Lecture 15

IIT Kanpur July 201
20 followers
Aug 2, 2024
[Music]
because that we'll continue discussing
from here okay and the looking at the
segmentation modules we are recapturing
it okay so what is this this is each
code block should be of size K Dash okay
which is B Das by c b Dash is the p Lo
which needs to be transmitted which
includes what B Das includes actual
payload plus transfer block CC Plus Code
block CRC right this is what you intend
to transmit okay now what we are saying
here is it is not done like this and the
primary Culp rate is the ldpc encoder
okay because it cannot encode
7072 okay and in order to handle this we
know now attached to this each code
block is basically the filler bit okay
and how many filler bits we need to
attach that's what we are going to see
now again and I'm taking you to this
next lecture okay so that's what we are
going to recap now okay right so
now we have to just like jog back into
our memory okay and then what happens
here is basically this as ldpc needs to
encode this and how does ldpc encoder
works okay and works by generating
something called this parity check
Matrix okay right and then this parity
check Matrix cannot be generated for any
arbitrary input length input code block
okay in order to generate this arbitary
uh to generate this parity check Matrix
what is defined is basically the base
graph Matrix okay and that needs to be
blown up to the size of what to blown up
to the required size okay and then
blowing up process is called lifting
okay lifting the base graph Matrix okay
and how much do you lift that's what is
decided by this table and it contains
these lifting size which is which the
standard uses okay and now so what is
this this lifting size is denoted by
this number z c okay so this is this
table which is included here for the
convenience and now what do this minimum
value of zc which you should calculate
okay right which should be sufficient to
basically encapsulate this K Das right
what am I saying here what is this KB KB
is basically the number of systematic
columns in which this ldpc encoder
arranges these input bits right this
number remains fixed the length of the
column will change but the number of
columns remain same okay which is equal
to 22 okay so this product should be
greater than K Das okay it can be equal
to K Das also okay at least k Dash okay
if it is equal to K Das then there is no
filler bit if it is greater than K Dash
then filler bits are required okay
because ldpc encoder can only encode K
not K Dash right so if you and as we
have discussed earlier you will have to
write a small piece of code to basically
see which one is sufficient okay is
basically searching how do you implement
the search again an IP okay where in
like how quickly you search through it
better your transmitter design okay or
the ldpc encoder design now if you do
this 22 zc turns out to be 352 one of
these numbers it should be defined
somewhere here you can see that it has
to come from here okay let's see where
is this right so 22 into 352 turns out
to be
7744 and you take any other number okay
it will be greater than 7744 you should
not pick that that zc because you have
to pick minimum zc which basically
ensures that K is greater than equal to
K Das okay that turns out to be 352
right you cannot have any other zc
lesser than this zc also because then K
will be less than K Das okay which
violates this condition now what is the
conclusion of this when segmented
transport block size is not matched to
suitable lifting size which is happening
here okay the segmented transport block
is how much K Das 7072 not matching with
the lifting size okay then filler bits
are added okay and how many the what is
the number of filler bits K minus K Das
672 okay right and we looked at another
example also where the transort block
size was small that was a second example
we remember filler bit were number of
filler bit was only four okay so it
depends on like what is the transer
block size okay right and this typically
okay so it will be in this range okay it
will keep reducing and so what is what
will be the output of this segmented
module okay segmentation module it will
give you four code block of size K okay
and which is k- plus F okay it's a 7744
filler bit bits with FAL 672 that's what
so you are ready to encode and how many
such modules you going to how many such
code blocks you are going to get it
comes here
right and now these numbers should be
comfortable to all of us okay right so B
this plus this is
28192 okay and then it goes through this
transfer block segmentation what you get
is four code blocks first one second
third and fourth okay right and now this
is 30976 because each one of the has a
length 7744 but if you look at carefully
okay what is the difference between the
last code block and the first all other
code blocks okay what are the difference
the transer block CRC sits in the last
code block okay because that's what you
are segmenting correct and naturally
this will go to the last code block okay
right and then so that's what it is so
this is code block CRC filler bit code
block CRC filler bit here transer block
CRC Cod block CRC and the filler bits
right okay and eventually like so this
is basically the input okay this input
is going to go to the ldpc encoder and
now it is ready to encode because you
have generated in such a way that it is
like ldpc encoder can handle this okay
so that's fine okay so what is happening
again okay so what are this ldpc encoder
input is K bits right that's what we
just calculated okay right ldpc encoder
length output length is basically a
function of base graph we have mostly
been focusing on this okay but the all
the ideas remain same for base graph 2
