14

since all lenses are round, full frame cameras' sensors can be 36 × 36 millimeters. Correct? Is it possible, and if so, why it has not been done yet? Would that qualify it as medium format?

mattdm
  • 143,140
  • 52
  • 417
  • 741
tony hage
  • 141
  • 1
  • 3
  • Are there any common uses for a square format? Most final applications I can think are one of 2,000 different rectangular formats. – JS. Jun 06 '16 at 21:52
  • 3
    @JS With square format you shoot first and crop later to your desired aspect ratio. Also no need to rotate the camera body to switch between portrait/landscape orientation, which can lead to using the camera in an unergonomic fashion. Why do you think the Hasselblad V System was so popular for so many decades? – osullic Jun 06 '16 at 23:00
  • 1
    I remember marking common aspect ratios on the focusing screen as a reminder to leave enough room in case your customer wanted to use a standard frame. Hasse is a fine camera, but a square format doesn't make aspect/cropping issues go away. – JS. Jun 07 '16 at 00:17
  • 1
    You use "roomy" framing for stock photography. Most commission photography sessions have a layout provided by the art director or designer to use for composing your miracle. – Stan Jun 08 '16 at 00:47
  • I just asked a similar question about making round sensors to accommodating any aspect ratio and allowing portrait orientation w/out rotating the camera if the sensor was round. I think this would be a great offering to the market, well worth a few extra bucks for extra sensor area. If most money is spent on lenses it would make sense to use the lenses to the fullest. – Michael Nov 29 '18 at 22:50
  • Like pointed out in the answer a square would not fit in the same circle, but I think a cross of 36x24 and 24x36 could be ideal so you could crop later and there would be no need to turn the camera. Maybe there would be a niche market for high end sports photography, considering the presumably much higher manufacturing costs? – Bjorn May 10 '22 at 17:22

5 Answers5

23

That's not correct. Look at this picture:

36x36 vs 36x24

The green rectangle is a 36x24 sensor. The green circle, which has a diameter of 43.3mm, is the minimal light spot needed for that size. The blue square is 36x36 sensor. The blue circle, which has a diameter of 50.9mm, is the minimal light spot needed for that size. As you can see a lens suitable for 36x24 does not necessarily cover the whole 36x36 frame.

As for the second part of your question - it is possible to qualify such sensors (in my opinion) as medium format because the most of available "full-frame" lenses will not cover these bigger sensors and so you'll need other lenses.

