LTE latency reductions: preparing for 5G

Now it’s time for a major leap in LTE latency reduction. Read on to learn about the new techniques that are the result of our leading research in this field.

As technology becomes more advanced so do the associated use cases. We have identified several that require low latency: for example, connected industrial plants or vehicular communications. Many other applications can also benefit from lower latency, including online gaming and file downloading. Going forward, we may need to address latency for applications like remote online surgery and tactile feedback.
There’s some more background in the blog post “LTE latency improvement gains” from late 2014.

With advances in LTE technology and use cases, we are ready for the next step: a great reduction in latency. The requirements on 5G, of which future LTE releases will be a part, are expected to be that the one-way latency for a downlink or an uplink transmission is less than a millisecond.

The first technology, being specified in Rel-14, is Fast uplink access, which makes it possible for a terminal to have an uplink grant available every millisecond, to be used only when there is data to transmit. By comparison, with scheduling based access, the terminal must transmit a request, wait for a grant, and then wait to use the grant. The persistent grant makes it possible to minimize the waiting time, and thereby the average radio delay is reduced to less than half.

Figure 1. Fast UL access (bottom) compared to scheduling request access (top).

The second technology is composed of two enhancements that are both targeted for specification in Rel-15. The first is designed to reduce the processing time but not the transmission durations. This is possible thanks to significantly increased processing capabilities in the hardware at both ends of the transmission since the first LTE release. The second component enables shorter transmission durations (Figure 2), which is a more radical change of the LTE frame structure. Here, the idea is to compress the whole transmission chain of waiting for a transmit opportunity and preparing for a transmission, transmitting the data, and finally processing the received data. The associated control and feedback should be done faster.

Figure 2. Illustration of an LTE downlink subframe with existing long and new short transmissions.

The compression is done by introducing transmissions with duration shorter than a subframe. In downlink this is done by splitting the data part of subframe into several parts. Each of these short transmission durations can be scheduled separately with a new in-band control channel. Also the uplink subframe is split into multiple shorter transmit durations and are scheduled from the same in-band control channel. The subframes are either split into two parts, four parts or into roughly six parts for the lowest latency mode. At the highest splitting level, a one-way transmission can be done in a total of about 0.5ms including processing of data. From simulations we have seen that this can improve the throughput for FTP download by 70% (Figure 3), an effect that comes from a faster TCP bitrate ramp-up thanks to the faster roundtrip of data and response.

Figure 3. Gain in downlink throughput for three different split levels of the subframe compared to the long transmission duration. File size 100kB and cell load 2Mbps.

In future extensions, we are adding robustness to Fast uplink access and Short TTI to get both reliable and fast transmissions, a combination that is required by the soon enabled use case of critical communications.