Canon sensor density

rutt · March 29, 2005

Is the number of pixels per square milimeter the right way to think about sensor density. I think I've been a little confused by this, because I was using the inverse of the crop factor instead in order to compare the various sensor density.

Supposing I am right:

20D - 22.5 x 15 mm sensor, 8.2mp. About 24k pixels / square mm.
1DS mark II - 36 x 24 mm sensor, 16.7mp. About 19k pixels / square mm.
1D mark II - 28.7 x 19.1 mm sensor, 8.2mp. About 15.5k pixels / square mm.

By this measure, the 20D is by far the best camera is you aren't actually going to use the larger FOV of the 1Series cameras. So if you were going to crop anyway, use a 20D. For tele work, the 20D is like getting a 1.6 TC that actaully improves quality (compared to the 1DS mark II, even more so compared to the 1D mark ii.) Similarly for macro work, etc. Put it another way, if Canon made a full frame sensor with the same density of the 20d, it would be 21mp.

I've been confused about this. Might still be confused. Or am I right now?

DoctorIt · March 29, 2005

rutt wrote:

Is the number of pixels per square milimeter the right way to think about sensor density. I think I've been a little confused by this, because I was using the inverse of the crop factor instead in order to compare the various sensor density.

Supposing I am right:
20D - 22.5 x 15 mm sensor, 8.2mp. About 24k pixels / square mm.

1DS mark II - 36 x 24 mm sensor, 16.7mp. About 19k pixels / square mm.

1D mark II - 28.7 x 19.1 mm sensor, 8.2mp. About 15.5k pixels / square mm.

By this measure, the 20D is by far the best camera is you aren't actually going to use the larger FOV of the 1Series cameras. So if you were going to crop anyway, use a 20D. For tele work, the 20D is like getting a 1.6 TC that actaully improves quality (compared to the 1DS mark II, even more so compared to the 1D mark ii.) Similarly for macro work, etc. Put it another way, if Canon made a full frame sensor with the same density of the 20d, it would be 21mp.

I've been confused about this. Might still be confused. Or am I right now?

Now I'm confused, what does crop factor have to do with sensor density?

rutt · March 29, 2005

DoctorIt wrote:

Now I'm confused, what does crop factor have to do with sensor density?

I didn't use crop factor in this logic, right. I just calculated the square mms of each sensor and then divided that into the number of pixels to get the pixels/mm. What I found surprised me. If you have a 1DS mark II, and you crop (manually) your image the same FOV as a 20D, your image will be lower res than the 20D's. I had been going along thinking that the 20D and 1DS mark II had the same sensor density, but the 1DS mark II just had more of it. If my reasoning is right now, this just isn't true.

Is this making sense now?

JamesJWeg · March 29, 2005

rutt wrote:

Is this making sense now?

James.

DoctorIt · March 29, 2005

I'm still confused on what you mean by crop "manually"...

One thing I'm sure of though, more pixels isn't always better, as I'm sure you know. My friend's Nikon D1h and its 4mp has the same if not better image quality as his 6mp D70. Bigger receptor sites are actually better, if done right. So I guess there's a tradeoff at some point. Obviously, you don't want one big awesome pixel, but once you're beyond the human distinguishing limit in terms of PPI, then it doesn't matter. Right?

I think I'm still confused though.

mercphoto · March 29, 2005

So what you're trying to say is if you take the fantastic new 1DsMkII (where do they come up with these names?) and crop the resulting image on your computer to the same FOV as a 20D that you end up with fewer pictures than if you took it with the 20D to begin with.

While I'm sure that's true, I'm not sure it says anything whatsoever about which camera is "better". If I need extreme telephoto I don't want a full-frame camera. If I need wide angle I don't want a 1.6X crop factor. Etc.

rutt · March 29, 2005

I must not be being clear. My point isn't really very deep. Suppose that you have both a 20D and a 1DS Mark II (ok, that's a happy enough thought that you might just want to spend some time there.) Now suppose that you want to go off and shoot some bird pictures. Suppose that you "only" have a 300mm lens and that with the 1DS Mark II, you will end up cropping to less than to roughly the same FOV as with the 20D. All I was saying is that you end up with more pixels with the 20D. So for long lens work where you don't have the mms you want, the 20D is a better choice than the 1Series cameras, just on this one consideration alone. There are other issues, of course.

I hope that's better.

rutt · March 29, 2005

mercphoto wrote:

So what you're trying to say is if you take the fantastic new 1DsMkII (where do they come up with these names?) and crop the resulting image on your computer to the same FOV as a 20D that you end up with fewer pictures than if you took it with the 20D to begin with.

While I'm sure that's true, I'm not sure it says anything whatsoever about which camera is "better". If I need extreme telephoto I don't want a full-frame camera. If I need wide angle I don't want a 1.6X crop factor. Etc.

Right on! That's exactly what I was trying to say. I thought I was being so clear, but I guess not.

mercphoto · March 29, 2005

rutt wrote:

Right on! That's exactly what I was trying to say.

