5D2 vs 7D Depth of Field

[commanderspike]

Yes, I knew that; my point was about why people associate larger sensors with better low-light performance.

People do that for sake of brevity. They may know the details and how it works, but for the sake of terminology talk about 'large sensors' when it comes to influencing low light performance.

The DPReview.com camera comparison chart has the pixel density for almost every DSLR on the market, inc. the up and coming 7D.

But even pixel size & density is not the whole story. It depends on numerous other factors including the read out, photosite design, micro lenses and technology (i.e. backlit, etc.)

The 7D has an improved microlens design over previous CMOS sensors, but more than double the pixel density compared to the 5D. So since it has a slightly higher density to the GH1 but an improved design and a higher number of megapixels on a larger chip, it will sit somewhere between the 5D and GH1 for low light performance.

The higher the total resoloution, the more noise is removed in downsampling the image to 1080p. So 18 megapixels crammed onto a APS-C sized sensor is not without it's advantages. I don't agree with the talk of huge 3 megapixel CMOS sensors. Meh.

 

[ydgmdlu]

Have you read this?


Myth busted: small pixels have worse performance than large pixels
http://www.reduser.net/forum/showthread.php?t=30076

It's not a myth in the most basic sense. After glancing over the posts in that thread, what becomes clear is that it's a very complicated issue. Notice, for example, that in the section where the OP talks about sensor size, he never really supports the claim that sensor size, in itself, is responsible for differences in performance; he mentions, as I've said in the past, that smaller sensors are typically used with lenses with far smaller absolute apertures, which means that far less light is being collected, which results in poorer performance.

Even the size of the lens points to this fact: the large sensor will require a lens that is perhaps 50 times larger and heavier for the same f-number and angle of view, and that lens will focus a far greater quantity of light than the very tiny lens on a digicam. When they are both displayed or used in the same way, the large sensor will have far less noise, despite the smaller pixels.

 

[ydgmdlu]

People do that for sake of brevity. They may know the details and how it works, but for the sake of terminology talk about 'large sensors' when it comes to influencing low light performance.

Actually, from what I've read, most people think that sensor size is a direct factor in low-light performance. I've even debated with people here at DVXuser who took that position.

 

[ydgmdlu]

This is written all over the place by people here and everywhere else, but its basically not true. Sensor size does make a difference in DOF using same focal length/stop/distance, due to circles of confusion. Theoretically and practically I've seen it in play. I don't want to get into a huge discussion about it because it's exasperating for me having to explain it again - I've done that in a previous thread somewhere at dvxuser.

Admittedly, it's actually a more complicated issue than either of us has made it seem.

What is meant by "depth of field"? Does it refer to a property of how an image looks, or is it something more abstract?

Depth of field is, basically, a property of the lens. It is wholly determined by the principles of optics. It refers to a distance around the subject within which things will appear sharp. This cannot change due to the size of the sensor.

However, when we talk about DOF, informally other things are at play. For example, when you use a longer lens, you will get a flatter image. When the background is out of focus, it will look very different from the OOF background of a deeper image. Because the image is flatter, background objects will appear larger in proportion to foreground objects, resulting in less overall scene detail. So the image would appear to have less DOF, in an informal sense, than a deeper image, even if the DOF is technically equal in both.

Using a crop sensor gets you essentially the same result as if you were to simply make a crop of the center part of a full-frame image in post. For that reason, in the most basic technical sense, the DOF between the image from the two sensors is exactly the same. The images themselves are essentially exactly the same, except that one has a narrower FOV than another. HOWEVER, most often the image from the smaller sensor is blown-up to more or less match the image from the larger sensor. For our purposes, we're trying to obtain a 1920x1080 image from both. When you enlarge the image from the smaller sensor, of course the circles of confusion in the resultant image will become larger as well. But the DOF of the original image, the image projected from the lens onto either sensor, has same-sized circles of confusion, no matter what the size of the sensor is. So that explains what you've seen in your experience.

So, basically, we're both right.

 

[studio1972]

When people say they want shallow DoF, in many cases, they actually want to blur the background. This is actually a different thing and can be achieved with a narrower field of view or by physically moving the subject further from the background, even if the DoF is actually not very shallow.

 

[Ian-T]

In regards to the almost identical image detail and sharpness of the 7D compared to the 5Dll, does the fact that it has a higher percentage of pixels in its smaller APC-S sized sensor contribute to that? Before I watched the test footage (from Bloom and Co.) that we've all seen I expected the 5Dll to be more detailed...but honestly.... their images look identical.

 

[ydgmdlu]

When people say they want shallow DoF, in many cases, they actually want to blur the background. This is actually a different thing and can be achieved with a narrower field of view or by physically moving the subject further from the background, even if the DoF is actually not very shallow.

But subject-to-camera distance is a factor in DOF...

 

[tnle2]

dadoboy is right, DOF depends on sensor size. Just look at the mathematical equations. DOF has the "c" term in the near and far equations and this "c" stands for circle of confusion. The circle of confusion is different for each sensor size because you're magnifying the image different amounts. It's simple--it's right in the equations. I have no idea why people argue that it isn't. They just end up confusing people. The math doesn't lie.

http://en.wikipedia.org/wiki/Depth_of_field
http://en.wikipedia.org/wiki/Circle_of_confusion

 

[commanderspike]

Actually, from what I've read, most people think that sensor size is a direct factor in low-light performance. I've even debated with people here at DVXuser who took that position.

