Questions about Brain and when and when not to use it?

[timh]

Hi Robin,

I am using the Brain and sensor analysis along with guiding and dithering in order to image (AS1294 Pro plus F5.6 f= 617 mm refractor). The problem is that while I may think that I understand the various parameters I may well in fact have my wires crossed. So I would be very grateful if I could check my understanding with you in order to address the likely origin of issues with banding and background pattern noise that I sometimes get. Apologies for the longish detailed note but others may experience the same and any light that you (or others) could cast would be very welcome. So here goes with the list of what I think I understand - and what may be going wrong?

1) The Brain recommends an 'optimum' gain and exposure value. 'Optimum' means that the Brain has set these two values about as low as they both can be (maximising the dynamic range of the data in the stack of 14bit ADUs to nearly 16 bits - according to the Y values of the Brain graphs) while at the same time also still sufficiently high that the signal detection limit (e/ pixel value) still remains within 5 or 10% of the asymptotic limit obtainable at longer exposures.

2) The Brain has two settings. Choose values to maximise dynamic range or unity gain (ca 1 electron per ADU unit). Is the latter especially for e.g. Photometry use? I choose the former - because it is desirable to, as far as possible, not bloat out most of the stars even whilst the primary objective may be to observe faint H II nebulae - although I know that some vistas will anyway require multiple exposures and high dynamic range processing.

3) Recently, setting up under rather bright skies - moonlight and passing thin high cloud in an urban environment - the sky reads about 5 e/ pixel/ s or so. Under the best conditions and with the UHC filter in place it can get down to 0.5 to 0.7e.

4) Thus with the UHC filter in place the Brain recommends e,g. a gain of about 120 (near unity) and exposure of about 50s - and without it about 5s.

5) The Brain recommends an offset of 4. I am guessing that it simply sets this value in order to be as close to the left hand axis as possible (to preserve ADU space) while remaining a couple of standard deviations a way from zero so that there are no negative values?

Recently (last night in fact) I ran with the UHC filter and conditions very close to Brain recommendations (offset 10, exposure 46s and gain 124 - simply because I already happened to have matching flats and darks for these particular values) and accumulated for about 50 min versus the Wizard Nebula and the Heart Nebula.

Given that the sky conditions were really quite poor the results were perhaps not too bad. BUT one problem that became all too evident was that there was quite a lot of background pattern noise and, in particular, banding relative to the signal.

I only have anecdotal data (i.e no systematic experiments) but I have previously only seen such a high ratio of pattern noise to signal on other occasions when I have reduced the gain down to 120 or so. At gain 285 - my previous 'norm' it wasn't such an obvious problem - or at least it was at a much less objectionable level.

A number of possible explanations occurred to me?

a) Not dithering properly (about every 3 frames and I was assuming that pixel unit dithering distance unit in Sharpcap refers to guide rather than main camera pixels?)?

b) I thought that the sky conditions were about OK. BUT may well have been misleading myself by just looking at the FWHM and brightness scores during live stacking? i.e. I was quite happy to see FWHM at about 2.4 (translating to FWHM arcsec about 3.8?) - but could low FWHM values just signify very low signal intensity? I thought that brightness was OK - but I guess that this filter isn't absolute - it is merely set rewlatively by the first frame?

c) So it possible that under certain sky conditions (e.g thin high cloud that you might not always see) the signal to sky brightness ratio drops off so far that there is then really very little signal relative to the general measurement of sky brightness? So then more gain is needed to get a signal that rises well above the background pattern noise

Basically I suspect therefore that I might be trying to make a silk purse out of a sow's ear? i.e. unrealistically ambitious for good photographs under conditions that are really quite hopeless.

The fact that there have been occasions when I have used lower gains without such a background problem also points to the sky being the main variable?

