Should Leica have gone Fovean?

[pico]

Advertisement (gone after registration)

Another strange thought (forgive, it's an ancestral trait)

 

Might a Fovean type of sensor be better able to cope with IR? There could be a forth plane of receptors to sense light in the 700nm and greater area. Or is the IR just too low a frequency to accomodate properly in a Fovean-type sensor? Or is four depths of sensor sites just too much?

 

If such a chip were possible, then it would offer more data to accept or reject in postprocessing the DNG format. One might choose pixel-grid specific areas to adjust or ignore, based upon the hits to the IR pixels - which is not the same as adjusting Purple because Purple can be legitimate on the Bayer grid. IR would be "known" in this case.

 

(It would have to be Purple, eh? Artists know that in subtractive color, Purple can be mixed using entirely different so-called Primary colors. But I digress...)

[Vivek Iyer]

Strange thoughts indeed, Pico.

 

Here are the problems with those thoughts:

 

1. Foveon does not make large chips similar to the one used in an M8.

 

2. The Foveon chip used in cameras such as SD-9 and the like can record

UV and IR when the low pass filter is removed (it is a removble filter in those cameras, BTW).

There is no indication that UV is registered only on the blue sensor and IR on the red sensor.

 

As long as Silicon is used as the sensor material, the sensor will be receptive to the spectral regions that crystalline Silicon is. There is nothing one can do other than cover the sensor with suitable filters to capture only the visible light (for normal photography).

[aj37]

Might a Fovean type of sensor be better able to cope with IR?

 

For all the hype about Foveon's sensor technology (heavily promoted by its parent company) I doubt that it would have had any specific advantage in IR, and might have introduced some additional disadvantages.

 

After all, the problem isn't really "coping" with IR -- it's striking a balance in the tradeoff between a really effective IR filter (which, apparently, would have had to be thick enough to risk color fringing) and a really pure image (which requires a thinner, less effective IR filter to let the lens' imaging qualities shine through.)

 

Leica initally chose to lean toward the side of the purer image -- which, if they had been surveyed beforehand, is probably exactly what its users would have chosen. Now that they're seeing the consequences, though, many evidently are open to a more practical, less idealistic approach.

 

This kind of engineering tradeoff is essential no matter what type sensor is being used, and undoubtedly a Foveon sensor would have required its own set of compromises (which might have been more difficult to determine, since there's less real-world experience available -- so far there's exactly one DSLR maker using this type of sensor, and their sales aren't exactly driving the competition from the marketplace.)

 

One drawback of the Foveon technology that's seldom discussed amid the hype is that since photons have to penetrate the sensor to different depths in order to be registered as different colors, the signals from the lower layers have to be amplified more than those from the top layers. It's the same situation as in multi-layer color films, in which the lower layers need to have a higher ISO speed than the upper layers, to allow for the fact that the filtration between the layers isn't 100% efficient. In the case of the Foveon sensor, the penalty for this is limited sensitivity: you can only turn up the overall gain so far before noise from the bottom sensor layers becomes unacceptable.

 

Many people already have expressed mild disappointment that the M8's high-ISO noise isn't quite as good as the Canon EOS 5D's (which has the advantage of on-chip noise suppression in its CMOS sensor) so using an inherently noisier sensor probably would have made them even more disappointed.