path: root/chapter07/content_ch07.tex

% SPDX-License-Identifier: CC-BY-SA-4.0
%
% Copyright (c) 2020 Philipp Le
%
% Except where otherwise noted, this work is licensed under a
% Creative Commons Attribution-ShareAlike 4.0 License.
%
% Please find the full copy of the licence at:
% https://creativecommons.org/licenses/by-sa/4.0/legalcode

\chapter{Spread Spectrum and Multiple Access}

\begin{refsection}
	
The electromagnetic spectrum is a sparse resource. It must be used as efficient as possible because it is shared with numerous users and applications. However, modern digital communication systems use spread spectrum technologies that occupy a relatively wide frequency band

The purpose of spread spectrum is amongst others:
\begin{itemize}
	\item Immunity against noise and disturbances
	\item Encryption and confidentiality of the communication
	\item Plausible deniability that the communication had ever taken place
	\item Coexistence with other services (multiple access)
\end{itemize}

Especially, the multiple access is an important reason for employing spread spectrum technologies. In modern communication systems, the resource \emph{frequency} is not only allocated to a single user. For example, \ac{LTE} allows many users to access the service with high data rates and low latency. The resource \emph{frequency} must be shared. Efficient medium access relies on spread spectrum to achieve this.

\section{Spread Spectrum}

In the modulation techniques considered so far, are \emph{plain} or \emph{non-spread spectrum}. For the definition of spread spectrum signals, the symbol duration and the transmission bandwidth is investigated.
\begin{itemize}
	\item The signal duration is the amount of time required to convey an information. In digital communication system, an information is a symbol modulated onto a carrier. The signal duration is the symbol period $T_{sym}$.
	\item The signal bandwidth $\Delta f_{sym}$ is the minimum bandwidth required by the receiver to receive the signal. In the case of modulation, it is the \emph{transmission bandwidth}.
\end{itemize}
In the \ac{ASK}, \ac{PSK} and \ac{QAM} modulation of Chapter 5, the signal duration and the bandwidth are inversely proportional.
\begin{equation}
	\Delta f_{sym} = \frac{1}{T_{sym}}
\end{equation}

The product of the bandwidth and the duration -- the \index{time-bandwidth product} \textbf{time-bandwidth product} -- is constantly $1$.\footnote{This is the ideal case. For real implementations, the constant may differ from $1$ depending on the modulation technique. However, it will be close to $1$.}
\begin{equation}
	T_{sym} \cdot \Delta f_{sym} = 1
\end{equation}

\paragraph{Time-Bandwidth Product.}

The time-bandwidth product is used for the definition of spread spectrum:
\begin{definition}{Spread spectrum}
	The time-bandwidth product of \index{spread spectrum} \textbf{spread spectrum} signals is significantly greater than $1$.
	\begin{equation}
		T_{sym} \cdot \Delta f_{sym} \gg 1
	\end{equation}
\end{definition}

Typically, this means that the bandwidth $\Delta f_{sym}$ is increased while the symbol duration and thereby the symbol rate is kept constant.
\begin{itemize}
	\item The symbol sequence (time-bandwidth product of $1$) is altered in a way which distributes the signal power over a wider frequency band. This process is called \index{spreading} \textbf{spreading}.
	\item The inverse process is \index{despreading} \textbf{despreading}. The original symbol sequence is reconstructed from the wide-band spread spectrum signal.
\end{itemize}

\begin{figure}[H]
	\centering
	\begin{tikzpicture}
		\begin{axis}[
			height={0.15\textheight},
			width=0.8\linewidth,
			scale only axis,
			xlabel={$\omega$},
			ylabel={$|\mathrm{S}_{XX}|$},
			%grid style={line width=.6pt, color=lightgray},
			%grid=both,
			grid=none,
			legend pos=north east,
			axis y line=middle,
			axis x line=middle,
			every axis x label/.style={
				at={(ticklabel* cs:1.05)},
				anchor=north,
			},
			every axis y label/.style={
				at={(ticklabel* cs:1.05)},
				anchor=east,
			},
			xmin=0,
			xmax=10.5,
			ymin=0,
			ymax=1.2,
			xtick={0},
			xticklabels={0},
			ytick={0},
			axis x discontinuity=parallel,
		]
			\addplot[blue, smooth] coordinates {(4.6,0) (4.7,0.02) (4.8,0.2) (4.9,0.71) (5,1) (5.1,0.71) (5.2,0.2) (5.3,0.02) (5.4,0)};
			\addlegendentry{Narrow-band signal};
			\addplot[red, smooth] coordinates {(2,0) (2.5,0.01) (3,0.05) (5,0.05) (7,0.05) (7.5,0.01) (8,0)};
			\addlegendentry{Spread spectrum signal};
		\end{axis}
	\end{tikzpicture}
	\caption[PSD of a narrow-band and spread spectrum signal]{\acs{PSD} of a narrow-band and spread spectrum signal. Both signals carry the same information and have the equal power. The narrow-band signal concentrates the whole signal power in a narrow frequency band. In contrast, the spread spectrum signal distributes the signal power over a wide frequency band.}
\end{figure}

\paragraph{Noise-like Signal.}

The signal power remains constant while the signal is spread.
\begin{itemize}
	\item The \ac{PSD} is reduced.
	\item But, the \ac{PSD} is integrated over a wider frequency range.
	\item The overall power remains constant.
\end{itemize}

The \ac{PSD} of the spread spectrum signal is flat in approximation.
\begin{itemize}
	\item The flat \ac{PSD} resembles the \ac{PSD} of noise.
	\item Spread spectrum signal are therefore \emph{noise-like}.
\end{itemize}

A third party who has no knowledge of neither the existence of the spread spectrum signal nor the technology used cannot detect the signal.
\begin{itemize}
	\item The spread spectrum signal looks like noise or a wide-band disturbance from the view of the receiver which does not participate in the communication.
	\item This circumstance can be used to conceal the existence of the signal (plausible deniability of its existence).
\end{itemize}

\paragraph{Despreading.}

\index{despreading} Despreading reconstructs the symbols -- and thereby the data -- from the spread spectrum signal.
\begin{itemize}
	\item The spreading is reversed.
	\item The symbols (time-bandwidth product of $1$) are reconstructed. This can be seen like re-concentrating the spread signal power in a narrow-band symbol sequence.
	\item The disturbances which are uncorrelated to the spread spectrum signal are converted into the wide-band noise with a low \ac{PSD}.
	\begin{itemize}
		\item The wide-band noise floor (like thermal noise or quantization noise) remains wide-band.
		\item Strong but narrow-band disturbing signals (like other users of the electromagnetic spectrum) are spread to low-\acs{PSD} wide-band noise during the despreading. The \ac{SNR} is increased by spreading the signal power of the disturbance.
	\end{itemize}
\end{itemize}

