This thread is about the art and science of looking at stains. It's about a sickness which I heard is hard to cure
Of course, this is about the stains which have recently been reported for the K-5. More and more reports are arriving.
I do not participate in posting examples and I don't know if my K-5 suffers from the problem as I didn't shoot at f/22 yet. And I refrain from commenting if the effect is of any significance and if other cameras shiw the same or not.
No, this thread simply is about an examination of what people are seeing. I try to relate an image effect to a possible physical cause.
Chapter 1. Measurements
As announced, I evaluate telfish's sample shots taken with the DFA 100/2.8 Macro. The originals have been posted here:
https://www.pentaxforums.com/forums/1291441-post54.html and
https://www.pentaxforums.com/forums/1291464-post56.html
The images copied here for easier reference:
- f/16:
- f/22:
- f/32:
- f/48 (effective, nominal: f/32 at 303mm for f=100mm)
All images (c) telfish
I evaluated the upper left stain. For higher precision, one could evaluate all stains but I stopped myself there
Chapter 2: Theoretical Stainology
I start from the most simple hypothesis I can make: that the stains are caused by opaque disk-shaped obstructions somewhere in the light path.
A disk casts a shadow onto the sensor. The shadow will be disk like and be divided into an outer (half) shadow (the
penumbra) with diameter
H and an inner full shadow (
umbra) with diameter
C. Let the obstruction diameter be
D and the distance of the sensor surface to the obstruction be
d.
Furthermore, let's define the following symbols: the effective numeric f-stop:
N. The light's wavelength
lambda=0.55µm and the pixel size
s=4.81µm. Finally, let
delta be the maximum phase shift of the photon wave field around the circumference of the obstruction, for any possible photon entering the lens.
Then, it holds true:
D = (H + C) / 2 (only if C >= 0)
d = (H - C) * N/2
delta / lambda = D / lambda / (2N)
Intensity I = (D / H)^2 ~ 1/4 N^2 (D/d)^2
Chapter 3: Measurement Results
Numbering the images as above, I obtain (for the upper leftmost stain):
- N = 16, 22, 32, 48
- H = 16, 15, 14, 12 pixels
- C = 0, 2, 4, 7 pixels
- D = ?, 41, 43, 46 µm
- d = ?, 690, 770, 570 µm
- delta/lambda = ?, 1.69, 1.23, 0.87
Chapter 4: Discussion
All results are compatible with the opaque disk-shape obstruction hypothesis. The results do roughly behave as anticipated, the average over the f/22 and f/32 result is probably the best predictor.
At f/48, the delta/lambda becomes smaller than 1. This means that the illumination (from the lens aperture) starts to behave like
coherent light (over an area as large as the obstruction). Coherent light is produced by lasers but obviously, lasers aren't the only method to produce it
Coherent light must be treated taking its wave nature into account. One consequence is the formation of what is known as:
Arago Spot or Poisson Spot
This spot is a bright spot in the middle of a dark area otherwise unreachable by classic ray optics. It is caused by diffraction.
We clearly see that the Arago spot only forms for f/48 or delta/lambda<1. Look at the last image and see the bright spot in the centers. The fact that this bright spot forms exactly as predicted by our opaque disk-shape obstruction hypothesis is a strong supportive argument. There is no evidence that the stain is from a liquid or oil particle.
Hystorically, the Poisson or Arago spot was among the first strong evidences that light cannot be particles (only).
Everything combined, this means that the upper left stain is created by a round opaque particle of a bit less than 1/20 mm diameter and sitting between maybe 0.68 mm and 0.78 mm above the sensor. So, this determines the vertical particle position with about 1/20 mm precision which is about its diameter. Other stains seem to come from somewhat smaller particles.
Chapter 5: Implication for the Pentax K-5
The attached image shows a cut through the sensor and the optical assembly on top of it. It is a magnification (4x in z-direction) of my cut model photography I made myself during Photokina 2010.
The red lines are identified surfaces and the green lines mark the area where I identified the stain to originate from (with upper and lower margins).
There seems to be a clear conclusion: the stains sit on the top (
outside) surface of the cover glass of the ceramic chip package housing the CMOS sensor and received from Sony.
Conclusion:On some K-5, there sit a few dust-sized particles of 30 µm - 50 µm size on top of the cover glass of the sealed CMOS chip.
There is a smaller but finite equal chance that the stains sit within the CMOS chip on the inner side (underside) of the cover glass, or on the bottom of the AA filter, resp.
Smaller particles probably would remain invisible. Larger particles are uncommon for dust. Even 40 µm is large. Typically, dust particles are between 1 µm and 20 µm. It may actually be particles other than dust.
I am wondering how the aggregation of dust on top of the sensor cover glass can be prevented at all. The cut model shows a spacer between chip and AA filter. So, the space in between may be sealed or not. Difficult to say.
In any case, to eliminate the effect, the AA filter must be removed and the cover glass be cleaned in a clean room environment.
And to prevent the problem from emerging, Pentax must seal the AA filter to the CMOS chip and do so in a clean room. They may have actually done so. And we see the effect of an incomplete cleaning of incoming Sony sensors.
There must be more cameras showing the same effect. ...
Enjoy the read
Last edited by falconeye; 12-07-2010 at 07:32 AM.