Sensor types - what cameras use which sensors?

[Dan Wells]

Both in full-frame and (especially) APS-C, there seem to be a lot of shared sensors. Canon sensors are almost always unique (sometimes shared among Canon models, but I've never heard of a non-Canon camera using a Canon sensor, and I've never heard of an interchangeable lens Canon using a non-Canon sensor (some of the point and shoots do)... Apart from Canon, there are a lot of Sony sensors around... The similarities are in the base sensor - things like AA filters, phase detection points and color filters (Fuji's X-Trans is a weird color filter on an otherwise standard sensor) are added on top.

As far as I know:

APS-C :

Newest generation Sony 24 MP (copper wiring, improved dynamic range) - may be "IMX 271"
Fuji X-Pro 2, X-T2 (unusual color filter),Fuji X-A3? (if so, it uses a standard Bayer filter, unlike the X-Pro 2 and X-T2) Sony a6300, a6500, Pentax K70?. Sony may have intended to release this sensor first in the a6300, but Fuji beat them to it, both announcing and shipping the X-Pro 2 ahead of the much-delayed a6300.

Sony 24 MP  IMX-193 (very long-lived standard 24 MP APS-C sensor)
Sony a6000, various Sony SLT models, Nikon D3300,D3400,D5300,D5500, Pentax K3,K3II, probably other Pentax

Toshiba 24 MP that seems to be a relative of the IMX-193 (Sony now owns Toshiba's sensor business)
Nikon D5200, D7100,D7200

"Nikon 24 MP" that may also be a relative of the IMX-193
Nikon D3200

Early Sony 24 MP (unique as far as I know) - less dynamic range and tricky color
Sony NEX-7

Sony IMX-071 16 MP - extremely long-lived standard 16 MP Sony sensor. Released in the Nikon D7000 in 2011, still in use in a couple of current Fuji models.
Nikon D5100, D7000, various Sony SLT and mirrorless, various Pentax, ALL Fuji X-Trans prior to X-Pro 2 (unusual color filter)

FF:

Sony IMX-128 (and perhaps close relatives) - 24 MP
Nikon D600, D610, D750, Sony A7, A7II, A99, A850? - essentially anything 24 MP FF except Leica and a couple of early models

Sony IMX-028 - 24 MP
Nikon D3x (unusual AA filter), Sony A900, A850?

Sony IMX-094 - 36 MP
Nikon D800, D800e, D810, D810a, Sony A7r, Pentax K1

Sony IMX-251? 42.4 MP (BSI, copper wiring)
Sony A7rII, A99II. It looks like this beast is the only sensor Sony won't sell on the open market, as there is no non-Sony camera with the unusually high resolution.

[SZRitter]

Being too lazy to look it up, aren't the Sigma sensors with Foveon a different sensor?

[scyth]

Sony A7rII, A99II. It looks like this beast is the only sensor Sony won't sell on the open market, as there is no non-Sony camera with the unusually high resolution.

why 'd Sony Semi not sell 42mp FF sensor provided that they (1) can manufacture enough after the earthquake and (2) there will be interested buyers ?

other than Sony Imaging buyers of FF sensors for consumer cameras are only Nikon and Ricoh/Pentax ... they might be either minding bottom line (for example Ricoh buying older and presumably cheaper 36mp sensor, which is not bad at all in a proper camera) or aware about production issues (not enough sensors) or waiting for the next iteration (may be Nikon for "70+ mp" sensor to go after 5Ds/r Canons) or actually simply are not yet ready with models using 42mp sensor

[Dan Wells]

