Panasonic Lumix G 20mm F1.7 II (H-H020A) - Banding

Started Aug 6, 2013 | Discussions thread
Anders W Forum Pro • Posts: 21,468
Re: Panasonic Lumix G 20mm F1.7 II (H-H020A) - Banding

tsi wrote:

Anders W wrote:

Yes, the banding can result not only in varying brightness but also varying color. Have you considered why? I have some ideas but since you may already have thought more or better about this, I let you shoot first.

Just a guess:

1.)

The banding is the same for all colors, but the color conversion misinterprets it as a color difference.

2.)

The banding noise is very high frequent and modulates different colors of the bayer pattern differently.

What is your opinion?

I don't think either of these two possibilities is very likely. In all likelihood, the sensor is read on a per-row basis with certain sets of rows differently affected by the interference from the lens than others. In the absence of any pattern noise (banding), the read-noise component of the signal should be purely random with a mean of zero (provided that we are sufficiently far from the black point where negative read errors are clipped so that the mean is no longer zero). The interference from the lens in all likelihood means means that the read-noise deviates from the zero mean on a per-row basis. Some rows systematically tend to have negative read errors on average and others positive (or some rows stay with an average of zero whereas others have positive or negative averages).

Now read errors are typically thought to be independent of the true signal. I would guess (without claiming to know) that this is true about the additional read errors caused by the interference too. In other words, for some rows, all the signals tend to be biased upwards or downwards by a certain absolute amount, i.e., a certain absolute number of electrons or ADU levels. This will cause color shifts because a change of any given number of ADU levels upwards or downwards means more if the true signal is weak than if it is strong. In order for color to remain unaffected by the pattern noise, the noise component added by the interference would have to be proportional to the true signal and I think it unlikely that such is the case.

Note that the above implies that we are likely to see color shifts together with the banding even on a white area in ordinary daylight. That the area is white doesn't mean that the RAW-level signals for red, green and blue are equally strong. The green signal is in this scenario far stronger than the other two on the RAW level (about twice as strong on the E-M5 and its siblings).

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +28 more
Post (hide subjects) Posted by
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
tsi
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow