I was invited by Uwe Steinmueller of Digital Outback Photo to join him in a shooting session with the new Canon 1Ds MK III, and of course this was an offer I couldn't refuse! We met at our usual location for testing new cameras or lenses, an abandoned building in Alviso, CA. We choose this particular building because it offers a multi-detailed surface texture to help ascertain resolution performance as well as a range of lighting and shadow values to help ascertain dynamic range and noise performance. Note that both Uwe and I prefer to evaluate camera performance by making “real” images instead of shooting test targets. I realize my evaluation methods will not satisfy die-hard pixel-peepers as it is not a fully controlled, scientific test. However, I feel it is a fair way to initially compare different cameras since it’s more representative how I use them in everyday shooting situations. Finally, please understand this is not intended to be a complete or even conclusive evaluation, but rather just my initial impressions about file quality. That disclaimer out of the way, here are my initial impressions.
The camera looks just like the 1D3 and the control layout appears to function identically too. I will leave specific feature discussions to others as I did not have enough time to fully explore all the operational aspects of the new body. Suffice it to say, I was able to set everything I wanted with it without having to resort to the owners manual; in other words, this is a pretty intuitive camera to use.
I shot the camera against a Canon 1Ds MK II, and a Canon 5D. All three cameras were shot at ISO 200, 400, 800, and 1600, and a tripod was used for every shot. When I compare cameras, I like to use the same lens on each camera --- and preferably one known to be a good performer. In this case, we used the excellent Canon 135 f2 L lens at an aperture of f8.
Okay, enough babble as I’m sure most of you have already skipped down to the images anyway <GRIN>.
All images presented here are the ISO 400 captures and here is the full frame capture area for reference:
Next the crops, but I need to clarify how I processed them. First off, I used Adobe Camera Raw for all conversions and processed all files identically except for sharpening. I chose “Cloudy” WB as we had a few light clouds in the area and exposure slider was kept at zero. I did however crank fill up to 30 and set blacks to 4 to help better see noise and into the shadows under the roof. I attempted to optimize the capture sharpening settings for each camera and used the following for these conversions: 5D, amount 30, detail 25; 1Ds2, amount 35, detail 30; 1Ds3, amount 50(!) and detail 50! (EDIT! I forgot to mention radius -- on ALL files it was set to the ACR minimum, 0.5 --- sorry for that omission! Also, I should mention I am using ACR 4.3.) I will use this opportunity to mention something that should now be obvious and likely disappoint more than a few of you: the 1Ds3 appears to have a quite strong AA filter, clearly stronger than the 1Ds2’s…
Here is the 1Ds3 crop:
Here is a similar crop from the 1Ds2:
And here is the crop from the 5D:
In looking at these ISO 400 crops, a few things become apparent. First, noise on the 1Ds3 is quite good – especially for a camera of that resolution – and comparable IMO to the 5D’s. Moreover, by cranking detail up to 50 to extract as much as possible, I actually enhance noise to a certain degree which is a disadvantage to the smaller pixels. In my crops, for some reason the 5D (and 1Ds2) appear a bit brighter in the shadows having used the same shadow and fill settings on both, but they are also showing a bit more defined noise; we’d expect this from being brighter, so I'd not infer too much from it. Comparing noise at the other ISO’s with more normal detail settings, I’d say the 1Ds3 appears to have noise characteristics similar to the 5D (the 1Ds2's are a bit higher) which is quite good for a camera of this resolution. While 200 and 400 will be the sweet spots for maximum performance from the 1Ds3, 800 is still quite usable, but in my opinion 1600 starts to get noisy enough to occlude detail, and moreover, showed a blocking up (loss of DR) in the shadows. Overall, I think the 1Ds3 shows very good noise performance for such a high-resolution camera.
As for detail, we can see the 1Ds3 has a very real advantage. Looking at the ends of the roof tiles, peeling paint chips and grain in the vertical piece of bare gray wood, we simply see more texture in each of these than either of the other Canon cameras offer. The telltale is the sand in the first main row old tan roof shingles --- only the 1Ds3 really resolves the granularity of the sand embedded in the surface of them. Obviously, more megapixels are working here. Suffice it to say there is detail to be had. But given the 1Ds2 was better than many Canon lenses, it is safe to assume that to get the most the 1Ds3 has to offer, one will need to use the best lenses available, proper capture technique and a sturdy tripod when shutter speeds dip.
The real question for many will be how significant is the 1Ds3’s detail difference? I’m sure this will be debated rigorously on the forums for the next year at least – and the irony is the answer is probably different for each of us, depending on how we'd use the camera. Nevertheless, I took the liberty of generating a few additional comparative crops from the 5D to let you each draw your own initial impressions about relative performance as it may pertain to you. I present them here with limited dialog:
First is a crop showing the 5D covering the same portion of the image the 1Ds3 captured. It is 700 x 583 pixels compared to the 1Ds3’s 900 x 750 above:
Here are the crops super-imposed to show the relative coverage differences:
Here is the smaller 5D crop above uprezzed to the same size as the 1Ds3:
And for comparative viewing convenience, here is the 1Ds3 crop again:
Finally, here are the two same-size crops viewed at 50%. I have found this 50% view to be a simpler and accurate method compared to the screen captures I usually post for conveying how detail might compare in a final print. It’s obviously not a perfect method, but coupled with the other crops it’s about the best I can do for you via web view -- again, I leave you to draw your own conclusions. Note that these crops are close approximations to actual sections from prints of about 16"x24":