|
|
|
|
|
|
We could adjust transmission power to maintain the error rate below threshold.
Error rate is usually measured in BLER, BLock Error Rate -
ratio of erroneous (decoding failed) transport blocks to the total number of transmitted transport blocks.
Better the signal conditions (lower BLER), lower the transmit power needed, lower will be the interference to other users and cells.
As error rate is kept in check, stable data rate is achieved.
This is one type of link adaptation.
Another type of link adaptation technique is to adjust MCS value. MCS stand for Modulation and Coding (redundancy) Scheme.
Better the signal conditions (channel quality), higher will be the MCS chosen (higher order modulation and lesser redundancies).
In a way, data rate keep changing and it follow channel quality.
This process is also known as "Adaptive MCS".
Adaptive MCS is usually preferred method over power control.
Adaptive MCS would always end up selecting only those UEs which have best channel quality among available UEs for selection.
This type of selection is called "Max C/I".
If we always select UEs with best channel quality, cell would always operate at best capacity.
However QoS may not get honoured for all UEs and/or some UEs may get starved.
Channel quality also vary over frequencies (apparent in wide band or higher bandwidth cases).
When allocating RBs, we may want to take this into consideration to get maximum cell capacity.
RR selection may not be most efficient whereas Max C/I may not be most fair.
There is an algorithm called "Proportional Fair Scheduling" (PFS) which bridge the gap between RR and Max C/I.
It tries to select UEs which has best channel quality and has not reached its needed average data rate.
This ensure that the UEs with better channel quality reach needed average data rate faster.
At the same time, as best channel UEs are not scheduled again and again, less likelihood that UEs will be starved.
Selection of time period is critical here, smaller than optimum value will make it look like Max C/I and
longer than optimum will push it near to RR performance.
Before an algorithm is chosen for scheduling, it is imperative to profile the user traffic.
If user traffic is bursty in nature with similar average data rates of all users,
even RR scheduling will work well with very good cell level data rates.
However, in practice, user traffic is not known or it keeps changing based on the time of the day or
it simply is changing due to various unpredictable unknown reasons.
So type of traffic is usually not considered in design, rather it is assumed that the user traffic
will compromise of both constant bit rate and bursty bearers.
|
|
|
|
|
|
|
|
|
|
