Week of 01/08/2018 - Progress Report 7

Tamara Jovanovic

Topics covered: Choosing of the Rotational Stage for the System, Design of the Cuvette Stand for the system in SolidWorks

Materials used: SolidWorks Software

On the hardware side, this week’s objectives were to find the best rotational stage for the system and order it, design a cuvette stand in SolidWorks and get it ready to be printed.

After some deliberation and discussion, we came to the conclusion that we need to purchase a rotational stage for our optical system. Because of really high sensitivity in free space, we need to be really careful and precise when we start working with actual substances. Laser light which will be shined through the cuvette with a substance inside of it has a certain refraction index that needs to be considered. Index of refraction of a certain material is a number that describes how light propagates itself through that material. In order for us to have good results, we need to maximize the index of refraction for maximum output power through glucose substances in the cuvette. A rotational stage will help when it comes maximizing the refraction index of a medium. By adjusting it with such high precision, we will be able to track our movements and set the cuvette with a substance in it at an angle where we can get maximum power retention and hopefully, get reasonable results.

The rotational stage shown below is ordered online and we now await its arrival so that we can implement it in our system and start the next chapter of our project.

Rotational Stage.

Based on the datasheet provided with this rotational stage, a cuvette holder was designed in SolidWorks. Measurements were taken with reference to our system and the cuvette stand was designed. Upon the approval of our advisor, the cuvette holder shown below will be printed using a 3D printer and will be implemented as soon as printing is completed.

The design on the cuvette stand consisted of making sure that its height doesn’t exceed the height at which our collimators were. Since the height of the rotational stage is 55mm, the cuvette holder which will directly be applied to it will be 33mm of height. The width was chosen based on the diameter of the rotational stage, which is 60mm. The cuvette stand is of rectangular shape, and will have studs which will be pushed into the holes of the rotational stage for maximum stability. There will be a rectangular hole in the middle of it, which is where the cuvette will go. The cuvette is of size 10x10mm. The cuvette stand design is shown in pictures below.

View of the stand from the top. The hole in the middle is where the cuvette will go.

View from the top with measurements.

View from the bottom.

For next week’s plans, the design of the cuvette stand should be finalized and the printing should start. Also, better optical alignment should be worked on, as well as trying to obtain as much power possible from the second collimator.

Ezequiel Partida

Topics covered: Implemented Correlation and RMSE% methods under dummy data for functionality verification

Materials used: Matlab

From last semester, we had implemented 2 methods for the determination of spectrum similarities when time came for glucose absorption readings: 1. Correlation coefficient analysis, which is derived from the equations below, comparing 2 equal-length sets of data.

2. RMSE%, which is the percent error of the root-mean-squared value for 2 sets of data.

From a short study performed last semester, we concluded that the first method was valid, since it correlated well with Matlab’s built-in function xcorr, giving us confidence that our method would be appropriate if the data permits. The tests performed included running the Matlab xcorr function vs our implemented method for 3 different sets of data: a sinusoidal vs an amplitude modulated sinusoidal, a sin vs a sinc, and a sin vs a sin+cosine. The results are shown below for reference.

In order to test the functionality of the RMSE% method, the same 3 test cases were ran for it since we knew that these test cases tested the 2 extreme cases for correlation (1 and 0) and a value in between. The results for the RMSE% comparison are as follows:

As it can be seen, the RMSE% does not function as well as the correlation method we previously implemented. For our method, results on the left, 2 sin signals are perfectly correlated since cross-correlation basically checks for the overall shape of a signal. However, RMSE% checks for overall discrepancies within 2 sets of data, so an amplitude change can greatly affect RMSE%. The first test case showed this. Correlation showed a 1, while RMSE% showed 70% error, since the 2 signals are different by an amplitude change. Similarly for the 2 other tests cases, RMSE% was high (which is bad) because the compared signals are not close to aligning perfectly. Therefore, other test cases need to be ran in order to verify if RMSE% works correctly.

The 3 new test cases compare a sin signal vs a sin signal with a small phase change of pi/30. The second test case tests a sin vs a tan. The third test case compares a sin vs a sin with a large phase shift of pi/4. The results were as follows:

As seen above, these test cases are more suitable for RMSE%. The first test case shown 2 signals which are almost identical, with a small phase shift almost visible. Correlation shows to be almost 1, which is visibly seen, and similarly, RMSE% shows a small error of 7%. This shows that correlation coincides with small error given that RMSE% takes into account vertical discrepancies. The second test case tests the extreme with a tangent signal. The correlation is almost 0, and as expected, RMSE% is very high. The final test case serves to reinforce that in between values hold true for both methods.

Additionally, this week, we captured spectral data using the femtosecond laser and our optical setup. The following plots and analyses show comparison between a pure laser spectrum in log scale and a laser spectrum refracted by a cuvette in log scale.

As shown, the pure laser output and the diminished signal hold a high correlation, since the red signal is still the same laser output, just with less power. Nonetheless, the RMSE% is very high, since the amplitudes definitely changed. However, this test served to show that the methods can be implemented with spectral data that will later be used once glucose readings are ready to be taken. Additionally, this helps to point out flaws in our methods. First of all, as discussed last semester, correlation only checks for overall shape, not x-axis discrepancies. RMSE% helps because this actually takes into account y-axis discrepancies. 1 idea is that cross correlation can be used to identify substances, since glucose has shown to only cause smooth, gradual changes in absorption spectra. Then, RMSE% can be used to get a rough estimate of the discrepancies of signals. However, more data needs to be taken in order to characterize the system and to see if the methods will be useful. The good part of these tests is that we know that the 2 methods function as expected. They can be reliable for what they are meant to do, and they even work well with our system.

Jonathan DeRouen

Topics covered: Confirmation of Progress from last semester, Confirmation of adequate power retention.

Materials used: Matlab Software, Spectrometer, Cuvette stand, GPIO cable, laptop computer

This week our objective was to prepare ourselves for acquiring glucose data to test our analysis techniques would work.

Firstly in order to get a proper signal to be read without losing too much power retention we decided to maintain approximately a 3 dB loss from our source to the output. As discussed in our earlier reports and sessions our laser outputs at a 4 mW range and as such to achieve a 3 dB loss that would be approximately 2 mW. We have managed to achieve a 1.5mW output which can be translated to approximate 4 dB loss. This can be calculated in equation 1 as shown below. This output can be maximized even further if necessary when data begins to be gathered, however we have discussed that due to the sensitivity of optical alignment we will try to work at the 4 dB loss of the system.

10log(4mw) - 10log(output system power) = dB loss

Confirming output: 1.5 mW from system.

Zoomed in value of output power.

Another system that we discussed was to test come of our signal processing techniques developed last semester and to incorporate them into our data analysis program. In order to achieve this test data, we decided to test using the GUI developed last semester on the computer located in the optics lab. However when I set up using the GUI on the new computer I was unable to have the computer find the GPIO connection we have been using since last semester. However this was not the major issue of this week and acquired some sample data using a physical USB drive.

The data I acquired was on CSV files in order to be analyzed. In order to achieve this data I used the absorption readings from the laser and while using a cuvette to create some noise interference in the spectrum readings. I had recorded data in the decibel and linear scales in order to have a full range of analysis.

Progress:

For the next upcoming week I hope to run this dummy data and confirm the operation in the algorithms implemented before.

Another problem is to finish moving our GUI interface onto the new system and revisit and eliminate the problem why the optics lab computer is not responding.

Wait for our rotational stage to be acquired in order to add it to our system.

Build using the 3-D printer the tool that Tamara developed to hold our cuvette onto the device

Continue research and development of data analysis methods