Revisited Milky Way Nightscape + MFT capability considerations

tjl66 wrote:

Interceptor121 wrote:

tjl66 wrote:

tradesmith45 wrote:

tjl66 wrote:

Interceptor121

Interceptor & tji66 are both right - PDAF pixels do impact image noise BUT in many cases the impact is too small to be visible.

Agree, but It is being made out to be an issue, when it is in fact a non-issue. Particularly with milky way images which are not particularly demanding.

again totally incorrect. As long lenses are guided and have bigger physical aperture deep space is actually less demanding on a small sensor than wide field with short focal

My actual experience is totally different. You seem quite sure of yourself, can you point me to the work you've done at longer focal lengths to draw your conclusions from?

For a start you will be working at Focal Ratios about 2-3 stops slower so any FPN (Fixed Pattern Noise) will also be boosted 2-3 stops for the equivalent image brightness. Furthermore, when guiding the FPN will be aligned from one image to the next, and so when you stack them the FPN patterns will reinforce. It is why people use autoguider dithering to prevent this from happening.

You made a statement about milky way images not being demanding which is broad generic and has nothing to do with fixed pattern noise. Milky way image dues to shorter focal and smaller physical aperture are more challenging

Second fixed pattern noise has nothing to with what lens you use or the fstop as it is a sensor pixel defect the lens does not impact it

Dedicated astrophotography cameras with MFT sensors are commercially available obviously they are pure contrast detection. A guiding system will use dithering this eliminates any need for dark frames to correct fixed pattern

This kind of guiding is not something you can do on a portable unguided tracker so generally landscape are much harder to process.

The important concern is visible FPN. No NR software can remove this problem. But not all cams w/ on sensor PDAF have visible FPN & not having PDAF is not guarantee the camera will not have visible FPN.

Absolutely, there are lot more sources for FPN than PDAF. Take a series of Bias frames average them and stretch them as much as possible and you will see all manner of fixed structures - on any sensor. Likewise do that with a flat field.

Fixed pattern noise is driven by defective pixels

defective pixels are those that do not produce an image

PDAF pixels do not produce an image so by definition they are defective pixel in fpn context that are arranged in row and therefore create patterns more than random hot pixels or cold pixels would do

this requires you to take dark frames for optimal results in all conditions which in turn means time and less optimal results

FPN comes mostly from small deviations in pixel black levels and small variations in sensitivity. This is what dark and flat field calibration are designed to address. It does not mean these pixels don't produce an image.

The pixels do not produce an image they are dedicated to phase difference the imaging is created by interpolation of neighbouring pixels this can create problems and ultimately banding

The previous poster makes a good point about sensitivity of cameras to fixed pattern noise. Those are two stacked frames one for the EM1MKII and one for the GH5M2

EM1MKII

GH5M2

Pen-F

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

We can go on with this for a long time however one of the most important criteria for a camera with a small sensor to do nightscapes (not just Milky Way) is to have good blacks and good uniformity so you know that you are unlikely to run into problems

The OM-1 with the lack of a fixed PDAF mask is more interesting than past models and has better metrics however it looks very easy to clip the highlights

Interceptor121 wrote:

EM1MKII

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

.The E-M1ii PDAF pattern doesn't induce horizontal stripes, if you're seeing them then the cause is something else. The desire to have cross-type points resulted in Olympus using a 16x16 pattern for the masked points that gives PDAF artifacts that are very different from what you see in say Sony or Nikon cameras.

whumber wrote:

Interceptor121 wrote:

EM1MKII

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

.The E-M1ii PDAF pattern doesn't induce horizontal stripes, if you're seeing them then the cause is something else. The desire to have cross-type points resulted in Olympus using a 16x16 pattern for the masked points that gives PDAF artifacts that are very different from what you see in say Sony or Nikon cameras.

