Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

AA filter, spatial filter and star colours

  • Please log in to reply
127 replies to this topic

#1 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 02 October 2018 - 07:21 PM

In my continuing review and analysis of the Canon 200D and Nikon D5300 I thought I would look at the effect of the antialiasing filter.  The Canon has one but the Nikon does not.

 

I was expecting the Nikon D5300 to show the weird star colours that Alan Dyer demonstrates on both the Sony A7III and the Nikon D750 - see the bottom section "Debayering Star Artifacts" here: https://amazingsky.n...trophotography/

 

I gave up waiting for a cloud-free night and used an artificial star instead - a white LED torch covered in baking foil with a tiny hole made by a pin.  I moved its position randomly between multiple exposures:

 

NikonSpatialFiltering.jpg

 

The 50mm lens I was using was not sharp enough to produce the pretty pink, purple, green and yellow stars.  However, something else was noticeable.  The Canon produced "stars" with more or less uniform colour.  However the Nikon produced stars having a mixture of colours - some the same as the Canon and some stars that were quite green.  Why is this?

 

I'm guessing this is an effect of the Nikon spatial filtering (which seems to kick in at around 0.5sec).  I think it's probable that, just like the later version of the Sony star eater algorithm, the spatial filtering preferentially attacks red and blue pixels.

 

Further testing and analysis is required.  I'm interested to know if anyone else has noticed the Nikon D5300 producing greenish stars?  If my hypothesis is correct, it would only affect stars that are small and unsaturated.  Or maybe you've noticed the pretty pink, purple, green and yellow stars that I was unable to produce?

 

Mark


Edited by sharkmelley, 02 October 2018 - 07:41 PM.

  • entilza and VincentD like this

#2 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 03 October 2018 - 05:34 PM

After some careful thought I designed a new experiment to help understand what is going on.  The approach was to generate a field of white dots on a computer monitor and photograph them.   The results are below at 400% scaling (click on the image to see it full size):

 

CanonNikonSpatialFilteringV2.jpg

 

Looking at the Canon panel and the first Nikon panel, there is actually very little difference between them, even though the Canon 200D has an anti-aliasing filter and the Nikon D5300 does not.  I would describe the filter as quite mild and it's not at all difficult to make the camera produce Moiré effects, despite the AA filter.  The overall colour of any particular star in the images above is governed by which colour pixel the peak brightness of the star hits.  The colour variations in the Canon stars are slightly less than the Nikon stars.  I think it conclusively demonstrates that you should not trust the star colour you see in a single exposure!

 

The really big difference is between the Nikon at 1/5sec and 1/4sec.  On the Nikon D5300, any exposure of 1/4sec and longer has spatial filtering (a type of hot pixel suppression and noise reduction) applied to it - this is typical of Nikon cameras though the exact exposure length at which it cuts in, varies from model to model.  It would appear that it's this spatial filtering that leads to the large number of green stars.  The spatial filtering is nowhere near as destructive as the algorithms that Sony cameras use (the infamous "star eater" algorithms) but it does still have some effect on star colour.  It seems that a bright pixel has its value capped only if it exceeds some threshold but it's not immediately clear what that threshold is.  On the other hand, the algorithms used by Sony cameras cap the value of every pixel that is brighter than its neighbours.

 

The conclusion seems to be that the anti-aliasing filter on the Canon 200D actually has very little effect on star colour, at least for the size of stars used in this experiment.  On the other hand, the spatial filtering algorithm used by the Nikon D5300 definitely has a noticeable effect, at least for the size of stars used in this experiment.

 

Mark


Edited by sharkmelley, 03 October 2018 - 05:48 PM.

  • nofxrx likes this

#3 SandyHouTex

SandyHouTex

    Skylab

  • *****
  • Posts: 4058
  • Joined: 02 Jun 2009
  • Loc: Houston, Texas, USA

Posted 04 October 2018 - 09:56 AM

After some careful thought I designed a new experiment to help understand what is going on.  The approach was to generate a field of white dots on a computer monitor and photograph them.   The results are below at 400% scaling (click on the image to see it full size):

 

attachicon.gif CanonNikonSpatialFilteringV2.jpg

 

Looking at the Canon panel and the first Nikon panel, there is actually very little difference between them, even though the Canon 200D has an anti-aliasing filter and the Nikon D5300 does not.  I would describe the filter as quite mild and it's not at all difficult to make the camera produce Moiré effects, despite the AA filter.  The overall colour of any particular star in the images above is governed by which colour pixel the peak brightness of the star hits.  The colour variations in the Canon stars are slightly less than the Nikon stars.  I think it conclusively demonstrates that you should not trust the star colour you see in a single exposure!

