23.4 Half-Duplex System Design Aspects
23.4.1 Accommodation of Transmit–Receive Switching
For TDD systems, switching between transmit and receive functions occurs on the transition from uplink to downlink (for the UE) and on the transition from downlink to uplink (for the eNodeB). For half-duplex FDD systems, switching only occurs at the UE, as the eNodeB is assumed to be full-duplex.
In order to preserve the frequency-domain orthogonality of the LTE uplink multiple access scheme, propagation delays between an eNodeB and the UEs under its control are compensated by means of timing advance as explained in Section 18.2.
At a half-duplex UE, the timing-advanced uplink transmission cannot be allowed to overlap with reception of any preceding downlink. For TDD, to prevent the overlap, a transmission gap or ‘guard period’ between transmission and reception at the eNodeB is created (TG1) to accommodate the greatest possible timing advance and any required switching delay (including power amplifier ramp-up or ramp-down to avoid excessive wideband emissions). A further guard period (TG2) is also required at the TDD eNodeB transition between uplink and downlink to cater for switching and power ramping delays only (this being independent of the propagation delay or timing advance). These are illustrated in Figure 23.8.
Four switching times are therefore of relevance in the case of TDD operation. These correspond to the transmit-to-receive and receive-to-transmit delays at the UE (denoted
T
TUE RxTx
eNodeB UE
p
TUE TxRx
TUE RxTx
High Delay UE (Tp = Tp-max)
TUE TxRx
time TG1
TG1
TG2
TG2 T
Tp-max
p-max p-max
UE
eNodeB Tp-max p-max T
Low Delay UE (
(downlink) p-max p-max
T ≈0)
(uplink) (downlink)
(uplink)
(uplink) (downlink)
(downlink) (uplink) T T
Figure 23.8: TDD signal timings in the presence of uplink timing advance.
as time intervals TUETxRx and TUERxTx respectively) and likewise at the eNodeB (denoted TeNBTxRx and TeNBRxTx). Figure 23.8 depicts two cases corresponding to the two extremes of propagation delayTPwithin a TDD cell (TP=0 for a UE physically close to the eNodeB and TP=TP_max for a UE at the border of a cell, where TP_max=dmax/c corresponds to the maximum one-way propagation delay supported by the cell, occurring at distancedmax). Note that the switching delays are exaggerated for diagrammatical clarity and that UE switching delays are assumed to be longer than those at the eNodeB.
It is apparent from Figure 23.8 that the time available at the UE for downlink to uplink transition is a function of the propagation delayTP(most stringent for the case of high delay) whereas the time available at the eNodeB for the same transition is constant and equal toTG1:
TG1=2TP_max+TUERxTx (23.9)
The time intervalTG2at the eNodeB is independent of the propagation delay. To support the case for whichTP→0 (i.e. a UE close to the eNodeB),TG2needs to be dimensioned such that
TG2=max(TUETxRx,TeNBRxTx) (23.10) In the case of HD-FDD,TeNBRxTx=0, soTG2is determined only by the timeTUETxRx. In order to support the case of lowTP, the (full duplex) eNodeB must still allow sufficient time for this UE switching delay if the uplink and downlink subframes surrounding the switching point are both active for a particular user. Hence in practice the uplink frame timing at the eNodeB should be advanced for the whole cell by an amountTUETxRxrelative to the downlink frame timing at the eNodeB even for a full-duplex eNodeB if it supports HD-FDD UEs in the cell.
Full duplex FDD UEs communicating with the same eNodeB will likewise need to have their timing advanced to maintain uplink orthogonality with the HD-FDD UEs.
It is important to note thatdmaxmay be significantly larger than the notional cell radiusr0 (i.e. half the ISD), due to propagation effects such as shadow fading. This effect is shown for
one example of a tri-sectored deployment with frequency reuse factor 1 in Figure 23.9 (the shadow fading is assumed to be log-normal with standard deviationσ). It can be observed that in order to accommodate, for example, at least 98% of UE locations, the guard period should be dimensioned in accordance withdmax≥γr0whereγis a factor between approximately 1.5 and 3 depending on the degree of shadow fading.
0 0.5 1 1.5 2 2.5 3 3.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Ratio of UE-eNodeB separation to nominal cell radius (γ = d / r0 )
Cumulative Distribution (CDF)
Standard Deviation of Log-Normal Shadowing
σ = 6dB σ = 8dB σ = 10dB σ = 12dB
Figure 23.9: Cumulative distribution of serving cell propagation delays relative to nominal cell radius.
Overall, the total guard timeTG at a TDD eNodeB per uplink–downlink cycle is equal to the sum ofTG1andTG2, as given by Equation (23.11). In the LTE specifications this is represented by a single guard period amalgamating both parts (with the uplink subframe timing advanced by an amountTG2at the eNodeB with respect to the downlink timing). This is only a matter of representation, however, and the end result is essentially identical to the presence of the two separate guard periods:
TG=TG1+TG2=2TP_max+TUERxTx+max(TUETxRx,TeNBRxTx) (23.11) By assuming reasonable values for the UE and eNodeB switching times (typically of the order of 10 to 20μs) the length of this amalgamated guard period can be dimensioned (in multiples of the OFDM4 symbol duration) for a particular deployment. Thus although
4Orthogonal Frequency Division Multiplexing.
guard periods represent an undesirable overhead for TDD, by allowing for a flexible and configurable guard period duration, systems can be tailored to the topology of the deployment whilst minimizing the spectral efficiency loss.
The LTE specifications [7] support a set of guard period durations ranging (non- contiguously) from 1 to 10 OFDM symbols for the normal CP (or from 1 to 8 OFDM symbols for the extended CP). A duration of 1 OFDM symbol should be sufficient for many of the anticipated cellular deployments of LTE (up to around 2 km nominal cell radius forγ=2), whereas at the other end of the scale, guard period durations of the order of 700μs support one-way propagation-path delays of the order of 100 km.
The guard period in LTE TDD is located within a mixed uplink/downlink subframe (known as a ‘special subframe’) as shown in Figure 6.2 and further discussed in Section 23.4.2.