So overall does this analysis seem reasonable ? In summary. Brain will set parameters to the best values to get the best dynamic range and faint feature selection BUT -caveat emptor - not if the conditions are really so bad -thin cloud afflicted etc - that the ratio of signal to sky brightness becomes too variable or low?

best wishes
Tim

[admin]

Hi,

I think that in theory the brain's recommendations should be valid regardless of the sky conditions, but the point that you raise is still very important – there are some things that the brain doesn't know about and therefore doesn't take into account in its calculations. For instance in some circumstances the recommendations might be for quite short exposures but that may not give bright enough stars for good stacking. Patton noise -like banding is another case – assuming the amount banding remains constant as the exposure changes then it's advantageous to take longer exposures because the ratio of the signal to the banding will be stronger if you take a small number of long exposures than taking a large number of short ones.

I think the rules of thumb would be

* If you think you need to take shorter exposures than recommended by the brain then increase the gain (because that would usually let you get away with shorter exposures – see the graphs).
* If you think you need to take longer exposures than recommended then nothing is really stopping you except the tracking capabilities of your mount and any saturation problems you might run into.

Another thing to look out for his to see if there are any changes you can make to the camera settings to minimise the landing noise – I have seen models of camera where using the fan at half speed gives dreadful banding and others where adjusting the USB parameters can help reduce the banding.

I wonder if it would be interesting for the central analysis to measure uniformity of longer dark frames – if the brain had some idea of the amount of banding that was likely to be present then it could take it into account.

Cheers, Robin

[timh]

Thanks Robin,

I looked at stretches of the darks and it revealed what looks to be the source of problem. There is significant banding at low gain(e.g. 122) that is much ameliorated at high (285) - which ties in with my general experience of using this camera. Exposure didn't make any difference to the banding in the darks - same at 95s as at 7.5s - just much more amp glow in the former than the latter as you would expect.

So now I wonder whether this is a general problem with these cameras, all CMOS cameras with the same chip or just peculiar to mine?

But otherwise ---as you suggest ---- I can use longer exposures to get more signal to try and submerge the patterning --or increase the gain which I can see works but then at the cost of limiting the dynamic range.

Also interesting suggestions about the USB connection etc. I will look into that. Maybe see if changing the cable length etc has any effect?

If this does turn out to be a more general problem than just with my camera then perhaps it is worth seeing if the Brain could incorporate it in some way? I have not yet played with the banding suppression settings. A quick search and it seems that the problem is widespread and across a number of cameras? cf

https://www.cloudynights.com/topic/6636 ... ing-issue/


A quick postscript. The banding appears to be nothing to do with cable length , power supplies (same with batteries), whether or not the fan is connected or not.

Fixes that appear most promising - based purely on examination of master darks - collected under various conditions - appear to be..

1) Use a compromise gain of about 150 - this is noticeably better than 120 - and still pretty good in terms of image depth according to Brain?
2) There seems to be a slight benefit in setting the USB speed down to 40

but thirdly and most dramatic in its effect is your banding suppression tool. It nicely suppressed banding in darks (e.e. at gain 67) when it was set to 50% suppression and at thresholds between 20 and 40% (I didn't try any lower)

Is there any liklihood of unwanted effects on the lights from using banding suppression or any reason why I shouldn't from here in just set up to now do all darks, lights and flats with banding suppression set at 50% and threshold at 35%? I am guessing that the highest threshold is the least aggressive. If this is a common problem with CMOS camera then maybe the Sharpcap banding suppression tool provides a general answer?


thanks once again for your comments and responding to all of this stuff - the advice to examine the darks seems to have been a good call


Tim

[admin]

Hi,

I think a lot of evidence is pointing towards banding being influenced by power supply noise issues – some cameras seem to control this better than others and some settings seem to help reduce it.

When setting up the banding suppression settings, you should set both settings to the lowest values that give you good results. The main banding suppression control is adjusting the strength of the smoothing applied to remove the horizontal bands. The threshold controls how much of the brightness range the suppression is applied to – basically from 0% up to the threshold value has banding suppression applied and anything brighter than the threshold value is not touched.

I don't think there should be any bad effects if you apply the banding suppression to both the darks and the light frames. There's no need to apply it to flat frames, but also probably no harm in doing so.

Cheers, Robin

[timh]

Hi Robin,

Just to continue this thread so as to help anyone that might read it because they also see banding at low gain with an ASI294 Pro camera

Firstly, many thanks for your advice once again.

