Why are there (at least to my knowledge) no rotatable sensors instead of grips for portrait photography. It seems like a great solution to obtain a lightweight DSLR for both landscape and portrait photography. Are there technical difficulties?

The reasons for not having a rotating sensor that I can think of are:

  • Technical difficulties
  • No demand of it since portrait photographers want a bigger camera anyway since most sessions are in studio and weight is not an issue.
  • The camera would be more fragile having the rotating construction.

On the other hand I see secondary advantages having a rotating sensor such as automatic horizontal leveling and even image stabilisation from roll movement.

Is there something I'm just missing here?


5 Answers 5


The camera would need to be quite a bit larger and heavier in addition to being more complex and fragile.

We've been there before, back in the film days, when the popular medium format cameras were 6x6 square format SLRs like the Hasselblad 503cx and the Rollei 6006, and Mamiya's rotating-back 6x7 format RB67 and RZ67. Because the Mamiya backs rotated, they actually had to be 7x7 cameras (the 6cm dimension was the width of the film roll). That meant that the mirror and focusing screen had to be big enough to support an image that was 7cm in either direction. The Mamiyas were huge compared to the 'Blads, and that was just to get a 4:5 aspect ratio working in both orientations.¹ It would be considerably worse with a 2:3 aspect ratio.

Because an APS-C-sized sensor is as wide as a full-frame sensor is tall, you could fit the works into a standard SLR mirror box without modifying the lens system, but to make it work with a full-frame DSLR, you would need to use a different lens mounting system with a longer register distance to accommodate the larger mirror. You would also need to worry about masking the viewfider for the format currently in use, and creating a viewfinder that covers the larger screen with a reasonable eyepoint and magnification (and VF info somewhere near the image without occulting anything).

In addition to building the reflex mirror and viewing screen big enough to support both horizontal and vertical images, a focal-plane shutter would need to travel the long dimension of the sensor as well. An APS-C-sized sensor would need, in essence, a full-frame shutter, and a full-frame camera would need the shutter from a Leica S. Either the sync speed is going to come way down, or the cost is going way up.

Mirrorless cameras would make this scenario a lot more practical, at the expense of making the body itself slightly bigger and heavier. And, of course, more mechanically complex and fragile. The lack of mirror and focusing screen, at least, means that the lens mount would not need to change.

Lens hoods, though, would need to be less effective. The current "tulip" shape in use is designed to closely fit the rectangular image without vignetting. If the sensor rotates, the lens hood would either have to be round or square to accommodate both formats.

¹ "6x6" was actually 56mm x 56mm, and "6x7" was 56mm x 70mm. The film was 6cm wide, but the image had a 2mm border.


The entire light box, including the rectangular reflex mirror would need to be rotated as well. You would also either need a second prism/viewfinder to catch the light coming off the rotated mirror or you would need to be able to rotate the prism/viewfinder as well. The same goes for the AF array in the floor of the light box that catches light off the reflection of the secondary mirror attached to the reflex mirror. At that point you've already rotated half the camera! It is far easier to place a set of vertical controls on the bottom of the camera (as top pro bodies such as the Canon 1D X or Nikon D4 do) or on an optional battery grip that does the same thing (for most other DSLRs).

  • \$\begingroup\$ Would it not be possible to use a quadratic mirror instead and an AF array that is diagonally symmetrical and actually not rotating anything apart from the sensor? \$\endgroup\$
    – Hugo
    Jan 22, 2014 at 5:22
  • \$\begingroup\$ It would increase the size of the light box which would in turn require a longer registration (sensor to mounting flange) distance which would in turn mean current lenses would not work with the new camera. \$\endgroup\$
    – Michael C
    Jan 22, 2014 at 6:09
  • \$\begingroup\$ As to the AF array, it must be oriented at precisely the same distance as the sensor AND most AF arrays are not square: they are also rectangular. \$\endgroup\$
    – Michael C
    Jan 22, 2014 at 6:11

I think the short answer is that you'd have to design a new mount specifically to achieve this, or use a smaller sensor than the mount would otherwise support (e.g. ASP-C sensor with an EF-only mount; no EF-S).

If you take the lens off a Canon 5D, for example, and look at the mirror/sensor (you can use Bulb mode or the sensor cleaning mode to see it) you'll see there's very little room around the sensor for anything else. The mirror folds up to just above the top of the sensor, and when down it is as far forward as the rear of an EF lens permits.

There's two ways you can go with this...

Allowing a rotating full-frame sensor, would require that there is effectively space for a square sensor (36x36 mm instead of the standard 36x24 mm for full-frame). This would thus also mean having a larger mirror to cover that whole space (and a larger viewfinder), and that the mirror would fold up higher to accommodate the sensor in portrait mode.

