Reading mode: Light Dark

EOS 1D BlackField test data

Go to last post
S

Steven Noyes

Forum Pro
Messages
12,383
Reaction score
31
Location
Chandler, US
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name) sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name) sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-) sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of frames together to create a better image with less noise. This is also a good experiment to determine if the noise is pattern based or random.

Basic conclusions: (Nothing really new)

ISO200. Very clean with at most 1 bit error (8 bit JPG). The error is much worse on the right side of the frame than on the left. Almost completely random. Averaging 3 files almost eliminates all noise (There are about 6 hot pixels at a 1 bit error level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO 200. Right side is 5 times worse than the left. There is a dark vertical band on the right side about 10% from the right side taking up 5% of the frame. There are a series of 10 pixel horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this indicates that there is a large random component to the noise. This also mean that frame averaging is a good plan to minimize the noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than ISO 400. Some of the random component is now gone.

Enjoy

Steven
 
Hi Steven,

can you tell me,what firmware you used and what firmware was installed,when you bought the camera.I'm asking,because I'm curious,wether the right side of the pictures is still worse than the left,even with the newest hardware/firmware.You can also notice this in many high ISO shots with dark background.If you look carefully,especially the banding is worse on the right side.
Stefan
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
 
No problem.

Camera serial #6659 delivered with Firmware 1.3.0. And yes the right
side is much worse (visually would say 4 times worse) than the left.

The sad part is the right side isn't that bad. It is just that the
left side is the absolute best I have ever seen. The right side is
very typical.

If you want I can put up the level enhanced images of ISO 200, 400 and
800. I hope that everyone could read the TarBall I put up.

Steven
Stefan said:
Hi Steven,
can you tell me,what firmware you used and what firmware was
installed,when you bought the camera.I'm asking,because I'm
curious,wether the right side of the pictures is still worse than
the left,even with the newest hardware/firmware.You can also notice
this in many high ISO shots with dark background.If you look
carefully,especially the banding is worse on the right side.
Stefan
Click to expand...
 
I found this very interesting as I have in the past toyed with the idea of taking more than one image and somehow combine them to minimize the random noise.

I gave up when I couldn't satisfactorally combine the images using PS6.

You say you "averaged" them. What was the process you used to do that?

I have a d30 btw, but I think the idea would be fundamentally the same.

Thanks
Steven Noyes said:
The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.
Click to expand...
--sean
 
What if you work with high-bit TIFFs instead? Use Bibble or something
where you can turn off ALL processing except perhaps WB and gamma
adjustment to produce the TIFFs. There's no telling how much has
been clipped at the bottom end and what the noise really looks like
from examining JPEGs. It also matters what color space is used; Adobe
RGB will almost certainly result in more noise than a smaller space.
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
 
So looking at these images in the 'Gimp' image editor on Linux, zooming into 100%, and turning the contrast way down I was able to see the noise (Not really visible to me before turning down contrast).

As you mentioned the noise jumps out at you and is clearly strong on the right side of the image, and pretty clear on the left side. There is a clearly defined demarcation between left and right split right down the middle of the frame.

So I have two questions:

1) What do you think is responsible for the noise diff from left to right. Is the sensor somehow logically divided into two halves? I know that it isn't physically divided. At least I think I know that! :-)

2) Since I was not able to see this noise on the black frame until I modified the contrast way way down, why do I care about this? Do you think it would affect the visible image either on screen or in print without artificially hammering the contrast? If I can't see it on a black frame with normal contrast how would I see it on an actual image?

Thanks for the interesting experiment. I may try it on my own 1D when it arrives tomorrow! :-)

Glenn
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
 
Thanks Steven,
btw,can you load up a 1600 ISO shot(standard curves) with dark background?
Stefan
Steven Noyes said:
Camera serial #6659 delivered with Firmware 1.3.0. And yes the right
side is much worse (visually would say 4 times worse) than the left.

The sad part is the right side isn't that bad. It is just that the
left side is the absolute best I have ever seen. The right side is
very typical.

If you want I can put up the level enhanced images of ISO 200, 400 and
800. I hope that everyone could read the TarBall I put up.

Steven
Stefan said:
Hi Steven,
can you tell me,what firmware you used and what firmware was
installed,when you bought the camera.I'm asking,because I'm
curious,wether the right side of the pictures is still worse than
the left,even with the newest hardware/firmware.You can also notice
this in many high ISO shots with dark background.If you look
carefully,especially the banding is worse on the right side.
Stefan
Click to expand...
Click to expand...
 
Glenn Rempe said:
So looking at these images in the 'Gimp' image editor on Linux,
zooming into 100%, and turning the contrast way down I was able to
see the noise (Not really visible to me before turning down
contrast).

