ISO and file size

[DanK]

Why does file size depend on ISO? I understand that boosting the ISO amplifies the signal and noise, but how does that affect file size?

I discovered this by accident. The following is a plot of the count of images my 5DIII can store on a 32GB SD card against the ISO setting:



This relationship is nothing obvious, e.g., not logarithmic in either variable.

Anyone know why this happens? Storage is so cheap that it doesn't matter (my camera holds two cards, and this one alone can fit 772 raw files at ISO=25,600), but I am curious about the process.

Thanks

Dan

[dabhand]

http://petapixel.com/2009/12/22/why-...er-file-sizes/

and that curve could be exponential.

[PhotomanJohn]

Wild Guess - May have to do with the compression algorithm that is trying to save more detail in the data. Are these RAW files using lossless compression or JPG?

John

[JPS]

Hi Dan,
I found this might explain things.
Cheers
John

http://petapixel.com/2009/12/22/why-...er-file-sizes/

Looks like Steve found the same site.

[ajohnw]

Weird link that John. PNG is lossless as I understand it.

http://en.wikipedia.org/wiki/Portable_Network_Graphics

John
-

[DanK]

Thanks, all. I guess I should have googled first.

The second reason in the petapixel posting--less compression because noise interrupts regularities in the data--makes sense to me.

The first reason--more noise takes more room--doesn't, but perhaps I misunderstand the way the sensor and electronics work. If I understand correctly, all that boosting the ISO does is amplify whatever the sensor has captured, both signal and noise. We do it to compensate for a weak signal (because we have too little light for the shutter speed and aperture). We want to boost the brightness of the signal data, but the price of doing that is we add luminance to the noise data as well. Hence, again if I am right, boosting ISO doesn't add new noise; rather, it increases the proportion of the data that is noise. I don't see why that would have a large effect on the size of an uncompressed file, but as Zhang points out, it will make a compressed file (including a raw file compressed with a lossless algorithm) larger. This logic assumes that the amount of noise generated by the process of amplification is minor.

Am I missing something here?

Dan

[Colin Southern]

Hi Dan,

Which bit is confusing you?

Basically, a RAW shot is still compressed (or every shot would be exactly the same size and would be (in MB) roughly twice the MP count -- so compression is a given. When numbers are all the same (low ISO) they compress better, but when there's more random variation they don't compress as well.

So in practice, bright areas of a scene (at base ISO) are so far from the noise floor they're not influenced by it -> clean data that compresses well, but higher and higher ISO settings mean less and less of the sensor's physical ability is used -> operating closer and closer to the noise floor -> the darker the tone the more noise that's mixed in with the signal -> less compression -> bigger file size.

[DanK]

Colin,

thanks. Careless reading on my part. Reading it again, I see that Zhang says that compression is why noise matters, so I do understand it. I misread the article the first time to mean that noise contributed separately from its effect on compression.

Dan