124 replies to this topic

[jimthompson]

Greetings all,

 

Over the past year I have done a lot of research into multi-narrowband filters, such as the OPT Radian Triad or Optolong L-eXtreme.  Through that work I have been made aware of an issue with trying to use narrow band pass filters on fast optics such as the Celestron RASA or SCT's with Starizona's Hyperstar system.  To get a better understanding of the problem I did an analysis using a set of generic H-alpha filters, ranging in bandwidth from 40nm down to 1nm.  I predicted how these filters would impact the performance of an 8" RASA telescope (f/2), as well as a 4" refractor of the same focal length (f/4).  The attached plot below is one of the main results from my analysis.  Because the pass band of the filters shift with the angle of the light passing through them, the filters act essentially like an aperture mask.  The narrower the filter, the more restrictive the mask.  This effect is very pronounced on a RASA scope because of the large central obstruction.  In fact, based on my calculations, you are much better off using the 4" refractor if you intend to use filters narrower than 6.5nm.  You can read my whole article on the matter at the link below.

 

http://karmalimbo.co...ics_Nov2020.pdf

 

Best Regards,

 

Jim T.

Attached Thumbnails

• 

[SilverLitz]

Very interesting!

 

It seems to show issues even with 7nm filters, and probably why I found myself using the same exposure times when shooting Cygnus Loop with my Samyang 135mm at f/2.8 and my Esprit 100 at f/4.13 shooting M16 when using Baader 7/8/8.5nm NB filters.  It also reinforced my decision to use f/2.8 and not f/2.0, and makes me wonder if it would have just as good to use f/4.

 

I always look at the histogram after a test shot to decide what exposure time to use.

Edited by SilverLitz, 03 November 2020 - 01:42 PM.

[jimthompson]

Thanks for adding your experience to the discussion.  I am always happy to hear that my analysis results have been observed in practice by others.

 

If you like, I can add a few different f-ratio refractors to the plot in my original post.  That would be more directly relevant to using camera lenses.  RASA and Hyperstar systems are especially sensitive to filter band width because of their central obstruction.

 

Regards,

 

Jim T.

[dangarnett]

Hey Jim, Great work on the paper. 

 

I was just thinking about this. It is really hard to find good information on how useful the High Speed filters are. Most info kinda stops at "its a steep light cone, so you'll notice an improvement with these high speed filters made for fast optics" But not a ton of technical information on how much better they are than normal filters. 

 

I took this shot last night of the Heart nebula with a Astrodon 3nm HA filter on C11HD - Hyperstar f/2. 1600mm @ 200 gain 75x180s

 

180s seemed like a ton of time for a f/2 system, so you paper mentioning that it is effectively a f/10+ makes a bit more sense why the exposures had to be so long to get some data. 

 

I'd love some comparisons on High speed vs non-high speed filters. 

 

[jimthompson]

Lovely shot Dan.  FYI, I just recently acquired a sample of the Astro Hutech IDAS-NBX filter.  It is specifically designed to be used on optics down to f/2.  I will get some analysis results posted on it as soon as I can.

 

Cheers,

 

Jim T.

[RogeZ]

Jim:

There are also Astrononiks MaxFR filters that are preshifted for fast scopes. Worth a test

[jimthompson]

I have added a bunch of additional refractor curves to my plot, each one having the same focal length as the 8" RASA (400mm) but different apertures in order to give different f-ratios.  The pattern that resulted is rather interesting I think.  It is interesting that my predictions suggest you are better off using a 50mm f/8 refractor than an 8" RASA if you intend to use filters 3.5nm or narrower.  One can also back out an optimum refractor aperture for each filter band width, i.e. what is the smallest scope I can use and still maximize my effective aperture?  (Worded another way: what is the slowest focal ratio I can use and still minimize my exposure time?)

 

Cheers,

 

Jim T.

Attached Thumbnails

• 

Edited by jimthompson, 03 November 2020 - 11:16 PM.

[alphatripleplus]

Thanks for another informative report, Jim. Definitely gives me food for thought when considering narrower filters with fast systems.

 

Although, the discussion is on H-alpha filters, I would assume your conclusions may also apply to OIII filters? I'm curious about the effect of fast optics there, as I've often read that many prefer using narrower OIII filters (e.g. 3nm) than they use for H-alpha.

[thecoldestnights]

Very interesting read!

However I'm wondering why you didn't take sensor size into account... You state that the angle at the edge of an f/2 light cone is 14°, but most people use a sensor that is much smaller than the image circle given by their scope. Therefore the relevant angle should be much lower than 14° and the effects would be mitigated.

I feel like this detail is especially important when interpreting Figure 5 : from what I understood, you integrated the transmissivity function over a theoretical image circle (whose area was normalized). The areas that affect the result negatively (where the light hits the image plane at a high angle) are precisely those that aren't necessarily covered by the sensor. Which is why, in my opinion, the sum should be done over the rectangle covered by the sensor, not over the whole image circle.

[jimthompson]

Very interesting read!

However I'm wondering why you didn't take sensor size into account... You state that the angle at the edge of an f/2 light cone is 14°, but most people use a sensor that is much smaller than the image circle given by their scope. Therefore the relevant angle should be much lower than 14° and the effects would be mitigated.

I feel like this detail is especially important when interpreting Figure 5 : from what I understood, you integrated the transmissivity function over a theoretical image circle (whose area was normalized). The areas that affect the result negatively (where the light hits the image plane at a high angle) are precisely those that aren't necessarily covered by the sensor. Which is why, in my opinion, the sum should be done over the rectangle covered by the sensor, not over the whole image circle.

The sensor size does not play a role in how much total illumination the sensor area receives.  It only impacts whether or not vignetting is observed due to diffraction at the edges of the baffle tube or other obstructions in the light path.  It is not totally intuitive but the fact is each pixel on your sensor receives light from the entire aperture of the scope.  A blockage of the scope aperture around the outside edge will affect the total amount of light getting to every part of the image circle, including the center.  You can confirm this is true by test.  Simply hold your hand in front of your telescope and see what happens to your camera image; the whole image gets darker.  That is why I have not made mention at all of sensor size in my calculation, it is not a factor in the total illumination of the image plane.

 

Thanks for the question!  Cheers,

 

Jim T.

[RogeZ]

Jim my only critique or advice on the paper is to use existing, commercial scopes for the comparison. Its fun to say 8” F/2 refractor and compare it to a RASA but again, that refractor does not commercial exist.

[jimthompson]

Thanks for another informative report, Jim. Definitely gives me food for thought when considering narrower filters with fast systems.

 

Although, the discussion is on H-alpha filters, I would assume your conclusions may also apply to OIII filters? I'm curious about the effect of fast optics there, as I've often read that many prefer using narrower OIII filters (e.g. 3nm) than they use for H-alpha.

Hi Errol,

 

I chose to do my calculations on generic H-alpha filters as the band shift issue is worse at long wavelengths.  The worst case then would be an S-II filter, but I figured that since H-alpha is the most common narrowband filter used it was more relevant to people's experience.  I did also characterize the band shift properties of the O-III band when I measured my Radian Triad filter.  The O-III band's center wavelength (CWL) shifts a similar amount to H-alpha (4.6nm at 14deg angle for O-III vs 6.2nm for H-alpha), but the drop in peak transmission is quite different.  I measured a peak transmission scale factor of 0.98 for the O-III band at a 14deg angle (i.e. practically no change), but for H-alpha the value is 0.61.  When you combine these two things together, you end up with the O-III band being much less sensitive to f-ratio than the H-alpha band.  You can see all my results from measuring the Radian Triad filter in my paper:

 

http://karmalimbo.co...tio_Oct2020.pdf

 

In terms of the curve above, using a 4nm wide filter for comparison, where the H-alpha filter gives around a 50mm effective aperture on the 8" RASA, a 4nm O-III filter would give more like a 75mm effective aperture.

 

Cheers,

 

Jim T.

[jimthompson]

Jim my only critique or advice on the paper is to use existing, commercial scopes for the comparison. Its fun to say 8” F/2 refractor and compare it to a RASA but again, that refractor does not commercial exist.

Hi Roger,

 

Agreed, an f/2 8" refractor isn't something that one can find for sale anywhere.  By the same token you aren't going to find a 1nm wide band pass filter either.  My objective in this analysis was to understand the overall trends, so to do that I have considered telescopes and filters that are outside what is currently available.  Doing this helps me to understand the bigger picture.  In particular, the theoretical 8" f/2 refractor was included to give people a better understanding of how much the RASA's central obstruction is hurting the scope performance when used with narrowband filters.  

 

Cheers,

 

Jim T.

[alphatripleplus]

  The O-III band's center wavelength (CWL) shifts a similar amount to H-alpha (4.6nm at 14deg angle for O-III vs 6.2nm for H-alpha), but the drop in peak transmission is quite different.  I measured a peak transmission scale factor of 0.98 for the O-III band at a 14deg angle (i.e. practically no change), but for H-alpha the value is 0.61.  When you combine these two things together, you end up with the O-III band being much less sensitive to f-ratio than the H-alpha band.  

Thanks, Jim. That makes sense.

[gatsbyiv]

There was an earlier thread on this where I asked: does focal ratio fully define the incident angle of light?  I don't think it does.  These analyses would seem true for a simple optical system with only an objective lens, but not for any system involving a reducer.  It is my understanding that the rays from a telephoto lens at f/2 are far more parallel than the focal ratio alone would indicate, and that the only way to truly define the incident angle is from the exit pupil distance of the optical system.  Is this not true?

[Der_Pit]

By the same token you aren't going to find a 1nm wide band pass filter either.  

Not for amateur use, yes.  But the filters in our solar observatory are all in the range 0.5-1.0nm.  There's even two 0.12nm ones.

Yes, custom made and bloody expensive (for a filter wheels load of filters you could buy a small appartement), but optically without flaws (>90% max transmission).  (They are used as prefilters for Fabry-Perot interferometers)

For the refractor I indeed doubt it's manageable.  But as you explain, it gives a good (theoretical) reference for the effects of central obstruction....

[Ken82]

This is from another forum -

http://www.astrosurf...les/filters.pdf

This is equation:

image.png.74a855f4c84c881ead5ebb2e71b5ac03.png

For 3nm filters, we need to have maximum shift of 1,5nm at 656.3nm.

654.8/656.3 = ~0.99771446

Now we square that and subtract from 1 we get ~0.0045658566.

Refraction index of air is 1.000293 and that of glass is around 1.5, so we have 0.445 * sin^2 angle = 0.0045658566

From this, we have sin angle = ~0.101356 = 0.1

Angle is 5.74 degrees.

This gives approx ratio of ~4.975 or F/5.

Interesting exercise. I expected longer F/ratio needed for such narrow filter, but it turns out that 3nm filters are good for F/5 and higher. They can certainly be used on even faster systems with slight loss of aperture due to filter acting as aperture stop. I also calculated for on axis beam - edge of field beam will have additional angle added that depends on focal length and size of sensor (but is relatively small in comparison).

I started this post believing that there would be significant difference between 3nm and 7nm on fast optics, but it turns out for regular F/ratios that we use in imaging, 3nm is just fine.

[Alex McConahay]

>>>>>>>This effect is very pronounced on a RASA scope because of the large central obstruction.

 

On what do you base this?

 

I had always heard that the difference is the F ratio, and had never before heard that it was the central obstruction.

 

Alex

[RogeZ]

Alex: A scope T ratio is affected by obstruction and it does impact the performance at the sensor. RASAs are T/2.5

[SilverLitz]

>>>>>>>This effect is very pronounced on a RASA scope because of the large central obstruction.

 

On what do you base this?

 

I had always heard that the difference is the F ratio, and had never before heard that it was the central obstruction.

 

Alex

I would expect the main cause is that the central obstruction takes out the straightest rays, the ones that would be the least affected by the narrow band passes, leaving the larger angle rays, which have the most shift. 

Edited by SilverLitz, 05 November 2020 - 02:25 PM.

[jimthompson]

There was an earlier thread on this where I asked: does focal ratio fully define the incident angle of light?  I don't think it does.  These analyses would seem true for a simple optical system with only an objective lens, but not for any system involving a reducer.  It is my understanding that the rays from a telephoto lens at f/2 are far more parallel than the focal ratio alone would indicate, and that the only way to truly define the incident angle is from the exit pupil distance of the optical system.  Is this not true?

Hi Charlie,

 

I am not an expert on camera lenses, but based on the simple schematic I found on Wikipedia (attached below), the answer to your question is: "yes, the focal ratio does fully define the incident angle of light".  From what I can see a telephoto lens is simply an objective with a Barlow after.  This arrangement allows for a much shorter lens.  The focal ratio is defined by the light cone passing through the sensor plane, and that corresponds to the reported f-ratio of the lens.  The implication here is that the f-ratio of the objective alone is much faster than that of the overall lens.

 

Regards,

 

Jim T.

Attached Thumbnails

• 

[jimthompson]

Not for amateur use, yes.  But the filters in our solar observatory are all in the range 0.5-1.0nm.  There's even two 0.12nm ones.

Yes, custom made and bloody expensive (for a filter wheels load of filters you could buy a small appartement), but optically without flaws (>90% max transmission).  (They are used as prefilters for Fabry-Perot interferometers)

For the refractor I indeed doubt it's manageable.  But as you explain, it gives a good (theoretical) reference for the effects of central obstruction....

You mention extremely narrow filters for solar observing.  In light of what I've presented above, the recommendation by H-alpha solar filter manufacturers to use these filters on scopes of at least f/30 f-ratio makes a lot of sense.  Band shift would very quickly move the filter off the H-alpha band when it is only a fraction of a nanometer wide.  I have even heard of some solar scope designs using a pair of telecentric lenses to create a section of the light path that is perfectly parallel, thus optimizing the filter performance over the entire aperture.

 

Cheers,

 

Jim T.

[jimthompson]

>>>>>>>This effect is very pronounced on a RASA scope because of the large central obstruction.

 

On what do you base this?

 

I had always heard that the difference is the F ratio, and had never before heard that it was the central obstruction.

 

Alex

Hi Alex,

 

Yes, f-ratio has the biggest impact on a scope's sensitivity to filter bandwidth.  However a scope design that has a central obstruction further amplifies the problem since (as mentioned above by SilverLitz) a larger percentage of the light passing through the filter has to do so on an angle.  Compare the red curve (RASA) to the dark purple curve (refractor) on my plot in post #7 above.  The difference between these two curves is the additional reduction in scope performance that is due to the central obstruction.

 

Regards,

 

Jim T.

[RogeZ]

Our telescopes are always focused at infinity and therefore there is no “straight and less straight” rays of incoming light, its all parallel at the entrance of the aperture.

The obstruction is just that, an area where light does not come in.

I would assume the larger the scope, the larger the percent of light that forms the steeper angles since the area of the mirror does not grow linearly with the radius. One of the knowledgeable folks will chime in. 

Jim: I think you have a great paper here, I would really encourage you to add real life scopes to the comparison like the Esprits and the Quattro scopes, the 11” RASA etc and that will provide a greater value to the community.

The 8” F/2 refractor data point is like comparing my bones to Wolverine’s adamantium....

Edited by RogeZ, 05 November 2020 - 04:40 PM.

[jimthompson]

Hi Roger,

 

I have the data for the RASA's and Hyperstar models, so I can work the curves out for them easily enough.  As for refractors, there are so many choices out there, perhaps it would be better if I generate a more generic plot that one can enter with a native scope f-ratio and find out what the predicted ratio of effective-to-physical aperture is?

 

By the way, your assumption about larger RASA scopes is not correct.  The relative size of the central obstruction to aperture is larger on the smaller RASA's (and other catadioptrics for that matter), so the central obstruction effect on filter sensitivity is mildly reduced as you increase scope size.  I show the 8" RASA because it is the worst of all the RASA models.

 

Best Regards,

 

Jim T.