DIGITAL NOISE AND DYNAMIC RANGE

Basically, digital noise is random dark and light pixels that are out of color range for the area they are found in. Digital noise is caused when the sensor and camera electronics are just at the edge of being able to accurately assign color to pixels. Although there is some random digital noise throughout all digital images, high ISO ratings and low light is what puts digital camera sensors on the edge of being 'noisy'.

In the first series below I used my Canon 50mm f2.5 macro lens and set the cameras to ISO 1600. The lighting was late afternoon partially shuttered window light and the exposure was 1/4 sec at f8. The RAW files were converted to 16 bit TIFF using Capture One Pro. I cropped a full resolution square out of the area below the pillow between the bear's legs. I then added 50 points of brightness control in Photoshop so it would be easier to see image shadow detail and random digital noise.

Surprisingly, the 1D Mark II, with it's larger photosites (lower pixel density) produced more noise than either of the other two cameras. This was not what I expected to see. Every thing else being equal, Larger photosites should be more effective in reducing random noise. Larger photosites should be able to collect light and discriminate colors more efficiently.

The amount of detail seen in the darkest areas was about equal between the 3 cameras. You must remember that the 1Ds Mark II was much closer to the bear and therefore had less depth of field. Also worth noting is the the perspective in the 1Ds Mark II is just beginning to show a nose and mouth area on the bear that is a little out of proportion to his head compared with the 20D photo. The 50mm is a great portrait lens on the 20D and not so good for this application on the full frame camera.

I did a backup image series to make sure the noise levels between the cameras in the first image series wasn't a fluke. I shot a white wall with the 50 macro from about 18 inches away. I focused the lens at infinity so there would be no texture at all in the image. I underexposed the image about 3.5 stops so the resulting image would only show 'dark noise'. I used Photoshop's ACR3 to convert the RAW file. ACR's default noise reduction is a little less aggressive than Capture One's default noise reduction. The 1D Mark II produced the nosiest image once again. Maybe the 1D Mark II I had on loan is not representative in this area. Nonetheless, It looks to me like there's not a significant difference between the 20D's 1.6 crop 3937 ppi sensor and the 1Ds Mark II's full frame 3540 ppi sensor. If you want to try to find a winner here, it just may be the 20D, exactly the opposite of what the mechanics of the sensors would predict. At least with Canon DSLR's it looks to me like the camera's firmware affects noise output more than the size of the photosites.

Mission San Luis Rey, Oceanside California

In the image series below we wanted to find out how these three cameras handled slightly blown highlights. The first 3 image set shows example exposures from the 20D at three different exposure levels at ISO 100.

For the three camera (same FOV) comparison series I used my 50mm macro lens at 1/320 sec, f8 and ISO 100. This exposure might be considered by some to be 1/3 stop overexposed for scene containing an average tonal range. It certainly is on the edge of being overexposed for a scene containing so much white. Some of the more subtle texture in the wall is starting to become less noticeable in this exposure.

20D -  8mp 1.6 crop, full image 1/320 sec @ f8

See full resolution crops from each camera below

In this image series, there's really no difference between the three cameras. If you want to split hairs, The 1Ds Mark II may have edged the the 1D Mark II by a couple of color points and the 1D edged out the 20D by one color point. Hardly worth noting. If photosite size matters for holding highlight detail, it certainly looks like the playing field gets leveled in each camera's firmware.

ARTICLE SUMMARY, CONCLUSIONS AND MORE OPINIONS

The 20D sensor's higher input resolution (higher pixel density) only helps with telephoto photography when your object is to fill the frame with your subject from as far away as possible. If you can zoom in, walk closer or get a longer focal length lens and fill the frame (same FOV), the 1D Mark II is on equal footing and the 1Ds Mark II's 16.7 mp resolution will rule.

When you are shooting normal to wide angle same field of view landscapes, the 20D is at a slightly noticeable disadvantage compared to the larger sensor equipped 1D Mark II. Again, when you have the same output resolution and the same FOV, pixel density is not a factor. The 20D is using the lens's central optical sweet spot. However, it looks like this factor is not enough to overcome the slightly lower optical performance of the wider angle lens and smaller image components projected on the 20D sensor. The 1D Mark II edges the 20D out in this application because it needs a little less wide angle focal length for the same FOV. The image components are larger on the 1D Mark II sensor and lens performance at the slightly longer focal length is slightly better. Again the 1Ds Mark II easily beats the two smaller sensor equipped cameras for this type of application.

Pixel density, at least in this comparative series of images, doesn't seem to have had the expected relational effect on random digital noise and dynamic range. The three cameras did equally well with slightly blown highlight reproduction. Even more confusing, the 20D seems to have edged out both the 1D Mark II and the 1Ds Mark II in producing slightly less dark area noise than either of these cameras. The 1D Mark II with its larger photosites should have been the winner here but turned out to be the worse noise producer.

Canon's newly announced 5D 12.7 mp DSLR probably would have compared favorably to the 1Ds Mark II if it had been included in this article. It has plenty of resolution and a full frame sensor..

My guess is that Canon will ultimately drop the 1.3 crop sensor camera from it's lineup. You can get a full frame 12.7mp prosumer DSLR now for slightly over $3k. That's less than half the price of two years ago for a Canon full frame DSLR. I Think Canon will move the sports and action oriented 1D Mark II up to full frame in that camera's next iteration. Due to manufacturing costs I suspect that the 1.6 crop cameras will be around for quite some time to satisfy the amateur and prosumer market needs. However, I seriously doubt canon will design a parallel line up of EF-S lens designed specifically for the 1.6 crop size sensor. In fact, Canon seems to be paving the way for acceptance of full frame sensors with some of the content on their 'World of Canon CMOS Sensors' website. Remember, this is just my opinion, I don't really have a crystal ball to see into the photographic future... ;^)

ARTICLE LINKS:
PAGE 1 - INTRODUCTION, TERMINOLOGY EXPLANATIONS AND PIXEL DENSITY WITH TELEPHOTO LENSES
PAGE 2 - DEPTH OF FIELD, PERSPECTIVE AND SAME FOV WITH TELEPHOTO LENSES
PAGE 3 - FOV AND LENS RESOLVING POWER WITH NORMAL TO WIDE ANGLE LENSES
PAGE 4 - DIGITAL NOISE AND DYNAMIC RANGE