[MJKoski]

P.S. Is there any chance you are getting mother-of-pearl-clouds which tend to produce strange "iridescent" CYMK colors and occur Scandinavia and other places. They might be especially visible in reflections in water due to polarization effects.

I have seen those two times in two years and it is so beautiful sight that I would only use single exposures to capture them. Certainly, those were not present at the time. We have an active Finnish site for sky observation and no one reported them at the time.

[jbinpg]

I have seen those two times in two years and it is so beautiful sight that I would only use single exposures to capture them. Certainly, those were not present at the time. We have an active Finnish site for sky observation and no one reported them at the time.

The official English name for these mother of pearl clouds is Polar Stratospheric Clouds. They have become more widely reported in recent years which is worrying since they have been implicated in ozone depletion. But your photos have tropospheric clouds so that is not what we see here.

If you have access to a super computer, you could try using pixel shift to remove any chance that debayering issues are causing the magenta cast. Perhaps you are hitting an onboard processing limit. Guaranteed that Ricoh did not envision this corner case in the lab.

[photoptimist]

You might also try ticking the "Save Process" checkbox in the Interval Composite parameter screen. That will save all the intermediate composite frames which might show you when the magenta cast appears or how it progresses over time. (It has the side advantage than if the battery dies, you'll have the composite up to the last frame before shutdown.)

[AstroDave]

I am struck by the color banding, particularly in the water.

What does the histogram look like for each exposure? In R, G, B?

As noted elsewhere above, it’s not clear what the averaging arithmetic is. If the camera is taking a simple average, by adding all the frames and then dividing by N, you need more than 9 extra bits of headroom in the processor (600 > 2^9 = 512) if any of the bits in the individual exposures approach full value. It may be that the processor has just a few extra bits, so things are better with fewer exposures (i.e. no overflow ever occurs).

If certain colors overflow in the averaging process, at certain brightness levels as the lighting changes, that could lead to the color banding in your image.

[MJKoski]

Okay some update! I think AstroDave is on to something. I just finished one 256x multiexposure in .PEF format with no issues. Then I increased the amount to 512 and bam! Instant magenta cast. Exposure times were 1/60 for the 256x and 1/25 for the 512x.

But it does not strickly explain why that other 150x has the same cast. It had much longer single exposures of 2 second though.

So is it a game of two factors? Amount & Exposure time or what?

[Racer X 69]

So is it a game of two factors? Amount & Exposure time or what?

I'm thinking you need to experiment some more.

[Gimbal]

I believe the averaging is done one frame at the time so no need for extra large numbers or processing power.

[Schraubstock]

Hi MJKoski

Just tone down the Magenta a bit and work some on the other colours and submit to the "Exclusive Gallery"

I shall not be surprised if it is accepted.

[Class A]

So is it a game of two factors? Amount & Exposure time or what?

If the underlying cause is an overflow error then the factors are

1. amount of images, and
2. exposure

with the latter ("exposure") comprising scene brightness, f-stop, shutter speed, ISO settting.

In other words, the higher the amount of images, the more likely it is that pixel values add up to numbers that don't fit the available bit depth anymore, but you can counteract that by stopping down, increasing the shutter speed, reducing ISO, using an ND filter, or shooting in darker conditions.

Unless the scene brightness changes significantly, you schould be able to compensate for more images in a run by reducing exposure accordingly. For example, if you double the number of images, you should halve the shutter speed or stop down by a stop.

The histogram of a single test shot should give you an estimate as to whether a long run should be successful. Try to record the rightmost non-zero values of the histogram (preferably looking at the RGB version and paying attention to the red channel in particular) for a shot that worked with a long run. If your single test exposure does not exceed that exposure, a long run should be successful.

[MJKoski]

Just tone down the Magenta a bit and work some on the other colours and submit to the "Exclusive Gallery"

Thank you for suggestion but I think that for now I will let this be. It is against my principles. An error somewhere caused grey clouds to appear in "sunlit" magenta colors.

[Racer X 69]

To me it looks like the image sensor is able to resolve wavelengths of light that we cannot, until the image has been composited and output by the software.

And even if it is something else, you still have a very dramatic image that is well composed. Celebrate that.

[MJKoski]

That is true, camera easily sees things we cannot. But this is unheard of Be it as it may, I decided to capture series of these for a collection of at least five.

[todd]

Not that this is going to help you but just food for thought... Happened to notice something in Rawtherapee while processing a pixel shift image that seems related.. There is an option to equalize the brightness of the four frames.. I was about to flag the setting and got a popup with the following advice


So makes me think something to do with the way the averaging/equalizing is done that has limitations related to changes in brightness (as others have already said better than I am able)...

Heckflosse is the RT developer who put pixel shift into RT (and is a member here on PF...) Would be interesting to get his input.. He talks about the equalizing function in the pixel shift thread on the RT board, (towards the bottom of the thread) but it's over my head..

[derekkite]

Typically magenta cast is caused by the processing engine not handling out of range values. A straight calculation will give you these results, so most software does an extrapolation of some sort to insert a valid result.

You may have bumped up against the limits of the processing engine.

[PentaMiCL]

Hi MJKoski, thank you for your input in the other threads. I wanted to experience this issue first hand so I took the plunge and spent a few thousand shutter actuations trying to get some useful data. I don't own a K-1, only a K-70, so the differences may be due to the different camera models.

I lit a white-ish wall with an LED light to create a surface with varying brightness and shot it using "Average" multiple exposure mode. I did a single shot, 2 exposures, 4, 8, 16, 32, 128, and 500 exposures. I wanted to see if the magenta casting is somewhat gradual or whether it only happens after a hitting a certain threshold. The results can be seen below.



The amount of magenta I see in the pictures slightly vary from picture to picture, but I did not see a consistent pattern of increasing magenta as the number of shots increases, nor a significant jump between the single exposure and the 500. I exaggerated the effect below, but my conclusion did not change.



A mistake, however, was made prior to this experiment. I first tried a single shot and compared it with a 128 shot multi-exposure, and there it was, the magenta casting. However, then, I realized that I had just shot the 128 shot multi-exposure using the "comparative brightness" mode. Here there is definitely an increase in magenta.



So, I must ask you, is there any chance that you were using the "comparative brightness" mode instead of averaging when you got the heavily magenta-casted shots? I only ask because I myself actually made that mistake when I conducted the experiment, as described above. In addition, the linear streaks and colorfully saturated highlights (clouds, reflections) in your sample shot is not that dissimilar from sample shots of Olympus's Live-Composite mode (e.g., https://www.creativeislandphoto.com/uploads/1/0/5/1/10515211/clouds-pb-small_orig.jpg), which is a similar type of exposure to comparative brightness multi-exposure. One more thing that gave me the idea of this possibility is that the picture of chunks of snow moving around in a body of water in your thread (Long exposure vs Average Interval Composite - PentaxForums.com) is described as an example of a 512 frame Averaging shot, but it seems to be a "comparative brightness" shot -- the chunks of snow at the time of each exposure should not be so opaque if they were averaged with 511 other exposures in which the chunk is no longer there.

For easier reference, I will embed the single shot Average, 500-exposure Average, single shot Bright, 128-exposure Bright samples below, all developed to exaggerate the effect and cropped. The last two images are further cropped in "single shot Average" vs "500-exposure Average", to demonstrate how Averaging does seem to give you advantages in shadow noise, even at extreme settings such as 500 shots. (In, K-70, at least).

single shot Average


500-exposure Average


single shot Bright


128-exposure Bright




single shot (above) vs 500 shot (below) averaging. Noise is smoothed out in the shot below.