20D and ISO 3200

wrote in message
...


They probably don't want people to use it carelessly, and blame the
noise on the camera, so they set it up in such a way as to keep the
"____ for dummies" crowd from using it.

The 20D's ISO 3200 is really its ISO 1600 under-exposed by a stop, with
a stop of highlights thrown away as it doubles the the RAW capture
values. Some offer this as the answer to your question, but the 10D did
not have ISO 1600 hidden, and its ISO 1600 was the same type of thing
(800 pushed to 1600), so that can't be the real explanation.


But it is interesting to know that it's not the same as 1600. I guess I
could shoot 1600 under exposed and push it myself in either the RAW
converter or Photoshop.

Film noise wouldn't bother me, but the 3200 noise is jpeggy.

--
Mark

Photos, Ideas & Opinions
http://www.marklauter.com

In message ,
"Mr. Mark" wrote:

But it is interesting to know that it's not the same as 1600. I guess I
could shoot 1600 under exposed and push it myself in either the RAW
converter or Photoshop.

Well, you get an extra stop of highlights like that.

When the light gets really low, I often just set my 20D to 1600, select
the longest shutter speed I feel comfortable with, and Tv mode or manual
mode if I'd rather stop the lens down a little.
--


John P Sheehy

In message , I,
wrote:

When the light gets really low, I often just set my 20D to 1600, select
the longest shutter speed I feel comfortable with, and Tv mode or manual
mode if I'd rather stop the lens down a little.

I forgot to mention, this can result in EI values of anywhere from 800
to 400,000 or so. I say 800, because I often set the EC to +1, so that
I get a pulled 800 if the lighting is sufficient. Pulling one stop is
usually not a problem in low contrast lighting with low contrast scenes.
--


John P Sheehy

In message ,
"Mr. Mark" wrote:

Yes, because it's the only ISO setting not actually based on analog
gain.

I dopn't think that's the reason, though, as the 10D's 1600 isn't
hidden, and is pushed 800.

When you say pushed you mean via on-camera software, right? I guess it's
time for me to read up on exactly how sensors emulate specific ISOs.

You're not likely to find any official documentation on this stuff; what
is known comes from users reverse-engineering the RAW data.

I know this from looking at the RAW data; an ISO 1600 image from a 10D
has basically three types of horizontal lines in it; ones that are even
all the way across, ones that are odd all the way across, and ones that
are alternating between odd and even. This means that the smallest bit
in the 12-bit data is totally meaningless, and derived from a 2x
multiplication, with a striping thrown over it.
--


John P Sheehy

wrote in message
...
In message ,
"Mr. Mark" wrote:

Yes, because it's the only ISO setting not actually based on analog
gain.

I dopn't think that's the reason, though, as the 10D's 1600 isn't
hidden, and is pushed 800.

When you say pushed you mean via on-camera software, right? I guess it's
time for me to read up on exactly how sensors emulate specific ISOs.

You're not likely to find any official documentation on this stuff; what
is known comes from users reverse-engineering the RAW data.

I know this from looking at the RAW data; an ISO 1600 image from a 10D
has basically three types of horizontal lines in it; ones that are even
all the way across, ones that are odd all the way across, and ones that
are alternating between odd and even. This means that the smallest bit
in the 12-bit data is totally meaningless, and derived from a 2x
multiplication, with a striping thrown over it.

I tried to understand a previous explanation of this from you but I never
got it. Can you explain how you come to those conclusions, step by step?

Why is the smallest bit meaningless? Why the 2x multiplication? And what
do you mean by striping?

Thanks,
Greg

In message ,
"G.T." wrote:


wrote in message
.. .
In message ,
"Mr. Mark" wrote:

Yes, because it's the only ISO setting not actually based on analog
gain.

I dopn't think that's the reason, though, as the 10D's 1600 isn't
hidden, and is pushed 800.

When you say pushed you mean via on-camera software, right? I guess it's
time for me to read up on exactly how sensors emulate specific ISOs.

You're not likely to find any official documentation on this stuff; what
is known comes from users reverse-engineering the RAW data.

I know this from looking at the RAW data; an ISO 1600 image from a 10D
has basically three types of horizontal lines in it; ones that are even
all the way across, ones that are odd all the way across, and ones that
are alternating between odd and even. This means that the smallest bit
in the 12-bit data is totally meaningless, and derived from a 2x
multiplication, with a striping thrown over it.

