Annoying EOS pixel count

[Cadenza]

Canon is often very thoughtful about usability details,
keeping the professional needs in mind. So why have
their R&D chosen to arbitrarily set the number of pixels
of EOS generated images at the numbers below?

3:2 ratio:

1D: 2464 x 1648
10D, 300D, D60: 3072 x 2048
20D, 1D MkII: 3504 x 2336
1Ds: 4064 x 2704


3:4 ratio:

S1: 2048 x 1536
G3: 2272 x 1704
G5: 2592 x 1944
G6: 3072 x 2304
Pro1: 3264 x 2448


Do these numbers make any sense to anyone,
from the usability standpoint? Why didn't they
develop sensors whereby the ultimate pixel count
delivered would be in nice round numbers like:

3:2 ratio:

2400x1600
2700x1800
3000x2000
3600x2400
4200x2800

3:4 ratio:

1500x2000
1800x2400
2100x2800
2400x3200


Considering that one often has to make calculations
on dpi, resolution, cropping and printing sizes, fiddling
with akward numbers is very annoying, and a serious
misuse of a photographer's precious time. Of course,
having round numbers would allow us to make most of
these calculations on our heads.

Does anyone know for sure why Canon chooses the
image pixel count that it does?

[Andy_T]

Most likely it will have some very boring technical reasons.

My suspicion would also be that

2400*1600=3840000 -> not quite 4 MP
2700*1800=4860000 -> not quite 5 MP
3000*2000=6000000 -> 6 MP, but maybe not compatible with the APS format?

So ... will we ever know? :roll:

I don't see it as much of a problem, on the other hand....

Best regards,
Andy

[jrm]

What you will notice is that al of Canon's resolution numbers are evenly divisible by 8. This is one reason (although I know why, I find it difficult to explain).

Actual chip dimensions/sensor area is going to be a factor as well. I am sure these are not arbitrarily decided numbers, but limitations/functions of the chip and the technology.

[BearSummer]

Hi Cadenza,

well they have done it with some of the values. When you realize that 1K isn't 1000 but 1024 then you will start to see some interesting values

1k= 1024
2k= 2048
3k= 3072
4k= 4096

as for why the other interesting values, your guess is as good as mine.

BearSummer

[CyberDyneSystems]

The post title say "EOS Pixel count"

But NONE of the 3:4 ratio cameras listed are EOS :wink:

[karusel]

When you think a little about the topic question it is really easy to understand. On the sensor the photosites have a certain density, so to reach 1.6 crop so and so many are needed. It would be unwise to develop a sensor that would have specific photosite density just so it would reach exactly the right number of resulting pixels at 1.6 crop (or, whatever the crop turns out to be). Which would be slightly pointless, they would not sell one camera more if the numbers were all rounded up. I imagine there are very small numbers to operate with when it comes to photosite size and density, a tiny change and there's 50 pixels change in one direction.

[OviV]

CDS,

Bob is starting to rub off on you dude! :lol:

[robertwgross]

To an engineer, 2048 and 3072 are nice, round numbers.

---Bob Gross---

[Scottes]

To an engineer, 2048 and 3072 are nice, round numbers.

Yep. 1024 is way better than 1000. 256 is a perfect number. And why have a million when you can have 1,048,576?

[tommykjensen]

To an engineer, 2048 and 3072 are nice, round numbers.

Yep. 1024 is way better than 1000. 256 is a perfect number. And why have a million when you can have 1,048,576?

He he, I hope I remember the correct phrase when I say:

There are 10 kind of people, those that understand and those that don't

[Mark Kemp]

Frankly the difference between 2000 and 2048 really won't have much effect on the quality of your prints etc. If you prefer round numbers then round them off, nobody will notice.

[mjordan]

It's that dang metric system that the rest of the world has been trying to force on us for years. They can't figure out our inches, feet, yards, miles, etc., so they are trying to get us to use cm, mm, liters and kilos.

So how many pixels are there to an inch anyway?


