Does this image show cosmic ray strikes?

[OzDave]

Hi,

Attach is a crop of a stack I took on Saturday with my QSI 683. The stack is a 7x10 minute exposure of M51 guided with PHD. This particular crop came from the Red stack. The subs used for stacking have been dark subtracted and flat-fielded and the stack has been stretched to reveal the artifacts that I am seeking answers on.

The other stacks (L, G, and are pretty similar. They contain stars that look reasonably ok, but also contain these odd trails all over the place.

I was reading about cosmic ray strikes in a book about imaging recently. I wondered if that's what these could be? They seem quite uniform in shape and all in the same direction. I might have thought it was star trails except that none of the actual stars are trailed, plus the guiding never showed more than around half a pixel deviation all night.

I've never done subs 10 minutes long before and this is only the third outing my QSI 683 has had since I bought it. Usually I am going for 3-5 minute subs having come from DSLR imaging. I've not seen trails like this on other imaging runs I've done, but maybe I'll go back to the last couple of sessions and see if there is anything similar.

I don't think they occur always in the same place either because the RGB combo image contains the trails in different colours.

Does anyone know what these trails could be?

David

[iwatkins]

It looks like the same sensor pixel on each sub, so would suggest hot pixels that your darks haven't picked up?

A cosmic ray hit on a single sub yes, but on the same sensor pixels on every sub? I don't think so.

Double check your darks and dark subtraction.

Cheers

Ian

[rfdesigner]

Hot pixels.

If your darks weren't taken at exaclty the same CCD temperature as your subs then the hot pixels will be different values.

You can get round this with a sigma clip stacking algorithum, or by using better darks or by using a hot pixel map.

Derek

[OzDave]

Hmmm, I just checked the master dark by moving the white point way over to the left. There are definitely no trails in that. Hot pixels yes, but no trails. Then I opened a Red sub (pre calibration) and did the same - no trails again that I can see under a heavy screen stretch. Then I went through each of the 7 red calibrated subs I have, and again I don't see a lot of evidence of trails, but maybe that is because the images need stacking to bring it out. However, looking at the brighter trails in the stacked image, I would have expected to see some evidence in the calibrated subs to explain why the trails are appearing.

To me it seems like this issue might be the result of stacking. I used Nebulosity 2.5.x on a Mac. I've not seen this before. I think I'll try stacking again and see what happens.

I'm not 100% sure I'm looking for the trails in the un(calibrated) subs properly. I mean I don't see any, but that doesn't prove they aren't there. It's a data point towards that conclusion, but I'd have to be sure my method of checking was right.

Should I post some other crops of subs to help analyse?

Anyone got other other theories?

[OzDave]

Hot pixels.

If your darks weren't taken at exaclty the same CCD temperature as your subs then the hot pixels will be different values.

You can get round this with a sigma clip stacking algorithum, or by using better darks or by using a hot pixel map.

Derek

My camera has temperature controlled cooling. I was running it at -20C when I took the darks and the lights. So I don't think it is a hot pixel issue. The reply I just posted earlier has some more info about analysis of the subs.

Basically I'm still confused where these trails are coming from. What should I post/look at next to try to figure this out?

David

[jgs001]

Looks like hot pixels to me... you'd see them as single point hot pixels in each individual subs, not trails, the stacking creates it. I don't know nebulosity, but the sigma stacking will remove them... not sure seven subs is enough, try it and see..

Looks like it's called standard deviation stacking in nebulosity

[Tom Janssen]

Try median instead of average at image integration.

That might resolve the problem.

Best regards,

Tom

[OzDave]

Looks like hot pixels to me... you'd see them as single point hot pixels in each individual subs, not trails, the stacking creates it. I don't know nebulosity, but the sigma stacking will remove them... not sure seven subs is enough, try it and see..

Looks like it's called standard deviation stacking in nebulosity

Yeah, I I think hot pixels is right after some further analysis. I was looking for complete trails in the calibrated subs, and of course there aren't any.

I just tried stacking two of the subs and the result showed "trails" in the places where the main stack had them, but only with two pixels in the trails. Then I stacked 3 subs, and got 3 pixel trails....

So it seems they are hot pixels getting shifted slightly as part of image alignment during stacking.

Now the question is why my darks didn't take care of them. I stacked 15 darks using Nebulosity's stddev (1.75) stacking mode. I don't usually use this but thought I'd try it out. It seems to "fix" pixels during the stacking so I'm thinking maybe it eliminated some hot pixels that lay outside 1.75 standard deviations from the mean, or something like that.

