Encoding and Interleavingof Information Signal

in GSM

Introduction

By

Rama Manohar

Objective

To give an insight in to the various steps and operations involved in the

information transfer in GSM networks .

Topics for discussion

• Speech Encoding

• Data Encoding

• Interleaving for Voice,Control and Data signals

Speech Encoding

• We shall start with a raw voice signal fed into the microphone, travel through the various stages involving vocoding, channel coding etc till it reaches the final burst format on the Air Interface.

Speech Encoding ckt

Voice Encoding

Channel coding

interleaving

RF Modulation

Raw Voicesignal

Speech Encoding ckt

• The voice is sampled at the rate of 50 samples per second.

• This results in 20 msec blocks of speech• Each of this 20 msec block is passed on to

the 13Kbps vocoder.• There are 260 information bits from the

output of the vocoder for every 20 msec input i.e.; 13Kbps *20msec = 260 bits.

Voice Encoding ckt

Vocoder I/p20 msec speech

blocks

13Kbps Vocoder Vocoder O/p260 bits

Channel coding

• Channel Coding is done to protect the logical channels from transmission errors introduced by the radio path.

• The coding schemes depend on the type of the logical channels, hence the coding can differ from speech, control and data .

Channel Coding for speech

Class class 1b class 21a

50 3 132 4 tailBits parity bits

Convolutional coder½ coder, k=5

456 bits=378 bits from Convolution coder + 78 class 2 bits

260 bits

Channel coding for Speech

• The 260 bits of speech info from the vocoder is broken down into three parts.

• Class 1a- 50 bits , these represent the filter coefficients of the speech and are the most important for proper detection of the speech at the receiver and hence are given maximum protection. 3 additional parity bits are derived from the class 1a bits for cyclic redundancy check (CRC).

Channel coding for Speech cont’d

• Class 1b - 132 bits are not parity checked but are fed into the convolutional coder along with 4 tail bits which are used to set the registers in the receiver to a known state for decoding purpose.

• Class 1b- 78 bits, these are not so important and are not protected but are combined with the output of the convolution coder.

Convolutional coder CC

• The Convolutional coder is a series of shift registers implemented using logic gates, where for every one input bit we get 2 output bits. Hence it is called ½ coder.

• Here k=5 is the constraint length, it means there are 5 shift register and each bit has memory depth of 4 , meaning it can influence the output of up to four next successive bits. This is useful during reception as bits can be derived even if a few consecutive bits are lost due to errors or corruption.

½ Convolutional coder

R1 R2 R3 R5R4

+

+

C0output

C1output

0110..Input bits

+ EX-OR R=register

Convolutional coder cont’d

• The output of the CC* is now 378 bits.

(50+3+132+4)*2=378

The total number of bits now is 378+78=456 bits.

*Note : The bit rate from the vocoder was 13Kbps for the 20 msec speech block, but after CC the bit rate increases to 22.8Kbps.

456 bits *20msecs=22.8Kbps * CC = Convolutional Coder.

Control Channel Coding

184 bits Control data

184 40 4 tailFire coded parity bits

½ Convolutional Coder456 bits output

Control Channel Coding

• The control information is received in blocks of 184 bits.

• These bits are first protected with a cyclic code called as Fire code, which is useful in correction and detection of burst errors.

• 40 Parity bits are added, along with 4 tail bits.• These 228 bits are given to the CC whose output is

again 456 bits at a bitrate of 22.8Kbps.• The control channels include the RACH, PCH,

AGCH etc.

Data Channel Coding

240 bits 4 tail Data bits

½ Convolutional CoderOutput= 488 bits

After PuncturingOutput=456 bits

Data Channel Coding

• The data bits are received in blocks of 240 bits. These are directly convolution coded after adding 4 tail bits.

• The output of the CC is now 488 bits, which actually increases the bitrate to 24.4 Kbps.

• To keep the bitrate constant on the air interface we need to puncture the output of the CC. Hence, we have a final bitrate of 22.8 Kbps again .

Channel Coding cont’d

• The above explanation was given keeping in view a full rate Traffic, Control, or Data channel.

