23 replies to this topic

[Stargazer3236]

I am puzzled. What does Full Width, Half Measure mean? How do I apply that to my imaging, when, for instance, I use SharpCap and have the option of applying FWHM to my stacking options?

[Bigdan]

I believe it relates to focusing.  When I am imaging, I have two mechanical methods for focusing:  Bahtinov mask, and a low FWHM number displayed by my camera control software.  Consistently, when I focus my CCD with Bahtinov mask, I notice when I come to focus, the FWHM is at its lowest value.  I haven't really checked it..... I use Deep Sky Stacker for stacking, and register my images first.  It gives a score, and I use the image with the highest score as the reference image.  I haven't checked to see if there is a correlation between highest score and lowest FWHM, but I think there are other factors taken into account, also.

 

A low FWHM is good.

[syscore]

I am puzzled. What does Full Width, Half Measure mean? How do I apply that to my imaging, when, for instance, I use SharpCap and have the option of applying FWHM to my stacking options?

https://www.youtube....h?v=UnnxzqMJuts

[tolgagumus]

A stars 3D profile somewhat looks like a bell. FWHM is the width of the bell at its midway point hence "half". For all practical purposes you want this number as small as possible. As the star goes out of focus, the shape of the bell gets fatter and the mid point gets larger in diameter.

Sent from my SAMSUNG-SM-G870A using Tapatalk

[Zebenelgenubi]

 

I am puzzled. What does Full Width, Half Measure mean? How do I apply that to my imaging, when, for instance, I use SharpCap and have the option of applying FWHM to my stacking options?

https://www.youtube....h?v=UnnxzqMJuts

 

Wow that's an excellent discussion of what is going on in an astro image.  Thanks for the reference.

[Jon Rista]

FWHM is actually Full Width at Half Maximum, which indicates it is the width of the star profile half way to the peak. Here is an example cross-section plot of a large star (and you can see much smaller stars as the spikes in the plot around the large bell curve), and I've marked where the FWHM is at half the height of the star:

 

 

The star peaks at 4000 ADU, the mean background level is about 700 ADU, which puts the half maximum at 2350. The "full width" is the width between the two red arrows, at that half maximum level. That is a standard measure of the size of a star. 

[freestar8n]

And for people who are learning about FWHM for the first time - the next question is - how does the fwhm vary with star intensity? You can look at stars in an image and the bright ones are much "bigger." But in fact - as long as the stars are not saturated, the fwhm is about the same - regardless of brightness. That's because a brighter star will be taller and "wider" overall - but it still maintains the same profile shape.

As the profile gets taller, the "waist" where you measure the width of the star (which is half way up to the max intensity) also rises - and it stays about the same.

So fwhm should be constant regardless of star intensity - and that's why it's a good measure of sharpness and focus/seeing/guiding quality.

Frank

[John Miele]

Frank,

That is a great explanation and answers that nagging question I have had about whether I should worry about the brightness when evaluating fwhm!

[Jon Rista]

It should probably be pointed out, since the example I shared deviates from Franks description...that very large stars, even if they are not saturated (the star in my example above is the large, VERY BRIGHT star Gamma Cas), will often have significantly larger FWHMs than the "average" star in the frame. You can see many other spikes in the plot in my screenshot above...those are smaller more average sized stars around Gamma Cas. You can see that at their FWHM, they are all about the same, which does line up with what Frank said.

Usually, when measuring FWHMs, the tool you use to do it will have some means of configuring what level of brightness to cut the calculation off at. Some also have a black point setting. These can be used to tune the algorithm to look a the right information to give you a useful result.

[freestar8n]

Jon - those look like noise spikes rather than stars.

Frank

[Jon Rista]

Jon - those look like noise spikes rather than stars.

Frank

I'm pretty sure they are stars, as there are a number of very small stars scattered about very close to Gamma Cas itself in the image. This is data that has been calibrated, cosmetically cleaned to eliminate any remnant hot pixels, and integrated with winsorized sigma clipping. The noise profile is pretty clean...so the only spikes left should be the stars. This was also from pretty short subs, 90 seconds, and most of the stars except Gamma Gas are pretty small...few pixels across. 

Edited by Jon Rista, 21 November 2016 - 04:30 PM.

[schmeah]

So do you guys get somewhat lower FWHMs with Ha starfields than with luminance, even when taken on the same night with the same conditions and optimal focus? And if so, what explains that? To be honest I'm not certain that I trust the reliability of FWHM, eccentricity/roundness measurements nor the correlation between these measurements and RMS guiding error. Or at least not with the software that I use (CCD stack and Maxim LE).

 

Derek

[Jon Rista]

Derek, I think the main reason NB stars tend to be smaller is they are a narrow sliver of the spectrum, so they can actually be focused better than the entire spectrum (i.e. an L filter). The L filter is focusing all the light from 700nm to 390nm, and even a very good scope is still going to have some dispersion at one wavelength or another. So even with the best possible L focus, the stars are still going to have larger FWHM than say from an Ha filter. With the Ha filter, you are focusing a very specific and narrow range of wavelengths, so you don't have to worry about the dispersion in other wavelengths. 

 

As for FWHM measurements and how they correlate with guide RMS. Guide RMS is one factor out of many that affect blur. I usually try to calculate what my best possible FWHMs are using this formula:

 

FWHM = SQRT(Seeing^2 + Dawes^2 + GuideRMS^2 + ImageScale^2)

 

There are additional blur factors, and they can come from a variety of things. The low pass filter on a DSLR can introduce significant blur. Wind can introduce blur. It may be better to separate GuideRMS from TrackingError as well, although I tend to roll them into one. Filters will introduce their own blur. And, optical aberrations will add some blur...I usually just go with Dawes as I don't know of any simple way to calculate what blur the various aberrations may be adding (although it could be non-trivial, and can be considered a key source of error in this calculation, when it tells you your FWHM  should be 2" and you measure 3".)

 

In my case, my seeing seems to be around 1" or so on average, the Dawes limit of my scope is 0.76", my guide RMS is usually around 0.65" on average (although it's been edging closer to 0.55" more these days, so I may have to change my calculation), and my image scale is 1.3". With these numbers, I calcuate a best-possible FWHM of 1.92". I measure FWHMs ranging from about 2.1" to 2.5" from most of my data, and in general they seem to fall around 2.3-2.4". I know that my lens has some spherical aberration, and it seems to have a little bit of CA, although not much. Those undoubtedly account for some of the extra blur in my actual measurements. The rest could be that seeing is worse than I think, and I am sure that wind plays a role, as I always have a light breeze at the very least. 

[sharkmelley]

So do you guys get somewhat lower FWHMs with Ha starfields than with luminance, even when taken on the same night with the same conditions and optimal focus? And if so, what explains that?

Longer wavelengths are less affected by the seeing (atmospheric turbulence) than shorter wavelengths and so they result in smaller FWHM.  Planetary and lunar imagers will often use H-alpha or near IR for that very reason.

 

Mark

[freestar8n]

Longer wavelengths are less affected by the seeing (atmospheric turbulence) than shorter wavelengths and so they result in smaller FWHM.  Planetary and lunar imagers will often use H-alpha or near IR for that very reason.
 
Mark

There is some evidence for benefit of longer wavelength in specific situations of turbulence and video imaging of planets - but for either guiding or deep sky imaging - I don't know of any benefit of longer wavelength. There is a slight improved steadiness of tip/tilt seeing effects at longer wavelength - but they are countered by the increased size of the Airy disk. That is why it is known as "the infra-red myth."

In my case - the primary cause for any difference due to wavelength in deep sky imaging is just the inherent chromatic aberrations of the optics. For my C11 with reducer, there was spherochromatism that tended to cause swelling of the blue channel - but with Edge and its better corrected optics I don't see that. With my recent hyperstar imaging I got smaller fwhm with green Oiii than red Ha.

I think a big reason the stars seem smaller with narrow band is a combination of the reduced chromatic aberration of the optics - and the fact that the stars are so suppressed in intensity and simply look smaller in long exposure images.

Frank

[schmeah]

So when most imagers on this site casually report their "FWHM" as say, less than 2" and attribute it to their excellent new premium mount, the filter type, seeing conditions etc are almost never reported. There never seems to be any standard, so I never really know what it means. Even in my own images, on occasion I can look at a sub that I might have tossed visually for star bloat, but the reported FWHM in CCD stack is smaller than a sub whose stars clearly are tighter. So like I said, I don't really trust these measurements except as a very basic guideline. Am I way off base here?

 

Derek

[Jon Rista]

So when most imagers on this site casually report their "FWHM" as say, less than 2" and attribute it to their excellent new premium mount, the filter type, seeing conditions etc are almost never reported. There never seems to be any standard, so I never really know what it means. Even in my own images, on occasion I can look at a sub that I might have tossed visually for star bloat, but the reported FWHM in CCD stack is smaller than a sub whose stars clearly are tighter. So like I said, I don't really trust these measurements except as a very basic guideline. Am I way off base here?

 

Derek

One thing that FWHM does not really handle well is halo bloat. Because it is the measure of the star at half the maximum, it effectively ignores any halo spreading. I don't know how many things can cause that...but certain patterns in guiding, certain aspects of dispersion etc. can result in a broader halo while the majority of the star's energy is concentrated within the centroid. The FWHM might increase a small amount, but not enough to account for the full spreading of the halo, which usually occurs well below the half maximum point. 

 

You can model a star and calculate the FWHM using different profiling algorithms. You can use a Gaussian profile (which would model saturated to near saturated stars better), Lorentzian profiles which might handle halo spread better, and Moffat profiles, which are considered to model stars better than either Gaussian or Lorentzian. If you have very soft, well sampled, round stars, then a Gaussian profile might actually be better as well...and this will usually give you the largest FWHMs. You may find that if you are currently modeling with Moffat, and switch to Gaussian, you get FWHMs that seem more reasonable. 

[Organic Astrochemist]

Can someone compare and contrast FWHM and eccentricity? How does one calculate eccentricity?

