Week 03b: Trial and Error

This week we continued our efforts in software development and hardware improvement.

Alex Esclamado:

This past week I tried to run some tests with the AOD in order to better understand its functionality. The proper setup is denoted in Figure 1 as per the instruction manual, with some included pictures of what was installed in the lab. For safety reasons, I opted to use an Infrared laser instead of the MenloSystems laser and plan to run experiments again with Dr. Asghari to further ensure safety by working with these sensitive components.

For the oscilloscope, we attempted to run the system in full but ran into coding errors. My hope was to familiarize myself with the FastFrame application of the scope and observe the data recording process with a live sample scanning. However, the Python code for this scope is still being worked upon. We are also still learning the additions made over break that allowed for the scope’s implementation. Primarily, a stage was added after the 50/50 coupler and before the photodetector. I have communicated with the graduate student in order to resolve these questions so that we may progress.

Trevor Wong:

This past week I successfully communicated with the GPU using PyCUDA. This was done by using a simple program that sends an array of data to the GPU to be doubled and then returned. With the successful use of the GPU, I turned my attention to understanding and testing various ways to implement FFT in with PyCUDA. Here is where several obstacles came to play. One of the more popular packages that perform GPU-accelerated FFT is known as PyFFT. The problem with PyFFT is that is not supported on Python 3 (our program runs on 3.7), only on Python 2. Alternatively, there are other packages that can perform FFT such as scikit.cuda and Reikna, which was another option I explored. With both these alternatives sample code and documentation is sparse. Given that, I have been testing all 3 modules for running FFT. The PyFFT code was able to produce a result, but the other two modules have been updated in a way that the sample code found no longer works. My plan is to get an FFT program working with a simple array, then an image, and finally data collected from the oscilloscope. For the coming week, I need to focus on not only finding what is compatible with our program but also a GPU-accelerated FFT module that is widely used and/or still supported. With that, I plan go in-depth on purely the processing portion of the current program and to speak with someone who is more familiar with the current program (Dr. Asghari or Max) to find out exactly what data is being collected and processed, what the data is after processing, and what type of processing is needed.