Optimization

Guideline Ref
Rajib Chakrabarti
Mobility parameter guideline

RJIL
MOBILITY
DEDICATED MODE:
There are three ways of optimizing handovers in LTE:

a) Via the modification of the parameters a3offset and hysteresisa3

b) By changing the parameter timetotriggereventa3

c) Via the modification of the parameter filtercoefficient for event a3.

These set of deliberations will dealt with parameter setting for Periodic Reporting of Event A3 only.
The intention is to deal with each of the cases mentioned above, one at a time. Hence, this deliberation
will concentrate in case a).

Definitions:

Event A3 is defined as a triggering event when a neighbour cell becomes an offset better than the
serving cell. The UE creates a measurement report, populates the triggering details and sends the
message to the serving cell. The parameters that define the trigger include:

 a3offset: This parameter can be found in 3GPP 36.331. It configures the RRC IE a3-Offset
included in the IE reportConfigEUTRA in the MeasurementConfiguration IE. The value sent
over the RRC interface is twice the value configured, that is, the UE has to divide the received
value by 2.The role of the offset in Event A3 is to make the serving cell look better than its
current measurement in comparison to the neighbor.
 Hysteresisa3: The role of the hysteresis in Event A3 is to make the measured neighbor look
worse than measured to ensure it is really stronger before the UE decides to send a
measurement report to initiate a handover.
 timetoTriggera3: The role of ttt in Event A3 is to avoid a ping-pong effect.
 CellIndividualoffsetEutran: This parameter is applied individually to each neighbor cell
with load management purposes. The higher the value allocated to a neighbor cell, the “more
attractive” it will be. This parameter can only be used if the neighbor list is broadcast in SIB4
or in an RRC connection reconfiguration.
Based on the picture above, event A3 will trigger when:

RSRP(target) > RSRS(Serving) +a3offset + hysteresisa3 – cellindividualoffsetEutran

And this condition is valid for timetotriggera3.

At the expiration of timetotriggera3, if the UE does not receive an RRC connection reconfiguration
message (handover command) from the eNodeB, then it will start a timer called reportingintervala3.
At the expiration of this timer, if the conditions for event A3 are still met and the eNodeB has not
responded, then another measurement report will be sent to the eNodeB. This process will continue
until the eNodeB responds or until a number of measurement reports given by the parameter
reportingamount have been sent.

Examples:

The table below assumes that cellindividualoffsetEutran is not used and shows when the
eventa3offset is triggered and when the UE ceases sending measurement reports.
 

As it can be seen from the table, eventa3 triggers at a3offset+hysteresisa3

However!!! After the first measurement result, subsequent measurement results can be sent if the
RSRP of the neighbor cell is only a3offset-hysterisisa3 dB stronger! Hence, weaker neighbors could
be reported in the measurements sent by the UE (this case is very rare but it exists in real systems).

Therefore, it is recommended to follow the optimization rules:

a) a3offset should always be larger than hysteresisa3 if we want UE to handover to cells with an
RSRP at least equal to the RSRP value of its serving cell.

b) Ensuring a3offset > hysteresisa3 avoids ping-pongs

c) The higher the value of a3offset+hysteresisa3 the more we drag the calls to neighboring cells. This
is very useful where we have coverage holes (not a one to one deployment scenario on top of 3G
cells)

d) The smaller the value of a3offset+hysteresisa3 the faster we release the calls to neighboring cells.
This is useful in those scenarios where a large number of LTE cells exists in a given geographical
area.

e) The higher the value of a3offset+hysteresisa3 the more difficult we make it for calls do handover
to other cells.

Remember, eventa3 triggers at a3offset+hysteresisa3. Subsequent message reports are sent when the
RSRP of the neighbor cell is a3offset-hysteresisa3 (See figure below).
TimetoTrigger Event a3 

As explained in part 1 of these deliberations, if the RSRP of a neighbor cell is a3offset+ hysteresisa3
dB stronger than the serving cell for a time period equal to timetotriggera3 then the UE sends the
first measurement report to the eNodeB indicating that eventa3 has occurred.  timetotriggera3 typical
values are [0, 40, 64, 80, 100, 128, 160, 256, 320, 480, 512, 640, 1024, 1280, 2560, 5120]
milliseconds.

Clearly, the utilization of timetotriggera3 is highly dependent on the parameters a3offset and
hysteresisa3. However, some general troubleshooting guidelines are provided here to minimize ping
pong effects.

Rules:

a) If a3offset+ hysteresisa3 is relatively large (i.e.: 6dB or stronger), then a value of timetotriggera3
under 100 ms is acceptable.

Explanation: Since the RSRP of the neighbor cell is already stronger than the value of the source cell,
the time to trigger should not be large.

b) If a3offset+ hysteresisa3 is relatively small (i.e.: 2dB), then a value of timetotriggera3 should be
around 320 to 640 ms.

Explanation: Since the RSRP of the neighbor cell is not much stronger than the value of the source
cell, the time to trigger should not large to ensure the value remains the same for a long period of
time.
c)     If a3offset = hysteresisa3, see b)

d)    If a3offset > hysteresisa3, see a)

e)     If a3offset < hysteresisa3, see a)

However, these recommendations depend much on the speed of the mobile and the coverage
scenarios.

The value allocated to timetotriggera3, hence, depends on:

 Parameter setting of a3offset and hysteresisa3,

 Morphology (dense urban, urban, suburban, rural)
 Speed of UE in the cells (freeways and or suburban roads).

So far, we have discussed two methods for optimizing event A3. In out next deliberation we will talk
about the benefits of optimizing another parameter called, filtercoefficient for event A3 that will allow
us to eliminate some of the effects of fast fading in the UE measurements.

Filter Coefficient for Event a3

Once the UE is configured to do measurements, the UE starts measuring reference signals from the
serving cell and any neighbors it detects. The next question is whether the UE should look at just the
current measurement value, or if the recent history of measurements should be considered. LTE, like
other wireless technologies, takes the approach of filtering the currently measured value with recent
history. Since the UE is doing the measurement, the network conveys the filtering requirements to the
UE in an RRC Connection reconfiguration message.

The UE filters the measured result, before using for evaluation of reporting criteria or for
measurement reporting, by the following formula:

where

 Mn is the latest received measurement result from the physical layer;

 Fn is the updated filtered measurement result, that is used for evaluation of reporting criteria
or for measurement reporting;
 Fn-1 is the old filtered measurement result, where F0 is set to M1 when the first measurement
result from the physical layer is received; and
 a = 1 / 2(k/4), where k is the filterCoefficent for the corresponding measurement quantity
received by the quantityConfig.

Then, the UE adapts the filter such that the time characteristics of the filter are preserved at different
input rates, observing that the filterCoefficent k assumes a sample rate equal to 200 ms.

The parameter “a” defines the weight given to current value and (1-a) (i.e., the remaining weight is
given to the last filtered value). For example, if filter coefficient k = 4, then a = ½^(4/4) =1/2. This
means that new measurement has half the weight and the last filtered measurement gets the other half
of the weight.
Example of Filter coefficient values are:

 Case 1: value k = 8 , a = ¼, Fn = ¾ Old + ¼ New

 Case 2: value k = 4, a = ½, Fn = ½ Old + ½ New

Optimization Rules:

a)     A high value of the parameter filtercoefficient will provide higher weight to old measurements
(more stringent filter)(the opposite is true)

b)    The higher the values of filtercoefficient the higher the chances of eliminating fast fading effects
on the measurement reports

1. This eliminates reporting a cell which RSRP was suddenly changed due to multipath or fast
fading
2. Which in turns eliminates the chances to handover to a cell which RSRP was strong for some
milliseconds
3. Therefore reducing the chances for Ping-Pong effects

c)     A value of 8 is typically used in the network although a value of 16 might also be used in dense
urban areas.

Event A3 Parameters Setting

The most important parameters involved in event a3 reporting are listed below:

-     eventA3offset
-     hysteresis
-     timeToTrigger
-     sMeasure
-     cellIndividualOffset
-     triggerQuantity
-     reportAmount
-     reportInterval
-     filterCoefficientRsrp

LTE R8 uses hard handover. Therefore, one of the main optimization concerns is to avoid ping pongs
between cells. Ping pongs significantly reduce user throughput and increases signaling in the E-
UTRAN (in the case of X2 handovers) and in the EPC (in the event of an S1 handover). The table
below shows an example with three different combinations for the parameters eventA3offset and
hysteresis.
Assuming a cellindividualoffset = 0, then:

Event a3 will trigger when

     RSRPsource + eventa3offset +hysteresis <= RSRPtarget + hysteresis 

Event a3 will not be further valid when

     RSRPsource + eventa3offset +hysteresis > RSRPtarget + hysteresis

Under these circumstances:

a)     CASE 1:

         a.     Event a3 will trigger when the RSRP of the target cell is 2dB stronger than the RSRP of the
serving cell

         b.     The UE will cease sending measurement reports when the RSRP of the target cell is less
than 2dB stronger than the RSRP of the serving cell

b)     CASE 2:

         a.     Event a3 will trigger when the RSRP of the target cell is 2dB stronger than the RSRP of the
serving cell

         b.     The UE will cease sending measurement reports when the RSRP of the target cell is
weaker than the RSRP of the serving cell

c)     CASE 3:

         a.     Event a3 will trigger when the RSRP of the target cell is 2dB stronger than the RSRP of the
serving cell

         b.     The UE will cease sending measurement reports when the RSRP of the target cell is -2dB
or weaker than the RSRP of the serving cell

Clearly, case 3 could be counterproductive since a candidate can be reported to the source cell when
the target is weaker than the source cell!!

A healthier approach is to provide a value of say, 3dB to a3offset and a value of 1 dB to the hysteresis
parameter (for core cells). This will ensure that the target cell is at least 4 dB to trigger the event a3
and the handset will not report a candidate when the target is not at least 2dB stronger than the source
cell (assuming that the number of measurement reports given by reportamount haven't expired).

Also,  in order to ensure that the target cell is strong enough than the source cell for a good amount of
time, the parameter timetotrigger should be set to values of 480, 512 or 640 miliseconds. However, a
drive test is recommended before and after these parameters have been modified along with the
creation of counter reports for X2 and S1 handovers.

RACH Preamble Planning

he total number of RACH preambles available in LTE is 64. These preambles are shared among users
for initial access and handover. Access to LTE systems can be classified based on the reservation or
not of preambles for access. When users have a reserved signature to access the system, they are said
to be using Contention Free Random Access (CFRA). On the contrary, when users don't have a
reserved signature for access they are said to use Contention Based Random Access. CFRA is
typically used during handover.  A number of the 64 preambles are normally reserved for handover
only.

The 64 preambles are not implicitly communicated to the UEs by the eNodeB but rather, the UE is
informed about the process of how to generate them via parameters broadcast in SIB2. These
parameters are:

a)     RootConfigurationIndex

In LTE, there are 838 root Zadoff-Chu sequences available for preambles. The length of each root
sequence is 839. RootConfigurationIndex, informs the UE via SIB2 which sequence is to be used.

b)    ZeroCorrelationZoneConfig.

One root sequence can generate several preambles by cyclic shift. One or more root sequences are
needed to generate all preambles in a cell. The UE starts with the broadcasted root index and applies
cyclic shifts to generate preambles. ZeroCorrelationZoneConfig points to a table where the cyclic
shift is obtained from.

The smaller the cyclic shift, the more preambles can be generated from a root sequence. Hence, the
number of sequences needed to generate the 64 preambles in a given cell is:

                        # of rows = ceiling (64 / (integer (sequence length/cyclic shift)))

For example, if the rootsequence index is 300 and the cyclic shift is 119, then, the number of rows
needed to generate the 64 preambles in a cell is:

# of rows = ceiling(64 /(integer(839/119))) = 10

This means, that if we allocated rootsequenceindex 300 to cell X, then cell Y must have
rootsequenceindex 310 and cell Y must have rootsequenceindex 320 as shown in the picture below.

