DIY low res slit spectrograph based on StarAnalyzer proposal / need a bit of help

[vlaiv]

To say that I'm fairly new to spectrography is a bit of an understatement .

I do have SA200 and all the imaging gear, and I did have a go couple of times using it both visually and capturing spectrum of a star.

Now I'm considering to give it a bit more time / dedication on those Moon lit nights when DSO imaging / observing is unfeasible.

I do understand drawbacks of SA200, and I was thinking how to overcome some of them, and came up with what I believe is feasible / cheap DIY solution for increasing resolution (not sure here but I believe so) and enhancing spectral graph extraction.

Drawbacks of SA as I see them are:

1. Grating is "operating" on converging / unfocused beam of light - this means that 0 order and higher order image is not in focus at the same time unless one tilts imaging plane (or use prism in grism configuration).

2. No way to eliminate other sources of light on chip where spectrum is recorded - like lp/sky background, background stars (0 order images), other spectra overlapping due to grating orientation (this can be worked around by rotating things until there is no overlap).

3. Resolution depends on seeing / FWHM of star image

So my obvious line of reasoning was that some sort of slit needs to be added to the system in order to fix issues 2 and 3. To fix issue 1, grating needs to be placed somewhere where light beam is collimated / in focus, and also primary focal plane needs to be free to accept the slit - slit will function well only if at focal plane.

So I came up with following setup that I think will work good:

CCD -> SA Grating -> Eyepiece -> Slit at focal plane of eyepiece - > Telescope / guider / whatever else

If I understand how eyepiece works (and this is big IF ,  I base my understanding of it on eyepiece projection technique and how it "behaves" when you move sensor from and away from eye lens).

So here is diagram that explains what I think is happening with eyepiece:

Now if this diagram is correct that means that in above arrangement light from a star would be "single collimated beam" hitting SA Grating after field lens and being diffracted into spectrum - also already focused beams so 0 order image on sensor and 1 order image will be in focus.

Now for the slit - we need some configuration that will let us easily position star on slit it self and also block everything else coming from sky / surroundings. Next diagram should explain how I think slit should be designed:

Diagram shows the view when looking down the barrel of 1.25" eyepiece. Now I opted for only half or a bit more of field lens to be covered because that provides a way to relatively easy place wanted star on slit. If we orient eyepiece so that RA is vertical to this diagram - we can observe on computer screen 0 order images (and first order spectra) of all object in FOV, detect our star and move it to be above slit opening, then by using RA move it down the slit until we are satisfied. Slit RA orientation has another benefit - that of PE error of mount, sometimes present even if guiding will keep star on slit. I also propose that slit be moved to one side so that 0 order image falls on side of sensor as opposed to center (both grating and short side of sensor would be parallel to slit) and leaves enough room on sensor for 1 order spectrum image.

Ok, arrangement seems quite simple, apart from couple of issues:

1. Choice of eyepiece with know focal point (or determining it) and way to make and mount slit at focal point - first thing that comes to mind is use of old type of razor blades (those exchangeable ones made of steel, really thin). Of course type of eyepiece might have an impact, should it be some sort of flat field eyepiece? Having T2 thread under eye guard is big bonus - it can be used to mount everything together after eyepiece without need for special eye projection adapter.

2. Eyepiece focal length, distance of sensor to grating, width of slit (not sure if this is something that we can control if we opt for simple razor type slit).

How do those interact together to give optimum resolution / dispersion for given sensor size / pixel size? We would need to devise some formulae like those on SA website that give you parameters for particular distance to sensor depending on seeing (here slit width), telescope focal ratio (here probably fl of eyepiece + distance eyepiece - sensor) and sensor size / pixel size.

I would really appreciate any suggestion / comment to this, and of course your view if it will work, and how well will it work compared to simple straight forward use of SA. Any feedback on calculations and impact of different variables is very welcome also.

If this happens to work and we create "framework" to calculate and easily build one - I think it would be great second step in spectroscopy after regular use of SA and before moving on to dedicated instruments.

 

[vlaiv]

Ok, so I did a bit of research and things got "complicated"

First of all, that diagram I made how eyepiece works is just wrong - and that is a good thing

If that diagram were right - spectrograph would have really poor resolution, it turns out that we need collimated (or near collimated) beam of light but not in single point (like laser, or wrong diagram above) but rather spread over greater surface. In fact, size of surface along with groove spacing determines resolution of spectroscope (it is actual number of slits that light goes thru, and that is lines per mm x beam width in mm).

