Idaho Airships, Inc.

There is-as can be expected-a bit of rivalry between Nikon and Canon. Frankly, that’s good for the industry and good for the consumer. The recent flagship releases of both lines (Nikon D3, Canon 1DsMIII) has shuffled the deck a bit.

The Nikon D3 is a full-frame model with stellar (from what I’ve been able to gather) high-ISO performance. The Canon 1DsMIII continues as the spatial resolution leader, and with its very advanced 14bit A/D (Nikon shares the bit spec) is most likely the color leader as well…but that remains to be seen.

Please keep in mind that you’re not just shooting with a camera. You’re also shooting with a filter, a lens, a technique, a selected aperture (lens performance varies greatly with aperture), under specific ambient conditions (light, temperature)…and that what comes out of the camera is not what you deliver to your customer (except in rare circumstances…I do have customers who insist on the RAW).

Therefore, when the spatial advantage of the Canon 1DsMIII or the high-ISO noise levels of the D3 are recognized, they should be kept in context as only…apologies for combining a bad pun and cliche’ in the same phrase…part of the picture.

I ended up sticking with Canon for this round, in part because I’ve got 15 lenses, flashes, and other assorted hardware to explain away to my wife, who is a degreed accountant (now a full time homeschooler) and not particularly sympathetic to what she interprets as frivolity. On top of that math we have two complete camera sets of two bodies each-we’re pretty well rooted in Canon. Still…I can’t help but compliment the D3 at half the price of the Canon ($4K vs. $8K). Each. Ouch.

The only handheld work I do (aside from the occasional snapshooting) is my aerial work, and I do have an established specialization in low light (early morning, late evening, and nighttime) aerials, so the high ISO question is critical and was not particularly easy to navigate around.

Why noise is important.

So, what do we really have in the 1DsMIII for high ISO performance?

I just profiled my first body, and I’ll share some of the process images here. I noise-profile my cameras because I have a rather sophisticated digital noise application that allows me to set up profiles for various ISO’s and batch-process my images when desired, or fine-tune the noise removal when appropriate. It works in the YCrCb schema and is very, very impressive at both noise mitigation and detail preservation. I strongly recommend the procedure-there are a number of such applets available.

When you view these images, please keep in mind that these targets are designed to capture and measure noise. They look rotten for a reason-quite a bit worse, to my experience, than an actual photograph of some thing or another. I tested with neutralized camera settings and at 67 degrees Farenheit (19.5C)-these images are from the ISO3200 (”H”) series. They are blurred by necessity-we wouldn’t want to pollute our noise with detail! They are also 100% crops of a much larger “patchwork” target of various colors.

The 16 bit .tif after fully neutral conversion from RAW, converted to the 8 bit sRGB .jpg necessary for display on your monitor.

Yeh, it’s ugly, but keep in mind that the purpose of this production is to feature noise. Relax.

As processed for noise, 16bit, aRGB, and converted for display as described above.

Just as an anecdotal exercise, imagine this as a 100% scan of Porta 160NC with no adjustments. Considering that this is still an image produced to feature a noise signature, it’s stunning. True, we’re not necessarily testing the software, but in this case it is part of the process, is batchable, and is not singular to Canon. We’ve come a long way in 4 years.

“Fine Tuned” with manual controls.

One of my more arrogant traits is that I don’t necessarily think that algorithms can optimize matters as well as the human mind, so as a control I fiddle with settings to see if I can “improve” the automated results. In this case, I could not appreciably increase the quality of the noise resolution even with a 9 band per Channel “Equalizer.” Thus, I trust the automated result.

While my results are anecdotal, it isn’t hard to relate to the result. This 1DsMIII is an order of magnitude “cleaner” at high ISOs (1600, H/3200) than any of my other DSLR’s, which are themselves quite remarkable.

One last word of caution in comparing noise signatures-it’s very difficult to do, whether using stops or decibels. Objective quantification is best left up to calibrated equipment in a controlled environment (these tests were produced under such conditions). A full stop is actually twice as much electromagnetic energy but we humans consider it a perceptible but minor increment. Eight or sixteen times the amount of camera (luma + chroma) noise might not be nearly as drastic as the linear math suggests.

L