High ISO = Less Card Space?

[Rodger]

I noticed this the other day on my 20D.


Shooting RAW at ISO 100 on a blank 4GB Sandisk cf card, I have 447 images left.


Up the ISO to say, 800, and it drops to 416.


That's a big jump. The number of images the card can hold decreases steadily as the ISO is hiked up. Is that normal? Does increasing ISO increase file size drastically enough that it would go from 447 to 416?


I'm more curious than worried. I figure I don't shoot much if at all, over ISO 800, and 416 images is still pretty many.


Just curious.


Thanks,


Rodger

[Jon Ruyle]

That's normal. Higher iso means more noise. Nose doesn't compress, so higher iso pictures tend to take more space.

[Rodger]

Ahh I see. Thanks!

[luck101]

yes, this is true. i notice a larger file size using higher ISO

[STL]

That's normal. Higher iso means more noise. Nose doesn't compress, so higher iso pictures tend to take more space.


Yes, but my XTi also increases the RAW size as I crank up the ISO while RAWs are uncompressed images AFAIK...

[peety3]

That's normal. Higher iso means more noise. Nose doesn't compress, so higher iso pictures tend to take more space.


Yes, but my XTi also increases the RAW size as I crank up the ISO while RAWs are uncompressed images AFAIK...
<div style="clear: both;"]</div>





Canon's CR2 file format uses a lossless compression algorithm. All of the 1s and 0s are retrievable from the .CR2 files, but areas with consistent patterns are compressed. (JPEG is a lossy compression algorithm, so if you have big patches of blue sky for example, pixels that are ALMOST the same value may get compressed to have EXACTLY the same value. That's why overly compressed images have bands of different blues through the sky.) CR2 also has a JPEG thumbnail embedded inside the file, which is therefore going to vary in size.


I think .NEF and I know TIFF are uncompressed (though NEF might have other embedded elements causing a change in size).


The camera also uses an expected/anticipated file size when calculating the number of pictures left. As you fill the card, you may see a picture be taken but not decrement the card space counter. I was playing with my Pocket Wizards the other night, with the 1D3 banging away at 10fps and watching the flashes trigger. With a 1GB card and L JPEG, I managed to drain the card space counter, causing the camera to pause, but as the files were created and written, the camera realized it had more space, so it took some more pictures on two separate bursts (and the images were moved into different folders). And yes, the PW Flex units were keeping up at 10fps...

[Dallasphotog]

Seems to be normal for all the Canon bodies. It's quite noticeable on the 5D MKII using 2G cards. When formatted, the camera will show the card with about 75 available images. If you make a big change to ISO settings, the number of available shots begins to drop away.

[Daniel Browning]

What's really funny is that it's unnecessary. For example:

ISO 1600 - f/4 - 1/500 - 25 MB

ISO 25,600 - f/4 - 1/500 - 45 MB



Both shots have the exact same exposure. The noise is the exact same too (after you correct the brightness in both images to be the same), which is no surprise given that ISO 25,600 is just a digital push of ISO 3200. It has 4 stops of clipped highlights, so it actually has less image data, yet it's almost twice as large. If Canon offered Metadata ISO, the filesize could remain the same, and the highlights would not be deleted needlessly.

[peety3]

What's really funny is that it's unnecessary. For example:

ISO 1600 - f/4 - 1/500 - 25 MB

ISO 25,600 - f/4 - 1/500 - 45 MB



Both shots have the exact same exposure. The noise is the exact same too (after you correct the brightness in both images to be the same), which is no surprise given that ISO 25,600 is just a digital push of ISO 3200. It has 4 stops of clipped highlights, so it actually has less image data, yet it's almost twice as large. If Canon offered Metadata ISO, the filesize could remain the same, and the highlights would not be deleted needlessly.
<div style="clear: both;"]</div>





They have the same exposure data, which suggests to me that the second shot had less light (maybe 16x less light). With less light, the sensor would be closer to its noise floor, as would the electronics further down the chain. More noise would then lead to a larger file.

[Daniel Browning]

They have the same exposure data, which suggests to me that the second shot had less light (maybe 16x less light).





