Why a Two-Sensor Camera Could Be a Fantastic Option

The concept of creating a mirrorless system is very similar in the case of all the camera manufacturers: get rid of the mirror and create a new lens mount closer to the sensor. What if there was another option to do it, arguably a better one?

When we started moving from DSLRs to mirrorless camera systems, we entered uncharted waters, not exactly knowing what to expect. The first instinct was to make the cameras smaller, more portable — tiny body with a pancake lens, a hiker's dream setup we still can get in many of the systems (e.g. the Fujifilm X-E4 or Sony a7C). But then came the realization: while we can make the bodies smaller, it will happen at the cost of ergonomics. And it's even more difficult with the lenses, as laws of physics cannot be bent to the will of even the best marketing division of a camera manufacturer. And so, our mirrorless journey, instead of smaller setups, led us to 2 lbs+ behemoths like the Panasonic Lumix S1 or Canon RF 85mm f/1.2 L — great and capable tools, but no different in size from their DSLR-era counterparts. Or even worse, they were bigger and heavier.

If the size of a mirrorless camera isn't its advantage, then what is? That's a fairly easy one: the autofocus system, with subject tracking, eye detection, animal detection, and so on. Modern electronic viewfinders are a technological marvel as well. The thing is: we didn't need new lens mounts for any of those. We even had Sony SLT cameras for a while, a crossbreed of a DSLR lens mount and an electronic viewfinder. As sensors got the on-chip PDAF systems, this could have been developed in a totally different direction.

The Two-Sensor Camera

Imagine we take a DSLR or SLT camera and replace the focusing screen (in a DSLR) or AF module (in an SLT) with a second sensor, leaving the mirror to split the light between them. Here's a diagram to illustrate it.

Above: a DSLR cross-section (Source - Wikimedia Commons, CC BY-SA 3.0), below: modified diagram for a twin-sensor camera
Such a device would need a modern sensor (two of them) with on-chip autofocus and an electronic viewfinder. IBIS would also be a welcome feature. It would technically be a mirrorless camera with a mirror.

The first advantage of such an approach is pretty clear: we can keep the old lens mounts and the lenses we've gathered over the years. But that's the least important one. Imagine what can be done with two sensors and a mirror between them that splits the light evenly or at a certain ratio. Here are some concepts; let's start with those where light is split evenly (1:1):

  • Two Bayer mask RGB sensors shifted by 1/2 pixel. This is a fairly simple option focused on increasing the resolution. Think of it as pixel shift technology without the limitations of multiple exposures, which means it would also work in situations like sport or wildlife photography. 
  • A Bayer mask RGB sensor and a monochrome one. Combined, they could provide for a better resolution and lower luminance noise. Such a camera could also be turned into a monochrome one with the flick of a button (that would lift the mirror up). No more dedicated monochrome models would be needed. 
  • A Bayer mask RGB sensor and an infrared one. A similar solution to the one described above, but allowing for simultaneous capture of color and IR information may be useful in some cases. 

If we go for a mirror that splits light unevenly (e.g. 16:1, which translates to a 4 EV difference), we could, on the other hand, achieve what computational photography is doing in smartphones: expand the dynamic range. And doing it with two sensors and simultaneous capture would introduce the benefits of HDR to genres of photography that weren't compatible with that technique before, like sports or action. Videographers could also harness this.

The sensors wouldn't need to be the same: one could be a high-resolution module wired to capture the details in highlights and midtones, and the second one could have bigger pixels and a base ISO chosen with the purpose of capturing shadow details in mind. Imagine having a Sony a7R IV and an a7S III in one body.

There's one last possible variant is a non-translucent mirror, that either reflects 100% of light to sensor one or moves up, revealing sensor two. Those sensors could differ in size, such as a 24-megapixel full-frame sensor and a 24-megapixel Micro Four Thirds sensor. Such a combination could be a nice option for bird photographers: a larger sensor with bigger pixels for low-light situations, and a smaller, more packed one, to get the extra reach when light is sufficient. No need for a second, crop-sensor body in your camera bag.

Potential Problems

While such a dual-sensor camera could widely expand the creative possibilities for both photographers and filmmakers, there are still some issues to be considered. Let's discuss them:

  • Synchronizing the sensors. If such a camera offered IBIS (which would be very welcome), the stabilization modules would have to work in perfect sync. The same goes for on-chip autofocus units, assuming both sensors had them.
  • Manufacturing. If we take several sensors with different resolutions and/or capturing range (RGB, monochrome, IR, etc.) and multiply that by several mirror variants (1:1 light split, uneven light split in several variants, no split), the resulting number of camera models would be overwhelming, both for the manufacturer and the customer. Getting it to work would probably require an approach used in the car industry, where the customer orders a certain engine, trim package, and color, and picks up the car several months later. Not everyone would be happy with such a sales model replacing a store shelf with cameras on it. 
  • Old lens mount. While keeping the EF, F or A mount would be great for the customer, it wouldn't be for the manufacturers. With new mounts, they can sell us the same (or better) lenses again, but with a higher price tag. 

