The purpose of block coding is to increase the Euclidean distance between the transmitted signals, thus improving so the performances (P(e)) for a given transmitted power at the price of a bandwidth expansion.

To compare codes, we can employ the Hamming distances since the Euclidean ones depend on Hamming distances.

There are two alternatives:

**Soft-decision decoding**(ML receiver).**Hard-decision decoding.**

Assuming that we are employing an antipodal signaling scheme.

The code word c_{i}=(c_{i1},c_{i2},…,c_{in}) is mapped into the sequence:

*with ψ (t)* a signal of duration T and energy *E*.

The hard demodulator makes binary decions on the components of the received vector **r **and then it delivers the received word to the decoder that estimates the code word by calculating the distance between the received word and all M possible code words.

Therefore it finds the code word that is closest to it in the Hamming distance sense.

In hard-decion decoding three basic steps can be recognized:

- Demodulation (r(t) passes througth the match filter and then it is sampled at T).
- Threshold decision device (quantization with two levels).
- Decoding by finding the code word that is closest to received one in the Hamming distance sense.

Decoding is realized on the base of standard array that identifies the 2^{k }decision regions.

The construction of the standard array allows us to divide the observation space, constituted by 2^{n} vectors, into 2^{k} decision regions.

A standard array is generated in the following way:

- The first row is constituted by all code words starting with the all-zero code word.
- The second row is built putting in the first position of the second row a sequence of length n that is not in the first row of the array and with minimum weight. The row is completed adding this sequence with the corresponding code words.
- The third row is constructed in the same way, putting in the first position a minimum weight sequence that is not in the array and then putting in the subsequent positions the sum of this sequence with the corresponding code words.

Each row of the standard array is called a coset and the first element of each coset is called the coset leader.

** Standard Array Properties**

**Theorem 1. All elements of the standard array are distinct.**

Proof: Let us assume two elements of the array are equal.

This can happen in two cases.

1. The elements belong to the same row (**coset** ). In this case we have:

*which* is impossible because the elements belong to different columns.

2. The two elements belong to two different cosets:

belong to the same coset which is impossible since by assumption k ≠ 1

**Theorem 2: if y _{1} and y_{2} belong to the same coset**

Proof:

From this theorem we conclude that each coset can be uniquely identified by the product . We can define the syndrome as

*with* s (called syndrome) a (*n-k*) – *vector* that depends only on the coset leader.

Each coset can be identified uniquely by a syndrome s

Decoding strategy:

- Let us calculate s: yH
^{t}= s - Find the coset leader corresponding to s by using the Syndrome Table
- Decode y as:

- if binary segnaling scheme is antipodal (es. BPSK)

- if binary segnaling scheme is orthogonal (es. BFSK)

- error probability for a binary modulation scheme with

**Since d _{ij}^{H} ≥ d_{min}^{H} and since Q(·) is decreasing function:**

Using the Union Bound we obtain:

Example: Compare the performances of an uncoded data transmission system with the performances of a code system using a (7,4) Hamming code with R=10 Kbit/sec, the received power is P=10^{-6}W, N0/2 =10^{-11}W/Hz and the modulation scheme is binary PSK.

**No coding**

** **

**coding**

Hamming code → d^{H}_{min}=3

Using this code the error probability decreases by a factor of 12.

In this case mod.+AWGN+ demod. hard can be modeled by a BSC:

because the code is linear and therefore they depend only on the weight distribution

The correct decision probability can be calculated as the probability that a correctable error occurs ( see standard array).

It can be also estimated with the union baud.

Hard-decision decoding scheme can obtain the same performances as the optimum one (soft-decision decoding scheme) paying in E_{b}/N_{0} <2 dB for common applications.

If we employ an 8-level quantizer, the performance difference with soft decision reduces to E_{b}/N_{0} < 0.1 dB.

Consider the example of the prvious example and we want to employ a hard decision decoding:

Hard-decision:

*P*_{e}*=P* (a non – correctable error configurationoccurs)=

*= P* (error configuration occurs with weight >1)=

In this case the error probability decreases by a factor 2 (12 in the soft-decision case).

*5*. Digital Transmission over AWGN chanel

*6*. Evaluation of P(e) for optimum RX in AWGN

*7*. Error probability for M-PSK

*8*. Noncoherent ML Demodulation of FSK signals in AWGN

*9*. Transmission through bandlimited AWGN channels

*13*. Cyclic Codes

Progetto "Campus Virtuale" dell'Università degli Studi di Napoli Federico II, realizzato con il cofinanziamento dell'Unione europea. Asse V - Società dell'informazione - Obiettivo Operativo 5.1 e-Government ed e-Inclusion