Bright spots in infrared photos (Gimp processed)


(ron) #21

@Jossie : Thanks for the pic. I didn’t look at them closely enough. What it tells me is that I did a bad job of overexposing the frame for Morgan’s test. You’re seeing the outline of the lightbulb, which clouds the picture. So, I’ll have to redo the shots, and shoot at a whiteboard in the sun or something like that.

And good evening to you! It’s 1:00 pm here on the east coast.

Nevertheless, if there is a blue signal I would expect that there must have been blue photons to produce that.

It seems that would be sensible. Could it mean that the blue photons are powerful enough in one location (high enough quantity) to penetrate the filter, for some reason? Perhaps I should do some tests on the filter.


(Hermann-Josef) #22

@Ronaldlees

Perhaps I should do some tests on the filter.

This is something I was going to suggest anyhow. Do you have a transmission curve for the filter, i.e. transmission as a function of wavelength? No filter is perfect so there might well be some leakage.

Hermann-Josef


(ron) #23

@Jossie ,

You know (well, I guess you don’t) - I tend to be cheap when I shouldn’t be. I have a filter that is supposed to be equal to this one:

http://www.hoyaoptics.com/pdf/R72.pdf

But …

I haven’t verified that the filter I have is as good as the Hoya Optics one. So, I should test it. I think I have a green laser pen around here. If I can’t find it, I’ll order a green, red, and blue one, and test with those. May take a few days for the red and blue one to get here.

  • Ron

(Hermann-Josef) #24

@Ronaldlees

according to this, the filter should be pretty dead at optical wavelengths. You could try a shot at bright sunlight outside and see what you get – although the test with the laser pen will give you also wavelength information.

Hermann-Josef


(Morgan Hardwood) #25

Thanks @Ronaldlees, they were not ideal but helped improve things. See https://github.com/Beep6581/RawTherapee/issues/5173


(ron) #26

@Morgan_Hardwood : Great! Sounds like an incremental help. What should I do to create the overexposed frames with improved quality?

  • Ron

(Morgan Hardwood) #27

@Ronaldlees well it’s as described in the “Adding Support for New Raw Formats”. You don’t need to shoot a whiteboard or anything fancy, just make sure every part of the frame is completely blown out, e.g. by pointing the camera at the sky, de-focusing everything and significantly overexposing (the histogram in the camera lies because it’s based on the embedded JPEG and not on the real raw file, so by overexposing by 2EV you work around that problem). But as this camera is from 2007 and your specific one is modified, I’m happy to leave things as they are and not spend more time on this :wink:


(ron) #28

@Morgan_Hardwood : OK, sounds good. I’m thinking about upgrading to a new model of the Sigma (maybe Quattro H) - so perhaps we could revisit this later.

BTW: None of the Sigmas need to be modified. The internal filter for IR is removable, and it’s easy peasy to put it back into the camera. This is much better than dedicating a DSLR to IR only service IMO.

  • Ron

(Hermann-Josef) #29

@Morgan_Hardwood

Hello Morgan,

now I am completely puzzled: What do you learn from completely overexposed images, where all the pixels have the same value?

Hermann-Josef


(Morgan Hardwood) #30

@Jossie you find the “white level” by examining at what point a pixel corresponding to a photosite clips. Clipping depends not only on the physical photosite (and new sensors can have various types of photosites of various sensitivity) but also on the electronics which follow it, and may change depending on e.g aperture. You can read more about it here:
https://raw.githubusercontent.com/Beep6581/RawTherapee/dev/rtengine/camconst.json


(ron) #31

One thing I’ll say about the SD14 is that it does not seem to show me any noticeable artifacts when I’m shooting normal visible-light photos. In fact, it seems to do quite well, and in some respects is even better than my Pentax K5, and in some other respects maybe not - but the difference in age between the cameras is pretty large, so it’s not exactly a fair fight. I think the Foveon does great with visible light photography - and the newer Foveon based cameras are probably even better.

Edit -> Just took a shot with ISO-200 and with the main lights turned on, and it had artifacts visible. Then took a shot at ISO-100, and the spots/artifacts were gone! So, at least for IR work, it looks like you need to stay at ISO-100 on this camera. I’ll take some more shots just to be sure, but it seems that at ISO-200 you could be getting into some sensor noise. I had figured 200 would be OK (usually is, right?).

So, sensor noise is not detected with the lens cap on? Maybe it takes adjacent photosites to energize the noise? I think I just answered my own question - but what a round-about way to do it! Thanks go to the respondents that helped.


#32

Random but unlikely guess: Could it have something to do with how the light is penetrating the sensor? Like X-Trans rectangles and related topics?


(ron) #33

Hi Afre,

Random but unlikely guess: Could it have something to do with how the light is penetrating the sensor? Like X-Trans rectangles and related topics?

I’m not sure I understand the X-Trans explanation, so I don’t know if that could relate to this thread. I do think I’ve been chasing a ghost or my tail, just because I didn’t take the admonitions given by others in other forums to run the camera at ISO 100.

