that in some cases can do a little
better, with a whole ****load of artifacts because it's guessing most
of the details.

The real resolution of a Bayer sensor is probably around 50%-70% of the
pixel count, depending on the scene.

nonsense. first of all, resolution is not measured by pixel count and
second of all, bayer is close to its theoretical maximum (nyquist).

This Foveon sensor, if it doesn't
have three complete colour layers as you claim,

that's not what was claimed.

will have a somewhat
higher real resolution - perhaps 70% to 90% of the pixel count.

only if you properly count the pixels, which sigma nor its followers do.