Each cell, then must have a different RootSequenceIndex to avoid the reception of false preambles in
adjacent eNodeBs and the planning could be linked (if desired) to the PCI planning. See figure below.
Cell Radius in LTE (Part 1 of 3)

The cell radius in LTE is affected and/or determined by three factors: a) The Preamble Format b) The Cyclic
Shift the corresponds to the ZeroCorrelationZoneConfig parameter and, c) The Cell Radius Parameter ..

in the previous part (1 out of 3) we discussed the relationship between the preamble format and the
cell radius. In this delivery, we will discuss how the ZeroCorrelationZoneConfig parameter affects the
cell radius.

The parameters ZeroCorrelationZoneConfig and RootSequenceIndex are used to generate 64 random

access signatures in each cell (all these access signatures should be different in each cell). Both, the
ZeroCorrelationZoneConfig and the RootsequenceIndex paramaters are broadcast in SIB2. The
random access sequences are built via the selection of a Zadoff-Chu sequence (one out of 839) given
by RootSequenceSequence and a cyclic shift (used 64 times to generate the 64 random access
signatures from the Zadoff-Chu sequence selected). The cyclic shift is indirectly given to the UE by
the parameter ZeroCorrelationZoneConfig, as shown in the table below (see columns two and three
and note that the cyclic shift has limited values). The available cyclic shifts are listed in 3GPP TS
36.211 table 5.7.2.-2.
The cyclic shift is also related to the cell size. The relationship between the cyclic shift and the cell
size is given by equation (1):

(NCS - 1) * (800 μs/839) ≥ RTD + Delay Spread                                                     (1)

In the equation, RTD stands for Round Trip Delay (twice the cell radius). Hence:

RTD = 2 R/c                                                                                         (2)

Then, the cell radius is given by:

R ≤ [c/2]*[(NCS - 1)*(800 μs/839)-Delay spread]                                                                (3)

For instance, if we assume that ZeroCorrelationZoneConfig is 12, then from the table above, Ncs =
119. Furthermore, if the delay spread = 6 μsec, then the cell size will be approximately 15.97km.
Note that the smaller the cyclic shift, the smaller cell size.

The delay spread in the equation above should be calculated by the RF engineer after a drive test is
carried out in the areas of interest.  The value of the delay spread is typically different for rural,
suburban, urban and dense urban environments.

he third factor that affects the cell radius in LTE is the parameter cell radius. Equipment
manufacturers typical offer a parameter called cellradius, that allows the modification of the cell
radius. The units of this parameter are typically Kilometers.

EXAMPLE:

Let's assume that the preamble format picked (or the only one currently available) is type 0 (which
offers a maximum cell radius of approximately 14 km). The possible values of the parameters
PrachconfigurationIndex are, therefore, 0 to 15. A network operator may decide to classify their
cells into rural, suburban, urban and dense urban cells. Furthermore, the operator may allocate a cell
radius to different morphologies, say: Rural = 14 km, Suburban = 8 km, urban = 5 km and dense
urban = 2 km. In this case, the values of the parameters associated with the cell radius could be:

PrachConfigurationIndex = Any number between 0 and 15 (Preamble Format 0).

Cellradius = 14 (rural), 8 (suburban), 5 (urban), 2 (Dense urban),

ZeroCorrelationZoneConfig =  12 (rural), 9 (suburban), 8 (urban) and 4 (dense urban).

Notes:

a) The value of PrachconfigurationIndex affects the RACH capacity (addressed in a future

deliberation).

b) The value of ZeroCorrelationZoneConfig affects the cell radius as explained in a previous

deliberation (See table below).

Throughput

CQI-PMI-ConfigIndex, RI-Configindex and Their Impact on Throughput

This deliberation analyzes the effect of the value of the parameters CQI-PMI-Configindex and RI-
configIndex on the Downlink throughput. This set of parameters is used when periodic CQI reporting
is used.

CQI-PMI-ConfigIndex determines how often the UE is supposed to report CQI (channel quality
Indicator) and PMI (pre-coding matrix indicator) on the PUCCH.

RI-ConfigIndex determines how often the UE is supposed to report RI (Rank Indicator) on the
PUCCH.

CQI, PMI and RI are transmitted to the eNodeB in format 2,2a or 2b in the PUCCH as shown below
(the picture below assumes a bandwidth of 10 MHz for the UL and 1 resource block for the PUCCH).

The following tables show the periodicity of reporting for CQI/PMI and RI in the units of sub-frames, based on
the configuration index for CQI/PMI and the RI that is sent to the UEs during RRC procedures. The tables are
extracts from TS36.213 (Tables 7.2.2-1A and 7.2.2-1B).

In this deliberation the capacity of the PUCCH for format 2 is assessed.

The following table shows the PUCCH formats used for channel feedback. The channel feedback
could carry channel quality indicator (CQI), Precoding Matrix Indicator (PMI) and Rank Indicator
(RI), depending on transmission mode configured for the UE. Code Division Multiplexing  and
Frequency Division Multiplexing is used to multiplex UE’s on the same RB (more accurately RB-
pairs) configured for PUCCH resources.
Format 2 carries CQI, PMI and ACK/NACKs. The multiplexing capacity could be 4, 6 or 12,
depending on parameter settings. In this example, for illustrative purposes only, let’s assume that the
multiplexing capacity is 4

As explained in the previous deliberation, CQI-PMI-ConfigIndex determines how often the UE is

supposed to report CQI and PMI  on the PUCCH. RI-ConfigIndex determines how often the UE is
supposed to report RI on the PUCCH.

CQI, PMI and RI are transmitted to the eNodeB in format 2, 2a or 2b in the PUCCH as shown below
(the picture below assumes a bandwidth of 10 MHz for the UL and 1 resource block for the PUCCH).

If we assume that CQI-PMI-Configindex is within the range 17 £ ICQI/PMI £ 36, then, the reporting
period is 20 sub-frames or 20 ms.
  

Mapping of cqi-pmi-ConfigIndex to Subframe Periodicity

If we assume that RI-Configindex is within the range 322 £ ICQI/PMI £ 482, then, the rank indicator
reporting period is 20 sub-frames*4 or 80 ms.

Mapping of ri-ConfigIndex to Subframe Periodicity

nRBCQI is a parameter that defines the number of resource blocks for CQI periodic reporting
(Format 2, 2a or 2b). If we assume its value is 2, then, the number of users than can report CQI
periodically in the PUCCH is:
NRBCQI * no. of UE multiplexed per sub-frame *periodicity of CQIreporting

=  (2RB)(4 UE/RB)(20) = 160 UE per CQI reporting period. 

See picture below. The number of UE per PUCCH can be increased by modifying the following
parameters:

a)     Increasing NRBCQI (but the PUSCH capacity will decrease)

b)    Decreasing the periodicity of CQI reporting (decreasing CQI-PMIConfigIndex) but this might
affect our uplink throughput.

c)     Increasing the number of UE multiplexed per RB.

It is up to the operator to decide how to play with this values and achieve the goals planned.
Mapping of cqi-pmi-ConfigIndex to Subframe Periodicity

Mapping of ri-ConfigIndex to Subframe Periodicity

For Instance, if CQI-PMI-ConfigIndex is set to a value between 17 and 36, the UEs are required to
send a CQI and PMI report every 20 sub-frames. That is, every 20 mili-seconds (See highlighted
row). 

On the other hand, if RI-ConfigIndex is set to a value between 322 and 482, then UEs are required to
send a RI report every 4*20ms = 80 ms.

Now, how do CQI-PMI-ConfigIndex and RI-ConfigIndex affect our downlink throughput?

Well… As usual, there is a tradeoff with parameter settings:

a) Low values of these parameters increase the periodicity of reporting CQI, PMI and RI. Hence, the
eNodeB has more accurate information of the downlink conditions each UE has. This allows the
eNodeB to provide better inputs to the scheduling algorithm and select the best choice of MCS as well
as power to transmit in the downlink.

b) The opposite is true.

The problem with setting these two parameters with low values is the need for more PUCCH capacity,
a topic that will be discussed in our next deliberation.

RSRQ…for A3 why not?

As the principal Standards Development Organization (SDO) responsible for developing LTE
Specifications, 3GPP is tasked with providing a framework within which Infrastructure vendors and
Device vendors can reliably build products built to said specifications, that should successfully
interoperate with each other. However there is always the dilemma that must be dealt with by any
Standards Body, which is to say, “how deeply and completely must specifications be written?” In my
past experience at Infrastructure vendors, in working with R&D teams and Standards people within
R&D, there was always a belief that Standards should not be overly defined so as to leave room for
vendor differentiation. Thus there is a balancing act that must take place – the specifications should be
written tightly enough to enable interoperability, but not so tightly as to preclude vendors from
building in their own functionalities and enhancements to add value above and beyond simply,
“building to the Standard.”

In light of this, the topic of this deliberation (to be delivered in two parts) is to discuss the possibility
that an important metric may be under-defined in 3GPP. There seems to be significant anecdotal
evidence coming from the LTE industry that there is an issue with device vendors measuring and
reporting a metric called Reference Signal Received Quality (RSRQ) in a standard fashion. It may be
that the interpretation of the definition of RSRQ in 3GPP is leading to significantly different
implementations by Device vendors such that the use of RSRQ as a mobility trigger is being called
into question, at least at this relatively early stage of LTE development.

Let’s examine the definition of the metric. Per 3GPP TSG RAN; EUTRA; TS36.214 Physical Layer –
Measurements, the following definition in blue is given:

Reference Signal Received Quality (RSRQ) is defined as the ratio N×RSRP/(E-UTRA carrier RSSI),
where N is the number of RB’s of the E-UTRA carrier RSSI measurement bandwidth. The
measurements in the numerator and denominator shall be made over the same set of resource blocks.

E-UTRA Carrier Received Signal Strength Indicator (RSSI), comprises the linear average of the total
received power (in [W]) observed only in OFDM symbols containing reference symbols for antenna
port 0, in the measurement bandwidth, over N number of resource blocks by the UE from all sources,
including co-channel serving and non-serving cells, adjacent channel interference, thermal noise etc.

The reference point for the RSRQ shall be the antenna connector of the UE.

If receiver diversity is in use by the UE, the reported value shall not be lower than the corresponding
RSRQ of any of the individual diversity branches.
At first glance the definition seems to make sense. In multiplying RSRP by the number of N Resource
Blocks in the measurement bandwidth (NxRSRP/E-UTRA carrier RSSI), the definition is striving to
make an apples to apples comparison of RSRP in the numerator versus the contribution of RSRP to
the RSSI in the denominator. Typically, RSSI is the measurement of all energy seen by a receiver in a
given measurement bandwidth and includes thermal noise, additional self-induced noise as
represented by the receiver noise figure, and all desired as well as undesired signals. The reference
signals of the serving cell would certainly qualify as desired signals thus they would also contribute to
the denominator. RSRP in itself is an average measurement of the signal strength of individual
Reference Signals over a measurement (Channel) bandwidth while RSSI is a cumulative measurement
over the total channel bandwidth. Multiplying RSRP by N Resource blocks would seem to be making
the numerator a cumulative quantity over the total channel bandwidth thus lending itself to a logical
comparison to the cumulative quantity expressed by RSSI in the denominator. Before going further,
we should discuss the definition of RSRP itself.

Reference Signal Received Power (RSRP) in itself is an average measurement of the signal strength
of Reference Signals only and the definition is given in TS36.214 as well, per the below text in blue.

Reference signal received power (RSRP), is defined as the linear average over the power
contributions (in [W]) of the resource elements that carry cell-specific reference signals within the
considered measurement frequency bandwidth.

For RSRP determination the cell-specific reference signals R0 according TS 36.211 [3] shall be used.
If the UE can reliably detect that R1 is available it may use R1 in addition to R0 to determine RSRP.

