Chemistry professor Alexander Scheeline at the University of Illinois made a spectrometer from a mobile phone to captivate students with analytical chemistry.
The professor assembled the chemist's basic scientific tool from inexpensive materials and a digital camera. Spectrophotometry is one of the most widely used means for identifying and quantifying materials. If, for example, you need to measure the amount of protein in meat, water in grain or iron in blood, you need spectrometer.
A student cannot evaluate the performance of spectrophotometry if he uses the mysterious "box" of a laboratory spectrometer. He does not understand what is happening inside and simply changes samples and records the results, - explains Alexander Shchilin. - It doesn't help the educational process. If you want to teach someone to use the tool creatively and improve it, you need something simpler and clearer. "
Figure: 1. This is all you need to make a spectrometer.
If you want to draw attention to the shortcomings of the instrument, it is much easier when these shortcomings are very large and are not compensated by the complexity of the devices and settings, "explains Alexander Shchilin.
In a spectrometer, white light passes through a sample of material that absorbs specific wavelengths of light. Then a diffraction grating decomposes the light into colors, and chemists can analyze the spectrum, determining the properties of the sample.
Figure: 2. Assembled spectrometer. The LED shines through the cuvette directly opposite the grate, which is secured with transparent tape.
As a light source, Professor Shchilin used one light-emitting diodepowered by a 3-volt battery. It is not difficult to buy a diffraction grating and sample cuvettes in the USA, and in the end all the equipment costs less than $ 3. It remains to find a suitable digital camera, and then the scientist remembered that every schoolchild and student has a mobile phone. After that, it remains only to solve the problem of data processing. To do this, the professor wrote a program for analyzing spectra from photographs in jpeg format and put it on the Internet along with the source codes.
For the first time, Alexander Shchilin demonstrated his invention while working on an exchange program in Hanoi (Vietnam). The Vietnamese students had no experience with scientific instruments, but enthusiastically set about experimenting with a cell phone spectrometer.
Figure: 3. A mobile phone will not replace an accurate spectrometer in serious scientific research, but not every student has $ 3,000 in pocket money for a hobby.
In the United States, a professor used a homemade spectrometer during his high school class. By the end of the 45-minute lesson, the students have learned things that elude most textbook-only students. For example, one student asked about the effect of scattered light on a camera's sensitivity and ability to read spectrum.
A senior pupil who knew almost nothing about spectrophotometry an hour ago discovered the main problem of all spectrometers, says Alexander Shchilin. - Since I started teaching, I have tried to explain to my students the concept of the effect of scattered light on a spectrometer and the effect of this problem on the performance of equipment. And suddenly I saw how the student himself understood the essence of this problem and asked me the right question! "
The scientist is happy to share his invention with school teachers and university professors at various seminars and using the Internet. He hopes that his invention will be improved, for example, will write an image processing program for smartphones, which will eliminate the need to use a computer. A mobile phone spectrometer can captivate a lot of people with analytical chemistry, which seems to many to be a complex and incomprehensible science. However, the invention of Alexander Shchilin demonstrates that the innate curiosity of a person can be easily awakened - it is enough to offer simple, understandable and exciting creative experiments.
Once I read an article on Wikipedia about a Fourier spectrometer, and I wanted to make one myself. This task is not at all simple, but we managed to make a working model of the spectrometer. I will warn you right away - this is not an infrared spectrometer, so they will not make especially interesting measurements.
About how a Fourier spectrometer works, and how it can be made at home - further (carefully, a lot of pictures!).
A bit of theory
Just in case - we will talk about optical spectrometers.I will try not to go deeply into the theory of spectrometers, although this topic is very broad.
The most common types of spectrometers are spectrometers with a dispersive element, which is capable of distributing radiation of different wavelengths in space. Diffraction gratings and prisms are examples of such elements.
Simplified diagram of a spectrometer with a semitransparent diffraction grating: 
In the diagram: 1 - entrance slit, 2 - collimating lens, 3 - diffraction grating, 4 - focusing lens, 5 - image plane (photodetector plane).
The investigated radiation passes through the entrance slit, is converted by objective 2 into a parallel light beam, which falls on the diffraction grating. The grating produces spatial separation of this beam - radiation with different wavelengths begins to propagate at different angles. The focusing lens 5 forms an image from the parallel beams in the plane 5, which can be recorded by a photodetector (for example, a CCD ruler).
These types of spectrometers are relatively simple, but they have their drawbacks.
One of the parameters influencing the spectral resolution of the spectrometer is the slit width - the smaller it is, the better the resolution. However, as the slit size decreases, the illumination of the photodetector decreases, which complicates the acquisition of spectra. Since the light is distributed over the image plane, the illumination of the image drops.
Prisms, although simple to manufacture and use, are not capable of providing high spectral resolution. Another drawback is that they can only work in a certain wavelength range determined by the prism material. Ordinary glasses are not capable of transmitting radiation with a wavelength longer than 3-4 microns.
Diffraction gratings are more difficult to manufacture, but provide much better spectral resolution. Reflective diffraction gratings can be used in a wide range of wavelengths - from ultraviolet radiation to far infrared. One of the disadvantages of diffraction gratings is that they give several different orders of the spectrum, which can distort the interferogram. To eliminate them, one has to use light filters that limit the radiation spectrum at the input or output of the spectrometer.
