ISO and 60D

[bobn2]

I need to shoot some fresh test shots. I'll push some that way and let you know how I go. Thanks a bunch.

It can be fun as well, though the PP strategies you adopt change.

They sure do.

Here are some test images. It's not a great scene but it illustrates what I've generally found.

This one is ETTR at 1/80th at 6400, processed in ACR with default settings (no NR or sharpening), and resized in PS without sharpening:



The appearance is realistic enough so I've used this image as my baseline - I tried to get the other images to match the tonal range.

This one is 1/80th at 1600, processed in ACR as above with Exposure = +2, and resized in PS:



The colour is similar, it has a bit more contrast, and appears noisier (easily fixed).

1/80th at 1600 again, processed in DPP with default settings (no NR or sharpening), Brightness = +2, and resized in PS:



Very similar to the 6400 shot, with slightly less apparent noise in the full size image.

1/80th at 1600 again, following your advice to do an uncorrected linear conversion to 16bit TIFF (I used DPP with Brightness = 0), and processed in PS:



To get the RGB levels similar to those of the 6400 image in all three areas of the WhiBal card, I used Levels with highlights input level = 32 to get a similar white point, midtone input level = 2.2 to get a similar gamma, then a curve with one point at input = 100, output = 128, to brighten the middle grey.

Contrast is low. The colour is not too bad on the WhiBal card and the picture, but it's a complete mess in the shadows. Because of the difference in appearance I'm not convinced that the apparent noise is any lower than in the other images.

But what if we do some correction in the raw converter? 1/80th at 1600 again, DPP with Brightness = +2, and processed the TIFF in PS:



RGB levels equalisation required highlights input level = 128, midtone input level = 2.2, and the same 100/128 curve.

This one is much better than doing all the correction in PS, but not better looking than any of the other three shots due to the aberrant colour in the shadows.

Thanks for all that, a very useful series and the first I've seen looking in detail at the processing implications, really useful.

So at least with the way I've approached it (trying to match the levels on the WhiBal card), shooting at 1600 and pushing doesn't give a decisive noise advantage, and there is a significant penalty in shadow colour.

I don't think anyone is proposing a significant noise advantage (just a slight one down to higher precision arithmetic) , the major advantage is highlight headroom.

I suspect you're doing something quite different to get a normal-looking image from the unpushed linear TIFFs. Would you tell me more about what you do Bob?

I do pretty much what I said, but get colour balance by eye (but I'm not very good at it)

I'll make these raw files available if you'd like,

Please, that would be great.

or we could work on some of yours... Thanks.

I don't have such a good example, and it would be nice to stick to the one to get as many variations as possible. But let me have them and I'll have a play.

Edit: the link to the first is broken, BTW.
--
Bob

 

[WilbaW]

Thanks for all that, a very useful series and the first I've seen looking in detail at the processing implications, really useful.

No worries. I always learn a lot doing things like that.

So at least with the way I've approached it (trying to match the levels on the WhiBal card), shooting at 1600 and pushing doesn't give a decisive noise advantage, and there is a significant penalty in shadow colour.

I don't think anyone is proposing a significant noise advantage (just a slight one down to higher precision arithmetic) , the major advantage is highlight headroom.

Sure, not imply anyone is, just stating that any such advantage is not in itself enough for me to justify using the strategy.

Now, this "highlight headroom" thing. When I hear that term I think of the margin of error you have between blown highlights in the JPEG image and saturation in the raw data. For instance, with the settings I use on my 450D, I can generally go 1 to 1.33 steps over the perfect ETTR exposure and still recover those highlights from the raw data, so my "highlight headroom" in that case is 1 to 1.33 steps. Is that what you mean, or is it more about the quality of the data in the highlights rather than absolute loss?

But if I can be sure that I'm not reaching saturation in the raw data, I don't need any more headroom than that...? Am I missing the point?

I suspect you're doing something quite different to get a normal-looking image from the unpushed linear TIFFs. Would you tell me more about what you do Bob?

I do pretty much what I said, but get colour balance by eye (but I'm not very good at it)

Yeah, but you didn't say much.

What steps do you take to get a normal tonal range from a linear converted 16bit TIFF? (Like, I pulled down the white point and increased the gamma value in the PS Levels dialogue, and so on.)

I'll make these raw files available if you'd like,

Please, that would be great.

No worries (big files) -
http://www.users.on.net/~alanw/DPReview/1600-80.CR2
http://www.users.on.net/~alanw/DPReview/6400-80.CR2

Edit: the link to the first is broken, BTW.

Yeah, sorry, I don't know how I missed that. Fixed now.

 

[pekr]

Well, AUTO is highest value can be set. I usually keep it at 800 value, because that is the value I still regard not compromising the quality. ISO 1600 might be bearable at some situations too.

But - Canon uses strange alghoritm. I found it using e.g. 1/60 and ISO 400. 1/60 in dimming conditions? Why? I would really welcome, if Canon would add the lowest required time setting along to AUTO ISO, as Nikon has.

Simply put - it works, works nicely, but check what camera mettering gives you as a resulting time. It might be tricky sometimes. Or - you your flash.

 

[bobn2]

Thanks for all that, a very useful series and the first I've seen looking in detail at the processing implications, really useful.

No worries. I always learn a lot doing things like that.

So at least with the way I've approached it (trying to match the levels on the WhiBal card), shooting at 1600 and pushing doesn't give a decisive noise advantage, and there is a significant penalty in shadow colour.

I don't think anyone is proposing a significant noise advantage (just a slight one down to higher precision arithmetic) , the major advantage is highlight headroom.

Sure, not imply anyone is, just stating that any such advantage is not in itself enough for me to justify using the strategy.

Now, this "highlight headroom" thing. When I hear that term I think of the margin of error you have between blown highlights in the JPEG image and saturation in the raw data. For instance, with the settings I use on my 450D, I can generally go 1 to 1.33 steps over the perfect ETTR exposure and still recover those highlights from the raw data, so my "highlight headroom" in that case is 1 to 1.33 steps. Is that what you mean, or is it more about the quality of the data in the highlights rather than absolute loss?

But if I can be sure that I'm not reaching saturation in the raw data, I don't need any more headroom than that...? Am I missing the point?

I suspect you're doing something quite different to get a normal-looking image from the unpushed linear TIFFs. Would you tell me more about what you do Bob?

I do pretty much what I said, but get colour balance by eye (but I'm not very good at it)

Yeah, but you didn't say much.

What steps do you take to get a normal tonal range from a linear converted 16bit TIFF? (Like, I pulled down the white point and increased the gamma value in the PS Levels dialogue, and so on.)

I'll make these raw files available if you'd like,

Please, that would be great.

No worries (big files) -
http://www.users.on.net/~alanw/DPReview/1600-80.CR2
http://www.users.on.net/~alanw/DPReview/6400-80.CR2

Edit: the link to the first is broken, BTW.

Yeah, sorry, I don't know how I missed that. Fixed now.

Thanks for that, I'll get on it some time. Not right now, I have 500 wedding photos that take priority.
--
Bob

 

[samWebster123]

But - Canon uses strange alghoritm. I found it using e.g. 1/60 and ISO 400. 1/60 in dimming conditions? Why? I would really welcome, if Canon would add the lowest required time setting along to AUTO ISO, as Nikon has.

Yeah I would really like to have the ability to set a min shutter speed in Auto ISO like the Nikon system. It's a fantastic option on the D3s and others but definitely a less than ideal implementation on the Canons!