It was more of a testing weekend. The wells on this camera are relatively deep, so I’ll definitely be trying longer exposure. That said, there is a trade off with CMOS between exposure duration and read noise that means you do well with relatively short exposures compared to CCD - I don’t fully understand that, a bit like my lack of understanding about gain, offset and readout mode. Many “knobs to twiddle” and much to learn!
DT
Hi David,
I spent a bit of time investigating CMOS cameras
over the last few years but didn't get very far.
In your case it would have been interesting to take 4 of your 5 minute Ha frames
and stack them and stretch them -
then compare with say one 20 minute Ha frame.
I believe you're right and the CMOS cameras have too high a
I find it weird that you can even vary these parameters let alone have insight to how they affect data gathering. If I have a Eureka moment I'll let you know
BTW I have found that QHY's stated back focus and thickness dimensions are rubbish. AP's Quad compressor has a +/- 1.00mm tolerance, which I found wasn't being met when I actually measured everything with a micrometer. I expect my AP spacer set is now correct but have yet to test it on the sky.....
Looking forward to your enlightenment!
Not impressed about the backfocus distances being out. Any hints as to where they’re wrong. Are you using the QHY filter wheel?
Hi David,
I spent a bit of time investigating CMOS cameras
over the last few years but didn't get very far.
In your case it would have been interesting to take 4 of your 5 minute Ha frames
and stack them and stretch them -
then compare with say one 20 minute Ha frame.
I believe you're right and the CMOS cameras have too high a
CMOS cameras have way less read noise compared to CCD. For example at high gain mode the 600 has read noise of around 1.2 electron whereas a good quality 16803 sensor is 9 electrons.
Dark current is also typically lower.
Think of gain as ISO and offset as the black point. Higher ISO on a digital camera means less dynamic range but the amount it drops gets complicated as these Sony sensors often have a 2 step gain where the read noise drops significantly at a certain gain(ISO).
High gain reduces well depth but lowers noise and also weakens Dynamic range.
Too small wells at high gain and you'll blow out your stars and get these all white star type images with no star colour.
The reason shorter exposures works is because the read noise is a fraction of CCDs so you get your signal above the read noise much sooner than you will with CCD.
If you want to do long exposures then you may want to use the mode that has the deepest wells (ie. best dynamic range).
I did 10 minute exposures in narrowband no problem with myASI183mm at a relatively high gain (111). I use 10 minutes with my CCDs as well. But I did start using 5 minute subs for narrowband as it seemed the sweet spot.
QHY gain and offset numbers seem to be different to ASI numbers so the settings aren't interchangeable.
Greg.
Last edited by gregbradley; 23-11-2020 at 03:47 PM.
CMOS cameras have way less read noise compared to CCD. For example at high gain mode the 600 has read noise of around 1.2 electron whereas a good quality 16803 sensor is 9 electrons.
Dark current is also typically lower.
Think of gain as ISO and offset as the black point. Higher ISO on a digital camera means less dynamic range but the amount it drops gets complicated as these Sony sensors often have a 2 step gain where the read noise drops significantly at a certain gain(ISO).
High gain reduces well depth but lowers noise and also weakens Dynamic range.
Too small wells at high gain and you'll blow out your stars and get these all white star type images with no star colour.
The reason shorter exposures works is because the read noise is a fraction of CCDs so you get your signal above the read noise much sooner than you will with CMOS.
If you want to do long exposures then you may want to use the mode that has the deepest wells (ie. best dynamic range).
I did 10 minute exposures in narrowband no problem with myASI183mm at a relatively high gain (111). I use 10 minutes with my CCDs as well. But I did start using 5 minute subs for narrowband as it seemed the sweet spot.
QHY gain and offset numbers seem to be different to ASI numbers so the settings aren't interchangeable.
Greg.
Hi Greg,
"I think you meant;
The reason shorter exposures works with CMOS is because the read noise is a fraction of CCDs so
you get your signal above the read noise much sooner than you will with CCD. "
Also - if the read noise is well below the sky noise from
a suburban location then I don't see how a long 20 minute subframe
would make any difference using CMOS?
Hi Greg,
"I think you meant;
The reason shorter exposures works with CMOS is because the read noise is a fraction of CCDs so
you get your signal above the read noise much sooner than you will with CCD. "
Also - if the read noise is well below the sky noise from
a suburban location then I don't see how a long 20 minute subframe
would make any difference using CMOS?
cheers
Allan
Yes quite right.
Well you probably could but I think you'll find that often bright areas will blow out due to the super high QE of these sensors.
The run of inclement weather in Sydney has made it difficult for me to explore the QHY600M's capabilities further, but I am coming to the conclusion the intrinsic noise of its CMOS sensor is about an order of magnitude higher than my KAF16803.