As you mentioned the noise jumps out at you and is clearly strong
on the right side of the image, and pretty clear on the left side.
There is a clearly defined demarcation between left and right split
right down the middle of the frame.

So I have two questions:

1) What do you think is responsible for the noise diff from left
to right. Is the sensor somehow logically divided into two halves?
I know that it isn't physically divided. At least I think I know
that! :-)

2) Since I was not able to see this noise on the black frame until
I modified the contrast way way down, why do I care about this? Do
you think it would affect the visible image either on screen or in
print without artificially hammering the contrast? If I can't see
it on a black frame with normal contrast how would I see it on an
actual image?
Click to expand...
You won't, so don't worry about it.
Glenn Rempe said:
Thanks for the interesting experiment. I may try it on my own 1D
when it arrives tomorrow! :-)

Glenn
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
Click to expand...
 
It is a little program that I am writing that allows you to drag in
any number (up to 65536 anyway) images and do one of 3 operations:

Sum and average (32 bits/channel)
Sum based off of brightness and adjust tone curve (Again 32 bit/channel)
This will assign a heavier weight to "brighter" images in such a
way that you can extend the dynamic range of an image.
Auto align and rotate and then sum images. This will be designed
for star photos only and I have no idea if it will work.

Program will also convert Canon RAW image formats. So I guess I will
have to add some color correction stuff as well.

I have used this averaging of images some in the past and when it can be
used, the results are nothing short of amazing.
SeanO said:
I gave up when I couldn't satisfactorally combine the images using
PS6.

You say you "averaged" them. What was the process you used to do that?

I have a d30 btw, but I think the idea would be fundamentally the
same.

Thanks
Steven Noyes said:
The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.
Click to expand...
--
sean
Click to expand...
 
Agree 100%. The JPG truncates the lower 4 bits of data since the
1D has a 12 bit A/D in it. What this indicates is that at ISO 200,
the 1D is getting about 9 bits of "real" data (The G1 gets about 6.5
bits of real data at ISO 50. To me this is a huge improvement). At
ISO the 1D is getting right at 8 bits so going to RAW will buy little
on a correctly exposed image. It will allow some exposure compensation
on an image that is not exposed right. This will come at the cost
of image quality if you have to "push" the exposer, however.

Color space was color matrix 1, JPG large level 10.
Jan Brittenson said:
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
Click to expand...
 
Glenn Rempe said:
1) What do you think is responsible for the noise diff from left
to right. Is the sensor somehow logically divided into two halves?
I know that it isn't physically divided. At least I think I know
that! :-)
Click to expand...
It almost certainly is. The logical structure is most likely an interline CCD with split readout lines. So each sampling site (pixel) is a light sensitive well coupled to a gated storage well. (Same thing as the optical well more or less except it is masked with black.) The gated storage wells in a given horizontal line of the image are connected together in a shift register (or "bucket brigade"). For readout speed reasons, each line is actually two shift registers, one going from the halfway point to the right side and the other going from the halfway point to the left side. There is a readout sense amp connected to each side.

Things can get a bit more complicated with interleaved readout as well. That would be where every other line goes to a different sense amp. (So four sense amps.) This sort of thing is done to afford getting the data off the chip fast enough.

I should add that I'm only guessing based on info from other chip designs and published behaviors of the EOS-1D. I do not know of a good source for low-level info on how the EOS-1D sensor is built.

-Z-
 
Well, that sounds very promising. I too was looking to use this technique for astronomy.

So, if you need a beta tester or whatever, you know where to find one :-)
Steven Noyes said:
Sum and average (32 bits/channel)
Sum based off of brightness and adjust tone curve (Again 32
bit/channel)
This will assign a heavier weight to "brighter" images in such a
way that you can extend the dynamic range of an image.
Auto align and rotate and then sum images. This will be designed
for star photos only and I have no idea if it will work.

Program will also convert Canon RAW image formats. So I guess I will
have to add some color correction stuff as well.

I have used this averaging of images some in the past and when it
can be
used, the results are nothing short of amazing.
SeanO said:
I gave up when I couldn't satisfactorally combine the images using
PS6.

You say you "averaged" them. What was the process you used to do that?

I have a d30 btw, but I think the idea would be fundamentally the
same.

Thanks
Steven Noyes said:
The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.
Click to expand...
--
sean
Click to expand...
Click to expand...
--sean
 
Steven

I am not sure but I think you will find this thread interesting:

http://www.dpreview.com/forums/read.asp?forum=1006&message=2222801

Read the whole thread and their are some good links to programs that allign images.

DJM
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
 
Glenn Rempe said:
So looking at these images in the 'Gimp' image editor on Linux,
zooming into 100%, and turning the contrast way down I was able to
see the noise (Not really visible to me before turning down
contrast).

