2
\$\begingroup\$

I found existing questions(1)(2)(3) about square sensors, but those would be excessive for the purpose described here.

Introduction

For low aspect ratio photos, e.g. 1:1, 6:5, 5:4, and 4:3, the 3:2 ratio of APS-C and 35mm sensors wastes a significant portion of the image circle. A fairly modest 28% increase in sensor area allows:

  1. All ratios from 3:2 to 1:1 to make native use of the lens, providing 61% more image area in the case of 1:1 (square) photos
  2. Electronic viewfinder (live-view) overscan for easier subject tracking
  3. Aspect ratio to be chosen in post-processing without additional cropping, if the full sensor readout is saved.
  4. Correction of barrel distortion with less or none of the cropping typically incurred

Formats like 5:4 and 1:1 may be less popular than 3:2 but they are hardly rare. The fractional increase in sensor area (and the mechanical and electronic changes needed to support it) shouldn't have a major effect on price; it wouldn't push a camera into a different class entirely (cf. the ~132% sensor area increase from APS-C to 35mm). Even at significant cost this would be a fine differentiating feature for flagship cameras; since it is often stated that lenses are the primary investment in a photographic system getting the most out of them should be a strong selling point.

Given the benefits listed above, illustrated below, I have to wonder why isn't this already commonplace on high-end cameras? Is there something that makes manufacturing a sensor, shutter, or IBIS unit 28% taller really that prohibitive? Are the benefits I believe exist incorrect or unappealing? Are people simply unaware of this technical option and therefore haven't voiced a demand for it?

Preemptively addressing possible concerns:

  • File size, buffer speed, etc. should remain for same for a selected 3:2 aspect
  • A standard 3:2 EVF could be used by simply displaying other formats a bit smaller
  • Vignetting need not be shown unless a full-sensor-readout option is specifically selected

Image area

For APS-C lenses my proposal would be a 23.50mm × 19.95mm sensor, shown in figure 1 in light blue along with standard APS-C and 35mm "full frame" sizes (wine and pink respectively) overlaid with the the APS-C minimum image circle. Figure 2 highlights the substantial difference in image area between a 1:1 crop of the APS-C sensor (green), and the native 1:1 frame (dashed).

Figure 1

Figure 2

Subject tracking

Referencing the figure 1, if a standard 3:2 image is desired there is unused area of the proposed sensor. This can be put to good use to provide a wider view in the electronic viewfinder which would make it easier to track a subject, e.g. birds in flight. (Appropriate crop guides for the target format would be overlaid.)

Ad hoc crop in post

If an option is provided to save the full sensor readout to RAW (or JPEG for that matter) one could defer the decision of aspect ratio to post without any additional cropping. This could be used to save a poorly composed photograph, or simply for changing artistic desire. It would also improve the ability to meet a client's request for printing a given image in a specific format, e.g. 8×10, 11×14 20×24.

Barrel distortion

Barrel distortion, common in wide-angle lenses, typically requires additional cropping (yellow frame) after correction—either in-camera or post—because the sensor (green frame) does not capture all of the distorted image. (fig. 3)

Figure 3

Without changing the image circle it is still possible to capture a wider usable angle of view if the sensor is extended. (fig. 4) Obviously the recoverable area depends on the aspect ratio, with only the top and bottom recoverable at 3:2 and only the sides at 1:1.

Figure 4

recoverable image
recoverable image

Addendum

Wafer die yield

An answer made the claim

A sensor area with more square aspect ratio leads to considerably worse wafer yields.

Attempting an approximation using CALY die yield calculator I see a 35% increase in cost for the proposed sensor over APS-C, compared to a 232% increase for 35mm.

Standard APS-C 23.50 × 15.60 die on a 300mm wafer:

Using the proposed 23.50 × 19.95 die with the same parameters:

As expected a far higher yield than 35.7 × 23.8 "full frame":

\$\endgroup\$
10
  • 1
    \$\begingroup\$ "multi-aspect sensor" seems like something you've made up / a sensor larger than the imaging circle. \$\endgroup\$
    – xiota
    Nov 2, 2019 at 23:55
  • \$\begingroup\$ "The fractional increase in sensor area... shouldn't have a major effect on price;" -- You're proposing the creation of a new sensor format with an aspect ratio no one else uses. The development and low volume would result in a higher price. \$\endgroup\$
    – xiota
    Nov 2, 2019 at 23:58
  • 2
    \$\begingroup\$ @xiota (1) Panasonic has this on some of their Micro Four Thirds cameras, though they cover 16:9 to 4:3 I think instead of 3:2 to 1:1, and I think that's what they call them. (2) If most flagship APS-C cameras used this format the economy of scale would exist; of course someone has to be first but it's not like introducing an entirely new lens mount etc. \$\endgroup\$
    – Mr.Wizard
    Nov 3, 2019 at 0:06
  • \$\begingroup\$ Which Panasonic models have this? \$\endgroup\$
    – xiota
    Nov 3, 2019 at 0:55
  • 1
    \$\begingroup\$ @Mr.Wizard Have you considered submitting this to PetaPixel as a guest article? Change it up from "why don't" to "why camera makers should..." \$\endgroup\$
    – mattdm
    Nov 3, 2019 at 19:53

2 Answers 2

1
\$\begingroup\$

Why aren't multi-aspect sensors commonplace on high-end cameras?

