Star Eater and A7rIII firmware 2 - encouraging test

Last night I got a few hours with pretty clear skies, with a 25% waning moon low in the sky and slightly hazy atmosphere, so I could test the A7rIII with the now almost two months old firmware version 2.00. I have not had the opportunity to do a reliable test earlier, just an initial test under so-so conditions.

This test was done with the camera on a tripod and 21 mm and 35 mm focal lengths. Exposures were 3.2 (no Star eater) and 4 seconds (Star Eater kicks in). The files were stacked to a total of 60 seconds to even out shutter speed differences. RegiStar was used for stacking, then the files were adjusted slightly to the same color and sky background weight with Photoshop.

Files taken at 3.2 second shows a tad crisper very faint stars, but there is surprisingly little differences in magnitude limit, checked against Starry Night Pro 7.

• 21 mm f:2.8 about mag 10.7
• 35 mm f:2 about mag 11.5

I would have expected more visible differences, especially with the 21 mm lens.

My conclusion: No worries using the A7rIII for night sky photography and even deep sky astrophotography, but not for serious astrometry or photometry.

Test photographs, 100% crop:

[ATTACH alt=""Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area."]2106133[/ATTACH]
"Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area.

Magnar W said:

Last night I got a few hours with pretty clear skies, with a 25% waning moon low in the sky and slightly hazy atmosphere, so I could test the A7rIII with the now almost two months old firmware version 2.00. I have not had the opportunity to do a reliable test earlier, just an initial test under so-so conditions.

This test was done with the camera on a tripod and 21 mm and 35 mm focal lengths. Exposures were 3.2 (no Star eater) and 4 seconds (Star Eater kicks in). The files were stacked to a total of 60 seconds to even out shutter speed differences. RegiStar was used for stacking, then the files were adjusted slightly to the same color and sky background weight with Photoshop.

Files taken at 3.2 second shows a tad crisper very faint stars, but there is surprisingly little differences in magnitude limit, checked against Starry Night Pro 7.

• 21 mm f:2.8 about mag 10.7
• 35 mm f:2 about mag 11.5

I would have expected more visible differences, especially with the 21 mm lens.

My conclusion: No worries using the A7rIII for night sky photography and even deep sky astrophotography, but not for serious astrometry or photometry.

Test photographs, 100% crop:

[ATTACH alt=""Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area."]2106133[/ATTACH]
"Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area.

Click to expand...

Good work Magnar.

I also was out the same time. I took some pixel shift images from a tracked A7r3 and Sigma Art 35 1.4 lens.

I notice the original images show the stars as quite bright and vivid but when I use Sony viewer and it makes a combined pixel shift image it dulls down the stars a large amount.

Have you ever seen this? Is there a setting in the software to prevent that?

Perhaps I should work out how to use RAW Therapee and see if it does the same thing. The Sony software washes out the stars and they are no longer bright and vivid.

Greg.

Astrophotographer 10 said:

Magnar W said:

Last night I got a few hours with pretty clear skies, with a 25% waning moon low in the sky and slightly hazy atmosphere, so I could test the A7rIII with the now almost two months old firmware version 2.00. I have not had the opportunity to do a reliable test earlier, just an initial test under so-so conditions.

This test was done with the camera on a tripod and 21 mm and 35 mm focal lengths. Exposures were 3.2 (no Star eater) and 4 seconds (Star Eater kicks in). The files were stacked to a total of 60 seconds to even out shutter speed differences. RegiStar was used for stacking, then the files were adjusted slightly to the same color and sky background weight with Photoshop.

Files taken at 3.2 second shows a tad crisper very faint stars, but there is surprisingly little differences in magnitude limit, checked against Starry Night Pro 7.

• 21 mm f:2.8 about mag 10.7
• 35 mm f:2 about mag 11.5

I would have expected more visible differences, especially with the 21 mm lens.

My conclusion: No worries using the A7rIII for night sky photography and even deep sky astrophotography, but not for serious astrometry or photometry.

Test photographs, 100% crop:

[ATTACH alt=""Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area."]2106133[/ATTACH]
"Star Eater" test Sony A7rIII, results with 21 mm and 35 mm lenses. For 21 mm: Gemini and M35 area. For 35 mm: Aldebaran and Hyades area.

