The State of LTE | Hacker News Reader

The State of LTE

opensignal.com

43 pointscleis41 comments

This was my first attempt at Data-visualisation using d3. I've done a few things with GoogleVis and R, the one is awesome for the web the other is awesome for its flexibility, D3 is awesome for both. I was also pretty much a novice to JS and CSS, I highly recommend AlignedLeft's tutorials http://alignedleft.com/ if like me you start from scratch with this.

Thanks for the tutorial rec, a client wants me to implement visual statistical analyses of what's going on in his business, and I could use these. All his data is in a proprietary dental office database (Patterson EagleSoft), and I was just thinking that D3 would be perfect for this project once I get it out of there.

P.S., Some of the scripts on the OpenSignal page could be moved from <head> to the end of <body>; other than that, it looks good!

Your graphs are pretty -- and it's an interesting page, but your ping section makes me wonder about the rest of your data.

You don't qualify what you're pinging, but if we're talking ping to the internet gateway, ping over a strong 802.11n connection on my iPhone 5 is 2 - 5 milliseconds on my local network. The same is not true of LTE which has more typically a 100ms latency to the carrier internet gateway.

LTE definitely brought down latency from 3G! But it's nowhere near as low latency as a good 802.11n connection!

True, we're pinging google.com (which resolves to its local sites) over the active data connection - whatever flavour of Wi-Fi or 3G. We could have also broken this down by 3G types e.g. UMTS vs HSPA etc. The idea is not to show the theoretical maximum of LTE or Wi-Fi, if we wanted to that we'd just use a dozen phones in lab conditions. Rather, we wanted to get a feel for the changes in user experience of the mobile web, so this data is drawn from a set of 9m speed tests run with the OpenSignal app.

This article leaves out a major issue if you're an iPhone user (no idea about Android phones): LTE frequencies have not been standardized. As usual, it's the US that is the outlier, and the rest of the world have agreed on a standard.

Here is Apple's list of models and their compatibility: http://www.apple.com/iphone/LTE/

For example, US iPhone models use GSM at 700b MHz, but won't work with European LTE (and vice versa) which use GSM at 850/1800/2100 MHz.

(And then there's my country, Norway, where we have only two networks which for some reason have decided on 1800/2600 MHz, which will work with none of the existing models.)

Disjointed channels and spectrum/bandwidth flexibility are features of LTE, hopefully the tune-able software defined radio is just around the corner.

"The LTE standard can be used with many different frequency bands. In North America, 700/800 and 1700/1900 MHz are used; 2500 MHz in South America; 800, 900, 1800, 2600 MHz in Europe; 1800 and 2600 MHz in Asia; and 1800 MHz in Australia."

It doesn't seem like anyone's particularly standardized on frequencies.

http://en.wikipedia.org/wiki/LTE_(telecommunication)#Frequen...

Rest of the world == Europe.

As a rule of thumb anything that's not standardized on iPhone is even less standardized on Android afaik there will be Android phones on all the bands that netwroks offer.

We're releasing an iOS app soon - hopefully within two weeks - and will be interesting to see how the LTE experience varies on exactly the same device across networks and countries.

Different parts of the world have allocated spectrum that's in use for other purposes, so the operators have to deploy LTE in the parts of spectrum that are available and they have a license for. All these(700/800, 1700/1900, 2500, 2600) are standarized bands as of today, though different bands have been standarized at different points in time.

This will be less of an issue as time goes on, newer phone models will support the various frequency bands.

I recently talked to an Intel engineer about why they don't have an Intel phone out yet. And he reached into his pocket and pulled out an Intel phone! Apparently, these phones are already available in China, but because Intel does not make the LTE chipsets, they have been unable to get it out. I asked him why Intel did not have LTE chipsets in their phones, Ans:"Because Qualcomm owns all the patents and makes the latest chipsets"

Intel/Infineon has an LTE baseband, but for whatever reason it isn't used in any phones. One factor is that Verizon and Sprint are locked into Qualcomm for backwards compatibility with their CDMA2000 networks and if you have to use Qualcomm for the Verizon version of the phone you might as well use Qualcomm for the GSM version as well.

HOW SPEEDS COMPARE doesn't really mean much. Japan has a higher density of users per square kilometer, so it seems to me that there should be some sort of normalization done to the numbers. I get better speeds in Kyoto (less simultaneous users) than I do in Shibuya (many simultaneous users!)

Thanks for the feedback. We aim to measure the actual user experience on LTE, rather than the theoretical capacity. If an area has a higher density of users then of course connections will be slower, but that is meant to be a part of what we're recording. Obviously countrywide averages are never going to say very much about localised area speeds, but they do give a general indication for the purpose of international comparison.

But on the other hand, higher density of users mean a large opportunity to use wifi offloading to complement the capacity. This of course depends on exactly where you measure and what exactly you are measuring due to short range of wifi (and on the capabilities of your measure equipment).

Also, how are you measuring speed? HTTP Download? Where are your servers located?

Sorry we missed this one, yep HTTP download from Cloudfront - so we use those servers. Others e.g. SpeedTest.net use the servers of the local ISPs which is better for measuring the max possible speed a user can get, but our focus is on the speeds users are more likely to get while actually browsing/downloading so Cloudfront works pretty well as a lot of global traffic is served through it - would love to see the % if anyone knows.

If density mattered, Hong Kong wouldn't have such good speeds.