To record the spectrum in the image plane of the spectrometer, multi-element photodetectors are installed, which make it possible to very quickly read the entire radiation spectrum. The most common silicon CCD and CMOS arrays are only suitable for VIS and NIR. To study radiation longer than 1.2 microns, receivers made of other materials are needed, for example, germanium, indium gallium arsenide, or even a line of microbolometers. Such multi-element receivers are produced by only a few companies in the world, they are very expensive and difficult to obtain.
To record spectra, one can use cheaper single-element photodetectors (photodiodes, bolometers), but in this case the image scanning must be mechanical - by moving one of the spectrometer nodes. This significantly increases the time taken to obtain the spectrogram and can reduce the accuracy of measuring the absolute values \u200b\u200bof wavelengths.
Fourier spectrometers use a completely different principle of operation - it is based on the phenomenon of interference.
In Wikipedia, as it seemed to me, the most simple and understandable description is given:
The main element of a Fourier spectrometer is a Michelson interferometer.
Let's say we have a coherent radiation source with a certain wavelength. When the path difference between the two beams arriving at the receiver is equal to λ / 2 (that is, the rays arrived in antiphase), the light intensity recorded by the receiver is close to zero. When the right mirror of the Michelson interferometer is moved, the difference in the path of the rays changes, and the intensity of the light recorded by the receiver also changes. Obviously, the light intensity is maximum when the difference in the path of the rays is a multiple of the wavelength.When the mirror is moved at a constant speed, a sinusoidal electrical signal will be observed at the output of the receiver. Moreover, the period of the sinusoid depends on the wavelength of the source, and the amplitude on the intensity of the source.
Now imagine that the input is an incoherent source. Each wavelength in the spectrum of the light source will give its own sinusoid at the output of the receiver. Thus, at the output of the receiver, we get a complex signal. When performing the inverse Fourier transform on the received signal, we obtain the spectrum of the input electrical signal, which is also the emission spectrum of the source (that is, the radiation intensity of the source at different wavelengths).
Scheme of radiation interference in a Fourier spectrometer:
In the diagram: 1 - radiation source, 2 - beam splitting (semitransparent) plate, 3 - fixed mirror, 4 - movable mirror, 5 - photodetector.
The layout of a real spectrometer is somewhat more complicated:
In the diagram: 1 - radiation source, 2,4 - collimating optics, 3 - entrance diaphragm, 5 - fixed mirror, 6 - movable mirror, 7 - mirror drive, 8 - beam splitter, 9 - reference channel laser, 10 - reference photodetector channel, 11 - focusing optics, 12 - signal photodetector.
In order to stabilize the speed of movement of the movable mirror, and to ensure the "binding" of the spectrometer to the absolute values \u200b\u200bof wavelengths, a reference channel is introduced into the spectrometer, consisting of a laser and its photodetector (9 and 12 in the diagram). In this case, the laser acts as a wavelength standard. High-quality spectrometers use single-frequency gas lasers for this purpose. As a result, the wavelength measurement accuracy is very high.
Fourier transform spectrometers have other advantages over classical spectrometers.
An important feature of Fourier spectrometers is that when even one photodetector is used, all spectral elements are simultaneously recorded, which gives an energy gain in comparison with element-by-element mechanical scanning (Falgett's gain).
Fourier transform spectrometers do not require the use of optical slits, which retain most of the luminous flux, which gives a large gain in luminosity (Jacquinot gain).
In Fourier spectrometers, there is no problem of overlapping spectra, as in spectrometers with diffraction gratings, due to which the spectral range of the investigated radiation can be very wide, and is determined by the parameters of the photodetector and the beam splitter plate.
The resolution of Fourier transform spectrometers can be much higher than that of traditional spectrometers. It is determined by the difference in the travel of the movable mirror Δ. The resolved wave interval is determined by the expression: δλ \u003d λ ^ 2 / Δ
However, there is also an important drawback - the large mechanical and optical complexity of the spectrometer. For interference to occur, both interferometer mirrors must be very accurately aligned perpendicular to each other. In this case, one of the mirrors must perform longitudinal vibrations, but the perpendicularity must be maintained with the same accuracy. In high-quality spectrometers, in some cases, to compensate for the tilt of a movable mirror during movement, a stationary mirror is tilted using piezoelectric actuators. To obtain information on the current tilt, the parameters of the reference beam from the laser are measured.
Practice
I was absolutely not sure that it was possible to make a Fourier spectrometer at home without having access to the necessary machines (as I mentioned, mechanics is the most difficult part of the spectrometer). Therefore, the spectrometer was built in stages.One of the most important parts of the spectrometer is the fixed mirror assembly. It is he who will need to be adjusted (smoothly moved) during the assembly process. It was necessary to provide the ability to tilt the mirror along two axes, and precisely move it in the longitudinal direction (why - lower), while the mirror should not tilt.
The base of the fixed mirror assembly is a single-axis stage with a micrometer screw. I already had these nodes, it was only necessary to connect them together. For a backlash-free connection, I used a simple clamping of the stage to a micrometer screw with a spring located inside the base of the stage.
I made this using three adjustment screws removed from a broken theodolite. A metal plate with a glued mirror is pressed by springs to the ends of these screws, and the screws themselves are fixed in a metal corner screwed to
table.
The design is clear from the photos:
The mirror adjustment screws and the micrometer screw are visible.
The mirror itself is visible in front. It is taken from a scanner. An important feature of the mirror is that the mirror coating must be in front of the mirror, and in order for the interference lines to be not crooked, the mirror surface must be of a fairly high quality.
View from above:
You can see the springs pressing the table in the micrometric screw and the fastening of the plate with the mirror to the corner.
As you can see from the photographs, the fixed mirror assembly is attached to a chipboard board. The wooden base of the interferometer is clearly not the best solution, but it was problematic to make it out of metal at home.
Now you can check the possibility of obtaining interference at home - that is, to assemble the interferometer. One mirror already exists, so a second test mirror and a beam splitter must be added. I had a beam splitting cube, and I used it, although the cube in the interferometer works worse than the beam splitter plate - its edges give additional reflections of light. The result is the following construction:
On one of the sides of the cube, not facing the mirror, you need to direct the light, and through the other you can observe the interference.
After assembly, the mirrors are not positioned too perpendicularly, and therefore an initial alignment must be performed. I did it with a low-power laser diode connected to a collimating lens of a fairly large diameter. A very small current must be applied to the laser - such that one can look directly at the crystal. The result is a point light source.
The laser is installed in front of the interferometer, and its reflections in the mirrors are observed through the cube. For ease of observation, I attached a prism to the cube, directing the radiation coming out of the cube upward. Now, turning the mirror adjustment screws, you need to combine the two visible laser reflections into one.
Unfortunately, I do not have photos of this process, and it does not look very clear - because of the glare in the cube, you can see many luminous points. Everything becomes much clearer when you start turning the adjusting screws - some of the points begin to move, and some remain in place.
After the mirrors have been aligned in the manner described above, it is enough to increase the laser power - and here it is, interference! It looks almost the same as in the photo at the beginning of the article. However, it is dangerous to observe laser radiation with your eyes, so to see the interference, you need to install some kind of screen after the cube. I used a simple piece of paper through which you can see the interference fringes - the power and coherence of the laser is enough to create a sufficiently contrasting image. By turning the mirror adjustment screws, you can change the width of the stripes - it is obvious that too narrow stripes are problematic to observe. The better the interferometer is aligned, the wider the fringes. However, as I have already mentioned, the slightest deflection of the mirrors leads to misalignment, and therefore, the lines become too narrow and indistinguishable. The sensitivity of the resulting interferometer to deformations and vibrations is enormous - just press the base board anywhere, and the lines begin to move. Even footsteps in a room cause lines to shake.
However, the interference of coherent laser light is not yet what is needed for a Fourier spectrometer to work. Such a spectrometer should work with any light source, including white. The coherence length of white light is about 1 micron.
For light-emitting diodes, this value can be higher - several tens of micrometers. The interferometer forms an interference pattern only when the difference in the path of the light beams for between each of the mirrors and the beam splitter is less than the radiation coherence length. For a laser, even a semiconductor one, it is large - more than a few millimeters, so interference occurs immediately after aligning the mirrors. But even from the LED, it is much more difficult to get interference - by moving the mirror in the longitudinal direction with a micrometric screw, you need to ensure that the difference in the path of the rays falls into the desired micron range.
However, as I already said, when moving, especially large enough (hundreds of microns), due to insufficiently high-quality mechanics of the stage, the mirror can turn slightly, which leads to the fact that the conditions for observing the interference disappear. Therefore, it is often necessary to reinstall the laser instead of the LED and correct the mirror alignment with screws.
In the end, after half an hour of trying, when it already seemed that it was not at all realistic, I managed to get the interference of light from the LED.
As it turned out a little later, instead of observing the interference through a piece of paper at the exit of the cube, it is better to install a matte film in front of the cube - this is how it turns out extended light source... As a result, the interference can be observed directly with the eyes, which greatly simplifies observation.
It turned out like this (you can see the reflection of the cube in the prism):
Then we managed to get interference in white light from the LED flashlight (the photo shows a matte film - it is facing the camera and you can see a dim spot of light from the flashlight on it):
If you touch any of the mirrors, the lines begin to move and fade until they disappear completely. The period of the lines depends on the radiation wavelength, as shown in the synthesized picture found on the Internet:
Now that the interferometer is made, we need to make a movable mirror assembly to replace the test one. Initially, I planned to simply glue a small mirror to the speaker, and apply current to it to change the position of the mirror. The result is the following construction:
After the installation, which required a new alignment of the fixed mirror, it turned out that the mirror swings too much on the speaker diffuser and warps it somewhat when current is applied through the speaker. However, by changing the current through the speaker, it was possible to smoothly move the mirror.
Therefore, I decided to make the design more robust, using a mechanism that is used in some spectrometers - a spring parallelogram. The design is clear from the photo:
The resulting unit turned out to be much stronger than the previous one, although the rigidity of the metal plate-springs came out somewhat high.
