Picked from the software, topic, and level of “Class MIMO-OFDM Channel Estimation Deep Learning #2: MIMO-OFDM System Simulation and Channel Dataset Generation”.
At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Here is the final result, a massive MIMO-FDM channel dataset with 32 antennas and 64 subcarriers. Behind 2048 complex numbers per channel, there are only 4 paths, clearly visible in the angular delay domain. This structure will later be used by deep learning. Let's start. Hello, welcome to episode two of the series on MIMO-FDM channel estimation with Deep Neural Networks. In massive MIMO, the base station uses dozens of antennas to serve users with large area coverage, but the channel for each antenna on each subcarrier must be known. Today we are building the channel model and dataset to train the estimator.
Step 1: Antenna array & channel geometry
Create file channel.py The base station has 32 antennas arranged as a uniform linear array, the user has 1 antenna. The pilot IDX function determines the pilot subcarrier with a complex pattern, 1 pilot every 4 subcarriers, so 16 pilots. The channel generation function creates the geometric channel. Each channel consists of 4 paths, each with an arrival angle, delay E, and gain rally. The steering vector of the ULA provides phase shift between antennas according to the angle, and delay E provides phase shift between subcarriers.
Step 2: Dataset, pilot signals & NMSE
The dataset function generates 12,000 training channels and 2,000 test channels with fixed SID, so the results are always the same and there is no need to store hundreds of megabit files. Training SNR is randomized between minus 10 to 20 decibels. Training SNR for massive MIMO antennas is indeed low because 32 antennas provide area gain of about 15 decibels. The receive pilot function adds noise to the channel on the pilot subcarriers, and the nemcdb function calculates the minimum square error in decibels, the main metric for channel estimation. Create file buat underscore dataset.py and run it.
Step 3: Generate dataset
Training dataset of size 12,000 x 32 antennas x 64 sub-carriers, with average channel power exactly 1. For one sample, the receiver only sees 32 x 16 pilot values. When the channel is transformed to the angular delay domain with FFT along the antenna axis and IFFT along the subcarrier axis, The 5% largest coefficients already account for 97.4% of the energy. This channel is sparse or rare.
Step 4: Antenna-frequency domain & angle-delay
Let's take a look at the graph. The channel magnitude at the antenna grid x sub-carriers forms a diagonal pattern, Interference results from 4 paths, and the red triangles mark the sub-carrier pilots. In the angular delay domain, the energy is concentrated at several bright points, one for each path. The classical per-antenna estimator does not utilize this structure, but CNN can.
Done!
Done. We have already modeled a massive MIMO OFDM channel with antenna array and geometric channel. Pilot pattern COM, continuous pilot signal, NMSI, and deterministic dataset. In the next episode, we build the classical score estimator and LMMSI. Currently working on a thesis, dissertation, or research in the field of telecommunications and artificial intelligence? Cerdas Inovasi Digital helps implement communication system research, from OFDM and multipath channels, massive MIMO simulation, score estimator list and LMMSI until deep learning-based channel estimation with PyTorch, as well as opening a Python programming course for electrical engineering and telecommunications. Want to master the entire series, from basic OFDM to NMSI evaluation and BER? All 6 episodes, complete with step-by-step PDF guides and source code, Available for premium tutorial subscribers. Subscribe starting at Rp69.000 per month, save 28% with the 12-month package, or one-time payment of Rp999.000 for lifetime access. Want a single class only? Buy single classes with lifetime access. Subscribe now on the premium page cerdasinovasidigital.com. Contact our WhatsApp, or visit cerdasinovasidigital.com. Thank you for watching, see you in the next episode.
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
53 video classes1,290 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