also and we understood why two different
base graphs are there okay I'm not
included that uh discussion here okay so
just to recapture base graph one is
optimized for higher code block length
and and higher rates okay right base
graph 2 is optimized for lower code
block length and lower rates okay so it
was difficult to design a single encoder
for both these extremes and that's why
standard has designed two different base
graphs okay right so in our case Okay so
ldpc
encoder output length is n okay see this
input is K output is n okay then I'm
writing this NR because we are looking
at the Earth code block okay so 66 into
352 okay 23 232 K is how much we just
calculated on the last slide okay and
then what happens here is okay so filler
bits okay are basically replaced with
zeros because ldpc encoder is a binary
encoder okay it can and handle only
either zeros or one okay they are null
bits okay so you need to insert them
with zeros okay while encoded and null
bits are added back after encoding
that's crucial okay because there are
dummy bits which need to be removed
before this data is
transmitted
okay so now what is the rate of each
code block rate of each code block is
7744 by this K and then it exactly comes
out to be rate
1x3 fine okay so mother code rate
right okay why is called mother because
all other code rates are derived from
this kids are derived from the mother
okay that's the notion okay so then but
this is 1x3 and we know user never ask
for 1x3 it will be some rate around 1x3
higher or lower okay so now and this is
what has been asked by the user in this
case 642 correct okay how do you derive
it that's where this rate matching comes
into picture
okay so rate matching what it does it
matches this rate to the one ldpc
encoder is generating and one which is
asked by the user okay and how does it
do that okay and for that you'll have to
see the go through this how many
resources are assigned to this user it's
basically one prb consisting of 14
symbols which has 168 subcarriers or
resource elements okay right six out of
168 sub carriers six are reserved for
pilots and so rest are used for
transmitting data 162 okay so total
number of bits which can be transmitted
for 70 prbs is basically 70 prbs each
prb has 162 resource element and each
resource element can transmit now four
bits because it a 16 qu okay right so
you can transmit these many bits okay
similarly if it is 256 this number will
increase correct okay because you can
trans double the number of bits okay
right not double sorry 256 then it will
be double okay so total number of
segmented code blocks is we saw c equal
to 4 these are to these are the total
number of bits you can transmit so for
each segmented code block okay this is
denoted as e and what is this this is G
G by C right this is 1 1 3 4 0 right for
each Cod block 113 40 bits can be
transmitted okay so this is these are
the numbers okay but how many bits ldpc
and coder is generating n correct 2 3 2
32 how many can you
transmit 1 1 3 4 0 that's precisely what
is rate matching okay
because you're generating at rate 1x3
you need to transmit at a higher rate
okay you're generating more parity bits
than what the user has asked for so what
is the
idea that you basically drop these par
bits at the transmitter itself so that
you can match to the rate what user has
asked for okay right so that's what is
happening and how is it implemented it
is implemented using the circular buffer
okay so you have generated n how many
are you supposed to be transmitting
e correct okay right e is 1 13 40 that
comes from here G by C correct 1 1 3 40
then there's the circular buffer so
First Rate match block will be 11 13 40
all of them will be same this is for the
second one second rate match block okay
right so you'll read 113 40 for the
second code block similarly for the
third code block and similarly for the
fourth code block right so rate matching
implements this is implemented using
this circular buffer please remember
this circular buffer has not kicked in
as of now okay okay the working of
circular buffer will kick in let's say
if the hybrid arq triggers if receiver
is not able to decode data and you have
to transmit some other redundency
version let's say you transmit rv3 okay
rv3 will begin from somewhere here we
saw these numbers we have calculated
them also okay right so what will happen
if you begin from this location you'll
wrap around and go back to the beginning
and read those old bits okay right so
that's where even you're dropping bits
circular buffer property will kick in
when you start transmitting different
Rundy versions typically rv3 you'll have
to wrap around right and that's what it
is so now once you have done the rate
matching I'm not going to go take the
another example where in like the number
the rate asked by the user is lower than
1x3 okay right so that is like will for
the because of the scarcity of time we
will not be able to do it here again
okay so now what happens after this is
you have rate matched okay then the next
process is interleaving so inter leing
what why is inter leing required in
order to handle the burst errors which
ldpc encoder finds it tough to handle
okay right so and what the key notion
here is this row column inter lever okay
and how is it implement content if it is
qpsk you divide the entire bitstream
into two parts stack them together okay
you you're writing like this in a memory
buffer okay let's say a two two
dimensional buffer just for our
understanding you write it like this
then you read it like
this right okay and if you're reading it
like this this is the first bit from the
second half this is second bit from the
first half this is second bit from the
second