Since the mirror must sit at 45º, and needs to fold up to allow the image to be captured., the space between the sensor and the lens flange would need to be increased. The current sensor-flange distance is minimised, allowing just enough space for the existing mirror here, and the lens design relies specifically on focusing light onto the sensor at the standard sensor-flange distance. You can't just change this parameter without extra optics (which would need to fit in that space too!).

In short, allowing a portrait sensor means a larger, higher mirror, and a longer sensor-flange distance than the lens mount was designed for. I guess if you were really keen you could design a mirror that gets out of the way by another mechanism (e.g. flips and slides), but this would be more complex, more prone to failure, and likely much slower.

However... if you want to keep the existing sensor-flange distance, you could limit the size of the sensor accordingly. The height of a full-frame sensor is 24mm, so an APS-C-sized sensor would fit in this space in portrait mode (Nikon ~23.6mm, Canon ~22.2mm).

However, you'd need to keep the same sized mirror as the full-frame camera, and thus could not use lenses that extend further into the body (like Canon's EF-S mount lenses). The pentaprism and viewfinder would also have to be full-sized to accommodate portrait-mode.

In short, you'd be just putting a rotating APS-C sensor in a full-frame camera. Or if you start from an APS-C camera, you'd be dropping the sensor size down to a bit smaller than Four Thirds size (i.e. to max ~15mm width). Either way it seems like you're losing more than you gain.

In a mirror-less camera this might be doable (since you don't have a taller mirror to get in the way), but then that's no longer a DSLR (and it may be cheaper to make square sensors than rotating ones, especially at the smaller size of typical mirrorless sensors).

Finally, an extra complication would be the precise calibration of the sensor-flange distance: a rotating sensor means this could easily vary as it rotates, or more easily get knocked out of alignment so it's not perfectly parallel with the lens mount flange.


You would need a view finder that could display the image in portrait orientation. You would need a shutter that can open and close the direction you want. You would need to be able to adjust the AF sensor. There is far more complexity and things to break. It is far, far easier and more reliable to turn the camera.

If there was a camera, even at the exact same price and quality (which wouldn't be possible, it would have to be much more expensive), that had a rotating sensor vs one that required rotating the camera, I would pick the one that requires rotating the camera every time. I'm fairly certain most professional photographers would choose the same for the reliability of a system with less moving parts and complexity.

There is simply no sufficient market for such a device to be financially viable.


You want to reach into a precision optics system, take its most delicate components, and turn them into moving parts? Yeowch! That's a horrible way to get an orientation-agnostic camera. Maybe we could have a better plan. Let's think about an orientation-agnostic camera that doesn't require precision mechanical rotation mechanisms.

We'll do it in electronics. We'll skip the mechanical rotation and make a camera that has a square sensor, then crop it and throw away data on the sides that we don't want. That's easy enough. Of course, we'll be paying more for the extra sensor area (and related components) that goes to waste. So it's modestly more expensive, but plausible.

We could also just switch to shooting squares... it worked for Instagram... :b

That still wouldn't buy you the automatic leveling, though: we need arbitrary rotation. For that, we need to think circles! Fortunately our lens is already circular. You could, in fact, ask why we just don't shoot circles for all our photos. (The reason is that circles are horrible and nobody wants them. Film cameras need to contend with the fact that the film is a straight line. And digital photos are still based on rectangular pixel data.)

But anyway. Circular lens. We'll also need a circular sensor1, and an effective way to turn our sensor data into rectangular pixel-data after being digitally rotated to arbitrary angles. We can do that, but we lose fidelity, because pixels aren't in 1-to-1 correspondence with sensor elements anymore. As soon as we invoke rotation, we lose effective resolution immediately.

We can turn to information theory and see how bad this information loss is. The Nyquist-Shannon sampling theorem says our sampling of a signal needs to be at least twice the frequency to capture it perfectly. So if our goal is a true 8 megapixel representation of a two-dimensional rectangle, then we need a sensor with 32 megapixels (that is, in the rectangle - we will need more outside of it). There will also be a lot of math rotating and scaling 32-megapixel matricies with high enough precision to avoid loss of fidelity. If you want to do this on-camera you'll want a dedicated chip for that, something GPU-like.

You can see how this starts to look unnecessarily expensive, at least given today's technology. Perhaps after many more years of camera technology development it will become common, though.

1 Okay, it wouldn't be a circular sensor, it'd be another square sensor where you just happened to only use the pixels in a circle in the sensor. Easier to manufacture and to read pixel data from.

  • \$\begingroup\$ Regarding mechanically moving sensors they already exist, albeit in another form; sensor-shift image stabilisation. Now that I'm digging into the literature I've actually found cameras that rotates the sensor to stabilise roll movement of the camera, Pentax K-7/K-5. \$\endgroup\$
    – Hugo
    Jan 23, 2014 at 6:47

Not the answer you're looking for? Browse other questions tagged or ask your own question.