fbpx

DIY Photography

Hacking Photography - one Picture at a time

Search

 

Submit A Story

Would a circular image sensor be the best thing ever? Heck yeah!

by 29 Comments

Have you ever wondered why camera lenses produce a circular image circle (in general) but sensors are rectangular? Of course the answer is mostly historical – the format of a 35mm photo is in 3:2 aspect ratio, a ratio that people have long known to work well due to human binocular vision.

But lenses, in general, are circular. they produce an image circle, a circle that, in general, allows the 35mm sensor to just fit inside it. You don’t want your image circle much bigger than this (except for tilt-shift lenses) because that makes your lenses heavier and more expensive.

A full frame sensor is 36x24mm, or 864mm²

So, the image circle has to be at least wide enough to fill the rectangle

so, we calculate the radius of the circle as √ (18² x 12²) = √ (324 + 144) = √(468) = 21.64mm.

This shows us the area of the whole image circle is 1,471mm². So by using the full frame sensor we’re not even close to getting the full image that the lens projects, we’re throwing away 41% of it.

Let’s put that in a different perspective.

Here’s a small table of some popular Canon camera sensors, the number of megapixels on each and how many megapixels the camera would have if it had a circular sensor of the same pixel density.

Yes, that’s right, using exactly the same sensor technology that is available today, you could push the EOS 80D up from 24.2 to 41.1 megapixels. Of course the 80D isn’t full frame, but there’s no reason the same system couldn’t work with APS-C as well. If anything it would be even more practical to do this.

Here’s a visual representation of how much more you get from a circular sensor compared to the regular one.

And remember. Not only are you getting more megapixels without increasing density or decreasing sensitivity, you’re also still using the same lenses with the same optical properties – although we’ll go into this a bit more later.

But wait! Who the hell wants circular photos?

No-one does. Circular photos are stupid.

But that’s not the point. We’re not going to make circular photos. We’re going to make regular rectangular ones. But so much better than before.

Here are five absolutely killer reasons that are so good that they probably could go on their own page each with a NEXT > button just to show more adverts. But no, I won’t do that to you.

Reason 1 – Never having to do this ever again

That’s right. Wave goodbye to having to turn your camera on the side to take portrait-oriented shots. With a circular sensor you can take normal cropped 36x24mm full frame images in either orientation just by switching your camera from landscape to portrait without having to hold your camera on the side.

But what’s so difficult about holding a camera sideways?

I guess you’ve never used one of these before…

That’s right. Switching camera orientation when shooting portraits with a flash is a real pain because your lighting angle changes, and switching between orientations on a tripod can be doubly-frustrating.

Reason 2 – Using different aspect ratios at full resolution

Have you ever needed to create a photo in a different aspect to the normal 3:2? Maybe you need to do a square image? Maybe one that is in 16:9 ratio? Currently all you can do is crop.

Using our circular 80D sensor again, what do we get? It’s time for another table:

Reason 3 – Changed your mind what orientation you needed? No problem!

Did you take a photo in portrait mode and then think you really should have taken that same shot in landscape? Well, now all you need to do is go back to your circular raw file and select the different orientation. Or even a completely different aspect ratio.

On the left, cropping a traditional landscape shot to convert it to portrait. On the right, cropping your circular image to portrait – It doesn’t matter what angle you were holding your camera at the time you took the shot, you still get the benefit of your full resolution.

And that means straightening horizons no longer means losing some resolution!

Reason 4 – Oh damn, I cut off the hand when I took this photo. I wish those missing pixels were hidden somewhere. Oh wait, they are!

Suddenly raw files become a lot more exciting. As well as changing your mind on orientation, you can essentially turn every photo into a shift lens photo by offsetting the rectangle a little.

Depending on how much shift you do you’ll get some vignetting and image loss in the corners, but in many cases that’s easily repairable in photoshop.