Then I agree with your post.

Now if Canon would build the auto-focus speed of a 1dMkII into a 20D...

pathfinder · March 29, 2005

rutt wrote:

Is the number of pixels per square milimeter the right way to think about sensor density. I think I've been a little confused by this, because I was using the inverse of the crop factor instead in order to compare the various sensor density.

Supposing I am right:
20D - 22.5 x 15 mm sensor, 8.2mp. About 24k pixels / square mm.

1DS mark II - 36 x 24 mm sensor, 16.7mp. About 19k pixels / square mm.

1D mark II - 28.7 x 19.1 mm sensor, 8.2mp. About 15.5k pixels / square mm.

By this measure, the 20D is by far the best camera is you aren't actually going to use the larger FOV of the 1Series cameras. So if you were going to crop anyway, use a 20D. For tele work, the 20D is like getting a 1.6 TC that actaully improves quality (compared to the 1DS mark II, even more so compared to the 1D mark ii.) Similarly for macro work, etc. Put it another way, if Canon made a full frame sensor with the same density of the 20d, it would be 21mp.

I've been confused about this. Might still be confused. Or am I right now?

The fly in the ointment in this logic, John, is that all pixels are not created equal.. and the noise varies as a result. Think of 8.2 Mpxl P&S 2/3s sensors. Their pixel density is even higher than a 20D's, but they do not have higher image quality as a result of higher pixel density, but lower quality due to higher noise levels.

Maybe we just need to shoot the same image with 20D and a 1DMkll with a 300mm lens, and then crop the 1DMlkll image to match the 20d and then compare them. Maybe we might borrow or find a 1DsMkll to compare with also...... Give me a few weeks to get my plate cleared of more pressing concerns right now. The government needs its share of our revenues soon you know:D

DoctorIt · March 29, 2005

rutt wrote:

I must not be being clear. My point isn't really very deep. Suppose that you have both a 20D and a 1DS Mark II (ok, that's a happy enough thought that you might just want to spend some time there.) Now suppose that you want to go off and shoot some bird pictures. Suppose that you "only" have a 300mm lens and that with the 1DS Mark II, you will end up cropping to less than to roughly the same FOV as with the 20D. All I was saying is that you end up with more pixels with the 20D. So for long lens work where you don't have the mms you want, the 20D is a better choice than the 1Series cameras, just on this one consideration alone. There are other issues, of course.

I hope that's better.

Got it. Whew, i thought I was losing all my math and science skills (this week sucks, NSF deadline).

wmsnyder · March 29, 2005

pathfinder wrote:

The fly in the ointment in this logic, John, is that all pixels are not created equal.. and the noise varies as a result. Think of 8.2 Mpxl P&S 2/3s sensors. Their pixel density is even higher than a 20D's, but they do not have higher image quality as a result of higher pixel density, but lower quality due to higher noise levels.

Maybe we just need to shoot the same image with 20D and a 1DMkll with a 300mm lens, and then crop the 1DMlkll image to match the 20d and then compare them. Maybe we might borrow or find a 1DsMkll to compare with also...... Give me a few weeks to get my plate cleared of more pressing concerns right now. The government needs its share of our revenues soon you know:D

I will throw in my two cents worth! The more pixels per mm you have the more light gathering ability you have. So in theory you should have better picture quality but this all depends on the associated circuitry used to process the photo in the camera. If you are manually croping a photo down in size you are throwing away pixels that the software thinks it does not need.

mercphoto · March 29, 2005

wmsnyder wrote:

I will throw in my two cents worth! The more pixels per mm you have the more light gathering ability you have.

No. You have less.

Think of it this way. You are talking about light gathering ability. This is a function of the lens and nothing more. More pixels per inch does not make more light fall on the sensor. More pixels per inch does nothing to bring more light in to gather. Given this, the light gathering ability should be identical regardless of the number of pixels per inch. But that is not quite the case.

Each light sensor on a chip is going to be surrounded by a certain amount of dead space. The more pixels you have, the more dead space you have. In other words, a sensor with a single giant pixel will be +99% sensor and very little dead space -- it gathers nearly all the light it falls on it. As you add pixels you also add dead space between pixels. The light that hits the dead space is not gathered.

For a given sensor size, the more pixels you cram in the less light you are actuallly recording.

ian408 · March 29, 2005

But you're supposing that pixels are the same size when comparing the
20d to the 1d{s}mkII. Are they? Is the process geometry the same for both?

Ian

mercphoto · March 29, 2005

ian408 wrote:

But you're supposing that pixels are the same size when comparing the 20d to the 1d{s}mkII. Are they? Is the process geometry the same for both?

I was not specifically comparing any two cameras in particular. I was responding to "The more pixels per mm you have the more light gathering ability you have." Which is false for the reasons I stated.

