#21  
Old 03-06-2016, 01:03 PM
Slawomir's Avatar
Slawomir (Suavi)
Amateur Photon Collector

Slawomir is offline
 
Join Date: Sep 2014
Location: Proserpine
Posts: 3,009
Quote:
Originally Posted by RickS View Post
I have been unfairly accused of adding diffraction spikes to stars in refractor images taken with a KAF-8300!
That accusation was a result of my ignorance- please forgive us!
Reply With Quote
  #22  
Old 03-06-2016, 01:18 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by Slawomir View Post
That accusation was a result of my ignorance- please forgive us!
I've heard it from quite a few people, Suavi. I wasn't trying to point the finger at anyone in particular
Reply With Quote
  #23  
Old 03-06-2016, 02:03 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 15,460
All modern CCDs and DSLRs use microlenses. They've been around for a while.

Microlenses have evolved over time as well. They were an issue with the ST3200ME as you say the 8300 has little diffraction spikes on bright stars. The Prolone 16803 has a little diffraction spike in one direction only on bright stars. Little oddities.

A few years back a few camera makers made marketing use out of gapless microlenses for a few years.

Another factor often talked about in digital camera circles but not in astro CCD is the thickness of the stack. That is how thick the sensor is plus the microlenses plus the cover slip. It can affect performance with some lenses.

Microlenses on the Sony A7r were custom designed so that those nearer the edges would slope the light over to prevent colour smearing on the edges of images with wide angles lenses.


Greg.
Reply With Quote
  #24  
Old 03-06-2016, 03:40 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,787
looks to me like we need some measurements to establish:
- (if/by how much) charge spreads laterally from full pixels
- how much light scattering there is from microlenses/coverglass etc.

anyone ever seen anything on any such measurements or how to do them? I found one paper on measuring charge diffusion in back illuminated pixels with a scanning laser, but we really need something that can be done without an optics lab.

It would be nice to know how much the star size is determined by the optics and how much by the detector - could help inform purchase decisions, particularly since the latest BSI chips may well lack the potential barriers that prevent lateral charge diffusion in FSI chips.
Reply With Quote
  #25  
Old 05-06-2016, 10:57 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Ray,

I dropped an email to Richard Crisp with a couple of simple questions. Will let you know if I hear back from him. If anybody knows this stuff it will be him!

Cheers,
Rick.
Reply With Quote
  #26  
Old 06-06-2016, 09:28 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by RickS View Post
I dropped an email to Richard Crisp with a couple of simple questions. Will let you know if I hear back from him. If anybody knows this stuff it will be him!
Richard was kind enough to respond with his comments and also wrote a little article (link included below.)

I think this in combination with Ray's thoughts on seeing explain bright star bloat, at least to my satisfaction.

Quote:
Originally Posted by Richard Crisp (in email to RickS)
the primary source of the bloat is the fact that there's a Detection Limit of the electronic imaging system.

If you could plot the intensity of individual stars versus the distance from the centroid of the star you get a curve reminiscent of a Gaussian (bell) curve.

At the top of this curve you have a small width of the curve, at the bottom it is very wide

The brighter the star is, then the lower in this curve is the Detection Limit: sort of like the base of a tree trunk versus the tip of the highest straight branch if that analogy makes sense.

Certainly there can be other contributing factors such as light scattering caused by any of a number of different sources: external to the system can be turbulence in the air, dust and so on. Internal to the system dirty optics, internal reflections within the optical path including microlens interactions could in principle contribute to bloat

Finally it is possible to have an image sensor that has poor diffusion MTF at the wavelengths of interest.

A common problem is encountered when one attempts to image at NIR wavelengths with a sensor not designed specifically for NIR such as a Deep Depletion device as made by E2V.

Diffusion MTF is simply a measure of how efficiently all of the charge is collected in the target pixel versus being collected by adjacent pixels. For visible wavelengths and KAF or KAI sensors statistically only a handful of electrons wind up in the wrong pixel. If very many did compared to the total charge in a given pixel, then the image would appear smeared or out of focus. That's a real issue for KAF/KAI and NIR imaging but this shouldn't be any impact on star bloat when we image using NIR blocking filters.


Here's a link to a PDF tutorial on the subject of Star Bloat etc

http://www.narrowbandimaging.com/inc...ters_crisp.pdf

Last edited by RickS; 06-06-2016 at 10:04 AM.
Reply With Quote
  #27  
Old 06-06-2016, 10:13 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,787
thanks very much Rick and many thanks to Richard as well.

the only thing I would add is that Richard's tutorial applies to diffraction limited stars and ignores atmospheric seeing. The star profiles presented earlier incorporate atmospheric seeing and that will be the primary determinant of resolution. However, the general point that the star size may be entirely explained by the optical physics of image formation is still appropriate. That includes the violet halos that occur when using a refractor that has been optimised for visual use.

My understanding is that a small amount of charge diffusion may occur in some BSI chips in the visible band, where there may not be any potential barrier to charge movement (it may not only be NIR that does this). However, the FSI chips that we use have potential barriers to prevent lateral charge diffusion. As Richard says, our star shapes do not depend on charge spreading over the surface of the detector.

His graph showing how the minimum detectable signal varies with read noise is also germane to the two threads currently going on the new low read noise CMOS chips.

Last edited by Shiraz; 06-06-2016 at 01:03 PM.
Reply With Quote
  #28  
Old 06-06-2016, 12:07 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 15,460
Great article. Thanks for posting Rick and thanks to Richard for the detailed write up.

The airy disc rings showing up on some brighter stars is kind of what I thought was going on. So when you have a higher QE cameras those outer rings of the airy disc can show up giving wider stars. The detection limit concept of wider stars when they are fully bright to expose the whole airy disc and get it above the read noise is quite a visible phenomenom when processing images from some cameras.

Also there is another question mark over how well antiblooming works on all chips.

So from this we get what? - with high QE small well low read noise sensors it makes more sense to do shorter subs and keep the airy disc outer rings below the detection limit to get tighter stars. Use longer exposures with sensors with deeper wells and higher read noise.

Also fast scopes get tighter stars. An interesting one. ASA Newts always seemed to me to have the smallest stars. Newts in general seem to have tighter stars than other systems. I guess they are often fast compared to other scopes.

Greg.
Reply With Quote
  #29  
Old 06-06-2016, 12:21 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,787
Quote:
Originally Posted by gregbradley View Post
Great article. Thanks for posting Rick and thanks to Richard for the detailed write up.

The airy disc rings showing up on some brighter stars is kind of what I thought was going on. So when you have a higher QE cameras those outer rings of the airy disc can show up giving wider stars. The detection limit concept of wider stars when they are fully bright to expose the whole airy disc and get it above the read noise is quite a visible phenomenom when processing images from some cameras.

Also there is another question mark over how well antiblooming works on all chips.

So from this we get what? - with high QE small well low read noise sensors it makes more sense to do shorter subs and keep the airy disc outer rings below the detection limit to get tighter stars. Use longer exposures with sensors with deeper wells and higher read noise.

Also fast scopes get tighter stars. An interesting one. ASA Newts always seemed to me to have the smallest stars. Newts in general seem to have tighter stars than other systems. I guess they are often fast compared to other scopes.

Greg.
it's aperture that determines how tight the stars are Greg, not FNo - FNo is important though if you have fixed pixel size and focal length (then a faster scope will give you a bigger aperture for the same focal length). The FSQ106 for example produces relatively fat bright stars, even though it is fairly fast - because it has a small aperture.

When you consider the effects of seeing, the Airy disk rings are filled in and you get basically a smoothly varying profile with most apertures. You also get a fatter core region, where seeing determines what resolution you get, not diffraction.

The bottom end star skirt size will depend on the degree of stretching - you will get bigger stars with more nearly saturated cores if you stretch more. If you need to expose deep and stretch hard to get faint stuff, you will get a lower detection limit and larger stars again. You get the same result regardless of QE or well depth, provided you choose appropriate sub lengths. However, if your small wells are from small pixels, the stars (and everything else) will be larger because you will have more of the smaller pixels under each one.

Last edited by Shiraz; 06-06-2016 at 02:03 PM.
Reply With Quote
  #30  
Old 06-06-2016, 03:01 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 15,460
Richard's paper is quite clear on that point. Its only F-ratio that determines the airy diameter. Its near the beginning of his paper about page 3.

Greg.
Reply With Quote
  #31  
Old 06-06-2016, 03:04 PM
Slawomir's Avatar
Slawomir (Suavi)
Amateur Photon Collector

Slawomir is offline
 
Join Date: Sep 2014
Location: Proserpine
Posts: 3,009
Really interesting thread, have definitely learnt a few new things

EDIT: it looks like my next scope will have to be a fast Newtonian- large aperture and fast f/ratio for a reasonable price!

What might also be affecting how stars appear on a screen is simply resolution of the image; that's where more megapixels may help in making stars look smaller. I bet that 51mp sensor would for most telescopes produce nice small stars (given normal sampling), at least on a screen...
Reply With Quote
  #32  
Old 06-06-2016, 03:30 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 15,460
Quote:
Originally Posted by Slawomir View Post
Really interesting thread, have definitely learnt a few new things

EDIT: it looks like my next scope will have to be a fast Newtonian- large aperture and fast f/ratio for a reasonable price!

What might also be affecting how stars appear on a screen is simply resolution of the image; that's where more megapixels may help in making stars look smaller. I bet that 51mp sensor would for most telescopes produce nice small stars (given normal sampling), at least on a screen...
It has been a good thread and answered some questions for me.
Yes I think it would. The only images I have seen with that FLI 50100 are from Wolfgang Promper and they were narrowband. Tiny stars but then stars are often small with narrowband.

Greg.
Reply With Quote
  #33  
Old 06-06-2016, 05:41 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,787
Quote:
Originally Posted by gregbradley View Post
Richard's paper is quite clear on that point. Its only F-ratio that determines the airy diameter. Its near the beginning of his paper about page 3.

Greg.
And that is true, but it doesn't tell the whole story. To prove that is the case, just imagine that both of the diffraction patterns in Richard's paper were sampled by 20 micron pixels. The resulting star shapes for both f8 and f2.8 would be the same - the energy for would both be contained in a single pixel. ie, the star shape in an image is not determined by the FNo. in isolation.

When you allow the pixel size and focal length to vary depending on sampling requirements etc. and then you incorporate seeing (which is measured in angular terms), the final determinant of tightness of star skirts is the aperture - properly sampled big scopes have inherently tighter star skirts than properly sampled smaller ones, as illustrated by the angular star profiles in the figure attached to post #9. Of course we already know this from viewing images taken by 24 inch f8 scopes - they look way tighter than those from FSQ106s, even though the 106 is a faster scope. This is not to say that a fast scope is no advantage - after you have settled the sampling by choosing pixel size and focal length, a faster scope will give you a bigger aperture, resulting in tighter star skirts and more signal.

Last edited by Shiraz; 06-06-2016 at 06:17 PM.
Reply With Quote
  #34  
Old 07-06-2016, 04:10 AM
ptc (Richard)
Registered User

ptc is offline
 
Join Date: Aug 2010
Location: Silicon Valley
Posts: 124
comment on analysis

It doesn't change the fact that there's a profile shape and that there's a detection limit and the two determine how fat the stars are.

We seldom experience diffraction limits but that's where I began because people are familiar with the Airy diameter and that there's an intensity versus radius dependence of stars.

--geez I haven't logged into this site for so long I had forgotten my user ID and PW!

rdc

here's the link to the latest version of the analysis/writeup

http://www.narrowbandimaging.com/inc...ters_crisp.pdf


If interested I can write about the BSI versus FSI sensors and how the MTF and QE are affected by the illumination strategy and how they can be improved by clever design of the IC

I have quite a bit I can say about that.
Quote:
Originally Posted by Shiraz View Post
thanks very much Rick and many thanks to Richard as well.

the only thing I would add is that Richard's tutorial applies to diffraction limited stars and ignores atmospheric seeing. The star profiles presented earlier incorporate atmospheric seeing and that will be the primary determinant of resolution. However, the general point that the star size may be entirely explained by the optical physics of image formation is still appropriate. That includes the violet halos that occur when using a refractor that has been optimised for visual use.

My understanding is that a small amount of charge diffusion may occur in some BSI chips in the visible band, where there may not be any potential barrier to charge movement (it may not only be NIR that does this). However, the FSI chips that we use have potential barriers to prevent lateral charge diffusion. As Richard says, our star shapes do not depend on charge spreading over the surface of the detector.

His graph showing how the minimum detectable signal varies with read noise is also germane to the two threads currently going on the new low read noise CMOS chips.
Reply With Quote
  #35  
Old 07-06-2016, 04:28 AM
ptc (Richard)
Registered User

ptc is offline
 
Join Date: Aug 2010
Location: Silicon Valley
Posts: 124
The airy diameter is determined solely by the f# and the wavelength of the light

Ray, you are confusing imaging of an Airy disk with the properties of the disk itself

if you want to image the disk you need to follow normal Nyquist sampling criteria

this is a very important zeroth-order issue.



Quote:
Originally Posted by Shiraz View Post
And that is true, but it doesn't tell the whole story. To prove that is the case, just imagine that both of the diffraction patterns in Richard's paper were sampled by 20 micron pixels. The resulting star shapes for both f8 and f2.8 would be the same - the energy for would both be contained in a single pixel. ie, the star shape in an image is not determined by the FNo. in isolation.