half right and and we saw how it is done
given in the standard very simple
instructions okay right so there inter
lever output okay I'm going to tell you
how it is handled at the receiver okay
again please don't search the standard
for the receiver okay so row size
depends on the modulation order okay
this row size depends on the modulation
order motivation is to avoid the burst
errors okay how bust errors are avoided
we have seen that okay burst errors will
happen here when you so all these
continuous bits will be in error but
when you de inter leave them the burst
error will get spread over right
so then let's go back so what what is
following this inter lever module is the
Cod block concatenation okay one of the
simplest modules okay half day job for
you to code what is this you put these
code blocks together and when will this
happen the number of code block is
greater than one okay right otherwise
this will always be a pass through okay
if c equal to one
now so the rate match output is four for
four Cod bit blocks is of length 1 1 3 4
0 total rate matched output is 45 36 0
and this is what we are supposed to be
transmitting correct Okay g that's what
it
is okay so G is this
okay we have come from 1 1 3 40 from
here only right this is e this is G okay
so we put them together we get back G
this is what you allowed to
transmit okay now this is the last part
which we did in last class also but we
are putting everything together okay so
inter lver avoids burst errors Scrambler
helps you in maintaining the unit energy
constraint okay two different tasks okay
and we understood that in great detail
okay now this what is the output length
of this code block concatenation G okay
right so we consider a block of G rate
masch bits and concatenated bits so what
are we doing here we scramble them using
this scrambling sequence okay and that
should also have length
G right okay so what is this output this
output is a xoring of bitwise xoring of
these two sequences okay so you have two
different bit streams you exort them bit
twice and that exort sequence is b i
okay that's the scrambled output okay
right so this is a pseudo random
sequence so we will see how this is
generated also this pseudo random
sequence C okay so that's important part
okay now 5G allows these modulation
schemes okay 4 16 64
256 now what
is what happens here is okay based on
this like what modulation scheme user
has asked for this will transmit two
bits and this will transmit 8 Bits okay
and then you take this
modulation there's a chain after this
also okay right I'm going to cover that
in the subsequent classes okay but this
is that chain is not standard specific
okay this is exactly defined in the
standard okay what comes after this is
purely implementation
okay how do you design your system
because what happens here is that this
is all this is done
digitally okay right it is as if like
you have a like a computer which is
doing it for you that's what happens in
your mobile phone the are the devices
Digital Signal process or fpgs which do
all this digital processing okay but
finally what you transmit is in analog
signal okay the waveform which is
actually transmitted on air okay right
standard doesn't specify how do you
convert it into that analog waveform
that's where your real engineering kicks
in okay and I'm going to show you that
we'll discuss that also and that
basically decides the complete Network
architecture also okay right so you'll
see that the way these systems are
implemented this part is implemented in
one part of the system okay till till
till okay to a certain form and then the
rest of the processing which comes after
this is implemented another part of the
system okay so this is all basically
digital okay then what comes after this
I'll just like jump and tell you it
comes what comes after this is basic is
the other modules but typically this
there's a digital to analog converter
okay right and then I'm taking Liberty
to like hide one or two modules that
will come later okay so there additional
to analog converter which converts this
digital bitstream into analog waveform
okay and that analog waveform basically
needs to be upconverted to the band
which is allocated to a particular
operator okay that is called the RF up
conversion okay this plus some other
modules plus this
DAC is basically implemented on on one
platform other is basically called the
RF front end which is implemented on
another platform okay now so what
happens at the receiver you will down
convert this RF signal okay then you
will what will follow after this RF down
converter is ADC analog to digital
converter because you have to do the
process okay right then that there's a
ADC sitting over here once it gives you
that okay there's few additional modules
you gotten back those these these
samples you can work with them correct
okay you do the complete chain and then
the data is basically if everything
works fine tbcrc gets validated then you
have these bits which you can give it to
the
Mac right okay so this is how typical uh
standard based system is implemented
we'll look at all these details also
what comes here okay so but just to make
sure that you don't get confused when it
comes later that sir only told earlier
till the transmit chain is still here
I'm simplifying it for you so that we
can understand the receiver chain which
follows next because it is for us to
understand how if you implement this
what goes in here okay right and that as
I said we have to like do it on our own
use our fundamentals which are which we
have learned earlier from other classes
okay so we'll see that I'll stop here
for now and then we'll do this receiver
bring in the next class okay thank you
[Music]
[Music]
[Music]
oh
[Music]
