Master Dark subtracts too much noise?

[wetcoaster]

I am using an ASI1600MM Pro. I live on the edge of a red zone known as Vancouver, As a consequence I use 6nm narrowband filters for Hα, Oiii and Sii. I shoot 5 min subs at gain = 200. Cooled temp is -20C. Earlier this week I captured the Bubble in Hα using those settings. I processed them in Pixinsight using a master dark created indoors. but at sensor temp - 20C. When I calibrated the subs using that master dark, aligned and stacked them, the integrated image showed the outer edges lighter than the core. That suggests to me that the master dark subtracted too much "noise".

My question is - should I take those subs in a cooler ambient temperature? Or is there another explanation?

[geordie85]

Did you use flats and dark flats in your calibration? Usually if the center is darker than the outer edges it's down to flats over correcting. 

Try taking flats with a lower ADU. 

[wimvb]

Since your camera is cooled to a set point, ambient temperature shouldn't matter, as long as the cooler can keep up. That said, personnaly I would shoot darks in a cool a place as possible.

There may also be other things happening;

1 how do you take darks? E.g. if you leave the camera on the scope, there's the risk of light leaks. That's why I take darks at night, with my camera covered with a lens cap and two layers of aluminium foil.

2 do you use "optimized" dark calibration? Cmos cameras such as zwo cameras, need matching darks that are used without scaling.

3 is the light edge possible residual amp glow. In that case the lighter areas should be conentrated in just a few places.

It would help to diagnose the problem if you post an example.

[Thalestris24]

Also maybe check your darks/master darks are what you expect. If not, check for light leaks.

Louse

[Laurin Dave]

Max cooling on the ASI1600 is 40-45c so it'll be at or near its limit indoors to get to -20c.. 

Dave

[symmetal]

My ASI1600MM with 5m subs shows noticeable 'amp glow' in 3 three of the corners, mainly top and bottom right, and some on the bottom left. If all four corners of your image are brighter than the centre then as geordie85 says it's the flats that are over correcting. If it's three corners that are brighter it implies your darks didn't correct enough. The darks are used on the 1600 primarily to combat 'amp glow' rather than improve noise. I did a check with mine with and without darks, and the addition of the darks made no noticeable difference to the noise but did correct the 'amp glow' on the three bright corners very nicely.

Alan

[wetcoaster]

First off, thanks for the comments. On to the answers.  I don't use flats because the ones I captured with electro luminessence  panel gave me terrible over-correction. I use Sharpcap which gives gives a readout of ADUs. I found suggestions from the web to use 28k ADUs, low 40s in % of histogram and  half of Full Well. The first two gave roughly the same outcomes but the tjhird option was different. ASI1600's Full Well is 20ke. That meant if  20ke = 20k ADUs, I should use 10k ADUs.  That gave totally different results but no solution.The most common was the outer edges of the image were much brighter than the centre

For reasons I do not understand, my Oiii captures have what looks similar to an Infrared hotspot (for those of you who shoot terrestrial IR like I do, i.e., central bright spot). The stars are always brighter with Oiii than for Hα. Focus problem?

The amp glow is in the upper left, lower left, and lower right corners. That is when the camera is oriented with the USB port at the 12 o'clock position. 300 seconds, gain 200.

I do not use "optimized" darks in PIS. I don't knowingly use scaling. I am trying to use darks captured in exactly the same environment as the lights. However, in most cases I have captured them with the camera only but the master dark I will provide was captured with the lens covered by a faux black leather bag. I also used black electrical tape to cover the opening of my ZWO manual filter wheel.

Regarding max temperature cooling range, I have run into the limits of the cooling range once in a while and was limited to -18.5°C. Last night it was 7°C outdoors so there was no problem achieving -20°C. I will post that dark on DropBox and provide a link here. I'll also upload the integrated image, unstretched,  6nm Astronomik Oiii filter, used with a Nikkor Ai 600/5.6 ED lens at f/5.6. It's a stack of 22 subs, 300s, gain 200. Sky was clear but light pollution limited the SQM reading to 18.9. Good seeing and transparency according to the Clear Sky Chart.

Narrowband is an adventure (translation - numerous problems for a new guy to overcome).

[wetcoaster]

Here are the links.

<https://www.dropbox.com/s/hxapx9k20u0dok6/00_18_21 Master Dark 300s g200.fits?dl=0>

<https://www.dropbox.com/s/91d35hdkc3ipr89/Bubble O3 600 integration.xisf?dl=0>

Don

[symmetal]

Hi Don,

If your flats did show a brighter edge than the centre that would explain why your images show this without flat calibration but why they over-corrected I can't say. Maybe someone using PixInsight to calibrate can help. I use Astroart for calibration and stacking and have no problems with over/under correction on flats or darks. The actual ADU level of the flat isn't that important as long as the histogram of the flat doesn't show any clipping at black or white and is on a linear section of the response curve of the camera. Modern cameras are pretty linear throughout their response as long as your not clipping. It's the difference in ADU value compared to the average ADU value that is used for flat calibration and not the absolute value. I usually just aim for a third to half of full well on the histogram.

The only narrowband imaging I do is in Ha but I find that flats are particularly necessary in Ha as the brightness variation can be quite considerable and is not a uniform fall off towards the corners like my LRGB flats are. The ZWO Ha I have posted is the original ZWO narrowband filter which can have bad light leakage around the edge giving a bright edge in places. The Baader one (from another scope and camera) is better but still isn't a uniform fall off. I replaced the ZWO with another Baader for the 1600 but haven't done any flat with it yet to compare.

