106 replies to this topic

[sharkmelley]

Continuing with the Sony A7CR, let's examine the potential issues of data scaling and the correction that causes the concentric colour polygons.

 

Firstly here is the RawDigger histogram of one of the flat files:

 

 

There are no spikes and no gaps!  This is excellent news because it means no data scaling is being applied that potentially causes concentric coloured rings.

 

Here are the plots of 3 flats with very different levels of exposure where the (bias-subtracted) red channel is divided by the (bias-subtracted) green channel and the (bias-subtracted) blue channel is divided by the (bias-subtracted) green channel:

 

 

Lens corrections were switched off and lossless data compression was used. There are no hints of the usual coloured concentric polygons that plagued earlier Sony mirrorless cameras - even though the Sony FE 20mm F1.8G lens was recognised by the firmware and the lens correction parameters were written into the EXIF header.

 

So now we have 2 Sony mirrorless cameras that look excellent for deep-sky astrophotography - the Sony A7R5 and the A7CR. The ZV-E1 is also a possibility except that it does not have bulb mode so there's no possibility of doing exposures longer than 30sec.

 

Mark

Edited by sharkmelley, 02 March 2024 - 07:51 AM.

[whwang]

Hi Mark,

 

In the flat B/G images, there is a circle of dark area and bright corners.  In R/G this is opposite and much more subtle but nevertheless it's there.  It's circular rather than polygon.  Do you think this still indicates some raw cooking?  Or it is such a low level effect that it doesn't have any practical impact?

[sharkmelley]

Hi Mark,

 

In the flat B/G images, there is a circle of dark area and bright corners.  In R/G this is opposite and much more subtle but nevertheless it's there.  It's circular rather than polygon.  Do you think this still indicates some raw cooking?  Or it is such a low level effect that it doesn't have any practical impact?

You raise an important point! 

 

What you are seeing there is colour shading i.e. a shift in colour towards the edge of the sensor.  This is almost certainly caused by pixel crosstalk, which is the effect where photons entering green pixels end up producing an electron in the adjacent red or blue and vice versa.  This effect is very dependent on the angle at which light strikes the sensor.  My conjecture is that the raw cooking on earlier cameras was intended to correct for this problem, for instance as described in this paper.  Sony used a polygon method, probably quite similar to what they patented here.

[sharkmelley]

I need to make an apology and a correction regarding the Sony A7R4.  Back in post#24 I showed this image of coloured rings in flats provided by Cloudy Nights member @xonefs:

 

 

I said back then that there is no sign of the dreaded Sony coloured polygons and this is still true.  But I've now determined the cause of the circular coloured rings and it's potentially a serious problem for deep-sky astrophotography.

 

The problem is immediately apparent in the RawDigger histogram of one of those flats:

 

 

Notice the regular histogram gaps.  This is caused by the camera digitally scaling the raw data and it's the direct cause of those rings.

 

 

The Sony A7R4 also damages stars, just like previous Sony mirrorless cameras.

 

I cannot recommend the Sony A7R4 for deep-sky astrophotography because of the damage to stars caused by the raw data filtering and because of the digital scaling that leads to coloured rings.

 

Mark

Edited by sharkmelley, 02 March 2024 - 05:11 AM.

[Digitaliz.se]

I thought astro flats were greyscale images.
I couldnt see gaps in Photoshop.
Are you doing anything extra i RAW digger?
Would like to test myself.

[sharkmelley]

I thought astro flats were greyscale images.
I couldnt see gaps in Photoshop.
Are you doing anything extra i RAW digger?
Would like to test myself.

The histogram you see in Photoshop is the histogram of data that has been digitally developed i.e. a significant amount of processing has been performed which will mix the colour channels and fill in the gaps.  RawDigger is the best app I know for displaying a histogram of the original raw data before any processing is applied.  Download the trial version (which is the full version but expires after the trial period) and display the histogram with Bin 1 resolution - you may need to restrict the data range displayed in order to display it with Bin 1

[sharkmelley]

One other thing, in RawDigger, make sure you use "Linear X-Axis".

Edited by sharkmelley, 03 March 2024 - 05:53 PM.

[astroboyabdi]

This is very good news.

