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

another real-world example of Sony's star-eater problem

  • Please log in to reply
51 replies to this topic

#1 whwang

whwang

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1682
  • Joined: 20 Mar 2013

Posted 13 July 2015 - 07:11 PM

Please see the attached images.  They are from Sony A7R and Sigma 50/1.4A @ F4.  One is stacking of 12 3-minute exposures (bulb), and the other is 90 30-second exposures (non-bulb).  The total exposure is slightly different, and this probably just compensates the excess of read noise in the 30-sec case.  So the depth of these two images should not be that different.

 

You can clearly see that many faint stars are GONE in the 3-min exposures, and the core of the bright stars are also strongly suppressed in the 3-min exposures.  Sony's star eater does not only only affect the cores of bright stars, but also faint stars.

 

To me, this is totally unacceptable.  Given what I see here, I can't recommend Sony to ANY astrophotographers.  It is just unacceptable.

 

I will have some struggling about whether to get rid of this camera or not.  Good luck to other Sony owners.

 

Cheers,

Wei-Hao

Attached Thumbnails

  • star_eater.jpg

Edited by whwang, 13 July 2015 - 07:12 PM.

  • sharkmelley likes this

#2 DuncanM

DuncanM

    Surveyor 1

  • -----
  • Posts: 1550
  • Joined: 03 Nov 2009
  • Loc: Arizona Sky Village or the rain forest

Posted 13 July 2015 - 07:21 PM

I would suggest that the astro community write to Sony, and complain. Maybe enough complaints will get them to make the very simple firmware changes needed to fix this problem.

 

This might be a place to start:

 

https://community.so.../ct-p/help-desk



#3 GaryCurran

GaryCurran

    Surveyor 1

  • *****
  • Posts: 1775
  • Joined: 14 Jun 2015
  • Loc: Poulsbo, Washington

Posted 13 July 2015 - 07:28 PM

Wei-Hao,

I've seen reference to the 'Star Eater' issue, but I'm not exactly sure what is meant by that.  I can easily see that the longer exposures have reduced or 'removed' stars from your images, but I'm not sure how or why.

 

Since I have two Sony DSLRs, an A77 and an A58, and these are the cameras I plan on using for my DSLR photography, could you explain a little bit more about what exactly is happening, and the reason for it?  Does it have anything to do with the 'compressed RAW' files that come out of the Sonys, and with their upcoming uncompressed RAW files, is that going to fix it?

 

Any information or links you can provide would be useful.

 

I thank you in advance.



#4 whwang

whwang

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1682
  • Joined: 20 Mar 2013

Posted 13 July 2015 - 07:38 PM

We did already, from a month ago:

https://community.so.../508740#U508740

However, so far no one in Sony responds to this.



#5 whwang

whwang

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1682
  • Joined: 20 Mar 2013

Posted 13 July 2015 - 07:43 PM

Wei-Hao,

I've seen reference to the 'Star Eater' issue, but I'm not exactly sure what is meant by that.  I can easily see that the longer exposures have reduced or 'removed' stars from your images, but I'm not sure how or why.

 

Since I have two Sony DSLRs, an A77 and an A58, and these are the cameras I plan on using for my DSLR photography, could you explain a little bit more about what exactly is happening, and the reason for it?  Does it have anything to do with the 'compressed RAW' files that come out of the Sonys, and with their upcoming uncompressed RAW files, is that going to fix it?

 

Any information or links you can provide would be useful.

 

I thank you in advance.

 

Hi Gary,

 

It has nothing to do with the raw compression.  (I actually do not think the raw compression is such a big deal, back to the old days when Nikon also did this.)  Just like what Nikon did on their DSLRs several years ago, Sony attempts to remove hot pixels in long exposures.  This removes many stars, both bright and faint ones.  On the A7 series, this happens on bulb exposures, but not the non-bulb exposures of 30 sec or shorter.  This is why I compared a long integration of 3 minute exposures and also a long integration of many 30 sec (non-bulb) exposures.  You can clearly see that in the 3-minute case, stars are either removed or strongly suppressed at their centers.

