Recently, 3GPP has shared the Non-StandAlone 5G New Radio (NR) details in Rel15. Naturally, everyone wants to know what changes and enhancements it holds to achieve the high throughput and low latency targets. So, I thought that I would write a short article on the new 5G specifications explaining the additions in the structure and how it provides gains over the current 4G specifications.
5G Major Targets : For 5G NR, there many targets that it needs to achieve but some of the most important targets are higher throughput and low latency.
URLLC (Ultra Reliable Low Latency Communication) : URLLC is one of the most important target for 5G NR. This target requires 5G to have an E2E latency of less than 1ms which sounds pretty tough to achieve considering that the current 4G or 4.5G LTE has a latency of around 25 to 30ms and it can even increase to above 100ms in case of congestion. In order to reduce the latency, the 5G NR introduces a scale-able or dynamic structure where the symbol length can change along with the slot length. Let’s have a look at this structure.
A shorter symbol duration causes an increase in subcarrier spacing Δf. The relation between symbol duration and subcarrier spacing is given by:
Symbol duration = 1 / Δf (plus a CP is added to it)
This shows that for a shorter symbol duration, we will need more bandwidth for the same number of carriers. As seen in the table, the usual 15 KHz that is used by 4G gives 14 symbols per slot with a symbol length of around 71 microseconds while the symbol length of 4.4 microseconds can be achieved with a subcarrier spacing of 240 KHz. This will provide around 16 slots in a single subframe.
Asynchronous HARQ : Since we have a flexibility of using a much smaller symbol duration so 5G NR also introduces a HARQ PDSCH Feedback Timing Indicator. This field is introduced in DCI Format 1_0 and 1_1 and tells the UE where it needs to send the HARQ feedback for the downlink data that it received. UE sends the HARQ retransmission based on this instruction over the PUCCH. Since, we have a slot length of only 0.0625 ms so if UE sends the ACK in slot N+4, the overall time would be around 0.0625*4 which will be around 0.25ms. However, since the HARQ feedback is asynchronous which means that the eNB informs the UE when and where to send the HARQ feedback, it will be mostly even lesser than this. So, this structure ensures a much smaller latency over the air which ensures that the URLLC target is achieved.
eMBB (Enhanced Mobile BroadBand) : eMBB is the throughput oriented part of the 5G NR. It is stated that 5G NR requires to achieve throughput targets above 1 Gbps initially and upto 20 Gbps in later phase and this also needs a lot of work.
Bandwidth : The current 5G specifications support 256 QAM but this was also supported by LTE, however, there aren’t many UEs in the market that support it commercially. Secondly, 5G supports a single carrier of upto 275 RBs while LTE only supported a single carrier of 100 RBs. This will give a big boost as this is about 2.75 times more bandwidth. Moreover, if Carrier Aggregation is done with a couple of such carriers, it will provide a huge throughput gain. As of now, around 16 carriers can be aggregation in 5G NR and a total aggregated bandwidth can go upto 1 GHz. This is much higher than the usual 3CC CA seen in LTE with 60MHz or 5CC CA with 100Mhz.
Massive MIMO : Another important factor is the MIMO functionalities for 5G NR which has introduced an increase in antenna ports which will provide much higher throughput gains based on SU-MIMO, MU-MIMO and beamforming. There are around 8 orthogonal DMRS sequences defined which can support SU-MIMO.
Reference Signals : The Reference Signals in 5G NR are also different from LTE. In LTE, there are cell specific reference signals which are always enabled and cause most of the downlink interference. In 5G NR, such reference signals are removed and RS are only used when the UE has data allocation. Basically, they are more like UE specific RS and act like DMRS. There are other types of RS as well like phase tracking reference signals which are used to counter phase noise at higher frequencies. This kind of structure will reduce the downlink interference and consequently improve the channel quality resulting in better spectral efficiency and capacity.
A question may arise that if there are no cell specific reference signals then how the UEs will get the RSRP during initial access or reselection since they will not any allocation at that time. The answer lies in the fact that 5G NR uses a PSS/SSS/PBCH block structure. The PBCH has it’s own set of reference signals which will always be present in the PBCH so the UE while reading the PSS/SSS/PBCH block should be able to ascertain a RSRP value from the PBCH DMRS.