There are two main techniques for dividing forward and reverse communication channels on the same physical transmission medium:
- Time Division Duplex (TDD)
- Frequency Division Duplex (FDD)
Time Division Duplex (TDD)
Using the TDD method, a single frequency channel is assigned to both the transmitter and the receiver. Both the uplink (UL) and downlink (DL) traffic use the same frequency f0 but at different times.
In effect, TDD divides the data stream into frames and, within each frame, assigns different time slots to the forward and reverse transmissions. This allows both types of transmissions to share the same transmission medium (i.e., the same radio frequency), while using only the part of the bandwidth required by each type of traffic (
Several inferences can be drawn from this description:
- Since the TDD scheme can allocate dynamically the amount of time slots assigned to each direction — transmit and receive– an operator can define the percentage of UL versus DL traffic. This is especially important for Internet-type traffic (the ratio for UL/DL is no longer constrained to a fixed 50/50).
- A guard band is not required to separate the UL and DL channels, because they both use the same frequency – hence, there is no loss in spectrum. A guard period, though, is necessary for synchronization purposes and to accommodate the turnaround time and the round trip delay whenever switching transmission from DL to UL, and vice versa.
- Because the UL/DL allocation is dynamic, there is very little waste of spectrum for asymmetric operations, i.e., last-mile applications, where typically the UL traffic is a fraction of the DL traffic. (Some spectrum is still lost for the guard periods, but this is negligible compared to the total length of data in a time slot).
Frequency Division Duplex (FDD)
Using the FDD method, a distinct frequency channel is assigned to both the transmitter and the receiver. At any particular instant in time, uplink (UL) traffic uses a frequency that is different from the frequency used by the downlink (DL) traffic.
The Base Station Unit (BSU) may receive uplink traffic while it simultaneously transmits on the downlink. To avoid the high design costs that FDD imposes on Subscriber Units (SU), WiMAX SUs use a hybrid duplex method called HFDD (half-duplex FDD). HFDD is very similar to TDD. An HFDD device transmits and receives at different times like a TDD device. The difference is that it also uses different frequencies for transmit and receive to communicate with an FDD Base Station. As a result, HFDD Subscriber Units offer only half the throughput capacity of a full duplex FDD Subscriber Unit.
FDD is typically used in applications that require an equal up- and downlink bandwidth, as all TDM voice applications do. Therefore, regulatory agencies grant up- and downlink channels of equal capacity for FDD-based systems.
Several inferences can be drawn:
- Due to the symmetric nature of FDD transmission channels, and the FDD legacy as duplex method of choice for TDM voice applications, FDD transmission channels are always of equal size (50% for UL and 50% for DL). In applications such as Internet access, which can be very asymmetric in nature, a large percentage of the available UL bandwidth remains unused and is, therefore, wasted.
- A guard band about two times the size of the UL or DL channel is required to separate the UL and DL channels. This amounts to an additional 50% loss in spectrum.
- Once the channel bandwidth is granted by the regulator, the UL/DL allocation cannot be changed. This leads to unused spectrum for asymmetric operations, i.e., for last-mile applications, where typically the UL traffic is a fraction of the DL traffic.
WiMAX and the Last Mile
Based on the characteristics of the two duplex methods (TDD and FDD), as analyzed in the previous section, it is clear that the TDD method provides the best network equipment for last-mile access service. However, this does not imply that FDD should be discarded completely. FDD has its place in the network, making a much stronger case for symmetric-type traffic, such as that found in a cellular/ISP backhaul. In this case, TDD tends to waste bandwidth during switchover from transmit to receive, has greater inherent latency, and may require more power-hungry circuitry.
- For last-mile applications, WiMAX uses a lower frequency range – 2 GHz to 11 GHz (similar to Wi-Fi). This allows Non Line of Sight (NLOS) connections, because lower-wavelength transmissions are not as easily disrupted by physical obstructions. They are better able to diffract, or bend, around obstacles. TDD is better here because of the asymmetry of traffic.