Mico
  • 113
  • 4
Zenit
  • 1,791
  • 1
  • 15
  • 21
  • 3
    Same comment as I wrote to Caleb: draw the diagonals :-) – Carl Witthoft Jun 06 '16 at 15:26
  • I'm very curious: Do you have a source for the statement that 'most available "full-frame" lenses will not cover these bigger sensors'? 2 additional cents: I think the 1:1 sensor would be a great idea, allowing for a workflow comparable to earlier 6x6 format photography. – kamuro Jun 06 '16 at 16:33
  • 1
    Even if one couldn't use a full corner-to-corner square, I would think that an image sensor captures a larger area than the light spot could give a photographer more options for cropping after the fact than one which is entirely within light spot, especially if the material would be masked to a non-rectangular shape after cropping anyway. – supercat Jun 06 '16 at 16:45
  • 2
    @kamuro No, i have no sources. But why would a manufacturer build a lens which covers more than needed? – Zenit Jun 06 '16 at 16:48
  • maybe because it is more work to bring the illumination circle down to a minimum than actually not caring about it ;) Also a greater illumination circle might be a side effect when trying to minimize vignetting. You also often see the effect of bigger illumination circle in zooms, which only get down to a minumum illumination circle on their widest end (which still might be bigger than the minimum needed). You can try that out yourself by putting a lens designed for crop sensor bodies on a full frame camera (or find examples on the internet). – kamuro Jun 06 '16 at 19:23
  • 1
    @Alex Manufacturers build lenses that cover more than needed for a very good reason: Lenses don't perform as well towards the edge of the image circle. So if you want good performance at the corners of your sensor, you need a lens that projects an image circle larger than your sensor. – osullic Jun 06 '16 at 23:23
  • @supercat That potential for a few users has to be balanced with the increased cost of producing a sensor larger than is useful to the vast majority of users. Building a sensor twice as large costs a LOT more than twice as much. Please see http://photo.stackexchange.com/a/54234/15871 – Michael C Jun 07 '16 at 06:45
  • Also: http://photo.stackexchange.com/a/53808/15871 – Michael C Jun 07 '16 at 06:59
  • A round-ish image would still be handy in permitting more flexibility when cropping out a rectangle at some angle. Might be useful for a gopro-like camera or scientific instrument. – JDługosz Jun 07 '16 at 07:29
  • @kamuro Yes, I understand that light through a lense covers more than the sensor behind that lens :) And some APS-C lenses cover a full frame sensor (I remember such a test about e-Mount lenses on then-new full frame Sony Nex). But a photographer would not want to use such a lens-sensor combination besause of poor performance on the edges (who wants fotos with dark blurry corners?). That's what I meant. – Zenit Jun 07 '16 at 07:42
  • @osullic Sure, I don't think that any lense covers EXACTLY the minimal possible circle. – Zenit Jun 07 '16 at 07:44
  • @MichaelClark: If one is using cheap optics, it makes sense to economize on the imaging sensor, but if one is using expensive optics I would think it would make sense to try to get the most out of them. Even if doubling the surface area of a sensor would quadruple the cost, that might still be a small fraction of the cost of the optics. – supercat Jun 07 '16 at 16:32
  • 2
    Doubling the linear size of the sensor doesn't quadruple the cost. It increases it by a factor of about 10. – Michael C Jun 07 '16 at 19:57
  • @MichaelClark: I was thinking of doubling the area, which would be roughly the effect of changing from having it inscribed in the light circle to having it enclose the light circle, so I would think the change in price should be less than 10x, but in any case sensors keep getting cheaper and cheaper relative to optics. If one needs a picture with a really wide aspect ratio and has a lens that can hold a certain level of visual acuity throughout a normal-format frame, that same lens would likely be able to hold that acuity across a wider but shorter frame, which would mean that the... – supercat Jun 09 '16 at 17:25
  • ...using the larger portion of the sensor but scaling down the result would result in a picture that was slightly sharper, at least at the middle, and would be no worse at any point than a standard-format frame would be at the corners). Since improvements in optical acuity get very expensive with higher-end lenses, using more of the image spot would seem like a possible win. Further, such a sensor would avoid the need to rotate the camera to select portrait/landscape shots, and could also be helpful when shooting unpredictably-moving subjects. – supercat Jun 09 '16 at 17:28
  • @supercat https://commons.wikimedia.org/wiki/File:Wafer_die%27s_yield_model_(10-20-40mm)-_Version_2-_EN.png – Michael C Jun 10 '16 at 09:43
  • The difference between a cheaper WA lens and a very expensive one is not a factor of 10. The Canon EF 11-24mm f/4 sells for $3K. The cropped EF-S 10-22mm f/3.5-4.5 sells for $650. That's only a factor of 4.6X. It is true that something like the EF 14mm f/2.8 is $2K and the Samyang/Bower/etc. 14mm f/2.8 is about $330 (6X), but there you're also giving up AF and camera control of the aperture (which sacrifices automatic exposure, stop down metering, etc.). – Michael C Jun 10 '16 at 09:50
  • @supercat Those specifically desiring an image sensor larger (and squarer) than a FF sensor do have options for such a sensor. http://www.bhphotovideo.com/c/buy/Medium-Format-Digital-Cameras/ci/16734/N/4259332394 – Michael C Jun 10 '16 at 09:55
10

Is it possible and why it has not been done yet ?

Not necessarily. A 24x36mm sensor will easily fit in an image circle that's too small for a 36x36mm sensor. Specifically, a 24x36mm sensor requires a minimum diameter of about 44mm to cover the sensor. A 36x36mm sensor would require an image circle of about 51mm diameter. A square sensor is certainly possible, but square frames were possible with film as well.

image circle

The diameter of the image circle has to be at least as large as the sensor diagonal. You can see from the diagram that the green 24x36mm rectangle fits within the yellow image circle, but a 36mm square does not -- the corners fall outside the image circle and will be dark.

There's a variety of aesthetic and practical reasons that non-square images are preferred. Indeed, people often prefer images that are even wider than the 3:2 aspect ratio that we have in the typical full frame sensor.

If you prefer square images for any reason, you can set many cameras to record a square image, or you can crop after the fact.

Caleb
  • 31,682
  • 6
  • 65
  • 120
  • 2
    It might be easier to understand this answer if you explain that the minimum image diameter is the length of the diagonal of the sensor region :-) . (and draw a picture) – Carl Witthoft Jun 06 '16 at 15:25
  • Sounds good @CarlWitthoft. – Caleb Jun 06 '16 at 16:00
  • 7
    Worth including: the square that does fit within the same image circle as 36x24mm is about 30.6mm. – hobbs Jun 06 '16 at 17:01
5

The sensors could be made to a square format (though the current diameter would not accommodate 36x36mm, it would need to be about 30mm) if there was a demand for it. But by that logic the question we may actually need to be answering is why aren't sensors circular given that lenses present a circular image?