That is because there are a lot more people using 4G in Japan then HK. Which only recently rolled out.

Random feedback:

On the THE GLOBAL ROLLOUT world map, the colors are not clear. They seem to be the same as the 2(?) pixel high lines below the labels below? Make them bigger! Or add a legend.

On HOW SPEEDS COMPARE I had to read the rotated y label first before I had any idea what the graph showed. Either add the unit to the numbers or add a clear title that is read normally.

LTE PING VS OTHER TECHNOLOGIES seems completely random. What are those numbers, how were they created? Ping to where?

Thanks for the comments, I agree we the key for the map could be clearer (yep it's those lines!). I'm also with you on the bar chart.

The pings are on google.com (which resolves to it's local site). And, extra clarification, we're using cloudfront CDN for the download tests.

OK will makes some changes then. Thanks!

Oh, I totally did not realise this was based on the measurements of your users.

I would add a forth colour on the map too: LTE from multiple carriers or not.

e.g. in the UK we only have LTE from one carrier at present, so for practical purposes most consumers (who are currently locked into contracts) don't have access to LTE, and those that do might choose not to since no competition means that currently LTE is very expensive here -- and that's if it's available at all; presently it's only in a few select cities. Listing "UK Broadband" as an additional LTE carrier is a bit misleading, since they're (to my knowledge) only a wholesaler and don't actually provide a service direct to customers.

I think a map showing which countries have multiple carriers would give an extra way of judging how developed a country's LTE infrastructure is. Great map otherwise though!

MetroPCS is using a 5x5 Mhz channel in most of its markets, that is the primary reason for low speed. This brings down the USA average. Most US network operators have the capability to reach the 100 Mbit/s downlink target; that requires a dedicated 20MHz channel (which no one has) and full-blown MIMO.

The Japan average is also bought down a lot by one network (NTT DOCOMO). As well as the usual factors age of the network could also be an issue here. Both SoftBank (16.2 Mb/s), KDDI (14.8 Mb/s) are under 1 year old, presumably under capacity waiting for people to upgrade handsets and contracts. DOCOMO (5.5 Mb/s) on the other hand launched in 2010. You see a similar effect in the US, where AT&T's network is regularly beating the older, more congested Verizon. We expect this to even out over time.

"WI-FI" is a bit ambiguous for a comparison like this. If your intention was to compare land lines to cellular then it should be called out as such. It would also be important to note that the land line performance in areas with LTE coverage may be significantly better than the national average used.

It's interesting how LTE doesn't always mean fast. Here in New Zealand carriers have only been testing LTE while upgrading their networks to 42mbps DC-HSPA. At home I can average about 30mbps on my iPad and iPhone over 3G. For me, I'd rather have constantly fast 3G rather than spotty LTE, especially with the battery situation. http://i.imgur.com/7k8LXe1.png

That's true. The fastest 3G speeds can exceed average LTE, so it's semi-defensible for carriers to brand some HSPA+ as 4G. In our tests though, the fastest speeds are always LTE. In the last week of the 50 fastest speeds tests all but 2 (#39 and #40) have been LTE. (Fastest was 69812 Mb/s - Oman Mobile!)

Great info! What are you pinging? Is it the default gateway or a specific host on the Internet? Also how do you determine download speed? The numbers for wi-fi are so disappointing. I wonder if you're not really measuring the speed of the DSL connection (or whatever) that the wi-fi router is sitting on.

If it's measuring public wifi, then I'm not surprised. I rarely see better than 2-3Mb. My HSPA+ phone easily beats almost every public wifi that I've seen, aside from the issue of data-caps.

Yep, it includes public Wi-Fi, though we are trying to distinguish between public and private for the purposes of making hotspot maps, anyway could also be cool to look at the breakdown in speeds between the two.

You may be interested in http://www.mobiperf.com/. Most of their data is available for research purposes, and there are many researchers who have used this for their tests.

That's part of the very cool M:Lab project, other apps running M:Lab http://measurementlab.net/measurement-lab-tools

Are HSPA+/LTE speeds really that "bad" generally? My HSPA+ phone typically gets 8-10Mbps down and roughly 50ms ping, and I've been led to believe that LTE here (the US) is considerably faster.

I use AT&T LTE in New York City, which from my research back in the fall was at the very top available in the US. Although I got 23 Mbps once, it is typically around 10 Mbps.

HSPA+ tends to have latency degrade far faster when you get farther from a tower than LTE, that's the big difference between the two.

LTE speeds are quite good on Verizon[1]. I get around 2-4MB/s normally. LTE pings for me are 50-100ms depending on where I am.

[1]Verizon uses the 700mhz LTE band.

I'm quite happy that this article clears up the real meaning of 4G.

Well, that's the ITU's definition. Unfortunately they didn't acquire the trademark for the 4G designation before carriers already had started to use it on their "slightly better 3G" networks, so there is no regulation on the use of the term in the market.

I think it's funny that AT&T called one of their earlier networks "4G" so now when I have an LTE signal, it doesn't say 4G and when I don't, it does...

Ex-Palm CEO Jon Rubinstein used to dismiss AT&T's "4G" network as "faux G."

Is T-Mobile going to have an LTE network in US?

Eventually yes. Right now it's just imaginary mostly, like Sprint's. However, both are supposed to have LTE soonish (they keep delaying it so can't be more definite than that).

Reply on news.ycombinator.com