 

The really big difference is between the Nikon at 1/5sec and 1/4sec.  On the Nikon D5300, any exposure of 1/4sec and longer has spatial filtering (a type of hot pixel suppression and noise reduction) applied to it - this is typical of Nikon cameras though the exact exposure length at which it cuts in, varies from model to model.  It would appear that it's this spatial filtering that leads to the large number of green stars.  The spatial filtering is nowhere near as destructive as the algorithms that Sony cameras use (the infamous "star eater" algorithms) but it does still have some effect on star colour.  It seems that a bright pixel has its value capped only if it exceeds some threshold but it's not immediately clear what that threshold is.  On the other hand, the algorithms used by Sony cameras cap the value of every pixel that is brighter than its neighbours.

 

The conclusion seems to be that the anti-aliasing filter on the Canon 200D actually has very little effect on star colour, at least for the size of stars used in this experiment.  On the other hand, the spatial filtering algorithm used by the Nikon D5300 definitely has a noticeable effect, at least for the size of stars used in this experiment.

 

Mark

Interesting stuff.  Thanks Mark.



#4 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 04 October 2018 - 05:01 PM

At this point the investigation takes a totally unexpected turn ... 

 

The Sony spatial filtering algorithms, let's call them "Star Eater" and "Star Eater v2" are well understood and documented:

http://www.markshell...ystareater.html

http://www.markshell...areater_v2.html

 

 

I thought it would be interesting to apply them to the Nikon raw data.  Here is the result:

 

NikonSonyStarEater.jpg

 

Look at the very remarkable similarity between the Nikon raw data spatially filtered with the Sony Star Eater v2 and Nikon's own spatial filtering (in the 3rd pane of the image I posted earlier - the 1/4sec ISO 100 pane).

 

The Sony Star Eater v2 is the spatial filtering algorithm used by all Mk2 and Mk3 versions of Sony's A7 series of full frame mirrorless cameras.  It appears that Sony implemented something similar (though not identical) to what Nikon was already using.

 

Mark


Edited by sharkmelley, 05 October 2018 - 12:40 AM.

  • whwang likes this

#5 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2831
  • Joined: 20 Mar 2013

Posted 04 October 2018 - 08:42 PM

D5300 is a five year old camera. Those earlier than D5300 can install the firmware hack to disable to raw cooking.  I wonder if the newer Nikon also has such a problem.  If yes, we need to warn their users.


  • sharkmelley likes this

#6 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 05 October 2018 - 12:35 AM

D5300 is a five year old camera. Those earlier than D5300 can install the firmware hack to disable to raw cooking.  I wonder if the newer Nikon also has such a problem.  If yes, we need to warn their users.

That's an interesting point.  Yes it was possible to disable this spatial filtering (a.k.a. hot pixel suppression or HPS) on earlier Nikons. 

 

Over on DPReview I found this description of reverse engineering the HPS algorithm for the Nikon D800 and D4, performed by Bernard Delley:

https://www.dpreview...s/post/50875417

 

It looks pretty similar to what is happening in the D5300 raw files.  If so, then it would certainly have the effect of preferentially attacking the red and blue pixels in a small star, turning it green.

 

Out of interest I might try to determine if the D5300 algorithm is identical to the D800 and D4.  With sufficient data e.g. dark frames and star fields it is possible to perform a statistical analysis in the same way that Bernard has done.

 

Mark



#7 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2831
  • Joined: 20 Mar 2013

Posted 05 October 2018 - 01:03 AM

I have a feeling that if this happens on D810A, many people would complain. Or 810A might have a special treatment in firmware?

#8 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 05 October 2018 - 01:33 AM

I have a feeling that if this happens on D810A, many people would complain. Or 810A might have a special treatment in firmware?

To be honest I don't think many people notice their star colours at all.  Probably because typical processing workflows bleach star colour. It took ages for anyone to accept that there was a star eater problem on the Sony cameras and most people haven't noticed that the latest version of Sony's algorithm turns stars green. 

 

It would be certainly be very interesting to see how the D810A behaves, since it was designed for astro-imaging.  All we would need is a few 5 minute dark frames at room temperature.  In fact all Nikon cameras can be diagnosed in a similar way.

 

Mark


Edited by sharkmelley, 05 October 2018 - 01:47 AM.


#9 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2831
  • Joined: 20 Mar 2013

Posted 05 October 2018 - 02:39 AM

I think I can provide a bunch of 3-min darks from D810A.  Will they be useful?



#10 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 05 October 2018 - 03:07 AM

I think I can provide a bunch of 3-min darks from D810A.  Will they be useful?

