Lecture
Cellular Network: Spectral Efficiency
In this lecture
Spectral efficiency and trunking capacity.
Spectral Efficiency
Spectral efficiency measures how effective the radio spectrum is
used.
Usually measures the spatial traffic density per bandwith unit.
The unit for spectral efficiency is2 Erlangs/m2 /Hz.
For circuit-switched call traffic, the spectral efficiency can be
expressed in terms of
Gc : offered traffic per channel (Erlangs/channel).
Nslot : number of channels per RF carrier (channels/carrier).
Nc : number of carriers per cell (carriers).
Wsys : total system bandwidth (Hz).
A: area per cell (m2 ).
2 More on Erlangs later
Spectral Efficiency
The spectral efficiency is
Nc Nslot Gc
ηS = Erlangs/m2 /Hz
Wsys A
Let Wchannel be the bandwidth allocated to 1 carrier, then for an
N-cell cluster system,
Wsys
Nc = .
NWchannel
Therefore
Gc Nslot
ηS =
NWchannel A
Spectral Efficiency
Spectral efficiency can be separated into three terms:
ηS = ηB ηC ηT
However these three efficiency terms are inter-related and needs to be
jointly optimized.
Bandwidth efficiency ηB = WNchannel
slot
: measure the efficiency of data
transmission. Maximizing by
Low rate voice coding, efficient data compression.
Bandwidth efficient modulation.
Efficient multiple access schemes.
Powerful error correcting codes.
1
Spatial efficiency ηC = NA : efficient frequency reuse. Maximize by
Reducing cluster size.
Reducing cell area.
Efficient channel assignment.
Trunking efficiency ηT = Gc : how much traffic each channel handles.
Spectral Efficiency
Interpretation:
Spectral efficiency (ηS ) measures how much offered traffic is handled
by 1Hz bandwidth in each m2 of area.
Offered traffic intensity (Gc ) indicate how much traffic needs to be
handled (not how well the system handle the traffic).
The spectral efficiency formulation does not take into account the
grade of service (GoS) provided.
Efficiency comparison among system is only meaningful when the
GoS is specified.
Trunking
Trunking allows cellular networks to use significantly less channels to
serve a large random user community.
Not all users require the service at the same time.
There is a tradeoff between
the number of channels;
traffic intensity: how many subscriber, how often and how long each
subscriber uses the system (on average);
Grade of Service (GoS): probability the service request is
denied/blocked.
Traffic Intensity
Traffic intensity is measured in Erlangs. A channel that is occupied A%
of the time carries A/100 Erlangs.
Offered traffic intensity per user
Au = λH
λ: the average number of call request per unit time per user
(calls/time unit).
H: average duration of a call (time unit).
Total offered traffic intensity
A = UAu
U is the number trunking users.
Grade of Service
There are many criteria for GoS, to name a few
Blocking probability: probability a new call is blocked. (block call
cleared systems)
Dropping probability: probability a hand over is blocked.
Average delay (blocked calls delayed system).
GoS computation: Erlang formulae assumptions
Large number of subscriber: the call arrival rate is (virtually)
independent of the number of existing users.
Poisson call arrivals with rate λ, i.e. inter-arrival time is
exponentially distributed.
call duration is exponentially distributed (with mean H).
Blocked Call Cleared Systems
Blocked call are removed from the system.
Blocking probability– Erlang B function
AK
K!
Pblock (A, K ) = PK Ai
i=0 i!
A: offered traffic in a cell.
K : the number of trunked channels, usually equals to the product of
number of frequencies (per cell) and number of channels per
frequency.
Extended Erlang B
Erlang B formula assumes blocked call are removed from the system.
In practice, blocked users reattempts. This increases the effective
offered traffic.
Assume a fraction α of blocked user reattempt (Recall factor).
Extended Erlang B
1 Set A0 = A, n = 0.
2 Compute the blocking probability according to Erlang B
Pb,n = Pblock (An , K ).
3 Compute the effective offered traffic: An+1 = A0 + αAn Pb,n .
4 Increment n and repeat step 2 until convergence (Pb,n and An is
stable).
Blocked Calls Delayed System
Blocked calls enter a queue and gets served when a channel is available.
Blocking probability - Erlang C function
AK
Pr(delay > 0) = A
PK −1 Ai
AK + K ! 1 − K i=0 i!
A: offered traffic in a cell.
K : the number of trunked channels.
Average delay
H
D = Pr(delay > 0)
K −A
H is the average call duration.
Blocked Calls Delayed System
Large delays are not desirable. Another important GoS indicator is
Probability delay > t
(K − A)t
Pr(delay > t) = Pr(delay > 0) exp −
H
Using the Erlang formulas
The Erlang formulae relate the number of channels, offered traffic
and GoS.
Computing the GoS for a given system is straight forward.
Some searching is needed to compute the number of channels
required or the traffic intensity for a given GoS.
Tables and charts can be used for traffic engineering.
Some online tools are available:
Erlang B formula:
http://www.erlang.com/calculator/erlb/
Erlang C formula:
http://www.math.vu.nl/~koole/ccmath/ErlangC/
https://planetcalc.com/3151/
Examples
Example 1
A cell contains 50 channels. The average call duration is 100 seconds.
How many calls per hour can be handled if a 2% blocking probability is
required?
Example 2
A hexagonal cell within a 4-cell system has a radius of 1.387 km. A total
of 60 channels are used within the entire system. If the load per user is
0.029 Erlangs, λ = 1 call per hour, and a probability of 5% of delayed
call, calculate:
1 number of users per square kilometer that this system can support.
2 the probability that a call will be delayed for more than 10s.
Next lecture
Path loss models.
Shadowing.
Small scale models: Rayleigh fadings, Ricean fading ...