\todo{Despreading in frequency-domain, suppression of noise and disturbances}

\subsection{Direct-Sequence Spread Spectrum}

A simple method for increasing the bandwidth systematically re-encoding the symbols using new symbols at a higher symbol rate.
\begin{itemize}
	\item The input symbols are at the rate of $f_{sym}$.
	\item The \emph{spreading} is achieved by re-encoding each symbol by $L$ new symbols, called \index{chip} \textbf{chips}.
	\item After re-encoding, the rate of the new symbols -- the \emph{chips} -- is $f_{chp}$. $f_{chp}$ is the \index{chip rate} \textbf{chip rate}. \nomenclature[Sf]{$f_{chp}$}{Chip rate}
	\item $L$ is the \index{spreading factor} \textbf{spreading factor}.
	\begin{equation}
		L = \frac{f_{chp}}{f_{sym}}
	\end{equation}
\end{itemize}

\begin{example}{\acs{IEEE} 802.11b (Wi-Fi)}
	\acs{IEEE} 802.11b is an early \ac{WLAN} standard from 1999, defining data rates of \SI{1}{Mbit/s} to \SI{11}{Mbit/s}.\footnote{In comparison to that, modern \ac{WLAN} standards like \acs{IEEE} 802.11ax have data rates of several \si{Gbit/s}. But they use other spread spectrum technologies.} Usually, the data rate is proportional to the transmission bandwidth. But, \acs{IEEE} 802.11b uses \ac{DSSS} to implement an adaptive data rate whilst retaining a constant bandwidth (approximately \SI{22}{MHz}).
	
	\begin{figure}[H]
		\centering
		\begin{circuitikz}
			\node[mixer](Spreader){};
			\node[draw,block,below=1.5cm of Spreader](Code){Code\\ generator};
			\node[draw,block,right=3cm of Spreader](Mod){\acs{BPSK}\\ modulator};
			
			\node[above=3mm of Spreader,align=center]{Multiplier as\\ the spreader};
			
			\draw[-o] (Spreader.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Input data\\ at $f_{sym}$};
			\draw (Code.north) -- node[midway,right,align=left]{Spreading code\\ at $f_{chip}$} (Spreader.south) node[inputarrow,rotate=90]{};
			\draw (Spreader.east) -- node[midway,above,align=center]{Chip\\ sequence} (Mod.west) node[inputarrow]{};
			\draw (Mod.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{Baseband\\ signal};
		\end{circuitikz}
		\caption{Simplified spreading and \acs{PSK} modulation signal chain of an \acs{IEEE} 802.11b conforming transmitter}
	\end{figure}

	Data is represented by bits. At a data rate of $f_{sym} = \SI{1}{Mbit/s}$, the data is spread by a factor $L = 11$. The data symbols are multiplied by the \emph{spreading code}. Because of the multiplication, the same block symbol as the mixer is used. The code has a \emph{chip rate} of $f_{chp} = L f_{sym} = \SI{11}{Mchp/s}$.\footnote{Physically, the units \si{bit} and \si{chp} are dimension-less. Therefore, $\SI{1}{Mbit/s} = \SI{1}{Mchp/s} = \SI{1}{MHz}$. However, the units shall refer to the quantity being considered.}
	
	\vspace{0.5em}
	
	For $L = 11$, the \index{Barker code} \textbf{Barker code} with $\vect{C}_{11} = \left[+1, +1, +1, -1, -1, -1, +1, -1, -1, +1, -1\right]$ is used as the spreading code. Let's consider a bit stream of $(10)_2$ as th data, which is encoded as the symbols $\vect{D} = \left[-1, +1\right]$. The spread sequence is:
	\begin{equation}
		\begin{split}
			\vect{S} &= \vect{D} \otimes \vect{C}_{11} \\
			 &= \left[\underbrace{-1, -1, -1, +1, +1, +1, -1, +1, +1, -1, +1}_{\text{Spread symbol } -1},\right. \\ &\qquad \left.  \underbrace{+1, +1, +1, -1, -1, -1, +1, -1, -1, +1, -1}_{\text{Spread symbol } +1}\right]
		\end{split}
	\end{equation}
	
	The \emph{chip sequence} $\vect{S}$ is at a rate of \SI{11}{Mchp/s}. In this case, the \emph{chip} have two discrete states $-1$ or $+1$. They can be modulated by a \ac{BPSK} modulator. It baseband is then mixed to the \ac{RF} band of \SI{2.4}{GHz} using an IQ modulator.
	
	\begin{figure}[H]
		\centering
		
		\subfloat[Data symbols]{
			\begin{tikzpicture}
				\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
				\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$D[n]$};
				\foreach \y in {-1, 0, +1}{
					\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
				}
				\foreach \x/\t in {5.5/1, 11/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\SI{\t}{\micro{}s}$};
				}
			
				\foreach \x/\n in {0/0, 5.5/1}{
					\draw[latex-latex] (\x,1.1) -- node[midway,above,align=center]{$D[\n]$} ({\x+5.5},1.1);
				}
				
				\draw[red,thick] (0,-1) -- (5.5,-1) -- (5.5,1) -- (11,1);
				
				\draw[latex-latex] (5.5,-1.1) -- node[midway,below,align=center]{$T_{sym} = 1/f_{sym}$} (11,-1.1);
			\end{tikzpicture}
		}
	
		\subfloat[Code chips]{
			\begin{tikzpicture}
				\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
				\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$C_l$};
				\foreach \y in {-1, 0, +1}{
					\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
				}
				\foreach \x/\t in {5.5/1, 11/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\SI{\t}{\micro{}s}$};
				}
				
				\draw[blue,thick] (0,0) \foreach \n in {0, 1} {
					\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
						-- ({(\n*5.5)+(\l*0.5)},\c) -- ({(\n*5.5)+(\l*0.5+0.5)},\c)
					}
				};
			
				%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
				%\draw[dashed] (11,-1.2) -- (11,1.2);
				\foreach \n in {0, 1} {
					\foreach \l in {0,1,...,10}{
						\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
						\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
					}
				}
			
				\foreach \x/\n in {0/0, 5.5/1}{
					\draw[latex-latex] (\x,-1.2) -- node[midway,below,align=center]{$\vect{C}_L$} ({\x+5.5},-1.2);
				}
			\end{tikzpicture}
		}
	
		\subfloat[Output chips]{
			\begin{tikzpicture}
				\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = m \cdot T_{chp}$};
				\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$S[m]$};
				\foreach \y in {-1, 0, +1}{
					\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
				}
				\foreach \x/\t in {5.5/1, 11/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\SI{\t}{\micro{}s}$};
				}
				
				\draw[olive,thick] (0,0) \foreach \n/\d in {0/-1, 1/+1} {
					\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
						-- ({(\n*5.5)+(\l*0.5)},{\c*\d}) -- ({(\n*5.5)+(\l*0.5+0.5)},{\c*\d})
					}
				};
			
