Sensor maximisation?

[Cap'n Fishy]

I was wondering why camera manufacturers still mostly produce 3 x 2 rectangular sensors...



That shape is wasting much of the incoming circular image. It also means you have to turn the camera back and forth between landscape orientation and portrait orientation. If you wanted to stay with a rectangle, then a square sensor would maximise the use of the circle. And it would negate the need to turn the camera back and forth and negate the need for the bulky double-grip arrangements built in to pro bodies and offered as expensive add-ons for lesser models...



You shoot a square and then extract whichever aspect ratio and orientation you want - either manually in post or by dialling it in to the menu. The aspect options are already being built in to cameras.

If you take it to its natural limits, you really want a circular sensor that captures the whole image area. That can then be cropped to any aspect ratio and orientation - again by dialling it in or in post.

It can't be a simple question of cost, as the high-end bodies already cost an arm and a leg! What's an extra kidney?

 

[paddler4]

I think it's just a carryover from the aspect ratio of 35mm film.

I've thought about this recently. To get a lighter kit, I am switching to an OM-1 Mark II, which is only 20 MPX, compared to the 24 of my R6 II. However, in practical terms, I think their pixel counts are more similar to that because in many cases, I expect to crop less off the sides of the 4:3 captures than from the OM-1 than I would from a 3:2 capture from the R6 II. However, there are times when the reverse is true...

However, since most people want images that aren't square (let alone circular), you would end up wasting a lot of expensive silicon to use a square or circular sensor.

 

[Cap'n Fishy]

I think it's just a carryover from the aspect ratio of 35mm film.

... which surprises me that it still persists, decades later.

However, since most people want images that aren't square (let alone circular)...

That's what I am saying - if you want images that are not square, you don't need to have them. You just dial in the aspect ratio and orientation that you want.

... you would end up wasting a lot of expensive silicon to use a square or circular sensor.

The square would not be much bigger than the 3 x 2.

The circular - you could restrict it to the high-end bodies.

On either, you would save expense by not needing to double-up the grip and duplicate controls. And it would be smaller and lighter.

 

[paddler4]

A square sensor that can capture as much detail as a 3:2 on the long side would require 1.5 times as much silicon, and it would require a larger (higher) and heavier body. I have no idea if it's still true, but back in the day, silicon was a major driver of the cost of cameras.

For most people, it makes no difference. I print a fair amount and print large occasionally, but most people don't print at all, let alone large. If you are only displaying on a screen or printing small, cropping off the ends of a 3:2 image won't matter for image quality. For me, when I am down in the 20 MPX range, it matters for my larger prints, but the fact is that the impact is very small given the quality of the up-resing algorithms in modern software.

I've been shooting SLRs since 1967 and have never owned a grip. For the amount I shoot in portrait orientation, I find that rotating the camera is fine.

 

[Cap'n Fishy]

I've been shooting SLRs since 1967 and have never owned a grip. For the amount I shoot in portrait orientation, I find that rotating the camera is fine.

So why do they add the expense and weight and bulk of putting grips on the high-end bodies?

 

[paddler4]

For people who do a great deal of portrait orientation shooting and are working fast, like sports photographers.

It's only done on a handful of bodies. For example, the R5 II is a very high end body, and it has no built in grip.

I've done a lot of shooting with other people, and I can't recall the last time I saw anyone with a built-in or added grip for portrait shooting.

 

[Cap'n Fishy]

For people who do a great deal of portrait orientation shooting and are working fast, like sports photographers.

It's only done on a handful of bodies. For example, the R5 II is a very high end body, and it has no built in grip.

I've done a lot of shooting with other people, and I can't recall the last time I saw anyone with a built-in or added grip for portrait shooting.

That's kind of what I'm getting at. I have not bothered buying a grip for my latest camera, as I can't justify the cost for the amount I go portrait-orientation. But there are odd occasions when I will have a session that is nearly all portrait-orientation, and it's a bit of a PITA to use the camera at right-angles. Hand-holding it, it's very easy to drift off the level - even with the spirit level in the viewfinder.

However - the R3 and R1 have a built-in grip - adding expense, weight and bulk - and they may not be used a great deal by some buyers. You could save the cost of building in a grip and put it into a circular sensor. Then you have 2 buttons assigned. One toggles between portrait and landscape orientation. The other cycles through the aspect options: 3:2, 16:9, 5:4, 1:1, etc. All of these will maximise the area of the sensor used to take the image.

 

[Mike_d]

I think Panasonic already has a camera sort of like this with an oversized sensor that can either be taller for stills or wider for video, but not at the same time.

 

[Wilt]

The increase in expense of sensor size is the increasing difficulty of producing sufficient manufacturing yields of sensors with more area without defects...why medium format sensors are still largely scarcely the size of 645 film format. That difficulty extends not only to the sensor but also to the LED display in the viewfinder, which also would need to support the choice of long vertical or long horizontal.

36mm x 36mm is not much smaller than the 40mm vertical that medium format has! And you know now expensive medium format digital still is.

 

[Cap'n Fishy]

The increase in expense of sensor size is the increasing difficulty of producing sufficient manufacturing yields of sensors with more area without defects...why medium format sensors are still largely scarcely the size of 645 film format. That difficulty extends not only to the sensor but also to the LED display in the viewfinder, which also would need to support the choice of long vertical or long horizontal.

36mm x 36mm is not much smaller than the 40mm vertical that medium format has! And you know now expensive medium format digital still is.

Still, in an ideal world, eh? There is a market for lenses costing north of £/$10,000. I think if they built it, people would buy it.