 

Here is another example:

http://www.dpreview....s/post/55841466

There are also examples on this forum, although I forget where to find them.

 

Cheers,

Wei-Hao



#6 ccs_hello

ccs_hello

    Cosmos

  • *****
  • Posts: 8695
  • Joined: 03 Jul 2004

Posted 13 July 2015 - 08:26 PM

Wei-Hao,

 

Just wondering if you run a histogram on an A7R dark,

where does the the camera perform the black clipping?  Does it clip on the right-hand side of the bell curve?

 

My question is about is it just SONY's strong star-eater anti-noise algorithm or due to black-clipping?  May be a combination from both?

 

Clear Skies!

 

ccs_hello



#7 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1203
  • Joined: 19 Feb 2013

Posted 13 July 2015 - 08:29 PM

Wei-Hao,

 

Thanks for producing that example.  I guess you used dithered acquisition in both cases?

 

You are quite right that it affects every unsaturated star - bright ones and faint ones.

 

It's all quite depressing and, like you, I am seriously asking myself whether I should continue to struggle with the A7S.  Or maybe I can get used to doing 30sec exposures only i.e. not using bulb mode.  But this creates huge amounts of data to be processed.

 

Gary,

There's another CN thread that discusses the issue here:

http://www.cloudynig...ater-algorithm/

 

I have an example of a real life single sub with actual holes in stars here:

http://www.cloudynig...ilon/?p=6637856

 

In essence Sony have implemented exactly the same hot-pixel-supression algorithm as the one Nikon used in the early days - the algorithm that made astrophotographers avoid Nikon cameras for years.  The algorithm attacks all pixels that are brighter than their neighbours.  Star centres generally satisfy this condition :(

 

Mark



#8 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1203
  • Joined: 19 Feb 2013

Posted 13 July 2015 - 08:35 PM

Just wondering if you run a histogram on an A7R dark,

where does the the camera perform the black clipping?  Does it clip on the right-hand side of the bell curve?

 

My question is about is it just SONY's strong star-eater anti-noise algorithm or due to black-clipping?  May be a combination from both?

 

The Sony A7 series of cameras do not clip blacks.  The star eating behaviour is totally a result of the undocumented (by Sony) bulb mode hot-pixel-suppression algorithm. 

 

Mark



#9 Jon Rista

Jon Rista

    Hubble

  • *****
  • Posts: 16483
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 13 July 2015 - 08:36 PM

Isn't this caused by LENR? Can't that be turned off?

#10 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1203
  • Joined: 19 Feb 2013

Posted 13 July 2015 - 08:42 PM

Isn't this caused by LENR? Can't that be turned off?

 

No, it's completely different to long exposure noise reduction and there is no way to switch it off.

 

Mark



#11 obin robinson

obin robinson

    Mercury-Atlas

  • *****
  • Posts: 2911
  • Joined: 25 Oct 2012
  • Loc: Jacksonville, FL

Posted 13 July 2015 - 08:47 PM

This "star-eater" problem looks like something that a firmware change could fix.

 

obin :)



#12 DuncanM

DuncanM

    Surveyor 1

  • -----
  • Posts: 1550
  • Joined: 03 Nov 2009
  • Loc: Arizona Sky Village or the rain forest

Posted 13 July 2015 - 08:50 PM

 

 

In essence Sony have implemented exactly the same hot-pixel-supression algorithm as the one Nikon used in the early days - the algorithm that made astrophotographers avoid Nikon cameras for years.  The algorithm attacks all pixels that are brighter than their neighbours.  Star centres generally satisfy this condition :(

 

Mark

Actually, Sony's algorithm is more destructive than Nikon's was because Sony's also effects bright stars as well.  



#13 DuncanM

DuncanM

    Surveyor 1

  • -----
  • Posts: 1550
  • Joined: 03 Nov 2009
  • Loc: Arizona Sky Village or the rain forest

Posted 13 July 2015 - 08:52 PM

This "star-eater" problem looks like something that a firmware change could fix.