        As I understand it, the Foveon is a VERY different sensor, complete with a different (and complex) fabrication method - I forgot all about it in the original list, because it has never sold enough to make any impact - it's appeared in a few fixed-lens cameras, but otherwise, it has only been in cameras using Sigma's wildly unsuccessful SA-mount (SA-mount is SO obscure that even B+H only bothers to stock three SA-mount lenses - a travel zoom, a 50 and a 70-200 f2.8 - everything else is special order). For B+H not to bother to keep a copy on the shelf, I suspect most of those lenses have total US sales of under 100 copies/year - I wonder if there are even 1000 Foveon bodies in the US (in 10 years teaching photography, I've never seen one in the wild - the only ones I've ever seen are at PhotoPlus - that probably makes them rarer than Phase One, which I've seen once or twice when taking a workshop or the like)?!?
      Some of the obscure lenses may sell a copy or two annually, or perhaps have never sold a copy at all (SA-mount is similar enough to Canon EF that Sigma can "announce" every lens they make in SA - if anyone actually wants one, they bolt a different bayonet on an EF lens). Something like a 500 mm f4 or a 300-800 f5.6 in SA mount almost certainly does not actually exist, despite appearing in Sigma's catalog, although it could exist in the 5 minutes it takes to swap mounts. Unless the new mirrorless SA-mount cameras take off, I suspect the mount and the terrible RAW conversion software make any virtues of the sensor mostly irrelevant...
     Despite its commercial irrelevance, the Foveon is an interesting design - it is essentially three sensors stacked on top of one another. In its oldest (and simplest) iteration, the three sensors all had the same resolution (one red, one green and one blue) - it was essentially the same principle as a 3-sensor video camera, except that the sensors were stacked (I'm not sure how light passes through two sensor layers to reach the bottom sensor, but it does), while a 3-sensor video camera (and one VERY old Minolta DSLR design) uses a beamsplitter prism so light hits all three color layers directly. The more recent Foveons put most of the resolution in the blue layer, and then augment with lower-resolution red and green layers to get color. And you thought X-Trans was a pain to convert...

       No other sensor is anywhere near as exotic as the Foveon - most "modified" sensors are standard Sony sensors with some changes in the AA filter or other "toppings", including IR-cut and phase detection. Manufacturers basically call Sony and say "give me 10,000 units of IMX-271 with no AA filter and this particular phase-detection overlay". Occasionally, someone will use an unusual AA filter that increases performance (at a financial cost) - Nikon's $8000 D3x used the same sensor as Sony's $3000 Alpha 900, but had an odd AA filter that seemed unusually sharp (and Nikon claimed it was very costly and helped explain the price).  Resolutions have now gotten high enough that manufacturers tend to omit the AA filter instead of using expensive designs.
      Even Fuji's relatively exotic X-Trans is just a custom color filter on a standard sensor - that takes MORE doing than just playing with the AA filter, since you couldn't "X-Trans" a sensor at home, while daring experimenters DO remove or replace AA filters themselves. Older Fuji designs did use modifications to the actual sensor, including some with hexagonal pixels or multiple pixel sizes on a single sensor. Even those used standard fabrication techniques, which the multi-layer Foveon does NOT.

[rdonson]

Image Sensors World

[Dan Wells]

On Sony selling the 42mp beast, the D810 is old enough that it predates the 42 MP sensor (and not yet old enough that Nikon would have logically replaced it). The Pentax K1, however, is significantly newer than the 42 MP sensor, and I would have thought Pentax would have used it if Sony was willing to sell it (its seems to be a mostly drop-in replacement for the 36 MP chip). Of course, the K1 is cheap for a full-frame camera, and Sony may charge a premium for the 42 MP chip...

How separate are Sony Semi and Sony Imaging? I'd think they'd try to get sensors out in their own cameras first? The latest 24 MP APS-C sensor was slated for the a6300 first, but that camera was badly delayed and the X-Pro 2 slipped out first... It looks like the Pentax K-70 got that chip, too (after Fuji and Sony) - it has the high-ISO performance characteristic of the new chip...

[shadowblade]

Foveon is a great technology that's, unfortunately, held by the wrong company. It completely eliminates some types of colour artifacts, improves colour resolution and has the potential to deliver even better high-ISO performance than current Bayer sensors, since it can make use of all the light hitting the sensor, rather than filtering out two-thirds of it before it even hits the photosites. That the current Foveon sensors deliver poor high-ISO output is a function of Sigma's design (much as how the early Exmors were poor at high ISO), not an inherent weakness of the Foveon concept. It would benefit greatly from Sony's BSI technology, which would put all the circuitry behind the light-collecting surface and drastically increase the amount of light reaching photosensitive areas.

The good thing is that both Canon and Sony have been working on multilayer sensors similar to Foveon, although they haven't commercialised them yet. So it may not be too long before we start seeing them in non-Sigma cameras. Probably in small sensors designed for phones first (a safer commercial experiment) but eventually trickling up to full-frame sensors.