The dark you posted looks very similar to mine with the corner 'amp glow' so I'd say is OK. Unfortunately I can't open your .xisf file to check it as I don't have Pixinsight.

ZWO Ha flat

Baader Ha flat

Alan

[wetcoaster]

Thanks Alan. Interesting stuff. I noticed the XISF format. Here is a link to a 32bit FITS file.

<https://www.dropbox.com/s/tok0tgm8bjtuxhg/Bubble O3 600 integration FITS.fit?dl=0>

When you say "in the linear portion", that is the 20-80% range according to something I've read.  I don't know if SharpCap's histogram for its Capture Flats function is showing the Full Well range. It does give a readout in ADUs. 28K is roughly 40% of the histogram but that was too bright.

Your flats are real interesting. I had one that simply blew me away. My Nikkor Ai 600/5.6 ED lens can go to f/32. During my manhandling of it the f-stop ring wound up at f/32. I joked with my frineds that I now knew how to find out if there was dust in a lens! It looked like something blasted by shotgun with fine pellets..... It also showed something very close to the sensor and a quick look confirmed that was so.

I had a better capture of the Bubble last night with Oiii. It appears my star bloating could be caused by over exposure! I cut the gain to 150 from 200 and ran 5 minute exposures. The subs looked a lot better. I intend to process them shortly.

 

Don

[symmetal]

Got your image OK Don.

There's no brightening towards the edges just the usual vignetting. If your flats show a bright edge it may be light leakage getting in from your light panel somehow so your getting extra light not via your filter.

The ASI1600 gives a 16 bit value output although it's only 12 bit resolution (the lower 4 bits are set to zero) so the histogram display in Sharpcap will be of that 16 bit value (the ADU). The full well or clipped output should display as 65535  ADU (FFFF in Hex). Your 28k is approximately 40% of 65535. When you say that was too bright what do you mean. 20-80% of the range is where the response would safely be linear, though it's probably still linear outside that range. Can you post one of your flats.

I noticed your image displays the image brightness as floating point values between 0 and 1. This must be Pixinsight's default save mode. I imagine you can select 16 or 32 bit integers if you wish.

Also your stars all have black holes inside them. Is this a stacking artifact? I've seen this when fits files in unsigned integer format are read as signed integers or vice versa without reading the header. (I'm looking at you RegiStar ). Also in this case the background appears as mid grey. However even your dim stars show this black centre so it's not a case of mis-reading the data. 

Alan

[symmetal]

The full well to ADU relationship needs a bit more clarification as it is dependant on the camera gain. As far as any external program (like Sharpcap) is concerned it reads a 16 bit ADU value from the camera and the value of 65535 is treated as peak white (threshold of clipping) hence the histogram is always shown as values between 0 and 65535.

For the ASI1600 (12 bit camera) the sensor output after analogue to digital conversion has a value between 0 and 4095, where 4095 corresponds to peak white. 4095 x 16 (to convert the 12 bit value to 16 bits) is 65520, which is the 12 bit resolution equivalent of 65535 (when expressed as a 16 bit value).

The ASI1600 has a full well of 20000 e- (electrons) which is greater than 4095 so the full sensor brightness resolution can't be expressed with a 12 bit sample. At unity gain (139) one e- corresponds to 1 ADU so a well value of 4095 is treated as effectively a full well and is output as a 16 bit value of 65520.

At twice unity gain (200) one e- corresponds to 2 ADU so a well value of 2047 is treated as a full well and is output as 65520. (Note digital resolution is now only 11 bits and not 12)

At gain 0 one e- corresponds to 0.2 ADU so a well value of 20475 is treated as a full well and is output as 65520. This is very close to the actual full well capacity of 20000 e-. But you're effectively throwing away 4 out of 5 electrons from the well, though will give the highest dynamic range so less clipping of bright stars.

I hope that makes the relationship between the ASI1600 well capacity and ADU a bit clearer. 

Alan

[wetcoaster]

Alan - Your detailed replies are much appreciated. I need to wrap my mind around the electron to ADU explanation. I'll post a flat when I have some time.

As for the floating point output, it is the default. I'm not clever enough to know the meaningful differences between the various options. Perhaps a signed integer might be better.

I'm focusing on tonight's run. I am enjoying a wonderful streak of good weather while battling some issues like why my Polemaster Polar Alignments are sub-standard. Last night I had 3 attempts that gave different locations for the NCP. In the past I used to single PAs but they seemed not to be accurate. So I went to a minimum of 2 and often 3. The NCP was sbtantially different. Not only that, using a different star for the mount's rotation gave a result that suggested the run failed.

I have to set up my mount each night because of where I live. Therefore a good PA is essential.

 

Don

[symmetal]

Don, make the most of the good weather. 

Don't worry too much about the electrons to ADU relationship. Just use the ADU value as that's all the programs outside the camera can see. I only mentioned it as in your earlier post you wondered if your flats histogram peak should be lower as the camera had a full well of 20k electrons.

The camera ADU output is an unsigned integer, while the default fits format is signed integer, but it doesn't really matter what the fits image number format is, as the fits header should give that information.

Good luck with your polar alignments. 

Alan

[wetcoaster]

Alan - I did my first LRGB combination using Hα and Oiii. It was of IC1396 (includes Elephant's Trunk). Very good detail but pink colour. I assigned the Oiii to G & B and Hα to L & R. All had the same weight. I can play with that to get a nicer looking image but it has another flaw. I had adapter problems and wound up switching them to get past a back focus issue. But the camera wound up being 90° different in orientation than the Hα orientation. Aligning them created a cross with only the overlapping areas properly represented.

I reread your post about full well vs ADUs and finally wrapped my mind around it. Very helpful.

Don