:lol:


Mike

[Mark Kemp]

Well since there are 144 poles to the perch and 723 2/3 miles to the foot on wednesday, as long its not a full moon, we must divide by 962.567 and adjust for the reflexion of the hypoteneuse. That obviously means that there are 156 pixels in an f8 picture, unless it was taken at more than 1/100th of a second or has a green cast. Now the picture will be printed 8 foot long by 2 thou wide on gloss toilet paper using non allergenic ink and sharpened for viewing at 376yards with 20x magnification binoculars and we must have at least a 6 foot border all around - leading to the straightforward conclusion that there are ... quite a lot of pixels per inch!

[robertwgross]

Can you translate that into furlongs per fortnight?

---Bob Gross---

[CyberDyneSystems]

CDS,

Bob is starting to rub off on you dude! :lol:

I'll take that as a good thing. :P

[defordphoto]

CDS,

Bob is starting to rub off on you dude! :lol:

I'll take that as a good thing. :P

So then you can do that furlong/fortnight calculation?

[Jon]

It's a binary thing. The branching and connecting circuitry design is much easier when you keep things in powers of 2.

Oh, and that's 16.438727509 furlongs/fortnight.

[Cadenza]

If it's a binary universe why did
Nikon engineers get the D70
to deliver 3008 x 2000 images?
(at least they got one dimension
in round numbers)

[PacAce]

If it's a binary universe why did
Nikon engineers get the D70
to deliver 3008 x 2000 images?
(at least they got one dimension
in round numbers)

Maybe Nikon doesn't know binaries. :? And if they do, then maybe they were afraid too many of their customers were going to ask the same question this thread is asking. :mrgreen: :lol: :mrgreen:

[Jon]

Hey, they're Nikon. Why should they do things the normal way?

Seriously, most sensors are larger than their rated areas, with the edges being used for things like noise assessment or performance monitoring. Nikon just uses more than most people do for that. Why? Possibly because they bought a standard sensor, and chose the area in use to match their "magical" 1.5x crop.

[robertwgross]

Nikon considers itself to be an optics company.

Canon considers itself to be an electronics company.

There is a big difference.

---Bob Gross---

[Cadenza]

Nikon just uses more than most people do for that. Why? Possibly because they bought a standard sensor, and chose the area in use to match their "magical" 1.5x crop.

Yeah, that's the other thing. Why did Canon give us a 1.6x crop?
1.5 crop would be slightly better. The 17-40 L would become a
25.5-60 instead of 27.2-64. 16-35 becomes 24-52.5.

[Scottes]

Yeah, that's the other thing. Why did Canon give us a 1.6x crop? 1.5 crop would be slightly better.

I want 2x crop. Then my 400mm becomes 800mm. NOW you're talking!

[Jon]

Yeah, that's the other thing. Why did Canon give us a 1.6x crop? 1.5 crop would be slightly better. The 17-40 L would become a 25.5-60 instead of 27.2-64. 16-35 becomes 24-52.5.

Canon's not using the same sensors as Nikon. They're smaller ones, so not large enough to support a 1.5x crop. Now why they're making sensors in that size is a totally different question. I think we're seeing their 3-pronged approach to sensor sizing: high-res 1:1 for the studio types (1Ds), 1.3X for the performance systems (1D Mk II) and 1.6x for the low end gives slightly better separation between the lines. Going all the way to 2x would absolutely kill wide angle for the low end, except with throttled lenses like the EFs series, only worse (not speaking in terms of lens quality here, you understand).

[s00pcan]

Most likely it will have some very boring technical reasons.

My suspicion would also be that

2400*1600=3840000 -> not quite 4 MP
2700*1800=4860000 -> not quite 5 MP
3000*2000=6000000 -> 6 MP, but maybe not compatible with the APS format?

So ... will we ever know? :roll:

I don't see it as much of a problem, on the other hand....

Best regards,
Andy

My G3 doesn't have quite 4mp.