#21  
Old 16-10-2016, 09:18 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
How does the stacking process deal with subs of different gains? Different gains yielding vastly different pixel values.
Reply With Quote
  #22  
Old 16-10-2016, 09:28 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Atmos View Post
How does the stacking process deal with subs of different gains? Different gains yielding vastly different pixel values.
definitely not recommended. recently tried stacking subs taken at gain 50, 100 and 200. PI eventually did a reasonable job with average + offset/scaling normalisation, but had to use severe min/max rejection with flux equalisation for the rejection normalisation - even then there are a few just noticeable warmish pixels left in the stack.

best bet is to decide on a gain and stick with it unless there is a very good reason to do otherwise.
Reply With Quote
  #23  
Old 16-10-2016, 09:39 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
I assumed that would be the case, figured it would follow the same concept line as tracking Ha and OIII, sometimes leaves some funky artefacts.

Ultimately I do still believe that per sub dynamic range is important as there are quite a few regions where there is a large dynamic range. Stacking a lot of subs may increase the dynamic range overall BUT you can only work with what each sub provides. This is fine when you have either bright or faint regions but it doesn't work with both ends of the spectrum.

This is one reason the 16803 is still king for wide fields, it can capture both faint and bright areas in a large FOV.
Reply With Quote
  #24  
Old 16-10-2016, 10:38 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
I am leaning the other way Colin - to the point of thinking that per-sub dynamic range is almost a secondary consideration with a low read noise camera - provided the top end is not overloaded (ie, you have to keep the subs short). When you have the signal spread out over 400+ subs, it can be mighty thin in any one sub, but still come up well on stacking - something like the 16803 will have ~25x the information in each sub (16 subs vs 400 for example), but produce the same result on stacking. ie, the 16803 needs to be able to handle vastly more signal in each sub, so it needs much more in-sub dynamic range than the 1600.

that is not to say that the 16803 is anything but a very good chip - it is clearly excellent. But the 1600 makes up for lower in-sub dynamic range by having lots of subs and relying on the magic of stacking to boost the final dynamic range. The 16803 has big pixels and will win out on big scopes or very wide fields, but other than that, I cannot think of any parameter where it will outdo the 1600 after stacking. I know that sounds a bit ridiculous, but I genuinely cannot think of anything.

edit: and the average 16803 camera is probably going to be a bit less touchy re USB download settings than the 1600.

Last edited by Shiraz; 16-10-2016 at 11:00 PM.
Reply With Quote
  #25  
Old 16-10-2016, 10:57 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Maybe my thinking is wrong on this but i see the dynamic range as basically being the amount of well depth above read noise. The 1600 will have considerably shorter subs than the 16803, 2.36x faster in fact at Gain 76 (full 12 bits) but the 16803 has 11x the well depth.

In Lum a 16803 on my scope at my dark site would be sky limited in ~230s which would equate to 97s with the 1600. Shooting any of the bright nebula like M42, M8, Tarantula the 1600 will undoubtedly struggle to cope with the dynamic range where as the 16803 will not likely have much of an issue. The resolution won't be too different either as I can drizzle the 16803 down to pretty much what I get with the QHY22.

With the 1600 if you're doing 97s subs to be sky limited, once any area is saturated, stacking cannot bring that back. This is where I am saying that individual dynamic range becomes important. I know I am using an extreme example as usually some 10s subs are thrown in there for that reason.
Reply With Quote
  #26  
Old 16-10-2016, 11:16 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Atmos View Post
Maybe my thinking is wrong on this but i see the dynamic range as basically being the amount of well depth above read noise. The 1600 will have considerably shorter subs than the 16803, 2.36x faster in fact at Gain 76 (full 12 bits) but the 16803 has 11x the well depth.

In Lum a 16803 on my scope at my dark site would be sky limited in ~230s which would equate to 97s with the 1600. Shooting any of the bright nebula like M42, M8, Tarantula the 1600 will undoubtedly struggle to cope with the dynamic range where as the 16803 will not likely have much of an issue. The resolution won't be too different either as I can drizzle the 16803 down to pretty much what I get with the QHY22.

