Difference between revisions of "Digital Signal Transmission/Viterbi Receiver"

From LNTwww
 
(55 intermediate revisions by 6 users not shown)
Line 1: Line 1:
 
   
 
   
 
{{Header
 
{{Header
|Untermenü=Impulsinterferenzen und Entzerrungsverfahren
+
|Untermenü=Intersymbol Interfering and Equalization Methods
 
|Vorherige Seite=Optimale Empfängerstrategien
 
|Vorherige Seite=Optimale Empfängerstrategien
 
|Nächste Seite=Signale, Basisfunktionen und Vektorräume
 
|Nächste Seite=Signale, Basisfunktionen und Vektorräume
 
}}
 
}}
  
== Betrachtetes Szenario und Voraussetzungen ==
+
== Considered scenario and prerequisites ==
 
<br>
 
<br>
Der [[Digitalsignalübertragung/Optimale_Empfängerstrategien#Matched.E2.80.93Filter.E2.80.93Empf.C3.A4nger_vs._Korrelationsempf.C3.A4nger|Korrelationsempfänger]] ist im Sinne der [[Digitalsignalübertragung/Optimale_Empfängerstrategien#Maximum.E2.80.93Likelihood.E2.80.93Entscheidung_bei_Gau.C3.9Fscher_St.C3.B6rung| Maximum&ndash;Likelihood&ndash;Entscheidungsregel]] optimal, das heißt, er führt bei gleichwahrscheinlichen Quellensymbolen zur minimalen Fehlerwahrscheinlichkeit. Nachteilig ist:
+
The &nbsp;[[Digital_Signal_Transmission/Optimal_Receiver_Strategies#Matched_filter_receiver_vs._correlation_receiver|"correlation receiver"]]&nbsp; is optimal in the sense of the&nbsp;[[Digital_Signal_Transmission/Optimal_Receiver_Strategies#Maximum_likelihood_decision_for_Gaussian_noise|"maximum likelihood decision rule"]],&nbsp; that is,&nbsp; it leads to the minimum error probability for equally likely source symbols.&nbsp; Disadvantage:
*Der Realisierungsaufwand steigt exponentiell mit der Länge $N$ der zu detektierenden Symbolfolge.<br>
+
[[File:EN_Dig_T_3_8_S1.png|right|frame|Block diagram of the Viterbi receiver|class=fit]]
*Da die Folge gemeinsam entschieden wird, kommt es bei großem $N$ zu langen Verzögerungen.<br><br>
 
  
In den 1970er Jahren hat [https://de.wikipedia.org/wiki/Andrew_J._Viterbi Andrew J. Viterbi] einen Maximum&ndash;Likelihood&ndash;Empfänger vorgeschlagen, der die Detektion von Teilen der empfangenen Nachricht erlaubt und bei dem sich der Realisierungsaufwand auch bei unendlich langen Folgen in Grenzen hält.<br>
+
*The realization cost increases exponentially with the length &nbsp;$N$&nbsp; of the symbol sequence to be detected.<br>
  
[[File:P ID1467 Dig T 3 8 S1 version1.png|center|frame|Blockschaltbild des Viterbi-Empfängers|class=fit]]
+
*Since the sequence is decided jointly,&nbsp; long delays occur when &nbsp;$N$&nbsp; is large.<br><br>
  
Zu den einzelnen Komponenten des Blockschaltbildes ist anzumerken:
+
In the 1970s, &nbsp;[https://en.wikipedia.org/wiki/Andrew_Viterbi Andrew J. Viterbi]&nbsp; proposed a maximum likelihood receiver that allows detection of parts of the received message and for which the implementation cost is limited even for infinitely long sequences.<br>
*Das an den Empfangsgrundimpuls $g_r(t)$  und die Störung angepasste [[Stochastische_Signaltheorie/Matched-Filter|Matched&ndash;Filter]] $H_{\rm MF}(f)$ dient der ''Störleistungsbegrenzung''. Das MF&ndash;Ausgangssignal $m(t)$ bzw. die Folge $\langle m_\nu \rangle$ der äquidistanten Signalwerte nach der Abtastung besitzt das bestmögliche Signal&ndash;zu&ndash;Stör&ndash;Leistungsverhältnis.<br>
 
  
*Aufgabe des Dekorrelationsfilters $H_{\rm DF}(f)$ ist es, aus der Folge $\langle m_\nu \rangle$ die Detektionsabtastwerte $d_\nu = d_{{\rm S}\nu} + d_{{\rm N}\nu}$ zu gewinnen, deren Störanteile $d_{{\rm N}\nu}$ unkorreliert sind. Dieses Filter wird deshalb auch ''Whitening&ndash;Filter'' genannt.<br>
+
It should be noted with regard to the individual components of the block diagram:
 +
*Adjusted to the basic reception pulse &nbsp;$g_r(t)$&nbsp; and the noise,&nbsp; the &nbsp;[[Theory_of_Stochastic_Signals/Matched_Filter|"matched filter"]]&nbsp; $H_{\rm MF}(f)$&nbsp; serves for&nbsp; "noise power limitation".&nbsp; The matched filter output signal &nbsp;$m(t)$&nbsp; or the sequence &nbsp;$\langle m_\nu \rangle$&nbsp; of equidistant signal values after sampling has the best possible signal-to-noise power ratio&nbsp; $\rm (SNR)$.<br>
  
*Der ''Viterbi&ndash;Entscheider'', der im Mittelpunkt der folgenden Betrachtungen steht, gewinnt aus der Folge $\langle d_\nu \rangle$ seiner wertkontinuierlichen Eingangswerte die binäre Ausgangsfolge $\langle v_\nu \rangle$ entsprechend der Maximum&ndash;Likelihood&ndash;Regel mit der kleinstmöglichen Fehlerwahrscheinlichkeit ${\rm Pr}(v_\nu \ne q_\nu)$.<br>
+
*The task of the decorrelation filter &nbsp;$H_{\rm DF}(f)$&nbsp; is to extract from the sequence &nbsp;$\langle m_\nu \rangle$&nbsp; the detection samples &nbsp;$d_\nu = d_{{\rm S}\nu} + d_{{\rm N}\nu}$&nbsp; whose noise components &nbsp;$d_{{\rm N}\nu}$&nbsp; are uncorrelated.&nbsp; This filter is therefore also called a&nbsp; "whitening filter".&nbsp; <br>
  
 +
*The&nbsp; '''Viterbi decision''',&nbsp; which is the focus of the following considerations,&nbsp; obtains the binary output sequence &nbsp;$\langle v_\nu \rangle$&nbsp; from the continuous-valued input sequence &nbsp;$\langle d_\nu \rangle$&nbsp; according to the maximum likelihood rule with the smallest possible error probability &nbsp;${\rm Pr}(v_\nu \ne q_\nu)$.<br>
  
Um den ''Viterbi&ndash;Algorithmus'' möglichst einfach beschreiben zu können, werden hier einige vereinfachende Voraussetzungen getroffen:
 
*Die Amplitudenkoeffizienten seien ''unipolar'' &nbsp; &#8658; &nbsp; $a_\nu \in  \{0,\hspace{0.05cm} 1\}$. Anzumerken ist, dass es bei der Verwendung bipolarer Koeffizienten &nbsp; &#8658; &nbsp; $a_\nu \in  \{-1,\hspace{0.05cm} +1\}$ nur weniger Modifikationen  bedarf.<br>
 
  
*Der Grundimpuls $g_d(t)$ besteht nur aus dem Hauptwert $g_0 = g_d(t = T_{\rm D})$ und einem einzigen Vorläufer $g_{-1} = g_d(t = T_{\rm D}-T)$.<br>
+
In order to describe the Viterbi algorithm as simply as possible,&nbsp;  some simplifying assumptions are made here:
 +
#Let the amplitude coefficients be unipolar &nbsp; &#8658; &nbsp; $a_\nu \in  \{0,\hspace{0.05cm} 1\}$.&nbsp; Note that using bipolar coefficients &nbsp; &#8658; &nbsp; $a_\nu \in  \{-1,\hspace{0.05cm} +1\}$&nbsp; requires only a few modifications.&nbsp; These difficulties concern the description rather than the implementation.<br>
 +
#The basic pulse &nbsp;$g_d(t)$&nbsp; after the decorrelation filter &nbsp;$H_{\rm DF}(f)$&nbsp;consists only of the main value &nbsp;$g_0 = g_d(t = T_{\rm D})$&nbsp; and a single precursor &nbsp;$g_{-1} = g_d(t = T_{\rm D}-T)$.&nbsp; There are no trailers&nbsp; ("postcursors")&nbsp; in our model: &nbsp;$g_{\nu} = g_d(t = T_{\rm D}+\nu \cdot T)\equiv0$.&nbsp; Also,&nbsp; let other precursors be excluded for now:&nbsp; $g_{-2} = g_{-3}= \text{ ...} = 0$. <br>
 +
#This gives for the samples of the continuous decorrelation filter output signal&nbsp; $d(t)$&nbsp; at time&nbsp; $\nu \cdot T$: &nbsp; $d_{\nu} = a_{\nu}\cdot g_{0} + a_{\nu+1}\cdot g_{-1}+d_{{\rm N}\nu}\hspace{0.05cm}$,&nbsp; where the noise component &nbsp;$d_{{\rm N}\nu}$&nbsp; is assumed to be Gaussian distributed with standard deviation &nbsp;$\sigma_d$.&nbsp;
  
*Damit ergeben sich für die wertkontinuierlichen ''Detektionsabtastwerte'' $d_{\nu} = a_{\nu}\cdot g_{0} + a_{\nu+1}\cdot g_{-1}+d_{{\rm N}\nu}\hspace{0.05cm}$, wobei die Rauschkomponente $d_{{\rm N}\nu}$ als gaußverteilt mit Streuung $\sigma_d$ angenommen wird.
 
  
 +
<u>Note:</u> &nbsp; The algorithm is not more complex with bipolar signaling. &nbsp; $\bullet$ &nbsp; In contrast, the computational cost increases when the basic detection pulse becomes broader and has more than one precursor &nbsp;$g_{-1}$. &nbsp; $\bullet$ &nbsp; The neglect of precursors in the description is not a fundamental limitation because any pulse can satisfy this condition by appropriate choice of the detection time &nbsp;$T_{\rm D}$. &nbsp; $\bullet$ &nbsp; In the following all signal values are normalized to &nbsp;$1$.&nbsp; <br>
  
Bei bipolarer Signalisierung ist der Algorithmus nicht aufwändiger. Dagegen steigt der Rechenaufwand, wenn der Detektionsgrundimpuls breiter wird und mehr als nur einen Vorläufer $g_{-1}$ aufweist. Die Vernachlässigung von Nachläufern stellt keine grundlegende Einschränkung dar, weil jeder Impuls $g_d(t)$ diese Bedingung durch geeignete Wahl des Detektionszeitpunktes $T_{\rm D}$ erfüllen kann. Anzumerken ist weiter, dass im Folgenden alle Signalwerte auf $1$ normiert werden.<br>
 
 
[[File:P ID1468 Dig T 3 8 S1b version1.png|right|frame|Abtastwerte zur Verdeutlichung des Viterbi-Algorithmus|class=fit]]
 
 
{{GraueBox|TEXT=   
 
{{GraueBox|TEXT=   
$\text{Beispiel 1:}$&nbsp; In der Grafik sind  als (blaue) Kreuze die Detektionsnutzabtastwerte $d_{ {\rm S}\nu}$ eingetragen, wobei die zugehörigen Amplitudenkoeffizienten $a_1 = 1$, $a_2 = 1$, $a_3 = 0$, ... aus dem grün eingezeichneten Quellensignal $q(t)$ abgelesen werden können.  
+
$\text{Example 1:}$&nbsp; In the graph,&nbsp; the samples &nbsp;$d_{ {\rm S}\nu}$&nbsp; of the noiseless detection  signal&nbsp;  are entered as&nbsp; (blue)&nbsp; crosses,&nbsp; where the corresponding amplitude coefficients &nbsp;$a_1 = 1$, &nbsp;$a_2 = 1$, &nbsp;$a_3 = 0$, ... can be read from the source signal &nbsp;$q(t)$&nbsp; plotted in green.
 +
[[File:P ID1468 Dig T 3 8 S1b version1.png|right|frame|Sampling values to illustrate the Viterbi algorithm|class=fit]]
  
Die Grundimpulswerte sind in diesem Beispiel zu $g_0 = 0.7$ und $g_{-1} = 0.3$ angenommen. Es ist weiter zu erkennen, dass $d_{ {\rm S}\nu}$ nur vier verschiedene Werte annehmen kann, nämlich $0$, $g_{-1}=0.3$, $g_0= 0.7$ und $g_0 +g_{-1}= 1$.
+
* The basic pulse values are assumed to be &nbsp;$g_0 = 0.7$&nbsp; and &nbsp;$g_{-1} = 0.3$.&nbsp; It can be further seen that &nbsp;$d_{ {\rm S}\nu}$&nbsp; can take only four different values, namely &nbsp;$0$, &nbsp;$g_{-1}=0.3$, &nbsp;$g_0= 0.7$,&nbsp; $g_0 +g_{-1}= 1$.
  
Die am Viterbi&ndash;Entscheider anstehenden Abtastwerte (rote Punkte) sind $d_0 = 0.2$, $d_1 = 0.7$, $d_2 = 0.5$, $d_3 = 0$, ... , wobei die Differenzen $d_{ {\rm N}\nu} = d_\nu - d_{ {\rm S}\nu}$ von einer AWGN&ndash;Rauschquelle herrühren.<br>
+
* The noisy samples present at the Viterbi decision&nbsp; (red dots)&nbsp; are &nbsp;$d_0 = 0.2$, &nbsp;$d_1 = 0.7$, &nbsp;$d_2 = 0.5$, &nbsp;$d_3 = 0$, ... ,&nbsp; where the differences &nbsp;$d_{ {\rm N}\nu} = d_\nu - d_{ {\rm S}\nu}$&nbsp; originate from an AWGN noise source.<br>
  
Ein Schwellenwertentscheider (mit der Schwelle bei $E = 0.5$) würde bei diesen dargestellten zehn Bit mindestens eine Fehlentscheidung treffen (zur Zeit  $t  = 4T$), und eventuell eine weitere bei $t  = 2T$, falls der Abtastwert $d_2 = 0.5$ doch geringfügig kleiner ist als der Schwellenwert $E = 0.5$. Dagegen wird der Viterbi&ndash;Empfänger diese Folge der Länge $10$ richtig entscheiden, wie auf den nächsten Seiten gezeigt werden wird.}}<br>
+
*A threshold decision&nbsp; $($with the threshold at &nbsp;$E = 0.5)$&nbsp; would make at least one wrong decision at these ten bits shown&nbsp; $($at time &nbsp;$t  = 4T)$,&nbsp; and possibly another at &nbsp;$t  = 2T$&nbsp; if the sample &nbsp;$d_2 = 0.5$&nbsp; is nevertheless slightly smaller than the threshold &nbsp;$E = 0.5$.  
 +
*In contrast, the Viterbi receiver will correctly decide this sequence of length &nbsp;$10$,&nbsp; as shown in the next sections.}}<br>
  
== Fehlergrößen und Gesamtfehlergrößen==
+
== Metric and accumulated metric==
 
<br>
 
<br>
Wie im Kapitel  [[Digitalsignalübertragung/Optimale_Empfängerstrategien#Betrachtetes_Szenario_und_Voraussetzungen|Optimale Empfängerstrategien]] gibt $Q \in \{Q_i\}$ die begrenzte, aus $N$ Binärsymbolen bestehende Quellensymbolfolge an.  
+
As in the chapter&nbsp; [[Digital_Signal_Transmission/Optimal_Receiver_Strategies|"Optimal Receiver Strategies"]],&nbsp; &nbsp;$Q \in \{Q_i\}$&nbsp; specifies the time-limited source symbol sequence consisting of &nbsp;$N$&nbsp; binary symbols.
*Die Anzahl der möglichen Symbolfolgen $Q_i$ ist somit $2^N$.  
+
*Thus,&nbsp; the number of possible symbol sequences &nbsp;$Q_i$&nbsp; is &nbsp;$2^N$.
*$V$ bezeichnet  die Sinkensymbolfolge der Länge $N$, die vom Viterbi&ndash;Entscheider gleich der wahrscheinlichsten Folge $Q_j$ gesetzt wird.<br>
+
 +
*$V$&nbsp; denotes the sink symbol sequence of length&nbsp; $N$&nbsp; set equal to the most probable sequence &nbsp;$Q_j$&nbsp; by the Viterbi decision.<br>
  
  
 
{{BlaueBox|TEXT=   
 
{{BlaueBox|TEXT=   
$\text{Definitionen:}$&nbsp; Die '''Fehlergröße''' $\varepsilon_{\nu}(i)$ bezeichnet die quadratische Abweichung zwischen dem tatsächlichen, verrauschten Abtastwert $d_\nu$ und dem zur Folge $Q_i$ gehörenden Nutzabtastwert $d_{ {\rm S}\nu}$:
+
$\text{Definitions:}$&nbsp;  
 +
*The &nbsp;'''metric'''&nbsp; $\varepsilon_{\nu}(i)$&nbsp; denotes the squared deviation between the actual noisy sample &nbsp;$d_\nu$&nbsp; and the noiseless sample &nbsp;$d_{ {\rm S}\nu}$ belonging to the sequence &nbsp;$Q_i$:&nbsp;
 
:$$\varepsilon_{\nu}(i) = \vert d_{\nu} - d_{ {\rm S}\nu}(i) \vert^2
 
:$$\varepsilon_{\nu}(i) = \vert d_{\nu} - d_{ {\rm S}\nu}(i) \vert^2
 
  \hspace{0.05cm}.$$
 
  \hspace{0.05cm}.$$
 +
:In the literature, this is sometimes referred to as&nbsp; '''error quantity'''.
 +
*The &nbsp;'''accumulated metric'''&nbsp; $\gamma_{\nu}(i)$ characterizes the sum of all metrics up to time &nbsp;$\nu$:<br>
 +
:$$\gamma_{\nu}(i)= \sum_{k=0}^{\nu}\varepsilon_{k}(i)
 +
\hspace{0.05cm}.$$
 +
:In this context,&nbsp; one also speaks of the&nbsp; '''total error quantity'''.}}
  
Die '''Gesamtfehlergröße''' $\gamma_{\nu}(i)$ kennzeichnet die Summe aller Fehlergrößen bis zum Zeitpunkt $\nu$:<br>
 
:$$\gamma_{\nu}(i)= \sum_{k=0}^{\nu}\varepsilon_{k}(i)
 
\hspace{0.05cm}.$$}}<br>
 
  
[[File:P ID1469 Dig T 3 8 S2b version1.png|right|frame|Darstellung der Fehlergrößen im Baumdiagramm|class=fit]]
+
The graph illustrates the quantities defined above in a tree structure,&nbsp; from which it can be seen that the accumulated metric can be calculated iteratively:
Die Grafik verdeutlicht die oben definierten Größen in einer Baumstruktur, woraus zu erkennen ist, dass die Gesamtfehlergrößen iterativ berechnet werden können:
+
[[File:P ID1469 Dig T 3 8 S2b version1.png|right|frame|Representation of the metrics&nbsp; $\varepsilon_{\nu}(i)$&nbsp; and accumulated metrics&nbsp; $\gamma_{\nu}(i)= \gamma_{\nu-1}(i\hspace{0.05cm}')+\varepsilon_{\nu}(i\hspace{0.05cm}'')
:$$\gamma_{\nu}(i)= \gamma_{\nu-1}(i')+\varepsilon_{\nu}(i'')
+
\hspace{0.05cm}$&nbsp; in the tree diagram. &nbsp; <u>Notes:</u><br>&nbsp; $\bullet$ &nbsp; $i$, &nbsp;$i\hspace{0.05cm}'$, &nbsp;$i\hspace{0.05cm}''$ are different index variables.<br>&nbsp; $\bullet$ &nbsp; Definition is valid for a basic pulse with main value &nbsp;$g_{0}$&nbsp; and precursor &nbsp;$g_{-1}$.<br>&nbsp; $\bullet$ &nbsp;  With &nbsp;$v$&nbsp; precursors the above sum would have to start at &nbsp;$k = 1 -v$.<br>&nbsp; $\bullet$ &nbsp; The parameter &nbsp;$i \in  \{0, \hspace{0.05cm}\text{...} \hspace{0.05cm}, 2^{N+1}-1\}$&nbsp;  is usually represented in binary.<br>&nbsp; $\bullet$ &nbsp; It describes the amplitude coefficients &nbsp;$a_1$, ... , &nbsp;$a_{\nu +1}$ $($each &nbsp;$0$&nbsp; or &nbsp;$1)$.|class=fit]]
 +
:$$\gamma_{\nu}(i)= \gamma_{\nu-1}(i\hspace{0.05cm}')+\varepsilon_{\nu}(i\hspace{0.05cm}'')
 
  \hspace{0.05cm}.$$
 
  \hspace{0.05cm}.$$
  
''Anmerkungen:''
+
It should be noted about this graph:
*$i$, $i'$, $i''$ sind unterschiedliche Laufvariable.
+
*The nodes of the tree diagram represent the&nbsp; "accumulated metric"&nbsp; $\gamma_{\nu}(i)$.&nbsp; Their number is doubled with each iteration step.&nbsp; At time &nbsp;$\nu$,&nbsp; there are &nbsp;$2^{\nu+1}$&nbsp; such nodes.&nbsp; For example,&nbsp; for &nbsp;$\nu = 3$,&nbsp; exactly &nbsp;$2^4 = 16$&nbsp; nodes can be seen.<br>
*Man bezeichnet $\gamma_{\nu}(i)$ auch als <i>Metrik</i>.  
 
*Obige Definition gilt für einen Grundimpuls mit Hauptwert $g_{0}$ und einem Vorläufer $g_{-1}$.  
 
*Bei $v$ Vorläufern müsste die obige Summe bei $k = 1 -v$ beginnen.
 
*Der Parameter $i \in  \{0, \text{...} , 2^{N+1}-1\}$  wird meist binär dargestellt und beschreibt so die Amplitudenkoeffizienten $a_1$, ... , $a_{\nu +1}$ (jeweils entweder $0$ oder $1$).
 
<br clear = all>
 
Weiter ist zu dieser Grafik noch Folgendes anzumerken:
 
*Die Knoten des Baumdiagramms stehen für die <i>Gesamtfehlergrößen</i> $\gamma_{\nu}(i)$; deren Anzahl wird mit jedem Iterationsschritt verdoppelt. Zum Zeitpunkt $\nu$ gibt es $2^{\nu+1}$ solcher Knoten. Beispielsweise sind für $\nu = 3$ genau $2^4 = 16$ Knoten zu erkennen.<br>
 
  
*Die zu den Gesamtfehlergrößen $\gamma_{\nu}(i)$ gehörigen <i>Amplitudenkoeffizienten</i> ergeben sich, wenn man den Weg vom Anfangsknoten bis zum betrachteten Knoten verfolgt. Es wird vereinbart, dass einem nach oben gerichteten Zweig der Amplitudenkoeffizient $a_\nu=1$ und einem nach unten gerichteten Zweig der Koeffizient $a_\nu=0$ zugeordnet wird.<br>
+
*The&nbsp; "amplitude coefficients"&nbsp; associated with the accumulated metric &nbsp;$\gamma_{\nu}(i)$&nbsp; are obtained by following the path from the initial node to the node under consideration.&nbsp; It is agreed that an upward branch is assigned the coefficient &nbsp;$a_\nu=1$&nbsp; and a downward branch is assigned the coefficient &nbsp;$a_\nu=0$. <br>
  
*Beispielsweise kennzeichnet der grün hinterlegte Knoten $\gamma_{3}(\rm 1100)$ die Gesamtfehlergröße unter der hypothetischen Annahme, dass die Symbole $a_1=1$, $a_2=1$, $a_3=0$, $a_4=0$ gesendet wurden. Diese Zuordnung kann auch aus den Richtungen der Pfeile im Baumdiagramm abgelesen werden: Zunächst zweimal nach oben, dann zweimal nach unten.<br>
+
*The green highlighted node &nbsp;$\gamma_{3}(\rm 1100)$&nbsp; denotes the accumulated metric under the hypothetical assumption that the symbols &nbsp;$a_1=1$, &nbsp;$a_2=1$, &nbsp;$a_3=0$, &nbsp;$a_4=0$ were sent.&nbsp; This assignment can also be read from the directions of the arrows in the tree diagram: &nbsp; First twice upwards,&nbsp; then twice downwards.<br>
  
*Aufgrund des Vorläufers muss bereits zum Zeitpunkt $\nu = 3$ der Koeffizient $a_4$ mitberücksichtigt werden. Alle Knoten $\gamma_{\nu}(i)$, die unter der Voraussetzung $a_{\nu +1}=1$ berechnet werden, sind im Baumdiagramm durch Rechtecke dargestellt, während die Hypothese $a_{\nu +1}=0$ jeweils durch ein abgerundetes Rechteck symbolisiert ist, zum Beispiel $\gamma_{2}(\rm 110)$ oder $\gamma_{3}(\rm 1100)$.<br>
+
*Due to the precursor,&nbsp; the coefficient &nbsp;$a_4$&nbsp; must already be taken into account at time &nbsp;$\nu = 3$.&nbsp; All nodes &nbsp;$\gamma_{\nu}(i)$ computed under the assumption &nbsp;$a_{\nu +1}=1$&nbsp; are represented by rectangles,&nbsp; while the hypothesis &nbsp;$a_{\nu +1}=0$&nbsp; is symbolized by a rounded rectangle in each case,&nbsp; for example &nbsp;$\gamma_{2}(\rm 110)$&nbsp; or &nbsp;$\gamma_{3}(\rm 1100)$.<br>
  
*Die Zweige im Baumdiagramm sind den <i>Fehlergrößen</i> $\varepsilon_{\nu}(i)$ zugeordnet. Beim vorausgesetzten Grundimpuls (nur $g_{0}$ und $g_{-1}$ sind ungleich Null) gibt es zu jedem Zeitpunkt mit Ausnahme des Startzustandes $(\nu> = 0)$ genau vier unterschiedliche Größen:
+
*The branches in the tree diagram are assigned to the&nbsp; "metrics"&nbsp; $\varepsilon_{\nu}(i)$.&nbsp; With the assumed basic pulse&nbsp; $($only &nbsp;$g_{0}$&nbsp; and &nbsp;$g_{-1})$,&nbsp; there are exactly four different metrics at each time point except for the initial state &nbsp;$(\nu = 0)$:&nbsp;
:$$\varepsilon_{\nu}(00) = |d_{\nu}|^2\hspace{0.05cm},\hspace{0.5cm}\varepsilon_{\nu}(01) = |d_{\nu}-g_{-1}|^2\hspace{0.05cm},\hspace{0.2cm}
+
:$$\varepsilon_{\nu}(00) = |d_{\nu}|^2\hspace{0.05cm},\hspace{0.5cm}\varepsilon_{\nu}(01) = |d_{\nu}-g_{-1}|^2\hspace{0.05cm},$$
 +
:$$
 
\varepsilon_{\nu}(10) = |d_{\nu}-g_{0}|^2\hspace{0.05cm},\hspace{0.2cm}\varepsilon_{\nu}(11) = |d_{\nu}-g_{0}-g_{-1}|^2\hspace{0.05cm}.$$
 
\varepsilon_{\nu}(10) = |d_{\nu}-g_{0}|^2\hspace{0.05cm},\hspace{0.2cm}\varepsilon_{\nu}(11) = |d_{\nu}-g_{0}-g_{-1}|^2\hspace{0.05cm}.$$
  
*Die Gesamtfehlergröße $\gamma_{\nu}(i)$ ist gleich der Summe aus dem vorausgegangenen Knoten $\gamma_{\nu-1}(i')$ und dem dazwischenliegenden Zweig $\varepsilon_{\nu}(i'')$. Beispielsweise gilt für die hervorgehobenen Knoten:
+
*The accumulated metric &nbsp;$\gamma_{\nu}(i)$&nbsp; is the sum of the preceding node &nbsp;$\gamma_{\nu-1}(i\hspace{0.05cm}')$&nbsp; and the intervening branch &nbsp;$\varepsilon_{\nu}(i\hspace{0.05cm}'')$.&nbsp; For example,&nbsp; for the highlighted nodes:
:$$\gamma_{1}(11)=\gamma_{0}(1)+\varepsilon_{1}(11) ,\hspace{0.05cm}
+
:$$\gamma_{1}(11)=\gamma_{0}(1)+\varepsilon_{1}(11) ,$$
\gamma_{2}(110)=\gamma_{1}(11)+\varepsilon_{2}(10) ,\hspace{0.05cm}
+
:$$ \gamma_{2}(110)=\gamma_{1}(11)+\varepsilon_{2}(10) ,$$
\gamma_{3}(1100)=\gamma_{2}(110)+\varepsilon_{3}(00).$$
+
:$$ \gamma_{3}(1100)=\gamma_{2}(110)+\varepsilon_{3}(00).$$
  
*Bei den ersten Knoten $\gamma_{0}(0)$ und $\gamma_{0}(1)$ wird berücksichtigt, dass vor der eigentlichen Übertragung ($a_1$, $a_2$, ...) vereinbarungsgemäß stets das Symbol $a_0 = 0$ übertragen wird. Daraus folgt:
+
*For the first nodes &nbsp;$\gamma_{0}(0)$&nbsp; and &nbsp;$\gamma_{0}(1)$,&nbsp; it is considered that, by convention, the symbol &nbsp;$a_0=0$&nbsp; is always transmitted before the actual transmission &nbsp;&nbsp;$(a_1$, &nbsp;$a_2$, ...$)$.&nbsp; It follows that:
 
:$$\gamma_{0}(0)=\varepsilon_{0}(00)= |d_{0}|^2 \hspace{0.05cm},\hspace{0.2cm}
 
:$$\gamma_{0}(0)=\varepsilon_{0}(00)= |d_{0}|^2 \hspace{0.05cm},\hspace{0.2cm}
 
  \gamma_{0}(1)=\varepsilon_{0}(01)=|d_{0}-g_{-1}|^2
 
  \gamma_{0}(1)=\varepsilon_{0}(01)=|d_{0}-g_{-1}|^2
 
  \hspace{0.05cm}.$$
 
  \hspace{0.05cm}.$$
  
Die beiden folgenden Beispiele werden hoffentlich diese etwas ermüdenden Aussagen verdeutlichen.<br>
+
The following two examples will hopefully clarify these somewhat tedious statements.<br>
  
== Fehlergrößen und Gesamtfehlergrößen (3) ==
+
{{GraueBox|TEXT=
<br>
+
$\text{Example 2:}$&nbsp; As in &nbsp;[[Digital_Signal_Transmission/Viterbi_Receiver#Considered_scenario_and_prerequisites|$\text{Example 1}$]],&nbsp; we consider the unipolar source symbol sequence of length &nbsp;$N=3$&nbsp; with the following parameter values:
{{Beispiel}}''':''' Nachzutragen ist die Berechnung der Fehlergrößen für die Zeitpunkte <i>&nu;</i> = 0 bis <i>&nu;</i> = 3. Die unipolare Quellensymbolfolge hat die Länge <i>N</i> = 3. Die Amplitudenkoeffizienten seien <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1, <i>a</i><sub>3</sub> = 0. Weiterhin gilt:
+
:$$g_{0}=0.7 \hspace{0.05cm},\hspace{0.2cm}
 
 
:<math>g_{0}=0.7 \hspace{0.05cm},\hspace{0.2cm}
 
 
  g_{-1}=0.3 \hspace{0.05cm},\hspace{0.2cm}
 
  g_{-1}=0.3 \hspace{0.05cm},\hspace{0.2cm}
 
  d_{0}=0.2 \hspace{0.05cm},\hspace{0.2cm}
 
  d_{0}=0.2 \hspace{0.05cm},\hspace{0.2cm}
Line 107: Line 104:
 
  d_{2}=0.5 \hspace{0.05cm},\hspace{0.2cm}
 
  d_{2}=0.5 \hspace{0.05cm},\hspace{0.2cm}
 
  d_{3}=0 \hspace{0.05cm}
 
  d_{3}=0 \hspace{0.05cm}
  \hspace{0.05cm}.</math>
+
  \hspace{0.05cm}.$$
  
Das Baumdiagramm mit den Knoten <i>&gamma;<sub>&nu;</sub></i>(<i>i</i>) ist unten dargestellt. Für die Zeitpunkte <i>&nu;</i> = 0 bis <i>&nu;</i> = 3 gilt:
+
Then,&nbsp; for the metrics &nbsp;$\varepsilon_{\nu}(i)$&nbsp; at time points &nbsp;$\nu = 0$&nbsp; to &nbsp;$\nu = 3$:  
 +
:$$\nu = 0\text{:} \hspace{0.2cm}\varepsilon_{0}(00)  =  \big [0.2- (0 \cdot 0.7 + 0 \cdot
 +
0.3) \big ]^2=0.04 \hspace{0.05cm},\hspace{0.4cm} \varepsilon_{0}(01)  =  \big [0.2- (0 \cdot 0.7 + 1 \cdot
 +
0.3) \big ]^2=0.01 \hspace{0.05cm};$$
  
:<math>\nu = 0 : \hspace{0.2cm}\varepsilon_{0}(00)  =  [0.2- (0 \cdot 0.7 + 0 \cdot
+
:$$\nu = 1{:} \hspace{0.2cm}\varepsilon_{1}(00)  =  \big [0.7- (0 \cdot 0.7 + 0 \cdot
  0.3)]^2=0.04 \hspace{0.05cm},</math>
+
  0.3)\big ]^2=0.49 \hspace{0.05cm},\hspace{0.4cm} \varepsilon_{1}(01)  = \big [0.7- (0 \cdot 0.7 + 1 \cdot
:::<math>\hspace{0.4cm} \varepsilon_{0}(01)  =  [0.2- (0 \cdot 0.7 + 1 \cdot
+
0.3)\big ]^2=0.16 \hspace{0.05cm},$$
  0.3)]^2=0.01 \hspace{0.05cm},</math>
+
:$$\hspace{1.4cm}\varepsilon_{1}(10) =  \big [0.7- (1 \cdot 0.7 + 0 \cdot
 +
0.3)\big ]^2=0.00 \hspace{0.05cm}\hspace{0.4cm}\varepsilon_{1}(11)  =  \big [0.7- (1 \cdot 0.7 + 1 \cdot
 +
  0.3)\big ]^2=0.09 \hspace{0.05cm};$$
  
:<math>\nu = 1 : \hspace{0.2cm}\varepsilon_{1}(00)  =  [0.7- (0 \cdot 0.7 + 0 \cdot
+
:$$\nu = 2\text{:} \hspace{0.2cm}\varepsilon_{2}(00)  = \big [0.5- (0 \cdot 0.7 + 0 \cdot
  0.3)]^2=0.49 \hspace{0.05cm},</math>
+
  0.3)\big ]^2=0.25 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{2}(01)  =  \big [0.5- (0 \cdot 0.7 + 1 \cdot
:::<math>\hspace{0.4cm} \varepsilon_{1}(01)  =  [0.7- (0 \cdot 0.7 + 1 \cdot
+
  0.3)\big ]^2=0.04 \hspace{0.05cm},$$
  0.3)]^2=0.16 \hspace{0.05cm},</math>
+
:$$\hspace{1.4cm}\varepsilon_{2}(10) = \big [0.5- (1 \cdot 0.7 + 0 \cdot
:::<math>\hspace{0.4cm}\varepsilon_{1}(10) =  [0.7- (1 \cdot 0.7 + 0 \cdot
+
  0.3)\big ]^2=0.04 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{2}(11)  =  \big [0.5- (1 \cdot 0.7 + 1 \cdot
  0.3)]^2=0.00 \hspace{0.05cm}</math>
+
  0.3)\big ]^2=0.25 \hspace{0.05cm};$$
:::<math>\hspace{0.4cm}\varepsilon_{1}(11)  =  [0.7- (1 \cdot 0.7 + 1 \cdot
 
  0.3)]^2=0.09 \hspace{0.05cm},</math>
 
  
:<math>\nu = 2 : \hspace{0.2cm}\varepsilon_{2}(00)  =  [0.5- (0 \cdot 0.7 + 0 \cdot
+
:$$ \nu = 3\text{:} \hspace{0.2cm}\varepsilon_{3}(00)  = \big [0.0- (0 \cdot 0.7 + 0 \cdot
  0.3)]^2=0.25 \hspace{0.05cm},</math>
+
  0.3)\big ]^2=0.00 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{3}(01)  = \big [0.0- (0 \cdot 0.7 + 1 \cdot
:::<math>\hspace{0.4cm}\varepsilon_{2}(01)  =  [0.5- (0 \cdot 0.7 + 1 \cdot
+
  0.3)\big ]^2=0.09 \hspace{0.05cm},$$
  0.3)]^2=0.04 \hspace{0.05cm},</math>
+
:$$\hspace{1.4cm}\varepsilon_{3}(10) =  \big [0.0- (1 \cdot 0.7 + 0 \cdot
:::<math>\hspace{0.4cm}\varepsilon_{2}(10) =  [0.5- (1 \cdot 0.7 + 0 \cdot
+
  0.3)\big ]^2=0.49 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{3}(11)  =  \big [0.0- (1 \cdot 0.7 + 1 \cdot
  0.3)]^2=0.04 \hspace{0.05cm},</math>
+
  0.3)\big ]^2=1.00 \hspace{0.05cm}.$$}}
:::<math>\hspace{0.4cm}\varepsilon_{2}(11)  =  [0.5- (1 \cdot 0.7 + 1 \cdot
 
  0.3)]^2=0.25 \hspace{0.05cm},</math>
 
  
:<math> \nu = 3 : \hspace{0.2cm}\varepsilon_{3}(00)  =  [0.0- (0 \cdot 0.7 + 0 \cdot
 
0.3)]^2=0.00 \hspace{0.05cm},</math>
 
:::<math>\hspace{0.4cm}\varepsilon_{3}(01)  =  [0.0- (0 \cdot 0.7 + 1 \cdot
 
0.3)]^2=0.09 \hspace{0.05cm},</math>
 
:::<math>\hspace{0.4cm}\varepsilon_{3}(10) =  [0.0- (1 \cdot 0.7 + 0 \cdot
 
0.3)]^2=0.49 \hspace{0.05cm},</math>
 
:::<math>\hspace{0.4cm}\varepsilon_{3}(11)  =  [0.0- (1 \cdot 0.7 + 1 \cdot
 
0.3)]^2=1.00 \hspace{0.05cm}.</math>
 
  
Wir betrachten wie im [http://en.lntwww.de/Digitalsignal%C3%BCbertragung/Viterbi%E2%80%93Empf%C3%A4nger#Blockschaltbild_und_Voraussetzungen_f.C3.BCr_Kapitel_3.8_.282.29 letzten Beispiel] die unipolare Quellensymbolfolge der Länge <i>N</i> = 3, mit den Amplitudenkoeffizienten <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1, <i>a</i><sub>3</sub> = 0 und die weiteren Parameterwerte
+
{{GraueBox|TEXT=
 
+
$\text{Example 3:}$&nbsp; With the metrics &nbsp;$\varepsilon_{\nu}(i)$&nbsp; determined in &nbsp;$\text{Example 2}$,&nbsp; the accumulated metrics &nbsp;$\gamma_{\nu}(i)$&nbsp; can now also be calculated. Below is the tree diagram with the accumulated metrics &nbsp;$\gamma_{\nu}(i)$&nbsp; as nodes for the time points &nbsp;$\nu = 0$&nbsp; to &nbsp;$\nu = 3$.&nbsp;
:<math>g_{0}=0.7 \hspace{0.05cm},\hspace{0.2cm}
 
g_{-1}=0.3 \hspace{0.05cm},\hspace{0.2cm}
 
d_{0}=0.2 \hspace{0.05cm},\hspace{0.2cm}
 
d_{1}=0.7 \hspace{0.05cm},\hspace{0.2cm}
 
d_{2}=0.5 \hspace{0.05cm},\hspace{0.2cm}
 
d_{3}=0 \hspace{0.05cm}
 
\hspace{0.05cm}.</math>
 
  
Nachfolgend ist das Baumdiagramm mit den Knoten <i>&gamma;<sub>&nu;</sub></i>(<i>i</i>) für die Zeitpunkte <i>&nu;</i> = 0 bis <i>&nu;</i> = 3 dargestellt. Die Berechnung der sich ergebenden Fehlergrößen <i>&epsilon;<sub>&nu;</sub></i>(<i>i</i>) folgt auf der nächsten Seite.<br>
+
[[File:P ID1470 Dig T 3 8 S2c version1.png|frame|right|Iterative calculation of the accumulated metrics (example)|class=fit]]
 +
<BR>
 +
*The minimum accumulated metric at time &nbsp;$\nu = 3$&nbsp; is &nbsp;$\gamma_{3}(\rm 1100) = 0.14$.&nbsp; From this,&nbsp; the coefficients of the most likely transmitted&nbsp; (unipolar)&nbsp; sequence according to the present signal values &nbsp;$d_0 = 0.2$, &nbsp;$d_1 = 0.7$, &nbsp;$d_2 = 0.5$&nbsp; and &nbsp;$d_3 = 0$&nbsp; are &nbsp;$a_1 = 1$, &nbsp;$a_2 = 1$&nbsp; and &nbsp;$a_3 = 0$&nbsp; (green path).  
  
[[File:P ID1470 Dig T 3 8 S2c version1.png|center|frame|Iterative Berechnung der Gesamtfehlergrößen (Beispiel)|class=fit]]<br>
+
*If the sequence length is &nbsp;$N = 3$&nbsp; $($that is: &nbsp; only three symbols are jointly decided by the Viterbi receiver$)$,&nbsp; the decision &nbsp;$a_4 = 0$&nbsp; is also certainly the correct one,&nbsp; since all coefficients &nbsp;$a_{\nu>3}$&nbsp; were assumed to be zero.<br>
  
Die minimale Gesamtfehlergröße <i>&gamma;</i><sub>3</sub>(1100) ist gleich 0.14. Daraus ergeben sich die Koeffizienten der nach den vorliegenden Signalwerten <i>d</i><sub>0</sub>, <i>d</i><sub>1</sub>, <i>d</i><sub>2</sub>, <i>d</i><sub>3</sub> mit größter Wahrscheinlichkeit gesendeten Folge zu <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1 und <i>a</i><sub>3</sub> = 0 (grün eingezeichneter Pfad). Weiter ist zu sagen:
+
*However,&nbsp; for longer sequences &nbsp;$(N > 3)$,&nbsp; it cannot necessarily be concluded from the minimum value &nbsp;$\gamma_{3}(\rm 1100)$&nbsp; that &nbsp;$a_1 = 1$, &nbsp;$a_2 = 1$,&nbsp; $a_3 = 0$&nbsp; is actually part of the most probable sequence.  
*Ist die Folgenlänge <i>N</i> = 3 (das heißt: nur drei Symbole werden durch den Viterbi&ndash;Empfänger gemeinsam entschieden), so ist auch die Entscheidung <i>a</i><sub>4</sub> = 0 mit Sicherheit die richtige, da alle Koeffizienten <i>a<sub>&nu;</sub></i><sub>>3</sub> als 0 vorausgesetzt wurden.<br>
 
  
*Bei längerer Folge (<i>N</i> > 3) kann aus dem minimalen Wert <i>&gamma;</i><sub>3</sub>(1100) nicht unbedingt geschlossen werden, dass <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1, <i>a</i><sub>3</sub> = 0 Teil der wahrscheinlichsten Folge ist. Bei Berücksichtigung weiterer Abtastwerte (<i>d</i><sub>4</sub>, <i>d</i><sub>5</sub>, ...) könnte sich dieses vorläufige Ergebnis durchaus noch ändern.<br><br>
+
*Considering further sample values &nbsp;$(d_4$, &nbsp;$d_5$, ...$)$&nbsp; this preliminary result could well change.}}<br><br>
  
Die Berechnung der Fehlergrößen für <i>&nu;</i> = 0 bis <i>&nu;</i> = 3 wird auf der nächsten Zeile gezeigt.<br>
 
  
 +
== Minimum accumulated metric==
 +
<br>
 +
We continue from the numerical values of the last examples:
 +
:$$d_{0}=0.2 ,\hspace{0.3cm}
 +
d_{1}=0.7 ,\hspace{0.3cm}
 +
d_{2}=0.5 ,\hspace{0.3cm}
 +
d_{3}=0 ,\hspace{0.3cm}
 +
g_{0}=0.7 ,\hspace{0.3cm}
 +
g_{-1}=0.3 .$$
  
 +
Thus,&nbsp; nothing changes in the tree diagram compared to &nbsp;$\text{Example 3}$.&nbsp; Some colored markings in this diagram only indicate in which way the tree diagram can be simplified according to Viterbi's suggestions.&nbsp; Important properties of the Viterbi decision can be seen in the graph,&nbsp; e.g. at time &nbsp;$\nu = 2$:&nbsp;
 +
[[File:P ID1471 Dig T 3 8 S3 version1.png|right|frame|Simplification of the tree diagram according to Viterbi|class=fit]]
 +
<br>
 +
*At time &nbsp;$\nu = 2$,&nbsp; the minimum accumulated metric is &nbsp;$\gamma_{2}(\rm 101) = 0.05$&nbsp; (highlighted in brown).&nbsp;
  
[[File:P ID1470 Dig T 3 8 S2c version1 (1).png|Iterative Berechnung der Gesamtfehlergrößen (Beispiel)|class=fit]]<br>
+
*This means: &nbsp; A decision at &nbsp;$\nu = 2$&nbsp; &ndash;&nbsp; based on &nbsp;$d_0$, &nbsp;$d_1$&nbsp; and&nbsp; $d_2$&nbsp; &ndash;&nbsp; would have been in favor of sequence &nbsp;"$\rm 101$"&nbsp; instead of the transmitted sequence &nbsp;"$\rm 110$".&nbsp; <br>
{{end}}<br>
 
 
 
== Minimale Gesamtfehlergröße und Trellisdiagramm (1) ==
 
<br>
 
Wir gehen weiterhin von den [http://en.lntwww.de/index.php?title=Digitalsignal%C3%BCbertragung/Viterbi%E2%80%93Empf%C3%A4nger#Fehlergr.C3.B6.C3.9Fen_und_Gesamtfehlergr.C3.B6.C3.9Fen_.283.29 Zahlenwerten des letzten Beispiels] aus.<br>
 
  
[[File:P ID1471 Dig T 3 8 S3 version1.png|center|frame|Vereinfachung des Baumdiagramms nach Viterbi|class=fit]]<br>
+
*It follows: &nbsp; A too early final determination should be avoided at all costs.&nbsp; However,&nbsp; at any time &nbsp;$\nu$&nbsp; one can already exclude several partial symbol sequences which need not be considered at later times.<br>
  
Wichtige Eigenschaften des optimalen Entscheiders (Maximum&ndash;Likelihood) lassen sich aus obiger Grafik anhand des Zeitpunktes <i>&nu;</i> = 2 erkennen:
+
*To &nbsp;$\gamma_{2}(\rm 111) = 0.35$&nbsp; the same metrics &nbsp;$\varepsilon_{3}(\rm 11)$&nbsp; and &nbsp;$\varepsilon_{3}(\rm 10)$&nbsp; are added as to &nbsp;$\gamma_{2}(\rm 101) = 0.05$.&nbsp; Because of&nbsp; $\gamma_{2}(\rm 111) > \gamma_{2}(\rm 101)$&nbsp; it is therefore already certain at &nbsp;$\nu = 2$&nbsp; that &nbsp;"$\rm 111$"&nbsp; cannot be part of the most probable sequence.  
*Die minimale Gesamtfehlergröße ist <i>&gamma;</i><sub>2</sub>(101) = 0.05 (braun hervorgehoben). Das bedeutet: Eine Entscheidung zu diesem Zeitpunkt &ndash; basierend auf den Werten <i>d</i><sub>0</sub>, <i>d</i><sub>1</sub> und <i>d</i><sub>2</sub> &ndash; wäre zugunsten der Folge &bdquo;101&rdquo; anstelle der gesendeten Folge &bdquo;110&rdquo; ausgegangen.<br>
 
  
*Daraus folgt: Für eine optimale Entscheidung sollte eine zu frühe endgültige Festlegung unbedingt vermieden werden. Allerdings kann man zu jedem Zeitpunkt <i>&nu;</i> bereits mehrere Teilsymbolfolgen ausschließen, die zu späteren Zeitpunkten nicht mehr berücksichtigt werden müssen.<br>
+
*The same is true for &nbsp;"$\rm 001$"&nbsp; and &nbsp;"$\rm 011$".&nbsp; Therefore,&nbsp; these nodes and all their successors need not be considered further&nbsp; $($brown coverages$)$.<br>
  
*Zu <i>&gamma;</i><sub>2</sub>(001), <i>&gamma;</i><sub>2</sub>(011) und <i>&gamma;</i><sub>2</sub>(111) werden jeweils genau die gleichen Fehlergrößen hinzuaddiert. Da diese drei Metriken alle größer sind als <i>&gamma;</i><sub>2</sub>(101) = 0.05, steht bereits bei <i>&nu;</i> = 2 fest, dass &bdquo;001&rdquo;, &bdquo;011&rdquo; sowie &bdquo;111&rdquo; nicht Bestandteil der wahrscheinlichsten Folge sein können. Diese Knoten und alle ihre Nachfahren müssen deshalb nicht weiter beachtet werden (braune Überdeckungen).<br>
+
*Also, the rounded nodes &nbsp;$\gamma_{2}(\rm 000) = 0.78$, &nbsp;$\gamma_{2}(\rm 010) = 0.24$,&nbsp; $\gamma_{2}(\rm 100) = 0.26$&nbsp; are not part of the most probable sequence,&nbsp; since they are larger than &nbsp;$\gamma_{2}(\rm 110) = 0.14$,&nbsp; which is marked in green.&nbsp; Also,&nbsp; these and their successors need not be considered from time &nbsp;$\nu = 3$&nbsp; (green covering).<br>
 +
<br clear=all>
  
*Gleiches gilt für die abgerundeten Knoten: <i>&gamma;</i><sub>2</sub>(000), <i>&gamma;</i><sub>2</sub>(010) und <i>&gamma;</i><sub>2</sub>(100) sind größer als der grün markierte Knoten <i>&gamma;</i><sub>2</sub>(110) = 0.14, so dass auch diese und ihre Nachfahren ab dem Zeitpunkt <i>&nu;</i> = 3 nicht mehr berücksichtigt werden müssen (grüne Überdeckungen).<br>
+
{{BlaueBox|TEXT= 
 +
$\text{Conclusion:}$&nbsp;
 +
#Of the eight nodes at &nbsp;$\nu = 2$,&nbsp; only two need to be followed up,&nbsp; namely the rectangle &nbsp;$\gamma_{2}(\rm 101) = 0.05$&nbsp; and the rounded rectangle &nbsp;$\gamma_{2}(\rm 110) = 0.14$.
 +
#These describe the minimum accumulated metrics assuming that &nbsp;$a_3 = 0$&nbsp; and &nbsp;$a_3 = 1$,&nbsp; respectively.}}
  
*Man stellt fest: Von den acht Knoten bei <i>&nu;</i> = 2 müssen nur zwei weiterverfolgt werden, nämlich das abgerundete Rechteck <i>&gamma;</i><sub>2</sub>(110) = 0.14 und das Rechteck <i>&gamma;</i><sub>2</sub>(101) = 0.05. Diese beschreiben die minimalen Gesamtfehlergrößen unter der Annahme, dass <i>a</i><sub>3</sub> = 0 bzw. <i>a</i><sub>3</sub> = 1 sein wird.<br><br>
 
  
Die Beschreibung wird auf der nächsten Seite fortgesetzt.<br>
 
  
== Minimale Gesamtfehlergröße und Trellisdiagramm (2) ==
+
== Representation with the trellis diagram ==
 
<br>
 
<br>
Verallgemeinern wir nun das Ergebnis dieses Beispiels. Unter der weiterhin gültigen Annahme, dass der Grundimpuls neben dem Hauptwert (<i>g</i><sub>0</sub>) nur einen Vorläufer (<i>g</i><sub>&ndash;1</sub>) aufweist, ergeben sich die beiden minimalen Gesamtfehlergrößen zum Zeitpunkt <i>&nu;</i> formal zu
+
We now generalize the result of this example.&nbsp; Under the still valid assumption that the basic pulse&nbsp; $g_d(t)$&nbsp; has only one precursor &nbsp;$(g_{-1})$&nbsp; besides the main value &nbsp;$(g_0)$,&nbsp; the two &nbsp;'''minimum accumulated metrics'''&nbsp; at time &nbsp;$\nu$&nbsp; are formally given by
  
:<math>{\it \Gamma}_{\nu}(0)  =  {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(00),
+
:$${\it \Gamma}_{\nu}(0)  =  {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(00),
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(10)\right] \hspace{0.05cm},</math>
+
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(10)\right] \hspace{0.05cm},$$
:<math> {\it \Gamma}_{\nu}(1)  =  {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(01),
+
:$${\it \Gamma}_{\nu}(1)  =  {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(01),
 
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(11)\right]
 
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(11)\right]
  \hspace{0.05cm}.</math>
+
  \hspace{0.05cm}.$$
 
 
Das Verfahren der <i>Gesamtfehlerminimierung</i> lässt sich im Trellisdiagramm anschaulich darstellen.<br>
 
  
[[File:P ID1472 Dig T 3 8 S3b version1.png|Beispielhaftes Trellisdiagramm|class=fit]]<br>
+
The procedure&nbsp; "minimization of the accumulated metrics"&nbsp;  can be clearly illustrated in the &nbsp;'''trellis diagram'''.&nbsp; <br>
 +
[[File:P ID1472 Dig T 3 8 S3b version1.png|right|frame|Example trellis diagram&nbsp; $($the numerical values are adapted to the previous examples$)$|class=fit]]
 +
A comparison with the figures in the last sections shows:
 +
*The lower branch represents the minimum accumulated metric &nbsp;${\it \Gamma}_{\nu}(0)$&nbsp; calculated at each time &nbsp; $\nu$&nbsp; under the hypothesis &nbsp;$a_{\nu + 1} = 0$&nbsp;  $($blue rounded squares$)$.<br>
  
Ein Vergleich mit den Bildern auf den letzten Seiten zeigt:
+
*The upper branch describes the minimum accumulated metric &nbsp;${\it \Gamma}_{\nu}(1)$&nbsp; under the assumption&nbsp; $a_{\nu + 1} = 1$&nbsp; $($red squares$)$.&nbsp; <br>
*Der untere Zweig stellt die minimale Gesamtfehlergröße <i>&Gamma;<sub>&nu;</sub></i>(0) dar, die zu jedem Zeitpunkt <i>&nu;</i> unter der Hypothese berechnet wird, dass <i>a<sub>&nu;</sub></i><sub>+1</sub> = 0 gelten wird (blaue abgerundete Quadrate).<br>
 
  
*Dagegen beschreibt der obere Zweig die minimalen Gesamtfehlergrößen <i>&Gamma;<sub>&nu;</sub></i>(1) unter der Annahme <i>a</i><sub><i>&nu;</i>+1</sub> = 1  (rote Quadrate). Auch hier sind die Zahlenwerte an das bisherige Beispiel angepasst.<br>
+
*Besides &nbsp;${\it \Gamma}_{\nu}(0)$&nbsp; and &nbsp;${\it \Gamma}_{\nu}(1)$,&nbsp; the maximum likelihood decision must also store the associated symbol sequences&nbsp; ("paths").&nbsp; These branches are highlighted in red and blue, resp.<br>
  
*Außer <i>&Gamma;<sub>&nu;</sub></i>(0) und <i>&Gamma;<sub>&nu;</sub></i>(1) muss der ML&ndash;Entscheider auch die dazugehörigen Symbolfolgen (Pfade) abspeichern. Diese Zweige sind in der Grafik rot bzw. blau hervorgehoben.<br>
+
*If &nbsp;${\it \Gamma}_{\nu}$&nbsp; arises from the node &nbsp;${\it \Gamma}_{\nu-1}(0)$&nbsp; &ndash;&nbsp; that is,&nbsp; if the lower of the two incoming branches is highlighted &ndash;&nbsp; the associated symbol is &nbsp;"$\rm 0$",&nbsp; otherwise the symbol is &nbsp;"$\rm 1$".<br>
  
*Falls <i>&Gamma;<sub>&nu;</sub></i> aus dem Knoten <i>&Gamma;<sub>&nu;</sub></i><sub>&ndash;1</sub>(0) hervorgeht &ndash; also wenn der untere der beiden ankommenden Zweige hervorgehoben ist &ndash; so ist das dazugehörige Symbol &bdquo;0&rdquo;, andernfalls die &bdquo;1&rdquo;.<br>
+
*For &nbsp; $\nu = 3$:&nbsp; Both, &nbsp;${\it \Gamma}_{3}(0) = 0.14$&nbsp; and &nbsp;${\it \Gamma}_{3}(1) = 0.23$,&nbsp; result from the precursor &nbsp;${\it \Gamma}_{2}(0)$,&nbsp; so that both selected paths point to the symbol &nbsp;"$\rm 0$"&nbsp; (blue branches).<br>
  
*Zur Zeit <i>&nu;</i> = 3 ergeben sich z. B. sowohl <i>&Gamma;</i><sub>3</sub>(0) = 0.14 als auch <i>&Gamma;</i><sub>3</sub>(1) = 0.23 aus dem Vorgänger <i>&Gamma;</i><sub>2</sub>(0), so dass beide ausgewählte Pfade jeweils auf das Symbol &bdquo;0&rdquo; verweisen (blaue Zweige).<br>
+
== Simplified trellis diagram ==
 
 
== Vereinfachtes Trellisdiagramm ==
 
 
<br>
 
<br>
Der Vorteil des Trellisdiagramms besteht darin, dass sich die Anzahl der Knoten und Zweige nicht bei jedem Iterationsschritt verdoppeln. Durch die Auswahl der minimalen Gesamtfehlergrößen werden nur noch diejenigen Symbolfolgen weiter betrachtet, die als Teil der wahrscheinlichsten Folge überhaupt noch in Frage kommen.<br>
+
The advantage of the trellis diagram is that the number of nodes and branches does not double at each iteration step.&nbsp; By selecting the minimum accumulated metrics,&nbsp; only those symbol sequences are considered further which could be part of the most probable sequence.<br>
  
Das Trellisdiagramm lässt sich weiter vereinfachen, indem man nur noch die ausgewählten Zweige einzeichnet. Dies ist im unteren Teil der Grafik an unserem Zahlenbeispiel verdeutlicht. Zur Erinnerung: Die tatsächlich gesendeten Amplitudenkoeffizienten seien <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1 und <i>a</i><sub>3</sub> = 0, und das oben gezeichnete Trellisdiagramm wurden unter der Annahme berechnet, dass aufgrund der Impulswerte <i>g</i><sub>&ndash;1</sub> = 0.3 und  <i>g</i><sub>0</sub> = 0.7 und des  AWGN&ndash;Rauschens die Eingangswerte <i>d</i><sub>0</sub> = 0.2, <i>d</i><sub>1</sub> = 0.7, <i>d</i><sub>2</sub> = 0.5 und <i>d</i><sub>3</sub> = 0 am ML&ndash;Entscheider anliegen.<br>
+
The trellis diagram can be further simplified by drawing in only the selected branches.&nbsp; This is illustrated in the lower part of the diagram with our numerical example: &nbsp;
 +
#The actual transmitted amplitude coefficients are &nbsp;$a_1= 1$, &nbsp;$a_2= 1$&nbsp; and&nbsp; $a_3= 0$.
 +
#The trellis diagram drawn above was calculated assuming that due to the pulse values &nbsp; $g_{-1} = 0.3$&nbsp; and&nbsp; $g_{0} = 0.7$&nbsp; as well as AWGN noise.
 +
# The noisy input values &nbsp;$d_0= 0.2$, &nbsp;$d_1= 0.7$, &nbsp;$d_2= 0.5$&nbsp; and&nbsp; $d_3= 0$&nbsp; are applied to the maximum likelihood decision.<br>
  
[[File:P ID1473 Dig T 3 8 S4 version1.png|center|frame|Trellisdiagramm und vereinfachtes Trellisdiagramm|class=fit]]<br>
+
[[File:P ID1473 Dig T 3 8 S4 version1.png|right|frame|Trellis diagram&nbsp; (above)&nbsp; and simplified trellis diagram&nbsp; (below)|class=fit]]
 +
<br><br>
 +
This&nbsp; '''simplified trellis diagram'''&nbsp; allows the following statements:
 +
*The decision &nbsp;$a_1= 1$&nbsp; can be made immediately at time &nbsp;$\nu = 1$,&nbsp; since under both hypotheses &ndash; the following symbol is &nbsp; $a_2= 0$&nbsp; or this symbol is &nbsp; $a_2= 1$&nbsp; &ndash; the same result  is obtained.<br>
  
Dieses vereinfachte Trellisdiagramm erlaubt folgende Aussagen:
+
*At time &nbsp;$\nu = 2$&nbsp; the final decision about &nbsp; $a_2$&nbsp; cannot yet be made.&nbsp; Assuming that &nbsp; $a_3= 0$,&nbsp; one would have to decide for &nbsp; $a_2= 1$,&nbsp; while &nbsp; $a_3= 1$&nbsp; would lead to the fixed choice of &nbsp; $a_2= 0$.&nbsp; <br>
*Die Entscheidung über <i>a</i><sub>1</sub> = 1 kann sofort zum Zeitpunkt <i>&nu;</i> = 1 getroffen werden, da sich unter beiden Hypothesen &ndash; nämlich das nachfolgende Symbol ist <i>a</i><sub>2</sub> = 0 bzw. das nachfolgende Symbol  ist <i>a</i><sub>2</sub> = 1 &ndash; das gleiche Resultat <i>a</i><sub>1</sub> = 1 ergibt.<br>
 
  
*Dagegen kann die endgültige Entscheidung über <i>a</i><sub>2</sub> zum Zeitpunkt <i>&nu;</i> = 2 noch nicht getroffen werden. Unter der Annahme, dass <i>a</i><sub>3</sub> = 0 sein wird, müsste man sich für <i>a</i><sub>2</sub> = 1 entscheiden, während die Hypothese <i>a</i><sub>3</sub> = 1 zur Festlegung auf <i>a</i><sub>2</sub> = 0 führen würde.<br>
+
*At&nbsp; $\nu = 3$&nbsp; the decisions for &nbsp; $a_1= 1$, &nbsp;$a_2= 1$, &nbsp;$a_3= 0$&nbsp; are final,&nbsp;  since both continuous paths&nbsp; $($blue and red$)$&nbsp; suggest this&nbsp; $($in this case correct$)$&nbsp; sequence.
 +
*If one were to postpone the decision to time &nbsp;$\nu = 4$,&nbsp; one would not have more usable information about &nbsp; $a_1$,&nbsp; $a_2$&nbsp; and&nbsp; $a_3$.<br><br>
  
*Zur Zeit <i>&nu;</i> = 3 ist die Entscheidung für <i>a</i><sub>1</sub> = 1, <i>a</i><sub>2</sub> = 1, <i>a</i><sub>3</sub> = 0 endgültig, da beide durchgehenden Pfade diese (die in diesem Fall richtige) Folge suggerieren. Würde man die Entscheidung auf den Zeitpunkt <i>&nu;</i> = 4 verschieben, so hätte man nicht mehr nutzbare Information über <i>a</i><sub>1</sub>, <i>a</i><sub>2</sub> und <i>a</i><sub>3</sub>.<br><br>
 
  
Alle Aussagen dieses Kapitels können mit dem folgenden Interaktionsmodul überprüft werden:<br>
 
[[:File:Viterbi.swf|Viterbi&ndash;Empfänger für einen Vorläufer]]<br>
 
  
== Erweiterung auf zwei Vorläufer ==
+
== Extension to two precursors ==
 
<br>
 
<br>
Wird der Grundimpuls durch die Abtastwerte <i>g</i><sub>0</sub>, <i>g</i><sub>&ndash;1</sub> und <i>g</i><sub>&ndash;2</sub> beschrieben, so gibt es im Trellisdiagramm zu jedem Zeitpunkt <i>&nu;</i> genau vier Metriken <i>&Gamma;<sub>&nu;</sub></i>(00), ... , <i>&Gamma;<sub>&nu;</sub></i>(11) zu bestimmen.  <i>&Gamma;<sub>&nu;</sub></i>(01) beschreibt dann beispielsweise die minimale Gesamtfehlergröße für die Detektion des Symbols <i>a<sub>&nu;</sub></i> unter der Hypothese, dass <i>a<sub>&nu;</sub></i><sub>+1</sub> = 0 und <i>a<sub>&nu;</sub></i><sub>+2</sub> = 1 sein werden, und es gilt:
+
If the basic pulse&nbsp; $g_d(t)$&nbsp; is described by the samples &nbsp; $g_{0} \ne 0$,&nbsp; $g_{-1} \ne 0$&nbsp; and&nbsp; $g_{-2} \ne 0$,&nbsp; then four metrics &nbsp; ${\it \Gamma}_{\nu}(00)$,&nbsp; ${\it \Gamma}_{\nu}(01)$,&nbsp; ${\it \Gamma}_{\nu}(10)$&nbsp; and&nbsp; ${\it \Gamma}_{\nu}(11)$&nbsp; must be determined in the trellis diagram at each time &nbsp;$\nu$.&nbsp;  
  
:<math>{\it \Gamma}_{\nu}(01) = {\rm Min}\left[{\it \Gamma}_{\nu-1}(00) + \varepsilon_{\nu}(001),
+
For example, &nbsp;${\it \Gamma}_{\nu}(01)$&nbsp; describes the minimum accumulated metric for the detection of the symbol &nbsp; $a_\nu$&nbsp; under the hypothesis &nbsp; "$a_{\nu-1} = 0$&nbsp; and&nbsp; $a_{\nu-2} = 1$".&nbsp; It holds:
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(10) + \varepsilon_{\nu}(101)\right]
+
:$${\it \Gamma}_{\nu}(01) = {\rm Min}\big[{\it \Gamma}_{\nu-1}(00) + \varepsilon_{\nu}(001),
  \hspace{0.05cm}.</math>
+
  \hspace{0.2cm}{\it \Gamma}_{\nu-1}(10) + \varepsilon_{\nu}(101)\big]
 +
  \hspace{0.05cm}.$$
  
In der Aufgabe  A3.12 wird noch im Detail auf die Berechnung der Fehlergrößen und die Minimierung der Gesamtfehlergrößen eingegangen. Hier betrachten wir nur ein beispielhaftes Trellisdiagramm, das die (fehlerfreie) Detektion von folgender Symbolfolge widergibt:
+
&nbsp; [[Aufgaben:Exercise_3.12:_Trellis_Diagram_for_Two_Precursors|Exercise 3.12]]&nbsp; discusses in detail the metrics calculation and the minimization of the accumulated metrics for the case of two precursors.
  
:<math>a_{1}=0 \hspace{0.05cm},\hspace{0.2cm}
+
{{GraueBox|TEXT= 
 +
$\text{Example 4:}$&nbsp; We consider here an exemplary trellis diagram, which reflects the (error-free) detection of the following symbol sequence:
 +
[[File:P ID1474 Dig T 3 8 S5 version1.png|right|frame|Trellis diagram for two precursors|class=fit]]
 +
 
 +
:$$a_{1}=0 \hspace{0.05cm},\hspace{0.2cm}
 
a_{2}=0  
 
a_{2}=0  
 
\hspace{0.05cm},\hspace{0.2cm} a_{3}=1
 
\hspace{0.05cm},\hspace{0.2cm} a_{3}=1
 
\hspace{0.05cm},\hspace{0.2cm} a_{4}=1
 
\hspace{0.05cm},\hspace{0.2cm} a_{4}=1
 
\hspace{0.05cm},\hspace{0.2cm} a_{5}=1
 
\hspace{0.05cm},\hspace{0.2cm} a_{5}=1
\hspace{0.05cm},\hspace{0.2cm} a_{6}=0 \hspace{0.05cm}, \hspace{0.05cm}...</math>
+
\hspace{0.05cm},\hspace{0.2cm} a_{6}=0 \hspace{0.05cm}, \hspace{0.05cm}\text{...}$$
 
 
[[File:P ID1474 Dig T 3 8 S5 version1.png|Trellisdiagramm für zwei Vorläufer|class=fit]]<br>
 
 
 
Dieses vereinfachte Trellisdiagramm erlaubt folgende Aussagen:
 
*Alle von <i>&Gamma;<sub>&nu;</sub></i><sub>&ndash;1</sub> (00) bzw. <i>&Gamma;<sub>&nu;</sub></i><sub>&ndash;1</sub> (01) abgehende Zweige sind dem Symbol &bdquo;0&rdquo; zugeordnet und blau gezeichnet. Die von den zwei oberen Zuständen abgehenden roten Zweige kennzeichnen die &bdquo;1&rdquo;.<br>
 
 
 
*Verfolgt man die durchgehenden Pfade, so erkennt man die angegebene Folge. Da zum Zeitpunkt <i>&nu;</i> = 6 nur blaue Zweige ankommen, liegen hier die ersten sechs Bit der Folge endgültig fest.<br>
 
 
 
*Teilfolgen könnten aber auch bereits zu den Zeiten <i>&nu;</i> = 1, <i>&nu;</i> = 3 und <i>&nu;</i> = 4 ausgegeben werden, da sich zu diesen Zeiten für alle vier Zustände die gleichen Teilpfade ergeben.<br>
 
  
*Dagegen darf bei <i>&nu;</i> = 2 und <i>&nu;</i> = 5 nicht sofort entschieden werden. Beispielsweise ist zum Zeitpunkt <i>&nu;</i> = 5 nur sicher, dass entweder <i>a</i><sub>5</sub> = 0, <i>a</i><sub>6</sub> = 1 oder <i>a</i><sub>5</sub> = 1, <i>a</i><sub>6</sub> = 0 sein werden.<br>
+
This simplified trellis diagram allows the following statements:
 +
#All branches departing from &nbsp;${\it \Gamma}_{\nu}(00)$&nbsp; or &nbsp; ${\it \Gamma}_{\nu}(01)$&nbsp; are assigned to symbol&nbsp; "$\rm 0$"&nbsp; and drawn in blue.&nbsp; The red branches departing from the two upper states denote symbol&nbsp; "$\rm 1$".<br>
 +
#If one follows the continuous paths,&nbsp; one recognizes the indicated sequence.&nbsp; Since at time &nbsp; $\nu = 6$&nbsp; only blue branches arrive,&nbsp; the first six bits of the sequence are finally fixed here.<br>
 +
#However,&nbsp; partial sequences could also be output already at times &nbsp; $\nu = 1$,&nbsp; $\nu = 3$&nbsp; and&nbsp; $\nu = 4$,&nbsp; since at these times the same partial paths result for all four states.<br>
 +
#In contrast,&nbsp; at &nbsp; $\nu = 2$&nbsp; and&nbsp; $\nu = 5$,&nbsp;decisions may not be made immediately.&nbsp; For example,&nbsp; at &nbsp; $\nu = 5$&nbsp; it is only certain that either&nbsp; "$a_5 = 0$,&nbsp; $a_6 = 1$" &nbsp; will be,&nbsp; or &nbsp; "$a_5 = 1$,&nbsp; $a_6 = 0$". }}<br>
  
== Fehlerwahrscheinlichkeit bei Maximum–Likelihood–Entscheidung==
+
== Bit error probability with maximum likelihood decision==
 
<br>
 
<br>
Die exakte Berechnung der Fehlerwahrscheinlichkeit <i>p</i><sub>B</sub> bei ML&ndash;Entscheidung (beispielsweise mit dem Korrelations&ndash; oder dem Viterbi&ndash;Empfänger) ist sehr aufwändig. Dabei müssen
+
The exact bit error probability calculation for binary maximum likelihood decision&nbsp; $($e.g. with the correlation or the Viterbi receiver$)$&nbsp; is very complex.&nbsp; Thereby
*die Differenzenergien zwischen allen möglichen Symbolfolgen <i>Q<sub>i</sub></i>, <i>Q<sub>j</sub></i> ermittelt werden, wobei die Fehlerwahrscheinlichkeit im Wesentlichen durch die minimale Differenzenergie bestimmt wird;<br>
+
*determine the difference energies between all possible symbol sequences &nbsp; $Q_i$&nbsp; and&nbsp;  $Q_{j \ne i}$,&nbsp; where the error probability is essentially determined by the minimum difference energy;<br>
  
*auch die Einflüsse von Matched&ndash;Filter <i>H</i><sub>MF</sub>(<i>f</i>) und Dekorrelationsfilter <i>H</i><sub>DF</sub>(<i>f</i>) berücksichtigt und der Effektivwert <i>&sigma;<sub>d</sub></i> des Detektionsstörsignals bestimmt werden.<br><br>
+
*also the influences of matched filter &nbsp; $H_{\rm MF}(f)$&nbsp; and decorrelation filter &nbsp; $H_{\rm DF}(f)$&nbsp; have to be taken into account and additionally the rms value &nbsp; $\sigma_d$&nbsp; of the detection noise signal is determined.<br><br>
  
Eine einfache Näherung für die (mittlere) Fehlerwahrscheinlichkeit bei ML&ndash;Entscheidung lautet:
+
{{BlaueBox|TEXT= 
 +
$\text{Without proof:}$&nbsp; A simple approximation for the&nbsp; $($average$)$ &nbsp;'''&nbsp; bit error probability in maximum likelihood decision'''&nbsp; is:
 +
:$$p_{\rm ML} = {\rm Q}\left(g_{\rm max}/{\sigma_d}\right)
 +
\hspace{0.3cm}{\rm with}\hspace{0.2cm}g_{\rm max}= {\rm
 +
Max}\hspace{0.15cm} \vert g_\nu \vert \hspace{0.05cm}.$$
 +
*The approximation is valid only for binary bipolar signaling.&nbsp; For unipolar signaling,&nbsp; the argument of the Q&ndash;function must be halved.}}<br>
  
:<math>p_{\rm ML} = {\rm Q}\left(\frac{g_{\rm max}}{\sigma_d}\right)
+
For the following interpretation and example,&nbsp; it is assumed that&nbsp; $v + 1$&nbsp; basic pulse values&nbsp; $($including the main value$)$&nbsp; are different from zero.&nbsp; Then holds:
\hspace{0.3cm}{\rm mit}\hspace{0.2cm}g_{\rm max}= {\rm
+
#The Viterbi decision must consider all these basic pulse values.&nbsp; This means that a trellis diagram with &nbsp; $2^v$&nbsp; states has to be processed.<br>
Max}\hspace{0.15cm}|g_\nu
+
#The precondition for the validity of the above equation is the uncorrelatedness of the noise at the decision,&nbsp; which is achieved by the decorrelation filter.<br>
|
+
#For the comparison with threshold decision &nbsp; $(p_{\rm ThD})$&nbsp; or decision feedback &nbsp; $(p_{\rm DFE})$,&nbsp; the noise rms value &nbsp; $\sigma_d$&nbsp; is assumed to be constant.<br>
\hspace{0.05cm}.</math>
+
#Optimization of the maximum likelihood system results in very narrowband filters,&nbsp; since all pulse tails can be factored out by the ML algorithm.<br>
 +
#Thus,&nbsp; with constant noise power density &nbsp; $N_0$&nbsp; $($at receiver input$)$,&nbsp; the noise rms value &nbsp; $\sigma_d$&nbsp; $($at decision$)$&nbsp; is for the ML system smaller than for other variants.<br>
 +
#This means: &nbsp; The signal-to-noise ratio gain by the ML decision may be even larger than the following example $($with &nbsp;$\sigma_d = \rm const.)$&nbsp; expresses.<br>
  
Diese Näherung gilt nur bei bipolarer Signalisierung. Bei unipolaren Amplitudenkoeffizienten muss das Argument der Q-Funktion halbiert werden.<br>
 
  
Für die folgende Interpretation und das anschließende Beispiel auf der nächsten Seite wird vorausgesetzt, dass <i>&upsilon;</i> + 1 Grundimpulswerte (inclusive des Hauptwertes) von 0 verschieden sind. Dann gilt:
+
{{GraueBox|TEXT= 
*Der Viterbi&ndash;Entscheider muss alle diese Grundimpulswerte berücksichtigen. Das bedeutet, dass ein Trellisdiagramm mit 2<sup><i>&upsilon;</i></sup> Zuständen zu bearbeiten ist.<br>
+
$\text{Example 5:}$&nbsp; We consider the basic pulse values &nbsp; $g_{0} = 0.6$&nbsp; and&nbsp; $g_{-1} = g_{+1} = 0.2 \ ( = g_1)$.&nbsp; In addition,&nbsp; we assume the constant noise rms value &nbsp; $\sigma_d = 0.1$.&nbsp; For simplicity,&nbsp; these quantities are normalized.<br>
 
 
*Voraussetzung für die Gültigkeit obiger Gleichung ist die Unkorreliertheit der Störungen am Entscheider, die durch das Dekorrelationsfilter erreicht wird.<br>
 
 
 
*Für den Vergleich mit Schwellenwertentscheider (<i>p</i><sub>SE</sub>) bzw. Entscheidungsrückkopplung (<i>p</i><sub>DFE</sub>) wird der Effektivwert <i>&sigma;<sub>d</sub></i> des Detektionsstörsignals als konstant vorausgesetzt.<br>
 
  
*Zu berücksichtigen ist aber, dass die Optimierung des ML&ndash;Systems zu sehr schmalbandigen Filtern führt, da alle Impulsausläufer durch den ML&ndash;Algorithmus herausgerechnet werden können.<br>
+
*For a binary receiver with simple threshold decision &nbsp;$\rm (ThD)$&nbsp; according to the chapter &nbsp; [[Digital_Signal_Transmission/Error_Probability_with_Intersymbol_Interference|"Error Probability with Intersymbol Interference"]],&nbsp; the&nbsp; (average)&nbsp; bit error probability for bipolar signaling is:
 
+
:$$p_{\rm ThD}  =  {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}+\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right)
*Unter der Voraussetzung konstanter Rauschleistungsdichte <i>N</i><sub>0</sub> (am Empfängereingang) ist deshalb der Störeffektivwert (am Entscheider) beim ML&ndash;System kleiner als bei den anderen Varianten.<br>
+
  +{1}/{2}\cdot {\rm
 
+
  Q}\left(\frac{g_{0} }{\sigma_d}\right)+ {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right)
*Das bedeutet: Der Störabstandsgewinn durch die Maximum&ndash;Likelihood&ndash;Entscheidung ist unter Umständen noch größer, als es das nachfolgende Beispiel (mit <i>&sigma;<sub>d</sub></i> = const.) ausdrückt.<br>
+
  \approx {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right) \frac{ {\rm
 
 
== Fehlerwahrscheinlichkeit bei Maximum–Likelihood–Entscheidung (2) ==
 
<br>
 
{{Beispiel}}''':''' Wir betrachten die Grundimpulswerte <i>g</i><sub>&ndash;1</sub> = 0.2, <i>g</i><sub>0</sub> = 0.6 und <i>g</i><sub>1</sub> = 0.2 und gehen vom konstanten Störeffektivwert <i>&sigma;<sub>d</sub></i> = 0.1 aus. Zur Vereinfachung sind alle Größen normiert.<br>
 
 
 
Bei einem Binärempfänger mit einfacher Schwellenwertentscheidung entsprechend [http://en.lntwww.de/Digitalsignal%C3%BCbertragung/Fehlerwahrscheinlichkeit_unter_Ber%C3%BCcksichtigung_von_Impulsinterferenzen#Gau.C3.9Ff.C3.B6rmiges_Empfangsfilter_.281.29 Kapitel 3.2] ergibt sich für die (mittlere) Fehlerwahrscheinlichkeit bei bipolarer Signalisierung:
 
 
 
:<math>p_{\rm SE}  =  \frac{1}{4}\cdot {\rm Q}\left(\frac{g_{0}+g_{1}+g_{-1}}{\sigma_d}\right)
 
  +\frac{1}{4}\cdot {\rm
 
  Q}\left(\frac{g_{0}+g_{1}-g_{-1}}{\sigma_d}\right)</math>
 
::<math> \hspace{0.15cm}  =  \frac{1}{4}\cdot {\rm Q}\left(\frac{g_{0}-g_{1}+g_{-1}}{\sigma_d}\right)
 
  +\frac{1}{4}\cdot {\rm
 
Q}\left(\frac{g_{0}-g_{1}-g_{-1}}{\sigma_d}\right)\approx \frac{{\rm
 
 
  Q}(2)}{4}= 0.57 \%
 
  Q}(2)}{4}= 0.57 \%
  \hspace{0.05cm}.</math>
+
  \hspace{0.05cm}.$$
  
Für den Empfänger mit idealer Entscheidungsrückkopplung erhält man unter Berücksichtigung von <i>g</i><sub>1</sub> = 0 (Kompensation des Nachläufers):
+
*For the&nbsp; [[Digital_Signal_Transmission/Decision_Feedback|receiver with decision feedback equalization]]&nbsp; $\rm (DFE)$,&nbsp; considering &nbsp; $g_{+1} = 0$&nbsp; $($full compensation of the "postcursor"),&nbsp; we obtain:
 
+
:$$p_{\rm DFE} ={1}/{2}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}+\hspace{-0.02cm}g_{-1} }{\sigma_d}\right)
:<math>p_{\rm DFE} = \frac{1}{2}\cdot {\rm Q}\left(\frac{g_{0}+g_{-1}}{\sigma_d}\right)
+
  +{1}/{2}\cdot {\rm
  +\frac{1}{2}\cdot {\rm
+
Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}g_{-1} }{\sigma_d}\right)\approx {1}/{2}\cdot {\rm
  Q}\left(\frac{g_{0}-g_{-1}}{\sigma_d}\right)\approx \frac{{\rm
+
  Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}g_{-1} }{\sigma_d}\right)=\frac{{\rm
 
  Q}(4)}{4}= 0.16 \cdot 10^{-4}
 
  Q}(4)}{4}= 0.16 \cdot 10^{-4}
  \hspace{0.05cm}.</math>
+
  \hspace{0.05cm}.$$
  
Dagegen führt die Anwendung der Maximum&ndash;Likelihood&ndash;Entscheidung (mit Korrelations&ndash; bzw. Viterbi&ndash;Empfänger) zur sehr viel kleineren Fehlerwahrscheinlichkeit
+
*In contrast,&nbsp; the application of the maximum likelihood decision with&nbsp; [[Digital_Signal_Transmission/Optimal_Receiver_Strategies#Matched_filter_receiver_vs._correlation_receiver|"correlation receiver"]]&nbsp; or &nbsp; [[Digital_Signal_Transmission/Viterbi_Receiver#Considered_scenario_and_prerequisites|"Viterbi receiver"]]&nbsp; $\rm (ML)$&nbsp; leads to the much smaller error probability
 +
:$$p_{\rm ML} = {\rm Q}\left(\frac{g_{0}}{\sigma_d}\right)
 +
\approx {\rm  Q}(6) = 10^{-9}
 +
\hspace{0.05cm}.$$
  
:<math>p_{\rm ML} = {\rm Q}\left(\frac{g_{0}}{\sigma_d}\right)
+
*This corresponds to a SNR gain of about &nbsp; $3 \ \rm dB$&nbsp; $($versus $\rm DFE)$&nbsp; and &nbsp; $7.5 \ \rm dB$&nbsp; $($versus $\rm ThD)$&nbsp; over the other two systems.&nbsp; The result of this simple approximation was essentially confirmed by simulations.}}<br>
\approx {\rm Q}(6) = 10^{-9}
 
\hspace{0.05cm}.</math>
 
  
Dies entspricht gegenüber den beiden anderen Systemen einem Störabstandsgewinn von ca. 3 dB (DFE) bzw. 7.5 dB (SE). Das Ergebnis dieser einfachen Näherung wurde durch Simulationen im Wesentlichen bestätigt.<br>
+
<u>Note:</u>&nbsp;
 +
#To apply the described Viterbi algorithm directly,&nbsp; the (normalized) basic pulse values &nbsp; $g_{0} =0.2$,&nbsp; $g_{-1} =0.6$&nbsp; and&nbsp; $g_{-2} =0.2$&nbsp; must be set in the equations.&nbsp;
 +
#Indeed,&nbsp; a time shift by multiples of the symbol duration&nbsp; $T$&nbsp; with respect to the coordinate system does not change the performance of the Viterbi decision.<br>
 +
#The maximum likelihood error probability according to the above equation depends solely on the largest basic pulse value.
 +
#It is quite possible that a "precursor" $($here:&nbsp; $g_{-1} =0.6)$&nbsp; is larger than the main value&nbsp; $g_{0}=0.2$.&nbsp; In case of threshold decision this would lead to a&nbsp;  "closed eye".<br>
  
Um den oben beschriebenen Viterbi&ndash;Algorithmus direkt anwenden zu können, müssen die (normierten) Grundimpulswerte <i>g</i><sub>0</sub> = 0.2, <i>g</i><sub>&ndash;1</sub> = 0.6 und <i>g</i><sub>&ndash;2</sub> = 0.2 eingestellt werden. Eine Zeitverschiebung um Vielfache der Symboldauer <i>T</i> gegenüber dem aus Darstellungsgründen gewählten Koordinatensystem ändert nämlich die Leistungsmerkmale der Viterbi&ndash;Entscheidung nicht.<br>
 
  
Die ML&ndash;Fehlerwahrscheinlichkeit nach obiger Gleichung richtet sich allein nach dem größten aller Grundimpulswerte. Dabei kann es durchaus sein, dass &bdquo;Vorläufer&rdquo; größer sind als der Hauptwert.{{end}}<br>
+
{{BlaueBox|TEXT= 
 +
$\text{Conclusion:}$&nbsp;  
 +
*Maximum likelihood decision is only advantageous in the presence of intersymbol interference.&nbsp;  
  
Aus der obigen Näherung erkennt man weiter, dass eine ML&ndash;Entscheidung nur bei Vorhandensein von Impulsinterferenzen von Vorteil ist. Bei Nyquistentzerrung (das heißt: nur der Grundimpulswert <i>g</i><sub>0</sub> ist von 0 verschieden) arbeitet auch der Maximum&ndash;Likelihood&ndash;Empfänger symbolweise und mit der gleichen Fehlerwahrscheinlichkeit Q(<i>g</i><sub>0</sub>/<i>&sigma;<sub>d</sub></i>) wie ein herkömmlicher Empfänger.<br>
+
*In the case of Nyquist equalization&nbsp; $($that is: &nbsp; only the basic pulse value &nbsp; $g_0$&nbsp; is non-zero$)$&nbsp; the maximum likelihood receiver also operates symbolically and with the same bit error probability &nbsp; ${\rm Q}(g_0/\sigma_d)$&nbsp; as the other receivers.}}<br>
  
== Aufgaben zum Kapitel==
+
== Exercises for the chapter==
 
<br>
 
<br>
[[Aufgaben:3.11 Viterbi–Empfänger und Trellisdiagramm|A3.11 Viterbi–Empfänger und Trellisdiagramm]]
+
[[Aufgaben:Exercise_3.11:_Viterbi_Receiver_and_Trellis_Diagram|Exercise 3.11: Viterbi Receiver and Trellis Diagram]]
  
[[Zusatzaufgaben:3.11 Maximum-Likelihood-Fehlergrößen]]
+
[[Aufgaben:Exercise_3.11Z:_Maximum_Likelihood_Error_Variables|Exercise 3.11Z: Metric and Accumutated Metric]]
  
[[Aufgaben:3.12 Trellisdiagramm für 2 Vorläufer|A3.12 Trellisdiagramm für 2 Vorläufer]]
+
[[Aufgaben:Exercise_3.12:_Trellis_Diagram_for_Two_Precursors|Exercise 3.12: Trellis Diagram for Two Precursors]]
  
[[Aufgaben:3.13 Vergleich SE – DFE – ML|A3.13 Vergleich SE – DFE – ML]]
+
[[Aufgaben:Exercise_3.13:_Threshold_Decision_vs._DFE_vs._Maximum_Liekelihood|Exercise 3.13: Threshold Decision vs. DFE vs. Maximum Liekelihood]]
  
 
{{Display}}
 
{{Display}}

Latest revision as of 12:20, 13 July 2022

Considered scenario and prerequisites


The  "correlation receiver"  is optimal in the sense of the "maximum likelihood decision rule",  that is,  it leads to the minimum error probability for equally likely source symbols.  Disadvantage:

Block diagram of the Viterbi receiver
  • The realization cost increases exponentially with the length  $N$  of the symbol sequence to be detected.
  • Since the sequence is decided jointly,  long delays occur when  $N$  is large.

In the 1970s,  Andrew J. Viterbi  proposed a maximum likelihood receiver that allows detection of parts of the received message and for which the implementation cost is limited even for infinitely long sequences.

It should be noted with regard to the individual components of the block diagram:

  • Adjusted to the basic reception pulse  $g_r(t)$  and the noise,  the  "matched filter"  $H_{\rm MF}(f)$  serves for  "noise power limitation".  The matched filter output signal  $m(t)$  or the sequence  $\langle m_\nu \rangle$  of equidistant signal values after sampling has the best possible signal-to-noise power ratio  $\rm (SNR)$.
  • The task of the decorrelation filter  $H_{\rm DF}(f)$  is to extract from the sequence  $\langle m_\nu \rangle$  the detection samples  $d_\nu = d_{{\rm S}\nu} + d_{{\rm N}\nu}$  whose noise components  $d_{{\rm N}\nu}$  are uncorrelated.  This filter is therefore also called a  "whitening filter". 
  • The  Viterbi decision,  which is the focus of the following considerations,  obtains the binary output sequence  $\langle v_\nu \rangle$  from the continuous-valued input sequence  $\langle d_\nu \rangle$  according to the maximum likelihood rule with the smallest possible error probability  ${\rm Pr}(v_\nu \ne q_\nu)$.


In order to describe the Viterbi algorithm as simply as possible,  some simplifying assumptions are made here:

  1. Let the amplitude coefficients be unipolar   ⇒   $a_\nu \in \{0,\hspace{0.05cm} 1\}$.  Note that using bipolar coefficients   ⇒   $a_\nu \in \{-1,\hspace{0.05cm} +1\}$  requires only a few modifications.  These difficulties concern the description rather than the implementation.
  2. The basic pulse  $g_d(t)$  after the decorrelation filter  $H_{\rm DF}(f)$ consists only of the main value  $g_0 = g_d(t = T_{\rm D})$  and a single precursor  $g_{-1} = g_d(t = T_{\rm D}-T)$.  There are no trailers  ("postcursors")  in our model:  $g_{\nu} = g_d(t = T_{\rm D}+\nu \cdot T)\equiv0$.  Also,  let other precursors be excluded for now:  $g_{-2} = g_{-3}= \text{ ...} = 0$.
  3. This gives for the samples of the continuous decorrelation filter output signal  $d(t)$  at time  $\nu \cdot T$:   $d_{\nu} = a_{\nu}\cdot g_{0} + a_{\nu+1}\cdot g_{-1}+d_{{\rm N}\nu}\hspace{0.05cm}$,  where the noise component  $d_{{\rm N}\nu}$  is assumed to be Gaussian distributed with standard deviation  $\sigma_d$. 


Note:   The algorithm is not more complex with bipolar signaling.   $\bullet$   In contrast, the computational cost increases when the basic detection pulse becomes broader and has more than one precursor  $g_{-1}$.   $\bullet$   The neglect of precursors in the description is not a fundamental limitation because any pulse can satisfy this condition by appropriate choice of the detection time  $T_{\rm D}$.   $\bullet$   In the following all signal values are normalized to  $1$. 

$\text{Example 1:}$  In the graph,  the samples  $d_{ {\rm S}\nu}$  of the noiseless detection signal  are entered as  (blue)  crosses,  where the corresponding amplitude coefficients  $a_1 = 1$,  $a_2 = 1$,  $a_3 = 0$, ... can be read from the source signal  $q(t)$  plotted in green.

Sampling values to illustrate the Viterbi algorithm
  • The basic pulse values are assumed to be  $g_0 = 0.7$  and  $g_{-1} = 0.3$.  It can be further seen that  $d_{ {\rm S}\nu}$  can take only four different values, namely  $0$,  $g_{-1}=0.3$,  $g_0= 0.7$,  $g_0 +g_{-1}= 1$.
  • The noisy samples present at the Viterbi decision  (red dots)  are  $d_0 = 0.2$,  $d_1 = 0.7$,  $d_2 = 0.5$,  $d_3 = 0$, ... ,  where the differences  $d_{ {\rm N}\nu} = d_\nu - d_{ {\rm S}\nu}$  originate from an AWGN noise source.
  • A threshold decision  $($with the threshold at  $E = 0.5)$  would make at least one wrong decision at these ten bits shown  $($at time  $t = 4T)$,  and possibly another at  $t = 2T$  if the sample  $d_2 = 0.5$  is nevertheless slightly smaller than the threshold  $E = 0.5$.
  • In contrast, the Viterbi receiver will correctly decide this sequence of length  $10$,  as shown in the next sections.


Metric and accumulated metric


As in the chapter  "Optimal Receiver Strategies",   $Q \in \{Q_i\}$  specifies the time-limited source symbol sequence consisting of  $N$  binary symbols.

  • Thus,  the number of possible symbol sequences  $Q_i$  is  $2^N$.
  • $V$  denotes the sink symbol sequence of length  $N$  set equal to the most probable sequence  $Q_j$  by the Viterbi decision.


$\text{Definitions:}$ 

  • The  metric  $\varepsilon_{\nu}(i)$  denotes the squared deviation between the actual noisy sample  $d_\nu$  and the noiseless sample  $d_{ {\rm S}\nu}$ belonging to the sequence  $Q_i$: 
$$\varepsilon_{\nu}(i) = \vert d_{\nu} - d_{ {\rm S}\nu}(i) \vert^2 \hspace{0.05cm}.$$
In the literature, this is sometimes referred to as  error quantity.
  • The  accumulated metric  $\gamma_{\nu}(i)$ characterizes the sum of all metrics up to time  $\nu$:
$$\gamma_{\nu}(i)= \sum_{k=0}^{\nu}\varepsilon_{k}(i) \hspace{0.05cm}.$$
In this context,  one also speaks of the  total error quantity.


The graph illustrates the quantities defined above in a tree structure,  from which it can be seen that the accumulated metric can be calculated iteratively:

Representation of the metrics  $\varepsilon_{\nu}(i)$  and accumulated metrics  $\gamma_{\nu}(i)= \gamma_{\nu-1}(i\hspace{0.05cm}')+\varepsilon_{\nu}(i\hspace{0.05cm}'') \hspace{0.05cm}$  in the tree diagram.   Notes:
  $\bullet$   $i$,  $i\hspace{0.05cm}'$,  $i\hspace{0.05cm}''$ are different index variables.
  $\bullet$   Definition is valid for a basic pulse with main value  $g_{0}$  and precursor  $g_{-1}$.
  $\bullet$   With  $v$  precursors the above sum would have to start at  $k = 1 -v$.
  $\bullet$   The parameter  $i \in \{0, \hspace{0.05cm}\text{...} \hspace{0.05cm}, 2^{N+1}-1\}$  is usually represented in binary.
  $\bullet$   It describes the amplitude coefficients  $a_1$, ... ,  $a_{\nu +1}$ $($each  $0$  or  $1)$.
$$\gamma_{\nu}(i)= \gamma_{\nu-1}(i\hspace{0.05cm}')+\varepsilon_{\nu}(i\hspace{0.05cm}'') \hspace{0.05cm}.$$

It should be noted about this graph:

  • The nodes of the tree diagram represent the  "accumulated metric"  $\gamma_{\nu}(i)$.  Their number is doubled with each iteration step.  At time  $\nu$,  there are  $2^{\nu+1}$  such nodes.  For example,  for  $\nu = 3$,  exactly  $2^4 = 16$  nodes can be seen.
  • The  "amplitude coefficients"  associated with the accumulated metric  $\gamma_{\nu}(i)$  are obtained by following the path from the initial node to the node under consideration.  It is agreed that an upward branch is assigned the coefficient  $a_\nu=1$  and a downward branch is assigned the coefficient  $a_\nu=0$.
  • The green highlighted node  $\gamma_{3}(\rm 1100)$  denotes the accumulated metric under the hypothetical assumption that the symbols  $a_1=1$,  $a_2=1$,  $a_3=0$,  $a_4=0$ were sent.  This assignment can also be read from the directions of the arrows in the tree diagram:   First twice upwards,  then twice downwards.
  • Due to the precursor,  the coefficient  $a_4$  must already be taken into account at time  $\nu = 3$.  All nodes  $\gamma_{\nu}(i)$ computed under the assumption  $a_{\nu +1}=1$  are represented by rectangles,  while the hypothesis  $a_{\nu +1}=0$  is symbolized by a rounded rectangle in each case,  for example  $\gamma_{2}(\rm 110)$  or  $\gamma_{3}(\rm 1100)$.
  • The branches in the tree diagram are assigned to the  "metrics"  $\varepsilon_{\nu}(i)$.  With the assumed basic pulse  $($only  $g_{0}$  and  $g_{-1})$,  there are exactly four different metrics at each time point except for the initial state  $(\nu = 0)$: 
$$\varepsilon_{\nu}(00) = |d_{\nu}|^2\hspace{0.05cm},\hspace{0.5cm}\varepsilon_{\nu}(01) = |d_{\nu}-g_{-1}|^2\hspace{0.05cm},$$
$$ \varepsilon_{\nu}(10) = |d_{\nu}-g_{0}|^2\hspace{0.05cm},\hspace{0.2cm}\varepsilon_{\nu}(11) = |d_{\nu}-g_{0}-g_{-1}|^2\hspace{0.05cm}.$$
  • The accumulated metric  $\gamma_{\nu}(i)$  is the sum of the preceding node  $\gamma_{\nu-1}(i\hspace{0.05cm}')$  and the intervening branch  $\varepsilon_{\nu}(i\hspace{0.05cm}'')$.  For example,  for the highlighted nodes:
$$\gamma_{1}(11)=\gamma_{0}(1)+\varepsilon_{1}(11) ,$$
$$ \gamma_{2}(110)=\gamma_{1}(11)+\varepsilon_{2}(10) ,$$
$$ \gamma_{3}(1100)=\gamma_{2}(110)+\varepsilon_{3}(00).$$
  • For the first nodes  $\gamma_{0}(0)$  and  $\gamma_{0}(1)$,  it is considered that, by convention, the symbol  $a_0=0$  is always transmitted before the actual transmission   $(a_1$,  $a_2$, ...$)$.  It follows that:
$$\gamma_{0}(0)=\varepsilon_{0}(00)= |d_{0}|^2 \hspace{0.05cm},\hspace{0.2cm} \gamma_{0}(1)=\varepsilon_{0}(01)=|d_{0}-g_{-1}|^2 \hspace{0.05cm}.$$

The following two examples will hopefully clarify these somewhat tedious statements.

$\text{Example 2:}$  As in  $\text{Example 1}$,  we consider the unipolar source symbol sequence of length  $N=3$  with the following parameter values:

$$g_{0}=0.7 \hspace{0.05cm},\hspace{0.2cm} g_{-1}=0.3 \hspace{0.05cm},\hspace{0.2cm} d_{0}=0.2 \hspace{0.05cm},\hspace{0.2cm} d_{1}=0.7 \hspace{0.05cm},\hspace{0.2cm} d_{2}=0.5 \hspace{0.05cm},\hspace{0.2cm} d_{3}=0 \hspace{0.05cm} \hspace{0.05cm}.$$

Then,  for the metrics  $\varepsilon_{\nu}(i)$  at time points  $\nu = 0$  to  $\nu = 3$:

$$\nu = 0\text{:} \hspace{0.2cm}\varepsilon_{0}(00) = \big [0.2- (0 \cdot 0.7 + 0 \cdot 0.3) \big ]^2=0.04 \hspace{0.05cm},\hspace{0.4cm} \varepsilon_{0}(01) = \big [0.2- (0 \cdot 0.7 + 1 \cdot 0.3) \big ]^2=0.01 \hspace{0.05cm};$$
$$\nu = 1{:} \hspace{0.2cm}\varepsilon_{1}(00) = \big [0.7- (0 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.49 \hspace{0.05cm},\hspace{0.4cm} \varepsilon_{1}(01) = \big [0.7- (0 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=0.16 \hspace{0.05cm},$$
$$\hspace{1.4cm}\varepsilon_{1}(10) = \big [0.7- (1 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.00 \hspace{0.05cm}\hspace{0.4cm}\varepsilon_{1}(11) = \big [0.7- (1 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=0.09 \hspace{0.05cm};$$
$$\nu = 2\text{:} \hspace{0.2cm}\varepsilon_{2}(00) = \big [0.5- (0 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.25 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{2}(01) = \big [0.5- (0 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=0.04 \hspace{0.05cm},$$
$$\hspace{1.4cm}\varepsilon_{2}(10) = \big [0.5- (1 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.04 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{2}(11) = \big [0.5- (1 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=0.25 \hspace{0.05cm};$$
$$ \nu = 3\text{:} \hspace{0.2cm}\varepsilon_{3}(00) = \big [0.0- (0 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.00 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{3}(01) = \big [0.0- (0 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=0.09 \hspace{0.05cm},$$
$$\hspace{1.4cm}\varepsilon_{3}(10) = \big [0.0- (1 \cdot 0.7 + 0 \cdot 0.3)\big ]^2=0.49 \hspace{0.05cm},\hspace{0.4cm}\varepsilon_{3}(11) = \big [0.0- (1 \cdot 0.7 + 1 \cdot 0.3)\big ]^2=1.00 \hspace{0.05cm}.$$


$\text{Example 3:}$  With the metrics  $\varepsilon_{\nu}(i)$  determined in  $\text{Example 2}$,  the accumulated metrics  $\gamma_{\nu}(i)$  can now also be calculated. Below is the tree diagram with the accumulated metrics  $\gamma_{\nu}(i)$  as nodes for the time points  $\nu = 0$  to  $\nu = 3$. 

Iterative calculation of the accumulated metrics (example)


  • The minimum accumulated metric at time  $\nu = 3$  is  $\gamma_{3}(\rm 1100) = 0.14$.  From this,  the coefficients of the most likely transmitted  (unipolar)  sequence according to the present signal values  $d_0 = 0.2$,  $d_1 = 0.7$,  $d_2 = 0.5$  and  $d_3 = 0$  are  $a_1 = 1$,  $a_2 = 1$  and  $a_3 = 0$  (green path).
  • If the sequence length is  $N = 3$  $($that is:   only three symbols are jointly decided by the Viterbi receiver$)$,  the decision  $a_4 = 0$  is also certainly the correct one,  since all coefficients  $a_{\nu>3}$  were assumed to be zero.
  • However,  for longer sequences  $(N > 3)$,  it cannot necessarily be concluded from the minimum value  $\gamma_{3}(\rm 1100)$  that  $a_1 = 1$,  $a_2 = 1$,  $a_3 = 0$  is actually part of the most probable sequence.
  • Considering further sample values  $(d_4$,  $d_5$, ...$)$  this preliminary result could well change.




Minimum accumulated metric


We continue from the numerical values of the last examples:

$$d_{0}=0.2 ,\hspace{0.3cm} d_{1}=0.7 ,\hspace{0.3cm} d_{2}=0.5 ,\hspace{0.3cm} d_{3}=0 ,\hspace{0.3cm} g_{0}=0.7 ,\hspace{0.3cm} g_{-1}=0.3 .$$

Thus,  nothing changes in the tree diagram compared to  $\text{Example 3}$.  Some colored markings in this diagram only indicate in which way the tree diagram can be simplified according to Viterbi's suggestions.  Important properties of the Viterbi decision can be seen in the graph,  e.g. at time  $\nu = 2$: 

Simplification of the tree diagram according to Viterbi


  • At time  $\nu = 2$,  the minimum accumulated metric is  $\gamma_{2}(\rm 101) = 0.05$  (highlighted in brown). 
  • This means:   A decision at  $\nu = 2$  –  based on  $d_0$,  $d_1$  and  $d_2$  –  would have been in favor of sequence  "$\rm 101$"  instead of the transmitted sequence  "$\rm 110$". 
  • It follows:   A too early final determination should be avoided at all costs.  However,  at any time  $\nu$  one can already exclude several partial symbol sequences which need not be considered at later times.
  • To  $\gamma_{2}(\rm 111) = 0.35$  the same metrics  $\varepsilon_{3}(\rm 11)$  and  $\varepsilon_{3}(\rm 10)$  are added as to  $\gamma_{2}(\rm 101) = 0.05$.  Because of  $\gamma_{2}(\rm 111) > \gamma_{2}(\rm 101)$  it is therefore already certain at  $\nu = 2$  that  "$\rm 111$"  cannot be part of the most probable sequence.
  • The same is true for  "$\rm 001$"  and  "$\rm 011$".  Therefore,  these nodes and all their successors need not be considered further  $($brown coverages$)$.
  • Also, the rounded nodes  $\gamma_{2}(\rm 000) = 0.78$,  $\gamma_{2}(\rm 010) = 0.24$,  $\gamma_{2}(\rm 100) = 0.26$  are not part of the most probable sequence,  since they are larger than  $\gamma_{2}(\rm 110) = 0.14$,  which is marked in green.  Also,  these and their successors need not be considered from time  $\nu = 3$  (green covering).


$\text{Conclusion:}$ 

  1. Of the eight nodes at  $\nu = 2$,  only two need to be followed up,  namely the rectangle  $\gamma_{2}(\rm 101) = 0.05$  and the rounded rectangle  $\gamma_{2}(\rm 110) = 0.14$.
  2. These describe the minimum accumulated metrics assuming that  $a_3 = 0$  and  $a_3 = 1$,  respectively.


Representation with the trellis diagram


We now generalize the result of this example.  Under the still valid assumption that the basic pulse  $g_d(t)$  has only one precursor  $(g_{-1})$  besides the main value  $(g_0)$,  the two  minimum accumulated metrics  at time  $\nu$  are formally given by

$${\it \Gamma}_{\nu}(0) = {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(00), \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(10)\right] \hspace{0.05cm},$$
$${\it \Gamma}_{\nu}(1) = {\rm Min}\left[{\it \Gamma}_{\nu-1}(0) + \varepsilon_{\nu}(01), \hspace{0.2cm}{\it \Gamma}_{\nu-1}(1) + \varepsilon_{\nu}(11)\right] \hspace{0.05cm}.$$

The procedure  "minimization of the accumulated metrics"  can be clearly illustrated in the  trellis diagram

Example trellis diagram  $($the numerical values are adapted to the previous examples$)$

A comparison with the figures in the last sections shows:

  • The lower branch represents the minimum accumulated metric  ${\it \Gamma}_{\nu}(0)$  calculated at each time   $\nu$  under the hypothesis  $a_{\nu + 1} = 0$  $($blue rounded squares$)$.
  • The upper branch describes the minimum accumulated metric  ${\it \Gamma}_{\nu}(1)$  under the assumption  $a_{\nu + 1} = 1$  $($red squares$)$. 
  • Besides  ${\it \Gamma}_{\nu}(0)$  and  ${\it \Gamma}_{\nu}(1)$,  the maximum likelihood decision must also store the associated symbol sequences  ("paths").  These branches are highlighted in red and blue, resp.
  • If  ${\it \Gamma}_{\nu}$  arises from the node  ${\it \Gamma}_{\nu-1}(0)$  –  that is,  if the lower of the two incoming branches is highlighted –  the associated symbol is  "$\rm 0$",  otherwise the symbol is  "$\rm 1$".
  • For   $\nu = 3$:  Both,  ${\it \Gamma}_{3}(0) = 0.14$  and  ${\it \Gamma}_{3}(1) = 0.23$,  result from the precursor  ${\it \Gamma}_{2}(0)$,  so that both selected paths point to the symbol  "$\rm 0$"  (blue branches).

Simplified trellis diagram


The advantage of the trellis diagram is that the number of nodes and branches does not double at each iteration step.  By selecting the minimum accumulated metrics,  only those symbol sequences are considered further which could be part of the most probable sequence.

The trellis diagram can be further simplified by drawing in only the selected branches.  This is illustrated in the lower part of the diagram with our numerical example:  

  1. The actual transmitted amplitude coefficients are  $a_1= 1$,  $a_2= 1$  and  $a_3= 0$.
  2. The trellis diagram drawn above was calculated assuming that due to the pulse values   $g_{-1} = 0.3$  and  $g_{0} = 0.7$  as well as AWGN noise.
  3. The noisy input values  $d_0= 0.2$,  $d_1= 0.7$,  $d_2= 0.5$  and  $d_3= 0$  are applied to the maximum likelihood decision.
Trellis diagram  (above)  and simplified trellis diagram  (below)



This  simplified trellis diagram  allows the following statements:

  • The decision  $a_1= 1$  can be made immediately at time  $\nu = 1$,  since under both hypotheses – the following symbol is   $a_2= 0$  or this symbol is   $a_2= 1$  – the same result is obtained.
  • At time  $\nu = 2$  the final decision about   $a_2$  cannot yet be made.  Assuming that   $a_3= 0$,  one would have to decide for   $a_2= 1$,  while   $a_3= 1$  would lead to the fixed choice of   $a_2= 0$. 
  • At  $\nu = 3$  the decisions for   $a_1= 1$,  $a_2= 1$,  $a_3= 0$  are final,  since both continuous paths  $($blue and red$)$  suggest this  $($in this case correct$)$  sequence.
  • If one were to postpone the decision to time  $\nu = 4$,  one would not have more usable information about   $a_1$,  $a_2$  and  $a_3$.


Extension to two precursors


If the basic pulse  $g_d(t)$  is described by the samples   $g_{0} \ne 0$,  $g_{-1} \ne 0$  and  $g_{-2} \ne 0$,  then four metrics   ${\it \Gamma}_{\nu}(00)$,  ${\it \Gamma}_{\nu}(01)$,  ${\it \Gamma}_{\nu}(10)$  and  ${\it \Gamma}_{\nu}(11)$  must be determined in the trellis diagram at each time  $\nu$. 

For example,  ${\it \Gamma}_{\nu}(01)$  describes the minimum accumulated metric for the detection of the symbol   $a_\nu$  under the hypothesis   "$a_{\nu-1} = 0$  and  $a_{\nu-2} = 1$".  It holds:

$${\it \Gamma}_{\nu}(01) = {\rm Min}\big[{\it \Gamma}_{\nu-1}(00) + \varepsilon_{\nu}(001), \hspace{0.2cm}{\it \Gamma}_{\nu-1}(10) + \varepsilon_{\nu}(101)\big] \hspace{0.05cm}.$$

  Exercise 3.12  discusses in detail the metrics calculation and the minimization of the accumulated metrics for the case of two precursors.

$\text{Example 4:}$  We consider here an exemplary trellis diagram, which reflects the (error-free) detection of the following symbol sequence:

Trellis diagram for two precursors
$$a_{1}=0 \hspace{0.05cm},\hspace{0.2cm} a_{2}=0 \hspace{0.05cm},\hspace{0.2cm} a_{3}=1 \hspace{0.05cm},\hspace{0.2cm} a_{4}=1 \hspace{0.05cm},\hspace{0.2cm} a_{5}=1 \hspace{0.05cm},\hspace{0.2cm} a_{6}=0 \hspace{0.05cm}, \hspace{0.05cm}\text{...}$$

This simplified trellis diagram allows the following statements:

  1. All branches departing from  ${\it \Gamma}_{\nu}(00)$  or   ${\it \Gamma}_{\nu}(01)$  are assigned to symbol  "$\rm 0$"  and drawn in blue.  The red branches departing from the two upper states denote symbol  "$\rm 1$".
  2. If one follows the continuous paths,  one recognizes the indicated sequence.  Since at time   $\nu = 6$  only blue branches arrive,  the first six bits of the sequence are finally fixed here.
  3. However,  partial sequences could also be output already at times   $\nu = 1$,  $\nu = 3$  and  $\nu = 4$,  since at these times the same partial paths result for all four states.
  4. In contrast,  at   $\nu = 2$  and  $\nu = 5$, decisions may not be made immediately.  For example,  at   $\nu = 5$  it is only certain that either  "$a_5 = 0$,  $a_6 = 1$"   will be,  or   "$a_5 = 1$,  $a_6 = 0$".


Bit error probability with maximum likelihood decision


The exact bit error probability calculation for binary maximum likelihood decision  $($e.g. with the correlation or the Viterbi receiver$)$  is very complex.  Thereby

  • determine the difference energies between all possible symbol sequences   $Q_i$  and  $Q_{j \ne i}$,  where the error probability is essentially determined by the minimum difference energy;
  • also the influences of matched filter   $H_{\rm MF}(f)$  and decorrelation filter   $H_{\rm DF}(f)$  have to be taken into account and additionally the rms value   $\sigma_d$  of the detection noise signal is determined.

$\text{Without proof:}$  A simple approximation for the  $($average$)$    bit error probability in maximum likelihood decision  is:

$$p_{\rm ML} = {\rm Q}\left(g_{\rm max}/{\sigma_d}\right) \hspace{0.3cm}{\rm with}\hspace{0.2cm}g_{\rm max}= {\rm Max}\hspace{0.15cm} \vert g_\nu \vert \hspace{0.05cm}.$$
  • The approximation is valid only for binary bipolar signaling.  For unipolar signaling,  the argument of the Q–function must be halved.


For the following interpretation and example,  it is assumed that  $v + 1$  basic pulse values  $($including the main value$)$  are different from zero.  Then holds:

  1. The Viterbi decision must consider all these basic pulse values.  This means that a trellis diagram with   $2^v$  states has to be processed.
  2. The precondition for the validity of the above equation is the uncorrelatedness of the noise at the decision,  which is achieved by the decorrelation filter.
  3. For the comparison with threshold decision   $(p_{\rm ThD})$  or decision feedback   $(p_{\rm DFE})$,  the noise rms value   $\sigma_d$  is assumed to be constant.
  4. Optimization of the maximum likelihood system results in very narrowband filters,  since all pulse tails can be factored out by the ML algorithm.
  5. Thus,  with constant noise power density   $N_0$  $($at receiver input$)$,  the noise rms value   $\sigma_d$  $($at decision$)$  is for the ML system smaller than for other variants.
  6. This means:   The signal-to-noise ratio gain by the ML decision may be even larger than the following example $($with  $\sigma_d = \rm const.)$  expresses.


$\text{Example 5:}$  We consider the basic pulse values   $g_{0} = 0.6$  and  $g_{-1} = g_{+1} = 0.2 \ ( = g_1)$.  In addition,  we assume the constant noise rms value   $\sigma_d = 0.1$.  For simplicity,  these quantities are normalized.

$$p_{\rm ThD} = {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}+\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right) +{1}/{2}\cdot {\rm Q}\left(\frac{g_{0} }{\sigma_d}\right)+ {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right) \approx {1}/{4}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}2 \cdot g_{1} }{\sigma_d}\right) = \frac{ {\rm Q}(2)}{4}= 0.57 \% \hspace{0.05cm}.$$
$$p_{\rm DFE} ={1}/{2}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}+\hspace{-0.02cm}g_{-1} }{\sigma_d}\right) +{1}/{2}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}g_{-1} }{\sigma_d}\right)\approx {1}/{2}\cdot {\rm Q}\left(\frac{g_{0}\hspace{-0.02cm}-\hspace{-0.02cm}g_{-1} }{\sigma_d}\right)=\frac{{\rm Q}(4)}{4}= 0.16 \cdot 10^{-4} \hspace{0.05cm}.$$
$$p_{\rm ML} = {\rm Q}\left(\frac{g_{0}}{\sigma_d}\right) \approx {\rm Q}(6) = 10^{-9} \hspace{0.05cm}.$$
  • This corresponds to a SNR gain of about   $3 \ \rm dB$  $($versus $\rm DFE)$  and   $7.5 \ \rm dB$  $($versus $\rm ThD)$  over the other two systems.  The result of this simple approximation was essentially confirmed by simulations.


Note: 

  1. To apply the described Viterbi algorithm directly,  the (normalized) basic pulse values   $g_{0} =0.2$,  $g_{-1} =0.6$  and  $g_{-2} =0.2$  must be set in the equations. 
  2. Indeed,  a time shift by multiples of the symbol duration  $T$  with respect to the coordinate system does not change the performance of the Viterbi decision.
  3. The maximum likelihood error probability according to the above equation depends solely on the largest basic pulse value.
  4. It is quite possible that a "precursor" $($here:  $g_{-1} =0.6)$  is larger than the main value  $g_{0}=0.2$.  In case of threshold decision this would lead to a  "closed eye".


$\text{Conclusion:}$ 

  • Maximum likelihood decision is only advantageous in the presence of intersymbol interference. 
  • In the case of Nyquist equalization  $($that is:   only the basic pulse value   $g_0$  is non-zero$)$  the maximum likelihood receiver also operates symbolically and with the same bit error probability   ${\rm Q}(g_0/\sigma_d)$  as the other receivers.


Exercises for the chapter


Exercise 3.11: Viterbi Receiver and Trellis Diagram

Exercise 3.11Z: Metric and Accumutated Metric

Exercise 3.12: Trellis Diagram for Two Precursors

Exercise 3.13: Threshold Decision vs. DFE vs. Maximum Liekelihood