|
|
|||
|
||||||||||||||||||||||||||||||||||||||||
 Test Scene |
The EOS-1D allows you to choose from 1/3, 1/2 or 1 EV steps betwen ISO 200 and 1600, at 1/3 EV steps this provides a total of 9 different ISO equiv. sensitivities. Additionally there are two extra sensitivities accessible through custom function 3; ISO 100 and ISO 3200. At ISO 100 dynamic range is less than 'normal' (ISO 200 - 1600), at ISO 3200 noise levels are quite high.
A full comparison of noise levels versus other digital SLR's is available in the 'Compared to' section of this review.
To give an impression of noise levels at different sensitivities the same scene was shot at 1 stop (1 EV) steps from ISO 100 to ISO 3200 in relatively low light (EV 5). We're also compariing the difference when using the EOS-1D's noise reduction option. The crops shown below are magnified 200% (nearest neighbour).
Settings: Parameters: Standard / Color matrix: 1 (sRGB)
| Noise Reduction: Off | Noise Reduction: On |
 |
 |
| ISO 100, 1.6 sec, F11 | |
 |
 |
| ISO 200, 1/2 sec, F11 | |
 |
 |
| ISO 400, 1/3 sec, F11 | |
 |
 |
| ISO 800, 1/6 sec, F11 | |
 |
 |
| ISO 1600, 1/13 sec, F11 | |
 |
 |
| ISO 3200, 1/25 sec, F11 | |
As you can see Canon has done a good job of keeping noise levels low. All the way up to ISO 800 skin tones remain relatively smooth with only a smattering of slight noise. At ISO 1600 noise is more visible but doesn't affect the overall look of the image, much of this noise would either not be visible in print or would look like slight film grain. At ISO 3200 things are a bit more extreme, I'd say it's only for those emergency situations. Interesting to note that the EOS-1D's noise reduction system does nothing to reduce ISO noise, indeed looking at the results above you should remember to DISABLE it unless you intend to shoot low ISO extremely long shutter exposures.
Many people have asked why the EOS-1D's ISO 100 sensitivity can only be used after enabling a custom function (therefore it isn't a "standard" sensitivity). I asked Canon this question during my Q&A session and they explained that the CCD has certain performance specifications. Using it at ISO 100 places some of those specifications outside the recommended range.
What does this equate to? Well, it's clear that ISO 100 is very clean (no noise) but in use it was also clear that ISO 100 does not have the same dynamic range (ability to capture a wide range of dark and light) as the rest of the sensitivity range. This is illustrated very nicely by the crops (below) from our test images (above).
|
|
| ISO 100 | ISO 200 |
As you can see, in the ISO 100 shot a large amount of the flower and candle wax has been 'blown out' (overexposed), where as at ISO 200 the detail and tone has been maintained.
 Test Shot |
There has been quite a lot of talk about the type of noise exhibitted by the EOS-1D (we're only talking about high ISO sensitivity here). And while it's lower than most of its competitors it does appear to take on the guise of horizontal lines in darker or shadow regions of an image. The four shots below were each taken at ISO 400, 800, 1600 and 3200 wth a deliberate exposure compensation of -2.0 EV. The crop is of a 480 x 240 area of the image (shown at 100%), the bottom half of the crop has been 'pushed' (using levels) approximately two stops in Photoshop to deliberately exaggerate any noise to make it more visible.
Settings: Parameters: Standard / Color matrix: 1 (sRGB)
 |
| ISO 400, 1/30 sec, F8.0 |
 |
| ISO 800, 1/60 sec, F8.0 |
 |
| ISO 1600, 1/125 sec, F8.0 |
 |
| ISO 3200, 1/250 sec, F8.0 |
It's fairly clear to see some kind of horizontal pattern to the noise in high ISO images, for most occasions at ISO 400 and 800 it's not visible without 'pushing' the image. Unfortunately at ISO 1600 and 3200 it's clearly visible, in a good percentage of the ISO 1600 shots I took at the Ice Hockey game (see the samples gallery) banding noise was visible in the darker background behind the players. This banding was both visible on-screen and in print.
Many people have looked through these samples and commented that as long as you don't underexpose the shot you won't see the banding. This is simply not true. The exposure is only related to the amount of 'signal' (brightness) present. Noise is visible where the signal is low and the noise can become visible, I deliberately used a black background in the shots above so that we can 'measure' the noise, clearly where there is a strong signal (bright area) this will overpower the noise and it won't be AS visible.
Thus the banding is only visible in medium / dark or shadow areas of the image. The trouble with this is that you're typically metering for the subject, not the background, and in many cases (such as sports events) the background will be several stops darker than the subject.