Definitely!

 

They must be raw frames and a single dark plus a single light (a long exposure with plenty of stars) should be adequate for now.  They don't need to come from the same session and don't need the same ISO or exposure length. 

 

I may not be able to look at them for a couple of days because I'm away this weekend.

 

For anyone interested in how it's done, take a look at my article on diagnosing spatial filtering:

Plus Bernard Delley's detective procedure:

 

Mark


Edited by sharkmelley, 05 October 2018 - 03:15 AM.


#11 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2831
  • Joined: 20 Mar 2013

Posted 06 October 2018 - 02:40 AM

Hi Mark,

 

Here are links to a bunch of D810A darks:

https://drive.google...6ao-h2wmnVSnQg9

https://drive.google...SvZlY3By9Dzy4h_

https://drive.google...BkjCjpU3Gm9CgRL

https://drive.google...pYIjgr_ICOo761_

https://drive.google...pn_ZrRlkjOzcKkm

 

They were taken at ISO 800, and exposure time was 180 sec.  The ambient temperature is between 11 and 12 deg C.

 

Here is a light frame with many stars, taken shortly after the above darks:

https://drive.google...RfU6-HajKxjx3x-

 

The files will be there permanently. So there is no need for you to rush. I prefer seeing all the analyses from the same person, so the quality and assessment can be consistent.

 

Cheers,

Wei-Hao



#12 entilza

entilza

    Soyuz

  • *****
  • Posts: 3744
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 06 October 2018 - 08:28 AM

Very interesting work Mark!  Regarding your artificial star, how far away was the light source from the lens?  Was it indoor or outdoor.  Thanks.



#13 AtmosFearIC

AtmosFearIC

    Apollo

  • -----
  • Posts: 1003
  • Joined: 10 Dec 2015
  • Loc: Melbourne

Posted 06 October 2018 - 09:18 PM

This is really interesting. I have run both a Nikon D7200 and D810 through SGP and I have always assumed that the reason I needed to do SCNR on the green channel on every image was due to there being more green pixels and there being a green bias from that.

From this though it may be that the red and blue pixels are being reduced, I do wonder if this could happen with longer exposures like the 300s ISO400 that I’m usually working with. I suppose they would.

I also wonder if software like MaximDL and SGP bypass some of the “processing” that the camera is doing.

#14 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 08 October 2018 - 12:14 AM

Hi Mark,

 

Here are links to a bunch of D810A darks:

 

The files will be there permanently. So there is no need for you to rush. I prefer seeing all the analyses from the same person, so the quality and assessment can be consistent.

 

Cheers,

Wei-Hao

Excellent - thanks!  I have downloaded them.  I'll now subject them to detailed analysis.

 

Very interesting work Mark!  Regarding your artificial star, how far away was the light source from the lens?  Was it indoor or outdoor.  Thanks.

In the first experiment, the torch covered in cooking foil with a pin-hole was 20 meters from the camera with its 50mm lens.  Both torch and camera were outside in a field.

 

If anyone wants to try out the approach of imaging a series of dots on a computer screen, my test file can be found here:

https://drive.google...HpGJDXR01drsSjI

 

Display the image at 100% on the monitor scale and adjust the distance of the camera from the screen so that the "stars" have a spacing of 10 pixels or less in your image.  Be careful not to let the "stars" saturate.

 

Mark


Edited by sharkmelley, 08 October 2018 - 12:28 AM.

  • entilza likes this

#15 SandyHouTex

SandyHouTex

    Skylab

  • *****
  • Posts: 4058
  • Joined: 02 Jun 2009
  • Loc: Houston, Texas, USA

Posted 08 October 2018 - 09:37 AM

At this point the investigation takes a totally unexpected turn ... 

 

The Sony spatial filtering algorithms, let's call them "Star Eater" and "Star Eater v2" are well understood and documented:

http://www.markshell...ystareater.html

http://www.markshell...areater_v2.html

 

 

I thought it would be interesting to apply them to the Nikon raw data.  Here is the result:

 

attachicon.gif NikonSonyStarEater.jpg

 

Look at the very remarkable similarity between the Nikon raw data spatially filtered with the Sony Star Eater v2 and Nikon's own spatial filtering (in the 3rd pane of the image I posted earlier - the 1/4sec ISO 100 pane).

 

The Sony Star Eater v2 is the spatial filtering algorithm used by all Mk2 and Mk3 versions of Sony's A7 series of full frame mirrorless cameras.  It appears that Sony implemented something similar (though not identical) to what Nikon was already using.

 

Mark

Or vice versa.



#16 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 08 October 2018 - 01:13 PM

Or vice versa.

No - not vice versa. 

 

The Nikon spatial filtering algorithm was reverse engineered back in Feb 2013 and the Nikon D5300 appeared later that year. 

Sony's algorithm was introduced 4 years later, in June 2017 (Firmware v 4.00 on the Sony A7RII).

 

Hence it appears that Sony implemented something similar (though not identical) to what Nikon was already using.

 

Mark


Edited by sharkmelley, 08 October 2018 - 01:15 PM.


#17 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 09 October 2018 - 04:01 PM

I had some clear sky tonight so I decided to take a genuine image of a real star field with the Nikon D5300, to see if my earlier experiments reflected the results of real astro-imaging.

 

Here's a single exposure of the Cygnus constellation: 3min ISO 400 with a 50mm lens at f/8:

 

NikonD5300_GreenStars1.jpg

 

All I've done is debayer, white balance (multipliers 2.12, 1.00, 1.49 from DXO's site) and background subtraction in PixInsight so this is still linear data.  Here's a small area of that image at 200% scale:

 

NikonD5300_GreenStars2.jpg

 

It's full of green stars caused by the spatial filtering.  Unfortunately this is building a case against the Nikon D5300 which is not pretty.

 

I've also had a chance to take a preliminary look at Wei-Hao's files from the Nikon D810A - the darks and the light.  It's quite clear that some kind of spatial filtering is still taking place but it's leaving the stars untouched!  I'm trying various statistical analyses to try and unlock its secret and hope to report back with evidence later this week.

 

Mark


Edited by sharkmelley, 09 October 2018 - 04:11 PM.

  • xiga and whwang like this

#18 xiga

xiga

    Viking 1

  • *****
  • Posts: 517
  • Joined: 08 Oct 2012
  • Loc: Northern Ireland

Posted 09 October 2018 - 04:36 PM

Mark I commend you for your knowledge and hard work in investigating all of this.
As a D5300 owner I will watch with interest.
  • sharkmelley likes this

#19 Jim Waters

Jim Waters

    Mercury-Atlas

  • *****
  • Posts: 2964
  • Joined: 21 Oct 2007
  • Loc: Phoenix, AZ USA

Posted 09 October 2018 - 06:02 PM

Yes Mark - Thanks for your hard work...!


  • sharkmelley likes this

#20 fmeschia

fmeschia

    Apollo

  • *****
  • Posts: 1032
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 10 October 2018 - 01:25 AM

Shouldn't spatial filtering be very recognizable by looking at the FFT of a dark frame? I tried with both Wei-Hao's D810A darks and my own D5300 darks but I can't see any recognizable pattern in the 2D FFTs. The noise seems to be white.

Francesco


Edited by fmeschia, 10 October 2018 - 01:25 AM.


#21 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 10 October 2018 - 04:01 AM

The spatial filtering on Sony cameras is severe enough to show up in an FFT. You are quite right that the Nikon D5300 and D810A is not obviously detectable, probably because the spatial filtering is less severe and fewer pixels are affected.

Mark

#22 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2831
  • Joined: 20 Mar 2013

Posted 10 October 2018 - 04:25 AM

The spatial filtering on Sony cameras is severe enough to show up in an FFT. You are quite right that the Nikon D5300 and D810A is not obviously detectable, probably because the spatial filtering is less severe and fewer pixels are affected.

 

And yet the green stars in the D5300 image are so obvious.  It's a red flag to me for all Nikon cameras.



#23 fmeschia

fmeschia

    Apollo

  • *****
  • Posts: 1032
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 10 October 2018 - 10:29 AM

And yet the green stars in the D5300 image are so obvious.  It's a red flag to me for all Nikon cameras.

I processed your light frame with bilinear debayering, but didn't see many green stars. Even fewer if I debayer it with VNG. I don't have the RAW file for the D5300 image so I can't repeat the experiment. I did it with some of my own D5300 light frames and didn't see green stars, but in my case each star was spread over several pixels (I should repeat the experiment with a 50 mm lens).

 

Another thought: given that each 2x2 Bayer cell has two green-filtered pixels, isn't it statistically expected that the number of green-filtered pixels in a given area will be twice as many as red or blue ones? In the asymptotical limit where the PSF of a star is spread over only one pixel, shouldn't the same apply to stars?

 

Francesco


Edited by fmeschia, 10 October 2018 - 10:29 AM.


#24 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 10 October 2018 - 03:24 PM

Here's a brief description of how I began to unlock the spatial filtering algorithm in the D5300 and then to compare it with the Nikon D810A.

 

Firstly you need to be familiar with the pattern of Red, Green and Blue pixels on a typical DSLR sensor:

 

StarEaterGreenPixels.JPG

 

Now if we consider the 5x5 block of pixels surrounding a Red pixel we have the following:

 

StarEaterAlgo.JPG

 

The pixels labelled V1 to V8 will all be Red.  Let's make the assumption for now that if V0 has an extreme value then its value will be replaced by the spatial filtering algorithm.  This is pretty standard stuff - it's how the original star eater algorithm worked.  We can test various possible replacement values e.g. the median of the values of the 8 neighbouring red pixels V1-V8 or the maximum of those values.  In fact it's the maximum and we can easily prove it - here's how ...

 

I took a long exposure Nikon D5300 raw dark frame.  Then for every single red pixel in the undebayered data, the function V0 - max(V1,V2,V3,V4,V5,V6,V7,V8) was calculated.  If the pixel values are truly random (i.e. unaffected by any spatial filtering algorithm) then we should obtain a nice histogram from this.  Here it is:

 

NikonD5300_Rchannel_fullpop.png

 

There is a slight spike at the zero point, where V0 - max(V1,V2,V3,V4,V5,V6,V7,V8) = 0, but we can't be certain it's statistically significant, even though we have a full population of 6 million red pixels.

 

So the next approach is to purposely select the more extreme pixels - I arbitrarily selected all the (unsaturated) red pixels whose value is more than 3 standard deviations above the mean.  There were 29500 of them.  Here's the histogram of V0 - max(V1,V2,V3,V4,V5,V6,V7,V8) for those pixels:

 

NikonD5300_Rchannel_3sigma.png

 

It's a nice distribution but with a giant spike where V0 - max(V1,V2,V3,V4,V5,V6,V7,V8) = 0.  What does this mean?  It's impossible for it to arise by chance.  Of the 29500 pixels, 5560 of them (i.e. nearly 20%) have a value which matches the maximum of the 8 neighbours.  It's the smoking gun - something (i.e. the spatial filtering) has deliberately set those values.  What we don't know for certain is exactly what criterion is used to select the offending pixels.

 

Exactly the same happens for the blue channel - nearly 20% of the 3-sigma blue pixels have a value which matches the maximum of the 8 neighbours.

 

For the green pixels the story is slightly more complicated.  Rather than the G1 and G2 channel being treated separately, the green pixels are treated all together: 

 

GreenPixels.JPG

 

Each green pixel is considered to have 12 neighbours: V1 to V12.  If the value of a green pixel is considered too extreme by the spatial filtering algorithm then it's value is overwritten by the maximum of those neighbours i.e. max(V1,V2,V3,V4,V5,V6,V7,V8,V9,V10,V11,V12)

 

So what's different about the D810A?  The basic algorithm is the same: extreme red and blue pixels have their value overwritten by the maximum of their 8 neighbours of the same colour and extreme green pixels have their value overwritten by the maximum of their 12 green neighbours.

 

Just as with the Nikon D5300, I started with a long exposure dark frame.  For the full population of 9 million red pixels, here is the histogram of V0 - max(V1,V2,V3,V4,V5,V6,V7,V8):

 

NikonD810A_Rchannel_fullpop.png

 

There is actually the tiniest spike at zero, halfway down the right-hand side of the distribution.

 

So instead, let's select the (unsaturated) red pixels whose value is more than 3 standard deviations above the mean:

 

NikonD810A_Rchannel_3sigma.png

 

Now the spike is very obvious.  There were 13810 red pixels with a value above 3-sigma and only 755 of them have a value which matches the maximum of the 8 neighbours i.e. around 5% of them.  It's a similar story in the green and blue channels.

 

So it appears that for one reason or another, far fewer D810A pixels are being affected by the spatial filtering.  The frustrating thing is that we don't yet know the criterion being used to select the offending pixels.  It's probably some kind of threshold but how is that threshold set?

 

In my next post (you must probably wait until this time tomorrow) I will demonstrate that a large number of stars in a D5300 image have their red and blue pixels "zapped" by the algorithm (turning the stars green) but it is apparently not the case in a D810A image.

 

Mark


Edited by sharkmelley, 10 October 2018 - 03:25 PM.


#25 sharkmelley

sharkmelley

    Vanguard

  • *****
  • topic starter
  • Posts: 2316
  • Joined: 19 Feb 2013

Posted 10 October 2018 - 03:30 PM

And yet the green stars in the D5300 image are so obvious.  It's a red flag to me for all Nikon cameras.

Yes, that's the really interesting thing.  The pixel values in small stars tend to be outliers and therefore they are self-selecting to be "zapped" by the spatial filtering algorithm.

 

Mark




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics