5G URLLC - Ultra Reliable and Low Latency? (part II)
My previous article on 5G URLLC attempted to provide high-level insight into the mechanisms that were responsible for lowering the user plane latency of 5G URLLC communications. This article attempts to complete the discussion by uncovering how 5G URLLC plans to make URLLC communications more reliable.
To make wireless communications reliable the signal power of the desired signal needs to be maximized relative to the noise and interference power that occupies the same passband as the desired signal. This ratio between the signal power and the interference + noise power is called "SINR"; the SINR formula is as follows:
SINR = Signal Power / (Interference + Noise Power)
To maximize SINR the signal power can be increased or the interference and noise power can be lowered or some combination of the two approaches can be employed. The power of the desired signal will experience attenuation at the receiver location as a normal result of fading. The interference power comes from usage of the same frequency at other cell sites and this is a normal cost associated with any channel re-use scheme. The discussion will start with techniques that maximize signal power, followed by techniques that minimize interference power.
Diversity techniques - this is a signal power improvement technique where a single stream transmission uses multiple transmit antennas at the base station and is received by multiple receive antennas at the UE (3GPP standards call the cell phone a "UE") in order to mitigate the effect of multipath fading. This technique requires sufficient spacing of the antenna elements so as to create independent (uncorrelated) copies of the transmit signal.
Receive diversity is based on the fact that using multiple receive antennas allows the various antennas to receive different versions of the same signal. The various versions of the signal will have taken different paths to the receive antennas and experience different fading conditions with the expectation that not all paths will experience a deep fade at the same time. Using spatial diversity to improve channel reliability is a technique that has been around for many years; when researching the topic I came across examples of radios with tubes (i.e. non-solid state) making use of spatial diversity techniques!
Macroscopic diversity - this is a signal power improvement technique implemented in the downlink-only (i.e. base station to UE) where multiple base stations transmit the same signal to a target UE. The received signal is combined at the UE. The approach has several benefits including combating shadowing effects (impact of obstacles on signal propagation), providing signal diversity, signal redundancy and increasing the total receive power of the desired signal. The cost of this approach is important to note since it consumes radio resources at all the sites that are transmitting the single user signal. Macroscopic diversity is also covered below in the "Coordinated Multipoint" section where it is used to lower interference power.
Improving control channel reliability - "PDCCH Aggregation" -- the physical downlink control channel (PDCCH) in LTE and 5G carries specific info for each user including defining where/when radio resources are dedicated to each UE, access coordination of shared channels, power control commands, scheduling info on where users can find system information and "pages" for incoming calls/data. It is a critically important channel in both LTE and 5G and important that this channel be reliably decoded by all users given its' central role in the basic performance of the LTE and 5G networks. 5G addresses this need for higher control channel reliability in URLLC use cases by supporting one or both of the following improvements: 1) the use of lower coding rates with more forward error correction, and 2) use of lower order modulation schemes. Item (1) offers coding gain and item (2) allows for greater signal reliability in lower signal to noise ratio environments. The 5G folks allowed for an "aggregation level of 16" which means that 16 Control Channel Elements (CCEs) can now be consecutively aggregated versus the previous value of 8 in LTE.
What does this mean? A CCE is a collection of "resource element groups" which is related to the minimum resource unit in both LTE and 5GNR; a minimum resource unit is one OFDM symbol in the time domain and one subcarrier in the frequency domain. Simply stated, the larger the aggregation level, the larger the PDCCH "bucket". A larger PDCCH bucket allows for the use of a lower coding rate and lower order modulation since a bigger bucket gives more room to fit the transmission of the same control channel information, increasing the reliability of the control channel. Of course the penalty for doing this is lower spectral efficiency.
Improving control channel reliability -- there are a collection of other techniques in 5G that will provide higher reliability of control channel signaling. Repetition of UE scheduling information is one technique. By having the base station repeat scheduling information, the UE can recover from missing an assignment since the remaining assignment information is automatically presented and available for the UE to use.
Another channel reliability improvement technique is to make the "channel quality indicator" information more reliable through the use of a lower coding rate. What is the "channel quality indicator" (CQI) and why does it need to be more reliable? The CQI is part of a feedback mechanism used by the UE to tell the base station how reliable the channel is and therefore what modulation and coding scheme the base station should use to send data to the UE. A wireless channel experiences several impairments at any point in time and providing a measurement of channel quality will give the base station the information it needs in order to send the most data that it can safely (reliably) send to the UE during the period of time for which the CQI measurement is considered valid.
If the base station decodes the UE's CQI incorrectly and ends up using a higher order modulation and coding scheme than it should, the UE will likely be unable to decode the transport block and a re-transmission will need to occur. Increasing the reliability of the "Channel Quality Indicator" report so that the base station doesn't mistakenly use a higher order modulation and coding scheme is a meaningful reliability and latency improvement for URLLC use cases.
Improving radio uplink reliability -- the LTE and 5GNR radio links are reliable but at the expense of latency. The protocol used to provide this reliability is called "hybrid ARQ" or "hybrid automatic repeat request". The process employed by hybrid-ARQ to guarantee delivery does consume time since it involves a series of messages between UE and base station requesting re-transmission of errored or missing data. The 3GPP folks have come up with a scheme to proactively repeat transmitted data rather than waiting for the hybrid-ARQ process to kick-in. By adopting a repetition factor of "K", a UE will be granted access for K transmissions; if no acknowledgment is received from the network after the initial UE transmission, the UE automatically re-sends the same packet again. This presumptive process repeats until the K repetitions are exhausted or an acknowledgment is received by the UE.
Now the discussion moves on to minimizing the SINR denominator -- lowering Interference power.
Mitigating co-channel interference - ICIC/eICIC -- by taking steps to lower the denominator in the SINR ratio discussed above, reliability will be enhanced. LTE defined a capability called "Inter-Cell Interference Coordination" (ICIC) in 3GPP Release 8 which allowed adjacent cells to communicate so that they were aware of what interference they were causing to other users not served by that cell. By coordinating time (eICIC, Rel 10), power and frequency assignment to their respective users, the participating base stations could lower the co-channel interference power that they were causing to eachother's users.
One drawback of ICIC/eICIC is that the interference information shared between cells was not very dynamic and therefore wouldn't help moment-by-moment experiences of users that are in a fast moving vehicle or other rapidly changing RF environments. These shortcomings paved the way for a more advanced interference mitigating technology known as "coordinated multi-point" or CoMP.
CoMP was part of 3GPP Release 11 and makes up for ICIC/eICIC shortfalls and then some. CoMP makes use of not only time and frequency to address interference but also uses the spatial domain and enhances spectral efficiency, something that ICIC/eICIC were not capable of. The spatial capabilities are provided by beamsteering. To see the highest gains requires sharing of UE data across all sites in what CoMP calls a "cooperation set" or "cooperation area". CoMP has a few different deployment options and is a very involved topic that could not possibly be covered in this article.
The most valuable modes of Coordinated Multipoint require high speed connectivity between sites in a "cooperation area" and this creates certain implementation difficulties due to the challenge in building the needed backhaul network. 3GPP Release 16 is discussing a feature called "Integrated Access and Backhaul" where cell site backhaul/fronthaul could be transported on the 5G RAN in segregated spectrum but in the same band (all above 6GHz). This consideration of backhaul by 3GPP is very welcome in my opinion, given the difficulty in finding fiber everywhere coverage is needed. With the 5G network edge soon to be carrying gigabits+ per second, backhaul/fronthaul planning can no longer be an afterthought.
Another option to resolve this need for communication between sites in a cooperation area is to deploy these sites in a "cloud RAN" type of deployment where the baseband processing is centralized and the radios are deployed at the sites, typically at the top of the tower as a remote radiohead. By collocating the control elements of the cooperating base stations, the transport problem becomes a simple LAN cabling effort with no WAN complexity whatsoever. The cloud RAN approach still requires transport of the fronthaul so the transport problem continues to exist although in a different format. The network capacity improvements offered by the most capable Coordinated Multipoint methodology (Joint Transmission) should easily outweigh the transport cost of building out cloud-RAN; the results of simulations are available and the SINR improvement is dramatic, meaning more users will operate at higher density modulation and coding schemes (i.e. more users experience faster service).
In summary, 5G URLLC will achieve high reliability by maximizing the signal to interference+noise ratio. While the signal improvement methods were relatively straight forward, the Interference mitigation is a complicated endeavor but a critical part of maximizing the value of the most constrained operator resource -- spectrum.
As I was researching coordinated multipoint and the tremendous value it brings in terms of maximizing the value of spectrum, I began to wonder how interference will be mitigated in the CBRS GAA deployments? The concern that I see is for those operators that attempt to build out wide area networks where co-channel interference will exist between multiple unrelated players that do not have a means to deploy CoMP with one another (I have assumed that there is no standard for the exchange of CoMP information across vendors). Will these operators have the sophistication/staff to handle this type of complexity? And if the SAS dynamically changes operator frequencies, how do you coordinate CoMP communications given your neighbors could change at any time? But then again, with the only spectrum cost being the SAS subscription fees to use the spectrum, stranding CBRS GAA capacity may not be a big concern.