				%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
				%\draw[dashed] (11,-1.2) -- (11,1.2);
				\foreach \m in {0,1,...,21}{
					\node[anchor=west,rotate=90,align=center] at({(\m*0.5)+0.25},1.1) {\scriptsize $S[\m]$};
					\draw[dashed] ({(\m+1)*0.5},-1.2) -- ({(\m+1)*0.5},1.2);
				}
			
				\draw[latex-latex] (6,-1.1) -- node[midway,below,align=center]{$T_{chp} = 1/f_{chp}$} (6.5,-1.1);
			\end{tikzpicture}
		}
	
		\caption{Data, spreading code and output chips of IEEE 802.11b symbols at a spreading factor of $L = 11$}
		\label{fig:ch07:wlan_dsss}
	\end{figure}
	
	At a data rate of \SI{11}{Mbit/s}, the data is \underline{not} spread ($L = 1$). The spreading code is $\vect{C}_{1} = \left[+1\right]$. The \emph{chip rate} equals the data bit rate. The \emph{chip rate} remains constant at \SI{11}{Mchp/s}.
	
	\vspace{0.5em}
	
	\textbf{Why adaptive data rate?} Spreading by $L = 11$ increases the \ac{SNR}. Each symbol is effectively repeated 11 times. This circumstance can be used to achieve a processing gain and increase noise immunity. The higher data rate at a lower spreading factor of $L = 1$ comes at the drawback of decreased noise immunity.
\end{example}

The \index{direct-sequence spread spectrum} \textbf{\acf{DSSS}} spreads the symbols by multiplying each data symbol with the whole spreading code.
\begin{itemize}
	\item The data symbol sequence is represented by the vector $\vect{D}$. A single data symbol is $D[n]$. \nomenclature[Sd]{$\vect{D}$}{Data symbol sequence}
	\begin{equation}
		\vect{D} = \left[D[0], D[1], \dots\right]
	\end{equation}
	\item The data comes at the rate of $f_{sym}$.
	\item The code sequence is represented by the $L$-dimensional vector $\vect{C}_L$. $L$ is the length of the code. A single chip of the code is $C_l$. \nomenclature[Sd]{$\vect{C}_L$}{Spreading code for direct-sequence spread spectrum}
	\begin{equation}
		\vect{C} = \left[C_0, C_1, \dots, C_{L-1}\right]
	\end{equation}
	\item The code repeats at $f_{sym}$. The rate of the single code chips is $f_{chp} = L f_{sym}$.
\end{itemize}
The process of multiplying each data symbol $D_n$ with the whole code sequence $\left[C_0, C_1, \dots, C_{L-1}\right]$ is represented by the \index{Kronecker product} \textbf{Kronecker product} $\otimes$. \nomenclature[Fk]{$\otimes$}{Kronecker product}
\begin{equation}
	\begin{split}
		\vect{S} &= \vect{D} \otimes \vect{C}_{L} \\
		S[m] &= S[n L + l] = D[n] \cdot C_l \qquad \forall 0 \leq l < L
	\end{split}
\end{equation}
$\vect{S} = \left[S[0], S[1], \cdots\right]$ is the sequence of output chips -- the \emph{spread spectrum signal}. The chips in $\vect{S}$ are at the chip rate of $f_{chp} = M f_{sym}$. \nomenclature[Ss]{$\vect{S}$}{Chip sequence of a spread spectrum signal}

\begin{figure}[H]
	\centering
	\begin{circuitikz}
		\node[draw,block](Spreader){Spreader\\ $\vect{D} \otimes \vect{C}_{L}$};
		\node[draw,block,below=1.5cm of Spreader](Code){Code\\ generator};
		\node[draw,block,right=2.5cm of Spreader](Mod){Modulation\\ (\acs{BPSK}, \acs{QPSK},\\ \acs{QAM}, ...)};
		
		\draw[-o] (Spreader.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Input symbol\\ sequence $\vect{D}$};
		\draw (Code.north) -- node[midway,right,align=left]{Spreading\\ code $\vect{C}_{L}$} (Spreader.south) node[inputarrow,rotate=90]{};
		\draw (Spreader.east) -- node[midway,above,align=center]{Chip\\ sequence\\ $\vect{S}$} (Mod.west) node[inputarrow]{};
		\draw (Mod.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{Baseband\\ signal};
	\end{circuitikz}
	\caption{Abstract \acs{DSSS}}
	\label{fig:ch07:abstract_dsss}
\end{figure}

Figure \ref{fig:ch07:abstract_dsss} depicts an abstract view on \ac{DSSS}. The \emph{spreading code} is a \index{pseudorandom code} \textbf{pseudorandom code}.
\begin{itemize}
	\item The code is generated by an algorithm and is thereby predictable if the algorithm is known.
	\item For a receiver that does not know the code generation algorithm, the code sequence is random. It is noise-like.
	\item \textit{The fact, that the code must be known to the receiver, can be used to implement data \texttt{}encryption.}
\end{itemize}

Time-bandwidth product:
\begin{itemize}
	\item The bandwidth of the signal is the \emph{chip rate} $f_{chp}$. It is $L$ times the bandwidth of a non-spread symbol.
	\item The symbol period is not touched and remains $T_{sym}$.
	\item The transmission bandwidth of a non-spread symbol $\frac{1}{T_{sym}}$.
	\item The time-bandwidth product is
	\begin{equation}
		T_{sym} \cdot \Delta f_{chp} = T_{sym} \cdot \frac{L}{T_{sym}} = L \gg 1
	\end{equation}
	\item The \ac{DSSS} fulfils the spread spectrum condition.
\end{itemize}

Example usage of \ac{DSSS}:
\begin{itemize}
	\item \acs{IEEE} 802.11b specification for \ac{WLAN}
	\item \ac{GPS}
\end{itemize}

\subsection{Frequency-Hopping Spread Spectrum}

Another straightforward method spreading a signal across the frequency spectrum is transmitting each symbol at another frequency. This technique is called \index{frequency-hopping spread spectrum} \textbf{\acf{FHSS}}.
\begin{itemize}
	\item The frequency band $\Delta f_{FHSS}$ is divided into $M$ \index{sub-band} \textbf{sub-bands} with a bandwidth of $\Delta f_{sub}$.
	\begin{equation}
		\Delta f_{sub} = \frac{\Delta f_{FHSS}}{M}
		\label{eq:ch07:fhss_sub_f}
	\end{equation}
	\item The bandwidth of the sub-band $\Delta f_{sub}$ is the transmission bandwidth of a single, non-spread symbol. $\Delta f_{sub}$ is wide enough to accommodate a symbol.
	\item Each symbol is transmitted in a different sub-band $k$ ($0 \leq k < M$).
	\item The sub-band n is selected according to the certain pattern, which is again called \index{spreading code} \textbf{spreading code}.
\end{itemize}

\begin{figure}[H]
	\centering
	\begin{circuitikz}
		\node[draw,block](Mod){Modulation\\ (\acs{BPSK}, \acs{QPSK},\\ \acs{QAM}, ...)};
		\node[draw,block](Code) at([yshift=-3.5cm]Mod) {Code\\ generator};
		\node[mixer,right=2.5cm of Mod](Mix){};
		\node[oscillator](LO) at([yshift=-3.5cm,xshift=5mm]Mix) {};
		
