0

What are the advantages and disadvantages of Foveon X3 sensor compared to sensors that use Bayer filters or other color filter arrays?

Using resolution, noise, and light gathering performance.

mattdm
  • 143,140
  • 52
  • 417
  • 741
jsldfjksd
  • 39
  • 2

1 Answers1

3

The X3 sensor records three colour values per photosite instead of one. This reduces ambiguity in rendering fine detail compared to Bayer or other one-colour-per-photosite sensors, reducing the need for an anti aliasing filter (which blurs the image so that light falls across more than one photosite). This results in great per pixel sharpness and very natural rendering of fine colour details and lack of colour moire, which is the advantage of the X3.

Unfortunately there are many disadvantages:

  • Resolution is on average slightly lower than Bayer sensors with the same total number of detectors. Why is this? Well imagine you have a 15MP X3 sensor and a 45MP Bayer sensor. Now if you're photographing a gig and there is only a red light pointing at the stage, then the green and blue photosites of both sensors will not record any signal as there is no green or blue light coming into the camera. So you'll get a 15MP image from the X3 as you'd expect, but the Bayer image will be interpolated from only 11.25 red pixels. Advantage Foveon. Now reverse the situation and imagine shooting an image of a grey cat against a white background. Now the incoming quantities of red, green and blue light are equal everywhere in the image. The X3 sensor produces a 15MP image as before, as there are only 15 million difference positions light is recorded across the image plane. The Bay sensor produces a true 45MP image as it records light from 45 million different positions across the image plane, and loses no information. These are extreme cases, but in practice it works out that a Foveon sensor is equivalent to a Bayer sensor with twice the number of pixel locations, i.e. a 15MP (45MP if you read Sigmas marketing) X3 ~= 30MP Bayer.

  • Limited adoption means fewer tools to process RAW images. Simga's own RAW converter has been widely criticised for ease of use issues. Also people may already have their own workflow for Bayer images.

  • The X3 is currently only available in APS-C size or smaller, 15MP max resolution so far. Bayer sensors have been commercially in medium format and larger with over 100 megapixels.

  • Development is not keeping pace with Bayer sensors which are growing larger all the time. Eventually Bayer sensors will be high enough resolution to use oversampling to get round the one colour per pixel problem, meaning eventually Bayer technology will win out due to it's simplicity (in my opinion).

  • Sensitivity is not as good, blue light has to penetrate both the red and green pixels to register. This results in very noisy images in anything other than very good light.

  • Colour reproduction is poor for certain colours for the same reasons.

Matt Grum
  • 118,892
  • 5
  • 274
  • 436
  • Maybe merge this into your older answer on the earlier question? – mattdm Apr 27 '13 at 16:56
  • Bayer sensors have topped out over the last few years, coming in either 16MP forms (which have significantly less resolution than modern Foveon cameras) and also higher resolution cameras like the D800e which is a bit below the new Foveon Quattro sensor in terms of detail captured. The D810e coming out soon does not have a higher resolution... as for color reproduction being poor, that is false. At higher ISO noise can be higher but at low ISO color reproduction in Fovoen cameras is excellent, to the point where they make excellent cameras for reference capture of paintings and fabrics. – Kendall Helmstetter Gelner Jul 08 '14 at 01:21
  • Also, you claim that the resolution is worse "on average". But the example at the far end you point out is a totally monochromatic subject, a grey cat on a white background. In reality most pictures people take will have color elements so "on average" the detail captured would in fact be much closer to the higher end of detail captured by bayer than the lower end. To claim it's a problem that the Foveon chip delivers constant resolution regardless of color seems much less of a problem than buying a "30 MP" camera that really gives you between 7.5 and 30MP depending on scene color. – Kendall Helmstetter Gelner Jul 08 '14 at 01:28
  • @KendallHelmstetterGelner You mention Bayer sensors "topping out" in resolution in the same sentence as the Foveon Quattro - the design of which was necessary to fight a fundamental limitation in the X3 design, in that light is defocused as it penetrates deeper into the sensor! A perfectly focused image will appear out of focus if you move the sensor backwards even slightly. The same effect happens inside the X3, which is why the Quattro reduced the resolution of the lower layers. The Bayer sensor is only limited by manufactures' unwillingness to push resolution due to consumer misconceptions. – Matt Grum Jul 08 '14 at 11:53
  • @KendallHelmstetterGelner With 45MP total, in the worst case (e.g. only red light) you get 11.25MP out of the Bayer and 15MP out of the X3. In the best case (e.g. a monochrome scene) you still get 15MP out of the X3 but get 45MP out of the Bayer! Yes most scenes have colour elements. A yellow object contains red & green light, when the brightness of yellow changes, red & green change by the same ammount, and demosaicing can fill in the missing red values using green of which there are 22.5MP! So it's really a case of fixed 15MP, or between 11.25MP and 45MP, averaging somewhere in the middle. – Matt Grum Jul 08 '14 at 12:49
  • @MattGrum - the comment on de-focusing is totally non-sensical. The X3F focuses exactly the same as any other sensor, the layer depths are not enough that focus is altered by the measurement depth. The Bayer sensor has depth too you know, as it uses similar wells (which have depth) to capture photons. – Kendall Helmstetter Gelner Jul 08 '14 at 19:54
  • @MattGrum - Yes in the best case for Bayer you get 45MP, but again that is a scene with only exactly neutral tones present (black to white and grey in-between) which is almost no real world shooting. The "Yellow object contains red & green light" is totally incorrect, there is a specific wavelength for yellow. And in the case where you are taking green and red sensors together - they are not measuring the same spatial location so that does not "add" to the accuracy. It's two 11MP images added together which is still 11 MP or real data. – Kendall Helmstetter Gelner Jul 08 '14 at 19:57
  • @KendallHelmstetterGelner I meant a yellow object in an image contains red and green. Also you missed the point entirely, red and green don't measure the same spatial location, but when the red and green response increase together then you can use the green values to reliably estimate what the missing red values are and vice versa. So you get 33.75MP of red and green data, and you only need to "guess" 1 in every 4 pixels where there is a blue sensor which doesn't record useful information for yellow objects. That's a lot more than the 15MP of red and green you get from the X3. – Matt Grum Jul 09 '14 at 14:43
  • @KendallHelmstetterGelner this isn't a discussion forum, so that's all for now I'm going to update my answer when I get a chance and provide sources for everything, I'm not making all this up - I very nearly bought a DP2M, and did a lot of research. – Matt Grum Jul 09 '14 at 14:45
  • I own a DP2M, and have a lot of experience. I'm just trying to make sure people have all the correct information. A lot of your points also do not take into account the Quattro sensor. – Kendall Helmstetter Gelner Jul 10 '14 at 03:17