With the 1600 if you're doing 97s subs to be sky limited, once any area is saturated, stacking cannot bring that back. This is where I am saying that individual dynamic range becomes important. I know I am using an extreme example as usually some 10s subs are thrown in there for that reason.
but won't the 16803 be getting about 6x as many photons/pixel/s as the 1600, so it will still saturate at lower flux?
Reply With Quote
  #27  
Old 17-10-2016, 06:11 AM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Quote:
Originally Posted by Shiraz View Post
but won't the 16803 be getting about 6x as many photons/pixel/s as the 1600, so it will still saturate at lower flux?
For whatever reason that didn't even enter my mind
Reply With Quote
  #28  
Old 17-10-2016, 07:49 AM
Camelopardalis's Avatar
Camelopardalis (Dunk)
Drifting from the pole

Camelopardalis is offline
 
Join Date: Feb 2013
Location: Brisbane
Posts: 5,425
Soooo....sounds to me like low read noise is the great equaliser, with a few caveats
Reply With Quote
  #29  
Old 17-10-2016, 10:49 AM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Quote:
Originally Posted by Shiraz View Post
but won't the 16803 be getting about 6x as many photons/pixel/s as the 1600, so it will still saturate at lower flux?
Not if you are imaging at the same number of arc-seconds per pixel.
Reply With Quote
  #30  
Old 17-10-2016, 01:58 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Camelopardalis View Post
Soooo....sounds to me like low read noise is the great equaliser, with a few caveats
nice summary

Quote:
Originally Posted by Slawomir View Post
Not if you are imaging at the same number of arc-seconds per pixel.
fnumber/aperture also comes into it, but Colin was comparing the cameras on his scope. If you look at the problem more generally, you still get the same sort of answer - unless you have a much faster scope for the 16803. Nothing beats extra photons (although higher QE helps).

Last edited by Shiraz; 17-10-2016 at 04:20 PM.
Reply With Quote
  #31  
Old 18-10-2016, 10:27 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Its hard to argue against the excellent images you have recently posted with your ASI1600.

A picture is worth a 1000 explanations and maths formulas!

Greg.
Reply With Quote
  #32  
Old 18-10-2016, 10:31 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Quote:
Originally Posted by Shiraz View Post
but won't the 16803 be getting about 6x as many photons/pixel/s as the 1600, so it will still saturate at lower flux?

No, its gets the same, just how that is divided up by the number of pixels. It gets a broader area but that's not per pixel but per area of sensor. What counts is what Suavi mentioned, light per pixel which would be more for a larger pixel and larger for a CCD as it does not have 40% of the sensor surface taken up with surrounding CMOS circuitry per pixel. So standard CMOS sensors unless they are BSI (backside illuminated) lose 40% of surface area to surrounding per pixel circuitry.

Later Sony BSI sensors like in the Sony A7r2 are designed to get that 40% back by flipping the sensor over and leaving the circuitry on the other side of the sensor. It started with the Sony RX100 ii or iii.

Greg.

Greg.
Reply With Quote
  #33  
Old 18-10-2016, 10:40 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by gregbradley View Post
No, its gets the same, just how that is divided up by the number of pixels. It gets a broader area but that's not per pixel but per area of sensor. What counts is what Suavi mentioned, light per pixel which would be more for a larger pixel and larger for a CCD as it does not have 40% of the sensor surface taken up with surrounding CMOS circuitry per pixel. So standard CMOS sensors unless they are BSI (backside illuminated) lose 40% of surface area to surrounding per pixel circuitry.

Later Sony BSI sensors like in the Sony A7r2 are designed to get that 40% back by flipping the sensor over and leaving the circuitry on the other side of the sensor. It started with the Sony RX100 ii or iii.

Greg.

Greg.
Nope, pretty sure it will get 6x the number of photons in each pixel because the bigger pixels intercept 6x as many photons. Nothing to do with surrounding circuitry - that determines the QE and the 1600 and 16803 seem to be about on par there.

the maths is not an end in itself, just a systematic way to work through the myriad of interacting parameters without spending a million bucks on hardware.