Incidentally, Sigma's ownership of Foveon almost screams out for it to be made into a MF mirrorless camera. Sigma is never going to compete in the FF and APS-C camera stakes - they're way behind and, unlike Sony in 2009, don't have the resources of a big company to let them catch up. But MF is a much less crowded field, particularly in the $5k-$10k range (knowing Sigma, possibly even cheaper). Most potential MF users are much more concerned with base-ISO image quality and high-ISO performance, which suits the current Foveon design. And Sigma also makes lenses with superb optics - this would let them overcome the lens deficiency that plagues many other MF systems. Their AF hasn't been the best in the past (who knows what the new 500 f/4 Sport will be like), but this matters far less for a MF camera than for an action-oriented full-frame or crop body/lens.

[ErikKaffehr]

Hi,

The other side is that the Foveon is very expensive to make, so making it in large sizes may be prohibitively expensive.

It is quite true that it collects all light as it does not have filters. The downside is that it does separate colours due to different diffusion depths. So the designers have very little control of the spectral response curves. So it takes more extreme processing to separate colours than with the Bayer design where developers have complete control of the CGA.

Best regards
Erik

Foveon is a great technology that's, unfortunately, held by the wrong company. It completely eliminates some types of colour artifacts, improves colour resolution and has the potential to deliver even better high-ISO performance than current Bayer sensors, since it can make use of all the light hitting the sensor, rather than filtering out two-thirds of it before it even hits the photosites. That the current Foveon sensors deliver poor high-ISO output is a function of Sigma's design (much as how the early Exmors were poor at high ISO), not an inherent weakness of the Foveon concept. It would benefit greatly from Sony's BSI technology, which would put all the circuitry behind the light-collecting surface and drastically increase the amount of light reaching photosensitive areas.

The good thing is that both Canon and Sony have been working on multilayer sensors similar to Foveon, although they haven't commercialised them yet. So it may not be too long before we start seeing them in non-Sigma cameras. Probably in small sensors designed for phones first (a safer commercial experiment) but eventually trickling up to full-frame sensors.

Incidentally, Sigma's ownership of Foveon almost screams out for it to be made into a MF mirrorless camera. Sigma is never going to compete in the FF and APS-C camera stakes - they're way behind and, unlike Sony in 2009, don't have the resources of a big company to let them catch up. But MF is a much less crowded field, particularly in the $5k-$10k range (knowing Sigma, possibly even cheaper). Most potential MF users are much more concerned with base-ISO image quality and high-ISO performance, which suits the current Foveon design. And Sigma also makes lenses with superb optics - this would let them overcome the lens deficiency that plagues many other MF systems. Their AF hasn't been the best in the past (who knows what the new 500 f/4 Sport will be like), but this matters far less for a MF camera than for an action-oriented full-frame or crop body/lens.

[shadowblade]

Hi,

The other side is that the Foveon is very expensive to make, so making it in large sizes may be prohibitively expensive.

Is that just because Sigma makes so few of them and can't really benefit from an economy of scale?

The technology needed to manufacture a Foveon sensor is pretty similar to that needed for BSI sensors, i.e. multilayer/3D circuitry.

It is quite true that it collects all light as it does not have filters. The downside is that it does separate colours due to different diffusion depths. So the designers have very little control of the spectral response curves. So it takes more extreme processing to separate colours than with the Bayer design where developers have complete control of the CGA.

Best regards
Erik

Is that any different to a Bayer sensor, though? In both cases, the designers have to decide how much weight to give photons from each of the three colour categories to turn raw sensor data into RGB values. All the sensor does - whether Foveon or Bayer - is count photons. It has no sense of colour in its own right. It's up to the designer whether a neutral grey is defined by a 2:2:1, 2:3:2, 1:2:1 or some other RGB photon ratio. And that's the same whether you're using a Bayer array or a Foveon sensor. Sure, with a Bayer sensor, they have the option of changing the filter in front of each photosite, but all that's really doing is filtering out light and reducing the quality of the data they collect in the first place.

[rdonson]

new Google Pixel phone sensor

[eronald]

new Google Pixel phone sensor

Here is the actual article, I think. This is interesting reading for anyone interested in where sensors are going, as everything now starts with phones and trickles down to larger camera sensors.


http://www.xda-developers.com/sony-imx378-comprehensive-breakdown-of-the-google-pixels-sensor-and-its-features/

Edmund

[SZRitter]

I just thought it was worth mentioning the Foveon as it is a radically different architecture. I'm a huge fan of fixed lens cameras, so I may attempt to pick up one of the Sigmas at some point. And it is a shame it hasn't trickled out to anyone else.

I didn't follow the announcement of the Pixel, but that chip sounds interesting. Will be interesting to see if that tech trickles up. Would love it in a Micro 4/3 body.