When you allow the pixel size and focal length to vary depending on sampling requirements etc. and then you incorporate seeing (which is measured in angular terms), the final determinant of tightness of star skirts is the aperture - properly sampled big scopes have inherently tighter star skirts than properly sampled smaller ones, as illustrated by the angular star profiles in the figure attached to post #9. Of course we already know this from viewing images taken by 24 inch f8 scopes - they look way tighter than those from FSQ106s, even though the 106 is a faster scope. This is not to say that a fast scope is no advantage - after you have settled the sampling by choosing pixel size and focal length, a faster scope will give you a bigger aperture, resulting in tighter star skirts and more signal.
Reply With Quote
  #36  
Old 07-06-2016, 06:36 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,787
Quote:
Originally Posted by ptc View Post
The airy diameter is determined solely by the f# and the wavelength of the light

Ray, you are confusing imaging of an Airy disk with the properties of the disk itself

if you want to image the disk you need to follow normal Nyquist sampling criteria

this is a very important zeroth-order issue.
We can equally accurately say that the angular diameter of the Airy disk is determined solely by the aperture and the wavelength of light.

You are looking at linear measures in the focal plane and I am looking at angular measures in the field.

Since we are talking about the appearance of stars in our images, I think that it is necessary to incorporate the image sampling process as well as the Airy pattern generation in the analysis. The "20 micron pixel" comment shows why that is important.

I chose to use the angular frame of reference because:
- It is universally used for specifying optical parameters - resolution, field of view, seeing, sampling etc are all angular terms. The standard Airy disk formulation is an angular one.
- I wanted to incorporate seeing (which completely changes the PSFs) and the convolution applies in the angular domain, not on the focal plane.
- I wanted to avoid the odd outcomes that occur with a linear focal plane analysis, without consideration of sampling. For example, such an analysis would suggest that an f1.4 23mm Samyang lens will have smaller stars and better resolution than the 3.9m f3.3 AAO Telescope, because it has a lower FNo and smaller Airy disk. But we all know that the AAO will vastly outperform the small lens. An angular analysis shows that the angular resolution depends on the aperture and, along with sampling considerations, this explains why the performance of the big scope far exceeds that of the small lens.

Despite the differences in our approaches, I am sure that we agree that star size and shape in our images is determined solely by the physics of the image formation process (and I would add, the sampling) and that it is not due to charge diffusion in the detectors. That at least is a major step forward - thank you.

Regards Ray

Last edited by Shiraz; 07-06-2016 at 09:21 PM.
Reply With Quote
  #37  
Old 08-06-2016, 12:14 AM
ptc (Richard)
Registered User

ptc is offline
 
Join Date: Aug 2010
Location: Silicon Valley
Posts: 124
you are talking around the point I made that you were mixing sampling of a spot with the properties of the spot.

my point is they are fundamentally different. The spot is a spot.

getting an accurate sample of the spot involves sampling theory and that essentially is going into things like Nyquist or not and so on.

I am simply dealing with the zeroth order issues. You can put all the filigree you want on it including debating similar-sounding but unrelated topics.

I am only saying don't mix apples with oranges unless you want to make punch.

ciao
rdc


Quote:
Originally Posted by Shiraz View Post
We can equally accurately say that the angular diameter of the Airy disk is determined solely by the aperture and the wavelength of light.

You are looking at linear measures in the focal plane and I am looking at angular measures in the field.

Since we are talking about the appearance of stars in our images, I think that it is necessary to incorporate the image sampling process as well as the Airy pattern generation in the analysis. The "20 micron pixel" comment shows why that is important.

I chose to use the angular frame of reference because:
- It is universally used for specifying optical parameters - resolution, field of view, seeing, sampling etc are all angular terms. The standard Airy disk formulation is an angular one.
- I wanted to incorporate seeing (which completely changes the PSFs) and the convolution applies in the angular domain, not on the focal plane.
- I wanted to avoid the odd outcomes that occur with a linear focal plane analysis, without consideration of sampling. For example, such an analysis would suggest that an f1.4 23mm Samyang lens will have smaller stars and better resolution than the 3.9m f3.3 AAO Telescope, because it has a lower FNo and smaller Airy disk. But we all know that the AAO will vastly outperform the small lens. An angular analysis shows that the angular resolution depends on the aperture and, along with sampling considerations, this explains why the performance of the big scope far exceeds that of the small lens.

Despite the differences in our approaches, I am sure that we agree that star size and shape in our images is determined solely by the physics of the image formation process (and I would add, the sampling) and that it is not due to charge diffusion in the detectors. That at least is a major step forward - thank you.

Regards Ray
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 12:09 AM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Advertisement
SkyWatcher Australia
Advertisement
Lunatico Astronomical
Advertisement
Celestron Australia
Advertisement
Bintel
Advertisement
Meade Australia
Advertisement
OzScopes Authorised Dealer
Advertisement
NexDome Observatories
Advertisement
Astronomy and Electronics Centre
Advertisement