Click to expand...

Good work Magnar.

I also was out the same time. I took some pixel shift images from a tracked A7r3 and Sigma Art 35 1.4 lens.

I notice the original images show the stars as quite bright and vivid but when I use Sony viewer and it makes a combined pixel shift image it dulls down the stars a large amount.

Click to expand...

I have just done a brief PixelShift test for astro, so I don't guite know how the results are, other than that it works! All my tracking devices are miles away for the moment ...

Astrophotographer 10 said:

Have you ever seen this? Is there a setting in the software to prevent that?

Click to expand...

Problem is that we need software that merge PixelShift files correctly, and at the moment I don't know others than Sony Edge and RAW Therapee - the first is slow and very contrasty at PixelShift default (might be this that make faint stars fainter?), and I have not yet tried the therapeut … ;-)

Astrophotographer 10 said:

Perhaps I should work out how to use RAW Therapee and see if it does the same thing. The Sony software washes out the stars and they are no longer bright and vivid.

Click to expand...

Might be an idea! I am using RegiStar for summing image data, and I have just started to look deeper into Astroart (handles dark & flat field, and removes gradients well), but this software is not good at stacking wide angle frames taken with camera on a tripod. Even Photoshop 'Statistics' function does better. ;-)

These look a tiny bit out of focus (or seeing was bad). So I wouldn't expect to see much difference since the stars aren't pinpoints

Magnar W said:

Problem is that we need software that merge PixelShift files correctly, and at the moment I don't know others than Sony Edge and RAW Therapee - the first is slow and very contrasty at PixelShift default (might be this that make faint stars fainter?), and I have not yet tried the therapeut … ;-)

Click to expand...

SonyPixelShift2DNG will create a DNG file. The advantage being that any raw convertor can used including various astro-processing software.

https://www.fastrawviewer.com/blog/SonyPixelShift2DNG-converter-Beta

Disclaimer: I don't have a A7RIII so I haven't tried this myself.

Mark

--
Takahashi Epsilon 180ED
H-alpha modified Sony A7S
http://www.markshelley.co.uk/Astronomy/

RicksAstro said:

These look a tiny bit out of focus (or seeing was bad). So I wouldn't expect to see much difference since the stars aren't pinpoints

Click to expand...

Focus is as good as possible with these lenses and the A7rIII loupe finder at highest magnification, and seeing was about average. I might squeeze a bit more out of these lenses a dry, cloudless night during the spring here up north.

Still, the results are good enough to say that the "star eater" isn't an issue for night sky photographers and even acceptable for much aesthetic deep sky photography.

sharkmelley said:

Magnar W said:

Problem is that we need software that merge PixelShift files correctly, and at the moment I don't know others than Sony Edge and RAW Therapee - the first is slow and very contrasty at PixelShift default (might be this that make faint stars fainter?), and I have not yet tried the therapeut … ;-)

Click to expand...

SonyPixelShift2DNG will create a DNG file. The advantage being that any raw convertor can used including various astro-processing software.

https://www.fastrawviewer.com/blog/SonyPixelShift2DNG-converter-Beta

Disclaimer: I don't have a A7RIII so I haven't tried this myself.

Click to expand...

Thanks! This might be worth a try!

I will do some PixelShift exposures next time we have clear skies here, and that might take some time …

Thanks Magnar for testing.

Magnar W said:

RicksAstro said:

These look a tiny bit out of focus (or seeing was bad). So I wouldn't expect to see much difference since the stars aren't pinpoints

Click to expand...

Focus is as good as possible with these lenses and the A7rIII loupe finder at highest magnification, and seeing was about average. I might squeeze a bit more out of these lenses a dry, cloudless night during the spring here up north.

Still, the results are good enough to say that the "star eater" isn't an issue for night sky photographers and even acceptable for much aesthetic deep sky photography.

Click to expand...

Well, it's just that the stars are all greater than a pixel or 2 across. So star eater wouldn't apply. Not saying your conclusion is wrong, just that it's inconclusive.

I'll be very surprised (but very happy!) if there is a difference between the behaviour of firmware version 2 and the previous version.

