06.06.2022 Views

B. P. Lathi, Zhi Ding - Modern Digital and Analog Communication Systems-Oxford University Press (2009)

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

13.3 Error-Free Communication Over a Noisy Channel 747

spheres of 2-unit radius drawn around each of these two vertices would be nonoverlapping.

In this case, even if channel noise causes two errors, we can still detect the message correctly.

Hence, the reason for the reduction in error probability is that we have not used all the available

vertices for messages. Had we occupied all the available vertices for messages (as is the case

without redundancy, or repetition), then if channel noise caused even one error, the received

sequence would occupy a vertex assigned to another transmitted sequence, and we would

inevitably make a wrong decision. Precisely because we have left the neighboring vertices

of the transmitted sequence unoccupied, are we able to detect the sequence correctly, despite

channel errors within a certain limit. The smaller the fraction of vertices used, the smaller the

error probability. It should also be remembered that redundancy (or repetition) is what makes

it possible to have unoccupied vertices.

Repetition Is Inefficient

If we continue to increase n, the number of repetitions, we will reduce P e , but we will also

reduce Rb by the factor n. But no matter how large we make n, the error probability never

becomes zero. The trouble with this scheme is that it is inefficient because we are adding

redundant ( or check) digits to each information digit. To give an analogy, redundant ( or check)

digits are like guards protecting the information digit. To hire guards for each information digit

is somewhat similar to a case of families living on a certain street that has been hit by several

burglaries. Each family panics and hires a guard. This is obviously expensive and inefficient.

A better solution would be for all the families on the street to hire one guard and share the

expense. One guard can check on all the houses on the street, assuming a reasonably short

street. If the street is too long, it might be necessary to hire a team of guards. But it is certainly

not necessary to hire one guard per house. In using repetitions, we had a similar situation.

Redundant (or repeated) digits were used to help (or check on) only one message digit. Using

the clue from the preceding analogy, it might be more efficient if we used redundant digits

not to check (guard) any one individual transmitted digit but, rather, a block of digits. Herein

lies the key to our problem. Let us consider a group of information digits over a certain time

interval of T seconds, and let us add some redundant digits to check on all these digits.

Suppose we need to transmit a binary information digits per second. Then over a period

of T seconds, we have a block of aT binary information digits. If to this block of information

digits we add (fJ - a)T check digits (i.e., fJ - a check digits, or redundant digits, per second),

then we need to transmit fJT (fJ > a) digits for every aT information digits. Therefore over a

T-second interval, we have

aT = information digits

fJT = total transmitted digits (fJ > a)

(fJ - a)T = check digits

(13.15)

Thus, instead of transmitting one binary digit every 1/a second we let aT digits accumulate

over T seconds. Now consider this as a message to be transmitted. There are a total of 2 aT

such supermessages. Thus, every T seconds we need to transmit one of the 2 aT possible

supermessages. These supermessages are transmitted by a sequence of fJT binary digits. There

are in a112f3T possible sequences of fJT binary digits, and they can be represented as vertices of

a fJT-dimensional hypercube. Because we have only 2 aT messages to be transmitted, whereas

2f3 T ve1tices are available, we occupy only a 2- <f3-a) T

fraction of the vertices of the {JTdimensional

hypercube. Observe that we have reduced the transmission rate by a factor of

a/ fJ. This rate reduction factor a/ fJ is independent of T. The fraction of the vertices occupied

(occupancy factor) by transmitted messages is 2 -( (3-a) T

and can be made as small as possible

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!