Sorry do disagree with you yet again But if you think about it, that is actually the dominant factor - sensor size.

If you have a large sensor and fill it with too many megapixels, the photosites may be small and densely packed but the image will scale down from a higher resolution to 1080p, and so the noise will be reduced in the process. Witness viewing a noisy 12 megapixel ISO 3200 photo at 640x480 as an extreme example.

Now if you had a large sensor with fewer megapixels, you also win, because the pixel density is so low.

Whatever way you slice it a small sensor is always a disadvantage when it comes to noise & low light performance.

The only way a small sensor can match a larger one is for the sensor technology itself to be less noisy and more sensitive to light on a photosite by photosite basis.

It is not 100% true that a small sensor with a low pixel density can out perform a large sensor with a high pixel density, either.

The Fuji F31 with it's 6 megapixel 1/1.7" has a pixel density of 14MP/cm2. The Fuji F100 had a slightly larger 1/1.6" chip with a silly high pixel density of 25MP/cm2 - nearly dounle the pixel density. Yet when the image is resized from 12 megapixels to 6, it can almost match the F31 for low light noise pixel by pixel.

This is part of the reason 1080p from a RED ONE looks better recorded at 4k rather than at 2k.

 

[ydgmdlu]

Sorry do disagree with you yet again But if you think about it, that is actually the dominant factor - sensor size.

See, now what we're doing is arguing over semantics. You make good points, and I don't disagree at all, nor have I ever contradicted them. But the problem is that sensor size is an indirect factor, however "important" it may be. The size of the chip is not going to make the photosites any more sensitive than they are. That's the point that I'm making.

Whatever way you slice it a small sensor is always a disadvantage when it comes to noise & low light performance.

How so? If you take the same photosites (same size, same technology, etc.) and spread them across a larger area, would you suddenly get a better image because the sensor is bigger?

It is not 100% true that a small sensor with a low pixel density can out perform a large sensor with a high pixel density, either.

The Fuji F31 with it's 6 megapixel 1/1.7" has a pixel density of 14MP/cm2. The Fuji F100 had a slightly larger 1/1.6" chip with a silly high pixel density of 25MP/cm2 - nearly dounle the pixel density. Yet when the image is resized from 12 megapixels to 6, it can almost match the F31 for low light noise.

Nobody here has disagreed with that, not even Barry.

Once again, the whole matter is very complicated. There are too many confounding factors such that no general statement can be made without a lot of qualification. I suggest that we stop talking about sensor size, photosite size, and pixel density altogether. It doesn't get us anywhere and can only cause confusion. We should just accept that some chips don't perform as well as others and leave it at that.

The 7D doesn't perform as well as the 5D in low light, and we don't have to say (nor should we say) that it's because of a smaller sensor, smaller photosites, or higher pixel density. It just is. After all, the 7D is considered a "lesser" model. Perhaps that's what we should say from now on: The 5D performs better because it's higher-end.

 

[ydgmdlu]

After doing some more thinking, I've decided to slightly reverse my position. But I still stand by most of what I've said. And even in admitting that sensor size is some sort of factor in low-light performance, I still say that it comes with a lot of caveats, such that it doesn't really make a difference.

If you think about it, there are no real tests that we can do to support any of the general statements that we're making. Every comparison that we might try to make is deeply flawed; the comparisons wouldn't be fair. For instance, you can try to compare the DVX with the EX1, but one's SD and the other is HD, so how do you determine which one has better low-light performance in any meaningful sense? The EX1 might have more noise, but it also has way more resolution to make that noise, as well the intended details, much more perceptible. You can try to compare the 7D with the 50D, but the 7D has better sensor technology to make up for its greater pixel density and (maybe) smaller pixels. The 5D may outperform the 7D, but it has bigger pixels and higher resolution, so what does that say about sensor size as a factor? And if you try to compare the big pixels from one sensor to the smaller pixels from another sensor, you have another problem: Where are you going to find two differently-sized pixels that share the same design but for the size? After all, camera makers are constantly improving sensor technology with every new camera, and different manufacturers and brands use different designs.

Ultimately, what's more important than numbers (be it sensor size, pixel density, etc.) is results. So as I suggested previously, that's what we should be talking about, rather than these silly numbers.

 

[Barry_Green]

ALL these discussions HAVE to be suffixed with "all other things being equal". And as vdgmdlu has pointed out, all other things are basically *never* equal. Comparing a DVX with an EX1, the DVX is more sensitive even though it has a smaller chip. But it's noisier, and then we have to factor in that one's HD and the other's SD, one's CCD and the other's CMOS, and one is from 2002 and the other's 2008. There's simply no way to do a direct comparison.

Shortly, hopefully, there will finally be a bit of a way; with the Epic and Scarlet we'll have different-sized chips using the exact-same technology. If we can compare a Mysterium-X of one size against a Mysterium-X of another size, then we'll finally have a chance at "all other things being equal."

Until then, it's all lips flapping to no definitive conclusion.

 

[mikezip007]

Here's a very small test I tried. I know it's not scientific, but it alludes to the answer I think I was out to get in the first place. Comparing the DoF of a full sized sensor to a crop with the same field of view. Both cameras set to:

ISO 200
f4.5
1/100

the 30D (crop sensor) at 70mm
5D2 (full frame) at 105mm

I eye-balled the framing to make it similar (this isn't rocket surgery so I know they are not identical) but you can see the DoF differences.

30D (crop sensor)

5D2 (full frame)

 

[mikezip007]

no?