In fact, in many cases you’ll be able to take the full circular image and extend it out into a square image by ‘content aware’ filling with some occasional manual editing to convert your 41.1 megapixel circular image into a 55 megapixel 7420×7420 square image. And let’s remember, this is with sensors and lenses designed to give us 24 megapixels currently.

Reason 5 – Circular photos!

There’s one canon lens that creates circular photos right now, the EF 8-15mm f/4L Fisheye, when set to 8mm on a full frame camera you get a 180 degree image filling a circle fitting inside your standard full frame image. This is, of course, an image circle in reverse.

With our new super circular sensor on a full-frame camera this lens would fill our entire image circle with an 180 degree image at around 10mm – giving a vastly higher resolution image.

Let’s switch to a 5DSR sensor type. On a regular 5DSR the actual circular part of a full 180 degree image from the fisheye would work out at 26.3 megapixels. But the same 180 degree image (with the lens set to about 10mm) would be the full 86 megapixels on the circular sensor. Again, this is the same lens, and exactly the same pixel density.

It sounds great! Why haven’t they done it before?

There have been three big reasons to stop this in the past.

Firstly, this is not that useful a tool for people who don’t edit their images. If you want to just take photos and print them out there are some benefits of convenience (landscape/portrait) but the rest of the benefits aren’t that significant to you. Fortunately more and more people are now comfortable editing their photos, and the benefits this new sensor would bring are clear.

Secondly, it doesn’t really fit in a DSLR. You’d need a square mirror and a bigger prism and then you hit a major problem – your mirror would be too big to flip up without hitting the lens. And you can’t move the lens away from the sensor. So it sounds like this will only work if you can design a camera that doesn’t need a reflex mirror assembly. Fortunately, we can now.

Finally, and the most significantly, the cost of creating the sensors. Sensors are created on silicon wafers and currently about 20 full-frame sensors can be made from a single wafer. Once you start trying to make circular sensors you get a lot more wastage and so you’d need to ensure you use your wafers more efficiently. The most efficient way to stack circles is essentially to stack them with hexagonal packing, and the very absolute best you could get out of that would be 14 circular sensors, but it’s more likely that 13 is the absolute limit. Needless to say this means your sensors are at least one and a half times more expensive than before.

Will all our lenses work?

Not all of them. First you have to ditch these

and replace them with these

You can be sure within 15 minutes of the new camera being launched there will be a dozen chinese companies offering circular-compatible lens hoods for the whole range of Canon lenses.

More seriously there are some other issues. Look at the back of your lens. Does it look like this? If so, then sorry, it won’t work properly with a full image circle. However, the camera will presumably have a database of lenses and capabilities so that it knows exactly how much to crop down for each lens to maximise available image from the lens.

Another issue is that currently lenses are tested and designed to perform well for rectangular sensors. Its’ quite possible an optical flaw within the glass area currently not covering the sensor area will go unnoticed during testing and the fault only show up during use with a circular sensor camera – and at that point who is liable for this? The lens wasn’t sold originally as suitable for a circular sensor – so although the likelihood is most of your glass will work, if there are problems with the lens you’re probably on your own.

Still, I’m sure you’ll agree the advantages seriously outweigh the disadvantages!

About the Author

Jolyon Ralph is a photographer and collector of Canon EOS film bodies, although he also shoots digital when forced to.  You can find out more about him and follow his work and writings on his website.  This post was also published here, and used with permission.

Related posts:

First Look at Sony’s New Curved Sensor – Will It Be A Game Changer? Here’s Our First Look at a Picture Taken With Sony’s New Curved Sensor A hipster’s guide to EOS film cameras

  • I like the idea of going back to square format…saving time in doing portraits/landscape….and easier when doing wedding albums…
    This sounds almost like it, so I could see it working….

    • So get yourself a Fuji GFX abs set it to 1:1.

    • Most DSLRs can be set to square format

    • Yes. But you lose megapixels doing so.

