2 stops better low light performance from Canon Rebel to 7D and from 7d to 5D Mk. II?

[Chris7]

I've heard that, as you go up the line of Canon bodies, you get about 2 stops more low light performance between three major models. That is, the 7D has 2 stops better low light performance than the Rebel series, and the 5D Mk. II has 2 stops better low light performance than the 7D.

I have also read that the 1D Mk. IV gives about an extra two stops over the 5D Mk. II.

Would the people here say this is about accurate?

Just pondering...

 

[FX120]

No, that's not accurate.

Low light performance is a very subjective thing, and a lot of it has to do with what the end user feels is an acceptable level of noise.

 

[Edge100]

I think that noise performance is an overrated metric of most cameras. Yes, there is a decrease in noise in the newer cameras vs. what we were seeing a few years ago, but to be honest, I've never found noise to be that objectionable, even on my 6 year old 1DmkII at ISO3200, at sizes that are commonly viewed on the web.

Unfortunately, we've become so used to seeing 100% crops and pixel-level; do you regularly hold a 16x20 print right up to your noise? Of course not.

There's no doubt that lower noise is a good thing, but I wouldn't stress about it too much.

 

[fiercetiger224]

Maybe, if you compare crop vs full frame cameras. Larger sensors equal better noise performance. Between just crop cameras, not so much. As camera manufacturers continue bumping up megapixels on crop cameras, noise won't improve much. The 7D is already pushing it in terms of pixel density, although it is the best in its class.

 

[Phrasikleia]

I went from a Rebel (XSi) to a 7D, and would agree that there is a difference of about two stops starting at ISO 800 (maybe one stop at most at ISO 400).

I would also agree that it matters a lot less if you don't have to submit to a stock agency that checks your images at 100%. The average hobbyist and the people who see his photos won't notice much difference in detail once the images are printed, unless they are printed very large.

As for the 7D-->5D Mark II, again there is a significant improvement, primarily at higher ISOs. It could be as much as two stops at the very highest ISOs.

 

[JFreak]

As far as my personal subjective testing goes, this is it:

xxxD series: ISO 400 is acceptable
xxD series: ISO 800 is acceptable
xD series: ISO 1600 is acceptable

But I am quite a bitch about noise, so most people can live with a full stop faster; however, this is how I feel about it and because I "only" have a 40D I also must have f/2.8 capable glass.

I've been dreaming about a 2nd hand 1DSmIII but my wife would like the smaller 5DmII better, so I'm quite torn about my next upgrade ;P

 

[toxic]

there is a difference, but it is not "x" stops, and it is not a constant between camera series.

there is a slight difference from Rebels to 40/50D
1-2 stop difference from 50D to 7D
1-2 stop difference from 7D to 5DII
who knows about the 1DIV

the 1D series is intended to have the best high-ISO handling among Canon cameras for its generation. the 1Ds does not - Canon decided it is primarily a studio camera, where low-ISO performance is favored.

 

[OreoCookie]

The 2-stop progression is non-sense. Before I say anything else, IMO noise is absolutely overrated as a measure as to whether a camera is `good.' Up to ISO 400, the difference is not practically relevant.

There are quite a few reviews that claim the Canon 40D has better noise behavior than the 7D.

That said, the 7D is still a great camera, but to say `every time you lose a digit, you gain 2 stops' is incorrect and probably irrelevant.

 

[anubis]

It took me a very long time, but I was able to find a comparison between the noise of the 40D and 5D that took into consideration everything that was important. The author took two identical pictures, one with a 40D and one with a 5D. He cropped the photos to be the same size and used manual exposure control to make sure the two photos were identically exposed. The purpose of the test was to look at a scene with a large dynamic range, do a "correct" exposure, and look at the noise level in the dark underexposed areas of the photo. The results were clear as day: the 5D had about a 2/3rd stop improvement in noise at 800 and 1600 ISO. 400 and below the effect was negligible.

Unfortunately, I can't find this article any more : /

 

[Phrasikleia]

There are quite a few reviews that claim the Canon 40D has better noise behavior than the 7D.

I think you got confused with the DxOMark results. Higher numbers are better. In the "low-light ISO" measurements, they scored the 40D at 703, versus 809 for the 7D.

