Lessons.txt

Lessons: 

I learned many things from this project, way more than when I started it a few years ago. First of all, it was definitely a good refresher of signal processing and the math behind it. Specially, the way filter works and how it implement it in hardware (or in this case in simulation).

For the things I don't want to forget, here they are:
	- Sampling frequency: Fs
			-> and so the Nyquist frequency is Fs/2
				This is the maximal frequency the signal can be at before being distorted by replicas of itself in the frequency domain
	- Carrier frequency: Fc
	- Message frequency: Fm
				This will depend of the symbol rate and how many bits the system sends per second
	- Demodulated signal frequency component: Fm + (2Fc - Fm) + (2Fc + Fm)
				Signal need to go through a low pass filter to remove the carrier frequency components
	- Cutoff frequency between Fm and 2Fc - Fm
	                     M-1
	- FIR filter: y(n) = sum [ h(k) x(n-k) ]
	                     k=0
	- order of FIR: M ~= Attenuation / (22 * (Fstop - Fpass) / Fs)
	- good website which explains pulse shaping: https://wirelesspi.com/pulse-shaping-filter/


Blocks in digital communication system:

bitstream -> QAM mapper -> Upsampling -> Pulse shaping -> Modulation ---------------------------------
																		  							  |
																									Noise
																		  							  |		Channel
																									Delay
																		  							  |
bitstream <- QAM demapper <- Downsampling <- Matched filter	<- Low Pass filter <- Demodulation	------
								  |													   |
						Symbol timing recovery								Frequency and phase recovery