Why Special Subframe is needed in LTE

As the single frequency block is shared in time domain between UL and DL the transmission in TDD is not continuous. All UL transmission need to be on hold while any downlink resource it is used and the other way around. 
Switching between transmission directions has a small hardware delay (for both UE and NodeB) and needs to be compensated. To control the switching between the UL and DL a guard period GP is allocated which compensates for the maximum propagation delay of interfering components.


Within a radio frame, the transmission direction changes several times between downlink and uplink. 

In special subframe DL to UL switching happens.
Special subframe includes DL,UL and a guard period.



Due to the different signal transit times between the eNodeB and the various mobile stations, a timing advance mechanism involving a time gap called “guard period” is needed when the transmission direction switches from downlink to uplink. However, no guard period is needed when the transmission direction switches from uplink to downlink. 


In the uplink,as shown in above figure the greater the distance between the eNodeB and the mobile station, the earlier the mobile station must start transmitting. This helps ensure that all signals reach the eNodeB in a frame-synchronous manner. When switching from downlink to uplink, a guard period is inserted between the DwPTS and UpPTS field in each special subframe. The duration of the guard period is configured by the network, based on the cell size. The maximum possible guard period length of ten OFDM symbols allows cell sizes with a radius of 100 km. 


When switching from uplink to downlink there is no need for a guard period, since the uplink signals all arrive at the eNodeB in a frame-synchronous fashion - thanks to the timing advance mechanism - and the downlink data is also transmitted in the form of a frame-synchronous OFDMA signal.


DwPTS : Downlink Pilot Time Slot
UpPTS :  Uplink Pilot Time Slot

Why Special Subframe is required in LTE?

In LTE Frame Type 2 (TD-LTE) there is a special sub-frame when switching from DL to UL but there is no special sub-frame or gap when switching from UL to DL. 

Different TDD modes

To understand this, it is important to know why a transmission gap is required when switching from DL to UL. The special sub-frame is made up of DwPTS, GP and UpPTS and all of these have configurable lengths while the sum of the lengths has to be 1 ms i.e. the length of the sub-frame. Now consider, the format 1, where the GP (guard period or TTG in WiMAX) is 4 symbols long which equates to 285 us approx. Consider a UE-A at a distance of 10 km from the eNB and UE-B at 50 km from the eNB. The time it takes the RF signals to reach the UE-A and UE-B will be

Time for UE-A = distance/velocity of light = 10000/3x10^8 = 33.3 us

Time for UE-B = distance/velocity of light = 45000/3x10^8 = 150 us

This means that after the eNB has transmitted the last symbol of DL data and it starts the GP, the last symbol will be received at UE-A after 33.3 us and at UE-B after 150 us. Now, every UE takes a small amount of time to switch from Rx to Tx mode and lets assume this switching period to be 50 us (it should be lesser for LTE UEs but this is just an assumption). So, the UE-A will start its switching period and will start transmitting after 33.5 + 50 = 83.5 us and it will take another 33.5 us to reach the eNB. This makes the total Round Trip Time for UE-A to be equal to 33.5 +50 + 33.5 = 117 us. Now we know that the GP at eNB is set 285 us so that means that the UE-A will be able to transmit the UL data within the GP. In actual practice, all the UEs know their Timing Advance from the eNB so the UE-A would wait that much amount before transmitting so that the UL data reaches the eNB at exactly the end of GP.

However, lets do the same analysis for UE-B. The total round trip time for UE-B would be 150 + 50 + 150 = 350 us which is greater than the GP (285 us) so, the UE-B would not be able to reach the first uplink symbol. Because of this, the GP is supposed to determine the maximum cell radius for a TDD system.

If there hadn't been a GAP or TTG (as in WiMAX) between the DL and UL transmission, these Over-The-Air timing delays and the switching period could not be compensated so we need to add a transmission gap when switching from DL to UL.

Now, lets consider the UL to DL switching. We will only consider UE-A for this example as UE-B has been proven beyond the cell range. The UE-A will transmit the last UL symbol and then start switching from Tx to Rx mode. The last UL symbol will reach the eNB after 33.5 us and the eNB would switch to Tx after receiving the last UL symbol. It will transmit the next DL symbol which will reach the UE-A after another 33.5 us and thus the UE-A will have a total of 33.5 + 33.5 = 67 us of time to switch from Tx to Rx mode without any presence of any RTG. So, while switching from UL to DL, a RTG or GP is not really required as the system already gets a virtual GP due to OTA delays.

There can be a query about the UEs that are very close to the eNB as they would have a very small Over-The-Air delay so it might not get enough time to switch to Rx mode. There can be two possible solutions for that

- In LTE, there is a 1 ms TTI so if the UE is too close to the eNB that it would not be able to switch to Rx mode in time, the eNB can allocate the DL resources in the next DL sub-frame so the UE will have 1 ms to make the switch. In WiMAX, this would not have been possible as it has 5 ms TTI and in case of No RTG, the UEs closer to WiMAX BTS would have to be scehduled in the next frame adding another 5 ms to the latency plane.

- Secondly, the UE switching from Rx to Tx and vice versa should now be extensively reduced. The 50 us limitation was there in beceem chipsets around 4 years back while I think that beceem and intel had made chipsets with switching periods of less than 20 us last year. SO, LTE UEs should have a much lower switching times.