    • I’m sure you use less than half the megapixels you have as it is. :) I know I do.

  • Unless the method of creating chips changes, the last point is maga valid. Think about making cookies out of a circular slab of dough and not being able to use the remains.
    The rectangular shape also allows the trimming from the edge of the lens that is less sharp.

  • Was actually talking to another action photographer about the idea about a year ago. The fact that shooting square-sensor would mean infinite cropping possibilities, and going to full optical circle would be even better. “Fix it in post” :D

  • Maciek

    What about increased size of RAWs that would store data that mostly wouldn’t be used? The data needs to be transferred from the sensor to card, 40% larger, as the article calculates, would slow down the photo burst rate (not mentioning the need of bigger buffers that I believe are quite expensive).
    The corners would be never used, so to save up some disk space a new file format should be introduced. It would take a while before existing software would implement and it would significantly slow down the adaptation process.
    Furthermore, the quality of image around the edge of lens is usually worse than in the center, which (it’s maybe not as strong argument) may be another reason for rejecting the data anyway.
    On top of that increased power usage and complexity in software and hardware (where the portrait/landscape should be, should the user be able to change their mind when browsing photos in camera etc) doesn’t seem to justify the few cases of cropped hand or rotating camera 90deg.

    • Jolyon Ralph

      Raw files use lossless compression (or lossy on some systems, but let’s not go there) so there would be almost no overhead on the size of the actual raw files and the normal raw file formats (eg CR2) would work fine.

      What would make a difference is that you can’t have circular layers in Photoshop :) So when editing these files there would be transparent pixels in the layer taking up precious memory – so you’d need significantly more resources than before to edit the raw files.

      The quality of the image in the corners is of no consequence if you’re using the circle to create a standard 3:2 image. Whichever way you crop it, landscape or portrait, or at ANY angle, you are still getting the same relative area of your image circle into your crop meaning you’re still getting the same levels of quality. If you choose to crop a different aspect ratio then of course you get differing results, but then I’m always of the opinion that having *something* to work with is better than nothing.

      • Maciek

        Well, lossless compression (I believe the systems claim to have fixed the issue but never mind them :)) does not change the fact that the raws would be larger unless we’d reduce the resolution. Ignoring the corner data, that would compress well, there would still be much more information that we potentially want to use. Except for most of it we wouldn’t as long as we crop to a rectangle.

        Handling extra data would slow down the whole platform or made it more expensive to produce something that would keep up to current standards. Look at how long it takes to introduce 4K video. Adding plenty more information that needs to be transferred form sensor, processed by CPU and pumped down to an sdcard would be a significant overhead.

        Moreover, photographers frequently know the orientation in advance (landscape are usually wider, portraits are usually taller) and when they don’t they more often than never crop tighter than necessary to achieve desired look. Therefore, the benefits would be limited to few cases.

        By deteriorated quality at the edges I meant going crazy and shooting wide panoramas in one frame (say 20:1). Then the edge quality would matter.

        Anyway, it is an challenging idea and I’m waiting for some company to prove all my arguments wrong ;)

  • Unless the price of the sensor go down,it’s not a good idea

  • The horizon will be ok every time

  • Eddy Kamera

    Bigger area = more expensive sensor… other than that, it’s a good idea.

  • Motti Bembaron

    Actually, square sensors make more sense. The area of a square sensor (the spans to the edges of the circular area allowed by the lens) has larger area than the traditional rectangular sensor.

    • GottaSay

      I agree. I have been waiting for one for a long time.

  • GottaSay

    You don’t have a clue how the sensors are made, do you? If you did you won’t waste time on this ridiculous article.

    • Jolyon Ralph

      Not only did you waste time reading it, you wasted time replying! I win :)

      But yes, to be fair, you’d probably need to cut a rectangle at a time rather than stacking circles as I suggested. In which case there are far fewer sensors that can be produced from a single wafer.