As always, Sony should really do one of two things:

• Give the ability to switch off the digital spatial filtering
• Update the algorithm to protect stars better. The decision to cap the value of a pixel should be based on the neighbouring pixels of all colours - not just those of the same colour. Some Nikon cameras still behave like Sony and produce the same green tinted stars but the good cameras (e.g. the D810A which is built for astrophotography) take this preferred approach.

Mark

--
Takahashi Epsilon 180ED
H-alpha modified Sony A7S
http://www.markshelley.co.uk/Astronomy/

RicksAstro said:

Magnar W said:

RicksAstro said:

These look a tiny bit out of focus (or seeing was bad). So I wouldn't expect to see much difference since the stars aren't pinpoints

Click to expand...

Focus is as good as possible with these lenses and the A7rIII loupe finder at highest magnification, and seeing was about average. I might squeeze a bit more out of these lenses a dry, cloudless night during the spring here up north.

Still, the results are good enough to say that the "star eater" isn't an issue for night sky photographers and even acceptable for much aesthetic deep sky photography.

Click to expand...

Well, it's just that the stars are all greater than a pixel or 2 across. So star eater wouldn't apply. Not saying your conclusion is wrong, just that it's inconclusive.

Click to expand...

If there were 1 pixel stars, hot pixel removal would have taken them away. What would such stars add to aesthetical images, except for looking like hot pixels?

Also, when do you get atmospheric conditions that allow you to register a star at one single pixel with a 42 Mp camera? With longer focal lengths the light from faint stars will be spread over more pixels anyway.

I think there is a technical part with spatial filtering, that has to be judged against the cleanness of shadow areas for long exposure photography in general. And there is an aesthetical part, where "star eater" at this level should not be a concern.

Camera technique, like ultra precise focusing and staying within the time span so that stars are not stretched when using a tripod, and also post processing skills and eventual use of specialized software for editing astrophotography, with their set of special filters, would probably have more impact on the end result than the A7rIII in-camera spatial filtering alone.

If I was going to work at the level you are indicating here, I would go for no less than a cooled monochrome astro camera with filter wheel, and software like MaxIm DL to handle dark, flats, bias frames, etc.

sharkmelley said:

I'll be very surprised (but very happy!) if there is a difference between the behaviour of firmware version 2 and the previous version.

Click to expand...

Hard to tell, since I have not tested side by side. I tend to agree with you, though.

sharkmelley said:

As always, Sony should really do one of two things:

• Give the ability to switch off the digital spatial filtering
• Update the algorithm to protect stars better. The decision to cap the value of a pixel should be based on the neighbouring pixels of all colours - not just those of the same colour. Some Nikon cameras still behave like Sony and produce the same green tinted stars but the good cameras (e.g. the D810A which is built for astrophotography) take this preferred approach.

Click to expand...

Sure! We all want the best starting point for our astrophotography.

What would also be intersting, was to see the differences with spatial filtering and no filtering at longer exposures, say some minutes, since this is more relevant for tracked astrophotography. How clean would the shadows be?

Magnar W said:

sharkmelley said:

I'll be very surprised (but very happy!) if there is a difference between the behaviour of firmware version 2 and the previous version.

Click to expand...

Hard to tell, since I have not tested side by side. I tend to agree with you, though.

Member said:

As always, Sony should really do one of two things:

• Give the ability to switch off the digital spatial filtering
• Update the algorithm to protect stars better. The decision to cap the value of a pixel should be based on the neighbouring pixels of all colours - not just those of the same colour. Some Nikon cameras still behave like Sony and produce the same green tinted stars but the good cameras (e.g. the D810A which is built for astrophotography) take this preferred approach.

Click to expand...

Sure! We all want the best starting point for our astrophotography.

What would also be intersting, was to see the differences with spatial filtering and no filtering at longer exposures, say some minutes, since this is more relevant for tracked astrophotography. How clean would the shadows be?

Click to expand...

Just a side note:

Have tried out the A7III (do not have the A7RIII) for astrophoto, latest firmware.