 

[anubis]

The 2-stop progression is non-sense. Before I say anything else, IMO noise is absolutely overrated as a measure as to whether a camera is `good.' Up to ISO 400, the difference is not practically relevant.

There are quite a few reviews that claim the Canon 40D has better noise behavior than the 7D.

That said, the 7D is still a great camera, but to say `every time you lose a digit, you gain 2 stops' is incorrect and probably irrelevant.

This probably doesn't take into consideration the fact that the 7D has almost twice as the pixels as the 40D and on identically sized prints, the 7D's noise will almost certainly be completely buried.

 

[OreoCookie]

I think you got confused with the DxOMark results. Higher numbers are better. In the "low-light ISO" measurements, they scored the 40D at 703, versus 809 for the 7D.

Confused?
The 7D offers higher ISO settings than the 40D, but the graph is quite clear, isn't it? Higher here means better signal to noise ratio. The 7D's curve is always below that of the 40D and the D300s (which are very, very close to each other).

In the linked article, the author juxtaposes the raw noise of various cameras. Note that dpreview only includes the 50D, 7D, D300 and the K7 in the article. He adds the nosie charts of the 40D:

I'm not sure how he makes the ranking, but at least up to ISO 1600, my eyes agree with his labelling.

This probably doesn't take into consideration the fact that the 7D has almost twice as the pixels as the 40D and on identically sized prints, the 7D's noise will almost certainly be completely buried.

It doesn't work like this, noise is measured on a per-pixel basis.

Note that I'm not saying that noise levels of the 7D are relevant as long as you avoid extreme settings.

 

[Edge100]

This probably doesn't take into consideration the fact that the 7D has almost twice as the pixels as the 40D and on identically sized prints, the 7D's noise will almost certainly be completely buried.

Exactly. If a 10MP camera and an 18MP camera have the same amount of pixel-level noise (i.e. in a 100% crop), then the 18MP camera is performing much better; when prints of the same size are made from both cameras, the prints from the 18MP camera will show less noise because you've binned far more pixels and averaged the noise out.

You can't just look at 100% crops of two different cameras and then say "Ah ha! The 40D has less noise than the 7D." Of course it does; it has 8 million fewer pixels! That's the problem we've gotten into with noise; we're not looking at prints anymore, by and large.

The best way to compare noise is to take the 18MP image and the 10MP image and make a JPEG or print of the same size from each (i.e. ignore 100% crops). Only then can you make an accurate comparison.

Attached are two (clickable) images I shot at a recent event here in Toronto (Canadians on the list will recognize the speakers as comedian Rick Mercer and astronaut Julie Payette). They're both shot with a 70-200 f/2.8L on a 1DmkII @ ISO 3200, 1/200s. Is there noise? Sure; especially in the shadows. And to be frank, it does not bother me in the least. And this is with a 6 year old camera with an 8MP sensor! I wouldn't choose this ISO if I could help it, and at 100% you can see quite a bit of pixel-level noise, but these photos are both featured on the client's website. And I wouldn't give up the AF performance of my 1DmkII just for a bit better high ISO performance.

The moral of the story? Pixel-level noise will inevitably increase as pixel density goes up; there's nothing much you can do about that. But I contend that unless you're making very large prints or have very, very picky clients, it's not really a big deal. There are far more important things to concern yourself with when it comes to taking a photograph.

 

[Edge100]

It doesn't work like this, noise is measured on a per-pixel basis.

You can measure it however you'd like. But why are you concerning yourself with pixel-level noise? Do you make a habit of viewing poster-sized prints from an inch away? That's what you're doing when you look at 100% crops from an 18MP camera.

I am FAR more concerned with how my prints (or final JPEGS) look at normal viewing distances (or output sizes).

 

[Phrasikleia]

It doesn't work like this, noise is measured on a per-pixel basis.

What Anubis meant was that it "works" (or rather, excels) on the multi-pixel level. That is, you end up with an image that has lots of detail and relatively imperceptible noise (the pixels are so small that the noisy ones do get "buried," as Anubis said, among the "good" ones). What happens at the single-pixel level matters very little to anyone except perhaps scientists.

