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

Is Sony Really Alpha?

  • Please log in to reply
1941 replies to this topic

#326 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 25 April 2015 - 06:45 PM

Hi Mike,

 

Thanks for your interest in this and thanks especially for posting those Canon 6D and 7DII images.  The images prove something very interesting: in both the JPGs and in the raw files I can see "stand-alone" bright pixels.  This is precisely what one would expect to see in a typical raw file with no pre-processing.  This is a normal characteristic of any sensor.  This absence of pre-processing is why the Canon cameras have always been held in high regard by the astro-imaging community. 

 

The issue I am seeing with the Sony A7S used in Bulb mode (it doesn't appear to affect other modes) is that bright pixels never appear singly but always in pairs with absolutely identical pixel values.  When I talk of "pixel doubling" I really mean the absence of single bright pixels - I now regret my inaccuracy of terminology because it might have led to confusion.  In any case, the absence of single bright pixels replaced by pairs of identical pixels is a clear indication that some pre-processing algorithm has been applied, because this simply cannot arise by chance across a whole image.  To see these identical paired values it is necessary to view the un-debayered raw file using an application such as RawDigger, PixInsight or IRIS where the actual stored pixel values can be viewed directly.

 

You suggest that the so-called "pixel-doubling" may be "no more than inherent characteristics of a given sensor and its design".  However, in the case of the Sony A7S "stand-alone" bright pixels never appear in Bulb mode but they do appear almost everywhere in other modes.  This indicates that it is not a characteristic of the sensor itself but of some algorithmic pre-processing performed only in Bulb mode.

 

My interest in astro-imaging extends beyond simply producing pretty pictures.  I am also interested in recording star magnitudes, looking a variable stars etc.  Any algorithm that punches a hole in my stars, thinking that the central peak value is a hot pixel, is not much use in that context.

 

As much as I want to like the Sony A7S (after all, I've paid good money for it and European prices are not cheap), I am seriously beginning to think I should trade it in for something like the Canon 6D or 7DII with its 14 bit unadulterated raw files or for one of the Nikons where the firmware can be hacked to produce an unadulterated long-exposure image.

 

Mark


Edited by sharkmelley, 25 April 2015 - 06:50 PM.


#327 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 25 April 2015 - 08:39 PM



Bwa, if anyone took a standard dark (without LENR) on any alpha (I tested a7S and a7II), one will see the following pattern and folks are starting to call it 'pixel doubling' [I am not sure what to call it just yet or if it is even a problem?]. Go ahead and do it on your own alpha and you'll see it.

 

 

Note: If one were just stacking lights (the modern trend) this pattern will be found in the stacked image as well.

 

 

With that said, I am not sure where it leaves things?

 

 

Plan: This is what I am going to do; I am going to take similar darks with 6D and 7DII and post results for comparison sake and take the discussion from there. At this point, I see it no more than inherent characteristics of a given sensor and its design (pixel size, etc.). Regards

 

 

Note: Following are about 600% zoom of in-camera JPGs.

I've shot 61 sec darks (in Bulb Mode).

 

What processing do I have to do to make these artifacts show up?  I've blown the dark up and can't see anything.  Do you push the exposure, up the gamma and/or brightness?  Tell me what you do to get those double pixels, or for that matter, any hot pixels to show up?

 

bwa



#328 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 25 April 2015 - 09:04 PM

a7S dark/s I show above were about 3min long in bulb mode and saved as RAW as well as JPG; sections posted were about 600% zoom into the camera JPG without any processing.

 

 

Note: Somewhat "paired" looking color mottle example/s I show above is more of a visual inspection of an in-camera JPG highly zoomed-in; what Mark is talking about is at much deeper level and are the actual pixel values.

 

 

IMPORTANT: What I am NOT sure is if the color mottle I am referring to at high level (visual inspection) is actually the same thing Mark is referring to when he talks of paired pixel values? In other word, are we talking same exact pixels in both cases? I am 'not' sure. May be Mark can confirm? Regards


Edited by mmalik, 25 April 2015 - 09:05 PM.


#329 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 25 April 2015 - 09:46 PM

The attached file is a 600% crop out of PixInsight of a 550 sec dark at ISO 6400:

(please see the full sized attachment of detail)

A7S_550sec_ISO6400_6xZoom.jpg

I could find no hot pixels in the image, i.e. Sony's bad pixel mapper is working as it is supposed to.

This is a screen capture of a 600% zoom from PixInsight of a Sony RAW image.

 

I seldom, if ever, shoot subs longer than 120 sec. so 550 sec is extreme.

 

If this is the problem being discussed, I have absolutely no concerns about image quality off my A7S or A7R...

 

bwa



#330 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 26 April 2015 - 04:41 AM

The attached file is a 600% crop out of PixInsight of a 550 sec dark at ISO 6400:

 

I could find no hot pixels in the image, i.e. Sony's bad pixel mapper is working as it is supposed to.