Now original SA works because it is placed in converging beam but away from focal plane (where CCD/CMOS is) and light beam there is still not fully converged but rather looks like defocused star - a circle of light (in fact that is exactly cross section of beam away from the focus).

Now on to the fun stuff. By using formula / calculation at http://www.wilmslowastro.com/software/formulae.htm#EPP

I found something really interesting for this case: you can use eyepiece projection to both increase and decrease image scale - this is really good news, and it is also good news that decreasing image scale also helps with resolution two fold:

Star FWHM is going to be smaller - less blurring and decrease of resolution, and to get decrease of image scale large focal length eyepiece is needed (well it depends on scope used as well) which in turn creates larger exit pupil - this is good because it increases theoretical resolution limit of grating.

Now onto some numbers for my particular case: Scope is 8" RC with FL of 1624mm - F/8 instrument with x0.67 reducer (effective fl of 1088 and F/5.4), with 32mm plossl eyepiece and ASI1600 placed 50 mm away from eyepiece (a bit of a problem here, because there is no real way to judge how far away from nodal point sensor is, but we can guesstimate) and grating to sensor distance of 40mm following is obtained (provided my calculations are ok).

Eyepiece magnification factor: x0.6

Star FWHM (given my average to good conditions and guiding is around 2.4") on sensor would be ~2.9 pixels

Dispersion would be: 4.8 angstrom/pixel

Exit pupil would be somewhere around 5.9mm (F/5.4 and 32mm eyepiece). And that in turn gives theoretical resolution with SA200 of R 1185. Of course real resolution will be much less and I estimate it to be around R 250 (due to 2.9 pixel size and 4.8 angstrom/pixel values but this is really only my guess rather than any calculation).

Now all of this is without the use of slit. Slit will of course add all of the benefits described above, but there is a twist to the story - since we are using eyepiece projection actual focal point when using eyepiece and looking thru it is not going to be the same as with eyepiece projection.

When placed at "true" focal point (one used for looking thru eyepiece) beam emerging from eyepiece is fully collimated, and can't be brought to a focus on Sensor. So when using eyepiece projection focus point is moved away from eyepiece (needs more out focus) - which can be a good thing, instead using slit somewhere inside 1.25" barrel close to field stop it could well be that empty filter cell can be used to hold slit elements and screwed onto eyepiece. But there is a problem - this depends on eyepiece - sensor distance, and as far as I can tell there is no way (for me at least) to calculate focal point position depending on eyepiece / sensor distance (and this is important since it dictates grating/sensor distance - we want grating to be as close to eye lens of eyepiece as possible where exit pupil beam is widest, and magnification factor - which we want to keep small).

So I guess that for slit part - which I'm not going to do for now, I'm just going to get myself eyepiece projection adapter like this one:

http://www.teleskop-express.de/shop/product_info.php/info/p7040_TS-Optics-1-25--eyepiece-projection-adapter-with-T-mount.html

and try above configuration to see how it works, it will be all about trial and error - I'll probably get empty filter cell, try to make a slit out of razor and see at what eyepiece/sensor distance is slit in sharp focus, and if that distance is usable in terms of dispersion and magnification - then great

[cletrac1922]

The guys I our club use old CD's for the pieces of material with slits cut in them

Just throwing some ideas into the ring

John

 

[vlaiv]

7 hours ago, cletrac1922 said:

The guys I our club use old CD's for the pieces of material with slits cut in them

Just throwing some ideas into the ring

John

 

Interesting idea, thanks for that, but can you describe it a bit better? How do they cut slit into CD plastics to be really narrow? Do you mean like regular slit cut or just CD grooves for recording or something?

[andrew s]

While can use an eyepiece to create a parallel beam for the SA you will then need a second lens/eyepiece to refocus the spectrum onto the ccd. I used two c-thread lenses to do this. The other issue you will have is how to get the star on the slit. This is always the big issue.

Have a look at this site  http://www.astrosurf.com/buil/spectrographs.htm

Regards Andrew

[vlaiv]

5 minutes ago, andrew s said:

While can use an eyepiece to create a parallel beam for the SA you will then need a second lens/eyepiece to refocus the spectrum onto the ccd. I used two c-thread lenses to do this. The other issue you will have is how to get the star on the slit. This is always the big issue.