I tried to understand a previous explanation of this from you but I never
got it. Can you explain how you come to those conclusions, step by step?

I just did that in the post you replied to.

Why is the smallest bit meaningless?

Because it is. How could the smallest bit be meaningful if it is
determined by a per-line pattern?:

xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1

or

xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0

or

xxxxxxxxxxx0 xxxxxxxxxxx1 xxxxxxxxxxx0 xxxxxxxxxxx1 xxxxxxxxxxx0

Or, if you prefer, from an actual image:

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

Why the 2x multiplication?

To hide what they're doing? To standardize the correlation between RAW
values and rendered output? I don't know exactly why they do it; I can
only speculate. All I can say is that it is done.

And what
do you mean by striping?

Lines of odd values, lines of even values, lines of alternating odd and
even values.


--


John P Sheehy

wrote in message
...
In message ,
"G.T." wrote:


wrote in message
.. .

I know this from looking at the RAW data; an ISO 1600 image from a 10D
has basically three types of horizontal lines in it; ones that are even
all the way across, ones that are odd all the way across, and ones that
are alternating between odd and even. This means that the smallest bit
in the 12-bit data is totally meaningless, and derived from a 2x
multiplication, with a striping thrown over it.

I tried to understand a previous explanation of this from you but I never
got it. Can you explain how you come to those conclusions, step by step?

I just did that in the post you replied to.

Why is the smallest bit meaningless?

Because it is. How could the smallest bit be meaningful if it is
determined by a per-line pattern?:

xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1 xxxxxxxxxxx1

or

xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0 xxxxxxxxxxx0

or

xxxxxxxxxxx0 xxxxxxxxxxx1 xxxxxxxxxxx0 xxxxxxxxxxx1 xxxxxxxxxxx0

Ok, since the last bit is the same all the way across the particular line
it's meaningless.


Or, if you prefer, from an actual image:

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

I'll ponder this image again when I have more time but I'm not getting the
correlation at the moment.


Why the 2x multiplication?

To hide what they're doing? To standardize the correlation between RAW
values and rendered output? I don't know exactly why they do it; I can
only speculate. All I can say is that it is done.

And what
do you mean by striping?

Lines of odd values, lines of even values, lines of alternating odd and
even values.


I understand a bit better now. I'm going to try out IRIS one of these days
especially since I'm going to do some ultra-beginner astrophotography this
weekend with my Rebel XT.

Greg

When the light gets really low, I often just set my 20D to 1600, select
the longest shutter speed I feel comfortable with, and Tv mode or manual
mode if I'd rather stop the lens down a little.

I forgot to mention, this can result in EI values of anywhere from 800
to 400,000 or so. I say 800, because I often set the EC to +1, so that
I get a pulled 800 if the lighting is sufficient. Pulling one stop is
usually not a problem in low contrast lighting with low contrast scenes.

I'm learning a lot. Thanks. What's an EI value?

--
Mark

Photos, Ideas & Opinions
http://www.marklauter.com

In message ,
"Mr. Mark" wrote:

When the light gets really low, I often just set my 20D to 1600, select
the longest shutter speed I feel comfortable with, and Tv mode or manual
mode if I'd rather stop the lens down a little.

I forgot to mention, this can result in EI values of anywhere from 800
to 400,000 or so. I say 800, because I often set the EC to +1, so that
I get a pulled 800 if the lighting is sufficient. Pulling one stop is
usually not a problem in low contrast lighting with low contrast scenes.

I'm learning a lot. Thanks. What's an EI value?

That's what you are exposing for, given a hypothetical standard subject,
like a grey card.

If the camera is set to ISO 3200, and you expose based on a grey card,
for 3200, then the EI (Exposure Index) is 3200. If you use +1 EC
(Exposure Compensation), then the EI is 1600; if you use -1 EC, the EI
is 6400; if you use -4 EC, then the EI is 51,200. You could do this
intentionally, or you could do it unintentionally by under-exposing. Of
course, for this to be meaningful, you have to render the output display
so that a grey card in the scene would look like 18% grey in the output,
I should think; IOW, it must be returned to a standard.
--


John P Sheehy

wrote in message
...

I'm learning a lot. Thanks. What's an EI value?

That's what you are exposing for, given a hypothetical standard subject,
like a grey card.

If the camera is set to ISO 3200, and you expose based on a grey card,

snip

Good explanation. Thanks. Now I have to experiment a little to make sure I
got it.

--
Mark

Photos, Ideas & Opinions
http://www.marklauter.com