This apply to the A7III:

- Spatial filtering is mild but still there, very faint stars turn green (easy to deal with)

- Red sensitivity in the Ha region is much improved

- The red channel is rather noisy in the shadow areas of the image, require a separate fix

Test images with a 106/530 mm f/5 Takahashi astrograph and the A7III show:

- A little less Ha sensitivity than modded or dedicated astro cameras

- Stars down to fainter than 19 mag in a 64x30 sec stacked image

The fun thing is that I did NOT record fainter stars with a similar total exposure when using a dedicated thermoelectrically cooled astronomical camera - most likely turbulence during the longer 16x4 min sub CCD exposures blurred the very faintest stars into oblivion.

Yes - the star eater issue is there.

No - the star eater issue do not affect ordinary photography and almost none will ever notice any degrading on starscapes and Milky way images - even when pixel peeping.

However: if the star eater is an issue seems to be a personal matter depending upon how your own work is affected. Personally the A7III is now included in my astronomy toolbox.

Trollmannx said:

Just a side note:

- Red sensitivity in the Ha region is much improved

Click to expand...

Is this due to firmware update, or are you comparing the A7III with previous Sony cameras?

Tristimulus said:

Yes - the star eater issue is there.

No - the star eater issue do not affect ordinary photography and almost none will ever notice any degrading on starscapes and Milky way images - even when pixel peeping.

Click to expand...

Great with supplementing "Star Eater" info on the A7III too!

Magnar W said:

Trollmannx said:

Just a side note:

- Red sensitivity in the Ha region is much improved

Click to expand...

Is this due to firmware update, or are you comparing the A7III with previous Sony cameras?

Member said:

Yes - the star eater issue is there.

No - the star eater issue do not affect ordinary photography and almost none will ever notice any degrading on starscapes and Milky way images - even when pixel peeping.

Click to expand...

Great with supplementing "Star Eater" info on the A7III too!

Click to expand...

Just adding the pudding:

If this kind of astrophoto satisfies then the A7III will do.

North America and Pelican nebulae.


Sony A7III + 106/530mm f/5 astrograph. 64x30sec stacked exposure at ISO800.

Magnar W said:

RicksAstro said:

Magnar W said:

RicksAstro said:

These look a tiny bit out of focus (or seeing was bad). So I wouldn't expect to see much difference since the stars aren't pinpoints

Click to expand...

Focus is as good as possible with these lenses and the A7rIII loupe finder at highest magnification, and seeing was about average. I might squeeze a bit more out of these lenses a dry, cloudless night during the spring here up north.

Still, the results are good enough to say that the "star eater" isn't an issue for night sky photographers and even acceptable for much aesthetic deep sky photography.

Click to expand...

Well, it's just that the stars are all greater than a pixel or 2 across. So star eater wouldn't apply. Not saying your conclusion is wrong, just that it's inconclusive.

Click to expand...

If there were 1 pixel stars, hot pixel removal would have taken them away. What would such stars add to aesthetical images, except for looking like hot pixels?

Also, when do you get atmospheric conditions that allow you to register a star at one single pixel with a 42 Mp camera? With longer focal lengths the light from faint stars will be spread over more pixels anyway.

I think there is a technical part with spatial filtering, that has to be judged against the cleanness of shadow areas for long exposure photography in general. And there is an aesthetical part, where "star eater" at this level should not be a concern.

Camera technique, like ultra precise focusing and staying within the time span so that stars are not stretched when using a tripod, and also post processing skills and eventual use of specialized software for editing astrophotography, with their set of special filters, would probably have more impact on the end result than the A7rIII in-camera spatial filtering alone.

If I was going to work at the level you are indicating here, I would go for no less than a cooled monochrome astro camera with filter wheel, and software like MaxIm DL to handle dark, flats, bias frames, etc.

Click to expand...

In your previous thread, there were clear demonstrations as to the potential destructive nature of the algorithm in certain conditions. And, like the other thread, the initial images wouldn't lend themselves to beaing "eaten", thus provide little in terms of a conclusion.

It's NOT just single pixels, please read Jim Kasson's page here for a better description than I could ever provide:


As I've said many times, I don't disagree with Sony on setting this spatial filtering on by default. It's best for the vast majority of the use cases of the camera. But allowing it to be turned off or tuning it like Nikon had done would make it so much more usable for Astro.