		\node[right=3mm of LO,align=left]{\acs{LO}};
		
		\draw[-o] (Mod.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Input symbol\\ sequence $\vect{D}$};
		\draw (Mod.east) -- node[midway,above,align=center]{Baseband\\ signal} (Mix.west) node[inputarrow,rotate=0]{};
		\draw (Code.east) -- node[midway,below,align=center]{Spreading\\ code $C_M[n]$} (LO.west) node[inputarrow,rotate=0]{};
		\draw (LO.north) -- node[midway,right,align=left]{Carrier} (Mix.south) node[inputarrow,rotate=90]{};
		\draw (Mix.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{\acs{RF}\\ signal};
	\end{circuitikz}
	\caption{Simplified block diagram of a \acs{FHSS} transmitter}
\end{figure}

Following can be said about the \emph{spreading code}:
\begin{itemize}
	\item The values of the \emph{spreading code} $C_M[n]$ define the sub-band $k$ where the symbol $D_n$ shall be transmitted. \nomenclature[Sd]{$\vect{C}_;$}{Spreading code for time-hopping or frequency-hopping spread spectrum}
	\begin{equation}
		k = C_M[n] \qquad , 0 \leq C_M[n] < M
		\label{eq:ch07:fhss_sub_index}
	\end{equation}
	One code sample $C_M[n]$ corresponds to the data symbol $D[n]$.
	\item The length of the \emph{spreading code} is generally not restricted. It can be repeating. But a repeating pattern is not required.
	\item The rule which generates the \emph{spreading code} (the algorithm) must be known to both the transmitter and receiver.
\end{itemize}

If $f_c$ is the centre frequency of the carrier, the exact transmission frequency $f_{n}$ of the symbol $S_n$ is:
\begin{equation}
	\begin{split}
		f_n &= \underbrace{f_c - \frac{\Delta f_{FHSS}}{2} + \frac{\Delta f_{sub}}{2}}_{\text{Centre frequency of the 1st sub-band}} + \underbrace{k \cdot \Delta f_{sub}}_{\text{Sub-band centre frequency offset}} \\
		 &\quad \text{Using \eqref{eq:ch07:fhss_sub_f} and \eqref{eq:ch07:fhss_sub_index}:} \\
		 &= f_c - \frac{M \cdot \Delta f_{sub}}{2} + \frac{\Delta f_{sub}}{2} + C_M[n] \cdot \Delta f_{sub} \\
		 &= f_c + \Delta f_{sub} \left(\frac{1-M}{2} + C_M[n] \right)
	\end{split}
\end{equation}

\begin{example}{\acs{FHSS}}
	Let's consider an example symbol sequence $\vect{D}$ with a length of $6$. The signal is spread by $M =4$.
	
	The spreading code generated by the algorithm is:
	\begin{equation}
		\vect{C}_4[n] = \left[C_4[0], C_4[1], \dots\right] = \left[1,0,3,2,0,1,\dots\right]
	\end{equation}
	
	The frequency usage can be plotted over time:
	\begin{figure}[H]
		\centering
		\begin{tikzpicture}[x=1cm,y=1cm]
			\draw[-latex] (0,0) -- (6.5,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
			\draw[-latex] (0,0) -- (0,0.4) (0,0.6) -- (0,5.5) node[left, align=right]{$f$};
			\draw (-0.3,0.4) -- (0.3,0.4);
			\draw (-0.3,0.6) -- (0.3,0.6);
			
			\draw (0,3) -- (-0.2,3) node[left,align=right]{$f_c$};
			\draw[latex-latex] (-1,2) -- node[midway,left,align=center,anchor=south,rotate=90]{$\Delta f_{sub}$} (-1,3);
			\draw[latex-latex] (-2,1) -- node[midway,left,align=center,anchor=south,rotate=90]{$\Delta f_{FHSS}$} (-2,5);
			\draw[latex-latex] (1,-0.5) -- node[midway,below,align=center]{$T_{sym} = 1/f_{sym}$} (2,-0.5);
		
			\foreach \k in {0, 1, 2, 3}{
				\draw[dotted] (0,{\k+1}) -- (8,{\k+1});
				\node[right,align=left] at(7,{\k+1.5}) {\small\itshape Sub-band $k = \k$};
			}
			\draw[dotted] (0,5) -- (8,5);
			
			\foreach \n/\c in {0/1, 1/0, 2/3, 3/2, 4/0, 5/1}{
				\node[fill=gray!50, draw=black, minimum size=1cm, anchor=south west] at({\n},{\c+1}) {\footnotesize $D[\n]$};
			}
		\end{tikzpicture}
		\caption[Time-frequency plot of a symbol sequence spread by a \acs{FHSS} with $M = 4$]{Time-frequency plot of a symbol sequence spread by a \acs{FHSS} with $M = 4$. The $M = 4$ sub-bands are distributed around the centre frequency $f_c$. Each sub-band has the bandwidth $\Delta f_{sub}$. The duration of one symbol is $T_{sym}$ (symbol period).}
		\label{fig:ch07:fhss_ex}
	\end{figure}

	Each symbol in Figure \ref{fig:ch07:fhss_ex} is modulated (\ac{BPSK}, \ac{QPSK}, \ac{QAM}, ...).
\end{example}

Time-bandwidth product:
\begin{itemize}
	\item The bandwidth of the signal is $\Delta f_{FHSS} = M \Delta f_{sub}$. It is $M$ times the bandwidth of a non-spread symbol.
	\item The symbol period is not touched and remains $T_{sym}$.
	\item The sub-band bandwidth is the transmission bandwidth of a non-spread symbol $\Delta f_{sub} = \frac{1}{T_{sym}}$.
	\item The time-bandwidth product is
	\begin{equation}
		T_{sym} \cdot \Delta f_{FHSS} = M \gg 1
	\end{equation}
	\item The \ac{FHSS} fulfils the spread spectrum condition.
\end{itemize}

Benefits:
\begin{itemize}
	\item The benefits are the same as for \ac{DSSS} and other \emph{spread spectrum} techniques.
	\item The noise and disturbance immunity is improved.
\end{itemize}

Example usage of \ac{FHSS}:
\begin{itemize}
	\item \acs{IEEE} 802.15.1 (Bluetooth) -- Bluetooth employs a more sophisticated frequency-hopping scheme. The sub-bands are not spaced equally. The sub-band are located in the gaps between the \ac{WLAN} bands in the \SI{2.4}{GHz} band. This reduces the mutual interference of Bluetooth and \ac{WLAN}. Both services can coexists with minimum disturbance.
\end{itemize}

\subsection{Time-Hopping Spread Spectrum}

A third way of spreading the signal power across the frequency is shortening the symbol width $T_{chp}$ while keeping time symbol repetition period $T_{sym}$ constant. This technique is called \index{time-hopping spread spectrum} \textbf{\acf{THSS}}.
\begin{itemize}
	\item The transmission bandwidth is increased to $f_{sym} = \frac{1}{T_{chp}}$.
	\item The symbol is now a pulse -- called \index{chip} \textbf{chip} -- in the time-domain with a width of $T_{chp}$. A rest of the time portion, until the next symbol is transmitted, is zero.
	\item The pulse/symbol width is related to the symbol period $T_{sym}$.
	\begin{equation}
		T_{sym} = M \cdot T_{chp}
	\end{equation}
	\item The time-bandwidth product is
	\begin{equation}
		T_{sym} f_{sym} = M T_{chp} \frac{1}{T_{chp}} = M \gg 1
	\end{equation}
	The requirement for \emph{spread spectrum} signals is fulfilled.
\end{itemize}

The shortened pulse is then shifted in time.
\begin{itemize}
	\item The symbol period $T_{sym}$ is divided into $M$ time slots of equal length $T_{chp} = \frac{T_{sym}}{M}$.
	\item The shortened symbol is shifted to the $n$-th time slot.
	\item The values of the \emph{spreading code} $C_M[n]$ define the time slot $m$ (pulse position) where the pulse of the symbol $D_n$ shall be transmitted.
	\begin{equation}
		m = C_M[n] \qquad , 0 \leq C_M[n] < M
		\label{eq:ch07:thss_sub_index}
	\end{equation}
	One code sample $C_M[n]$ corresponds to the data symbol $D[n]$.
	\item The length of the \emph{spreading code} is generally not restricted. It can be repeating. But a repeating pattern is not required.
	\item The rule which generates the \emph{spreading code} (the algorithm) must be known to both the transmitter and receiver.
\end{itemize}

\begin{figure}[H]
	\centering
	\begin{circuitikz}
		\node[draw,block](Mod){Modulation\\ (\acs{BPSK}, \acs{QPSK},\\ \acs{QAM}, ...)};
		\node[draw,block,right=1.5cm of Mod](Shifter){Time-shifter\\ and pulse shaper};
		\node[draw,block,below=2cm of Shifter](Code){Code\\ generator};
		
		\draw[-o] (Mod.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Input symbol\\ sequence $\vect{D}$};
		\draw (Mod.east) -- (Shifter.west) node[inputarrow,rotate=0]{};
		\draw (Code.north) -- node[midway,right,align=left]{Spreading\\ code $C_M[n]$} (Shifter.south) node[inputarrow,rotate=90]{};
		\draw (Shifter.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{Baseband signal\\ (chip sequence $\vect{S}$)};
	\end{circuitikz}
	\caption[Simplified block diagram of a \acs{THSS} transmitter]{Simplified block diagram of a \acs{THSS} transmitter. The time-shifter and pulse shaper play a key role. The pulse shaper forms the shortened pulse with a width of $T_{chp} = \frac{T_{sym}}{M}$. The time-shifter shifts the pulse to the $C_n$-th time slot.}
\end{figure}

The output chip sequence $\vect{S} = \left[S[0], S[1], \dots\right]$ is at the rate $f_{chp} = M \cdot f_{sym}$. It is defined by:
\begin{equation}
	S[m] = S[n M + l] = \begin{cases}
		D[n] &\quad \text{if } l = C_M[n], \\
		0 &\quad \text{if } l \neq C_M[n].
	\end{cases}
\end{equation}

\begin{example}{\acs{THSS}}
	A data symbol sequence is spread by the spreading code:
	\begin{equation}
		\vect{C}_4[n] = \left[C_4[0], C_4[1], \dots\right] = \left[1,3,\dots\right]
	\end{equation}
	
	\begin{figure}[H]
		\centering
		
		\subfloat[Data symbols]{
			\begin{tikzpicture}[x=1cm,y=1cm]
				\draw[-latex] (0,0) -- (9,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
				\draw[-latex] (0,0) -- (0,2.5) node[left, align=right]{$D[n]$};
				\foreach \x/\t in {4/1, 8/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
				}
			
				\foreach \n in {0, 1}{
					\node[fill=gray!50, draw=black, minimum width=4cm, minimum height=2cm, anchor=south west] at({\n*4},0) {$D[\n]$};
				}
			\end{tikzpicture}
		}
	
		\subfloat[Spreading code]{
			\begin{tikzpicture}
				\draw[-latex] (0,0) -- (9,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
				\draw[-latex] (0,0) -- (0,2.5) node[left, align=right]{$C_4[n]$};
				\foreach \x/\t in {4/1, 8/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
				}
				
				\foreach \n/\c in {0/1, 1/3}{
					\node[fill=gray!50, draw=black, minimum width=4cm, minimum height=2cm, anchor=south west] at({\n*4},0) {$C_4[\n] = \c$};
				}
			\end{tikzpicture}
		}
		
		\subfloat[Output chips]{
			\begin{tikzpicture}
				\draw[-latex] (0,0) -- (9,0) node[below right, align=left]{$t = m \cdot T_{chp}$};
				\draw[-latex] (0,0) -- (0,2.5) node[left, align=right]{$S[m]$};
				\foreach \x/\t in {4/1, 8/2}{
					\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
				}
			
				\foreach \n in {0, 1}{
					\foreach \m in {1, 2, 3, 4}{
						\draw[dashed] ({(\n*4)+(\m)},0) -- ({(\n*4)+(\m)},2.5);
					}
				}
			
				\foreach \n/\c in {0/1, 1/3}{
					\node[fill=gray!50, draw=black, minimum width=1cm, minimum height=2cm, anchor=south west] at({(\n*4)+(\c)},0) {\footnotesize $D[\n]$};
				}
			
				\draw[latex-latex] (1,2.4) -- node[midway,above,align=center]{$T_{chp} = 1/f_{chp}$} (2,2.4);
			\end{tikzpicture}
		}
		
		\caption[\ac{THSS} signal]{\ac{THSS} signal. The symbol is shortened to a chip whose position is chosen by the spreading code.}
	\end{figure}
\end{example}

Benefits are the same as in other spread spectrum techniques. The noise immunity is improved.

Example usage of \ac{THSS}:
\begin{itemize}
	\item \ac{UWB} systems conforming to \acs{IEEE} 802.15.4
\end{itemize}

\subsection{Symbol Reconstruction}

\index{despreading} \textbf{Despreading} reconstructs the symbol sequence from the received \emph{spread spectrum} signal.

A despreader uses a cross-correlator to detect the spread symbols in the received signal. Remember that this is one of the purposes of cross-correlation.
\begin{itemize}
	\item The received signal is correlated with the \emph{spreading code}.
	\item The symbols spread with the same \emph{spreading code} can thereby be detected in the received signal.
\end{itemize}

\begin{remark}
	You may wonder why the autocorrelation is not used. Both receiver and transmitter use the same \emph{spreading code}. In deed, this a common misunderstanding. The received signal is different from the code and from the transmitted signal. The received signal contains noise and is subject to the transfer function of the transmission channel.
\end{remark}

\begin{figure}[H]
	\centering
	\begin{adjustbox}{scale=0.7}
		\begin{circuitikz}
			\node[draw,block](Demod){Demodulation\\ (\acs{BPSK}, \acs{QPSK},\\ \acs{QAM}, ...)};
			\node[draw,block,right=3cm of Demod](Corr){Cross corelation};
			\node[draw,block,below=1.5cm of Corr](Code){Code\\ generator};
			\node[draw,block,right=1.5cm of Corr](Det){Detector};
			
			\draw[-o] (Demod.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Received signal $r(t)$};
			\draw (Code.north) -- node[midway,right,align=left]{Spreading\\ code $\vect{C}$} (Corr.south) node[inputarrow,rotate=90]{};
			\draw (Demod.east) -- node[midway,above,align=center]{Received chip\\ sequence $\vect{R}$} (Corr.west) node[inputarrow]{};
			\draw (Corr.east) -- node[midway,above,align=center]{$\mathrm{R}_{RC}$} (Det.west) node[inputarrow]{};
			\draw (Det.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{Data};
		\end{circuitikz}
	\end{adjustbox}
	\caption{Abstract despreader using a cross-correlator}
	\label{fig:ch07:abstract_spreader}
\end{figure}

\subsubsection{Despreading of DSSS Signals}

\begin{figure}[H]
	\centering
	\begin{adjustbox}{scale=0.7}
		\begin{circuitikz}
			\node[draw,block](Demod){Demodulation\\ (\acs{BPSK}, \acs{QPSK},\\ \acs{QAM}, ...)};
			\node[mixer,right=3cm of Demod](Mul){};
			\node[draw,block,right=1cm of Mul](Acc){Accumulator};
			\node[draw,block,below=2cm of Mul](Code){Code\\ generator};
			\node[draw,block,right=1cm of Acc](Det){Detector};
			
			\node[above=3mm of Mul,align=center]{Multiplier as\\ despreader};
			
			\draw[-o] (Demod.west) node[inputarrow]{} -- ++(-1cm,0) node[left,align=right]{Received signal $r(t)$};
			\draw (Code.north) -- node[midway,right,align=left]{Spreading\\ code $\vect{C}_L$} (Mul.south) node[inputarrow,rotate=90]{};
			\draw (Demod.east) -- node[midway,below,align=center]{Received chips\\ $\vect{R}$} (Mul.west) node[inputarrow]{};
			\draw (Mul.east) -- (Acc.west) node[inputarrow]{};
			\draw (Acc.east) -- (Det.west) node[inputarrow]{};
			\draw (Det.east) -- ++(1cm,0) node[inputarrow]{} node[right,align=left]{Data};
			
			\draw[dashed] ([shift={(5mm,2cm)}]Acc.north east) node[anchor=south east,align=right]{Cross-correlation} -- ++(0,-3.5cm) -- ++(-7cm,0) -- ++(0,3.5cm) -- cycle;
		\end{circuitikz}
	\end{adjustbox}
	\caption{Simplified \acs{DSSS} despreader}
	\label{fig:ch07:abstract_dsss_spreader}
\end{figure}

\begin{itemize}
	\item The spreading, which was a multiplication of the data with the \emph{spreading code}, is reversed by multiplying the received chips again with the \emph{spreading code}.
	\item The accumulation then sums up the correlated chips.
	\item When the full code length has been accumulated:
	\begin{itemize}
		\item The detector decides which data symbol is received based on the correlator output.
		\item The correlator is reset in order to receive the next symbol.
	\end{itemize}
\end{itemize}

Let's consider three situations based on the IEEE 802.11b examples (Figure \ref{fig:ch07:wlan_dsss}):
\begin{itemize}
	\item Reception of a noise-free signal
	\item Reception of pure noise
	\item Reception of a noisy signal
\end{itemize}

\begin{figure}[H]
	\centering
	
	\subfloat[Received chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = m \cdot T_{chp}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$R[m]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[olive,thick] (0,0) \foreach \n/\d in {0/-1, 1/+1} {
			\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
				-- ({(\n*5.5)+(\l*0.5)},{\c*\d}) -- ({(\n*5.5)+(\l*0.5+0.5)},{\c*\d})
			}
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \m in {0,1,...,21}{
			\node[anchor=west,rotate=90,align=center] at({(\m*0.5)+0.25},1.1) {\scriptsize $R[\m]$};
			\draw[dashed] ({(\m+1)*0.5},-1.2) -- ({(\m+1)*0.5},1.2);
		}
		
		\draw[latex-latex] (6,-1.1) -- node[midway,below,align=center]{$T_{chp} = 1/f_{chp}$} (6.5,-1.1);
		\end{tikzpicture}
	}
	
	\subfloat[Code chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$C_l$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[blue,thick] (0,0) \foreach \n in {0, 1} {
			\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
				-- ({(\n*5.5)+(\l*0.5)},\c) -- ({(\n*5.5)+(\l*0.5+0.5)},\c)
			}
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		
		\foreach \x/\n in {0/0, 5.5/1}{
			\draw[latex-latex] (\x,-1.2) -- node[midway,below,align=center]{$\vect{C}_L$} ({\x+5.5},-1.2);
		}
		\end{tikzpicture}
	}
	
	\subfloat[Accumulation of cross-correlation. The grey circles are the sampling (detection) and reset points.]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$\sum \mathrm{R}_{RC}$};
		\foreach \y/\v in {-1/-11, 0, +1/+11}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\v$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[green,thick] (0,0) -- ++(5.5,-1);
		\draw[fill=gray!50,draw] (5.5,-1) ++(0:0.1) arc(0:360:0.1);
		\draw[green,thick] (5.5,0) -- ++(5.5,1);
		\draw[fill=gray!50,draw] (11,1) ++(0:0.1) arc(0:360:0.1);
		
		\foreach \y in {-0.5,0.5}{
			\draw[orange] (0,\y) -- (11,\y) node[anchor=south east,align=right]{Data detection threshold};
		}
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				%\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		\end{tikzpicture}
	}
	
	\subfloat[Data symbols]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$D[n]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\foreach \x/\n in {0/0, 5.5/1}{
			\draw[latex-latex] (\x,1.1) -- node[midway,above,align=center]{$D[\n]$} ({\x+5.5},1.1);
		}
		
		\draw[red,thick] (0,-1) -- (5.5,-1) -- (5.5,1) -- (11,1);
		\end{tikzpicture}
	}
	
	\caption{Despreading of a noise-free \acs{DSSS} signal}
\end{figure}

\begin{figure}[H]
	\centering
	
	\subfloat[Received chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = m \cdot T_{chp}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$R[m]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[olive,thick] (0,0) \foreach \m/\r in {0/-1, 1/-1, 2/+1, 3/+1, 4/-1, 5/-1, 6/-1, 7/+1, 8/+1, 9/+1, 10/+1, 11/-1, 12/+1, 13/+1, 14/+1, 15/+1, 16/-1, 17/-1, 18/-1, 19/+1, 20/+1, 21/-1}{
				-- ({(\m*0.5)},{\r}) -- ({(\m*0.5+0.5)},{\r})
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \m in {0,1,...,21}{
			\node[anchor=west,rotate=90,align=center] at({(\m*0.5)+0.25},1.1) {\scriptsize $R[\m]$};
			\draw[dashed] ({(\m+1)*0.5},-1.2) -- ({(\m+1)*0.5},1.2);
		}
		
		\draw[latex-latex] (6,-1.1) -- node[midway,below,align=center]{$T_{chp} = 1/f_{chp}$} (6.5,-1.1);
		\end{tikzpicture}
	}
	
	\subfloat[Code chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$C_l$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[blue,thick] (0,0) \foreach \n in {0, 1} {
			\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
				-- ({(\n*5.5)+(\l*0.5)},\c) -- ({(\n*5.5)+(\l*0.5+0.5)},\c)
			}
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		
		\foreach \x/\n in {0/0, 5.5/1}{
			\draw[latex-latex] (\x,-1.2) -- node[midway,below,align=center]{$\vect{C}_L$} ({\x+5.5},-1.2);
		}
		\end{tikzpicture}
	}
	
	\subfloat[Accumulation of cross-correlation. The grey circles are the sampling (detection) and reset points.]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$\sum \mathrm{R}_{RC}$};
		\foreach \y/\v in {-1/-11, 0, +1/+11}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\v$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[green,thick] (0,0) \foreach \v in {-1,-1,+1,-1,+1,+1,+1,-1,-1,+1,-1}{ -- ++(0.5,{\v/11})};
		\draw[fill=gray!50,draw] (5.5,{-1/11}) ++(0:0.1) arc(0:360:0.1);
		\draw[green,thick] (5.5,0) \foreach \v in {-1,+1,+1,-1,-1,+1,-1,+1,-1,+1,+1}{ -- ++(0.5,{\v/11})};
		\draw[fill=gray!50,draw] (11,{+1/11}) ++(0:0.1) arc(0:360:0.1);
		
		\foreach \y in {-0.5,0.5}{
			\draw[orange] (0,\y) -- (11,\y) node[anchor=south east,align=right]{Data detection threshold};
		}
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				%\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		\end{tikzpicture}
	}
	
	\subfloat[Data symbols]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$D[n]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[red,thick] (0,0) -- node[midway,above,align=center]{No data} (11,0);
		\end{tikzpicture}
	}
	
	\caption{\acs{DSSS} despreading of pure noise}
\end{figure}

\begin{figure}[H]
	\centering
	
	\subfloat[Received chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = m \cdot T_{chp}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$R[m]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
	
		\draw[olive,thick] (0,0) \foreach \m/\r in {0/-1, 1/-1, 2/-1, 3/-1, 4/+1, 5/+1, 6/-1, 7/+1, 8/+1, 9/+1, 10/+1, 11/+1, 12/+1, 13/+1, 14/-1, 15/+1, 16/-1, 17/-1, 18/-1, 19/-1, 20/+1, 21/-1}{
			-- ({(\m*0.5)},{\r}) -- ({(\m*0.5+0.5)},{\r})
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \m in {0,1,...,21}{
			\node[anchor=west,rotate=90,align=center] at({(\m*0.5)+0.25},1.1) {\scriptsize $R[\m]$};
			\draw[dashed] ({(\m+1)*0.5},-1.2) -- ({(\m+1)*0.5},1.2);
		}
		
		\draw[latex-latex] (6,-1.1) -- node[midway,below,align=center]{$T_{chp} = 1/f_{chp}$} (6.5,-1.1);
		\end{tikzpicture}
	}
	
	\subfloat[Code chips]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$C_l$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[blue,thick] (0,0) \foreach \n in {0, 1} {
			\foreach \l/\c in {0/+1, 1/+1, 2/+1, 3/-1, 4/-1, 5/-1, 6/+1, 7/-1, 8/-1, 9/+1, 10/-1}{
				-- ({(\n*5.5)+(\l*0.5)},\c) -- ({(\n*5.5)+(\l*0.5+0.5)},\c)
			}
		};
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		
		\foreach \x/\n in {0/0, 5.5/1}{
			\draw[latex-latex] (\x,-1.2) -- node[midway,below,align=center]{$\vect{C}_L$} ({\x+5.5},-1.2);
		}
		\end{tikzpicture}
	}
	
	\subfloat[Accumulation of cross-correlation. The grey circles are the sampling (detection) and reset points.]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.3) -- (0,1.5) node[left, align=right]{$\sum \mathrm{R}_{RC}$};
		\foreach \y/\v in {-1/-11, 0, +1/+11}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\v$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\draw[green,thick] (0,0) \foreach \v in {-1,-1,-1,+1,-1,-1,-1,-1,-1,+1,-1}{ -- ++(0.5,{\v/11})};
		\draw[fill=gray!50,draw] (5.5,{-9/11}) ++(0:0.1) arc(0:360:0.1);
		\draw[green,thick] (5.5,0) \foreach \v in {+1,+1,+1,+1,-1,+1,-1,+1,+1,+1,+1}{ -- ++(0.5,{\v/11})};
		\draw[fill=gray!50,draw] (11,{+9/11}) ++(0:0.1) arc(0:360:0.1);
		
		\foreach \y in {-0.5,0.5}{
			\draw[orange] (0,\y) -- (11,\y) node[anchor=south east,align=right]{Data detection threshold};
		}
		
		%\draw[dashed] (5.5,-1.2) -- (5.5,1.2);
		%\draw[dashed] (11,-1.2) -- (11,1.2);
		\foreach \n in {0, 1} {
			\foreach \l in {0,1,...,10}{
				%\node[anchor=west,rotate=90,align=center] at({((\n*11+\l)*0.5)+0.25},1.2) {\scriptsize $C_{\l}$};
				\draw[dashed] ({((\n*11+\l)+1)*0.5},-1.1) -- ({((\n*11+\l)+1)*0.5},1.2);
			}
		}
		\end{tikzpicture}
	}
	
	\subfloat[Data symbols]{
		\begin{tikzpicture}
		\draw[-latex] (0,0) -- (12,0) node[below right, align=left]{$t = n \cdot T_{sym}$};
		\draw[-latex] (0,-1.2) -- (0,1.5) node[left, align=right]{$D[n]$};
		\foreach \y in {-1, 0, +1}{
			\draw (0,\y) -- (-0.2,\y) node[left,align=right]{$\y$};
		}
		\foreach \x/\t in {5.5/1, 11/2}{
			\draw (\x,0) -- (\x,-0.2) node[below,align=center]{$\t T_{sym}$};
		}
		
		\foreach \x/\n in {0/0, 5.5/1}{
			\draw[latex-latex] (\x,1.1) -- node[midway,above,align=center]{$D[\n]$} ({\x+5.5},1.1);
		}
		
		\draw[red,thick] (0,-1) -- (5.5,-1) -- (5.5,1) -- (11,1);
		\end{tikzpicture}
	}
	
	\caption[Despreading of a noisy \acs{DSSS} signal]{Despreading of a noisy \acs{DSSS} signal. The noise shows up as flipped chips in the received chip sequence.}
\end{figure}


\subsubsection{Processing Gain}

In fact, the \ac{DSSS} despreading consists of
\begin{itemize}
	\item \textbf{Down-sampling}: The received and cross-correlated chips are down-sampled by $L$, the length of the \emph{spreading code} length.
	\item \textbf{Filtering}: The accumulator is in fact a \ac{FIR} low-pass filter.
\end{itemize}

The \emph{down-sampling} provides a \index{processing gain} \textbf{processing gain} of $L$.
\begin{itemize}
	\item The processing gain improves the \ac{SNR} by $+ \SI{10}{dB} \cdot \log_{10} \left(L\right)$.
	\item The spread spectrum signal is more immune against noise.
\end{itemize}

\begin{fact}
	Despreading of spread spectrum signals provides a processing gain.
\end{fact}

The processing gain is not only limited to wide-band (\ac{AWGN}, thermal, quantization, etc. noise). Narrow-band disturbances are suppressed as long as they are uncorrelated with the \emph{spreading code}. The cross-correlation has only a peak when the code matches and will be low if the code does not match. So, all uncorrelated superimposed signals can be removed. This is the strength of spread spectrum techniques.

\begin{figure}[H]
	\centering
	
	\subfloat[\acs{PSD} of the received signal \textbf{before} despreading]{
		\centering
		\begin{tikzpicture}
			\begin{axis}[
					height={0.15\textheight},
					width=0.8\linewidth,
					scale only axis,
					xlabel={$\omega$},
					ylabel={$|\mathrm{S}_{RR}|$},
					%grid style={line width=.6pt, color=lightgray},
					%grid=both,
					grid=none,
					legend pos=north east,
					axis y line=middle,
					axis x line=middle,
					every axis x label/.style={
						at={(ticklabel* cs:1.05)},
						anchor=north,
					},
					every axis y label/.style={
						at={(ticklabel* cs:1.05)},
						anchor=east,
					},
					xmin=0,
					xmax=10.5,
					ymin=0,
					ymax=1.2,
					xtick={0},
					xticklabels={0},
					ytick={0},
					axis x discontinuity=parallel,
				]
					\addplot[red, smooth] coordinates {(2,0) (2.5,0.01) (3,0.05) (5,0.05) (7,0.05) (7.5,0.01) (8,0)};
					\addlegendentry{Spread spectrum signal};
					\addplot[blue, smooth] coordinates {(4.6,0) (4.7,0.02) (4.8,0.2) (4.9,0.71) (5,1) (5.1,0.71) (5.2,0.2) (5.3,0.02) (5.4,0)};
					\addlegendentry{Narrow-band noise};
					\addplot[green, smooth] coordinates {(0,0.03) (10,0.03)};
					\addlegendentry{Wide-band noise};
			\end{axis}
		\end{tikzpicture}
	}

	\subfloat[\acs{PSD} of the received signal \textbf{after} despreading]{
		\centering
		\begin{tikzpicture}
			\begin{axis}[
				height={0.15\textheight},
				width=0.8\linewidth,
				scale only axis,
				xlabel={$\omega$},
				ylabel={$|\mathrm{S}_{\tilde{R}\tilde{R}}|$},
				%grid style={line width=.6pt, color=lightgray},
				%grid=both,
				grid=none,
				legend pos=north east,
				axis y line=middle,
				axis x line=middle,
				every axis x label/.style={
					at={(ticklabel* cs:1.05)},
					anchor=north,
				},
				every axis y label/.style={
					at={(ticklabel* cs:1.05)},
					anchor=east,
				},
				xmin=0,
				xmax=10.5,
				ymin=0,
				ymax=1.2,
				xtick={0},
				xticklabels={0},
				ytick={0},
				axis x discontinuity=parallel,
			]
				\addplot[red, smooth] coordinates {(4.6,0) (4.7,0.02) (4.8,0.2) (4.9,0.71) (5,1) (5.1,0.71) (5.2,0.2) (5.3,0.02) (5.4,0)};
				\addlegendentry{Spread spectrum signal};
				\addplot[blue, smooth] coordinates {(2,0) (2.5,0.01) (3,0.05) (5,0.05) (7,0.05) (7.5,0.01) (8,0)};
				\addlegendentry{Narrow-band noise};
				\addplot[green, smooth] coordinates {(0,0.03) (10,0.03)};
				\addlegendentry{Wide-band noise};
				
				\draw[latex-latex] (axis cs:4,0.05) -- node[midway,left,align=right]{Improved \acs{SNR}} (axis cs:4,1);
			\end{axis}
		\end{tikzpicture}
	}

	\caption[Processing gain of despreading]{Processing gain of despreading. The received signal is composed of the spread spectrum signal and two noise components (narrow-band and wide-band noise). After despreading the spread spectrum signal is reconstructed while both (uncorrelated) narrow-band and wide-band noise are attenuated. Despreading spreads the \acs{PSD} of the narrow-band noise, while the \ac{PSD} of the wide-band noise is not changed. The narrow-band noise does not interfere, even if it at the centre frequency of the spread spectrum signal like in this case.}
\end{figure}

\section{Multi-carrier Modulation}

\subsection{Inter-Carrier Interference}

\subsection{Orthogonal Frequency-Division Multiplex}

\todo{OFDM}

\section{Multiple Access}

\todo{Multiplexing}

\todo{Sharing Resources}

\subsection{Space-Division Multiple Access}

\subsection{Time-Division Multiple Access}

\subsection{Frequency-Division Multiple Access}

\subsection{Code-Division Multiple Access}

\subsection{Orthogonal Frequency-Division Multiple Access}

\section{Orthogonal Codes}

\nocite{ipatov2005}

\phantomsection
\addcontentsline{toc}{section}{References}
\printbibliography[heading=subbibliography]
\end{refsection}