Read noise looks to be a red herring. It is simply the consistency of the signal your read-up, and says nothing about the pixel to pixel variations.
Problem is CMOS have outliers that while consistent in value, are spread over a much wider bell curve....
The run of inclement weather in Sydney has made it difficult for me to explore the QHY600M's capabilities further, but I am coming to the conclusion the intrinsic noise of its CMOS sensor is about an order of magnitude higher than my KAF16803.
Read noise looks to be a red herring. It is simply the consistency of the signal your read-up, and says nothing about the pixel to pixel variations.
Problem is CMOS have outliers that while consistent in value, are spread over a much wider bell curve....
Well, swings and roundabouts. I have yet to use this sensor on my RC16, where the sampling would favour larger pixels. The big difference between this chip and the KAF16803 is the noise...note that the pixel values of the median combined CCD vary just 8 ADU, the CMOS however has a bell curve well over 100 ADU wide. A bit like 1600 grit sandpaper vs 80 grit...the CCD gives you a much smoother finish...
Well, swings and roundabouts. I have yet to use this sensor on my RC16, where the sampling would favour larger pixels. The big difference between this chip and the KAF16803 is the noise...note that the pixel values of the median combined CCD vary just 8 ADU, the CMOS however has a bell curve well over 100 ADU wide. A bit like 1600 grit sandpaper vs 80 grit...the CCD gives you a much smoother finish...
Hi Peter,
you are getting a very narrow range of ADUs and
also a very low value too.
I'm not getting that with my KAF8300 CCD.
See the pic below.
From a dark frame I get a range of values from 898 to 1149
which is 251 ADUs wide.
You are getting only 8 ADUs wide with your KAF16803 CCD.
You are also way down at an average 100 ADUs out of a possible 65536
whereas I'm at 1019 ADUs.
You're an order of magnitude better in base noise.
Hi Peter,
you are getting a very narrow range of ADUs and
also a very low value too.
I'm not getting that with my KAF8300 CCD.
See the pic below.
From a dark frame I get a range of values from 898 to 1149
which is 251 ADUs wide.
You are getting only 8 ADUs wide with your KAF16803 CCD.
You are also way down at an average 100 ADUs out of a possible 65536
whereas I'm at 1019 ADUs.
You're an order of magnitude better in base noise.
What's going on here?
cheers
Allan
As I mentioned on the web page, the ADU levels shown are from 12 median combined, dark calibrated darks taken from both sensors (if you just used uncalibrated darks, all you'd be looking at is thermal noise).
I used ALL of the KAF16803 pixels for the upper graph, but only a 190k pixel sample from the CMOS....not entirely valid sampling I know, but when I increased the sample size for the CMOS the bell curve actually got wider.
I am not aware of any systemic error I would have made, but if there has been a blunder on my part, please let me know.
P.S. You can download my KAF16803 median combined dark calibrated dark here (right click, save as)
for completeness the CMOS calibrated dark is here right click, save as.. careful it's 122Mb
Last edited by Peter Ward; 24-11-2020 at 02:35 PM.
As I mentioned on the web page, the ADU levels shown are from 12 median combined, dark calibrated darks taken from both sensors (if you just used uncalibrated darks, all you'd be looking at is thermal noise).
I used ALL of the KAF16803 pixels for the upper graph, but only a 190k pixel sample from the CMOS....not entirely valid sampling I know, but when I increased the sample size for the CMOS the bell curve actually got wider.
I am not aware of any systemic error I would have made, but if there has been a blunder on my part, please let me know.
P.S. You can download my KAF16803 median combined dark calibrated dark here (right click, save as)
OK thanks Peter,
I think it's because you're using 12 median combined, dark calibrated darks -
they are processed - not the RAW data coming straight off the sensor.
The median combine is smoothing out your real values.
OK thanks Peter,
I think it's because you're using 12 median combined, dark calibrated darks -
they are processed - not the RAW data coming straight off the sensor.
The median combine is smoothing out your real values.
cheers
Allan
The rub is: I did the same with the CMOS and got a spread of 100ADU
The goal was to see how consistent calibrated data would be.
The rub is: I did the same with the CMOS and got a spread of 100ADU
The goal was to see how consistent calibrated data would be.
Hi Peter,
that's true.
I think the reason that my average value is 1019 ADUs and yours is at 100 ADUs
is the offset in my camera.
I set it to 110 which I don't quite understand but it has probably placed
the average where it is. ( I copied a known good value)
see here: http://www.stark-labs.com/help/blog/...nAndOffset.php
By the way - my gain is set to 10 which is supposed to mean 10%
but that's not true either as my maximum peak is at 54,570 ADUs
(not 2 to the 16) = 65536.
I find it weird that you can even vary these parameters let alone have insight to how they affect data gathering. If I have a Eureka moment I'll let you know 
