Eos M ISO 1600 to 6400 tests and review...

[Ivan Muller]

FYI, I have just completed some high iso tests and comparisons made with the Eos M. I also shot a jazz concert with the Eos M and a Canon 70-300L attached via the lens adapter with one of the images shown here made at 5000 iso including a 100%crop. More images and tests here on my blog at...http://thelazytravelphotographer.blogspot.com/.

 

[Rienzphotoz]

Well done Ivan!

 

[Pi]

This statement

More pixels add more resolution but also more noise at higher ISO's...

is generally speaking incorrect.

I enjoyed the pictures, BTW.

 

[bdunbar79]

This statement

More pixels add more resolution but also more noise at higher ISO's...

is generally speaking incorrect.

I enjoyed the pictures, BTW.

Oh dear...

First off, great review! But, I agree with Pi, the statement about pixels and noise is just not correct.

 

[Ivan Muller]

Thanks for your reply! Re my statement, would you care to explain why you thinks its incorrect? I am happy to change it if you can give some more detailed explanation....

 

[Pi]

Thanks for your reply! Re my statement, would you care to explain why you thinks its incorrect? I am happy to change it if you can give some more detailed explanation....

Read this, for example:

http://www.josephjamesphotography.com/equivalence/#8

There are also heated discussions on other forums, you can google them. I like Joe's essay. He also happens to be a very talented photographer.

In short, the answer is that the high ISO (to be more precise, low exposure, photon) noise is a part of the signal projected on the sensor, and determined by the discrete nature of light.

Here is a more technical explanation:

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/

 

[Ivan Muller]

thanks Pi! I will read the article from Joseph James...but being more practically inclined than theoretically its going to take some time!

 

[Pi]

thanks Pi! I will read the article from Joseph James...but being more practically inclined than theoretically its going to take some time!

Do not read Emil's article then!

BTW, check out Joe's pbase galleries.

 

[Ivan Muller]

I thought as much! will check it out....

 

[yakman]

Thanks for the review. Nice read.
I decided to buy an OVF (certainly not Leica) after reading your review!
Please keep it updated with your findings on using the OVF with 22/F2

 

[seekthedragon]

Thanks for your reply! Re my statement, would you care to explain why you thinks its incorrect? I am happy to change it if you can give some more detailed explanation....

Read this, for example:

http://www.josephjamesphotography.com/equivalence/#8

There are also heated discussions on other forums, you can google them. I like Joe's essay. He also happens to be a very talented photographer.

In short, the answer is that the high ISO (to be more precise, low exposure, photon) noise is a part of the signal projected on the sensor, and determined by the discrete nature of light.

Here is a more technical explanation:

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/

Sounds good for a mathematician, but last time, when I was taught microelectronics, total CMOS sensor noise was almost a linear function of pixel count, due to the number of active components required. You can double the SNR by a 2x2 averaging, but using the same technology, the on-chip pixel binning is just better. I don't really see, why all this papers missing this point.

 

[Pi]

Sounds good for a mathematician, but last time, when I was taught microelectronics, total CMOS sensor noise was almost a linear function of pixel count, due to the number of active components required. You can double the SNR by a 2x2 averaging, but using the same technology, the on-chip pixel binning is just better. I don't really see, why all this papers missing this point.

The second author is actually a physics professor at the U. of Chicago. Not that there is anything wrong being a mathematician, I am.

I think that you are confusing noise per pixel with noise per unit area. You might also be talking about read noise, which depends on the technology. This http://www.sensorgen.info/ is a website computing the read noise from DXO data (per pixel). As you can see, smaller pixels usually have lower read noise (measured in (e-). On image level, it may be lower of higher, depends.

Emil's article contains a lot of info. The data is pretty old, 2008, but the theory has not changed since that. Why don't you read it. This is the most relevant part: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#pixelsize but you cannot disconnect it from the rest.