There were some attempts at circular sensors in the phone space when Nokia were working on the PureView cameras like the 808 before PureView just became a marketing thing for their best cameras. The camera then used the surface area it needed to make the chosen aspect.

Going back to squares - other square sensors are available in the 35mm size range outside of the SLR space (like the SBIG stx-16803 aimed at astronomers - thanks Dan.)

One of the things is that inside a DLSR you also need space for a mirror to move into and also for a shutter mechanism. While the shape of cameras could change to accommodate it, it's a lot of R&D for a small payoff.

It's also telling that even the medium format cameras use a non-square sensor with the Leica S, Pentax 645D, Leaf AFi, Hasselblad H5D and PhaseOne medium-format systems all adopt a rectangular aspect-ratio similar to the 6x4.5 systems than 6x6.

Any increase in sensor area also means a non-linear drop in wafer yield (the number of useful sensors you can get out of a silicon wafer) as demonstrated graphically in this answer - again it's extra cost.

Little payoff is where the idea really falls down - Canon, Nikon, Sony etc all put huge amounts of money into R&D, if any of them thought there was a viable market then they'd go for it. I'm not seeing one and the lack of such DSLR cameras on the market would back that up.

Community
  • 1
James Snell
  • 9,709
  • 25
  • 39
  • Some specialty cameras do use square sensors. eg the SBIG stx-16803 has a 36.8x36.8mm 16 megapixel sensor. Square sensors are relatively common in high end cameras designed for astrophotography (less expensive ones are more likely to use a commodity sensor in more traditional aspect ratios). – Dan Is Fiddling By Firelight Jun 07 '16 at 11:21
2

Just an additional to the above answers, as I don't see anyone mentioning it:

  • aside from the others notes on lens and mount coverage, a square format sensor was produced and marketed and sold. It was superseded quite quickly as the market in which it was sold was looking for larger sensors, and the 36x36 was just too cropped. After the 2007 release of the upgraded P20+ the square format was dropped.

cameras' sensors can be 36 × 36 millimeters. Correct?

36.9 x 36.9 mm

why it has not been done yet?

It has, the first consumer variant was seen in the tethered H20 in 2001. The Kodak DCS 645 Pro Back came out in 2002. The Phase One P20 came out in 2004.

Would that qualify it as medium format?

It did, as you can see below, went on to fit various Medium Format cameras with interchangeable back ability:

Phase One P20 (also Kodak DCS 645 Pro Back)

  • Sensor: KAF-16802CE Kodak sensor
  • 16 megapixel CCD
  • 4,145 x 4,145 total pixels
  • 36.9 x 36.9 mm (1.5 x 1.5 in)
  • 9 x 9 µm pixel pitch
  • 1:1 ratio
  • 16 bits per pixel ADC
  • Single exposure: 4,080 x 4,080
blobbymcblobby
  • 734
  • 3
  • 8
-1

Square is not necessarily larger. 24 megapixels will hold:

3:2 6000x4000 pixels
4:3 5657x4243 pixels
1:1 4899x4899 pixels

And "pictures" have mostly always been rectangular, from most film, and paintings too (like Rembrandt). We did have square medium film, but the print paper was rectangular . It seems a matter of strong preference.

WayneF
  • 12,879
  • 1
  • 16
  • 30
  • 5
    This is not 100% accurate: my parents' venerable Kodak Brownie camera was square format, and so were the prints. (now get offa my lawn you young punks, etc) – Carl Witthoft Jun 06 '16 at 15:22
  • 1
    @CarlWitthoft - And let's not forget good ol' Hasselblad's square 6cm x 6cm film format. :-) The actual dimensions of the negatives and slides were 56mm x 56mm, if I remember correctly. – Mico Jun 06 '16 at 19:48
  • I don't see how this answers the question. – David Richerby Jun 06 '16 at 22:26
  • 1
    @CarlWitthoft My first camera shot square using 126 cartridges: 28mm x 28mm. –  Jun 07 '16 at 01:52
  • In this case, the question is suggesting making a 6000×6000 sensor. Corners will be shaded in most cases, but you could decide which of the above compromises you want for every shot. The answer to Why Not That? is basically "size and cost". You're answering a different question — Why Not Fit a Square in the Circle?, and it's a good answer to that... just not the question asked. – mattdm Dec 22 '16 at 16:18
  • 1
    Come on Matt, you ought to know about vignetting. Vignetting reaches into diagonal corners, and affects regions of any shape sensor that far from center of lens (the diagonal corners). Sensor shape is NOT a factor, only extent of diagonal is (i.e., distance from center of lens). In the case of 6000x6000, the height will be greater and the width will be less, but the DIAGONAL will NECESSARILY be exactly the SAME (for the same lens to cover it). – WayneF Dec 22 '16 at 16:32