There is something going on for sure as the behaviour changes if you are at low or high gain. For that particular image the frame is at ISO 2000 and the patterns are horizontal however if you look at low gain the patterns are vertical

The same happens for the EM1X when it works at low gain the patterns are vertical then they become horizontal

The EM1MKIII instead as a situation where the patterns are vertical at low gain and then at high gain they almost disappear

I have no idea why but what I know is that I do not like this variable behaviour

The Panasonic cameras that I have owned instead (with the exception of the GH6 which is weird) have the same behaviour regardless of the ISO with some mild variation of black level

I prefer to have cameras that have the same behaviour at any ISO and that keep black levels at bay this is my preference if someone else wants to do else that is fine by me but this is my reasoning and direct experience (I have used the EM1MKII for a period of time before i started looking into this and in addition to poor ergonomics on the field I was faced with patterns)

I am conscious that when someone says something that is not positive about any piece of Olympus equipment it must go under scrutiny so here we go with hijacking the thread that was focussed on how to take good images and how I do it

Interceptor121 wrote:

whumber wrote:

Interceptor121 wrote:

EM1MKII

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

.The E-M1ii PDAF pattern doesn't induce horizontal stripes, if you're seeing them then the cause is something else. The desire to have cross-type points resulted in Olympus using a 16x16 pattern for the masked points that gives PDAF artifacts that are very different from what you see in say Sony or Nikon cameras.

There is something going on for sure as the behaviour changes if you are at low or high gain. For that particular image the frame is at ISO 2000 and the patterns are horizontal however if you look at low gain the patterns are vertical

The same happens for the EM1X when it works at low gain the patterns are vertical then they become horizontal

The EM1MKIII instead as a situation where the patterns are vertical at low gain and then at high gain they almost disappear

I have no idea why but what I know is that I do not like this variable behaviour

The Panasonic cameras that I have owned instead (with the exception of the GH6 which is weird) have the same behaviour regardless of the ISO with some mild variation of black level

I prefer to have cameras that have the same behaviour at any ISO and that keep black levels at bay this is my preference if someone else wants to do else that is fine by me but this is my reasoning and direct experience (I have used the EM1MKII for a period of time before i started looking into this and in addition to poor ergonomics on the field I was faced with patterns)

I am conscious that when someone says something that is not positive about any piece of Olympus equipment it must go under scrutiny so here we go with hijacking the thread that was focussed on how to take good images and how I do it

The pattern noise contribution from the PDAF masks always looks like this

Courtesy of BClaff

What you're seeing is that as you move towards higher ISO settings you get additional pattern noise that's typical of all sensors due to the minor imperfection in the lithography. It's not visible as lower ISO setting but shows up once you really start pushing things. This has nothing to do with the PD masks, it's just standard CMOS behavior.

whumber wrote:

Interceptor121 wrote:

whumber wrote:

Interceptor121 wrote:

EM1MKII

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

.The E-M1ii PDAF pattern doesn't induce horizontal stripes, if you're seeing them then the cause is something else. The desire to have cross-type points resulted in Olympus using a 16x16 pattern for the masked points that gives PDAF artifacts that are very different from what you see in say Sony or Nikon cameras.

There is something going on for sure as the behaviour changes if you are at low or high gain. For that particular image the frame is at ISO 2000 and the patterns are horizontal however if you look at low gain the patterns are vertical

The same happens for the EM1X when it works at low gain the patterns are vertical then they become horizontal

The EM1MKIII instead as a situation where the patterns are vertical at low gain and then at high gain they almost disappear

I have no idea why but what I know is that I do not like this variable behaviour

The Panasonic cameras that I have owned instead (with the exception of the GH6 which is weird) have the same behaviour regardless of the ISO with some mild variation of black level

I prefer to have cameras that have the same behaviour at any ISO and that keep black levels at bay this is my preference if someone else wants to do else that is fine by me but this is my reasoning and direct experience (I have used the EM1MKII for a period of time before i started looking into this and in addition to poor ergonomics on the field I was faced with patterns)

I am conscious that when someone says something that is not positive about any piece of Olympus equipment it must go under scrutiny so here we go with hijacking the thread that was focussed on how to take good images and how I do it

The pattern noise contribution from the PDAF masks always looks like this

Courtesy of BClaff

What you're seeing is that as you move towards higher ISO settings you get additional pattern noise that's typical of all sensors due to the minor imperfection in the lithography. It's not visible as lower ISO setting but shows up once you really start pushing things. This has nothing to do with the PD masks, it's just standard CMOS behavior.

The cameras I have/had or rented GH5/GH5M2/G9/GH5S do not exhibit any sudden jump between low and high gain worse case some deterioration of black

Is it because they do not have any imperfection or because there is something different compared to Olympus cameras with the same identical sensor? Other than PDAF I cannot think of any such reason that would induce patterns seeing that the sensors are made by the same supplier on the same production lines.

Either way I would not go anywhere near any of those cameras especially as this is not the only issue the other is flare in backlit scene which I had regularly on sunset or sunrise shots

Interceptor121 wrote:

The cameras I have/had or rented GH5/GH5M2/G9/GH5S do not exhibit any sudden jump between low and high gain worse case some deterioration of black

Is it because they do not have any imperfection or because there is something different compared to Olympus cameras with the same identical sensor? Other than PDAF I cannot think of any such reason that would induce patterns seeing that the sensors are made by the same supplier on the same production lines.

Either way I would not go anywhere near any of those cameras especially as this is not the only issue the other is flare in backlit scene which I had regularly on sunset or sunrise shots

Not sure what to tell you. Both my by GH5 and G9 exhibited the same behavior at high ISO settings.

whumber wrote:

Interceptor121 wrote:

The cameras I have/had or rented GH5/GH5M2/G9/GH5S do not exhibit any sudden jump between low and high gain worse case some deterioration of black

Is it because they do not have any imperfection or because there is something different compared to Olympus cameras with the same identical sensor? Other than PDAF I cannot think of any such reason that would induce patterns seeing that the sensors are made by the same supplier on the same production lines.

Either way I would not go anywhere near any of those cameras especially as this is not the only issue the other is flare in backlit scene which I had regularly on sunset or sunrise shots

Not sure what to tell you. Both my by GH5 and G9 exhibited the same behavior at high ISO settings.

Nope and never had a flare on a back-lit shot either. To be honest that is even more apparent as you can see the little squares coming up red

As I said if someone else has some good results and uses other kit well done to them as far as I am concerned I know what gives me the best results after a few years of trying

tjl66 wrote:

Interceptor121 wrote:

tjl66 wrote:

Interceptor121 wrote:

tjl66 wrote:

Interceptor121 wrote:

Sorry I do not need to show anything to you. I have done my diligence and am not going to get back there I have already the best equipment for the job

Great you have the best equipment for your job. But this still doesn't demonstrate the existence of Olympus "PDAF issues".

The issues are with stretching and in some conditions the pixels flare this is a known fact a single exposure will not bring the issue to life

How long the exposures are etc is irrelevant is stacking and stretching that brings out the issues this was already noted long time ago by Bill Claff when the camera came out

I posted a 24 minute stack above, made up of 48 exposures, that has been aggressively stretched. Surely this should show the Olympus "PDAF issues"?

It would help if you stop hijacking my thread what other people do it is not my concern however getting the best performance is and I have sorted it out no need for more investigations

This seems like a very relevant discussion under "MFT capability considerations". I am questioning the statement you made regarding Olympus cameras and issues with PDAF pixels since I cant find any evidence that this is a problem on any of my images.

Images talk by themselves where are yours that look like mine?

No problem, even though most of my astrophotography is at longer focal lengths I had this image taken last year with an em1.2 and 17mm f1.2 lens at f2.0 which is of the same region of sky as yours. Its a 10 stack of 60 second images. I've also reprocessed with ASinh (way more saturation than I normally like) but it gives us a basis of comparison. Being at 17mm f1.2 set at f2.0 has more resolution, but the physical aperture being 8.5mm is similar to the 12mm f1.4. I assume the total exposure is similar.

Is this image with or without dark frame subtraction?

