Motivated by the book “Using Commercial Amateur Astronomical” by J.L. Hopkins I got hold of the Science Surplus Spectrometer, which provides a digital spectrum (2048 points) from the light received through a optical fiber. The spectroscope (USD 200) is shipped from the US and had to go through German customs, which actually applies a specific custom tariff for spectroscopes. Now the work started with a rather long and colorful road to achieve first spectral light:
The main issue is the small amount of available stellar light! This requires to hook the spectrometer up to a telescope, and even then we will only be able to analyze the brightest stars.
The spectrometer spreads the stellar light across 2048 pixels, which corresponds to roughly reducing the light by 2000 times compared to registering the light with a normal camera on a single “dot”. In addition there are losses from coupling the light out of the telescope into the fiber and the spectroscope.
Calibrate the spectrometer (well explained in the manual) for instance using a fluorescent lamp and save the spectrum to file to get the 1..2048 range converted to nm.
I used this information later to write my own data acquisition and image processing program.
Get a serial to USB cable (long enough to be later useful to hook up the telescope outside) and get Windows XP running in a virtual machine
to test the spectrometer needs to be connected to the serial interface and the accompanying Windows software needs to run on my Macintosh compute (VirtualBox with an old Windows XP came to the rescue, some fiddling with the settings required to make it recognize the USB serial adapter as valid COM port)
Now the tricky part: “how to get starlight out of the telescope and into the spectroscope”?
Hopkins suggest to drill a hole in a diagonal mirror, but that option looked too tricky to me and I wanted to go for the beam splitter option. But how to avoid that half of the precious star light gets waisted?
I found a nice presentation and solution by fellow astronomer J. Hubbell, who describes his “cold mirror” solution (and gives very helpful advice about the best collimation into the fiber, more about that later)
Building a cold mirror (materials: about EUR 200)
This requires to repurpose some other equipment. My part list from any astronomy shop:
Skywatcher or Telescope Service 1.25″ flip mirror which can be easily taken apart (but beware, the screws are small and not very sturdy)
several adapters: one to put in other 1.25″ equipment (and eventually a camera, which has a C-mount thread requiring a second adapter)
be aware that the relatively long optical path added by the flip mirror requires to shift-back the focus by 7-8 cm! This worked for my Newton telescope only, because the tube can be shortened.
finally: the FM03 – 1″ Visible Cold Mirror, AOI: 45°, 1 mm Thick, which I ordered from Thorlabs to be put into the filter CP02/M mount
more adapters: S120-SMA Fiber Adapter Cap with Internal SM1 (1.035″-40) Thread, and SM1A10 Adapter with External SM1 Threads and Internal C-Mount Threads
the C-mount thread is then screwed into a short 1.25″ C-mount adapter (I ordered this short C-Mount adapter from the UK) and on the other end a 0.5 focal reducer lens for collimation.
finally, some do-it-yourself work with a fret saw (and lots of attempts to get the right measures) to replace the supplied flip mirror with the mounted cold mirror, attached with M6 plastic screws on a 1mm thick wooden plate with a circular hole cut out.
Note that the final setup will NOT allow you to flip the mirror anymore and that in the through-direction the intensity is 90% reduced (the transmission/reflectance characteristics is available at Thorlabs site)
To be continued… image processing and computer guiding the telescope with the star centered on the fiber for minutes.
Just came in: spectra of Mirach
Machine learning techniques (“neural networks”) are presently explored in a wide range of applications, with the standard showcase being image recognition. In most scenarios the input data is “user generated” (for example handwritten digits) or comes from automatic sensors. The neural network gets trained with (Key–Value) pairs which are often tagged before used for “supervised learning”.
But what if we want to apply machine learning to scientific data sets generated by demanding simulations for instance on supercomputers? In that case, the input data does not come “for free”, but is the outcome of state-of-the art simulations, for instance of the optical properties of photosynthetic complexes, discussed before.
The advantage of the Machine Learning technique in this case is that the input parameter are known and the training works with very reliable information. This allows one to find very small-sized (in terms of storage) neural network representations of huge data sets (several gigabytes). We (Rodriguez & Kramer 2019, arxiv version) have explored this method for encoding the information of “two-dimensional optical spectra” and to relate the spectra to the molecular structure, such as the dipole orientations and the fluctuating energy states.
From a “physics perspective”, machine learning provides a way of automatic parameter fitting and could be seen as minimizing a variational parameter space. The problem: a variational principle always gives happily an answer, even if that answer is wrong. While we cannot solve this problem, we have studied how good different network layouts perform under the constraint of fixing the number of fitted parameters. This determines the size of the resulting parameter file of the network, which becomes surprisingly small. You can explore it by downloading the ancillary data we deposited on the arxiv.
The lunar eclipse yesterday (July 27, 2018) provided an excellent opportunity to take pictures and to study the color of Earth’s shadow. I snapped some images with my smartphone (Huwai Mate 10 Pro, Leica camera) through a 130mm Newton telescope. The moon was pretty low and within the city barely visible. Just after leaving the central shadow cone, the bluish color of the umbra/penumbra region became very noticeable. Today, I analyzed the picture a bit and drew the relative red/green/blue components along the shadow. Indeed blue wins within a narrow stripe bordering the penumbra. According to the literature, this is a signature of the ozone layer (which absorbs orange and red light, while blue passes and arrives at the moon).