Latency: LTE’s Achilles’ Heel?

By | December 19, 2011

While much attention has been focused on LTE data rates, another important parameter, latency, has gone largely unnoticed. Yet, latency is critical to enable a number of applications particularly voice services (VoLTE). I review in this post some of the publicly reported results I found on LTE latency.

But first, I like to clarify that latency will depend on a number of parameter. It is a statistical measure and through the reported results, we sort of get a feel for what LTE can provide. Some of the parameters impacting latency include traffic and subscriber load, the type of traffic and the channel radio frequency condition.

LTE includes quality of service management with up to 9 classes of service. Conversational voice is very susceptible to delay and packet error loss, particularly for low bit-rate vocoders. Real-time gaming is another highly demanding application both in terms of delay budget and packet error rate.

QCI

Resource Type

Priority

Packet Delay Budget

Packet Error Loss Rate

Example Services

1

GBR

2

100 ms

10-2

Conversational Voice

2

4

150 ms

10-3

Conversational Video (Live Streaming)

3

3

50 ms

10-3

Real Time Gaming

4

5

300 ms

10-6

Non-Conversational Video (Buffered Streaming)

5

Non-GBR

1

100 ms

10-6

IMS Signalling

6

6

300 ms

10-6

Video (Buffered Streaming)
TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)

7

7

100 ms

10-3

Voice,
Video (Live Streaming)
Interactive Gaming

8

8

300 ms

10-6

Video (Buffered Streaming)
TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)

9

9

I summarized below some of my findings. Unfortunately, most of these reported values were not qualified for the general conditions under which the measurements were taken.

Reported Latency Network / Reference Comments
Average: 49 msecMin: 40 msec AT&T Houston (Signals Research) Latency measured to servers in the operators local market.
Average: 23 msecMax: 38 msec TeliaSonera Finland (Epitiro) These are reported as “Network Latency” and defined as “the time it takes for a network to respond.” I believe this measurement was done under very controlled conditions since they reported relatively stable average.
Average: 143 msecMin: 79 msec Verizon (BTIG) Tests were conducted in-building and near a window.

According to LTE system specification and requirements, user plane latency is defined to be sub 10 msec for two way radio delay. Although this value excludes the latency to the core, the bulk of latency typically occurs on the air interface. Hence, there is much more room for improvements and optimization when it comes to LTE latency. In fact, I see this as a key issue and even failure of the LTE standard if latency is not improved. After all, LTE was designed with a 1 msec sub-frame to meet the low latency requirement. So far, this goal has not been achieved.

LTE Latency measurements in AT&T's Houston network.

AT&T Houston LTE Network Latency Test Results (Source: Signals Research Group, LLC)

TeliaSonera Epitiro LTE Network Latency Test Results

TeliaSonera Finland LTE Network Latency Test Results (Source: Epitiro)

BTIG Verizon LTE Network Test Results

Verizon LTE Network Test Results (Source: BTIG)

5 thoughts on “Latency: LTE’s Achilles’ Heel?

  1. mayur

    Is PACKET DELAY Budget between UE ENB or ENBPDN

    1. Frank Rayal

      The main delay is between the UE and eNB. The delay to the core is generally small in comparison.

      1. mayur

        I agree,but im asking specifically about packet Delay Budget

        1. Frank Rayal

          The packet delay budget is end-to-end: it’s the total allowed for the application to run properly.

Comments are closed.