Dependencies

- numpy
- matplotlib.pyplot
- soundfile
- sounddevice
- scipy.io.wavfile

Digital FSK Modulation

FSK (frequency-shift keying) modulation is a form of Passband modulation in witch digital information is transmitted through frequency changes of a carrier signal. The simplest FSK is binary FSK (BFSK). BFSK uses a pair of discrete frequencies to transmit binary (0s and 1s) information. With this scheme, the "1" is called the mark frequency and the "0" is called the space frequency.

This type of modulation is commonly used for transmissions in Radio Frequency Band, but FSK also alows us to implement it in low frequencies for a carrier with a low efficiency loss. This makes digital transmission through channels normally utilized to transmit voice possible. On top of that it also presents a versatile communication protocol in situations in which transmitting through radio waves or optical fibre is not an option or possibility

Syncronism

The message which shall be transmitted is asynchronous. Therefore, the moment at which the actual message begins is unknown, which implies that the detector must be designed to identify correctly the start and end point of the message within the transmission. For this to be possible, every message will be transmitted with a pre-defined bit sequence concatenated before the message to indicate where to begin the decoding process. This sequence we will call the "header" of the message and can be found in the file header.txt. Let it be known that this sequence was generated using a "maximum lenght sequence" algorithm, which increases the definition of the autocorrelation.

Data Format

The data transmitted may be of different kinds (audio, text, image, etc..) and of varied sizes. The program is designed to be able to detect the bits and feed them (without the header) to the app_decoder function. This function accepts a binary array and returns the number of bits lost in the detection process. Attention should be directed towards the fact that the input of app_decoder is in fact an array of int. Also, let it be stated that it is paramount that the correct first position of message be identified, for an error in synchronism will reduce the acurracy to aproximately 50%.

Types of Signals to be Processed

Two different modulations can be processed: BFSK (2-FSK) and 4-FSK. Knowing which is being used beforehand, one must chose before beginning signal reception. Below are the above mentioned modulations:

• 2-FSK

  • 800 Hz - 0
  • 1200 Hz - 1

• 4-FSK

  • 600 Hz - 00
  • 800 Hz - 01
  • 1000 Hz - 11
  • 1200 Hz - 10

Although other detection methods exist and can be applied in this case, in this project we use a Matched Filter

For further information on how the transmission works, see TX_script.py and other files in Examples folder.

Symbol Rate

The symbol rate used may be defined moments before transmitting. Therefore, the detector presents the option of defining the rate at the beginning of the decoding process.

Test Files

In the repository, you will find two test files: test_2fsk.wav and test_4fsk.wav. These files were generated with a 20 bit/sec rate and were used to help in the process of programming the decoder. The test.txt file contains the respective bit sequence (without the header). Let it be stated that these test files, naturally, are noise free. If testing with noise (to simulate a recording) is desired one must only uncomment the correct section of the code, which adds random bits before the actuall recording and gaussian noise to the whole array. Also in this repository are the two files which were recorded when presenting the project in Unicamp's EE881 (Telecommunications) class, Nov/2019. Naturally, these files include noise. The 2-FSK (20 baud rate) file works with a high precision considering the absence of error correcting algorithms in this project. The 4-FSK (50 baud rate) file can't be decoded due to the elevated baud rate together with the not so good quality of the recording. Let it be noted that the second transmission was with a 50 baud rate do to its size (being an image, it would take aproximately 9 minutes to transmit at a baud rate of 20).

Recording

If recording the transmission with your computer is desired, the current "audio obtaining" section should be commented and the above section uncommented. For this section to function properly, the user must know the time duration of the transmission or guess a value that will guarantee recording of the full audio file. You must also make sure that your microfone is of good quality, or else you will be unable to decode the transmission.