Best Canon bodies for low light shooting?

[Chris Brennan]

Recently I've found myself shooting some subjects at dusk. I use a 7D with a 400mm f4 DO. I've gotten decent results at ISO 800 and even some at 1600 but am reluctant to go higher with the smaller sensor. Am I going to realize significantly better high ISO performance if I step up to a 1D MkIV or should I be thankful for what I have?

I look forward to your comments.

[Grant Atkinson]

Hi Chris, I shoot a lot of wildlife that almost always wakes up and gets moving just after sunset like African wild dogs, lions and leopards. I shoot a 70-200f2.8, and a 300f2.8. My fiance Helena and I have a 7D, a 1DmkIII, and a 5Dmk2. I use the 7D in the good light of day for it's reach when trying to get close ups or for small subjects. When the light gets low I put it away, and use the other two bodies. Both the 1DmkIII and the 5Dmk2 give me about a one stop improvement over the 7D, and enable me to keep shooting with a much better chance of getting decent results.
I also agree though with Peter Kes advice above, test yours out, it all depends on what you want to do with the images..large prints, computer screen or web usage..
Hope that helps
Grant

[Grant Atkinson]

Hi Chris, what I forgot to add was that although your question was a comparison of the low light performance of the 7D against the 1DmkIV, I responded with my thoughts on two different cameras, though from what I have seen from using the mkIV, it's low light capabilities are similar to the 5Dmk2, or perhaps a little better, depending on which reviews you read...
cheers for now
Grant

[Roger Clark]

Chris,

This is a confusing subject. What is important is lens delivering light and pixel size as projected on the subject. If you equal that, in the Canon line, the 7D has the top performance followed by the 1DIV. You can see this in Figure 10 at;
http://www.clarkvision.com/articles/digital.sensor.performance.summary/

So if you had two different lenses, each with the same diameter, but different focal lengths so that they deliver the same number of pixels on the subject, then the 7D would just barely win out over the 1DIV, the next camera down would be the 50D followed by the 5DII.

This may come as a sock to people, but that is what the sensor data show. The reason experience may be different is that we tend to use the same telephoto lenses on different cameras, and different lenses for landscapes between different cropped cameras. When we use different lenses to equalize field of view (e.g. landscape photo), we are skewing our perceptions. When we use the same lenses on telephoto images, we are skewing our perceptions.

So when one uses a 400 mm f/4 lens on a 7D and complain about noise, switching to a camera with larger pixels will give lower noise per pixel, but less pixels on the subject. Which you prefer,
A) more noise with more pixels, or
B) less noise and less detail,
is for you to decide.

But, if you changed to a 1DIV and 500 mm f/4, you would get about the same detail on the subject, but the larger aperture lens would deliver more light. So to maintain the detail, and improve the signal-to-noise ratio, the key is more light, meaning a larger clear aperture lens (e.g. 400 mm f/2.8 on a 7D).

So changing the camera (and pixel size) will change noise for detail. If you want to maintain detail with lower noise, you need a larger lens. If you buy a camera with larger pixels and want to maintain the detail with lower noise, you need an even larger lens with a longer focal length.

lens, diameter, and increase in light over 100 mm diameter lens:
400 f/4 = 100 mm diameter lens, 1.0x increase
300 f/2.8 = 107 mm diameter lens, 1.14x increase
400 f/2.8 = 143 mm diameter lens, 2.04x
500 f/4 = 125 mm diameter lens, 1.56x increase
600 f/4 = 150 mm diameter lens, 2.25x increase
800 f/5.6 = 143 mm diameter lens, 2.04x increase.

So better low light performance and maintaining detail on the subject (compared to your current lens+body) would be a 600 f/4 on a body with 600/400 = 1.5 times the pixel size, or 4.3 microns *1.5 = 6.4 microns. Cameras with 6.4 micron pixels include the 1DsIII and 5D2. Of course, AF performance and desired frames per second are other factors to consider.

Roger

[Desmond Chan]

Not a scientist here, and it seems to me all these calculations look great, explain a lot. But, when will all those differences...noise and especially the details, etc. become noticeable to our naked eyes, really make or break an image without the viewers going pixel peeping? A 4x6 print? 8x10, 16x24, 20 ft x 30ft or much larger print? Or, for web viewing, resolution of 800, 1280, 1920 or much larger?

Just wondering from the practical side of thing, when those calculations and their results actually matter.

Then there is the application of noise reduction and its effect on the final image.

[Alan Lillich]

Roger,

Thanks for another great science lesson. But I'm not sure if I have the basic physics right. I measured the front elements of my 70-200 f/2.8 and 300 f/2.8, and indeed the ratio of the diameters is 1.54 - close enough to 1.5 to be within the error of my eyeball measurements. Assume same body, same position, just swapping lenses.

The 300 will capture 2.25 times as many photons coming off the subject. But it will also spread those photons over 2.25 times as many pixels. So the per-pixel S/N is about the same, right? But the granularity of the noise compared to subject features is less with the 300 image. So if I then print the subject to the same size, the 300 image will appear less noisy - in terms of noise per subject feature.