They have the exact same "exposure", which means the exact same amount of light. (The "data" value is not part of exposure; it is part of "brightness"). The ISO 25,600 shot has a four-stop (16X) increase in gain, so the data values are 16X higher than the ISO 1600 shot, even though the amount of light is the same. The SNR is the same too, but one must equalize the brightness for a visual comparison.

[peety3]

What's really funny is that it's unnecessary. For example:

ISO 1600 - f/4 - 1/500 - 25 MB

ISO 25,600 - f/4 - 1/500 - 45 MB



Both shots have the exact same exposure. The noise is the exact same
too (after you correct the brightness in both images to be the same),
which is no surprise given that ISO 25,600 is just a digital push of
ISO 3200. It has 4 stops of clipped highlights, so it actually has less
image data, yet it's almost twice as large. If Canon offered Metadata
ISO, the filesize could remain the same, and the highlights would not
be deleted needlessly.





You never said whether these two bulleted examples were of
the same subject matter. You only said this was two different
images/files. You said both shots have the exact same exposure, which
we could see plainly by reading the text.







They have the same exposure data, which suggests to me that the second shot had less light (maybe 16x less light).





They have the exact same "exposure", which means the exact same amount of light. (The "data" value is not part of exposure; it is part of "brightness"). The ISO 25,600 shot has a four-stop (16X) increase in gain, so the data values are 16X higher than the ISO 1600 shot, even though the amount of light is the same. The SNR is the same too, but one must equalize the brightness for a visual comparison.
<div style="clear: both;"]</div>





Jesus, do you have to pick apart every post word by word? If you had a tool that could show you the 1s and 0s related to a particular picture/image/whatever YOU want to call it on your memory card, there'd be some data in that file that would tell you the exposure 'specifications'. Based on your previous posts, you define 'exposure' to be shutter and aperture. I consider that to be data (it's expressed in 1s and 0s, and varies picture-by-picture), and it happens to be data about the exposure, so I called it 'exposure data'. There'd be millions of 1s and 0s that represent what the camera considers to be the actual image, I'd consider that to be 'image data'.


And the same "exposure" in no way signifies that it's the exact same amount of light. I can set my camera to manual exposure, f/4 and 1/500, and shoot all day long. There's no inherent likelihood that any two pictures would have the same amount of light. I could also shoot in Av f/4, and when I see a shutter speed of 1/500th I could lock exposure, to take just two pictures. However, if I'm shooting in Av f/4 and take a shot when in ISO 1600 and see a shutter speed of 1/500th, I'm going to assume that the camera chose that speed based on what the meter interpreted of the scene. If I'm shooting in Av f/4 and take a shot when in ISO 25,600 and see a shutter speed of 1/500th, I'm going to assume that the camera chose that speed based on what the meter interpreted of the scene. Chances are, the second example shot (i.e. f/4 1/500th with an ISO selection of 25,600) was taken in 1/16th the light (within the parameters of the metering mode selected) as the first example shot (i.e. f/4 1/500th with an ISO selection of 1600). But guess what? They've got the same exposure (and you'll find the same exposure data in each image file). Big whoopdedo. As a result of the difference in light seen by the meter and sensor, I'm expecting higher noise and larger file size, OK?


Why do I come here?

[Daniel Browning]

You never said whether these two bulleted examples were of
the same subject matter.


It doesn't have to be. I used the word exposure to make such disclaimers unnecessary. If the subject is two stops brighter, and one adds a two stop ND, then exposure stays the same. Exposure relates to the total amount of light per area.






You only said this was two different
images/files. You said both shots have the exact same exposure, which
we could see plainly by reading the text.


No, what you can see plainly by reading the text is that they have the same f-number and shutter speed. If flash was used on one, or ND filter on the other, or different scene luminance on one, then exposure would be different. I specifically stated that exposure was the same, therefore all those other factors (flash, ND filters, scene luminance) are the same.



Based on your previous posts, you define 'exposure' to be shutter and aperture.


No, I define exposure the same way as the dictionary and photography textbooks (e.g. "Photography", by London and Upton): total amount of light per area.