 

[MMP]

I imagine a larger, and especially a circular sensor, would be a nightmare for the engineers designing the lens. The optics would have to be near perfect from edge to edge, or the glass would need to be significantly wider in diameter. Even the absolute best lenses show some degree of vignetting, distortion, and a drop in sharpness on our rectangular sensors. This is why using a full-frame lense on a crop body can make the lens appear much better...you are sampling less of the peripheral glass and looking through a smaller portion straight down the center where most lenses perform best.

 

[Trespassers W]

I think pixel addressing in circular sensor would also be more complicated requiring higher processing power and different algorithms.

 

[Cap'n Fishy]

I imagine a larger, and especially a circular sensor, would be a nightmare for the engineers designing the lens. The optics would have to be near perfect from edge to edge, or the glass would need to be significantly wider in diameter. Even the absolute best lenses show some degree of vignetting, distortion, and a drop in sharpness on our rectangular sensors. This is why using a full-frame lense on a crop body can make the lens appear much better...you are sampling less of the peripheral glass and looking through a smaller portion straight down the center where most lenses perform best.

Sure, but nothing much is changing. You are not taking a circular image. You are still taking a rectangular image with the majority of it occupying the good area in the centre and with just 4 corners going to the edge - same as a full size sensor does now. It's just that you would have the option to rotate the rectangle and change the aspect ratio.

 

[Anton Largiader]

I've thought about this in the past, also, but the square sensor really isn't that efficient.

A rectangle in a circle can be wider than a square in a circle. So if you use the square sensor and crop it to your desired rectangle, it's smaller in the image circle than if you'd just started with the rectangle. With the conventional rectangle you're not just throwing away some top & bottom pixels, you're gaining some side pixels. Since most of us don't want to end up with square images, the rectangular sensor allows us to optimize for that.

EDIT: if you are advocating for a square that is the long dimension, so that there are always dead corners, then that would be functional.... but still not efficient cost-wise.

 

[paddler4]

To follow up on Anton’s point:

A full frame sensor is 24x36mm, 864 sq mm. To capture all of the detail that sensor provides in the long dimension, a circular sensor would need a diameter equal to the diagonal of the FF sensor. Since it would have to be cut from a square portion of the fabricated sensor sheet, it would require using 1872 sq mm of material, almost 2.2 times as much as is used for the FF sensor. That’s a whopping increase in cost.

Then here is the extra cost, weight, and nuisance of a larger body for the large majority of photographers who don’t have a camera with a built in grip or generally shot without an add-on grip.

Not going to happen.

 

[Nick5]

So why do they add the expense and weight and bulk of putting grips on the high-end bodies?

I shoot Architecture and in Portrait Orientation a lot. Years ago when I bought the 7D in 2009, I was thinking of adding a Battery Grip. What sold me was standing in front of a mirror and looking at your posture in Landscape Orientation and then in Portrait Orientation. Landscape and your body is aligned. In Portrait, your right hand is above your left eye on top, shoulder is raised and your Hip kicks to the right.
The addition of the Battery Grip keeps me aligned like Landscape while capturing Portrait Orientation images.
For me, a body without a grip hits me in a spot on the palm of right hand the causes irritation in minutes. The grip alleviates that completely as I’m holding the side of camera and Battery Grip.
Also the Battery Grip acts as a counter Balance with longer lenses.

 

[Cap'n Fishy]

I shoot Architecture and in Portrait Orientation a lot. Years ago when I bought the 7D in 2009, I was thinking of adding a Battery Grip. What sold me was standing in front of a mirror and looking at your posture in Landscape Orientation and then in Portrait Orientation. Landscape and your body is aligned. In Portrait, your right hand is above your left eye on top, shoulder is raised and your Hip kicks to the right.
The addition of the Battery Grip keeps me aligned like Landscape while capturing Portrait Orientation images.
For me, a body without a grip hits me in a spot on the palm of right hand the causes irritation in minutes. The grip alleviates that completely as I’m holding the side of camera and Battery Grip.
Also the Battery Grip acts as a counter Balance with longer lenses.

Right, so it is good to be able to turn the image to portrait mode for all the reasons you say. But you can do that with a circular sensor, without the need for the weight, bulk and expense of an added grip.

 

[Cap'n Fishy]

I've thought about this in the past, also, but the square sensor really isn't that efficient.

A rectangle in a circle can be wider than a square in a circle. So if you use the square sensor and crop it to your desired rectangle, it's smaller in the image circle than if you'd just started with the rectangle. With the conventional rectangle you're not just throwing away some top & bottom pixels, you're gaining some side pixels. Since most of us don't want to end up with square images, the rectangular sensor allows us to optimize for that.

EDIT: if you are advocating for a square that is the long dimension, so that there are always dead corners, then that would be functional.... but still not efficient cost-wise.

OK, so let's ditch the idea of a square sensor, and go for a circular sensor. Then it can be cropped to a rectangle of any aspect ratio, each of which will occupy the maximum number of pixels in the circle. One button to toggle between portrait and landscape orientation; one button to cycle through the aspect ratio options.

 

[Mike_d]

Trying to make a circular sensor makes no sense to me. So much would be unused. You'd need a 36x36mm sensor (purple) to be able to capture a 36x24mm frame in either orientation (red/green). That's a 50% larger sensor (36x36 vs 36x24), increasing cost. 11% of the sensor (purple squares) never gets used. Bad pixels in those corners don't kill the whole unit, which would help improve yields a bit. I don't know enough about chip making to know if it's even possible to make a + shaped sensor with the unneeded corners then going to an adjacent chip. Doing this with an APS-C sensor (yellow) would keep the cost down. Or maybe it would be cheaper to develop a mechanism for rotating the entire sensor 90 degrees on command. Maybe an IBIS on steroids could do the job.

 

[Anton Largiader]

This is all in the name of making rectangular photos. But what if we grew up in a culture where circular photos were the norm? Maybe we are kind of getting there with the use of round avatar photos.