I think that graph you posted is measuring on the single-pixel level, if I'm not mistaken. Regarding the other graphic, I don't know why all roads of this sort inevitably lead to dpreview (my least favorite site for forums regarding photography). At any rate, those crops give little chance for the extra resolution of a camera to come into play.

I can only speak from my own experience with a Rebel (the XSi) and a 7D: big difference in image quality between them. If some camera with nearly half the resolution (e.g. a 40D) shows a bit less noise, then that's good for anyone who is fine with that lower resolution...which brings us back to the complexity of comparing cameras of different classes: it's rather pointless to attempt some kind of general assessment when there are so many variables and so few of them matter to everyone in the market for a camera.

 

[OreoCookie]

You can measure it however you'd like. But why are you concerning yourself with pixel-level noise?

You measure the signal to noise ratio in various situations (black, gray, white, etc.). I'm concerning myself with this, because this is the standard way to measure noise.

Do you make a habit of viewing poster-sized prints from an inch away? That's what you're doing when you look at 100% crops from an 18MP camera.

I am FAR more concerned with how my prints (or final JPEGS) look at normal viewing distances (or output sizes).

I think you have missed the sentence in my original post saying that on today's cameras, noise levels are by and large irrelevant. I'm agreeing with you

 

[Chris7]

By "2 stops better low light performance" I meant to say that I had heard that, for example, the Rebel series at ISO 400 would have about as much noise (and other high ISO artifacts) as the 7D at ISO 1600; and the 7D would have about as much noise at ISO 1600 as the 5D Mk. II at ISO 6400.

But this is not not what I'm seeing from the screenshots in reviews of the individual cameras. The Rebel series, for example look OK to me at ISO 400. The 7D has quite a bit of noise at ISO 1600. And the 5D Mk. II just starts to barely show grain at ISO 1600 (but still looks great), for example.

But I know of no reviews comparing screenshots of these cameras.

Since the Rebel series and 5D Mk. II have been around longer and are more familiar than the 7D and 1D Mk. IV, maybe I could ask:

What ISO could I shoot a 5D Mk. II and a have a similar amount of noise as a Rebel XS at ISO 200?

What ISO could I shoot a 5D Mk. II and a have a similar amount of noise as a Rebel XS at ISO 400?

(I know the 5D Mk. II is better, period, but this comparison would still be useful to me).

 

[Edge100]

You measure the signal to noise ratio in various situations (black, gray, white, etc.). I'm concerning myself with this, because this is the standard way to measure noise.

Fair enough, but it's a misleading way to measure noise. All things being equal, if pixel-level noise levels are maintained on a higher MP camera, then you've improved noise performance.

I think you have missed the sentence in my original post saying that on today's cameras, noise levels are by and large irrelevant. I'm agreeing with you

My mistake.

 

[Phrasikleia]

By "2 stops better low light performance" I meant to say that I had heard that, for example, the Rebel series at ISO 400 would have about as much noise (and other high ISO artifacts) as the 7D at ISO 1600; and the 7D would have about as much noise at ISO 1600 as the 5D Mk. II at ISO 6400.

But this is not not what I'm seeing from the screenshots in reviews of the individual cameras. The Rebel series, for example look OK to me at ISO 400. The 7D has quite a bit of noise at ISO 1600. And the 5D Mk. II just starts to barely show grain at ISO 1600 (but still looks great), for example.

But I know of no reviews comparing screenshots of these cameras.

For starters, you can't lump the whole Rebel series together. You'd have to pick one model for a meaningful comparison.

I've seen a whole lot of ISO 1600 shots from a 7D at this point, and they are not comparable with ISO 6400 on a 5D Mark II, so what you heard is just plain wacky. ISO 1600 is very clean on either camera. The differences really start to come out above that.

 

[OreoCookie]

What Anubis meant was that it "works" (or rather, excels) on the multi-pixel level. That is, you end up with an image that has lots of detail and relatively imperceptible noise (the pixels are so small that the noisy ones do get "buried," as Anubis said, among the "good" ones). What happens at the single-pixel level matters very little to anyone except perhaps scientists.

I fully understand what Anubis and you are saying. And I'm not even disagreeing.