The reference point for the RSRP shall be the antenna connector of the UE.

If receiver diversity is in use by the UE, the reported value shall not be lower than the corresponding
RSRP of any of the individual diversity branches.

Let’s go through some quick examples of how RSRP should be measured based on some assumptions
for a single transmit antenna at the eNodeB and a single receive antenna at the UE. Let’s assume that
the operating parameters are:

FDD Operation, 10 MHz wide carrier bandwidth (50 Resource Blocks), Transmit Power settings of
the downlink Reference Signals = +20 dBm, Path Loss between eNode B transmit antennas and UE
receive antennas = 100 dB
In LTE on the downlink, we set an equal Energy per Resource Element (EPRE) for the different
resource elements, with the reference signals being set to an absolute value of transmit power, and all
other resource elements carrying other channels and signals set to an offset value versus the reference
signal setting. Assuming flat fading for the channel, and not taking into account antennas gains and
coaxial feeder/jumper losses, in our example the RSRP would be simply:

+20 dBm – 100 dB = -80 dBm

For our 10 MHz wide carrier, there would be 50 resource blocks with 2 reference signals per symbol
(always in symbols 0 and 4 in each Resource Block for two antenna operations in the downlink). If
each reference signal is transmitted at +20 dBm (equal to 100 milliwatts in linear terms), the total
power consumed in the eNode B power amplifier by the reference signals is 100 Reference Signals x
100 mW/RS = 10 Watts (+40 dBm) for a single transmit branch. However since RSRP is the received
signal strength averaged over all the reference signals in the measurement bandwidth of 10 MHz, we
use +20 dBm in our example calculation and we assume a path loss of 100 dB applied against each of
the reference signals. What’s interesting to note here is that if we were to calculate a cumulative
RSRP at the UE receiver it is logical to think that the cumulative RSRP would be:

+40 dBm – 100 dB = -60 dBm

In reality, for a non-flat fading scenario, the UE would take into account the fading characteristics of
the channel and this would result in the equivalent of a frequency-dependent path loss over the
channel bandwidth, and the measured RSRP would then differ somewhat from our example, but
again, we are assuming flat fading for simplicity.

Given that this deliberation will extend into multiple parts, before concluding we can draw the first
interesting observation. In looking at the cumulative RSRP versus the average RSRP of the example,
we see there is a difference of 20 dB based on the fact that the cumulative RSRP is measuring the
total contributions of 100 reference signals over one symbol time in 50 resource blocks versus an
average number. So why does the definition use N RB’s X RSRP in the numerator instead of N
Reference Signals X RSRP? We notice in the definition of RSRQ that there is a definition for how
RSSI is measured as well, which itself is an averaged measure.  Stay tuned for Part 2 of this
deliberation as we go deeper into our interpretation.

In Part 1 of this deliberation series, the question was posed as to whether Reference Signal Received
Quality (RSRQ) might be underdefined in standards. The 3GPP Standards definitions of RSRP and
RSRQ were discussed, and a simple example of UE calculation of RSRP was shown. I have decided
to expand this deliberation into three parts due to the length of discussion, and in Part 2 of this
deliberation, we shall go into more discussion of the variables that go into the calculation of RSRQ
and how the determination of these variables might lead to differing results of RSRQ for different
UE’s. Specifically in Part 2 we shall discuss the contribution of the serving cell reference signals to
the numerator and denominator of RSRQ.

As in Part 1, per 3GPP TSG RAN; EUTRA; TS36.214 Physical Layer – Measurements, the following
definition for RSRQ in blue is given:

Reference Signal Received Quality (RSRQ) is defined as the ratio N×RSRP/(E-UTRA carrier RSSI),
where N is the number of RB’s of the E-UTRA carrier RSSI measurement bandwidth. The
measurements in the numerator and denominator shall be made over the same set of resource blocks.

E-UTRA Carrier Received Signal Strength Indicator (RSSI), comprises the linear average of the total
received power (in [W]) observed only in OFDM symbols containing reference symbols for antenna
port 0, in the measurement bandwidth, over N number of resource blocks by the UE from all sources,
including co-channel serving and non-serving cells, adjacent channel interference, thermal noise etc.
The reference point for the RSRQ shall be the antenna connector of the UE.

If receiver diversity is in use by the UE, the reported value shall not be lower than the corresponding
RSRQ of any of the individual diversity branches.

Our example used assumptions of  FDD Operation, 10 MHz wide carrier bandwidth (50 Resource
Blocks), Transmit Power settings of the downlink Reference Signals = +20 dBm, Path Loss between
eNode B transmit antennas and UE receive antennas = 100 dB and we had an interesting initial
observation in comparing the contribution of RSRP in the numerator and the denominator of RSRQ.
The numerator is NxRSRP, and for each of the 50 Resource Blocks in our 10 MHz wide carrier, if the
signal experiences flat fading across the carrier bandwidth, the results would be 50 x the average
RSRP per Resource Block. We computed the average RSRP of – 80 dBm in our example. In order to
make the NxRSRP calculation we  need to express – 80 dBm in linear terms of milliwatts which is
antilog10(-80/10), which yields a value of 1 x 10-8 milliwatts and thus the numerator would be equal to
50 x (1 x 10-8 milliwatts) = 5 x 10-7 milliwatts which is equal to -63 dBm.

Having calculated the numerator we remember that the denominator is equal to the E-UTRA Carrier
Received Signal Strength Indicator (RSSI) and recall the definition of RSSI below:

E-UTRA Carrier Received Signal Strength Indicator (RSSI), comprises the linear average of the total
received power (in [W]) observed only in OFDM symbols containing reference symbols for antenna
port 0, in the measurement bandwidth, over N number of resource blocks by the UE from all sources,
including co-channel serving and non-serving cells, adjacent channel interference, thermal noise etc.

As the reference signals of the serving cell (in LTE Release 8 and 9, there is only one serving cell at
any given time for a UE) will also contribute to the RSSI, we can calculate their contribution as twice
that of the numerator of RSRQ. Why is that? This is because we must take into account that for the
cumulative calculation there are 100 reference signals being transmitted over the carrier bandwidth in
symbols 0 and 4 (for the two antenna transmit case). So, for the flat fading case with each reference
signal transmitted at +20 dBm, and with each reference signal experiencing 100 dB of path loss prior
to reception at the UE, the UE would measure 100 x (1 x 10 -8 milliwatts) = 1 x 10-6 milliwatts which is
equal to -60 dBm.

In summary, we see that the serving cell reference signals contribution to the numerator is – 63 dBm
and their contribution to the denominator is -60 dBm. Is there anything wrong with this? No, it is
simply the way that the definition of RSRQ is constructed that leads to my interpretation of this
seemingly built-in difference of 3 dB because of choosing N Resource blocks instead of 2N Resource
Blocks in the expression in the numerator.

In Part 2 of this deliberation, we looked at the contribution of the downlink Reference Signals to the
numerator and denominator of RSRQ. I noted at the end of part 2 that there seemed to be a 3 dB
greater contribution to the denominator versus the numerator due to the use of NxRSRP in the
numerator whereas in the denominator, there would be 2N discrete reference signals each of which
would lend a contribution to the overall RSSI. Let’s revisit the definition of RSRQ again (3GPP TSG
RAN; EUTRA; TS36.214 Physical Layer – Measurements).

Reference Signal Received Quality (RSRQ) is defined as the ratio N×RSRP/(E-UTRA carrier RSSI),
where N is the number of RB’s of the E-UTRA carrier RSSI measurement bandwidth. The
measurements in the numerator and denominator shall be made over the same set of resource blocks.

E-UTRA Carrier Received Signal Strength Indicator (RSSI), comprises the linear average of the total
received power (in [ W ]) observed only in OFDM symbols containing reference symbols for antenna
port 0, in the measurement bandwidth, over N number of resource blocks by the UE from all sources,
including co-channel serving and non-serving cells, adjacent channel interference, thermal noise etc.

The reference point for the RSRQ shall be the antenna connector of the UE.

If receiver diversity is in use by the UE, the reported value shall not be lower than the corresponding
RSRQ of any of the individual diversity branches.

Coming back to our original issue, which is the postulation that there may be issues with using RSRQ
as a mobility trigger due to different vendor’s mobiles reporting RSRQ differently, let’s now examine
where the possibilities for error lie. Please see the figure below. We will assume a network made up
of only three cells with a single UE under consideration.

Let’s assume that the UE will measure the cells around it and will calculate RSRP and RSRQ for each
of cells 1, 2, and 3. A UE measurement report in LTE will include the RSRP and RSRQ for the PCI of
each measured cell. For RSRP measurements, this is very straightforward. As previously discussed in
part 1 of this deliberation, RSRP is a signal strength measurement made only on a cell’s downlink
cell-specific reference signals and the UE will make RSRP measurements for cells 1, 2, and 3
independently based upon the path loss to each of those cells. For RSRQ which is a signal quality
measurement, the UE will make a measurement for each of the cells using N×RSRP/(E-UTRA carrier
RSSI) which means the numerator will be different for each cell, but the denominator will be the same
per the following:

Cell 1 RSRQ = NxRSRPcell1/RSSI

Cell 2 RSRQ = NxRSRPcell2/RSSI
Cell 3 RSRQ = NxRSRPcell3/RSSI
RSSI, which stands for Received Signal Strength Indication is a measurement of all the received
power in a measurement bandwidth, which includes reference signal energy but also the energy of
subcarriers carrying user data, RRC Signaling, and other control signaling. The definition of RSRQ
states that RSSI is only measured in symbols that contain Reference Symbols for antenna port 0,
therefore an apples-to-apples comparison effort is made here as well to only contain the RSSI
measurement to the same time period that the RSRP measurement is made.

Remember that we do not have soft handoff in Release 8/9 LTE and a UE is only in communication
with one serving cell at a time. In our diagram above we show cell 1 as the serving cell and cell 2 and
cell 3 as non-serving cells. Therefore any data or signaling being sent to the UE from cell 1
contributes useful power to the RSSI. However, intercell interference will be created on the downlink
to the UE receiver and will also contribute to the RSSI measurement if the non-serving cells 2 and 3
are sending data or signaling to their own served UE’s during the same time period and their
transmission’s RF energy is sufficiently high to be captured by our UE receiver making
measurements. From our UE’s perspective above, any RF energy captured of non-serving cell
subcarriers carrying other user’s data or signaling in other cells is pure interference and will be
reflected in worsened RSRQ measurements. This is an example of LTE network-generated self-
interference and leads to the realization that higher traffic loads carried in an LTE network will lead to
worsened average RSRQ measurements. Thus RSRQ lends itself to be another handover trigger that
takes into account network load/interference, whereas RSRP does not. In today’s world a major issue
is also external interference from non-LTE sources but we are not considering that in this example.

So now that we have had a fairly thorough discussion on the definition of RSRQ and how it is
measured, where are the possibilities for interpretational differences between different UE vendors?

There are two possibilities in the standard and these are “measurement bandwidth” and “linear
average” which are highlighted in red in the RSRQ definition. From my perspective, I will consider
that the measurement bandwidth always means the full carrier bandwidth (i.e. 1.4/3/5/10/15/20 MHz)
and is constant across a network (but could change at network borders) and therefore should not be a
contributing issue. The remaining culprit is “linear average”. Due to the length of part 3 of this
deliberation, I will create Part 4 which will be the final part of this overall deliberation, and we shall
discuss this linear averaging that may point to the possible interpretational differences in RSRQ
reporting.