You've listed plenty of technical reasons this seems like a good idea. The reason probably isn't technical.

For years, a few camera enthusiasts online were calling for a compact, fixed-lens camera with a large sensor. And for years, there weren't any. Eventually, Sigma took the risk and it turned out there was a decent market — despite plenty of quirks and foibles in their design, they sold that camera as fast as they could make it. So, others followed — Fujifilm, Sony, Pentax, even Canon and Nikon.

Once the market was proven, and consumer demand demonstrated, the category took off.

Meanwhile, Panasonic has put out several cameras with multi-aspect-ratio sensors. Unlike with the large-sensor compacts, press and blog and forum response has been "meh" (if it's mentioned at all) — and market response has been likewise tepid.

I have no insider data, but given this I can see why the camera makers wouldn't prioritize this in feature development.

On the technical front, there are two things I can think of which you haven't mentioned. Both relate to the sensor height. The first is shutter travel; the extra distance means faster travel to get the same sync speed, and probably in general comparable prices to full frame. Second, on DSLRs, that height equates to room the mirror needs to get out of the way. A high-end APS-C DSLR would probably be about as big as a full-frame one in order to accommodate this, at which point they might as well just make that model full-frame to capitalize on the ineffable mystique that sensor size has somehow captured in many people's minds.

\$\endgroup\$
1
  • \$\begingroup\$ Good update. Thanks. \$\endgroup\$
    – Mr.Wizard
    Nov 3, 2019 at 22:53
0
\$\begingroup\$

If you take a look at the raw images of typical compact cameras these days, you'll find that they contain considerably vignetted corners. So indeed the wide angle range is only attained by working from an oval imaging area that is inflated back to a rectangular picture. As a result, corner sharpness at wide angles tends to be lacklustre since there are fewer active sensor pixels than pixels in the distortion-corrected image.

A sensor area with more square aspect ratio leads to considerably worse wafer yields. For similar reasons, laptop screens these days tend to be "wide" which is marketing speak for vertically challenged.

Since it is expensive to provide more sensor area than will often be used, multi-aspect sensors that will for any given image use only a subset of the available pixels are comparatively rare. In practice, also for sensor protocol and processing reasons, "multi-aspect" sensors tend to be comparatively standard 4:3 sensors with rather little (if any) of the sensor lying outside the image circle. When they actually produce 4:3 aspect ratio images, they deliver fewer pixels than using the same sensor in a fixed-aspect camera would.

\$\endgroup\$
7
  • 1
    \$\begingroup\$ "A sensor area with more square aspect ratio leads to considerably worse wafer yields." Would you explain this in more detail? Is it more than simply the 28% increase in area, and something about the aspect ratio itself? \$\endgroup\$
    – Mr.Wizard
    Nov 2, 2019 at 22:35
  • \$\begingroup\$ I don't know if you will be able to respond (it seems you did not register an account) but I attempted to address your second point in an addendum; is it reasonable? I am curious about "also for sensor protocol and processing reasons" too. \$\endgroup\$
    – Mr.Wizard
    Nov 2, 2019 at 23:28
  • \$\begingroup\$ @Mr.Wizard The larger the sensor size, the lower the yield for the same number of defects on a silicon wafer from which the sensors are cut. Each single defect affects a larger area of the total wafer as being unusable for larger sensors than for smaller sensors. \$\endgroup\$
    – Michael C
    Nov 3, 2019 at 17:03
  • \$\begingroup\$ @Michael Yes, that's true, which is why in the addendum I get a ~35% increase in price for a 28% increase in sensor area. (I wonder if unusable defect rates are that high in reality?) I don't see that as prohibitive however considering the same estimate puts the cost of "full frame" sensors ~230% higher. \$\endgroup\$
    – Mr.Wizard
    Nov 3, 2019 at 19:12
  • \$\begingroup\$ 35% price increase for 28% area increase is miniscule and nowhere near the actual difference in reject rates. 230% is closer to reality. For very large sensors, sometimes there are zero usable chips cut from a single wafer, as every single one of them has unacceptable defects. \$\endgroup\$
    – Michael C
    Nov 4, 2019 at 14:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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