ISO settings

[Subyn00by]

So why not just use ISO100 all the time? I mean, I know in low light you want to use a little higher ISO to allow a faster shutter to reduce shake, but I see a lot of photographers using ISO200, 300, and higher for scenery/landscape. What (if any) benefit is acheived from this?

[Doug Pardee]

Like all things in photography, it depends on the situation.

There is almost no noticeable difference in image quality between ISO 100 and ISO 200 on most Canon DSLRs. On the other hand, shooting at ISO 200 gives you twice the shutter speed, making for sharper pictures. Or lets you stop down more to increase Depth of Field. If you're using a polarizing filter, the 2+ stops of light loss can make for slower shutter speeds and/or wider apertures than you'd like.

Another consideration is that for the Raw shooter, some Canon DSLRs—the 20D, 30D, and XT/350D for sure, and I suspect the XTi/400D and 40D—have a sensor that saturates early at ISO 100, limiting highlight headroom slightly. It's not much—maybe 1/3 stop—but using ISO 200 will give you just a pinch more highlight headroom in Raw on those cameras. (The JPEG conversion doesn't use those extreme highlights and so is unaffected.)

[Cody21]

It's all about the "triangle" - the Shutter Speed, the Aperture, and the ISO ... the 3 of them work together for your final exposure. If you need a higher (e.g., f/22) aperture, you may not be able to get a fast enough shutter speed, hence you would up the ISO to compensate for the otherwise slow SS thereby allowing you to get a faster SS.

Suggestion: buy Bryan Pertersen's "Understanding Exposure". on Amazon for like $15. This book will REALLY bring it all together for you ... at least it did for me... YMMV

[therealmr]

Sensor saturation is a false problem if you're working in the correct color space when converting RAW. Just FYI.

Like Cody said, ISO is only one part of a conglomeration of settings that work together to produce an image. And, there are external factors you don't see in the picture that could affect these (like doug said, polarizer, desire for increased shutterspeed, plus the fact that its true - Canon's noise at 100-200 looks about the same).

[Doug Pardee]

Sensor saturation is a false problem if you're working in the correct color space when converting RAW. Just FYI.

Not on certain Canon DSLRs. The sensor used in the 20D, 30D, and XT/350D is not capable of reaching the full "4095" value when set at ISO 100.

It's not the colorspace. The Raw data simply doesn't go all the way up because the sensor fully saturates. Again, it's not a huge amount—maybe 1/3 stop. But the limitation is there, and is easily observed by taking grossly overexposed photos (all white) at various ISO settings and bringing the Raw files into DPP. At all ISOs except 100, the spike will be at the right edge of the Raw histogram. At ISO 100, it'll be a bit short of the right edge.

[therealmr]

Interesting... interesting. Anywhere I can see this, or should I just take your word for it? :)

[Doug Pardee]

Interesting... interesting. Anywhere I can see this, or should I just take your word for it? :)

You can try it yourself. I already gave the instructions. It's a simple test to run. On my XT/350D, maximum Raw output is about 3150 at ISO 100, or about 0.35 stops below the maximum of 4095 that I get at other ISO settings.

Here's a visual sample (http://forums.dpreview.com/forums/read.asp?forum=1019&message=21743848), from a 20D, showing the difference in tonality range between ISO 100 and ISO 200. The ISO 100 photo on the left is blown in extreme highlight areas where the ISO 200 photo on the right is still showing contrast.

[tonylong]

You can try it yourself. I already gave the instructions. It's a simple test to run. On my XT/350D, maximum Raw output is about 3150 at ISO 100, or about 0.35 stops below the maximum of 4095 that I get at other ISO settings.

Here's a visual sample (http://forums.dpreview.com/forums/read.asp?forum=1019&message=21743848), from a 20D, showing the difference in tonality range between ISO 100 and ISO 200. The ISO 100 photo on the left is blown in extreme highlight areas where the ISO 200 photo on the right is still showing contrast.

That's an interesting observation -- I know that many if not most Nikon DSLRs are designed to saturate at ISO 200. I haven't read specs from Canon about this as a deliberate design. It may be, or it may be a characteristics of particular sensor classes. It would be good to know how this applies to, for instance, the 5D, the 40D and the 1D cameras.

[Doug Pardee]

it may be a characteristics of particular sensor classes.

It is indeed specific to the sensor design. Essentially the same 8 Megapixel APS-C CMOS sensor is used on the 20D, 30D, and XT/350D, and that sensor has this characteristic. I don't know if the same thing occurs in the XTi/400D, 40D, and the upcoming XSi/450D, which use similar but different APS-C CMOS sensor designs.

It would be good to know how this applies to, for instance, the 5D, the 40D and the 1D cameras.It's simple to test. As far as I know, this limitation has not been seen in Canon's APS-H and "full frame" sensors.

[mabas9395]

You guys seem to be pretty good with my dumb questions in the past. So what values are you refering to when you talk about the "full 4095 value" or "max RAW output is about 3150"? I've never heard of values like this before.

[Doug Pardee]

what values are you refering to when you talk about the "full 4095 value" or "max RAW output is about 3150"?

On 12-bit sensors, the range of a Raw data point is… well, 12 bits. It goes from 0 to 4095.

Similarly, in 8-bit file formats such as JPEG the range of the data is 0 to 255.

On the new 14-bit sensors the range is 0 to 16383.

1 bit: 0-1
2 bits: 0-3
3 bits: 0-7
4 bits: 0-15
5 bits: 0-31
6 bits: 0-63
7 bits: 0-127
8 bits: 0-255
9 bits: 0-511
10 bits: 0-1023
11 bits: 0-2047
12 bits: 0-4095
13 bits: 0-8191
14 bits: 0-16383
15 bits: 0-32767
16 bits: 0-65535

[Thelonius]

I'm pretty unsteady so I keep my cameras at ISO 400 and adjust accordingly.

[mabas9395]

That makes sense. Thanks.

On 12-bit sensors, the range of a Raw data point is… well, 12 bits. It goes from 0 to 4095.

Similarly, in 8-bit file formats such as JPEG the range of the data is 0 to 255

[E-K]

You can try it yourself. I already gave the instructions. It's a simple test to run. On my XT/350D, maximum Raw output is about 3150 at ISO 100, or about 0.35 stops below the maximum of 4095 that I get at other ISO settings.

I'm not seeing the same results with my 350D.

For my first test, I tried a gray card being lit by a single CF bulb. I took a long exposure of 30s and to ensure the image was grossly overexposed.

What I saw was that the green channel was pretty much pegged at 4095, the red channel averaged around 4070 and the blue channel averaged around 3500. Keeping the same exposure by increasing ISO and halving the shutter time resulted in RGB averages of about (4095, 4095, 3200) for ISO 200.

In the second test, I used a flash at full power instead of the CF bulb. RGB averages were about (4074, 4084, 4095) for ISO 100 and (4095, 4095, 4095) for ISO 200 (although I did not adjust the exposure for this test which I should have).

I'm not saying that this means that there isn't a slight reduction in the dynamic range at ISO 100 versus ISO 200, just that ISO 100 seems capable of providing values in the 4095 range.

e-k

[Longwatcher]

Certain cameras have a base ISO setting, some of Canon cameras are at ISO 200 and some are at 100, this is the stting that they will provide the best quality as far as SNR is concerned.

I just can't remember which are which.

[E-K]

If I remember correctly, the Canon dSLRs with a base ISO of 200 were CCD based (e.g. the original 1D). As far as I know all of the CMOS dSLRs were a base of 100.

e-k