As you mentioned the noise jumps out at you and is clearly strong
on the right side of the image, and pretty clear on the left side.
There is a clearly defined demarcation between left and right split
right down the middle of the frame.

So I have two questions:

1) What do you think is responsible for the noise diff from left
to right. Is the sensor somehow logically divided into two halves?
I know that it isn't physically divided. At least I think I know
that! :-)
Click to expand...
My guess (and that is all it is) is signal routing. The 1D does split the
circuit that gets the data off of the CCD down the middle. Apparently,
the circuit on the left has better routing or placement.
Glenn Rempe said:
2) Since I was not able to see this noise on the black frame until
I modified the contrast way way down, why do I care about this? Do
you think it would affect the visible image either on screen or in
print without artificially hammering the contrast? If I can't see
it on a black frame with normal contrast how would I see it on an
actual image?
Click to expand...
@ ISO 200
At 8X10 this will be 100% invisible. At 20 X 30 it might make a huge
difference. One thing is that it will allow sharpening without causing
the noise to be sharpened as well.

But for the most part, it is a fun academic experiment.
Glenn Rempe said:
Thanks for the interesting experiment. I may try it on my own 1D
when it arrives tomorrow! :-)

Glenn
Click to expand...
 
Look at http://www.panoguide.com . Good sight.

Close but not quite. That and I run on OS X I don't know of any
thing out there for this. There are some really good programs for
astro-photography for Windows but they only work on B&W.

I know how to align star field images and do an average. The routine,
however, falls apart on normal images. Time time time.

Steven
David James Miller said:
I am not sure but I think you will find this thread interesting:

http://www.dpreview.com/forums/read.asp?forum=1006&message=2222801

Read the whole thread and their are some good links to programs
that allign images.

DJM
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
Click to expand...
 
Keep at it and let us know, interested in this too

DJM
Steven Noyes said:
Close but not quite. That and I run on OS X I don't know of any
thing out there for this. There are some really good programs for
astro-photography for Windows but they only work on B&W.

I know how to align star field images and do an average. The routine,
however, falls apart on normal images. Time time time.

Steven
David James Miller said:
I am not sure but I think you will find this thread interesting:

http://www.dpreview.com/forums/read.asp?forum=1006&message=2222801

Read the whole thread and their are some good links to programs
that allign images.

DJM
Steven Noyes said:
NOTE: You only want to do this test if you want to say. "Boy this is
horrible." NO camera is perfect. The 1D is 1 to 2 orders of
magnitude
better than the G1.

The following link is a link to data taken on the 1D with the lens
cap on. The files are then summed up and averaged.

http://www.noisetech.com/BlackField200-400-800.tiff.tgz

(Note: I have tested this in OS X OmniWeb and OS X IE)

There are a series of files. This will decompress with a program
like WinZip/PKZip/StuffIt or OpenUp. Or for you Linux people:

tar -xzf BlackField200-400-800.tiff.tgz

The Files:

Sample100.jpg -- This file is ISO 200 (Sorry about the name)
sample image.
A simple autocontrast will show the noise image.

Sample200.jpg -- This file is ISO 400 (Sorry about the name)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

Sample800.jpg -- This file is ISO 800 (GOT THE NAME RIGHT :-)
sample image.
A level adjust of loweset 1% map to black and brightest
1% map to white will show noise image.

The composites are the result of average a specific number of
frames together to create a better image with less noise. This is
also a good experiment to determine if the noise is pattern based
or random.

Basic conclusions: (Nothing really new)
ISO200. Very clean with at most 1 bit error (8 bit JPG). The
error is much worse on the right side of the frame than on the
left. Almost completely random. Averaging 3 files almost
eliminates all noise (There are about 6 hot pixels at a 1 bit error
level). There is no repeated pattern noise.

ISO400. Still clean but the noise level is much higher than ISO
200. Right side is 5 times worse than the left. There is a dark
vertical band on the right side about 10% from the right side
taking up 5% of the frame. There are a series of 10 pixel
horizontal bands about 40% up from the bottom ONLY on the right.

With 20 frames averaged, however, the noise is much better and this
indicates that there is a large random component to the noise.
This also mean that frame averaging is a good plan to minimize the
noise :-)

ISO800. Average 11 frames. ISO 800 looks about 150% worse than
ISO 400. Some of the random component is now gone.

Enjoy

Steven
Click to expand...
Click to expand...
Click to expand...
 
I have not done 1600 and 3200 but will and will post the results.

Steven

PS. I have to send it in for repair to get the focusing calibrated :-(

Steven
Stefan said:
Thanks Steven,
btw,can you load up a 1600 ISO shot(standard curves) with dark
background?
Stefan
Steven Noyes wrote:
Click to expand...
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top