On the left is a hardboard board with a diaphragm hole. Protects the spectrometer from external light.
A collimating lens is installed between the hole and the beam splitting cube, glued to the metal frame:
A special plastic holder is visible on the frame, into which you can insert a matte film (located in the lower right corner).
Installed lens for photodetector. A small mirror on a swivel mount is installed between the lens and the cube. It replaces the previously used prism. The photo at the beginning of the article was taken through him. When the mirror is turned to the observation position, it overlaps the lens, and the registration of the spectrogram becomes impossible. In this case, it is necessary to stop sending a signal to the speaker of the movable mirror - because of too fast oscillations, the lines are not visible to the eye.
Another one-axis table is visible in the lower center. Initially, a photo sensor was fixed on it, but the table did not give any special advantages, and later I took it off.
I installed a focusing lens from the camera in front:
To simplify the alignment and testing of the spectrometer, I installed a red photodiode near the diaphragm.
The diode is mounted on a special swivel holder, so that it can be used as a source of test radiation for the spectrometer, while the light flux from the objective is blocked. The LED is controlled by a switch installed under the holder.
Now it's worth telling a little more about photo sensors. It was originally planned to use only one common silicon photodiode. However, the first attempts to make a high-quality amplifier for the photodiode turned out to be a failure, so I decided to use the OPT101 photosensor, which already contains an amplifier with a conversion factor of 1,000,000 (1 μA -\u003e 1V).
This sensor worked quite well, especially after I removed the aforementioned table and adjusted the height of the sensor precisely.
However, a silicon photodiode can only receive radiation in the 400-1100 nm wavelength range.
The absorption lines of various substances usually lie farther, and a different diode is needed to detect them.
There are several types of photodiodes available for NIR applications. For a simple homemade device, germanium photodiodes are most suitable, capable of receiving radiation in the range of 600 - 1700 nm. These diodes were produced back in the USSR, so they are relatively cheap and affordable.
Photodiode sensitivity:
I managed to get photodiodes FD-3A and FD-9E111. In the spectrometer, I used the second one - it has a slightly higher sensitivity. For this photodiode, we still had to assemble an amplifier. It is made using the TL072 operational amplifier. In order for the amplifier to work, it was necessary to supply it with a voltage of negative polarity. To obtain this voltage, I used a ready-made DC-DC converter with galvanic isolation.
Photo of a photodiode with an amplifier:
On both photodiodes, the beam of light from the interferometer must be focused. A beam splitter could be used to separate the light from the lens, but this would attenuate the signals from the diodes. Therefore, after the lens, another swivel mirror was installed, with which you can direct the light to the desired diode. The result is the following photosensor assembly:
In the center of the photo is the lens, on top of it is the reference channel laser. The laser is the same as in the rangefinder, taken from the DVD drive. The laser begins to form high-quality coherent radiation only at a certain current. In this case, the radiation power is quite high. Therefore, in order to limit the power of the beam, I had to cover the laser lens with a light filter. On the right, there is a sensor on OPT101, at the bottom - a germanium photodiode with an amplifier.
In the reference channel for receiving laser radiation, an FD-263 photodiode is used, the signal from which is amplified by an LM358 operational amplifier. In this channel, the signal level is very high, so the gain is 2.
The result is the following construction:
Under the test LED holder is a small prism that directs the laser beam towards the photodiode of the reference channel.
An example of an oscillogram obtained from a spectrometer (a white LED serves as a radiation source): 
The yellow line is the signal fed to the movable mirror speaker, the blue line is the signal from the OPT101, the red line is the result of the Fourier transform performed by the oscilloscope.
Software part
Without software processing, a Fourier spectrometer is impossible - it is on a computer that the inverse Fourier transform is carried out, which converts the interferogram received from the spectrometer into the spectrum of the original signal.In my case, it is especially difficult that I control the mirror with a sinusoidal signal. Because of this, the mirror also moves in a sinusoidal manner, which means that its speed is constantly changing. It turns out that the signal from the interferometer output turns out to be frequency modulated. Thus, the program must also correct the frequency of the processed signal.
The entire program is written in C #. Working with sound is done using the NAudio library. The program not only processes the signal from the spectrometer, but also generates a sinusoidal signal with a frequency of 20 Hz to control the movable mirror. Higher frequencies are less well transmitted by the mechanics of the movable mirror.
The signal processing process can be divided into several stages, and the results of the signal processing in the program can be viewed on separate tabs.
First, the program receives a data array from the audio card. This array contains data from the main and reference channels:
Above - the reference signal, below - the signal from one of the photodiodes at the interferometer output. In this case, a green LED is used as a signal source.
The reference signal processing proved to be quite difficult. One has to look for local minima and maxima of the signal (marked on the graph with colored dots), calculate the speed of the mirror (orange curve), and look for points of minimum speed (marked with black dots). The symmetry of the reference signal is important for these points, so that they do not always coincide exactly with the actual minimum speed.