I'm saying that the 7D has more noise than the 40D. (And yes, I know why.)
You're trying to explain to me why it's not relevant. (Which I also don't disagree with.)

However, the OP's question was whether there is always less noise going up in Canon's line-up. The answer is no. Whether this is practically relevant is a different question. (In my opinion, it's not.)

 

[Phrasikleia]

What ISO could I shoot a 5D Mk. IV and a have a similar amount of noise as a Rebel at ISO 200?

What ISO could I shoot a 5D Mk. IV and a have a similar amount of noise as a Rebel at ISO 400?

Well, since I reckon the 5D Mark IV will come out in about 2013, the answers could be as much as ISO 300,000 or more.

 

[Edge100]

Here's a good example:

The 50D and the Rebel T1i share the same sensor. Check out the review of the T1i here. You can quite clearly see that the T1i has less noise than the 50D at ISO3200.

Thus, it does not always follow that the Rebel camera have "more noise" (whatever that means) than the XXD cameras (of the same vintage).

This is a reduction in pixel-level noise, but is still a fair comparison because the sensor is the same. Whether this translates into differences in visible noise in final output (JPEG or print) is not clear.

When sensor size differs, however, pixel-level noise is largely meaningless.

 

[OreoCookie]

Fair enough, but it's a misleading way to measure noise. All things being equal, if pixel-level noise levels are maintained on a higher MP camera, then you've improved noise performance.

There is no other way to measure noise.

I don't want to dredge on an argument when we agree, but the fact of the matter is that the pixel level noise level is not maintained. You're correct that keeping the noise levels constant while increasing pixel count leads to lower overall noise in practical applications.

But in this case, pixel-level noise increases and thus it's not at all clear whether you gain anything in terms of image quality by trading higher pixel density for higher noise levels. If you interpolate, then you get all sorts of artifacts from interpolating pixels. Then there is the software side: just compare the Nikon D3x to the Sony Alpha 900: they are based around the same sensor, but Nikon has managed to keep noise under much better control.

Comparing cameras is a tricky business.

My mistake.

No prob

 

[Edge100]

Since the Rebel series and 5D Mk. IV have been around longer and are more familiar than the 7D and 1D Mk. II, maybe I could ask:

What ISO could I shoot a 5D Mk. IV and a have a similar amount of noise as a Rebel XS at ISO 200?

What ISO could I shoot a 5D Mk. IV and a have a similar amount of noise as a Rebel XS at ISO 400?

(I know the 5D Mk. IV is better, period, but this comparison would still be useful to me).

Typos aside, this question is hard to answer.

Do you mean, at what ISO could I shoot with a 5DmkII and have the same pixel-level noise as a Rebel XS at ISO 400? If so, it's probably something like ISO 800 (i.e. one stop). ISO400 is pretty clean, even on a Rebel XS.

But remember, the 5DmkII is a 21 MP camera, while the Rebel XS is a 10 MP camera; you're talking about more than twice as many pixels. If you resize a 5DmkII image and a Rebel XS image both to 800 pixels, you will notice far less noise based purely on binning twice as many pixels; those cameras could have precisely the same pixel-level noise and you'd still see that difference.

This is why looking at 100% crops is meaningless for real-world use. Look at the 640 pixel images I posted. Those are from an 8MP camera at ISO3200. Does the noise bother you?

 

[Edge100]

There is no other way to measure noise.

Perhaps not objectively, but one can evaluate prints or JPEGs and see if pixel-level noise translates into real-world differences.

I agree that Nikon seems to have done a better job of maintaining detail at very high ISO levels (i.e. ISO6400 and up); there's no doubt that Canon has fallen behind in this respect (though the 1DmkIV images I've seen are very encouraging). Hopefully, Canon realizes the error of ramping up MP counts to no end. 21MP is very nice, and at low ISOs, translates into levels of detail that have been previously reserved for MF cameras (though the D2X is even slightly better than this at 24MP). But there is a real need for a 12MP, full-frame camera with a pro-level AF system.

Basically, the same pixel pitch as the old 5D (which produces very usable images at ISO3200), but with the 7D's AF system. Or, put another way, a Nikon D700.