Page 2 of 9

Re: First starlight for my new echelle

Posted: Sun Jun 18, 2017 8:59 am
by Andrew Smith
Hi Tim, I was looking at your flat blaze plots and at one level they are as I would expect. If the flat is indeed flat across the order (which it will be more or less) then you would expect more noise to show at the peak and drop off as the attenuation from the falling efficiency kicks in. You would also expect sampling a wider aperture to smooth it out. What does the noise look like after the blaze is removed?

However, your noise is not symmetric about the peek of the blaze being stronger on the blue wing. Is this the same across all orders or just the one shown?
It made me think it could be either PRNU or possibly fringes. Another possibility is that the trace of the order is not so accurate on the blue side.
This could all irrelevant as you are getting good results and that is what counts.

Still very impressed with your work and the way you made the folding mirror. I did look at these slitshttps://lenoxlaser.com/shop/optical-apertures/standard-slits/glass-air-slit/ with the idea of getting one coated but went the fiber route in the end.

Regards Andrew

Re: First starlight for my new echelle

Posted: Sun Jun 18, 2017 9:53 am
by Christian Buil
Hi Tim,

Using a wide slit with ISIS (eShel mode) to make flat-field is not usual in fact (ISIS search to subtract the inter-order background). The problem can come from that.

How you exploit MaximDL for extract geometrical distorted flat traces.

Can you display a picture of the flat-fied or provide one. A spectral lamp image would also be appreciated (you use one, model?).

Christian

Re: First starlight for my new echelle

Posted: Sun Jun 18, 2017 11:02 am
by Olivier GARDE
Hi Tim,
Congratulation for this spectrograph, very nice work specially the hole in the mirror....

Have you try to mesure the efficiency of your spectrograph ?

I Think the design of your spectrograph is more efficiency (with a slit) than a fiber spectrograph du to losting flux in the fiber, but I think there are 2 disadvantages in fixing this kind of spectrograph directly on the telescope :
- No thermal regulation (necessary when measuring very small radial velocities, for exo planet search : 0,1 deg celcius is 100m/s velocity shift)
- Mechanical bending problems depending on the position of the telescope, can shift the wavelength during exposure.

Did you try to mesure this kind of mechanical shift vs telescope position ? For example taken Calibration lamp at different position of the telescope and compare the shift between each position ?

Re: First starlight for my new echelle

Posted: Mon Jun 19, 2017 1:51 am
by Tim Lester
Hi all,

Thanks for all the feedback. I'll try and answer all the questions.
First Christian:
With MaximDL I have been looking at the flat_##.fit image compared to the flat_##.dat profile that ISIS generates not the raw distorted flat image.
I do not think the problem is the inter-order background subtraction.
Here is the profile in the cross dispersion direction for flat _32.fit. The inter-order region is quite wide and very near zero so ISIS has done a good job of removing the inter-order background.
order32 separation.PNG
and here is a profile generated in MaximDL - it is smooth, no ripple.
order32 maxim profile.PNG
With the eShel mode I always get ripple no matter what binning I use all the way from 8 (the minimum allowed) to 20. Larger than 20 and the orders either side start to get included.
Here is the ISIS plot for a binning of 10 and 18 compared. They are identical.
order32_ISIS.PNG
However if I process the flat_32.fit image in ISIS as though it were a regular spectral image and generate a profile I get smooth profiles just like the Maxim profile.
This is the case with a "sky" subtraction zone set in the inter-order region or without "sky" subtraction.
order32_ISIS_2.PNG
order32_ISIS profile.PNG
I believe there must be a problem with the eShel processing. Could it be with the optimum binning routine? Unlike with regular spectra there is no option to turn this off in eShel mode.
At the moment I can't think of anything else that might explain the ripple.
I find the ripple quite concerning because it does not appear to be in the data. This may be happening to the star spectra as well as to the flats but just not visible for most spectra due to low S/N.

I had posted a link to my working directory but realize that was a rather large download. Here is a link to just the flat and relco images as well as flat_32.fit, falt_32.dat and also the full log so you can identify orders: https://1drv.ms/f/s!AvXqhKI5kEGRlhO4W57pp4yk6pWN

Now Olivier:
I realized before I started construction that I will never get the spectral stability that is possible with a fiber fed spectrograph but I wanted the better efficiency (not measured yet) and hence S/N. A slit also allows one to design a higher resolution spectrograph when restricted to a 25x50 mm grating.
I know form experience with my last R=9000 spectrograph (which is very similar mechanically) that the main flexure problem is at the camera to lens mounting. The front plate on the QSI 583 housing is quite thin and it flexes. It is not shown in the photograph but there will be some cross bracing on either side of the camera down to the spectrograph. I just haven' completed it yet and was anxious to give the instrument a try. My R=9000 spectrograph did not have this cross bracing and I achieved about 1 km/sec repeatability, So I am hopeful. I plan on trying to control temperature to better than 1 degree C over the integrations for one object but certainly not to 0.1 C !
I should point out that in the end spectral shifts due to differing placement of the star image on the slit can be a large source of errors (>1 km/sec). A fiber feed avoids this problem since the signal is averaged over the exit end of the fiber - but only if the fiber modes are randomized (by constant random bending of the fiber for example).

