A User Review of EE's 5G Network in London
In May 2019, UK mobile operator EE launched one of the first 5G networks to go live anywhere in the world. The network was available in six cities at launch, including London, Europe's largest city, where I've been testing the service for a few weeks.
Having covered 5G as an analyst for five years, and seen quite a few pre-commercial demonstrations, you could say I was eager to get my hands on a commercial service.
So how does it measure up?
In short, the EE network team has done an excellent job of introducing 5G to its 4G/LTE network.
However, the answer to such a question will depend on any individual's expectations. I regard myself to be a fairly clued-in user, so the service performed more or less as anticipated, with coverage actually a little better than expected.
In summary, it's a good service that augments the mobile broadband experience we have today. What it isn't is an entirely new experience that blows you away.
Faster mobile broadband is the very least that most people expect and, at this stage, speed tests do seem to be the killer app for 5G. The first speed test I ran after picking up the OnePlus7 5G phone from EE's public relations team resulted in a 650 Mbit/s downlink. This was recorded maybe 100-200 meters from one of EE's flagship 5G sites near St Paul's Cathedral, running massive MIMO. On its own terms, 650 Mbit/s is impressive for a commercial mobile service. To compare it with LTE, I ran the same speed test, at the same time and location, using my personal Pixel 3 device on the Vodafone UK network and managed a 65 Mbit/s downlink. So, there you have it -- by a fortunate coincidence, 5G downlink speeds are exactly ten times faster than LTE!
This wasn't a like-for-like comparison, of course: I wasn't directly beneath a Vodafone antenna (as was the case with EE); the OnePlus7 device is a newer, higher-spec device than my Pixel 3; and so on. And yet, despite these caveats, this result is indicative of how 5G differs from 4G. Where you have good 5G signal, it is faster than 4G, and occasionally appreciably so, to the extent that when browsing or downloading, you don't always need a speed test app to notice the difference. The challenge to this analysis is that EE's LTE network in London is world class and frequently generates >100 Mbit/s downlink connections, making it hard at times to discern the difference between the LTE and 5G user experience. This is consistent with EE's aim of using 5G to expand capacity in areas of high demand, so that all users have a good experience.
The top downlink speed I experienced was 750 Mbit/s in the heart of London's media and advertising land (Golden Square, Soho) on a very quiet Sunday morning. That was very much an outlier, however: Speeds of 400-500 Mbit/s are more common. But mostly, in zones with a 5G signal, the downlink speeds were in the range of 100-300 Mbit/s, which, in my experience, is typical of EE's 5G network in London. Here's a selection of speed test screenshots:
It's also fair to note that in quite a few places where the phone showed a 5G icon, downlink performance was lower than you would expect -- sometimes in the 20-50 Mbit/s range, and occasionally even less. This typically happened indoors (most of my tests were outdoors), in streets overshadowed by tall buildings or when traveling on a train or bus.
The picture below shows an example from the south side of London Bridge -- a busy road junction right outside a major rail station -- outdoors mid-morning. In the screen shot to the right, you can see the phone shows a 5G icon and the speed test app records a downlink speed of 33 Mbit/s. To the left, the screen shot from the Cell Mapper app (an unofficial crowd-sourced cell site database) helps explain why the recorded speed is nearer to 50 Mbit/s than 500 Mbit/s -- the serving site for this speed test was located across the river and a couple of blocks behind the waterfront.
Delivering LTE performance when the phone shows a 5G icon isn't the behavior EE wants, and the company is clear it doesn't want to be regarded as a company that claims to offer 5G in more places than it actually does, because this could lead to customer dissatisfaction with 5G. It is, however, an almost unavoidable situation, given the nature of wireless networks, that performance will get worse towards the edge of the cell. Displaying the 5G icon while delivering LTE-like performance may, on occasion, also be due to the way EE has deployed coverage and implemented what's known as enhanced dual connectivity in its network to manage the connections to the primary (LTE) and secondary (5G) cells. The average customer won't want to hear about this kind of explanation, of course: The good news is that as more 5G sites are deployed in contiguous coverage zones, this issue should fade away.
There were a couple of other glitches when the device showed as connected to 5G. One that happened quite a few times was a version of the "spinning wheel" that can show when you hit a link in a browser or an app -- for some reason, the connection just did not come life. Sometimes it seemed like this was associated with uplink speed tests running in the background: Maybe it was something else. Who knows. It happened enough to be mildly annoying. I haven't heard whether any other users have had this issue.
To put this review into context, here's some background on the EE 5G network deployment. This is based on my current understanding, and is not official EE information:
EE's 5G network uses 40 MHz of TDD spectrum at 3.5 GHz. It is an overlay macro cell deployment that uses the existing LTE sites, on a 1.8 GHz grid.
London is predominately rooftop deployments. In busier areas, it is already a relatively dense macro network.
5G is very tightly integrated with LTE. As 5G is deployed, LTE equipment is also refreshed.
The network uses enhanced dual connectivity (EN-DC) to combine LTE and NR service. This is to maximize data rates and make efficient use of the spectrum. Downlink is both on LTE and NR (unlike the first US networks, which were NR downlink-only at launch). Currently, uplink is on LTE or NR.
Most of the EE 5G base stations are configured for 8x8 MIMO, with a few higher-capacity sites using 64T64R massive MIMO active antenna systems for 5G. The use of 8x8 as opposed to 32T32R and 64T64R systems for 5G is interesting.
Each 5G site has a minimum 100 MHz of downlink spectrum deployed per sector, comprising 2x60MHz of LTE (3 carriers of FDD) and 40 MHz of 5G at 3.5 GHz (a single TDD carrier). Busier sites have 125 MHz deployed comprising 2x85 MHz of LTE and 40 MHz of 5G.
EE's 40 MHz of 5G spectrum is less than the mid-band deployments of other operators in South Korea, parts of Europe and even Three UK, all of which have 100 MHz. With another mid-band spectrum auction next year, EE is hopeful it will be able to secure adjacent spectrum which it could deploy on the same equipment it has today, without having to deploy extra hardware.
EE's LTE network is based on 1.8 GHz with extra capacity provided by 2.1 GHz and 2.6 GHz. It has upgraded to LTE-Advanced Pro in many places.
- Backhaul is typically 10-Gig Ethernet provided by BT Openreach, the incumbent, national regulated wholesale provider in the UK.
5G coverage in London
EE has about 20,000 sites in the UK and will upgrade its busiest 3,000 sites to 5G first. This is sometimes known as a "pepper pot" strategy because it refers to investing in 5G at sites where demand is highest rather than deploying for contiguous coverage. In the London area, where I've been using the service, EE has around 1,500 sites and had deployed 5G on a couple of hundred sites at launch, with more being added on an ongoing basis. Since the launch more than two months ago, it feels like 5G coverage has been added in a lot more locations: It's hard to tell, and it could just be I didn't notice before, but it seems like EE's 5G coverage is expanding quickly.
At the time of writing, coverage across London is still patchy. Some areas don't have 5G at all, quite a few areas have 5G in hot spots, and some areas have pretty OK coverage. In context, considering the service has only been live for a short period, coverage is actually pretty good. In parts of central London, you don't have to hunt for 5G; the icon pops up all over the place. In the City of London financial district (a.k.a. the "Square Mile"), for example, 5G coverage isn't yet ubiquitous, but it is commonplace. The 5G icon also shows up fairly frequently indoors in cafes and lobbies; although, even with the icon showing, indoor downlink speeds are generally closer to LTE speeds (for the reasons discussed above).
On balance, then, a good start by EE on 5G coverage. You can understand why EE is investing in its busiest sites first and its focus on delivering a consistently good mobile broadband experience across LTE and 5G for all users makes complete sense. From a user perspective -- I guess, a power-user perspective -- I would like contiguous coverage zones in more areas, without flitting between 4G and 5G so frequently. Already there are some areas of contiguous coverage and I'm sure more of these zones will exist in London in the near future.
It's worth noting that when I reviewed the first GPRS, UMTS, HSPA and LTE services in London, coverage in each case was better than we see now with 5G. This reinforces that, in the first instance, operators will use 5G to augment the mobile broadband experience in areas of high demand. There is no 5G coverage, for example, at my house in "Zone 2" (a few miles from central London), whereas for 3G and 4G introduction, my area was covered immediately. However, like much of London, my neighbourhood is blanketed by LTE-Advanced and because I live not much further than a cricket ball throw from an EE cell site, I consistently get 200 Mbit/s in my house and can get close to 300 Mbit/s on my street. This is one of the reasons it's hard for 5G to differentiate at launch -- EE's LTE is excellent in lots of places.