 

obin :)

Absolutely, and Nikon did change their firmware while the Nikonhacker group has been able to issue firmware mods to remedy the problem on some older Nikons:

https://nikonhacker.com/



#14 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1203
  • Joined: 19 Feb 2013

Posted 13 July 2015 - 08:53 PM

 

 

 

In essence Sony have implemented exactly the same hot-pixel-supression algorithm as the one Nikon used in the early days - the algorithm that made astrophotographers avoid Nikon cameras for years.  The algorithm attacks all pixels that are brighter than their neighbours.  Star centres generally satisfy this condition :(

 

Mark

Actually, Sony's algorithm is more destructive than Nikon's was because Sony's also effects bright stars as well.  

 

 

They are equally destructive - it's the same algorithm. Both affect all unsaturated stars.



#15 ccs_hello

ccs_hello

    Cosmos

  • *****
  • Posts: 8695
  • Joined: 03 Jul 2004

Posted 13 July 2015 - 09:47 PM

Old "Nikon star eater" algorithm was initially cured by "Mode 3" work-around.

 

Any hope on doing the same (i.e., before a SONYhacker comes to reality :) )?

 

P.S. dpreview "Con" findings will trigger camera mfgs to change their practice.  Second alternative is DPR's SONY user forum.

 

Clear Skies!

 

ccs_hello



#16 whwang

whwang

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1682
  • Joined: 20 Mar 2013

Posted 13 July 2015 - 09:59 PM

P.S. dpreview "Con" findings will trigger camera mfgs to change their practice.  Second alternative is DPR's SONY user forum.

 

Will try that.  However, the first example of this problem on Sony that I am aware of actually came form dpreview.  It hasn't triggered anything yet.  Here we are probably looking at a long term effort, not a solution any time soon.

 

Cheers,

Wei-Hao



#17 Jon Rista

Jon Rista

    Hubble

  • *****
  • Posts: 16483
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 13 July 2015 - 10:19 PM

Isn't this caused by LENR? Can't that be turned off?

 
No, it's completely different to long exposure noise reduction and there is no way to switch it off.
 
Mark


Hmm, bummer. Sounds like we really need some Sony hackers to come along and clean up their mess. :p
  • calypsob likes this

#18 GaryCurran

GaryCurran

    Surveyor 1

  • *****
  • Posts: 1775
  • Joined: 14 Jun 2015
  • Loc: Poulsbo, Washington

Posted 13 July 2015 - 10:26 PM

Thank you, all.  I read some of Mark's thread over on DP Review, and saw this star test.  I guess that limits me to 30 second exposures.  Crap, I'm not ready to buy a CCD camera yet.



#19 Jon Rista

Jon Rista

    Hubble

  • *****
  • Posts: 16483
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 13 July 2015 - 10:33 PM

You can use pretty high ISO settings on Sony cameras, and still get excellent results. That will mitigate the problem. I've often stacked hundreds of subs, and even with my 5D III (arguably inferior to the A7 series of cameras) I have gotten some great results. I would stick with 30 second exposures, and just get TONS and TONS of them. ;) If you expose for your hours using 30s subs, that would be 480 subs. I am not sure if DSS could handle that...but PixInsight definitely can. You can get pretty good results with four hours of total exposure.

#20 whwang

whwang

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1682
  • Joined: 20 Mar 2013

Posted 13 July 2015 - 11:49 PM

You can use pretty high ISO settings on Sony cameras, and still get excellent results. That will mitigate the problem. I've often stacked hundreds of subs, and even with my 5D III (arguably inferior to the A7 series of cameras) I have gotten some great results. I would stick with 30 second exposures, and just get TONS and TONS of them. ;) If you expose for your hours using 30s subs, that would be 480 subs. I am not sure if DSS could handle that...but PixInsight definitely can. You can get pretty good results with four hours of total exposure.