 

If this is the problem being discussed, I have absolutely no concerns about image quality off my A7S or A7R...

 

 

Yes, I really think you are really seeing the pixel pairing symptom there.  I'm sure Sony's bad pixel mapper does a good job but under normal circumstances you would still expect to see some random single bright pixels arising purely by chance.  My hypothesis is that the reason you are not seeing these random single bright pixels is that the Sony hot pixel suppression algorithm has removed them (truncating the value of the hot pixel to that of a neighbour, thus artificially creating an identical pair of pixels with the same value).  This HPS algorithm works very differently to hot pixel mapping because it is applied regardless over the whole image.

 

Remember though, this is not the real problem I am concerned about.  The real problem is that HPS algorithm might be removing stars from the image.  The absence of single bright pixels in the image is simply an important indicator that the algorithm was activated for that image.

 

I hope to do some tests with the camera attached to my 500mm f/l scope tonight.  The scope has capable of producing 6 micron stars, which is smaller than the pixel size.  I expect to find that such stars will be very severely affected.

 

Mark



#331 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 26 April 2015 - 06:54 AM

I suspect the problem is related to how SONY is sacrificing some pixels just for contrast-difference AF thus these pixels are replaced by the normal same-color pixel near-by.
When the "normal, ordinary pixel" is a hot pixel, its high value gets replicated to the contrast-detect special pixel(s).

 

I am seeing similar pixel pairing (see above...) in 7DII and 6D, more in 7DII, less in 6D. I don't think this is only happening in Sony; I can see the same effect more or less in Canon [of course subtleties are there given different manufacturers]. What I don't know if this is a problem or an intended feature/design? Regards


Edited by mmalik, 26 April 2015 - 10:29 AM.


#332 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 26 April 2015 - 08:19 AM

I think I'm probably not explaining this very well.  Of course it is always possible to see pixel pairing occurring by chance in any image: Canon, Nikon or Sony.  The eye will always be drawn to such things.

 

The key difference is the absence of single bright pixels in the A7S image.  Look at the difference in characteristic between the two cropped darks below:

 

6D Dark (courtesy of mmalik):

6D_Dark.JPG

 

A7S (Bulb mode) Dark:

A7S_Dark.JPG

 

The 6D Dark has a liberal "salt and pepper" sprinkling of single bright pixels and single dim pixels.  In the A7S (Bulb mode) dark, the bright pixels and dim pixels only appear in pairs or sometimes triples because of some Sony algorithm that has been internally run on the data.  The pairs or triples are always two pixels apart in the undebayered files above because the pairs are always the same colour pixel i.e. R, G or B.

 

But in the end, I don't think any of us really care what a dark looks like. The thing I really care about is that this algorithm severely damages or even destroys small stars.

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

The lack of single bright pixels in the A7S dark is merely a tell-tale symptom of this problem.

 

Mark


Edited by sharkmelley, 26 April 2015 - 08:23 AM.


#333 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 26 April 2015 - 10:34 AM

Thanks for the dark analysis. Mark, can you do the same for 7DII; I feel 7DII may be bit more closer to a7S in such pairings. Regards



#334 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 26 April 2015 - 11:08 AM

The Canon 7DII is just like the Canon 6D.  The A7S is different to both.

 

7DII Dark (courtesy of mmalik):

7DII_Dark.JPG

 

Mark



#335 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 26 April 2015 - 12:02 PM

 

The attached file is a 600% crop out of PixInsight of a 550 sec dark at ISO 6400:

 

I could find no hot pixels in the image, i.e. Sony's bad pixel mapper is working as it is supposed to.

 

If this is the problem being discussed, I have absolutely no concerns about image quality off my A7S or A7R...

 

 

Yes, I really think you are really seeing the pixel pairing symptom there.  I'm sure Sony's bad pixel mapper does a good job but under normal circumstances you would still expect to see some random single bright pixels arising purely by chance.  My hypothesis is that the reason you are not seeing these random single bright pixels is that the Sony hot pixel suppression algorithm has removed them (truncating the value of the hot pixel to that of a neighbour, thus artificially creating an identical pair of pixels with the same value).  This HPS algorithm works very differently to hot pixel mapping because it is applied regardless over the whole image.

 

Remember though, this is not the real problem I am concerned about.  The real problem is that HPS algorithm might be removing stars from the image.  The absence of single bright pixels in the image is simply an important indicator that the algorithm was activated for that image.

 

I hope to do some tests with the camera attached to my 500mm f/l scope tonight.  The scope has capable of producing 6 micron stars, which is smaller than the pixel size.  I expect to find that such stars will be very severely affected.