You asked "Images talk by themselves where are yours that look like mine?" So I did. Because I am in the tropics I usually run darks to remove the random scattering of hotpixels - but that has nothing to do with PDAF pixels.

if you can’t see the wood from the trees is difficult to have a conversation

This is what I wrote you may want to read it again

A special mention are PDAF pixels, those show up with the process I use when the image is stressed so if you ever considered using an Olympus camera with PDAF pre OM-1 (there should not be a problem there) you need to take a whole set of flat, bias, dark frames in addition to your exposures otherwise you may end up with posterization issues and other nasty problem. I recommend not using a camera with PDAF and using one with nothing on your sensor. Please note the PDAF issue show up in any situation and are aggravated by thermal effects

Yes read it and still don't agree.

You don't know what my process is hence it is not possible for you to agree or disagree

OK, let me be more specific, everything after that point where you make a recommendation not use an Olympus with PDAF. Since you don't specify a process, we can only assume you are making a general recommendation - a recommendation I disagree with.

You are free not to agree and don't need to do a crusade about it

I have used them and decided never again for this purpose if this works for you happy days

tjl66 wrote:

Interceptor121 wrote:

tjl66 wrote:

Interceptor121 wrote:

tjl66 wrote:

tradesmith45 wrote:

tjl66 wrote:

Interceptor121

Interceptor & tji66 are both right - PDAF pixels do impact image noise BUT in many cases the impact is too small to be visible.

Agree, but It is being made out to be an issue, when it is in fact a non-issue. Particularly with milky way images which are not particularly demanding.

again totally incorrect. As long lenses are guided and have bigger physical aperture deep space is actually less demanding on a small sensor than wide field with short focal

My actual experience is totally different. You seem quite sure of yourself, can you point me to the work you've done at longer focal lengths to draw your conclusions from?

For a start you will be working at Focal Ratios about 2-3 stops slower so any FPN (Fixed Pattern Noise) will also be boosted 2-3 stops for the equivalent image brightness. Furthermore, when guiding the FPN will be aligned from one image to the next, and so when you stack them the FPN patterns will reinforce. It is why people use autoguider dithering to prevent this from happening.

You made a statement about milky way images not being demanding which is broad generic and has nothing to do with fixed pattern noise. Milky way image dues to shorter focal and smaller physical aperture are more challenging

How would you know? You speak with authority on this yet its clear you have no experience in long exposure/long focal length guided astrophotography to draw a comparision with.

Second fixed pattern noise has nothing to with what lens you use or the fstop as it is a sensor pixel defect the lens does not impact it

What happens with the visibility of that noise when you boost the image 2-3 stops?

Dedicated astrophotography cameras with MFT sensors are commercially available obviously they are pure contrast detection. A guiding system will use dithering this eliminates any need for dark frames to correct fixed pattern

This kind of guiding is not something you can do on a portable unguided tracker so generally landscape are much harder to process.

The important concern is visible FPN. No NR software can remove this problem. But not all cams w/ on sensor PDAF have visible FPN & not having PDAF is not guarantee the camera will not have visible FPN.

Absolutely, there are lot more sources for FPN than PDAF. Take a series of Bias frames average them and stretch them as much as possible and you will see all manner of fixed structures - on any sensor. Likewise do that with a flat field.

Fixed pattern noise is driven by defective pixels

defective pixels are those that do not produce an image

PDAF pixels do not produce an image so by definition they are defective pixel in fpn context that are arranged in row and therefore create patterns more than random hot pixels or cold pixels would do

this requires you to take dark frames for optimal results in all conditions which in turn means time and less optimal results

FPN comes mostly from small deviations in pixel black levels and small variations in sensitivity. This is what dark and flat field calibration are designed to address. It does not mean these pixels don't produce an image.

The pixels do not produce an image they are dedicated to phase difference the imaging is created by interpolation of neighbouring pixels this can create problems and ultimately banding

The previous poster makes a good point about sensitivity of cameras to fixed pattern noise. Those are two stacked frames one for the EM1MKII and one for the GH5M2

EM1MKII

GH5M2

Pen-F

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

We can go on with this for a long time however one of the most important criteria for a camera with a small sensor to do nightscapes (not just Milky Way) is to have good blacks and good uniformity so you know that you are unlikely to run into problems

The OM-1 with the lack of a fixed PDAF mask is more interesting than past models and has better metrics however it looks very easy to clip the highlights

That's all great, but how do these impact in actual real astroimages. I've presented a number of images with an em1.2 and you don't seem unable to point out "PDAF issues". Yet I have no problem finding technical issues with your images like chromatic aberation and oversharpened stars with dark rings around them.

