High ISO Stuff!!

Hey All!

We all go around and around about the different cameras and their high ISO performance, and it can be educational and entertaining!

So, I got inspired today to take the time to run some tests and evaluations using my three older bodies, the 30D, the 5D Classic, and the 1D Mk III.

The purpose was not to show "awesome" high ISO performance, because newer bodies/technology has led the way, and of course the upcoming 1Dx is slated to be the "awesomest"!

What I wanted was to give a glimpse at the effect a high ISO has on images that are properly exposed, and then to show the effects of increasing the exposure by one stop of light by opening the aperture. So I shot with each at its highest "native" ISO -- 1600 for the 30D and 5D, then 3200 for the 30D. And then, I went One Step Beyond, and shot with the cameras at their "High" ISO setting, so a simulated ISO 3200 for the 30D and 5D and 6400 for the 1D3.

I showed that the fully framed images from all three cameras at Web viewing size look quite nice! And then, to the "dirty work": I show sets of 100% crops of two different areas with combinations of shadow and detail to compare/evaluate noise and detail handling.

Here are a couple of the sets -- you may be surprised at how "un-amazing" the differences are!

All three at ISO 1600, 100% crops:

Sorry, here's the link to the PBase gallery:

http://www.pbase.com …_2011_real_worl​d_high_iso

I'll need more time to study that, but at first look you've done a terrific job, Tony.

I hadn't realised what a difference it makes if you go to a higher ISO and use a "full" exposure rather than pushing the exposure at a lower ISO...am I correct in deducing this?

Canon 1D4 vs 1D3 noise test

Yeah, I saw that thread, Frank! Sweet stuff from the 1D4, and the 1DX sounds awesome!

Ricardo222 wrote in post #13465873
I'll need more time to study that, but at first look you've done a terrific job, Tony.

Hey thanks, Ricardo!

It's quite a task to do something like this -- it's not taking a bunch of snapshots but it's actually setting up and doing a "real" shoot, and this was with three separate cameras having to be all coordinated! But it was...I don't know that "fun" is the word, but at least it gave me something meaningful to do!

Well, yeah, the amount of light makes all the difference! If you look at the photos, check out the brighter areas, especially on the "boosted exposure" shots, and you will see virtually no noise in those 200% crops, compared to the low-light shadows!

Good stuff Tony

I think a lot of us folks get snickered at for being pretty anal when it comes to pushing the maximum exposure limits. But it really does make quit a difference in reducing noise and therefore being able to retain detail for sharpening at higher ISOs.

Yeah, just pushing the exposure by a stop above "medium" can make a real difference -- those "H" ISO shots could use some selective NR in the shadows, but overall they come out quite nicely! In fact, even though the "normal" exposures could use a bit more, they are still pretty impressive if you do in fact nail a good normal exposure! Even the trusty ol' 30D did pretty well!

You know, I've been really busy for the past couple of days, and my "fancy" setup is all askew, but it occurs to me that I could have pushed those tests a bit more by pushing the exposure to the limits of the Raw format -- ETTR to the max, and finding a way to really analyze the noise difference between a "normal medium" exposure and an ETTR exposure. The 1 stop push I did really wasn't pushing things!

I think the problem with pulling back to much in post can be hue changes, depending on the camera profile one uses. I really prefer to only go 1/3 stop.

windpig wrote in post #13480228
I think the problem with pulling back to much in post can be hue changes, depending on the camera profile one uses. I really prefer to only go 1/3 stop.

Well, as you can see from the crops, there are slightly different hues between the three cameras, but that's a bit of a different issue and I didn't fuss about it! I was more interested in how much you could gain in the signal-to-noise ratio as you let in more light...

It would be interesting to see comparison shots if you could show us what you can come up with!

Hi Tony

I'm going to try an untwisted profile when I get the chance. I've been very curious as to how well they work in not creating color shifts.

That's a great comparison, Tony; and a perfect example of why folks, in general, should worry more about GETTING a proper exposure than about trying to hold their ISO down to reduce noise or about going 'FF' for better noise performance.

Snydremark wrote in post #13480465
That's a great comparison, Tony; and a perfect example of why folks, in general, should worry more about GETTING a proper exposure than about trying to hold their ISO down to reduce noise or about going 'FF' for better noise performance.

Yeah, good things to bear in mind to keep "perspective"!

I do appreciate the incremental increases from the full frame as well as the 1D bodies because, for much of what I shoot it does matter (for example wildlife/bird shooting where I often have to shoot at high ISOs and then crop closely). But for much shooting, I have no problem grabbing my trusty ol' 30D! And, as that project shows, well-exposed fully-framed images at high ISOs can do great from any of the cameras!

tonylong wrote in post #13480504
Yeah, good things to bear in mind to keep "perspective"!

I do appreciate the incremental increases from the full frame as well as the 1D bodies because, for much of what I shoot it does matter (for example wildlife/bird shooting where I often have to shoot at high ISOs and then crop closely). But for much shooting, I have no problem grabbing my trusty ol' 30D! And, as that project shows, well-exposed fully-framed images at high ISOs can do great from any of the cameras!

Yup, which is why I'm dubious that the 1Dx will somehow be quite as good as it's being made out to be or that upgrades to the 5D or 7D will somehow double (or substantially increase) the MP while retaining or improving noise and DR.

Unless someone makes a chip with double the Quantum Efficiency of the sensor I don't see any behemoth improvement that suddenly makes everything else look like a POS.

Check out some of the graphs in here, particularly figure 2. We've dicked around with more or less pixels, slightly better QE, slightly better micro-lenses, different pixel gaps, improved read noise, better filters etc.., but the fundamentals have stayed more or less the same. We need a completely new type of sensor next.
http://www.clarkvision​.com …rmance.summary/​index.html