In Part 3 of this deliberation, we discussed the definition of RSRQ, applied it to a miniature network
of 3 cells and looked at how RSRP and RSSI were captured in the formula for RSRQ. I pointed out
that RSSI captures the effects of network loading and interference whereas RSRP does not. I then
pointed out that two terms – “measurement bandwidth” and “linear average” in the definition might
be possibilities for different implementations for different UE vendors, but then ruled out
measurement bandwidth by assuming that measurement bandwidth always means the full carrier
bandwidth (1.4/3/5/10/15/20 MHz) of a deployed LTE RF Carrier. Thus we are down to the final part
of this deliberation in which we discuss the linear average terminology and what that means.

LTE UE’s perform Layer 3 (RRC layer) time-based signal averaging on the measurements gathered at
the physical layer, based upon filter coefficients that are configured on them by the serving eNB via
the RRC Connection Reconfiguration message. The filter coefficient is configurable and may be
turned off or on, and if turned on, the value is configurable. These filter coefficients are used to apply
a weighting value “a” upon the current measurement being conducted, and also a weighting value (1 –
a) applied to a previously calculated measurement and the two weighted measurements are then
summed for the resulting single measurement value. Thus if these filter coefficients are used, LTE
UE’s essentially never provide instantaneous measurement values of RSRP or RSRQ in a
measurement report, but are always providing a “rolling” average value which is always affected by
the previous average measurements. In my experience these filter coefficients are turned on and
configured in real networks and provide a smoothing function to minimize excessive signal variation
for measurement reporting. According to the LTE RRC specification TS36.331, in the Layer 3
Filtering section, the filter coefficient assumes a sample rate of 200 milliseconds, which has to do
with how often Layer 3 is sampling the physical layer measurements and then using that input to
calculate the filtered measurements every 200 milliseconds, after which those filtered measurements
actually get used for measurement reporting. So in essence, the LTE UE will calculate Layer 3 RSRP
and RSRQ for measured cells every 200 milliseconds and these are averaged measurements per the
above discussion. These measurements may be reported in periodic measurement reporting, or event-
based measurement reporting, or event-based periodic reporting or a combination of the two. Ok, so
how are the UE Physical Layer measurements made which get reported to Layer 3 every 200
milliseconds?

The requirements for Physical Layer measurements and measurement reporting are found in LTE
Specification TS36.133, Requirements for Support of Radio Resource Management. This
specification contains the relevant information for UE Measurements Procedures in the RRC
Connected state for both RSRP and RSRQ for intra frequency, inter frequency, and IRAT reporting. I
will restrict the discussion in this deliberation to the intra frequency case, although certainly the other
cases could be discussed in the future as well.

In the E-UTRAN FDD Intra frequency measurements section of the specification, it is stated that the
measurement period for intrafrequency measurements shall be 200 milliseconds. Also, it is stated that
when no measurement gaps are activated, the UE shall be capable of performing RSRP and RSRQ
measurements for 8 identified intra-frequency cells, and the UE physical layer shall be capable of
reporting measurements to higher layers with the measurement period of 200 milliseconds, which is
in keeping with the definition of the Layer 3 sampling period of 200 milliseconds.

Finally, in the Measurements Reporting Requirements for UE section of TS36.133, it is stated for
EUTRAN measurements that “The reported measurement result after layer 1 filtering shall be an
estimate of the average value of the measured quantity over the measurement period.” This statement
covers both RSRP and RSRQ measurements.

Based on these requirements from the two specifications my conclusion is this: The UE will make
Physical Layer measurements of RSRP and RSRQ over a 200 millisecond measurement period. The
UE shall calculate an average value of RSRP and RSRQ for that period and report the value to upper
layers (Layer 3 RRC) at 200 millisecond intervals. If enabled, Layer 3 filtering shall be applied to the
averaged physical layer measurements which means that the physical layers undergo a second stage of
averaging per the description in TS36.331 at Layer 3. I have tried to draw this below.

In the diagram the layer three filtered result (in green) is a combination of the previous layer 3
averaged result multiplied by (1-a), plus the averaged Layer 1 measurement (in blue) multiplied by a.
For example, for the second L3 measurement result this is calculated as M2 = M1(1-a) +
M2(a).
Ok, so finally where are the possibilities for different vendor implementations? It appears that it is up
to vendor implementation to decide how to do the Layer 1 averaging in the defined 200 millisecond
measurement intervals. In 200 milliseconds, there are 200 subframes. A vendor might choose to make
measurements over all of those subframes or a subset of those subframes, even down to as few as a
single subframe per 200 subframes before computing an average and reporting it to Layer 3. The
reality is that traffic conditions can change drastically in an LTE network in a very short period of
time, given that the scheduling interval is 1 millisecond. A sharp rise or drop in traffic can generate a
corresponding sharp rise or drop in RSSI which directly affects the RSRQ measurement! Given that
the physical layer averaging is done across a span of 200 milliseconds, which is a relatively long
period versus the scheduling interval of 1 millisecond, and knowing that the averaging at the physical
layer is up to UE vendor implementation, in my opinion this is where there could be a possibility for
different interpretations of the definition of RSRQ. Whew! Any input to this deliberation is quite
welcome, please contact me if you have any insights you can add to my own.

IDLE MODE
One of the important activities performed by UE during RRC Idle mode is listening to paging
message. The UE should listen to the paging message to know about incoming calls, system
information change, and ETWS (Earthquake and Tsunami Warning Service) notification for ETWS
capable UEs. In LTE, the UE needs to monitor PDCCH continuously (every 1ms) for P-RNTI to
listen to paging message and this drains UE battery power heavily. Instead, if the UE is allowed to
monitor P-RNTI in PDCCH, only at predetermined period say every 60ms or 100ms, it will reduce
UE’s power consumption. The process of monitoring PDCCH discontinuously for P-RNTI to listen to
paging message during RRC idle state is known as Discontinuous Reception (DRX) in RRC_Idle.

What is Idle mode for LTE UE?

 No NAS Signaling connection exists between UE and core network

 At least one PDN Connection exists
 UE is registered with MME & EMM context exists at MME
 UE location is known to the MME with an accuracy of a list of tracking areas
 No RRC connection exists between UE and eNB
 The eNB has no context information

Paging in Idle Mode

Paging is a procedure to transmit paging message to the UE in RRC Idle mode or EMM registered
state. Paging message may be triggered either by MME in core network or eNB in Radio Access
Network.

