RTL-SDR V4 workflow

1. Réception RF via l’antenne :

 Ce qui se passe : L'antenne capte les ondes radio analogiques présentes dans
l’environnement.

2. Syntonisation avec la puce Rafael Micro R828D

 Ce qui se passe : Le tuner sélectionne une plage de fréquences souhaitée.

 Sure Rafael Micro R828D : le R828D capte et décode les signaux
radio et TV.
Il permet à des appareils comme les téléviseurs ou les clés RTL-SDR de
recevoir des chaînes TV, écouter la radio, ou surveiller des
communications sans fil (avions, navires, etc.).
 SOURCE : https://www.jotrin.com/product/parts/R828D

3. Filtrage passe-bande (BPF)

 Ce qui se passe : Des filtres matériels éliminent les fréquences indésirables.

4. Amplificateur à faible bruit (LNA)

 Ce qui se passe : Amplifie les signaux faibles avec peu de bruit ajouté.

5. Conversion analogique-numérique (ADC)

 Ce qui se passe : La puce RTL2832U convertit le signal analogique en données

numériques (échantillons I/Q).

We will have to demodulate FM, frequency shift, filter, decimate, resample,

synchronize, decode, and parse the bytes.

when using quadrature/IQ sampling, your received bandwidth equals your

sampling rate
Example
FM Radio Basics

 FM radio stations (like 95.5 FM) transmit audio using frequency modulation.
 But there's more than just audio in an FM signal — there are extra signals hidden inside.

What’s Inside an FM Signal (after demodulation)?

When we look at the Power Spectral Density (PSD) after FM demodulation, we see:

1. 0–17 kHz → The main mono audio signal.

2. 19 kHz → A pilot tone that helps decode stereo sound.
3. 38 kHz → The stereo audio (difference between left and right).
4. 57 kHz → RDS (Radio Data System) — a small digital signal that gives info like station name,
song, time, etc.
5. 67 kHz → DirectBand — a now-unused Microsoft service for GPS and traffic data.

 Carson’s Rule estimates FM needs ~250 kHz of bandwidth.

 We sample at 250 kHz to cover the whole signal.

🧾 RDS Focus

 RDS sits at 57 kHz, takes up about 4 kHz of bandwidth.

 It’s a low-rate digital signal carrying:
o Station ID
o Song info
o Time/date
o Traffic alerts

RDS data (e.g., song name, station info) is packaged into:

o Groups of 104 bits → 64 data bits + 40 check bits.

o Sent continuously, no gaps, so the receiver must find the group boundaries.

 Modulation: BPSK (Binary Phase Shift Keying).

 Differential coding is used to avoid issues with phase shifts.
 Symbol rate: 1187.5 symbols/sec → ~1.2 kbps data rate.
 No advanced error correction, just CRC for error detection.

Where is it placed?

 The BPSK signal is centered at 57 kHz, added to the FM signal before transmission.
 Then the entire FM signal (audio + RDS + others) is FM modulated and broadcast.
cyclic redundancy check (CRC)

Steps to Receive and Decode RDS:

1. Receive the FM signal:

o Tune to the FM station’s frequency (e.g., 95.5 MHz) and sample the signal at 250 kHz.
o The signal is extracted from a file
2. FM Demodulation:
o Extract the frequency from the phase,
o  You're calculating the phase difference between adjacent elements in the signal, which
corresponds to the frequency change between them.
o  Signal with delay and conjugation: The code multiplies the received signal with a
delayed and conjugated version of the signal. It isolates the phase difference while ignoring
the magnitude.
o  Why conjugate and delay?: Conjugating the signal flips its phase, and delaying it shifts
it by one sample. When you multiply these two, you get a result where the frequency of the
signal becomes the main feature, while the original phase (which is irrelevant for frequency)
gets canceled out.
o  Phase extraction: After multiplying, the phase of the result is calculated. The phase
change directly relates to the frequency, so by measuring the phase, we can recover the
frequency variations from the FM signal.
o x = 0.5 * np.angle(x[0:-1] * np.conj(x[1:]))
3. Frequency Shift:
o Shift the RDS signal to 0 Hz by removing the 57 kHz carrier frequency of the RDS. And
making the center the new carrier frequency
o Positive frequency shift shifts the signal to higher frequencies, while a negative frequency shift
shifts the signal to lower frequencies.
o The time vector t gives the time (in seconds) for when each of the N samples.the frequency
shift is applied over time, we shift each sample at a time. T1 we shift sample 1 time 2 we shift sample
2 etc
4. Low-Pass Filtering:
o Filter out everything except the RDS signal, We want to remove the high-frequency
components of the signal and keep only the low-frequency RDS signal. acting as a matched filter
to isolate the desired information.
o The convolution operation is used to apply the filter to the signal.
o Design the Low-Pass Filter: then convole the filter to the signal to apply the filter on it
o Firwin to creat a filter

Filter type Cutoff(s) pass_zero

Lowpass [f] True (default)
Highpass [f] False
Bandstop [f1, f2] True
Bandpass [f1, f2] False

5. Decimation:
o Reduce the sample rate by 10, since we only need the lower-frequency part of the signal
after filtering you only keep samples at the rhythm of the beat — you don’t need the
extra ones anymore, because the melody is already slow and smooth
o take only each 10 element
o
 6. Resample to 19 kHz:
o Resample the signal so that we have an integer number of samples per symbol, making it
easier to process. Up and down
o resample_poly ()
o now your signal is no longer sampled at 25 kHz. It’s now 19 kHz! So need to change the sample
rate
7. Time Synchronization:
Use a Mueller and Muller method for symbol-level synchronization, aligning the data stream
correctly.
"Adjust the timing of our sampling so we always grab the signal right at the middle of each
symbol."

 GMSK at 9600 baud (symbols per second).

 yes, technically it is. Anything > 1 sample/symbol is oversampled.
 1 sps=symbol rate/sampling rate check if between 2 AND 10 IN ORDER FOR MULLER TO WORK

How it works mu :In the algorithm, mu is updated with each iteration, allowing the
synchronization process to track the exact position within each symbol.
For instance:

o If mu = 0.0, you're at the beginning of the symbol.

o If mu = 0.5, you’re in the middle.
o If mu = 1.0, you've passed the symbol boundary and are at the start of the next symbol.

8. Frequency Synchronization:
o Use a Costas loop for fine frequency synchronization, ensuring the signal’s frequency is
correctly adjusted.
o The Costas loop adjusts the frequency of the receiver's local oscillator (LO) to correct any frequency
offset in the received signal.The
o Costas loop is used to correct small frequency and phase errors in a received signal. When a signal is
received, it may have a frequency offset, causing the constellation (a plot of the signal) to spin. The
Costas loop helps sync the signal by adjusting the receiver's local oscillator until the signal’s
frequency and phase match the original one. This stops the spinning and ensures the signal can be
decoded correctly. The process is important for accurate data reception, especially in noisy
environments.
9. BPSK Demodulation:
o Decode the BPSK signal back into the binary 1's and 0's.
o  You use the real part of the complex signal (np.real(x)).
o  Then, you check whether the real part is greater than 0 (indicating a binary '1') or less
than 0 (indicating a binary '0').
o  The astype(int) converts the result from boolean (True/False) to integers (1/0).
o
10. Differential Decoding:

 Undo the differential encoding applied during transmission (which encoded the data as changes
between bits), if change then 1 if not then 0 between two bits.
 This eliminates the need to worry about phase shifts (such as 180-degree rotations) because the change in
the data is what's important.
 bits = (bits[1:] - bits[0:-1]) % 2
 bits = bits.astype(np.uint8)


11. Decode to Bytes:

 Group the 1’s and 0’s into 8-bit bytes (since each RDS data unit is 8 bytes, or 64 bits).
 12. Parse the Data:

 Finally, the decoded bytes are parsed into readable output like station name, song title, and other
data

1. What is a Sample Rate?

In digital signal processing (DSP), the sample rate refers to how frequently we sample a continuous signal
to convert it into a digital signal.

 For example, if you are sampling a continuous waveform, the sample rate tells you how many times
per second you capture a data point.
 For BPSK signals, where you transmit discrete symbols (either +1 or -1), sample rate defines how
many discrete samples you take per symbol to represent that signal digitally.

What is AIS?

AIS is used by ships to broadcast information like position, heading, and speed. It uses:

 GMSK modulation (a type of FM)

 VHF frequencies (around 161.975 MHz and 162.025 MHz)
 Messages are sent in TDMA slots (time-division).