Have a look at this site  http://www.astrosurf.com/buil/spectrographs.htm

Regards Andrew

Yes, I just figured that one out last night by reading some of the theory explained on astrosurf.

There is really no need for second lens, but then possible resolution is compromised again by converging beam (chromatic coma being the worst according to astrosurf). I wonder how to properly calculate F/ratio of converging beam from eyepiece projection, it might not be so bad after all. I guess proper way would be to divide eyepiece / sensor distance with exit pupil. In above calculation that would be something like F/8.5 which is not bad, it is even greater than original scope's F/8. But that is only if my assumption about beam is correct. Also not sure if exit pupil size changes because of focus position change when using eyepiece projection (not the same for observing / afocal). Not sure how much field curvature is there in eyepiece projection either, but I guess that again I won't be able to focus full spectrum at the same time.

I did think of using c-thread lens because in that configuration it will effectively be Afocal recording, which means that beam will be fully collimated after the eyepiece (c-thread lens focused to infinity) but I wonder how to calculate sensor angstrom/pixel resolution (plate scale is it?) when using such setup. How will distance from eyepiece to camera lens impact magnification? How well will it behave in this configuration if star is moved away from optical axes (then again we have beam although collimated hitting grating at an angle, not sure what sort of aberration this introduces, if any)?

I do have c-thread lens that I've got with CMOS camera and it has adapter to T2 so it can easily be mounted but I know that this lens, although fast, gives rather small fully illuminated circle. It can't even illuminate ASI185 chip and ASI1600 is much much bigger.

Also I wonder how this afocal setup is going to behave in terms of aberrations, and what would be usable FOV, but no idea how to calculate it.

It should how ever work the best in terms of resolution, as it is closest to configuration of classical spectrograph: focal plane (slit) -> collimation lens (eyepiece) -> dispersion element (grating) -> focusing lens (c-mount lens) -> sensor.

I guess now it is time for good old trial and error, and checking different configurations to figure out how each behave.

[andrew s]

My approach is to copy existing designs with some mods depending on what I have or can get. Here is an example from Mr Star Analyser himself  http://www.threehillsobservatory.co.uk/astro/spectroscopy_19.htm

Regards Andrew

PS I intended to say you don't need a slit you can still use it slitless which make it much simpler to find the target. Robin's is like this. The only advantage of the slit is that it defines the resolution of the spectra rather than the seeing disk.

Regards Andrew

[vlaiv]

Yes, I would leave slit aside for now, although "half slit" version could make centering target on slit a bit easier.

According to that web page (which just means I reinvented wheel with all of this ) it looks like proper sized lens is a way to go rather than using small size c-mount lens.

[robin_astro]

Hi Vlaiv,

If  you have not come across them already (thanks for the hat tip Andrew :-)   ), here are my two configurations using the Star Analyser which can incorporate a slit :

SEPSA. This keeps the converging beam but adds a focal plane to place the slit at

http://www.threehillsobservatory.co.uk/astro/spectroscopy_18.htm

Fully collimated "junk box"design (without slit though one could be added

http://www.threehillsobservatory.co.uk/astro/spectroscopy_19.htm

You might also be interested in this grism design (note the mirror slit guider which (almost) all commercial slit spectrographs use thse days)

http://www.burwitz-astro.info/spectrographs/tragos/

and my modification of the similar Shelyak ALPY spectrograph using a 200 l/mm grism

http://www.threehillsobservatory.co.uk/astro/spectroscopy_20.htm

As Andrew said though, adding a slit is the easy part, finding the target, focusing it and guiding it on the slit are the hard parts.  With bright targets though you can get away with defocusing the target into a large blob so it is impossible to miss the slit. (See my Vega spectrum using the SEPSA for example)

 

Cheers

Robin

[vlaiv]

1 hour ago, robin_astro said:

Hi Vlad,

If  you have not come across them already (thanks for the hat tip Andrew :-)   ), here are my two configurations using the Star Analyser which can incorporate a slit :