I consider that to be data (it's expressed in 1s and 0s, and varies picture-by-picture), and it happens to be data about the exposure, so I called it 'exposure data'.


It appears that we differ on the definition of "exposure".






And the same "exposure" in no way signifies that it's the exact same amount of light. I can set my camera to manual exposure, f/4 and 1/500, and shoot all
day long. There's no inherent likelihood that any two pictures would
have the same amount of light.





A change in the amount of light *is* a change in exposure.



Chances are, the second example shot (i.e. f/4 1/500th with an ISO selection of 25,600) was taken in 1/16th the light (within the parameters of the metering mode selected) as the first example shot (i.e. f/4 1/500th with an ISO selection of 1600). But guess what? They've got the same exposure (and you'll find the same exposure data in each image file).


That's incorrect. They may have the same brightness, and the same raw data values, but they have very different exposures.

[peety3]

Pick one: "exposure is the total amount of light per area" or "exposure is shutter speed and aperture".


Seriously, please pick one.

[Daniel Browning]

Pick one: "exposure is the total amount of light per area" or "exposure is shutter speed and aperture".


Seriously, please pick one.





It's the total amount of light per area, as I said above several times.

[peety3]

Pick one: "exposure is the total amount of light per area" or "exposure is shutter speed and aperture".


Seriously, please pick one.





It's the total amount of light per area, as I said above several times.
<div style="clear: both;"]</div>





OK, then saying this:






What's really funny is that it's unnecessary. For example:

ISO 1600 - f/4 - 1/500 - 25 MB

ISO 25,600 - f/4 - 1/500 - 45 MB



Both shots have the exact same exposure. The noise is the exact same
too (after you correct the brightness in both images to be the same),
which is no surprise given that ISO 25,600 is just a digital push of
ISO 3200. It has 4 stops of clipped highlights, so it actually has less
image data, yet it's almost twice as large. If Canon offered Metadata
ISO, the filesize could remain the same, and the highlights would not
be deleted needlessly.





just doesn't fly. Nowhere in either of those bulleted points is there a measure of light per area. There is merely a length of time, a diameter value, and an approximate sensitivity of an imaging sensor.


If those two bulleted points were shots of the same static 'thing' (which you never stated in that post), they might represent the same optical exposure, but the camera is not "programmed" to interpret those "exposures" the same, and therefore there's no reason to say the noise is the same.

[Daniel Browning]

just doesn't fly. Nowhere in either of those bulleted points is there a measure of light per area. There is merely a length of time, a diameter value, and an approximate sensitivity of an imaging sensor.


I think we still have a misunderstanding. I think the confusion started when my statement made it seem like exposure means the same as "f-number and shutter". I should have made it more clear that they are not the same (nor did I mean to imply they were the same.)


I think this is how you interpreted my statement:

The camera settings (f-number, shutter speed) are the same, therefore the total light (exposure) is the same.



That would be wrong, because as you mentioned, it fails to account for the many other factors that can affect total light.Here is what I was trying to say:

Exposure is the same, therefore the total light is the same. Also, The camera settings happen to be X, Y, and Z.



I could have put it this way:

Both shots have the same exposure, because they have the same f-number, same shutter speed, same lens transmissivity, same 0.3 ND filter, same linear polarizing filter, same 50 GN flash, same 300 lux/s scene luminance, same aperture activation variance, and same shutter activation variance.



But it's a lot simpler to just say they have the same exposure, because it means all those other factors are the same as well.


Sorry for the confusion.

[Jon Ruyle]

Both shots have the exact same exposure. The noise is the exact same too (after you correct the brightness in both images to be the same), which is no surprise given that ISO 25,600 is just a digital push of ISO 3200. It has 4 stops of clipped highlights, so it actually has less image data, yet it's almost twice as large.


The snr is the same, but the higher iso picture has more noise (that is, standard deviation of noise is a larger number). Thus it takes more data to store.


One could say "we should have metadata iso" and I agree heartily. But the real problem here, I believe, is a compression algorithm so stupid it doesn't even notice all those trailing zeros at high iso. As you say, the higher iso picture contains less data. If compressed reasonably, it would be smaller.