Tim

Re: First starlight for my new echelle

Posted: Mon Jun 19, 2017 3:21 pm
by Joan Guarro Flo
Hi Tim,

I hope that will soon find a solution to these problems.

There is a question that has me intrigued. What material is the slit? What is its origin? The've bought in a shop? What reference number? Te've made yourself? How ?

Many thanks, Joan.

Re: First starlight for my new echelle

Posted: Mon Jun 19, 2017 9:05 pm
by Tim Lester
Hi Joan,

The slit was the hardest part. In the end I was able to get it fabricated as a favour by the Nanofabrication Lab at our national research facility here in Ottawa. My career was in semiconductor fabrication and I had contacts there.
The substrate was a 1mm thick glass plate with antireflection coatings both sides. NanoFab applied an aluminum film with the 23x75 um aperture. This was done with e-beam lithography and the aperture is accurate to 0.1 um which was way better than I actually needed!

I had been looking at National Aperture's reflective slits as a possible source: thttp://www.nationalaperture.com/reflective.htm
but I never actually made any inquires so don't know what the cost would be.

Tim

Re: First starlight for my new echelle

Posted: Mon Jun 19, 2017 10:21 pm
by Christian Buil
Hi Tim,

I think ISIS problem fixed.

The difficulty arises from the fact that ISIS can not find exact center of orders trace. ISIS eShel function is designed to work with optical fibers, i.e. narrow order. You are using a wide slit along the spacial axis and ISIS is here struggling.

It is easy to simulate your default from the command line (Tools tab). You can use your flat_32.fit image. Enter the command:

l_opt flat_32 p 25 33

(25 - 33) are the binning coordinate along Y-axis.

Display the p.fit generated profile. You can observe the ripple.
L_OPT is the function used for eShel processing, but the problem is the same is you use a simple binning:

l_add flat_32 p 25 34

If the binning zone is centered on the trace, all is ok:

l_add flat_32 p 32 38

or better, is you include all the trace signal (similar to your maximDL test):

l_add flat_32 p 28 46

This edge problem on the trace comes from the interpolation algorithm when geometric correction is applied (banding correction), here a spline interpolation (cardinal spline) that does not work perfectly in your case. It is easy for me to provide you a ISIS version with bilinear interpolation, less sophisticated, but which will produce less artefact.

But before, it is important to verify that ISIS does not correctly find the trace of the orders. I recommend that you look the check.fit and check2.fit images and control the overlay of the trace found by calculation and the actual trace (select "Rotation test" before). Details are here:

http://www.astrosurf.com/buil/isis/eshe ... chelle.htm

Try also to modify the "Order detection threshold".

Give me the result, the solution is close!

Christian

Re: First starlight for my new echelle

Posted: Tue Jun 20, 2017 1:20 am
by Tim Lester
Christian,

Thanks a lot for your help so far! We are close.
I had initially used the check,1,2.fit images to make sure ISIS was finding the orders correctly - it was.
I repeated this and confirmed that ISIS is finding the center of the orders perfectly and fitting a very good line. I did this for a variety of detection thresholds, always with excellent results.
In any case the straightened order images (flat_##.fit) that ISIS produces are all perfectly centred in y. This must be confirmation.

I do not think the problem lies with ISIS not correctly determining the center of the orders. It seams to be doing a very good job.

However,
When I process with a binning of 12 then flat_32.dat shows ripple. A binning of 12 is more than enough to capture the whole order in the y direction.
(Note that I had accidentally uploaded a double size image for flat_32.fit - sorry! The link: https://1drv.ms/f/s!AvXqhKI5kEGRlhO4W57pp4yk6pWN
now has the correct flat_32.fit image. I have also uploaded a tung_check overlay.)

The center of the order 32 profile in the y direction is 18.
IF I process flat_32.fit with the command: l_opt flat_32 p 12 24 (a binning of 12 centered on 18) then I get the identical rippled profile (perfect match).
BUT if I process with the command: l_add flat_32 p 12 24 I get a well matched profile but with NO ripple.

Surely this means that the optimized binning is causing the problem and not the interpolation method.

Tim

Re: First starlight for my new echelle

Posted: Tue Jun 20, 2017 10:12 pm
by Christian Buil
Tim,

You can download here a complete version of an intermediate update (unoficial V5.8.2), test and give me the status :

http://www.astrosurf.com/buil/isis/down ... _5_8_2.zip

Christian

Re: First starlight for my new echelle

Posted: Wed Jun 21, 2017 2:59 am
by Tim Lester
Christian,

Thank you very much.
Everything works perfectly with the new version :D

Tim