Mid-band 5G rooftop deployments, such as those used by EE, will be the dominant deployment model in much of Europe and parts of Asia in the first phase of 5G. This differs from the mmWave street-level deployments in the US. Both have their merits: The peak data rates reported on US mmWave are significantly higher than on mid-band networks; however, mid-band gives better coverage than mmWave on a per site basis. I like the coverage you get with mid-band 5G.
However, from using 5G in London, I think mid-band operators are going to have ramp investment in street-level small cells in the coming years, perhaps using mmWave, to deliver 5G in a truly excellent way. Mid-band spectrum and rooftop deployments are a fantastic way to introduce 5G, but to provide service to urban canyons and areas of hyper-dense demand, this strategy will only take you so far. There are just too many of these hard-to-reach locations in large, busy cities to rely on the existing macro site grid and mid-band spectrum. It's fine for now, but targeted densification and indoor coverage seem sure to follow in later phases of 5G. It is also clear that low-band 5G, either via fresh spectrum, or by using dynamic spectrum sharing in existing low-bands, is going to quickly become important.
In all the 5G speed test screenshots I posted on Twitter -- @gabeuk, thanks for the follow! -- most of the questions were about latency, which at around 40ms for a roundtrip to the Internet is… exactly the same as LTE!
The promise of low-latency 5G is a subject where the industry (media and analysts included) has let expectations run ahead of reality: Yes, 5G is specified for ultra-low latency, but only in specific deployments and configurations. In today's commercial mobile networks, which are deployed non-standalone using dual connectivity LTE-5G, the lower limit on latency is bound by LTE. Korea Telecom is currently the world leader in low latency mobile broadband, claiming a sub-10 ms round-trip due to optimizations in the radio domain (a different twist of dual-connectivity) and in the architecture (including edge cloud deployment). Sub 10ms latency mobile broadband performance will come to more networks in time.
A note on speed test apps
I don't understand in detail how these apps measure Internet speed. There appears to be quite a bit of variability in methodology and they often produce quite different results, even accounting for the inherently variable nature of mobile connections. The screen shots in this review are from the app Fast, a free service provided by Netflix and pre-installed by the EE public relations team on my test device. It is predictable and consistent, without ads or, so far as I can tell, user tracking, or indeed anyway to record your measurements other than screenshots. I also like the Meteor speed test app from Open Signal, which records location history and similarly doesn't have ads. Both these apps seemed to generate results within similar ranges.
The OnePlus 7 5G smartphone
Those new to 5G will also be using (for now) a new, high-end smartphone. These devices nearly always feel fast if you're coming from a device that's a year or two old, and it's interesting to think about how that dovetails with the 5G network to generate the overall user experience.
The OnePlus7 5G smartphone is a nice device, with a good screen, fast charging, a fast processor, lots of RAM and so on. Other than 5G, the best feature, in my opinion, is the 90-frames-per-second screen refresh rate, which makes scrolling smooth and fast.
The battery held up pretty well. I'm not sure how this phone is designed, but apparently a phone with dual-connectivity 4G/5G radios typically consumes more power than a single radio device. In practice, it's not obvious that this makes much difference. I suspect the large, high-definition screen has a bigger impact on battery life. In any case, it was a new device and quite large, so battery was never a problem over a full day.
My biggest issue was the occasional "hang," as noted above -- that may be a device issue, or a problem with the network, or some combination. Maybe even user-error...
Coming from a Pixel 3, which runs the latest Google Android, the OnePlus version of Android felt a little old-fashioned. Hardware-wise, on paper, it's higher spec than the Pixel, but who looks at phone specs anymore? In practice, I prefer the Pixel. I'd love to try the Samsung S10 5G phone.
The independent review process
The device and SIM was loaned to me free of charge by EE's public relations team on my request. They didn't place any restrictions on what I could write, or how I could test the service. It is in my interests to maintain relations with EE, and because my analyst work is focused on 5G I have vested interest to see it succeed; however, it is more important that the Light Reading audience can read a genuinely independent review.
A final note: Since working for Heavy Reading (part of the Light Reading group), I've tested GPRS (2.5G), UMTS (3G), HSPA (3.5G) and 4G mobile data services at launch in London. Each time I've been impressed with the service, and EE's 5G launch is no different in that respect. Equally, I've been able to extrapolate a launch service into future potential, and with the 5G cycle just getting underway, it is clear that, at a minimum, we're going to get really excellent downlink performance -- and that's just the beginning!
I am, of course, looking forward to seeing how EE's rivals perform on 5G.
— Gabriel Brown, Principal Analyst, Mobile Networks & 5G, Heavy Reading