I think I'll restack my darks using average only and recalibrate the subs to see if that makes a difference.

David

[Tom Janssen]

Try median instead of average at image integration.

That might resolve the problem.

Best regards,

Tom

[OzDave]

Ok, so restacking the darks and calibrating again and stacking didn't help. The trails are still there. So I picked a couple of prominent trails in the images and went in search of the actual pixels that created them.

Pixel A has the following values:

Raw sub - 10259

Master dark - 4346

Dark subtracted sub - 5913

Pixel B has the following:

Raw sub - 5389

Master dark - 2236

Dark subtracted sub - 3153

The average background pixel value in the dark subtracted sub is around 680, so consequently these improperly handled pixels stand out and make a trail as the subs are shifted during alignment for stacking.

So, now I'm wonderign why the master dark is so out of line compared to the actual raw subs. Why would the raw sub have a pixel value of 10259 and the master dark only 4346? The darks and lights were definitely taken with the camera at -20C at least according to the camera temperature monitor of the capture software. However, I took the darks inside in the afternoon before it got dark. The heatsink temperature was around 37C if I remember well and the cooler power was at 87%. During the imaging run it was cooler outside and the cooler power was maybe in the mid-sixties, with heatsink just under 30C. The CCD was -20C in both cases.

The individual dark frames (I didn't check them all, but checked some) seem to have pixel values around the values quoted above for master dark. So the master dark average seems correct. The dark subtraction also seems to be correct according to the values, it's just that the pixels in question seem to be hotter in the raw subs than they were in the dark and so the pixel values are not adequately handled.

Any ideas?

David

[OzDave]

This might have to be put down to one of those unexplainable things, but I decided to redo the darks. I did a set in Nebulosity and another set in a different capture utility. The new darks contain more reasonable values for the pixels that I manually inspected in my previous post.

<pause>

I've now recalibrated my subs with the new darks and the trails have vanished!

I don't understand why my original darks were not particularly accurate. I did them using the same settings as now. I do not think I made a mistake in this as I know they took 10 minutes each to expose and I was checking the CCD temperature while it was all happening.

So unless the pixels in question are prone to random variation in hotness levels, or their characteristics somehow changed suddenly, I don't have any other explanation.

Anyway, it appears my problem is now solved. Thanks to all responders! The diagnosis was right from the start except that I didn't see how that could be until I went digging for evidence. Live and learn as they say...

Cheers

David

[iwatkins]

Glad you got it sorted David

I don't have set point cooling on any of my cameras, but I learnt the lesson a long time ago: darks follow lights, always. Or better still, interspersed.

Just because your CCD reports the same temp, it doesn't mean the rest of the device is acting the same as when you captured the subs

Cheers

Ian

[martin_h]

If you load one of your original darks into Nebulosity and goto Edit-Image Info it will tell you the exposure time and the temp.(it does for mine)

[Jessun]

David, you can also do a bad pixel map in Nebulosity, and it you link it up with DSS dithering is also something to be recommended - did you try that?

I guess you'll find that even with the hot pixels gone in the image above, some signs of the same type trailing will show up pretty much everywhere. DSLR's are worse for this though.

Dithering and even the odd knock in ra/dec between subs randomizes things in your favour. At least that's what I find.

[OzDave]

I found that the new darks cleared up the trails completely. No sign of any hot pixel artefacts at all, which is great.

I did find that doing a stddev stack of my darks in Nebulosity produced a much "brighter" dark than doing a simple average stack. So much so that I thought subtracting the stddev master dark would cause more damage than it would solve. So I went for the simple average and my calibrated images look just fine now.

My next issue is trying to process my data in photoshop. This is where the real learning curve is at the moment for me. Colour balance in particular for LRGB.

David

[swag72]

I found something silimar to this when I stacked in DSS - Must redo my darks then. glad this was raised - Thanks

[rfdesigner]

Glad to see you're on top of it.

I've tried using sigma clipping, averaging and median for darks and I found I got slightly less resulting sub noise with a very very slight sigma clip (sigma of 4) median worked reasonably well but I really didn't like the resulant histogram in my dark corrected subs.. didn't get to the bottom of that but it just wasn't right. Averaging didn't work due to cosmic ray hits during my darks.

I do everything in IRIS, mainly because I learnt it 12 years ago and haven't felt a need to move on. It is also very powerful and allows you to keep an excellent grip on what is actually happening, unlike some other applications which seem to try and hide everything which leads to confusion and errors. (just my opinion)

Derek