Discontinuous Reception for Paging

Paging information for the UE is being carried on the PDSCH in the resource blocks indicated by the
PDCCH. Paging indication on the PDCCH is single fixed indicator (FFFE) called the Paging RNTI
(P-RNTI). Different group of UEs monitor different sub frames for their paging messages.

UEs in idle mode monitor the PDCCH channel for P-RNTI used for paging to receive paging message
from eNB. The UE needs to monitor the PDCCH only at certain UE specific paging occasion, i.e.
only at specific subframe in specific radio frame. At other times the UE may go for sleep mode to
save battery power.

Let us discuss how can UE find out paging frame and paging occasion for its paging message.

Paging Frame (PF): One Paging Frame (PF) is one Radio Frame in which the UE is looking for
paging message. One Paging Frame may contain one or multiple Paging Occasion(s).

Paging Occasion (PO): It is a subframe where there may be P-RNTI transmitted on PDCCH
addressing the paging message for the UE. There is always only one paging occasion for each UE in a
DRX cycle.

To find out PF (Paging Frame) and PO (Paging Occasion), the UE uses DRX parameters broadcasted
in the system information SIB2.
Significance of DRX Parameters:

DRX cycle: Indicates the number of radio frames in the paging cycle. Larger the value of this
parameter, lesser the UE battery power consumption. Whereas smaller the value of DRX cycle,
increase in the UE battery power consumption. DRX cycle may be cell specific or UE specific
parameter. DRX cycle is cell specific when it is configured at eNB and broadcasts to all UEs in a cell
in system information SIB2 as default paging cycle. DRX cycle is UE specific when MME may
configure DRX cycle and provide it to UE in NAS signaling as’ UE specific DRX cycle’ and to eNB
in PAGING S1 AP message as ‘Paging DRX’ for MME initiated paging message. Both UE and eNB
use minimum of default and dedicated DRX cycle.

An UE with DRX cycle of 128 radio frames needs to wake up every 1.28 sec and look for paging
message.

nB: This cell specific parameter indicates number of Paging occasions in a cell specific DRX cycle.
Configuration of nB value depends on paging capacity required in a cell. The larger the value of nB is
configured to, the larger the paging capacity. Likewise, the smaller the paging capacity when the
value of nB is smaller.

With this we can infer that a maximum of 1 radio frame is available per DRX cycle of the UE for
paging. Also, the minimum of 1 sub frame per radio frame (nB ≤ 1T) and maximum of 4 sub frames
(for nB=4T) are available for paging message for the UE.

Calculation of Paging Frame Number

Both the eNB and the UE calculate the Paging Frame number by the following relation:
  

The eNB will receive UE_ID from the MME in PAGING S1AP message as “UE Identity Index
Value” for MME initiated paging message.

The UE will find out the UE_ID by using the relation IMSI mod 1024 (Ref: TS 36.304 section 7.1)

The IMSI is given as sequence of digits of type Integer (0..9). The IMSI shall, in the formulae above,
be interpreted as a decimal integer number, where the first digit given in the sequence represents the
highest order digit.

For example: IMSI = 12 (digit1=1, digit2=2). In the calculations, this shall be interpreted as the
decimal integer "12", not "1x16+2 = 18".

Calculation of Paging Occasion (Paging Sub frame Number in a Paging Frame)

Paging Occasion for the UE is obtained from the predefined subframe pattern (36.304 section 7.2,
shown below for quick reference) based on the value of Ns, i_s. For example, if Ns=2 and i_s = 1, UE
interprets that sub frames 4 and 9 are configured for paging in the cell and subframe 9 is paging
occasion for it.

Subframe Patterns (Ref: TS 36.304 section 7.2)

FDD

TDD (All DL/UL Patterns)

UEs in idle mode monitor the PDCCH channel for P-RNTI used for paging to receive paging message
from eNB. The UE needs to monitor the PDCCH only at certain UE specific paging occasion, i.e.
only at specific subframe in specific radio frame. At other times, the UE may go to sleep mode to save
battery power. The DRX parameters and the significance of each DRX parameter is discussed in first
part of this deliberation. One of the DRX parameters ‘nB’ is the deciding factor for paging capacity in
a cell. Let us discuss different scenarios by considering different values of nB in this deliberation and
understand impact of this parameter on paging capacity.

Scenario 1: nB = T/2

Observations:

 The UE needs to wake up every 32 ms and look for paging message (UE’s DRX cycle of 32
radio frames)
 One out of every 2 radio frames (nB=16) available for paging within UE’ s DRX cycle. That
is total of 16 paging frames are configured in the cell
 One paging subframe (Ns=1) per paging frame available for paging
 According to the pre-defined pattern for FDD, for Ns=1 ,i_s=0, sub frame 9 is used for paging
in a paging frame
 Total of 16 paging occasions (nB=16) per UE’s DRX cycle across which UEs are distributed
based on the UE identity.
  

Scenario 2: nB = 2T

Observation:

 Every radio frame is Paging Frame

  2 sub frames per PF (Ns = 2) are used for paging within UE’s DRX cycle
 According to the pre-defined pattern for FDD, for Ns=2 and i_s=0, sub frames 0 and 4 are
used for paging in a paging frame and Paging opportunity for the UE in discussion, is
subframe 4
 32 Paging Frames and 64 Paging occasions (nB=64) are available within UE’s DRX cycle
across which UEs are distributed based on the UE identity.

Scenario 3: nB = 4T
Observations:

1. Every radio frame is Paging Frame

2. 4 sub frames per PF (Ns = 4) are used for paging within UE’s DRX cycle
3. According to the pre-defined pattern for FDD for Ns=4, sub frames 0,4,5 and 9 used for
paging in a paging frame and paging opportunity for the UE in discussion is sub frame 5
4. 32 Paging Frames and 128 Paging occasions(nB=128) are available within UE’s DRX cycle
across which UEs are distributed based on the UE identity