Anyone test the a7cii which is the sony a7iv sensor but with the ai chip etc. If the a7cr is good then perhaps this maybe also good

[xonefs]

I suspect the c ii is good as well, but I don’t see the value for the lower res bodies when the R is not much more

[astroboyabdi]

it was more to double up as bird photography camera with the added bonus of being better in low light for nightscapes. What we need is is an a7v or a1 mark 2, the aiv is good but has star eater and no ai chip making it worse for both of the scenarios.

[Digitaliz.se]

Those that bought/are planning to buy A7R 5 or A7C R for DSO, are you also going to modd them? 

I'm thinking of the cost of buying state of the are daytime photography camera for the same price as a cooled FF astrocamera is crazy. Why not skip all the fuzz and just go all they way.
If I hade the money to buy a Sony A7R 5 for DSO photography, I would definitly buy a ASI6200MC instead. 
But since I am money aware, I went the route of getting A7III, A7R and A7R 4, all bought used and modded. Quite happy with them.

 

/Stefan

[xonefs]

Yes I will absolutely mod it. Mirrorless is for nightscapes and widefield (135mm and less) dso. Astro cams are a pain for this.

If you are looking for something mostly for a scope absolutely just get a zwo 6200.

[Shubham]

Waiting for someone to test A7C2. It was released along with A7CR so (fingers crossed ) should not have star eater or concentric polygon ring problems. Hopefully this is a game up from Sony for astrophotography from now onwards.

 

I’ve been looking at wide field nightscape astro options lately and it seems like Canon has been behind in sensor tech and its RF lens lineup screaming “we don’t care about astrophotography” from their lack of quality lens options and dumb gatekeeping not allowing companies like Sigma to add quality glass to their lineup. I’ve also heard about the disaster it is with their EF to RF converter when trying to use Sigma Arts with their R series mirrorless cameras.

 

So unless someone’s a real Canon fanboy and wants to struggle with all that, I see no point in not going with Sony with their improved non-star-eater cameras. Sony by far has the best lens lineup for astrophotography including the likes of Sigma and others. My first camera will be a Sony. Excited to see what A7S4 would bring to the table.

Edited by Shubham, 19 April 2024 - 10:23 AM.

[V0LDY]

Thanks for the Sony A7CR test files you sent me.  They are very encouraging.  I need to split the plots across 2 posts because of file sizes.

 

Firstly, let's discuss the raw data filtering (a.k.a. "star eater").  Here is my usual diagram of raw (non-bias-subtracted) pixel values plotted against the maximum of neighbouring values found in the 5x5 block surrounding them:

 

SonyA7CR_ISO1600_m24_small.jpg

 

You'll notice how similar the plot is to that of the A7R5 in post#3 and how different it is from many of the plots of various "star-eating" cameras here.

 

The important thing about this plot is the strong vertical arm which represents isolated bright pixels.  We are no longer seeing the very destructive effect where pixels brighter than their neighbours all have their value capped to the maximum of their (same colour) neighbours.

 

Just like the A7R5, there are some residual inter-colour-channel effects in the green and blue leading to those fan type "rays".  However, in the case of the A7R5 it did not lead to star damage and I'm confident that the same will be observed with the Sony A7CR.

 

Mark

Hmm, I've also run my test which is really similar after hearing the issue was solved, and I also got the spikes for the A7RV.
So... they might not impact star colors anymore (is there any field test about it? An head 2 head comparison with the RIV would be great), but what is going on then? Cuz those straight lines in the blue channel still have to come from somewhere.

Does it still mean that it's better not to use dark and bias frames since the data isn't really RAW anyway (even if stars aren't disrupted?).
 

 

[sharkmelley]

Hmm, I've also run my test which is really similar after hearing the issue was solved, and I also got the spikes for the A7RV.
So... they might not impact star colors anymore (is there any field test about it? An head 2 head comparison with the RIV would be great), but what is going on then? Cuz those straight lines in the blue channel still have to come from somewhere.

Does it still mean that it's better not to use dark and bias frames since the data isn't really RAW anyway (even if stars aren't disrupted?).

 

That's a great plot you produced!  I love it when other people produce their own analysis because it helps verify that I'm not making a stupid error.  It also confirms that the effects seen are not limited to a single (possibly defective) camera.

 

I have no explanation for those blue "rays" except that it is obvious there is still some raw-data processing going on.  A raw data filtering algorithm has 2 distinct stages:

• Identify the pixels that need their value capped
• Cap their values

The "rays" tell us about the second part.  The rays tell us that when blue pixels are capped (for whatever reason) they are capped to a value that is approximately the value of the brightest of the immediate blue neighbours or to twice the value (bias subtracted) of the brightest of their immediate blue neighbours.  But we don't know why those particular blue pixels are selected for capping.