I do not know if the process geometry for those two cameras is the same or not.

rutt · March 29, 2005

I started this, but I was just looking for a place to admit that my math was wrong. I understand why it's better to have larger pixels. I understand why it's better to have more of them. I understand that there are reasons to like cameras with less dense sensors better than the reverse. I just had my math wrong before and wanted to correct that.

ian408 · March 29, 2005

mercphoto wrote:

I was not specifically comparing any two cameras in particular. I was responding to "The more pixels per mm you have the more light gathering ability you have." Which is false for the reasons I stated.

I do not know if the process geometry for those two cameras is the same or not.

Indirectly, there is some comparison between the 20d and 1dmkII. But I
just wanted to pick on rutt

Regardless, I think the discussion is productive and I'd thank John for
bringing it up in the first place.

Ian

rutt · March 29, 2005

ian408 wrote:

But you're supposing that pixels are the same size when comparing the
20d to the 1d{s}mkII. Are they? Is the process geometry the same for both?

Ian

Ian, you and I both know that you know more about this particular kind of thing than I do. Since you put your finger it it, I get to ask someone who actually might know the answer to some of the questions I have about this.

The mistake I made was to assume that the 20D and 1Ds Mark II sensors were made from the same "stuff" but the 1DS Mark II just had a larger chunk of it. But when I did the math correctly, I found that that wasn't true. The 20D has more pixels per square mm. Doesn't that mean that it has a different process geometry? Am I using the words correctly?
Why do you think they didn't just make the 1DS Mark II sensor from roughly the same stuff and the 20D sensor. That would have given it a resolution of roughly 21MP. A yield issue? Didn't the cameras come out at roughly the same time? Can they afford to do something more for the 20D and Rebel because they are higer volume?
Maybe the 1 Series cameras have better low light performace becaues they have larger pixels? The 1D Mark II would be the winner here if that was all there was to it. But it came out about 6 months before the other two cameras. Is that long enough to matter?

3rdPlanetPhotography · March 29, 2005

mmmm
huh?

ian408 · March 30, 2005

rutt wrote:

Ian, you and I both know that you know more about this particular kind of thing than I do. Since you put your finger it it, I get to ask someone who actually might know the answer to some of the questions I have about this.

I don't know that I know more than you in this regard. Only that pixel size
is not a constant and that different process geometries have different
device sizes.

And I know that Maker's Mark is pretty good stuff and tomorrow, I'm gonna
pay :lol

Ian

wmsnyder · March 30, 2005

ian408 wrote:

I don't know that I know more than you in this regard. Only that pixel size
is not a constant and that different process geometries have different
device sizes.

And I know that Maker's Mark is pretty good stuff and tomorrow, I'm gonna
pay :lol

Ian

I found this info as part of canon's 20d description:
The sensor features high pixel density made possible by the introduction of smaller area photo-diodes with a better S/N ratio and equivalent sensitivity to those of the EOS 10D.

ian408 · March 30, 2005

wmsnyder wrote:

I found this info as part of canon's 20d description:
The sensor features high pixel density made possible by the introduction of smaller area photo-diodes with a better S/N ratio and equivalent sensitivity to those of the EOS 10D.

That sounds like smaller pixels and a reduced area between each pixel. Be
neat to see the layout or die photo for an image sensor.

wxwax · March 30, 2005

DoctorIt wrote:

one big awesome pixel

You should put that under your name.

DoctorIt · April 1, 2005

wxwax wrote:

You should put that under your name.

Good idea

DoctorIt · April 1, 2005

rutt wrote:

Ian, you and I both know that you know more about this particular kind of thing than I do. Since you put your finger it it, I get to ask someone who actually might know the answer to some of the questions I have about this.
The mistake I made was to assume that the 20D and 1Ds Mark II sensors were made from the same "stuff" but the 1DS Mark II just had a larger chunk of it. But when I did the math correctly, I found that that wasn't true. The 20D has more pixels per square mm. Doesn't that mean that it has a different process geometry? Am I using the words correctly?

Why do you think they didn't just make the 1DS Mark II sensor from roughly the same stuff and the 20D sensor. That would have given it a resolution of roughly 21MP. A yield issue? Didn't the cameras come out at roughly the same time? Can they afford to do something more for the 20D and Rebel because they are higer volume?

Maybe the 1 Series cameras have better low light performace becaues they have larger pixels? The 1D Mark II would be the winner here if that was all there was to it. But it came out about 6 months before the other two cameras. Is that long enough to matter?

some good questions... a bump to maybe see answers...

wxwax · April 1, 2005

DoctorIt wrote:

Good idea

ian408 · April 1, 2005

DoctorIt wrote:

some good questions... a bump to maybe see answers...

Here is an interesting paper on the subject of how the sensor works. There's a
table that gives pixel sizes on page 3. The range is 3.3um x 3.2um for
a TSMC .18um process (doesn't say which one--there are several optimized
for either performance or power consumption) to 11.2um x 5.6um for a
Panasonic/Matsushita 1.0um process.

I would say that 1 series cameras are probably better in low light because
they have larger pixels. Which might also answer why the two use different
sensor designs.

Ian

Canon sensor density

Comments