Thursday, 8 April 2010

still gnawing on that: exploring camera ISO effects

yes, I'm still gnawing on this

The last few bog posts (combined with my being in the middle of packing up my house) have given me cause to gnaw further at this topic of open source vs closed Intellectual Property and what effects the move to digital is having on photography.

Make no mistake, I like digital. I like computers (notice where you're reading this?) and have spent quite an amount of energy getting to know my new media (I really only changed seriously to digital 5 or 6 years ago).

snowyJarviRuokoThis is not to say that I do not appreciate film and continue to use 120 roll and 4x5 sheet substantially in my photography.

I don't use film because I'm a luddite, I use it because I know it will give me the look I want.

How do I know? well by knowing my materials.

A friend of mine over in Soundimageplus seems to have started gnawing the same pithy issue, and came up with the thought:

OK - at the moment Canon offer great choices and good value, but what happens when they eliminate their main rival? Will they be so customer friendly when they don't have to try so hard?
and given that he's a self affesed Lieca fan and a Nikon owner this is no trivial statement (though I think personally he's a card carrying camera agnostic with a strong bent towards impulse shopping, but don't tell him I said that).

Well some time ago I was digging around in the bowels of RAW files produced by the Fuji S5 camera (with an eye to moving towards that) and found they were simply stunning ... even if only (and perhaps particularly at) 6MP. You see their superCCD is really supper ... but any wedding photographer out there reading this already knows that ...

To the point I discovered some interesting points. Canon it seems, (they won't admit anything) do some skulduggery in (probably the analog) signal processing pre writing to RAW on their incremental ISO that results in predelivery of combed histograms in their RAW data. For example:

So, before you introduce any banding into the picture yourself you start out with it. As my surfie mates would say back in the 80's
This was brought to attention on a forum back in 2006 after one programmer who was a photographer noticed the phenomenon. The response from Canon was via Chuck Westfall was this:
Hi, Peter:

I appreciate your interest, but we do not comment on our image processing algorithms. Our cameras are basically offered "as is," and we do our best to make sure that image quality is as high as it can be. Canon's official statement on EOS 30D image quality is as follows:

"The image quality of the EOS 30D and all other EOS Digital SLRs conforms with Canon's internal quality standard at all available ISO speed settings. We have no further comments to offer on this issue."

Best Regards,

Chuck Westfall
Director/Media & Customer Relationship
Camera Marketing Group/Canon U.S.A., Inc.

it was nice of them to discuss this to any extent ... wasn't it.

To be fair how many people have put Kodak or Fuji to task on their film stocks and got very far?

Probably noone (unless you buy 10 Km of 35mm stock at a time ... say in LA somewhere)

Of course the difference (in my view) is that as a film camera user I can say "no I don't like this film stock" and go use another one ... after all, its just a cassette of film I load into my camera isn't it... Heck I could (and did) mix n match brands within my camera system (especially the 4x5 LF one) but with a Digital I'm really trully wedded to a system.

As some may know a divorce is expensive

Now, Peter did some interesting analysis on the EOS cameras back then, and has published his data here. I'd like to take a moment to discuss the significance of the 5D and the the 30D data.

First, this is the data he got of signal to noise ratio of each of the ISO settings of the camera. As always higher signal levels (over noise) is better ... so a higher number is better. We know (or should know) that a camera goes "down" in quality as we push over its "base ISO". On the 5D that is 100 ISO

Peters data supports that notion and further suggests that by using fractional ISO you will drop down significantly in quality. In fact many fractional ISOs are lower in quality than 800, so you may as well set that.

Bet you won't find that in the manual.

This indicates to me that using auto ISO on the camera will quite likely result in lower quality images (assuming its not biased to choose the best ISO ... but then who'd know?)

The situation is even more interesting with the 30D. Where it seems the fractional ISO's are in fact the best with 160 being the best:

and a rapid trail off towards the 1600ISO end (with an advantage to be using 1250). These data are in "counts" which is a linear representation, some people may prefer seeing this as DB so he also presents that:

This just makes it more obvious that you don't use anything past 1600 on that camera (and if you took some test shots you'd probably see that too ... DPReview seems to have)

This is the step between 1600 and 3200 on their review

significant isn't it...

Peter reports
that 100 looks worse than 200 because at ISO 100 the pixels saturate at only 3398 counts and even though the dark noise is lower (StdDev = 2.11) the dynamic range ends up equal to (3398-128)/2.11 = 1550. At ISO 200 on the other hand the saturation is 4095

Thanks mate, re-reading this helps me with the observations I had with RAW testing on my G1 (some time recently).

Pete thinks that the intermediate levels are constructed ones, but looking at the data he's presented here I think that the intermediate levels (like 160) are the real base levels and the "proper" levels (100, 200) are the constructed ones .. perhaps for some fine efficiency tuning.

So as Peter observes:
Considering the above and that the 5D was released in September 2005 it is an impressive achievement and the camera still holds it own against many current (as of February 2008) cameras.
What I see is that technology has not really developed significantly since the 5D (and perhaps before) and that all the new stuff is just about regurgitating the stuff that the subscription editors already had in their pockets to dish out gradually.

the more things change, the more they stay the same


Noons said...

Fantastic post, mate. Great stuff.
Lots of reading there.
Confirming a lot of my suspicions about "raw" pre-processing, not just from Canon either!
I'm tempted to do some of these with my D200, I suspect something amiss there as well with the ISO steps.

Why on earth isn't this information published by the makers? It's in their interest to let folks know where the best conversion is happening!

stefan kozma said...

this is interesting. I read up on some of this kind of stuff back when i was buying my first digital. companies will capitalise on any way they can hide truth to gain profit. still i try not to worry. i doubt raw will ever be raw. film doesnt have hidden / manipulated information.