This is not a Sony sensor, but with a QE of around 0.6 (guess) and read noise that can be <1e, it is surprisingly close to "Nirvana" of QE=1 and RN=0. There is not much more opportunity for anyone to develop sensors much beyond where they are, when they are already so close to perfection. BSI can give much better QE with the small pixels on mobile phones, but isn't anywhere near as significant with 3.8 micron pixels. Maybe there will be a bit more QE from BSI and more dynamic range is possible, but the days of new astro sensors being waaay better than their predecessors are over. Maybe the way of the future will be much larger sensors, either with lots more pixels or with bigger pixels, but the Panasonic chip in the 1600 is the only largish CMOS mono chip out there for now. There are a few very interesting CMOS chips in the medium format world (where the 16803 came from), but they are not mono and are 5 figure $.

edit: for interest, the following comes from an Aptina white paper(no longer available) that suggests that BSI and FSI technologies are about equal at 1.4micron pixel size - it doesn't mean that BSI will not be used with larger pixels, just that it will not be quite as significant in that regime. In particular, it is pointed out that BSI pixels have worse crosstalk (lateral blooming) presumably because there is no backside electrode structure to stop it:

"For 1.4 micron BSI pixels, QE is typically in the 50-60 percent range with crosstalk in the 15-20 percent range. The combination of BSI’s high QE and somewhat degraded crosstalk at 1.4 micron results in a net overall image quality that is comparable to FSI for 1.4 micron pixels.

Today, 1.1 micron BSI pixels are still in the early stages of development, but when they are production-ready, they would be expected to have a QE approaching 50-60 percent with crosstalk in the 10-30 percent range. These 1.1 micron BSI pixels should be outperforming 1.1 micron FSI pixels at that point due to the fabrication challenges in shrinking FSI pixels to 1.1 micron.

A tipping point for BSI will be the 1.1 micron pixel node where FSI will likely be unable to achieve the market-required performance – necessitating a transition to BSI for applications that require this smaller pixel."

Last edited by Shiraz; 18-10-2016 at 12:16 PM.
Reply With Quote
  #34  
Old 18-10-2016, 12:33 PM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Ray, other than the widely published 60% QE i can not actually find any calculated QE for the mono Panasonic MN34230ALJ sensor. Senorgen provides the calculation methodology but its way over my head.
And can you, or someone else, explain the difference between Relative QE and the calculated QE?

Last edited by glend; 18-10-2016 at 12:45 PM.
Reply With Quote
  #35  
Old 18-10-2016, 12:59 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
relative QE just gives the relationship between the QEs at various wavelengths - eg, it might show that the QE at Ha is 2/3 that at 500nm. It does not tell you what either one actually is - ie how many photoelectrons you will get for a given number of photons at that wavelength. To measure that you need a calibrated source and geometry that is known, so that you can say with confidence that the illumination on a pixel is nn photons and then by measuring the signal, you can work out the absolute quantum efficiency = photoelectrons/photons.

Forget the Sensorgen stuff - it is based on poorly documented measurements that may or not have anything at all to do with absolute quantum efficiency. For example, in one of primary the source documents, the author states "The calculated QE is denoted as relative QE, since there might be an multiplicative factor for all QEs. But my opinion is that the range 23 – 40% is quite true for CMOS sensors." uuurk!

The only way to measure absolute QE is with a calibrated light source and precise geometry.
Reply With Quote
  #36  
Old 22-10-2016, 07:24 AM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Ray i see Sam (ZWO) is claiming true 14 bits for the just released (Beta Testing), APC-C colour camera the ASI071MC. I don't see it as a threat to the 1600 but would like to know if it is actually delivering 14bit. Read noise is a little higher than the 1600 it seems.
Reply With Quote
  #37  
Old 22-10-2016, 07:40 AM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
At Gain 0 the ASI1600 has about 12.45 bits of dynamic range. At 14 bits that puts it up there with the Sony A7R which has a huge dynamic range.

EDIT: Just noticed it has the same sensor as the Nikon D7000, good sensor If they'd made an FX version of this sensor (D800?) a number years ago I'd have pulled the trigger on that
Reply With Quote
  #38  
Old 22-10-2016, 08:33 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
according to this https://landingfield.wordpress.com/tag/imx071/, the true dynamic range of the chip in the new camera is 12.4 bits, which is about the same as the asi1600. That isn't a problem - the data will fit in 14 bits and give you some read noise in the lower bits....plus a marketing tool.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 08:58 AM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement