Exercise 3.11Z: Metric and Accumutated Metric

From LNTwww

Calculation of the minimum accumulated metrics

For the maximum likelihood constellation with bipolar amplitude coefficient   $a_{\rm \nu} ∈ \{+1, –1\}$  discussed in   "Exercise 3.11",  the metrics   $\varepsilon_{\rm \nu}(i)$  and the minimum accumulated metrics   ${\it \Gamma}_{\rm \nu}(–1)$ and ${\it \Gamma}_{\rm \nu}(+1)$  are to be determined.

  1. The basic pulse is given by the two values   $g_0$  and  $g_{\rm –1}$. 
  2. These,  as well as the detection samples   $d_0$  and  $d_1$,  can be taken from the following calculations for the metrics   $\varepsilon_{\rm \nu}(i)$  at times   $\nu = 0$  and  $\nu = 1$. 
  3. Note that the symbol  $a_0 = 0$  is always sent before the actual message   $(a_1$,  $a_2$,  $a_3)$. 


For time  $\nu = 0$  holds:

$$\varepsilon_{0}(+1) \ = \ \big[-0.4- 0.4\big]^2=0.64 \hspace{0.05cm},$$
$$\varepsilon_{0}(-1) \ = \ \big[-0.4+ 0.4\big]^2=0.00 \hspace{0.05cm}.$$

From this it could be concluded already at time   $\nu = 0$  that with high probability   $a_1 = -\hspace{-0.05cm}1$. 

For time   $\nu = 1$,  the following metrics result:

$$\varepsilon_{1}(+1, +1) \ = \ \big[-0.8- 0.6 -0.4\big]^2=3.24 \hspace{0.05cm},$$
$$\varepsilon_{1}(+1, -1) \ = \ \big[-0.8- 0.6 +0.4\big]^2=1.00 \hspace{0.05cm},$$
$$\varepsilon_{1}(-1, +1) \ = \ \big[-0.8+ 0.6 -0.4\big]^2=0.36 \hspace{0.05cm},$$
$$ \varepsilon_{1}(-1, -1) \ = \ \big[-0.8+ 0.6 +0.4\big]^2=0.04 \hspace{0.05cm}.$$

The minimum accumulated metrics   ${\it \Gamma}_{\rm \nu}(-\hspace{-0.07cm}1)$   and   ${\it \Gamma}_{\rm \nu}(+1)$   that can be calculated with these six metrics are already plotted in the graph.  The other detection samples are   $d_{2}=0.1 \hspace{0.05cm},\hspace{0.1cm} d_{3}=0.5 \hspace{0.05cm}.$


Notes:

  • All quantities here are to be understood normalized.
  • Also assume bipolar and equal probability amplitude coefficients:   ${\rm Pr} (a_\nu = -\hspace{-0.05cm}1) = {\rm Pr} (a_\nu = +1)= 0.5.$



Questions

1

What detection samples   $d_0$  and  $d_1$  were assumed here?

$d_0 \ = \ $

$d_1\ = \ $

2

Which basic pulse values were assumed here?

$g_0\ = \ $

$g_{-1} \ = \ $

3

Which of the listed detection samples are possible for   $\nu ≥ 1$? 

$±0.2,$
$±0.4,$
$±0.6,$
$±1.0.$

4

Give the minimum accumulated metrics for time   $\nu = 2$   $(d_2 = 0.1)$.

${\it \Gamma}_2(+1)\ = \ $

${\it \Gamma}_2(-\hspace{-0.05cm}1)\ = \ $

5

Calculate the minimum accumulated metric for time   $\nu = 3$  $(d_3 = 0.5)$.

${\it \Gamma}_3(+1) \ = \ $

${\it \Gamma}_3(-\hspace{-0.05cm}1) \ = \ $


Solution

(1)  From the equations on the information section one can see  $d_0 = \underline{–0.4}$  and  $d_1 = \underline {–0.8}$.


(2)  The metric  $\varepsilon_0(i)$  include the basic pulse value  $g_{\rm –1}$,  which is used to establish the relationship between the amplitude coefficient  $a_1$  and the detection sample  $d_0$  $(g_0$  is not included in these equations$)$.

  • One can see  $g_{\rm –1}\ \underline {= 0.4}$.
  • From the equations for  $\nu = 1$,  the main value $g_0 \ \underline {= 0.6}$  can be read.


(3)  The correct solutions are  1 and 4:

  • The possible useful samples are  $\pm g_0 \pm g_{\rm –1} = \pm 0.6 \pm0.4$,  i.e.  $\underline {±0.2}$  and  $\underline {±1.0}$.
  • In contrast,  unipolar signaling   ⇒   $a_\nu \in \{0, \hspace{0.05cm} 1\}$  would result in values of  $0, \ 0.4, \ 0.6$  and  $1$.
  • The relationship between bipolar values  $b_i$  and unipolar equivalents  $u_i$  is generally:   $b_i = 2 \cdot u_i - 1 \hspace{0.05cm}.$


(4)  The metrics are obtained for  $\nu = 2$  considering the result from  (3)  as follows:

$$\varepsilon_{2}(+1, +1) \ = \ [0.1 - 1.0]^2=0.81,\hspace{0.2cm} \varepsilon_{2}(-1, +1) = [0.1 +0.2]^2=0.09 \hspace{0.05cm},$$
$$\varepsilon_{2}(+1, -1) \ = \ [0.1 -0.2]^2=0.01,\hspace{0.2cm} \varepsilon_{2}(-1, -1) = [0.1 +1.0]^2=1.21 \hspace{0.05cm}.$$

Thus,  the minimum accumulated metrics are:

$${\it \Gamma}_{2}(+1) \ = \ {\rm Min}\left[{\it \Gamma}_{1}(+1) + \varepsilon_{2}(+1, +1), \hspace{0.2cm}{\it \Gamma}_{1}(-1) + \varepsilon_{2}(-1, +1)\right] = {\rm Min}\left[0.36 + 0.81, 0.04 + 0.09\right]\hspace{0.15cm}\underline {= 0.13} \hspace{0.05cm},$$
$${\it \Gamma}_{2}(-1) \ = \ {\rm Min}\left[{\it \Gamma}_{1}(+1) + \varepsilon_{2}(+1, -1), \hspace{0.2cm}{\it \Gamma}_{1}(-1) + \varepsilon_{2}(-1, -1)\right] = {\rm Min}\left[0.36 + 0.01, 0.04 + 1.21\right]\hspace{0.15cm}\underline {= 0.37} \hspace{0.05cm}.$$
Calculation of the minimum accumulated metrics


In the adjacent trellis diagram,  the state  "$1$"  is to be interpreted as  "$+1$"  and  "$0$"  as  "$–1$".

Then holds:

  • ${\it \Gamma}_2(+1) = 0.13$  is the minimum accumulated metric under the hypothesis that the following symbol will be  $a_3 = +1$.
  • Under this assumption,  $a_2 = \ –1$  is more likely than  $a_2 = +1$,  as shown in the trellis diagram  (the incoming path is blue).
  • A realistic alternative to the combination  "$a_2 = \ –1, a_3 = +1$"  is  "$a_2 = +1, a_3 = \ –1$",  which lead to the minimum accumulated metric  ${\it \Gamma}_2(–1) = 0.37$. Here, the incoming path is red.


(5)  For time  $\nu = 3$,  the following equations hold:

$$\varepsilon_{3}(+1, +1) \ = \ [0.5 - 1.0]^2=0.25,\hspace{0.2cm} \varepsilon_{3}(-1, +1) = [0.5 +0.2]^2=0.49 \hspace{0.05cm},$$
$$\varepsilon_{3}(+1, -1) \ = \ [0.5 -0.2]^2=0.09,\hspace{0.2cm} \varepsilon_{3}(-1, -1) = [0.5 +1.0]^2=2.25 \hspace{0.05cm}.$$
$$\Rightarrow \hspace{0.3cm}{\it \Gamma}_{3}(+1) \ = \ {\rm Min}\left[0.13 + 0.25, 0.37 + 0.49\right]\hspace{0.15cm}\underline {= 0.38} \hspace{0.05cm},\hspace{0.8cm} {\it \Gamma}_{3}(-1) \ = \ {\rm Min}\left[0.13 + 0.09, 0.37 + 2.25\right]\hspace{0.15cm}\underline {= 0.22} \hspace{0.05cm}.$$
  • In both equations,  the first term in each case is the smaller,  with  ${\it \Gamma}_2(+1) = 0.13$  included in each case.
  • Therefore,  the Viterbi receiver will certainly output  $a_3 = +1$,  no matter what information it will still get at later times  $(\nu > 3)$.
  • If we follow the continuous path in the trellis diagram from the right to the left,  the other amplitude coefficients are also fixed by fixing  $a_3 = +1$:
$$a_1 = a_2 = \ –1.$$