• For Half rate or Lesser rates the same principle of channel coding holds good, with slight differences in the encoding process.

Interleaving

• Having encoded the logical channel information, the next step is to build its bit stream into bursts that can be transmitted within the TDMA frame structure. This is the stage where the interleaving process is carried out.

• Interleaving spreads the content of one information block across several TDMA timeslots or bursts.

Interleaving cont’d

• The following interleaving depths are used :• Speech – 8 blocks• Control – 4 blocks• Data – 22 blocks• The interleaving process for a speech block is

shown wherein which a 456 bit speech block is divided into 8 blocks of 57 bits each and each of these odd and even 57 bit blocks are interleaved diagonally on to alternate bursts on the TDMA frame.

Speech Interleaving

8* 57 bits each = 456 bitsOf Speech block N

57Even

Of N-1

57EvenOf N

Speech blockN-1

57odd

Of N-1

57odd

Of N

The speech is spread over 8 such normal burstsEach normal burst consists of two blocks of 57 bit speech

from different 20msec blocks (say N, N-1) along with26 bit training sequence T and 2 flag F plus 6 start stop bits .

T+FT+FT+F

456 bit speech data

Control Data Interleaving

114 114 114 114

456 bits control data

The control data is spread over 4 blocks using rectangular interleaving instead of diagonal interleaving as in speech the receiver will have to wait for at least

2 multiframes before being able to decode the controlmessage

TDMABurst blocks

Data Interleaving

114 114 114 114

Burst 1 Burst 22Burst 2 Burst 3 Burst 4 Burst 19

First 6bits

First 6bits

Last 6bits

Last 6bits

456 bit data block

Data Interleaving cont’d

• Here the data block of 456 bits is divided into 4 blocks of 114 bits each.

• The first 6 bits from each of the 114 bit blocks is inserted in to each frame, the second 6 bits from each of the 114 bits into the next frame and so on spreading each 114 block over 19 TDMA bursts while the entire 456 bits is spread over 22 TDMA bursts.

• Thus the data interleaving is said to have a depth of 22 bursts.

Data Interleaving cont’d

• The reason why data is spread over such along period of time is that if data burst is corrupted or lost, only a small part of it is lost which can be reproduced at the receiver.

• This wide interleaving depth does produce a time delay during transmission but that is acceptable since it does not affect the data signal quality at the receiver, unlike speech where delay could result in bad quality of signal to the subscriber.

• *Note – The interleaving used in data is diagonal interleaving.

Before Deinterleaving3 successive bursts corrupted

After DeinterleavingThe corrupted bursts are spread over a length equal to the

interleaving depth so that the effect of the errors isminimized.

Interleaving Advantage

Air Interface Bitrate

• The information which is now coded and interleaved at 22.8 Kbps now has to be transmitted over the Air interface to the BTS.

• The information burst is not sent directly , but is sent in ciphered form within a burst envelope. This ciphering is done using ciphering keys and algorithms known both by the mobile and the BSS.

Air Interface Bitrate cont’d

• The Kc is the ciphering key and A5 algorithm are applied to the information(speech or data) which increases the bitrate to a final rate of 33.8 Kbps from/to each mobile.

• If we assume all 8 timeslots of the cell to be occupied then the bitrate of the Air interface comes to 33.8 * 8= 270.4 Kbps/channel.

Air Interface Bitrate cont’d

A5 Algorithm

Kc Information Block 22.8 Kbps

Sent on Air interface

Ciphered information burst33.8 Kbps

Air Interface Bitrate cont’d

1 2 3 4 5 6 7 8

Mobile Tx’s at

33.8 Kbps

Cell rx’s 8*33.8 KBps = 270.4 KbpsPer TDMA frame

Cell coverage area

TDMA Fn TDMA Fn+1

Decoding and Deinterleaving at the Receiver

• At the receiver the reverse process of Deinterleaving and decoding have to take place respectively, so as to recover the information from the signal.

• After Deinterleaving the signal will be decoded which is the reverse process of the Convolutional coding, using Viterbi decoders.

• The decoder can recover lost or corrupted data up to 4 successive bits, because the memory depth of the CC is 4(for k=5).