SLT Contrast Loss...

[Greg Beetham]

Tungsten has a very strong spike at around 3000-4000K normally, it's really terrible for influencing other colours to exhibit a degree of colour change, but if you say it's like that regardless, ok, I still would like to see if it's like that under a +10000K light just for the sake of completeness, then I'd be happy that all bases were covered.
Greg

[David Kilpatrick]

Are you entirely sure that's not simply the standard reflection given at that angle by a hot mirror infra red cut filter? It looks exactly the colour I get if I tilt an OCLI hot mirror to get a reflection. The glass thickness, visible by the dual reflection, seems to indicate what you are seeing there is not the sensor, but the hot mirror/low-pass filter assembly.

David

[agorabasta]

Well, the closest to 10K I could easily get is the 'cool' white from an LCD screen. The reflection simply appears less pink and more to violet, but no big change in general. I'll try to make a good sample pic later.

And tungsten has 'spikes' in the sense that it's as smooth a black-body spectrum as it gets.

[agorabasta]

David, so far I'm not entirely completely sure of the source of that reflection. What I'm pretty sure of is that a strong purple reflection anywhere from the sensor surface to the higher layers of its toppings may and should be a source of purple fringing. Only the topmost surface reflections would not produce fringing as the reflection is not reflected back immediately.

[Greg Beetham]

I was speaking 'K' as in Kelvins, I should have been more precise, sorry. I meant tungsten has a spike (from memory) around 600nm in the colour spectrum.
Greg

[David Kilpatrick]

Agorabasta - take a magnifying glass, almost any one will do but a plastic one will be most representative of microlenses. Turn it slowly between normal viewing, and side-on. Observe the colour visible in the lens as you turn it to a more acute angle. This holds an important clue as to why purple fringes are stronger with some sensors, and why they tend to affect the outer region, and affect faster or wider angle lens more.

The Foveon sensor has a completely different structure, but it too suffers from purple fringes as well as rather strong green-magenta CA.

David

[agorabasta]

Greg, you were absolutely right putting it as 10000K. I should have put it as 10kK to be precise, but chose to put it as 10K (like 'ten thousand') as long as we talk Kelvin temps anyway

[agorabasta]

David, the chromatic dispersion in microlenses does happen, and it sure may add to purple fringing. But the plastic used to make those lenses has rather controllable properties, and I'd bet they try to optimise it as much as possible.
Also, the purple fringing is quite often much wider than 1 pixel. That actually contradicts the microslenses primary role in creating PF - too much dispersion to expect from even the poorest plastic...

On the other hand, the same effect the microlenses offer for the near-normal cone in directing that light towards the active area of the sensel, turns to direct more oblique/acute angle incident light towards the metal wiring around the sensel active area. That parasite reflection is then reflected back towards the sensor from the microlenses' outer surface. So the microlenses definitely make the angle dependence of PF more strong. If we had a well collimated beam of white light hitting the sensor, the sensor output would be changing from near perfect white spot to a strong purple haloed one in a very abrupt manner as the angle of incidence is making the microlenses to project it onto the photocells wiring.

P.S. And why do you mention the Foveon? It's not a BIS, AFAIK it's got the wiring on the top of the chip anyway. Still, IIRC, some earlier Foveons had no microlenses. My point was the BIS (Back Illuminated Sensors) should have no PF regardless of microlenses presense.