Part Five: Uplink Capacity
Uplink capacity is crucial to how many devices can be supported per base station. We agree with the cellular industry’s assessment that the numbers can be very high. 3GPP predicts as high as 40 devices per household. Cellular LPWA claims to be able to support these devices, but they seem to impose some rather unpleasant constraints on the data model. As the requirements laid out in the 3GPP document, TR45.820 : Cellular system support for ultra-low complexity and low throughput Internet of Things (CIOT), endpoint data is assumed to be buffered over long periods of time and sent infrequently to achieve sufficiently high capacity for LPWA. For a number of reasons, this greatly limits the usefulness of the cellular LPWA system.
Fundamental Cellular Issues
There are two fundamental issues when it comes to uplink capacity.
Issue 1: FDD Spectrum. A scheduled protocol in FDD spectrum (e.g. cellular LPWA) is the right approach for high data rate cellular links, but an unscheduled protocol in TDD spectrum (e.g. RPMA) is the best technology choice for small, unscheduled packets. Note that even though some LTE deployments are in TDD spectrum, there is additional work to take advantage of “channel reciprocity” which the current cellular LPWA approaches do not employ. Accurate power control is essential for good capacity. LTE leverages closed-loop power control, which works great for large blocks of data. For small data transactions, closed loop power control fails to work because there is not sufficient time for the feedback to take effect since the packet is too short for the loop to converge. Thus, the accuracy of the power control is very poor resulting in poor spectral efficiency.
Issue 2: Scheduling Overhead. Scheduling makes perfect sense for the large voice/data packet sizes. The scheduling overhead of initiating a large data or voice session is not much of an issue. However, when initiating a session for a message size that is on average 32 bytes (which is consistent with the assumed data model in TR45.820 : Cellular system support for ultra-low complexity and low throughput Internet of Things (CIOT)), the scheduling overhead becomes greater than the message itself. As stated in An Overview of 3GPP Enhancements on Machine to Machine Communications published in IEEE Communications Magazine:
- “After the random access procedure is successfully completed, the UE can establish radio resource control (RRC) connection with the eNB.”
- “Once the RRC connection is established, the UE blindly decodes the MPDCCH in the configured search space to obtain uplink and downlink data assignments.”
These overhead transactions consume both uplink (on the PUSCH channel) and downlink (on the MPDCCH channel) capacity that exceeds that of the small data transaction itself and must occur every time an endpoint communicates information (no matter how small).
The buffering of data (called transmit packet aggregation) required for Cellular LPWA to achieve sufficient capacity introduces latency that greatly reduces the utility of the system. This is where 3GPP employs some sleight of hand. They will claim to support the small payload transaction, and they will claim to support a large number of endpoints. The only problem: they will not support both small payload and high capacity at the same time.
- Exception Reports which from “are delivered in near real time, with a latency target of 10s.”
- Periodic Reports which from “the MAR periodic reporting traffic model is used in system level simulations for capacity analysis.”
All the 3GPP capacity simulations are relying upon bundling periodic reports into arbitrarily large transmissions in a method called transmit packet aggregation. This puts the rate of communication transactions within the realm that cellular technology can handle efficiently.
The problem with this based on our experience:
- Bursts of latency sensitive alarms must be handled reliably which stresses instantaneous capacity. The Smart Meter application is just one example where alarm information may be rapidly sent during a large outage.
- Real-time situational awareness is important to many applications. Data must be received in real time so that it’s actionable.
- The cellular billing structure will be onerous to the customers in order to dissuade applications from small, frequent, transactions.
How RPMA Solves the Issues
RPMA has been designed for tremendous uplink capacity for frequent small packet data models that are dominant in the LPWA space.
Issue 1: RPMA Uses a TDD Approach. Since RPMA was designed precisely for these small data transactions, RPMA uses a Time Division Duplex (TDD) approach. This removes the need for closed loop power control because the uplink channel is known precisely by measuring the downlink channel due to a fundamental phenomenon is known as channel reciprocity. Channel reciprocity is only valid in TDD approaches where the uplink frequency is the same as the downlink frequency.
Issue 2: RPMA Was Designed Without the Need of Scheduling. RPMA does not squander capacity communicating in scheduling overhead. This helps with capacity and power consumption.
As a result of solving both of these issues, RPMA does not need to distinguish between Exception Reports and Periodic Reports. RPMA handles these transactions the exact same way. In this sense, all RPMA transactions are low-latency.
Another aspect to point out is that based on TR45.820 : Cellular system support for ultra-low complexity and low throughput Internet of Things (CIOT), all 3GPP LPWA simulations use the TU-1Hz which, in our experience, is unrealistically benign compared to the real world. It is not fully known the impact of a more realistic channel model in regards to uplink capacity, but it’s very likely to dramatically reduce capacity in the field.
Quantifying this: as far as RPMA capacity, we have enough in-field expertise (and the corroborating analysis) to convincingly show that we can support 500k devices per sector even given a factor of several safety margins for peak traffic. By contrast, the top-line number of the contribution in R1-156624: Capacity evaluation for in-band operation (which are the typical numbers mirrored in lots of other contributions) is 50k messages per hour per 200 kHz slice of bandwidth which corresponds to about 25k supported devices at the 3GPP data model (which we use as well in our 500k number). But back off by a factor of 2 for headroom, they’re at 12k. And that’s assuming the very benign channel assumption as well as transmits packet aggregation. So there is lots of room for that 12k number to crater further.
This post is a part of the series Is the cellular standard roadmap (3GPP/GSMA) the answer to Low Power Wide Area (LPWA) Connectivity? Click a link below to learn more, or download our free eBook, How RPMA Works: The Making of RPMA.
- Part 1: Introduction
- Part 2: Cellular LPWA Availability
- Part 3: 3GPP/GSMA is NOT Providing a Graceful Evolution Path for Machines
- Part 4:. Cellular LPWA Complexity
- Part 5: Cellular LPWA Performance Issue 1: Uplink Capacity
- Part 6: Cellular LPWA Performance Issue 2: Downlink Capacity
- Part 7: Cellular LPWA Performance Issue 3: Firmware Download
- Part 8: Cellular LPWA Performance Issue 4: Robustness
- Part 9: Cellular LPWA Performance Issue 5: Power Consumption