I have now run quite a few experiments with the camera and also contacted FLO the supplier who suggested that I see whether banding at low to unity gain was equally as evident when I captured frames using ASIstudio. Perhaps not surprisingly the results from using ASIstudio - at gains 0, 120 and 390 - were indistinguishable from Sharpcap except for a slightly different colour cast. Probably not surprising since both programmes were using the same native ZWO driver?

Overall I observed that ....

1) Banding varies with gain and is always evident at gain zero and is always much ameliorated at gain 390.

2) Banding increases significantly when the cooling is switched on (don’t know whether it is the fan or temperature per se)

3) The effect of cooling on banding is especially obvious at intermediate unity gain 120 - I.e banding is not that obvious at ambient temperature but becomes quite apparent when the camera is cooled (e.g. to – 5C which is the sort of temperature that I typically use)

Banding appeared not to be an effect of power supplies/ mains modulation etc. Both the laptop and cooling were run off battery power. No additional USB cabling was used other than the short USB3 cable that came with the camera. The laptop was connected via a USB3 socket.

So - at the moment - the SC banding suppression tool looks to be the most practical way forward (unless it does turn out that I have an atypically bad camera that ZWO will agree to replace). A setting of 50% suppression at a threshold of 10% appears effective. I haven't tried lower values yet.

best wishes
Tim

[donstim]

I use a gain of 121 with my ASI 294MC Pro camera and have never seen this banding issue.

Here's an example dark at gain 121: https://photos.app.goo.gl/dcqwnNptrufDwfKK7

Don

[timh]

Many thanks Don,

Your post makes me think that I might have a problem with my camera - although went I went to your link I couldn't see the dark stretched.

Here are are my PixInsight stretched masterdarks. Two at gain 0 and two at gain 120. Two at -5C and two at ambient temperature. The only one that doesn't show banding is the masterdark at gain 120 and at ambient temperature.

https://onedrive.live.com/?id=330584F8B ... F8B152FA32

(assuming my link works)
best wishes
Tim

[admin]

Hi,

the link to cooling is quite possible – most of these cooled cameras run the Peltier using pulse width modulation (PWM) to control the power of the cooling system – since there is quite a large current going into the Peltier cooler, switching it on and off rapidly for PWM is likely to cause ripple on pretty much every other power supply voltage in the system unless the design of the electronics is really really good.

Cheers, Robin

[donstim]

Hi Tim,

Yeah, I didn't stretch it since I didn't know how much of a stretch you were talking about. I figured you could download it and stretch it however much you want.

Uploading this file made me realize that all my darks are .png files rather than .fts files. Does this make a difference if I am only using them within SharpCap?

And now that I am actually examining my dark frames, I don't see any hot pixels at all. So, there is really nothing to stretch except the black background. Makes my wonder if my darks were done correctly.

Don

Edit: Okay, I've taken some sample new darks to match yours (well almost - used a gain of 121 instead of 120) - gains of 0 and 121 and temps of ambient and -5C. Here is the link (hope it works); there is the unstretched fits file and a stretched tif file of each - https://1drv.ms/u/s!Ajsy7Ey4_8uFxWnUIaD ... u?e=CiutLK. I used FITS liberator to stretch the files, and the only option for saving them was as .tif files, which changes how they look, so you may want to work with the unstretched fits files and stretch them yourself.

Would be interested in your comments as to what you see in these. It looks to me that like yours, the only one without any banding at all is the gain 121 at ambient temperature.

I would also like to know why the .png files look radically different and whether they worked at all as darks in SharpCap livestacking. I'm obviously still learning. Maybe that's why I didn't see any banding previously (although I did encounter it last night - maybe I just didn't notice it before - but it seemed to lessen as the stacking progressed).

[Menno555]

The horizontal banding of the 294MC Pro is known and expected. I had some mail about this with Zwo support, send them some example and they say it is normal.
On the Zwo site at https://astronomy-imaging-camera.com/pr ... -pro-color there is a set of Dark files for download to see what you can expect. All this under Dark Current.
To counter this, is to use more darks. If I use for example 20 darks and make a MasterDark in DSS, then the banding is still there.
But if I use 40 or 50 darks, the banding is mostly gone.
Hope this helps a bit.

Menno