frustrated Spectral analysis on CFD temporal signals is an elusive craft with many boobytraps. Here is the training we do at Cerfacs.

Initial evaluation

We kindly ask our esteemed trainee to take the test on spectral analysis. This test will help you to understand the kind of knowledge you should expect from this training. No pressure, you will re-take this test at the end to measure your progress (or your demise…)

Group discussion

If you follow our training in person, the trainer will probably ask you to listen several sounds and classify them by ear.

Then each sound spectrum will be showed on screen. We use the application RiFFT on our presentations. You can however install a spectrum analyzer app on your phone, like spectrum analyzer or spectroid.

The purpose of this torture is to observe how our audition can focus on one sound among many others concurrent sounds. The nature of harmonic versus broadband noise is the most critical part to get. This is the reason we have have two different spectrums normalizations : the Amplitude and the PSD spectra.

Training on Notebooks

We will use jupyter notebooks to get down to spectral analysis flesh and bones. To get your terminal ready, create a virtual environment with jupyter:

> python3 -m venv test_spectral
> source test_spectral/bin/activate
(test_spectral)> pip install jupyter

Then we can install the necessary libraries for spectral analysis (numpy, matplotlib, pandas, etc…) . Fortunately, our in house package uses all these. So simply install satis and you will be ready to go.

(test_spectral)> pip install satis

Now we are ready to proceed. Get the notebook you want from our spectral analysis notebooks. Choose the notebook and download.

You can start your notebook as:

(test_spectral) >jupyter notebook InteractiveSpectrumAnalysis.ipynb 

it should pop up in your browser. Try the notebooks in increasing order.

Introducing SATIS,

The package satis is easily available on PyPI. On the landing page, you can see, and try , the CLI of Satis.

Now that you got an edge on the Amplitude and PSD spectra, try to understand why satis insights could be useful.

  • First alone
  • Then two by two
  • Then by groups of four

Final evaluation

We kindly ask our esteemed trainee to take again the test on spectral analysis. Let see if the trainer managed to give you some useful ideas on the spectrums.

Like this post? Share on: TwitterFacebookEmail

Antoine Dauptain is a research scientist focused on computer science and engineering topics for HPC.

Keep Reading





Stay in Touch