One of the found velocity minima is taken as the origin of the interferogram (marked with a red vertical line). Further, one oscillation period of the mirror is distinguished:
The number of oscillation periods of the reference signal per one pass of the mirror (between the two black dots in the screenshot above) is indicated on the right: "REF PERIODS: 68". As I already mentioned, the resulting interferogram is frequency modulated and needs to be corrected. For the correction, I used the data on the current period of signal fluctuations in the reference channel. The correction is carried out by interpolating the signal using the cubic spline method. The result is visible below (only half of the interferogram is displayed):
The interferogram is obtained, now you can perform the inverse Fourier transform. It is done using the FFTW library. Conversion result:
As a result of this transformation, the spectrum of the original signal in the frequency domain is obtained. In the screenshot, it is converted to inverse centimeters (CM ^ -1), which are often used in spectroscopy. But I'm still more used to the scale in wavelengths, so the spectrum has to be recalculated:
It can be seen that the resolution of the spectrometer decreases with increasing wavelength. You can slightly improve the shape of the spectrum by adding zeros to the end of the interferogram, which is equivalent to performing interpolation after performing the transformation.
Examples of the obtained spectra
Laser radiation:On the left, the rated current is supplied to the laser, on the right, a much lower current. As can be seen, with decreasing current, the coherence of laser radiation decreases, and the spectrum width increases.
The sources used were: "ultraviolet" diode, blue, yellow, white diodes, and two IR diodes with different wavelengths.
Transmission spectra of some light filters:
The emission spectra are shown after interference filters, taken from the densitometer. In the lower right corner - the spectrum of radiation after the IR filter, taken from the camera. It is worth noting that these are not the transmittances of these filters - to measure the transmission curve of the filter, you need to take into account the shape of the spectrum of the light source - in my case it is an incandescent lamp. With such a lamp, the spectrometer had certain problems - as it turned out, the spectra of broadband light sources are obtained somehow clumsily. I have not been able to figure out what this is connected with. Perhaps the problem is related to the nonlinear motion of the mirror, possibly to the dispersion of the radiation in the cube, or poor correction of the uneven spectral sensitivity of the photodiode.
And here is the resulting emission spectrum of the lamp:
The teeth on the spectrum on the right are a feature of the algorithm that compensates for the uneven spectral sensitivity of the photodiode.
Ideally, the spectrum should look like this:
While testing the spectrometer, one cannot help but look at the spectrum of a fluorescent lamp - it has a characteristic "striped" shape. However, when registering the spectrum with a Fourier spectrometer of the spectrum of a conventional 220V lamp, a problem arises - the lamp flickers. However, the Fourier transform allows you to separate the higher frequency oscillations (units of kHz), given by the interference, from the low frequency ones (100 Hz), given by the network:
The spectrum of a fluorescent lamp obtained by an industrial spectrometer:
All spectra above were obtained using a silicon photodiode. Now I will give the spectra obtained with a germanium photodiode:
The first is the spectrum of the incandescent lamp. As you can see, it is not very similar to the spectrum of a real lamp (already given earlier).
To the right is the transmission spectrum of a copper sulfate solution. Interestingly, it does not transmit infrared radiation. A small peak at 650 nm is associated with the re-reflection of laser radiation from the reference channel to the base.
This is how the spectrum was filmed:
Below is the water transmission spectrum, to the right of it is a graph of the real water transmission spectrum.
Next are the transmission spectra of acetone, ferric chloride solution, isopropyl alcohol.
Finally, I will give the spectra of solar radiation obtained by silicon and germanium photodiodes:
The uneven shape of the spectrum is associated with the absorption of solar radiation by substances contained in the atmosphere. On the right is the real spectrum shape. The shape of the spectrum obtained by the germanium photodiode differs markedly from the real spectrum, although the absorption lines are in their places.
Thus, despite all the problems, I still managed to get the interference of white light at home and make a Fourier spectrometer. As you can see, it is not without its drawbacks - the spectra are somewhat curves, the resolution is even worse than that of some home-made spectrometers with a diffraction grating (this is primarily due to the slow motion of the moving mirror mirror). But nevertheless - it works!
Friends is approaching Friday evening, this is a wonderful intimate time when, under the cover of an alluring twilight, you can reach your spectrometer and measure the spectrum of an incandescent lamp until the first rays of the rising sun, and when the sun rises, measure its spectrum.
How do you still don't have your own spectrometer? It doesn't matter, let's go under the cat and correct this misunderstanding.
Attention! This article does not claim to be a full-fledged tutorial, but perhaps within 20 minutes after reading it, you will expand your first radiation spectrum.
Man and spectroscope
I will tell you in the order in which I went through all the stages myself, one might say from worst to best. If someone is aiming at a more or less serious result at once, then half of the article can be safely skipped. Well, people with crooked hands (like mine) and just curious will be interested in reading about my ordeals from the very beginning.There are enough materials on the Internet on how to assemble a spectrometer / spectroscope with your own hands from scrap materials.
In order to acquire a spectroscope at home, in the simplest case, you will not need much at all - a CD / DVD blank and a box.
On my first experiments in the study of the spectrum I was prompted by this material - Spectroscopy
Actually, thanks to the author's developments, I assembled my first spectroscope from a transmissive diffraction grating of a DVD disc and a cardboard box from tea, and even earlier before that I had enough of a dense piece of cardboard with a slot and a transmissive grating from a DVD disc.
I can't say that the results were stunning, but the first spectra were quite successful, miraculously saved photos of the process under the spoiler
Photo of spectroscopes and spectrum
The very first option with a piece of cardboard 
The second option with a tea box 
And the captured spectrum
The only thing for my convenience, he modified this design with a USB video camera, it turned out like this:
photo spectrometer
I must say right away that this modification saved me the need to use a mobile phone camera, but there was one drawback the camera could not be calibrated to the settings of the Spectral Worckbench service (which will be discussed below). Therefore, I was not able to capture the spectrum in real time, but I could fully recognize the already collected photographs.
So let's say you bought or assembled a spectroscope according to the instructions above.
After that, create an account in the PublicLab.org project and go to the SpectralWorkbench.org service page. Next, I will describe to you the spectrum recognition technique I used myself.
To begin with, we will need to calibrate our spectrometer.To do this, you will need to take a snapshot of the spectrum of a fluorescent lamp, preferably a large ceiling lamp, but an energy-saving lamp is also suitable.
1) Press the Capture spectra button
2) Upload Image
3) Fill in the fields, select a file, select new calibration, select a device (you can select a mini spectroscope or just custom), select which spectrum you have, vertical or horizontal, so that the spectra in the screenshot of the previous program are clear - horizontal
4) A window with graphs will open.
5) Checking how your spectrum is rotated. There should be a blue range on the left, red on the right. If this is not the case, select the button more tools - flip horizontally, after which we see that the image has rotated but the graph is not, so press more tools - re-extract from foto, all the peaks again correspond to the real peaks.
6) Click the Calibrate button, click begin, select the blue peak directly on the chart (see screenshot), click LMB and the pop-up window opens again, now we need to click finish and select the extreme green peak, after which the page will refresh and we will get the calibrated by wavelengths image.
Now you can fill in other investigated spectra, when requesting a calibration, you need to indicate the graph that we have already calibrated.
Screenshot
Type of configured program 
Attention! Calibration assumes that you will subsequently take pictures with the same apparatus that calibrated the change in the image resolution apparatus, a strong shift of the spectrum in the photo relative to the position on the calibrated example can distort the measurement results.
Honestly, I slightly corrected my pictures in the editor. If there was a flare, darkened the environment, sometimes rotated the spectrum a little to get a rectangular image, but once again I repeat the file size and the position of the spectrum itself relative to the center of the image, it is better not to change.
With the rest of the functions such as macros, auto or manual brightness adjustment, I suggest you figure it out yourself, in my opinion they are not so critical.
The resulting graphs are then conveniently transferred to CSV, while the first number will be a fractional (probably fractional) wavelength, and the average relative value of the radiation intensity will be separated by a comma. The obtained values \u200b\u200blook nice in the form of graphs built for example in Scilab
SpectralWorkbench.org has smartphone apps. I haven't used them. therefore I cannot evaluate.
Have a colorful day in all the colors of the rainbow friends.
UPD: At the request of DrZugrik, I will additionally write that the option with SpectralWorckbench equipment is one of the most budgetary ones, it can cost 500 perpetually conventional units.
In previous articles I described how I tested various plant LEDs. For the analysis of the spectrum, I and taken from a familiar physics teacher.
But the need for such a device appears periodically and the spectroscope, or even better, the spectrometer would like to have at hand.
My choice is a grating jewelry spectroscope
Once a thing for jewelers, the set included a "leather" case
The dimensions of the spectroscope are small
What else was clear from the store description
Everything is assembled tightly, so there will be no dismemberment.
Let us also believe that there is a lens-objective on one side of the tube, and a diffraction grating and protective glass on the other.
And inside is a beautiful rainbow. Having admired it to his heart's content, he began to look for what to look at on the spectrum.
Unfortunately, it was not possible to use the spectroscope for its intended purpose, since my entire collection of diamonds and precious stones was limited to a wedding ring, which is completely opaque and does not give any spectrum. Well, perhaps in the flame of the burner))).
But the mercury fluorescent lamp honestly gave a lot of beautiful stripes. Having admired different light sources to my heart's content, I was puzzled by the question that the picture needs to be somehow fixed and the spectrum measured.
A little DIY
A picture of a camera attachment was spinning in my head for a long time, and under the table there was one that had not undergone the last modernization, but quite successfully coped with PVC plastic.
The design is not very beautiful. All the same, the backlashes in X and Y were not completely won. Nothing ball screws are already in the assembly and are waiting for the supporting linear rails to arrive.
But the functionality turned out to be quite acceptable so that the rainbow was displayed on an old Canon, which has been lying idle for a long time.
True, here I was disappointed. The beautiful rainbow was becoming discrete.
All the fault is the RGB matrix of any camera and camera. After playing around with the white balance settings and shooting modes, I came to terms with the picture.
After all, the refraction of light does not depend on what color to fix the image. A black-and-white camera with the most uniform sensitivity over the entire width of the measured range would be suitable for spectral analysis.