Here's a recent narrowband image I took where the FWHM, was around 4, but the stars looked much better and rounder than some other images where FWHM was less than 3.

 

I don't think those stars are too bad for 60s subs, unguided, on a non-premium mount, but I'd be glad to have someone evaluate them (if possible).

I can see why FWHM is important, but maybe other measures are also important.

[tolgagumus]

Maxim, CCD inspector, Pixinsight all give eccentricity measurements. I'm sure others will too. My typical FWHM with the broadband filters is in average about 2.75". However when using NB filters the FWHM drops 1.75". This is at 0.49" pixel. This is measured with CCDI. Pixinsight gives me a different number for the same subs. I don't think you can compare FWHM numbers from one system to another unless they are identical and shot from the same location using same software to measure. My cut off for eccentricity is 20% according to CCDI or 0.70 according to Pixinsight.

Sent from my SAMSUNG-SM-G870A using Tapatalk

Attached Thumbnails

• 

Edited by tolgagumus, 23 November 2016 - 04:27 AM.

[syscore]

Can someone compare and contrast FWHM and eccentricity? How does one calculate eccentricity?

Here's a recent narrowband image I took where the FWHM, was around 4, but the stars looked much better and rounder than some other images where FWHM was less than 3.

 

IC63 2 (2).jpg

I don't think those stars are too bad for 60s subs, unguided, on a non-premium mount, but I'd be glad to have someone evaluate them (if possible).

I can see why FWHM is important, but maybe other measures are also important.

FWHM and Eccentricity are independent of each other. The former measures the width of your stars and the latter, their roundness. You can have small non-round stars. Eccentricity is generally a function of your mount's tracking/guiding ability, while FWHM is generally a function of your seeing, assuming in both cases that your optics (and focus) are in good shape. MaximDL uses Flatness in place of eccentricity, which can be converted to eccentricity and vice versa. PixInsight uses eccentricity. An eccentricity of 0.42 (or a flatness of 0.10) is considered perceptibly "round". The eccentricity (and flatness) of a perfect circle is 0.

[schmeah]

In reading the responses in this and other similar threads, I wonder why we obsess so much with numbers that are not standardized and vary so much depending on the software, algorhythm, user input etc so as to render them marginally useful at best. I can understand if someone is doing actual science, astrometry where the numbers actually matter, but in that instance rigid standards would need to be adhered to and statistical validity upheld. IMO our eyes are far better at consistently telling us which subs are "better" , in terms of composing a pleasing astrophoto, and all this number quoting is little more than mental .....

 

Derek

[Jon Rista]

In reading the responses in this and other similar threads, I wonder why we obsess so much with numbers that are not standardized and vary so much depending on the software, algorhythm, user input etc so as to render them marginally useful at best. I can understand if someone is doing actual science, astrometry where the numbers actually matter, but in that instance rigid standards would need to be adhered to and statistical validity upheld. IMO our eyes are far better at consistently telling us which subs are "better" , in terms of composing a pleasing astrophoto, and all this number quoting is little more than mental .....

 

Derek

If you personally use the same software and the same algorithm to evaluate your stars, then FWHM does make a useful gauge as to the quality of your own data. If you use the same software and model to compare to other's stars (i.e. PixInsight's FWHMEccentricity script), then you would be comparing consistent results as well.

 

FWHM and eccentricity can also play a key role in helping you optimize the quality of your data to meet your own personal goals. If you want maximum resolution with the roundest stars, you can rank and weight your subs by FWHM and eccentricity, and discard subs that fall outside of some relative or absolute range. PixInsights SubframeSelector script allows you to do this. 

 

In my case, I went from FWHMs over 4" to as high as 9" or so with my 5D III without and with an LP filter, to FWHMs of 2-2.5" with the ASI1600. The improvement was massive and amazing. I have also been able to draw parallels between what I see as softer subs while I am acquiring the data, and how the FWHMs actually measure...with these tiny pixels and monochrome sensor, smaller differences are more obvious than they were with my 5D III with it's blurring low pass filter and mosaiced sensor array. I use FWHM much more as a relative guide than an absolute guide of how sharp my subs are...relative to other data acquired with the same scope.

[acaballero]

Excellent discussion. Good references and facts. No opinions and speculations. I wish all topics in all forums could be like this one.

[terry59]

In reading the responses in this and other similar threads, I wonder why we obsess so much with numbers that are not standardized and vary so much depending on the software, algorhythm, user input etc so as to render them marginally useful at best. I can understand if someone is doing actual science, astrometry where the numbers actually matter, but in that instance rigid standards would need to be adhered to and statistical validity upheld. IMO our eyes are far better at consistently telling us which subs are "better" , in terms of composing a pleasing astrophoto, and all this number quoting is little more than mental .....

Derek

We can all quote specs or other people. Happens here all the time to the extent that relating real experiences is treated as a bad thing because it must include the prevailing environment, equipment and the operator.

These things are never secondary except in the classroom or lab while teaching or designing. Even then you need HFE

Edited by terry59, 18 July 2018 - 01:21 PM.