 

 

Hi Jon,

 

Yes, I can do tons of 30-sec exposures, except that it won't be as efficient as 3-min exposures because of the read noise and overheads.  In the above example, I had already push the ISO to 1600.  Going to higher ISO will not give me lower read noise, and 30-sec exposures under F4 is still affected by read noise.  Then there is a few seconds of pause time between exposures.  The time wasted on this will increase a lot if I do 30 sec instead of 30 min.

 

The example I showed above was stacked with PixInsight.  So yes, PixInsight can do it, except that it is slooooooooow.  Essentially I will need to spend 3x to 4x more time on calibration and stacking than the actual exposure.  This means my image processing cannot catch up with the speed I take images.

 

Overall, this 30-sec limit just make everything much much more painful.  Good results are still possible, but doing this makes me wonder why do I have to torture myself like this?

 

 Cheers,

Wei-Hao



#21 jhayes_tucson

jhayes_tucson

    Aurora

  • *****
  • Posts: 4569
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 14 July 2015 - 12:02 AM

Good work Wei Hao!  That's really not good...like at all.  I think that I'll stick with my Canon a while longer.

 

John



#22 GaryCurran

GaryCurran

    Surveyor 1

  • *****
  • Posts: 1775
  • Joined: 14 Jun 2015
  • Loc: Poulsbo, Washington

Posted 14 July 2015 - 12:24 AM

I wanted to stay at ISO 800 or less for my exposures, which would have meant longer exposure times.  But, for now, it's a moot point, at least until I can get several sessions under the stars and learn the basics, and get a feel for what the cameras can do, and what my exposure times need to be for certain objects and such.  So, I'm not really worried right now.  But, as I progress and get better, unless Sony comes out with a fix for this, I'll end up looking for a new camera, which will probably be a dedicated imager.  That's a ways off, though.



#23 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 9006
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 14 July 2015 - 04:30 AM

John, hope you wouldn't mind, I am going to borrow your Dumbbell nebula to draw a star comparison. Following are somewhat better quality examples than above.


All, except that John's stars are bloated a bit probably due to well overflow, I don't see much difference in stars; rather I see more discrete and well defined stars in a7S due to improved FWC. Regards


•24x4min, ISO1600 (6D jhayes_tucson...)
•96min integration

•14x3min, ISO1600 (a7S mmalik...)
•42min integration

•30x2min, ISO 1600 (6D mmalik...)
•60min integration

 

 

Attached Thumbnails

  • jhayes_tucson-6D.jpg
  • M27,NGC6853-DumbbellNebula_a7S_CR2_12c.jpg
  • M27,NGC6853-DumbbellNebula_CR_12c.jpg

Edited by mmalik, 14 July 2015 - 04:55 AM.


#24 vdb

vdb

    Viking 1

  • -----
  • Posts: 835
  • Joined: 08 Dec 2009

Posted 14 July 2015 - 05:12 AM

I'm so happy I cancelled my Sony A7S order last minute and went for the Nikon D810A.

Also people claiming the low noise of this camera are actually looking at in camera noise reduction that cannot be turned off ones over 30 sec subs.

 

Every time I see a Sony image I feel it is overly smooth because off this algorithm. Yes you can stretch more and maybe get to fainter stuff, but people trying to compare make the fatal flaw of

a) not matching the same sampling and

b) if I do a 4 pixel HPS in PS I get an as "clean" plastic image ... 

 

I really hope Sony will fix this as the large pixel and "native" low noise are very promising, just not for the moment though.

 

/Yves



#25 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1203
  • Joined: 19 Feb 2013

Posted 14 July 2015 - 05:19 AM

John, hope you wouldn't mind, I am going to borrow your Dumbbell nebula to draw a star comparison. Following are somewhat better quality examples than above.


All, except that John's stars are bloated a bit probably due to well overflow, I don't see much difference in stars; rather I see more discrete and well defined stars in a7S due to improved FWC. Regards

 

 

I think that those examples were taken at longer focal lengths and have well sampled stars so they will be relatively unaffected by the hot pixel supression ("star eater") algorithm.  The problem is noticeable at shorter focal lengths where the very tight (and undersampled) stars appear as individual bright pixels to the 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.







Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics