Jump to content


ChipScar

Member Since 13 Jun 2004
Offline Last Active Dec 21 2008 08:01 PM
-----

Topics I've Started

D3 vs D3x Noise Comparison: First Tests

21 December 2008 - 07:45 PM

The D3's low noise signature is perhaps its most appreciated feature - especially for those of us who shoot it extensively underwater. Now comes the D3x, with roughly double the size of the sensor. We all know the key issue this raises - more but smaller pixels crammed onto a sensor of the same physical size as that of the D3 are theoretically bound to make for more noise. The only question is ... how bad is it?

With early adopters like myself having gotten the camera only yesterday (December 20, 2008), we don't have any answers to this question in the real world of underwater photography. However, we can make a few simple tests above water to get preliminary answers.

First, let me say that I DID NOT try to judge how much better the D3x was in terms of sharpness or resolution. Everybody and their brother will be doing this over the weekend, and I can assure you that casual inspection of side-by-side images made with both cameras, the same lens, etc. etc. showed that the resolution of the new sensor has ... well ... significantly higher resolution than that of the D3. I'll let the real camera geeks tell us just how much higher, but it's enough for me to a) feel good about paying a ridiculous price for the camera, and b) want to preferentially shoot this camera underwater, if the noise signature can be tamed. Note the word "tamed". More about this later.

If you would like to see the actual images I obtained from my tests, I suggest you go to the following link, which will take you to one of my older websites. I have uploaded all of the images there, and you can read everything that I’ve posted here in the proper sequence.

Pbase D3 vs D3x Noise Comparisons

Basic Methodology - D3 vs. D3x Direct Noise Comparison

My methodology to assess the D3x’s noise signature was pretty simple. First, I made a variety of similar images using a tripod, and a new Nikon 24-70 mm f2.8 lens set at 70mm. The idea was to vary the ISO on each camera, shoot on aperture priority (f5.6, just to make it interesting) letting the shutter speed adjust as required by the change in ISO, but to have everything else stay the same. The original reference image was of a backyard scene.

For the initial direct comparison test, I turned off just about everything I could in both the D3 and D3x. In short, there was NO in-camera noise reduction or sharpening applied. As it turns out, this is NOT the way I think we'll be shooting the D3x (are you still interested in this little test?) but it's the way many of us have shot everything from the D100 through the D2x to the D3. At least underwater.

I made similar images with each camera at ISO 100, 200, 400, 640, 800, 1000, 1250 and 1600. I'm sure some terrestrial photographers are checking out ISO 3200+ even as we speak, but for underwater photography, I personally try to keep the D3 at 640 and below if possible. I'll ramp up to 800 if required without too much angst, and I never shoot above 1250 even under dire circumstances (like Alcyone at 7:30 am in the rain). It's all about the noise, even with the magnificent D3. So, my personal shooting style dictated my ISO test parameters.

The biggest issue I faced this morning was how to process the RAW images. As we all know, Nikon doesn't even give you a raw converter in exchange for your $8000 (I plan to throw red paint on the Chairman of Nikon and shout "Shame! Shame!" if I ever meet him). Besides, I hate Capture NX2. However, the folks at Adobe are still struggling to keep ACR up to date with the raw conversion coding for every bloody new point-and-shoot that's come out in the last quarter, Phase I is behind as well, forget about Aperture, and even Bibble isn't ready for the D3x. Which led me to discover a rather marvelous little raw converter called ... you guessed it ... "Raw Converter", by Iridient Digital, Inc. For $125 I downloaded a license that is good for 15 months, and opens D3x raw files just fine. If fact, it may be hard to get me back to ACR. This is a very sweet little raw converter. But that's another story .... although I would suggest you check it out when you can:

Raw Converter Website

In any event, to keep everything as consistent as possible, I turned off everything I could in Raw Converter, INCLUDING SHARPENING AND NOISE REDUCTION. I also used all the default settings. NOTHING was optimized. I basically did batch conversions for all the images at once.

After doing the raw conversion, I cropped to the red rectangle and converted (via Raw Converter) each image to a TIFF. The images from each camera were then sized to 345 x 345 pixels and placed in conjunction with the corresponding images in Photoshop CS4. There was a little downsampling required for the D3x images compared to the D3 images – I did that in Photoshop using the bicubic sharpening sampling protocol (found in the image size dialog box).

The resulting images were very usable, especially when viewed at 100 - 200% on my screen. Unfortunately, when I first uploaded all of these images, I found that at the size I prepared for my original post, you really couldn’t see the noise patterns very. So, for purposes of this test, I rezzed up each image in Alien Skin Blowup 2 to 1600 pixels across the horizontal margin. This makes for very nice, big, juicy images that show the noise patterns well.

Results

Part I. D3 vs D3x Direct Comparison

The direct comparison images are shown in the gallery here:

Part I

With every noise reduction aid (and sharpening) turned OFF in both cameras, there is no doubt that the D3x is noisier than the D3. This can be seen on inspection, and also by quantitating the amount of noise using Noise Ninja (run at default settings). Again, I’m sure some of our noise geek friends will have ways of adding precision to all of these measurements I’ve made, but for real world purposes, I think we could say that the amount of noise generated by the D3x sensor (IN THIS TEST) is roughly the same as the D3 at ISO 100, about double at ISO 640, and about 3x at ISO 1600. Mind you, those are multiples of what is a small noise signature to begin with. But they are multiples.

Part II. Effect of In-Camera Noise Reduction

If you read everything said by Nikon about the D3x and noise, there has always been a discussion about improved NR routines built in to the camera. To test this claim in a rough and ready manner, and more importantly, to see if the in-camera NR could significantly ameliorate the noticeably greater inherent noise signature of the D3x, I re-ran the tests several hours later with the D3x in-camera noise reduction turned to “Low”, “Normal” or “High”. I didn’t repeat the D3 tests in a similar fashion – my own experience is that in my real world usage (e.g. underwater) the D3 in-camera NR isn’t very helpful until you go above ISO 1250, and then comes at the cost of a meaningful loss of sharpness. Put another way, if I could get the D3x noise signature down to that of the unaided D3, I would be a very happy camper.

The results again speak for themselves:

Part II

At ISO 640, 800, 1000, 1250 and 1600, for which I have made comparative images, turning on the in-camera noise reduction is VERY HELPFUL … especially at the higher ISOs. Furthermore, it doesn’t look like it makes much difference, under my test conditions, whether the in-camera NR is set to low, normal or high … either in terms of noise reduction or loss of sharpness. In fact, to my eye, there is little or no loss of sharpness in any of these NR’d images. I’m sure the guys who do this for a living will have expert commentary on this within a few weeks, but at first blush, it all looks like the D3x is very usable within the ISO range that I’ve tested, so long as you leave some degree of in-camera noise reduction turned ON. I’m sure purists will moan and whine about this, but I don’t really care how we get there so long as the final result meets my admittedly high standards.

Part III. Effect of Noise Ninja on Images Captured with In-Camera Noise Reduction Turned On

As my final initial look at the issue of noise and the D3x, I processed several sets of images captured at ISO 800 and ISO 1600 with a) low, normal or high in-camera NR turned ON, and b) then subjecting the resulting images to post-processing noise reduction via the Noise Ninja plug-in in Photoshop CS4. The results were again very gratifying:

Part III

The already improved noise signature associated with in-camera NR turned ON could be further improved in the final image, without adding any significant loss of sharpness, by running the Noise Ninja plug-in filter in Photoshop CS4. Quantitatively, this strategy brought the noise signature at ISO 800 and 1600 to very low levels normally associated with ISO 400 and below, and almost comparable to that achievable with the use of Noise Ninja in low ISO images captured by the D3.

The attached table shows the noise signatures as reported by Noise Ninja’s profile of each image. Again, Noise Ninja was run in full default mode, with NO tweaking of any image.

Noise Comparison Chart

Final Thoughts and Comments

Well, hot diggety-dog. The D3x is, as expected, inherently noisier than it’s noise-optimized good twin, the D3. However, it appears to me that Nikon wasn’t just blowing smoke when they claimed that the noise processing engine in the D3x has tamed much of the noise generated by the more densely-packed D3x sensor.

With in-camera NR turned on, and under these limited test conditions, I found the “noise-to-resolution” tradeoff to be strongly in the D3x’s. I also found that using Noise Ninja in a post-processing environment had additional benefit, seemingly with – to my untrained eye – very little further loss of sharpness. I would also comment that performing additional noise reduction in the raw converter of your choice may also yield very satisfactory results in lieu of using Noise Ninja … we’ll have to wait and see once all the engineers have enabled their converters to handle D3x files.

As for the underwater world … this will be the real test of the D3x vs. the D3 for myself and my colleagues. Digital cameras optimized for terrestrial photography (e.g. ALL of them) show their worst sides from a noise perspective when taking essentially monochromatic images of blue water that has not-so-subtle light gradients and lots of junk floating around. The D2x was a big improvement on the D100 in making underwater images, and the D3 was an even bigger improvement on the D2x. I suspect that when we all get the D3x underwater in the deep blue wide-angle world, we’ll find we have to give up a smidgen of noise and smoothness of those pesky blue and cyan gradients, but will get the great benefit of being able to crop down our wide-angle images for compositional purposes, and still be able to print at huge sizes. That’s worth the trouble and, for some of us, the expense of this new and promising camera.

As for macro photography, where you have control of the light, can shoot at ISO 100, and are on top of your subjects … hoo boy! I predict those 3ft by 5ft metallic prints of pygmy seahorses will never look more dazzling than when captured on this remarkable camera.

Chip Scarlett
John.scarlett@tercica.com