I've had my old 10D and Rebel astro-photographs on covers of books and featured on Nasa's APOD, mainly because I followed the same techniques with DSLR as I did with Cooled cameras. It was a great challenge. My advise to everyone is to treat your captured data as gold. Never execute a process that destroys good data, and refine the "desired" data and minimize "noise" data . Repeatable noise data can be removed with various careful techniques. Random noise data can only be minimized by overcoming it with good data.

This algorithm breaks that golden rule. I used hot pixel removal techniques as well, but I mapped the pixels using many averaged dark exposures, looking for inhenently hot or dead pixels and using SW to "remove" them. This is similar to what Sony does every month or so, taking a few seconds on power down to map out "bad" pixels.

Trollmannx said:

Magnar W said:

Trollmannx said:

Just a side note:

- Red sensitivity in the Ha region is much improved

Click to expand...

Is this due to firmware update, or are you comparing the A7III with previous Sony cameras?

Tristimulus said:

Yes - the star eater issue is there.

No - the star eater issue do not affect ordinary photography and almost none will ever notice any degrading on starscapes and Milky way images - even when pixel peeping.

Click to expand...

Great with supplementing "Star Eater" info on the A7III too!

Click to expand...

Just adding the pudding:

If this kind of astrophoto satisfies then the A7III will do.

North America and Pelican nebulae.


Sony A7III + 106/530mm f/5 astrograph. 64x30sec stacked exposure at ISO800.

Click to expand...

Beautiful image. The nebulae really pop compared to most dedicated setups aside from those using narrowband Ha filters, especially for only 30 minutes!

Love the 106 Tak...a superb Astrograph!

--
Rick Krejci

Thank you for kind words!

Taken with a stock A7III - especially satisfied with capturing the red parts and the blue parts of the nebulae like this - some H-beta in there too, usually lost with modded or dedicated one shot color astro cameras.

Also note the bright yellow and blue star, nicely differenciated.

In my case recording this color palette is cherished and of much more importance than mulling over some residual star eater issue - yes it still is there - but absolutely no problem in my case. Happy to include the A7III in my astronomical toolbox.

After all the A7III and A7RIII are camerar initially made for ordinary photography.

RicksAstro said:

In your previous thread, there were clear demonstrations as to the potential destructive nature of the algorithm in certain conditions. And, like the other thread, the initial images wouldn't lend themselves to beaing "eaten", thus provide little in terms of a conclusion.

Click to expand...

In the tread from my initial firmware 2 post, I also posted quick and dirty sample taken with the 21 mm lens. I was in doubt of this sample, as bad conditions very well could have caused the differences between my 3.2 and 4 sec exposures.

I am way more confident in this second test, under decent but not perfect conditions. I took several series with the 21 mm lens, and also the 35 mm lens, refocused for new areas of the sky, and all frames show the same result as the one published here. Same star spread, same magnitude limit, same sky background structure.

RicksAstro said:

It's NOT just single pixels, please read Jim Kasson's page here for a better description than I could ever provide:

Click to expand...

I know. I was referring to you mentioning "Well, it's just that the stars are all greater than a pixel or 2 across".

RicksAstro said:

My advise to everyone is to treat your captured data as gold. Never execute a process that destroys good data, and refine the "desired" data and minimize "noise" data.

Click to expand...

I agree, but there are also practical limits for what you can do, and what your gear are capable of capturing. My take is probably some more pragmatic than yours, so I accept that the A7rIII is what it is. So I will use it for lower end deep sky photography. I also have a SBIG cooled astro camera, but using this is impractical when away from home.

RicksAstro said:

This algorithm breaks that golden rule. I used hot pixel removal techniques as well, but I mapped the pixels using many averaged dark exposures, looking for inhenently hot or dead pixels and using SW to "remove" them. This is similar to what Sony does every month or so, taking a few seconds on power down to map out "bad" pixels.

Click to expand...

Then use a more suited camera for astrophotography. ;-)

Member said:

I would have expected more visible differences, especially with the 21 mm lens.

Click to expand...

I'm surprised why you expected any differences at all. I can see that the stars on the bottom right image are elongated (very tiny star trails). In any situation with visible star trails, star eater would not engage, be it old firmware or new firmware.