Have I got it? Very interesting, I never thought of the noise reduction effect of increased focal length at the same aperture. Makes sense though, noise comes from individual pixels, if you look more closely at them then you'll make the noise more evident.

Alan

[Chris Brennan]

Thank you all for your comments!

For starters, I'm going to experiment more with higher ISOs and see what I can get. I use the Nik Dfine NR software, and will work with it to see how much I can get out of the images I shoot at those higher ISOs.

Roger, my goal is to get more pixels on the subject so perhaps rather than look at another body, I should consider a longer lens... certainly something to think about.

Most of my images I just view on screen but occasionally will print up to 13x19 prints...

[Elliotte Rusty Harold]

Not a scientist here, and it seems to me all these calculations look great, explain a lot. But, when will all those differences...noise and especially the details, etc. become noticeable to our naked eyes, really make or break an image without the viewers going pixel peeping? A 4x6 print? 8x10, 16x24, 20 ft x 30ft or much larger print? Or, for web viewing, resolution of 800, 1280, 1920 or much larger?

Just wondering from the practical side of thing, when those calculations and their results actually matter

If you can get close enough to the bird with a long enough lens to fill the frame, then you're right: all those differences probably don't matter. However, what better high noise performance, etc. gives you is the ability to shoot smaller birds from further away, then crop most of the frame. Sure I'd love to shoot that Mourning Warbler from my 3.5m minimum focusing distance, but realistically I'm going to have to crop. Even if I'm shooting for the Web with a 21 megapixel body and only need about 1024 by 768, sometimes all I can get on the bird is about a megapixel, and then all the pixel peeping details become very noticeable. Practically, the better the low level details are, the longer your lens is.

[Roger Clark]

Roger,

Thanks for another great science lesson. But I'm not sure if I have the basic physics right. I measured the front elements of my 70-200 f/2.8 and 300 f/2.8, and indeed the ratio of the diameters is 1.54 - close enough to 1.5 to be within the error of my eyeball measurements. Assume same body, same position, just swapping lenses.

The 300 will capture 2.25 times as many photons coming off the subject. But it will also spread those photons over 2.25 times as many pixels. So the per-pixel S/N is about the same, right? But the granularity of the noise compared to subject features is less with the 300 image. So if I then print the subject to the same size, the 300 image will appear less noisy - in terms of noise per subject feature.

Have I got it? Very interesting, I never thought of the noise reduction effect of increased focal length at the same aperture. Makes sense though, noise comes from individual pixels, if you look more closely at them then you'll make the noise more evident.

Alan

Alan,

Let's work through it.

70-200 f/2.8: at 200 mm, the aperture is 200/2.8 = 71 mm.

300 f/2.8: aperture = 300/2.8 = 107 mm.

Light collection ratio = 107/71 squared = 1.51^2 = 2.27 times more light.

So yes, the 300 spreads the light out more, (300mm / 200mm) which reduces the signal per pixel but the focal length ratio squared (1.5*1.5 = 2.25), making the signal-to-noise ratio per pixel the same. But you have more detail with the 300 f/2.8 image. So with more pixels of the same noise per pixel and the 200 mm f/2.8 image, when you print images at the same subject size with the two images, the one with the 300 mm will show more detail, and because there would be more pixels on the subject, it would also appear less noisy.

If you add a 1.4x TC on the 70-200 and work at the 200 end, you get close to the same detail on the subject, but 2.27 times less light and correspondingly lower signal-to-noise ratio.

So the increase in S/N is not due to the same aperture, but the increase in clear aperture.

Roger

[Roger Clark]

Roger, my goal is to get more pixels on the subject so perhaps rather than look at another body, I should consider a longer lens... certainly something to think about.


Chris,
For more pixels on the subject with better signal-to-noise, you need not only a longer focal length, but a larger aperture. As you have a 400 DO lens now, a 300 f/2.8 will be only a small increase in aperture, so the remaining lenses that would be a significant step up are the 400 f/2.8, 500 f/4, 600 f/4, and 800 f/5.6.

Unfortunately, the price and weight goes up a lot.

Roger

[Chris Brennan]

Roger -

I initially bought the DO to serve as a walk-around lightweight tele with the hopes of adding a not-so-lightweight lens at some point in the future... that's still my plan, and chances are I'll go with the 500/f4. Truth be told, I'd love to have one of the new 400/2.8s due out this spring, but my better half isn't particularly fond of selling a vehicle to do it...

[Morkel Erasmus]

Hi Chris

I use the 7D, and unlike Grant, I don't have other pro or semi-pro backup bodies...that being said I have been getting very good to acceptable results at ISO 1600-6400 in different situations. I shoot with the 100-400mm L IS USM and I also use the 10-22mm USM for star-scape shooting.

I've found you just need to ensure you expose to the right of the histogram in order to minimise noise...