".... I've just carried out what I regard as a reasonably robust quantification of the effect of Sony's star-eater algorithm on an A7RII, as follows:
1. Take two sets of ten images of the same set of the Milky Way, exposed a minute or so apart, and with the first set exposed using 3.2 second exposures (no star-eater) and the second set exposed using 4 second exposures (with star-eater);
2. Separately load each set of images into DeepSkyStacker as raw files, setting the raw-read brightness setting in DSS to compensate for the difference in exposure, i.e., to 4.0 for the 4 second exposures, and to 4.8 for the 3.2 second exposures. That is the 20% increase in read brightness for the 3.2 second exposures compensates for their 20% shorter exposure compared to the 4 second exposures.
3. Set the star detection threshold to 90%, register each set, and calculate the average number of stars detected.
4. The results:
- average number of stars for the 3.2 second set - 640.3;
- average number of stars for the 4.0 second set - 472.8;
5. The conclusion: Sony star-eater reduces the number of detectable stars in this equivalent images by 167 or by about 26%. ..."
p.7 #5 · Star Eater Algorithm and getting in contact with Sony
MJKoski wrote:
Well arska, the resolution loss with small details is more than that, lower than 24MP. Sharp edges may be contained but surface texture lost for ever.
I agree. And this was my point too. The loss of stars may be relatively less than the actual resolution/details loss.... But this is not so unambiguous thing. It is very difficult to determine what is the real effect of the algorithm, for example to resolution. It depends on what kind of surfaces/texture the scene that is being photographed, contains. Star-sky is only one kind of scene.
But the most important point is: We should not even have to think about this issue. This is Sony's mistake, not ours! Sony should fix it without our hassle.
p.7 #6 · Star Eater Algorithm and getting in contact with Sony
thrice wrote:
@mreynolds767@ Your stars look a lot bigger than a single pixel. The noise profile changes noticeably in your 'bluer' sky images between 3.2s and 4.0s.
But do you see even one less star
I don't and these are 100% crops
I personally think my tests show as others in the past have that this IS a 3.3 firmware issue and not previous to that.
May help those trying to perform a downgrade until Sony gets their act together and comes up with a firmware update to correct their oversight.
I hope they will also update so bulb is not affected, but fairly certain that will not happen and their fix will just make the current standard the same as I have it on 3.1 firmware
p.7 #7 · Star Eater Algorithm and getting in contact with Sony
thrice wrote:
Sorry everyone, despite rolling back the firmware I can confirm spatial filtering still occurs at 4.0s.
Probably a microcode block that the 3.20 firmware didn't write whereas the 3.30 firmware does.
Back to waiting for Sony.
thrice, I think that your previous speculation was quite right. Modified v.3.20 firmware does not write over that "microcode block ", or something similar... What a shame indeed.
It can't be a complicated piece of code, just drive star eater algorithm to work at 4sec and longer exposures, as it previously worked only in BULB mode. What was a much better way to work, BTW.
p.7 #8 · Star Eater Algorithm and getting in contact with Sony
mreynolds767 wrote:
But do you see even one less star
Yeah quite a few of them on the periphery where they're reduced to nearly 1px are greatly diminished.
This isn't an algorithm that says “oh you're a star I'm going to eat you”. It says here is a single pixel of noticeably increased brightness vs surrounding pixels, I'm going to map it out.
Because your stars are defocused to increase their size they are mostly left alone.
I'm not sure what the agenda of pushing that firmware 2.0->3.2 doesn't eat stars is.
Perhaps maintaining value of those bodies to allow higher resale value.
We already have very visible and quantifiable evidence that 2.0+ eats stars.
End of story.
".... I've just carried out what I regard as a reasonably robust quantification of the effect of Sony's star-eater algorithm on an A7RII, as follows:
1. Take two sets of ten images of the same set of the Milky Way, exposed a minute or so apart, and with the first set exposed using 3.2 second exposures (no star-eater) and the second set exposed using 4 second exposures (with star-eater);
2. Separately load each set of images into DeepSkyStacker as raw files, setting the raw-read brightness setting in DSS to compensate for the difference in exposure, i.e., to 4.0 for the 4 second exposures, and to 4.8 for the 3.2 second exposures. That is the 20% increase in read brightness for the 3.2 second exposures compensates for their 20% shorter exposure compared to the 4 second exposures.
3. Set the star detection threshold to 90%, register each set, and calculate the average number of stars detected.
4. The results:
- average number of stars for the 3.2 second set - 640.3;
- average number of stars for the 4.0 second set - 472.8;
5. The conclusion: Sony star-eater reduces the number of detectable stars in this equivalent images by 167 or by about 26%. ..."
For those interested, I've just redone this test with DeepSkystacker on two sets of ten images, one exposed at 3.2 seconds, and the second exposed at 4 seconds - however, this time I dropped the detection threshold in DSS to 10% to allow detection of stars with wide variation in brightness, i.e., not just the bright stars as previously. Results are as follows:
1. for the 3.2 second images DSS detected on average 21,264 stars;
2. for the 4 second images DSS detected on average 12,061 stars;
Conclusion: DSS on average detects 9200 less stars in the 4 second exposures than in the 3.2 second exposures, despite their longer exposure. This is a reduction of 43% or getting close to half!
p.7 #10 · Star Eater Algorithm and getting in contact with Sony
thrice wrote:
Yeah quite a few of them on the periphery where they're reduced to nearly 1px are greatly diminished.
This isn't an algorithm that says “oh you're a star I'm going to eat you”. It says here is a single pixel of noticeably increased brightness vs surrounding pixels, I'm going to map it out.
Because your stars are defocused to increase their size they are mostly left alone.
I'm not sure what the agenda of pushing that firmware 2.0->3.2 doesn't eat stars is.
Perhaps maintaining value of those bodies to allow higher resale value.
We already have very visible and quantifiable evidence that 2.0+ eats stars.
End of story....Show more →
I have no agenda and my camera is not for sale.
Excuse me for being an owner and doing some testing and not believing everything you read on the Internet.
I personally can not see one less star nor anything that bothers me in the slightest between my 3.2 and 4.0 second exposures and again these are 100% crops so as to what would be seen on screen or in print there is no difference to my eyes at all.
These means one of two things:
1.) It was a firmware update after 3.1 and you are incorrect about that
2.) I am too blind or too dumb to see it which I honestly hope is the case because then Star Eater does not affect me and I could care less if Sony fixes it or not. If #2 is your answer I am blissfully ignorant to the issue and encourage all other owners to take similar comparisons and see if this issue affects them at all as well.
Personally I still believe #1 is the case based on other example tests others have presented which show what appear to me much different results than my tests.
I encourage others to look at my examples and see if you can see any difference at all in the quality between 3.2 and 4.0 secs.
The first set was with a 135mm lens, the 2nd with a 21mm lens. First set of 3.2 vs. 4.0 were on a tracker , 2nd set of 3.2 vs. 4.0 sec. were not on a tracker.
I am not fighting with you just trying to have a real conversation about the true effects of this noise reduction you and others are complaining about.
I have yet to see a serious test before of pre-3.3 update of the Star Eater. Have seen a few pics posted here and there where people complain about it but nothing I would call a test. until recently it was accepted that it was a 3.3 update so just try to get facts straight so the pitch to Sony to correct this is accurate and not Internet hyperbole.
p.7 #11 · Star Eater Algorithm and getting in contact with Sony
mreynolds767 wrote:
I have no agenda and my camera is not for sale.
Excuse me for being an owner and doing some testing and not believing everything you read on the Internet.
I personally can not see one less star nor anything that bothers me in the slightest between my 3.2 and 4.0 second exposures and again these are 100% crops so as to what would be seen on screen or in print there is no difference to my eyes at all.
These means one of two things:
1.) It was a firmware update after 3.1 and you are incorrect about that
2.) I am too blind or too dumb to see it which I honestly hope is the case because then Star Eater does not affect me and I could care less if Sony fixes it or not. If #2 is your answer I am blissfully ignorant to the issue and encourage all other owners to take similar comparisons and see if this issue affects them at all as well.
Personally I still believe #1 is the case based on other example tests others have presented which show what appear to me much different results than my tests.
I encourage others to look at my examples and see if you can see any difference at all in the quality between 3.2 and 4.0 secs.
The first set was with a 135mm lens, the 2nd with a 21mm lens. First set of 3.2 vs. 4.0 were on a tracker , 2nd set of 3.2 vs. 4.0 sec. were not on a tracker.
I am not fighting with you just trying to have a real conversation about the true effects of this noise reduction you and others are complaining about.
I have yet to see a serious test before of pre-3.3 update of the Star Eater. Have seen a few pics posted here and there where people complain about it but nothing I would call a test. until recently it was accepted that it was a 3.3 update so just try to get facts straight so the pitch to Sony to correct this is accurate and not Internet hyperbole. ...Show more →
Even if you want to argue on a print you can't see it (and I would argue that depends on print size), do you really condone Sony a. changing how a camera works after it's out and b. manipulating image quality in RAW files? I can't imagine you do. And if you don't condone it, then we are all on the same side.
p.7 #12 · Star Eater Algorithm and getting in contact with Sony
Ugh? It was the update in 2015 October which ruined the game. As Mr. Grum demonstrated earlier. And calling this "Star Eater" is kind of wrong. It smudges every photo with high amount of small details. That is what spatial filtering does.
Defocusing star field even by a small tiny bit will drop its clarity drastically and causes sharp points to flatten out to small blobs. Running a filter over such blob returns whatever ~average value the blob has and nothing vanishes.
It is unknown whether this filter was even powered up more at some later point.
p.7 #13 · Star Eater Algorithm and getting in contact with Sony
I am on the same side and want Sony to correct this, no doubt.
If I am shooting RAW with LENR turned off it should be OFF ; no debate what so ever from me on that point.
I already have in camera options to turn on and off other types of noise reduction to add one automatically is not acceptable.
I would like to clarify exactly at what point Sony implemented it so Sony also realizes the ability to turn this off should be for all images, not just over 4.0 secs ; it should not be there on Bulb mode either, which affects me no matter the firmware my camera is on.
p.7 #15 · Star Eater Algorithm and getting in contact with Sony
I am a Canon shooter. I don't shoot with a Sony…yet. The a9 has caught my attention because I am a wildlife photographer. But I also do some night photography, and this star-eater filter that erases dimmer stars in exposures over 3.2 seconds is pretty much a show stopper for me. I hope that Sony responds to its current customers with a fix for this issue, and will eliminate it for all future Sony bodies. And by the way, why on earth did Sony not include an intervalometer and GPS in the a9?
p.7 #18 · Star Eater Algorithm and getting in contact with Sony
I see clear smudging of detail in that example, yes.
Not sure the magnification that example is using but my rounder stars are not showing that smudge effect which is really what this Star Eater is.
I am really not seeing the difference, especially with the shot on a tracker (the darker sky color images) I can see no difference when it should clearly be there based on what most have said so it has me confused if nothing else.
p.7 #20 · Star Eater Algorithm and getting in contact with Sony
I really hope Sony fixes this... I have a couple A7r's and I'd love to upgrade to the A7rII. But I shot a lot of astro and so will hold off until then.