- For cellular/ISP backhaul applications, a line-of-sight (LOS) connection is stronger and more stable, so it is able to send much more data with fewer errors. LOS transmissions use higher frequencies, with ranges reaching a possible 66 GHz. At higher frequencies, there is less interference and much more bandwidth. FDD is better in this application because of the symmetry of traffic.
The choice of TDD or FDD may be dictated by the regulatory agency. Each country and/or regulatory body can specify if one or more duplex methods are permissible in a given frequency band. The operator must first determine if the frequency band requires a specific duplex method or accepts either.
The determining factor for the operator making a duplex choice should be determined by customers (audience) and the applications that customers expect of the service. With the heavy desire for WiMAX Forum Certified equipment to enable broader deployment of last-mile access systems, there are excellent reasons to select the TDD method for this application.
A comparison of technical issues regarding FDD versus TDD is shown in table 1.
TDD is ideally suited to the transport of asymmetric traffic, as is typical with Internet access, and it allows service providers to define accordingly the percentage of bandwidth allocated to each direction. In addition, TDD makes more efficient use of spectrum, allowing network operators to achieve greater returns on their investments in infrastructure. As for FDD, it is the scheme of choice when traffic is symmetric, as in carrier backhaul and enterprise data transfer applications.
Therefore, in the near term – as WiMAX is adopted for last-mile applications – network operators are advised to choose TDD in order to achieve the flexibility required for managing divergent traffic patterns.
Table 1: A Comparison of FDD versus TDD
Multiplexing Method in Advantage
|FDD requires a guard band to separate the DL and UL channels. In MMDS, two RF channels are used to separate the UL and DL channels, which amount to a substantial loss in spectrum.||No guard bands are required.|
|No guard time is required at the end of DL transmission. However, guard time is required at the end of UL transmission because typically the SUs are HFDD units that need to turn around from Tx to Rx to receive the new BSU schedule information for the next downlink.||Guard time is required between Tx and Rx and vice versa. The guard time is equal to a unitÂ’s turn around time plus the round trip delay. A unitÂ’s turn around time is in the order of 50 us. The round trip delay is in the order of 66 us. Thus the round trip delay can absorb the transmitterÂ’s turn around time whenever the direction of traffic switches. The loss in throughput due to guard time for a 5 ms frame is about 2%.|
|Frequency Plan and Reuse|
|The adjacent channel interference is much lower than in a TDD scheme.||Frequency planning is required only for one channel. If all TDD-based systems are synchronized to GPS, using the same frame size and DL/UL partitioning can mitigate interference.|
|FDD requires one transmitter and a separate receiver. Further a diplexer and shields are required to isolate the DL and UL.||As the transmitter and receiver use the same filters, mixers etc the cost of a TDD scheme is substantially less than an FDD scheme.|
Table 1: A Comparison of FDD versus TDD (Cont’d)
Multiplexing Method in Advantage
|Dynamic Bandwidth Allocation|
|Once the channel bandwidth is granted by the regulator the UL/DL allocation cannot be modified. This leads to unused spectrum for asymmetric operations such as Internet traffic.||Where cell interference is not a problem, adaptive UL/DL allocation allows dynamic bandwidth allocation for UL and DL traffic. This is especially important for Internet traffic.|
|The average FDD latency in a PMP system is 1 frame and the best case latency is about 0.5 frame.||The average TDD latency in a PMP system is 2 frames and the best case latency is about 1 frame.|
|Adaptive Antenna System/ Multiple Input-Multiple Output (AAS/ MIMO) advantages|
|For closed loop beam forming, FDD requires the SU to provide the channel response for the DL direction. This increases the latency and reduces the performance of the beam former.||TDD allows the BSU to estimate the DL channel as both DL and UL are operating on the same frequency. The performance of the beam former is therefore better.|