 

Mark

 

If one considers the debayering process (depending upon which of the at least four different methods is used: http://www.stark-lab...ayering_API.pdf), a hot/cold photosite is going to impact more than one pixel in the final image.  The same will be true for stars that totally fit within one photosite, i.e.: undersampled.  In undersampled images the best way to recover the information is via Drizzling.

 

bwa



#336 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 26 April 2015 - 12:15 PM

 

If one considers the debayering process (depending upon which of the at least four different methods is used: http://www.stark-lab...ayering_API.pdf), a hot/cold photosite is going to impact more than one pixel in the final image.  The same will be true for stars that totally fit within one photosite, i.e.: undersampled.  In undersampled images the best way to recover the information is via Drizzling.

 

bwa

 

 

Agreed.  But even Drizzle can't undelete stars!

 

Mark



#337 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 26 April 2015 - 12:53 PM

 

 

If one considers the debayering process (depending upon which of the at least four different methods is used: http://www.stark-lab...ayering_API.pdf), a hot/cold photosite is going to impact more than one pixel in the final image.  The same will be true for stars that totally fit within one photosite, i.e.: undersampled.  In undersampled images the best way to recover the information is via Drizzling.

 

bwa

 

 

Agreed.  But even Drizzle can't undelete stars!

 

Mark

 

Yup, can't argue with that.

 

A comparison of the three debayering algorithms used in PixInsight on a 100x100 pixel crop from my previous 550 sec @ ISO 6400 dark:

(click on the image for the full sized attachment)

debayering comparison.jpg

As would be assumed the VNG shows the most detail (and emphasizes paired hot/cold pixels).

 

Now we can only hope the hot/cold pixel mapping doesn't apply to stars...

 

bwa



#338 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 April 2015 - 10:52 AM

I tried the A7S on my Tak Epsilon tonight (focal length 500mm). It was a typical cloudy night, so I placed an artificial star at the end of the garden. The results are pretty much similar to the example in the first post.

In both examples below the left hand star is non-Bulb mode and the right hand star is Bulb mode with the same exposure length.

10 sec ISO 125:
post-216852-0-76201800-1430180819.jpg

2 sec ISO 2000:
post-216852-0-65933600-1430180830.jpg

I can't totally rule out a sub-pixel scope movement between the exposures but the loss of the bright pixels is quite dramatic. These are two representative examples of a very consistent pattern.

To be fair, I should point out that on a longer focal length scope, where the star is adequately sampled (say 1000mm or more), this would not be such an issue - the central peak of the star would be flattened but the main body of the star would remain more or less intact.

Mark


Borrowing medical terminology :), this is a good in vitro test. Next step would be to perform an in vivo test, i.e., test of both very small and bit larger "actual" stars. Regards


In retrospect, I feel most folks are cognizant of pixel size and focal length interplay to avoid such an artifact [...if this were to turn out an issue in real life]. Read more here... and here.... Regards


Edited by mmalik, 28 April 2015 - 10:54 AM.


#339 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,549
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 28 April 2015 - 12:54 PM

In retrospect, I feel most folks are cognizant of pixel size and focal length interplay to avoid such an artifact [...if this were to turn out an issue in real life]. Read more here... and here.... Regards

 

 

I agree to a certain extent.  However, everyone with a DSLR/Mirrorless camera will also do some imaging with a wide angle lens at some point.  After all, it's one of the strengths of having a DSLR :)



#340 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 28 April 2015 - 06:01 PM

 

In retrospect, I feel most folks are cognizant of pixel size and focal length interplay to avoid such an artifact [...if this were to turn out an issue in real life]. Read more here... and here.... Regards

 

 

I agree to a certain extent.  However, everyone with a DSLR/Mirrorless camera will also do some imaging with a wide angle lens at some point.  After all, it's one of the strengths of having a DSLR :)

 

And I often do image with a wide-angle lens...  But I have an awful hard time determining whether I'm actually missing any stars!?  How do you know when you look at something like:

South%2BMilky%2BWay%2B-2014Jun27-1.jpg

The whole darn image is essentially stars.  Missing a few might be an improvement!?

 

Based on the forums where the A7 series cameras' handling of hot/cold pixels has been discussed, I don't think Sony's approach is all that great; however, I'm not sure if it has a significant impact on the cameras' ability to turn out great images?  Over the past week or so since this topic has been discussed I've done a fair bit of pixel peeping and can't determine whether the problem impacts my images or not.  If the problem isn't obvious, is it a problem?

 

bwa



#341 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 09:28 AM

Short of doing RAW analysis, I guess one way would to compare the actual image against another camera with same/similar wide angle lens. I don't think algorithm would have any significant impact on over all image quality. In the grand scheme of things this may turn out to be a red herring for all we know? In short, keep at it; don't let the discussion distract you. Regards


Edited by mmalik, 30 April 2015 - 07:43 AM.