Envy? You have not posted a single image worth commeting yours are like test shot with no concept of creativity or else. Who are you to ask anyone to prove anything with that?

I recommend you go and drink some camomile and from now you go into the ignore list

What a total troll!

bclaff wrote:

Interceptor121 wrote:
...

The previous poster makes a good point about sensitivity of cameras to fixed pattern noise. Those are two stacked frames one for the EM1MKII and one for the GH5M2

EM1MKII

GH5M2

Pen-F

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

...

These images appear to be from the PhotonsToPhotos Sensor Heatmaps .

It's very important to view such images at "original size"

You don't indicate which settings you used or provide any links.

Here is a portion of the table on that page:

What stands our are the column metrics rather than the row metrics.

This image (must be viewed "original size") makes it obvious

Regardless of whether the FPN is horizontal or vertical or otherwise it can definitely affect image quality. I don't consider PDAF pixels that show as FPN to be defective but you may consider that semantics.

That's right they are your images. However in some camera the trend changes in high gain and the metric that is high is row. The example I put was ISO 2000 not 200.

The OM-1 does not have such issues as I understand all pixels are used for imaging after focus is achieved and neither the value jumps at ISO 1000

bclaff wrote:

Interceptor121 wrote:

bclaff wrote:

Interceptor121 wrote:
...

The previous poster makes a good point about sensitivity of cameras to fixed pattern noise. Those are two stacked frames one for the EM1MKII and one for the GH5M2

EM1MKII

GH5M2

Pen-F

You can see how the EM1MKII has much more evidence of horizontal stripes than the other two that is because the row metrics are higher due to PDAF

...

These images appear to be from the PhotonsToPhotos Sensor Heatmaps .

It's very important to view such images at "original size"

You don't indicate which settings you used or provide any links.

Here is a portion of the table on that page:

What stands our are the column metrics rather than the row metrics.

This image (must be viewed "original size") makes it obvious

Regardless of whether the FPN is horizontal or vertical or otherwise it can definitely affect image quality. I don't consider PDAF pixels that show as FPN to be defective but you may consider that semantics.

That's right they are your images. However in some camera the trend changes in high gain and the metric that is high is row. The example I put was ISO 2000 not 200.

The OM-1 does not have such issues as I understand all pixels are used for imaging after focus is achieved and neither the value jumps at ISO 1000

Yes, interesting that the High Conversion Gain (HCG) FPN is more horizontal than vertical.

Must view "original size"

If happens for the EM1X row and for the EM5III there is no HCG but I bet it would happen too

Some pixels may change behaviour for autofocus or else

Either way those pixels present problems and you don't need astrophotography to bring them out in fact most of my issues where shooting sunset where the PDAF pixels would flare with certain lenses and show red boxes

tjl66 wrote:

No problem, even though most of my astrophotography is at longer focal lengths I had this image taken last year with an em1.2 and 17mm f1.2 lens at f2.0 which is of the same region of sky as yours. Its a 10 stack of 60 second images. I've also reprocessed with ASinh (way more saturation than I normally like) but it gives us a basis of comparison. Being at 17mm f1.2 set at f2.0 has more resolution, but the physical aperture being 8.5mm is similar to the 12mm f1.4. I assume the total exposure is similar.

tji66 got a question for you.  Did you find similar field curvature in your copy of the M.Z 17mm f1.2 Pro?

Thanks

Interceptor121 wrote:

Either way I would not go anywhere near any of those cameras especially as this is not the only issue the other is flare in backlit scene which I had regularly on sunset or sunrise shots

I'm very curious about this comment.  Can you post an example?

Do you see this defect in this image?

Thanks

Very very impressive !