 

It's clear that Sony is still not giving us the true raw data but this new filtering algorithm is far less destructive than the previous "star-eater" versions.  I have seen a raw astro-imaging exposure taken with the A7R5 using a very sharp lens and I see no obvious star damage.  By contrast, the star damage from previous Sony cameras was very obvious indeed in the raw file.  Importantly, blue pixels at the core of stars are not systematically affected by the capping of blue pixel values.

 

Using previous "star-eater" Sony cameras, calibration with darks was ineffective because the raw-data filtering was already attacking single pixels brighter than their neighbours i.e. the pixels that cause the fixed pattern noise.  With the Sony A7R5 it is definitely worthwhile to use darks and the impact on the fixed pattern noise is quite dramatic.

 

Mark

[Paulimer]

I do have an A7cii available for testing, but I won't be under the stars any time soon. So I'll just shoot some darks in bulb and ask Mark for help with the test of star eating? Thanks!

 

Paul

[sharkmelley]

I do have an A7cii available for testing, but I won't be under the stars any time soon. So I'll just shoot some darks in bulb and ask Mark for help with the test of star eating? 

Thanks for the A7CII darks. As expected, the resulting plot has exactly the same characteristics as the A7CR and the A7R5:

 

 

This is very good news.

 

Mark

[gregbradley]

If I'm using an FSQ 106 at its native f5, will a Sony A7RV be useable or will severe vignetting occur?

I have had several FSQs and full frame CCDs. Vignetting is minor as the corrected image circle is quite large in an FSQ. It should handle a medium format camera.

 

Greg.

[Shubham]

I have had several FSQs and full frame CCDs. Vignetting is minor as the corrected image circle is quite large in an FSQ. It should handle a medium format camera.

 

Greg.

What do you think about the star shapes in the edges of a FF sensor with small pixels on an FSQ tho Greg?? I’ve read on multiple forums people complaining about the difficulties in making a FF sensor work with the FSQs… Astro-Physics also says they specifically designed their 110GTX to improve the correction to the edges for these newer FF sensors with small pixels.

 

Would like to hear your thoughts on this.

Edited by Shubham, 22 May 2024 - 02:36 AM.

[nhmorgan79]

I'm curious if anyone with an A1 has updated to the newest v2 firmware and found that it is behaving like the newer a7RV and A7Cii

[nhmorgan79]

Mark, I think I followed your testing protocol. If you have a chance, I would love to see if the a1 with the newer 2.0 firmware has resolved the issue. I've created 10 flats at f/2.5 ISO100 using a flat panel and the 135mm GM lens attached. Files are uncompressed RAW with all corrections turned off. I created 5 for the electronic shutter and 5 for the mechanical (efcs) shutter. I then created a 30second dark both the electronic and mechanical shutter, and a 60 second dark using the bulb mode (camera only allows bulb mode in mechanical shutter).

Here is a link to the files:

https://www.dropbox....t=db6dv8xr&dl=0

 

I could do the same for the new A9iii but I doubt anyone is going to try and use that camera for astro with its diminished DR and noise levels.

[sharkmelley]

Here is a link to the files:

https://www.dropbox....t=db6dv8xr&dl=0

Thanks for these files - they're exactly what I need.  There is some good news and some bad news.

 

The good news is that I see no evidence of channel scaling, histogram gaps, concentric polygons and concentric rings.

 

The bad news is that the raw-data filtering ("star eater") is the same as on earlier models:

 

 

As usual, a large number of blue pixels appear to escape the algorithm, which is apparent from the vertical blue arm on the left representing blue pixels that are brighter than their neighbours.  It not clear why this is the case.

 

Mark

[xonefs]

Does anyone have an FX3 to test in photo mode? There is a new firmware supposed to be released for it soon with a lot of features so I do wonder if they would change anything. I expect the original to have star eater since it came out in 2021 but with a major firmware update coming who knows. 

[xonefs]

Also does anyone have an A7 iv with updated firmware to test?

[primeshooter]

So let me get this straight, the star eater issue is gone? Are the green stars gone and the coloured rings?

Edited by primeshooter, 15 August 2024 - 07:59 PM.