Are we ever going to see such a camera? With the industry shifting so rapidly towards new lens mounts, it's not very likely. But it's nice to have dreams. Maybe Pentax will come to the rescue?

Log in or register to post comments

14 Comments

David Yoon's picture

They’ve had 3 CCD video cameras that work in a similar manner. One thing to consider is that you’ll need to calibrate the mounting of that second sensor like the autofocus sensor in DSLRs. But in an age where one can design sensors with dual gain outputs and sub pixels like Canon does, does a dual sensor architecture get you much in terms of performance?

Hilbren van Doorn's picture

Agreed alignment of senors is a big problem! Especially when knowing this problem had just been solved with mirror-less cameras.
But the concept is interesting, i think the same could be achieved with a Light-field camera, even the suggestions that Kelly hofer gave, with using different polarization's. If you add everything together you get: higher dynamic range, polarity selectivity, selective resolution and selective focus. All in post processing. The only downside is the all these functionalities is degraded light sensitivity.

Robert Edwardes's picture

They still have 3 CCD camcorders, most broadcast systems are that still that with sony coming out with 2 in the last 5 years. 1" camcorders are still relatively new, and aren't as good in some aspects since a 3 CCD has 1 sensor per color so the camcorder had less processing while a 1 sensor system has to run a demosaicing algorithm

Lee Christiansen's picture

Yes, my Sony PMW-500 broadcast camera is a 3 CCD camcorder and it delivers stunning images.

Whilst the world has gone nuts over single CMOS, larger sensor cameras - often recording in Log modes, the PMW-500 sticks with the concept of a knee curve which helps get large dynamic ranges with ease and using zoom lenses that can open up to f1.9 it means the difference in depth of field with a C300 type camera and a f2.8 lens is only truly about 1/2 stop.

I still have a great love for 3 CCD cameras, and they handle some things a lot better than CMOS, (no bending verticals, no issues with flash lighting...)

kelly hofer's picture

I believe there is something to be explored here. Adding polarizers to both with opposing rotations, would open up the world of comparative capture. One could also have a nd filter and capture at a much higher or lower exposure to give you up to twice as much dynamic range.

David Greenberg's picture

Love the dream, but I'd worry that the slightest knock could misalign the sensors by a pixel or two. Maybe that could be rectified with a constant self-check and instant firmware recalculation.

Dave Haynie's picture

The back sensor is on an IBIS gimbal, allowing it to lock up to the top sensor any time both are used... no IBIS of course. The question I have is the mirror... can we really make a variable beamsplitter, or is it always fixed split if top sensor is used for anything?

Lawrence Huber's picture

Interesting concepts.

Michael Gerbracht's picture

I think it would also be interesting to use film as th second "sensor". So you could have a hybrid camera or just use the sensor for AF and a backside display but shoot film only.

Michael M.'s picture

I think alignment is not a huge issue, you don't need much computing power to realign two images. So every camera could be calibrated and re-calibrated once in a while. But you may loose increased resolution if the shift is exactly the 1/2 pixel you used to offset the sensors.

Anyway, it's a great concept!

Dave Haynie's picture

You have two criteria for alignment that matter. One is 3D alignment, aka focus. They have to exactly at the aame focal point, a problem never well solved by DSLRs. The other is higher resolution, combining both sensors. While you could incorporate multiple unaligned in a Drizzle, but for the single shot you need a half-pixel interstitial offest between the two chips.

Tim van der Leeuw's picture

You need pixel alignment (X and Y axis), but also depth-alignment (Z axis) and ensure that there's no rotation of any of the sensors wrt. the plane of focus.

The last 2 will be much harder to achieve than just pixel alignment, since incorrect Z axis alignment of one of the sensors means that that sensor will record a slightly out-of-focus image.

I think it's a solvable problem using the same mechanisms as used for IBIS, but it has to be done by the camera while focusing, before taking the shot.

Michael Parsons's picture

Interesting!
I suggest two mounts each with its own sensor is the way to go - there was a discussion on dpreview (Micro 4/3 forum) a while go about a new Micro 4/3 body which has many of these advantages:
* Portraits with *two* focal points, e.g. a near and far subjects (e.g. mother and daughter).
* Combine the images in some clever way. Obviously stereoscopically, but I think you could do some really interesting computational photography with the two images taken at different focal lengths.
* With two mounts I could use two fast primes, say a 17mm F/1.2 and a 45mm F/1.2 - I don't know of a zoom this bright.
* With two mounts and the same lenses I can combine two images, say for panoramas, anamorphic look or for stereoscopic images
* I could use one vintage lens and one modern lens for lovely bokeh effects and also focus on something moving quickly.
* Take one shot in focus (for the portrait subject) and a second shot defocussed (for the background) simplified as they can be done at the same time.
Mike.

AJ L's picture

Interesting idea. Really complicated. I don’t know if these designs would work in the market.

What would have been interesting would have been if, say, Nikon has just taken the D850, D750 and D5xxx designs and replaced the mirror and prism with an EVF, making F mount mirrorless cameras. They’d have been the mirrorless cameras with the best lens selection immediately and nobody would have caught up.