I mean, the thing takes such darned good pictures at ISO 100. So, for my “dark room” tests, I thought I’d push it up (but only a little, to ISO-200). I didn’t really believe the admonitions about ISO, I guess. But, I believe them now. The Foveon loves the low end of the ISO scale. When it’s down there, it does great in this camera. Lesson learned. The fact that it has ISO-50 I suppose is a hint.

As mentioned a couple messages prior to this one - the spots/artifacts are gone when I use ISO-100, with near-darkness or with the lights turned on - either one.


(Hermann-Josef) #34

@Ronaldlees

Good morning Ron,
as far as I understand cameras with a CCD detector, changing the ISO setting just changes the conversion factor from CCD counts to the output signal. This is way higher ISO produces more noisy pictures. In CCDs there is no amplification e.g. as in image tubes. If this also holds for the detector in your camera, I do not know.

I would be interested to see one of you ISO-100 (without lens cover) shots to compare to the example above.

Hermann-Josef


(ron) #35

Guten Tag Jossie,

This is my take on the CMOS sensor design:

With the CMOS designed sensors there is a built-in amplifier under each photosite. The properties of this amplifier might be different (amplification increased or decreased), along with the integration/sampling rate. With amplification, there is always a noise-power relationship. Increased power typically hurts the signal-to-noise ratio. So, perhaps the result of a photon hit is more likely to be read out by the sensor/amplifier setup when its values are tweaked by higher ISO, which comes with a penalty of poorer noise figure, and/or the integration/sampling rate could be changed (I’m not sure how the latter item plays out). It’s been a long while since I looked at any of this kind of stuff, so don’t write it down.

As far as the difference between CMOS and CCD sensors, I found this:

https://www.cse.wustl.edu/~jain/cse567-11/ftp/imgsens/index.html

Now I have to move into the realm of conjecture instead of just bad memory. Maybe the Foveon needs higher amplification and/or different sampling schemes from the start, due to its layering structure versus the matrix of Bayer. I don’t know any Sigma people, so this is total conjecture.

Here is another article that speaks to the Foveon sensor, but I haven’t decoded it yet, except to say that maybe it implies that the simpleton explanation for the ISO differences given in my previous couple paragraphs may fall short :

The latter link is an eye-waterer.

Sigma is thought to have much improved noise figures in the newer Quattro line. Another article I read implies that the noise is not really higher on the Foveon, but the Bayer scheme “averages it out” by interpolation. Is the better noise performance of the later Sigma models partially the result of averaging the stuff out in firmware? Or is it physical sensor change that improves the picture, or both? This may explain why Sigma people want you to use their software for post processing. The noise may not really exist.

IMO - The Foveon color is unbeatable, but maybe it needs more post processing to bring it out.


(ron) #36

Enough already! The length of my post is maybe eye-watering. Sorry.


(Hermann-Josef) #37

@Ronaldlees
thanks for these links. This is the stuff I was involved in with our CCD observations, especially if we commissioned a new camera / detector.

But I did not find anything about various ISO settings.

If you are talking about amplification, this seems to be the same in CCD and in CMOS detectors: It is the conversion factor between number of detected photons and the digital number in the output. This is not an amplification but just a multiplication, because you cannot change the number of photons detected. This is what I meant above. However, I never used a CMOS detector. They did not find their way into astronomy. IR-detectors seem to be related at least in the sense that you can address individual pixels for read-out.

Still I would be interested to take a look at an ISO-100 image without the spots (lens not covered) for comparison (if possible as the raw file).

Hermann-Josef


(ron) #38

@Jossie : Can any of the detectors detect a single photon? I’m not an expert in this realm - I’m a long-time electronics hobbyist and a (very recent) photography hobbyist/novice. Anyway, I’ll take a no-spot raw photo and post it ASAP …

  • Ron

#39

Sorry, you have the wrong Ron.


(Hermann-Josef) #40

@ron sorry for this!

@Ronaldlees

Can any of the detectors detect a single photon?

I am not sure what quantum mechanics says about detection of a single photon. In both, CCD and CMOS, charge is built up by illumination with photons. The lower the read-out noise is the lower photon flux can be detected (all well above 1 photon :slight_smile: ). In the analog-to-digital-converter (ADC) the charge is converted to a digital number. And here enters what is called gain: you can specify, how many detected photons correspond to 1 count (some call it digital unit). We called this factor EPC for electrons-per-count. I think, this is what the ISO setting in CCD-cameras does: For low levels of illumination you do not want to get low numbers. Instead you decrease the conversion factor EPC and end up with decent numbers, i.e. the image appears brighter. But the noise is also multiplied.

George Rieke (2003, Detection of light) only briefly mentions CMOS detectors. He talks about “providing gain in the amplifier for each pixel”. I do not understand enough about electronics to understand if it is the ADC he is talking about.

In reading I now learned, why CMOS-detectors did not make it into astronomy: Due to the complex circuitry required for each pixel, the area-filling-factor (i.e. the fraction of light sensitive area) is much less for CMOS than for CCDs. With the low light levels in astronomy, it is the photon collecting area where the emphasize is.

Hermann-Josef