Spectral analysis technique.
Through trial and error, such a technique was drawn
1. A picture of the scale of the visible range of light (400-720nm) is drawn, the main lines of mercury for calibration are indicated on it.
2. Several spectra are taken, always with a reference mercury one. In a series of surveys, it is necessary to fix the position of the spectroscope on the lens in order to exclude a shift in the spectrum from the series of images horizontally.
3. In the graphic editor, the scale is adjusted to the mercury spectrum, and all other spectra are scaled without horizontal shift in the editor. It turns out something like this
4. Well, then everything is driven into the Cell Phone Spectrometer analyzer program from this article
We check the technique on a green laser, for which the wavelength is known - 532nm
The error turned out to be about 1%, which is very good with the manual technique of fitting mercury lines and drawing a scale practically by hand.
Along the way, I learned that green lasers are not direct radiation, like red or blue, but use solid-state diode pumping (DPSS) with a bunch of secondary emissions. Live and learn!
Measuring the wavelength of the red laser also confirmed the correctness of the method.
For interest, I measured the spectrum of the candle
and burning natural gas
Now you can measure the spectrum of LEDs, for example "full spectrum" for plants
The spectrometer is ready and running. Now I will use it to prepare the following review - a comparison of the characteristics of LEDs from different manufacturers, whether the Chinese are fooling us and how to make the right choice.
In short, I am satisfied with the result. Maybe it made sense to connect the spectroscope to a webcam for continuous spectrum measurement, as in this project
Spectrometer testing by my assistant
Be sure to watch the video on the channels (there are thematic playlists):
https://www.youtube.com/channel/UCn5qLf1n8NS-kd7MAatofHw
https://www.youtube.com/channel/UCoE9-mQgO6uRPBQ9lsPZXxA
Please help to gain 1000 subscribers on the first channel and at least 4000 hours of views over the last year on each of them, for this, watch at least one video in full!
This beautiful picture is a photograph of the light and infrared spectrum emitted by high pressure sodium lamp NLVD type DNaT (Arc Sodium Tubular). To view and photograph various spectra, it is enough to have a digital camera and a specially prepared CD-R or DVD-R. The latter underestimates the brightness, especially of red. CD-R reduces the brightness of blue and produces lower resolution. The first photo was taken via DVD-R.
The two yellow lines are the sodium doublet with wavelengths of 588.995 and 589.5924 nm. The second doublet is infrared 818.3 and 819.4 nm.
Spectrum graph.
Now a few words about preparing disks. A part must be cut out of the disc to completely cover the lens.
In the photo the DVD-R is purple. We need transparent diffraction grating, therefore, on the CD-R we glue a wide adhesive tape from the side of the inscriptions. We tear it off and together with the scotch tape the disc cover is removed. With DVD-R, it's even easier, the cut piece easily splits into two parts, one of which we need.
Now, using double-sided tape, you need to glue the diffraction grating to the lens, as in the photo below. You need to glue on the side opposite to the one from which the layer is torn off, because the surface under the layer is easily contaminated by the lens, and after cleaning the image quality of the spectrum will be worse.
The result is the simplest spectroscope, best suited for studying light sources from a certain distance.
If we want to study not only the visible spectrum, but also the infrared, and in some cases the ultraviolet, then it is necessary to remove the filter from the camera that blocks the infrared rays. It is worth noting that part of the IR and UV spectrum visible to the eye at a sufficiently high radiation intensity (laser points 780 and 808 nm, a 940 nm LED crystal in the dark). If it is necessary to provide the same visual sensation for the wavelengths of 760 nm and 555 nm, then the radiation flux for 760 nm must be 20,000 times more powerful. And for 365 nm it is a million times more powerful.
Let's go back to the filter called Hot Mirror, which is in front of the matrix. You need to open the camera body, unscrew the screws that attach the matrix to the lens, pull out the filter, and assemble the camera in reverse order. Hot Mirror looks like this:
2 left filters from cameras. They have a pink sheen, and the turquoise color appears at a different angle. In addition to IR, they can also partially or completely block ultraviolet rays. Therefore, their removal opens up opportunities not only for infrared photography, but also ultraviolet, if the optics and the matrix of the camera allow. For UV photography, UV-pass filters are used to block visible light.
Now we turn to the very process of photographing the spectra. The room should be dark, in addition, you can use a black screen near the camera, a point or slit light source that minimally illuminates the room. Turning on the camera, we will see the following image using the example of a 405 nm laser, shining through a narrow slit between two blades:
The center point is the laser itself. Two lines are its spectrum. You can use any of them. To do this, you need to turn the camera and zoom in. If we continue to move the camera, we will see several other lines of the second, third, etc. orders of the spectrum. In some cases, they will interfere, for example the second-order green line will be superimposed on the 1064 nm infrared line. This occurs in the spectrum of a green laser, unless an IR cut filter is installed. It is on the bottom right in the filter photo. To remove the overlap, I used a red filter. Photo of this example with signed wavelengths:
As you can see, the green line of the second order completely covered the 1064 nm line. And the next photo with blocked green light, where only two IR lines, 808 nm and 1064 nm, remain. I didn’t sign since. the location is identical to the previous photo.
From the image, where there is a radiation source, one known wavelength and several unknowns, they can be easily identified. For example, open a photo with captions in Photoshop. Through ruler tool measure the distance from the laser to the line 532. It is equal to 1876 pixels. We measure the distance from the laser to the line, the wavelength of which we want to know, up to 808. The distance is 2815 p. We consider 532 * 2815/1876 \u003d 798 nm. Inaccuracy occurs due to distortion of the lens optics. At maximum optical approximation, the error decreases. It has also been observed that the 808 nm laser emits a shorter wavelength, around 802 nm, and its wavelength decreases as the supply current decreases.
And without a source of radiation, the photo can be determined by knowing the other two wavelengths. We measure the length from the line 532 to 1064, there are 1901 p. From 532 to 808, we get 939 p. We consider (1064-532) / 1901 * 939 + 532 \u003d 795 nm.
But the easiest way is to compare a photograph with two known lines with scale... In this case, nothing no need to count.
Further incandescent lamp spectrum, which is very similar to the spectrum of the Sun, but does not contain Fraunhofer lines... Interestingly, the camera displays infrared radiation up to 800 nm as orange, and more than 800 nm looks like violet.
White LED spectrum also continuous, but has a dip in front of the green region and a peak in the blue region of 450-460nm, which is caused by the use of a corresponding blue LED covered with a yellow phosphor. The higher the color temperature of the LED, the higher the blue peak. It lacks ultraviolet and infrared rays, which were present in the spectrum of an incandescent lamp.
And here cold cathode lamp spectrum from the backlight of the monitor. It is linear and repeats exactly fluorescent lamp spectrum... The IR portion of the spectrum is taken from CFLs for better image quality.
Now go to ultraviolet black light, or, as it is also called, Wood's lamp. It emits soft, long-wavelength ultraviolet light. The photo turned out like this:
Infrared spectrum for fluorescent lamps, CCFL, Wood is almost the same. Only the latter lacks several lines that are closest to the visible range. Infrared rays are most intensely emitted from those parts of the lamps where the filaments are located. The photo was taken through a paper spectroscope, more on which below.
Paper spectroscope.
Such a spectroscope is well suited for viewing the spectrum with the eye. It can also be used with different cameras, such as a telephone. There are two varieties.
2. Operates on reflection from a diffraction grating. It is possible not to delaminate the disks, but then pale duplicates will appear next to the bright lines from the lasers, due to re-reflections inside the disk, which should not be in the spectrum. It is very difficult to transfer the shiny CD layer to another surface to keep it as smooth. Therefore, you need to use a CD that has the same iridescent surface on both sides. From the side where there are inscriptions on ordinary discs, you need to tear off the transparent layer using adhesive tape. It is important that the shiny layer remains on the disc. I managed to do this with half the disk (from edge to center), this was enough for the spectroscope. If you do not tear off the transparent layer, the uniform spectrum will appear discontinuous with alternating dark stripes.
Glued to the spectroscope additional ring, with which it is held on the camera lens. Between the light source and the spectroscope, it is recommended to place matte film or prism with two matte edges as shown for better light distribution. The inner part of the spectroscope is made of black paper without shine, the second layer is made of foil, and on top is plain paper on which the drawing is printed. The side where the light enters can be painted black so that UV and violet radiation does not cause the paper to glow white, distorting the image.
With the help of this spectroscope, it was possible to clearly and vividly photograph spectrum of neon indicator light... They are used to illuminate switches, in indicators of the operation of kettles, stoves and other devices.
Not only lasers produce one thin line of the spectrum. If the wire is dipped into a solution of NaCl salt, and then brought into the fire of a gas turbo burner or lighter, then yellow glow with wavelengths of 588.995 and 589.5924 nm.
Some turbo lighters have a lithium plate. It colors the flame in red with 670.78 nm line.
Below is a photo of these spectral lines along with laser lines: green 532 nm, red 663 nm, infrared 780 nm and 808 nm.
It is convenient to use the above yellow light for determination of the period of the diffraction grating in the absence of a laser, and calculating the wavelength of light sources... The simplest device in the figure below consists of two rulers, on one of which a diffraction grating is fixed, and a narrow slit of two blades rises above the other. The distances in millimeters from the diffraction grating to the screen (ruler) with a slit and from the slit (maximum of the zero order) to the maximum of the first order are used. In the first figure, you need to look through a diffraction grating at a light source with a known wavelength. Thus, you can calculate the period of the diffraction grating by the formula under this image, and then, in the same way, you can determine the wavelength, but using the formula from under the second figure. It shows the determination of the laser wavelength in a slightly different way: the laser shines through a diffraction grating onto a ruler. In this case, the gap is not needed. I used a diffraction grating from the Starry Sky attachment that came with the laser pointer. There are two grates, but the nozzle was disassembled and one grate was pulled out. The CD diffraction grating did not fit at all, because gave a huge error of 100 nm.
The next photo of a rare light source is lightning. The spectrum goes into the UV range up to about 373 nm, which is the limit for this camera.
The spectrum of a white discharge lamp that illuminates a football field.