      But then I’m thinking that wafer space in the corners may not need to be wasted – maybe there is the opportunity to use that space for some of the support circuitry for the sensor seeing as that is fabricated using the same processes as the CMOS sensor.

  • Sounds to me one would only gain a fraction of height for portraits and a little bit of width for panoramic shots, presuming current sensors were enhanced size-wise (looks like it from the picture). I am not opposed. Most pictures have to be cropped anyway. But, I have a feeling this won’t happen. I’d rather have an MF and call it a day. :)

  • Thom

    You can experience a similar feeling by mounting an APS-C lens on a full frame camera. To some extent at least.

  • Tj Ó Seamállaigh

    hmm food for thought, though it needs lot of work to make it practical. If it gets practical and people find it useful, be sure that it will be the Thing soon.
    It might make my panorama stitching a bit harder though, but I wouldn’t know until I try.

  • .. with circular noise :D

  • phillip mccordall

    It wold be great for me, so many clients ask for different crops, this would be a superb advantage . Except for price and of course the size of the camera required. I’ve read some remarks that show very little understanding about sensors and quality like ” set the camera to 1:1″ something that will only limit even more the possibilities , another about lenses , a lot of readers are totally missing the point of why this would be great. If it’s possible I’ve no idea though.

  • KC

    It’s been done – in film days. The Kodak 1 (1888) made circular images.

    It’s an interesting idea, but if the point is a multi-aspect sensor, and not having to flip a camera for verticals, why not a square sensor? Why not a rotating sensor? There’s precedent there. The Rollei TLR’s with their 6X6 format, the Mamiya RB67 with its rotating back.

    It’s something I’ve always been curious about – but why are we still using 3:2? With the shift from film to digital there was a choice of any format, even a new format, but 3:2? How did that get to be the “one”? 4:3 is closer to square.

    The point is we could have broken free of 3:2 during the transition and it could have changed camera design. Instead we locked into a 35mm SLR metaphor

    I like the idea of an oversized sensor, though.

  • Richard Anderson

    Circular, not sure. But 4:3 would be better in APS FF than 3:2, that ancient motion picture film format from 1909 or something. Wastes the light cone.

    • KC

      Whoa. That’s sacrilege to some people. 24X36 is a “holy” relic. Not that 4:3 isn’t. All kidding aside, an opportunity was missed with digital to come up with an entirely new format and form factor.

      A circular sensor might be problematic from a production point of view, but not a square sensor. The RB/RZ67 had a rotating back, so did view cameras. (Imagine the dynamics of flipping them on their side.) So, a square sensor that you can use 1:1 and also “flip” horizontal and vertical for 3:2 or 4:3 makes a bit of sense. Like a 6X6 roll film SLR. Better still, it would allow the battery to be moved to the bottom (and be bigger), and a rotating grip could be used on the side for ergonomics.

  • It’s done. In digital era. Nokia did it in PureView 808 Symbian device and later in Lumia 1020 WP8 device. These had 41 Mpx circular sensors… … …

  • Rob

    Sounds good for different crops, but isn’t logical for production reasons. It would be extreme inefficient, and therefore expensive, to make round sensors for manufacturers and waste a lot of silicon.

  • IshmaelDNS

    Great story but I am sorry circular images are not stupid. They are great if you know with to do with them. I wish my camera could create circular images.

  • Atrox Donnerkail

    You forgot to take into account the processing power would need to be increased! Its a big difference to save those 41% more megapixles on the sd card. For some it doesnt matter I know
    Also where do you cut those Circles? I mean on a Pixel level it could be really hard to make curved edges with those Pixels! I think the cost would increase more than 3x!
    Also take into account that the image-quality is more than bad in those extreme corners of a standard lens. So the main area you are gaining is just poor quality…

    The easiest way you could get Circular Aspect ratio ist to just make the Sensor much bigger, like using a APS-C lens on a FF Body…. But you wouldnt want to do that right? ;)

Recent Posts