SEPSA. This keeps the converging beam but adds a focal plane to place the slit at

http://www.threehillsobservatory.co.uk/astro/spectroscopy_18.htm

Fully collimated "junk box"design (without slit though one could be added

http://www.threehillsobservatory.co.uk/astro/spectroscopy_19.htm

You might also be interested in this grism design (note the mirror slit guider which (almost) all commercial slit spectrographs use thse days)

http://www.burwitz-astro.info/spectrographs/tragos/

and my modification of the similar Shelyak ALPY spectrograph using a 200 l/mm grism

http://www.threehillsobservatory.co.uk/astro/spectroscopy_20.htm

As Andrew said though, adding a slit is the easy part, finding the target, focusing it and guiding it on the slit are the hard parts.  With bright targets though you can get away with defocusing the target into a large blob so it is impossible to miss the slit. (See my Vega spectrum using the SEPSA for example)

 

Cheers

Robin

Hi Robin,

Yes, I've had a look at your pages after Andrew posted the link, and my instinct reaction was: "Oh no, I've reinvented the wheel, again ..."

But on the other hand it is good to see that I was not so far off in my reasoning, being new to all of this.

I plan to use OAG for recording spectra, and so far I've been able to achieve something like 0.5" RMS with my setup (might even improve, I'm eyeballing Berlebach Planet to upgrade tripod on my HEQ5 and I might also upgrade default saddle to better one - that should give me additional stability against wind). Not sure if that is good enough, but for some reason I'm thinking about 1-2" slit (which would be 7.5 - 15 um with my scope). Do you think that is anywhere good value?

I'm also trying to figure out some formulae to calculate best setup for both eyepiece projection and afocal method. For eyepiece projection method, I'm wondering if my reasoning for F/ratio of converging beam is sensible - it is eyepiece to sensor distance / exit pupil? Exit pupil will of course be eyepiece FL / scope F/ratio. I'm going to use that value and some formulae from Astrosurf pages to calculate approximate effective resolution (depending on star FWHM and chromatic coma due to converging beam), or rather optimize eyepiece FL / distances to get best recording.

Also for afocal (or as you call it Junk Box arrangement ), I was wondering how to obtain following two values: Star (or slit if later added) size, and spectrum size on chip. For Star size (in terms of FWHM) I was thinking something along the lines:

If we have extended object of 10 arcsecond and we apply x60 magnification with eyepiece (scope fl / eyepiece fl) - it will appear as 10 arcminute object - so collimated beams from each end of object will form an angle of 10 arcminutes when exiting eyepiece. But result of that is somewhat worrying, if I take for example 32mm eyepiece and 2.4" star FWHM, I will end up with star size of 120" or 2' and when I image that with let's say 50mm lens and ASI1600 I will end up with star size on sensor of 7.65 pixels  but my original setup has 5 pixel star image for 2.4" FWHM star - this means that I might loose resolution? Maybe by using shorter FL lens at the end I might get smaller star image, but shorter FL lens will also shorten the spectrum?

That is the other bit of this approach that I'm figuring out how to calculate - I'm thinking to take formula for angle of dispersion for grating with 200 l/m and see which angle am I going to get (at max dispersion for lets say 900nm) and then see how many pixels that is going to cover given that 50 mm lens and 3.8 um pixels provide resolution of 15.7"/pixel.

I have so many questions

If there is collimated beam on grating but at an angle (not fully perpendicular but at an angle) will that introduce aberrations? Do I need to keep star centered on optical axes? This will in effect half available sensor size to record spectrum on.

Cheers,

Vlad

[Merlin66]

Vlad,

Check out the SimSpec and TransSpec design spreadsheets available on the website:

http://www.astronomicalspectroscopy.com/

I think you're trying to over complicate things........

[SteveBz]

I'm sure I saw a cheap french spectrometer for about 100 euros on this site, but I can't see it now. It looked a bit like an external hard disk drive and the guy was making them himself.

Does anyone have the link?

I'm also trying to make a spectroscope with the eBay 4 Pound diffraction grating from China 500 or 1000 lines per mm.

https://www.ebay.co.uk/p/Diffraction-Grating-Slide-Linear-1000-Lines-mm-Holographic-Physics-Spectrum/572461349

 

However 100 euros probably saves a lot of heartache.

Steve.

[Merlin66]

Vlad/ Steve,

where did you end up on these projects????

[SteveBz]

Just made this today for testing tonight.

 

[Merlin66]

I'll be interested to see how you go....

Most of those cheap (and high l/mm) gratings are un-blazed and very inefficient.

The Paton Hawksley gratings are much better.

 

[SteveBz]

It is a bit high, 500 l/mm.

I might need to saw the end off my nose piece to reduce the distance to the sensor. I'm planning to use a DSLR because of its wide chip.

Had to look up unblazed:

http://www.vikdhillon.staff.shef.ac.uk/teaching/phy217/instruments/phy217_inst_blaze.html

Are the Star Analysers blazed?

 

[Merlin66]

Steve,

Yes, both the SA100 and SA200 are blazed for 550nm.

Far more light into the spectrum, less into useless zero image.

 

 

[SteveBz]

23 minutes ago, Merlin66 said:

Steve,

Yes, both the SA100 and SA200 are blazed for 550nm.

Far more light into the spectrum, less into useless zero image.

 

 

So my plan for tonight is to use the zeroth order spectrum to help with focusing. In order for it to work at all, the focal plan needs to be on the grating, and I thought I'd use the zeroth order for that with my bahtinov mask to help, then orient the grating and the DSLR sensor to make sure the first order spectrum is on the chip, finally, with the handset, to try to move the zeroth order point to the edge of the image and off the chip and try to take a prolonged exposure of Vega, which will be up and calling out for attention.

What do you think?

[Merlin66]

Focus on the zero order will get you close, but for success you really need to focus on the spectrum.

[SteveBz]

I just tried it with the Sun, and when I panned off centre a little, with the sun at the edge of the field of vision, if I held my head at a certain angle, stood on one leg and hummed "Clementine" to the tune of "Waltzing Matilda" I could see a continuous spectrum.  But as you say, quite faint.

I'll try with Vega this evening, but I may need to stand on the other leg!!

[SteveBz]

I don't have my pc here so I'll download and process in the morning. This is Vega.

 

[SteveBz]

Ok, so here it is in all its glory.  I suspect I was out of focus and that the double-strip appearance is due to the central support on the secondary mirror running down the length of the spectrum.  This is a five-second exposure on a Celestron 114. 

The optical train is:

OTA->2x Barlow->Nosepiece->500l/mm Grating on inside of nosepiece->Nikon D5000.

Vega was in focus, but as you suggested, I think that the spectrum was not.  I'm not sure whether I could just have adjusted the focuser, or whether the grating was just too close to the sensor, I suspect the latter.  I wonder if I shaved the rim off the filter attachment whether I could screw the grating further into the nose-piece and therefore create an adjustable positioning?

What do you think overall, guys?  What are my options?  I've spent less than a tenner so far, in any currency.

Regards

Steve.

[Merlin66]

Yeah,

you need to focus on the spectrum. Adjust the focuser.

 

[robin_astro]

Hi Steve,

Lose the Barlow. It is only making the star image bigger and losing you resolution

The Star Analyser etc are only 100-200l/mm for a reason.  At 500l/mm the  angle and curvature of the spectrum is so large where it hits the camera sensor that there is a big change in focus from one end of the spectrum to the other.  (There are other aberrations in the converging beam arrangement  that make lower resolution gratings at larger distances work better too. ) 

If you want to try building a "Star Analyser on the cheap" you could go for the Paton Hawksley school 100 l/mm grating that I developed the Star Analyser from.

http://www.patonhawksley.co.uk/transmissiongratings.html

(TE218)

http://www.threehillsobservatory.co.uk/astro/spectroscopy_jeulin.htm

It is not quite as efficient as the SA100 and can vary in quality as the QC is not as strict but will perform much better than the grating you have. (Note it is  on glass so cannot be cut like the cheap plastic film gratings so you will need to make a mount for it. This is what I used for a web/video cam.

http://www.threehillsobservatory.co.uk/astro/spectroscopy_3.htm

Aim for a larger spacing than this though, giving around 10A/pixel as a starting point. Use the PH calculator as a guide

http://www.patonhawksley.co.uk/calculator/

Robin

 

[Stub Mandrel]

On 16/08/2017 at 13:18, andrew s said:

My approach is to copy existing designs with some mods depending on what I have or can get. Here is an example from Mr Star Analyser himself  http://www.threehillsobservatory.co.uk/astro/spectroscopy_19.htm

You trust a man who labels his graphs in Angstroms?