  • bwallan likes this

#342 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 10:10 AM

An actual star field comparison [tiny section of the Cygnus Wall] of a7S and 6D. Single RAW was used for each of the images below. After STF, zoom (+) button was clicked six times in PixInsight; star size/field differences were NOT adjusted; hence some subtleties in star size/field. Both images were taken on the same equipment (~2300mm FL refractor) but on different nights so some seeing and other atmospheric factors might also be at play. Regards 

 

 

Note: a7S was IDAS LPS-D1 filtered; 6D was IDAS HEUIB-II filtered

 

Attached Thumbnails

  • SonyA7S_DSC00105Zoom.JPG
  • Canon6D_IMG_0611Zoom.JPG


#343 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 11:09 AM

Some real-life comparison of integrations from 60Da, 6D and a7S; all were taken on the same equipment (254mm APO) but on different occasions:

 

  • 60Da
  • 7x10min, ISO400, LENR
  • 70min integration, guided
  • Complete frame (APS-C)

 

High resolution...

 

Attached Thumbnails

  • NGC7000-CygnusWall_12c.jpg


#344 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 11:11 AM

  • 6D (modified)
  • 18x5min, ISO1600, Multiple LENR
  • 90min integration, guided
  • Minor crop

 

High resolution...

Attached Thumbnails

  • NGC7000-CygnusWall_12c.jpg

Edited by mmalik, 29 April 2015 - 11:32 AM.


#345 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 11:13 AM

  • a7S (modified)
  • 13x3min, ISO2000, LENR
  • 39min integration, guided
  • Minor crop

 

High resolution...

Attached Thumbnails

  • NGC7000-CygnusWall_12c.jpg

Edited by mmalik, 29 April 2015 - 11:32 AM.


#346 bwallan

bwallan

    Viking 1

  • -----
  • Posts: 563
  • Joined: 08 Jul 2010
  • Loc: Alberta, Canada

Posted 29 April 2015 - 01:53 PM

I'd be quite happy with either the 6D or A7S image...

 

Not that many years ago I'd be happy with the 60Da image as well.

 

bwa



#347 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 29 April 2015 - 06:26 PM

If you compare high res. of 6D with a7S (above), a7S is much better than 6D [barring any acquisition subtleties]. I think that was the message in my comparison above :). I have not done any point objects like planetary nebulas or galaxies with a7S yet, but for distributed nebulas, a7S wins leaps and bounds ahead of 6D.

 

 

6D is a great camera but a7S has a class of its own. For example, consider single exposure comparison of 6D and a7S below; one can make out quite a bit of detail in Cygnus Wall even in one exposure on a7S while it is not the case in 6D.

 

 

Same is evident on the back of camera (BoC) preview of the image after acquisition.

Note: BoC 6D image is ISO1600/5min; BoC a7S image is ISO2000/3min

 

 

Last but not lest, noise is simply much much lower in a7S. Regards

 

Note: Difference in scale of the two samples is stemming from different pixel size/file size of each camera and how processing software renders it. Both RAWs were independelty/identically processed from start to finish.

Attached Thumbnails

  • SingleExposureComparison.jpg
  • DSC02805.JPG
  • DSC02773.JPG

Edited by mmalik, 01 May 2015 - 12:04 PM.

  • Relativist likes this

#348 Relativist

Relativist

    Cosmos

  • *****
  • Posts: 8,152
  • Joined: 11 Oct 2003
  • Loc: OC, CA, USA

Posted 01 May 2015 - 11:38 AM

Very nice comparisons. Any chance you can show us what the A7s can do with the ISO cranked up with much shorter exposures on the same objects?



#349 vdb

vdb

    Surveyor 1

  • -----
  • Posts: 1,556
  • Joined: 08 Dec 2009

Posted 01 May 2015 - 11:59 AM

I beg to differ on opinion, at over 2000 mm the A7S is better matched to the focal length ...

on wider field equipment the 6D will resolve more detail ... for fairness you should also resize the 6D to same pixel scale as the A7S, the sony will still be better due to the pixel scale, but difference will be less dramatic, downscaling when done properly is also kond of noise reduction ...

 

At the focal length the 6D has a harder time getting away from the readnoise, and so a longer exposure will take of course more noise thermal noise ... a better comparison should be the A7R ... 

 

/Yves



#350 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 10,664
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 01 May 2015 - 12:34 PM

...should also resize the 6D to same pixel scale as the A7S

 

...a better comparison should be the A7R

 

If folks would like to conduct pixel scale/level analysis, links below:

 

 

6D: Download RAW file named IMG_0611.CR2...

a7S: Download RAW file named DSC00105.ARW...

 

Note: Both of the RAWs are LENR enabled!

 

 

Yves, going by the pixel size, 6D "is" the closest to a7S. Regards

 

post-205769-0-99191000-1425407067.jpg


Edited by mmalik, 01 May 2015 - 12:42 PM.



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