Untitled 3
Untitled 3
Noise Source
Figure 1.4: Basic block diagram of the communication process including channel code and CRC
code blocks
message called an acknowledgement when it does not detect any errors in the decoded bit
string. The source starts a timer after transmitting a information bit string and waits for an
acknowledgement. If the acknowledgement is not received within a predetermined time duration,
a timeout event occurs and the source decides that the received message was corrupted. It then
retransmits the information bit string. Thus ARQ involves implicitly requesting the source to
resend the corrupted information.
An error detecting code like the CRC is essential for the correct functioning of an ARQ system
but an ECC is not. In fact, an ARQ system is often used to provide reliable communication
without using an ECC across channels where bit errors are rare, i.e. when the BER is low. This
is because using an ECC results in the transmission of redundant bits even when the channel
does not introduce any bit errors, reducing the throughput. On the other hand, using ARQ will
result in retransmissions and hence throughput reduction only when the channel introduces bit
errors. In this sense, an ARQ-based communication scheme trades throughput for reliability
adaptively. The throughput reduction due to the presence of the CRC code occurs irrespective
of whether we use an ECC in conjunction with the ARQ scheme or not. Of course, an ECC
becomes essential when the channel introduces frequent errors, i.e. when the BER is high. This
is because a purely ARQ-based scheme relies on the chance that when an error is detected in
the demodulated bit string one of the retransmissions will pass through the channel without any
errors being introduced. Such an event is likely only when the BER is low but will require a
very large number of retransmissions when the BER is high. In the latter case, using an ECC
will result in a reduction in the BER and the subsequent ARQ mechanism will require fewer
number of retransmissions.
Now we have enough background to discuss the first advantage of dividing long information bit
strings in to smaller bit strings or packets before transmission. Suppose we want to communicate
a million bits across a channel which has a BER of 10−6 . Such a channel introduces, on the
average, one error in every million bits which are transmitted across it. If we transmit the one
million bits at once and if an error is detected by the CRC code, we will have to retransmit the
million bits. If we assume, for simplicity, that the second transmission contains no bit errors the
throughput is halved because we had to transmit twice the number of bits to communicate the
million bits. Now consider a packetized system where the million bits are divided into packets
of size 100,000 bits each and transmitted one packet at a time. Since the channel introduces
one error every million bits, only one of the packets will contain an error and will need to be
retransmitted. Once again if we assume the second transmission contains no bit errors, the
throughput reduction is only due to the retransmission of the 100,000 bits long packet and is
approximately 10%.
In this course, we will discuss several ARQ schemes which represent a tradeoff between through-
put, link utilization and buffer size requirements at the transmitter and receiver. The last two
parameters in this tradeoff will be introduced and discussed in detail in a later chapter.