5G researchers manage record connection speed | Hacker News Reader

5G researchers manage record connection speed

bbc.co.uk

"We need to bring end-to-end latency down to below one millisecond"

The incredible bandwidth is surely great but they don't mention how to improve latency here, they forget to tell what was the latency of this 100m transfer from an antenna to a receiver.

One of the big issues with today apps is latency, this is easily frustrating. It also prevents apps such as real time multiplayer games to be viable.

As a user, my impressions is that the latency involved with routing mobile data around is a lot higher (more like a second, even with LTE) than the latency involved with the physical transmission. Mobile data connections are generally more reliable than WiFi, but usually there is a significant lag before results are returned. Is that not correct?

This matches my personal experience with tethering being my main internet connection outside of work/office internet

Any omnidirectional microwave/RF network technology will have a short enough range that the signal propagation time is negligible for the purposes of human-perceived latency. This is a consequence of the need for cells to be small enough that they are only serving a manageable number of users. (Directional links like microwave backbones and backhaul can have latency worth caring about.)

Any latency you perceive comes from your packets sitting in a buffer waiting for their turn to be transmitted or waiting to be re-transmitted after a collision. The problem of queue management for wired networks has only had a satisfactory solution for a few years. When you throw in things like collisions due to the shared medium and interference from other users of the spectrum (less of a problem for cellular than WiFi), transmission rate and power selection covering multiple orders of magnitude, interference that can be problematic for some users but undetectable for users elsewhere in the cell, and the almost complete unwillingness of the hardware vendors to sell anything hackable enough to do research with, it's clear that wireless will continue to suck for the foreseeable future. However, a lot of these problems get a lot easier when you're working at a frequency that can never reach from the inside of your house to the inside of the neighbor's house.

Latency, latency, latency. Everything else is pretty much irrelevant at this point.

It's a simulation [1]. The claim is that the technique is applicable to center frequencies below 6GHz and the 1Tbit/s used a bandwidth of 100MHz (for a spectral efficiency of 10^7 bit/s/Hz???). I'm guessing that whilst the technique might be applicable below 6Ghz, the 1TB/s rate isn't.

There are fundamental limits on the information capacity of an antenna using the EM spectrum [2], based on the surface area of the volume of space it occupies, in units of wavelength. (Related to the Holographic Principle?) I haven't done the calculation, to see if the claimed rate is within this limit, but a spectral efficiency of 10^7 bit/s/Hz is about 5 orders of magnitude beyond what others have reported (less than 100bit/s/Hz [3]). It will be interesting to see the details!

[1] http://www.ispreview.co.uk/index.php/2015/02/university-of-s...

"UPDATE 25th Feb 2015

We’ve been finding the 1Tbps claim a little difficult to digest and so have been prodding Professor Rahim Tafazolli for further details, specifically a greater clarification of how the performance was achieved.

According to Tafazolli, the new class of Detector (a completely new approach) was tested through computer simulations (these simulated a real mobile/wireless environment) and were found to achieve the 1Tbps rate claimed. In our view that’s quite a bit different from conducting a practical test.

Next year Tafazolli said that his team would work to implement this in a proper hardware/software platform and test it in a real environment in the 5GIC outdoor testbed. Hopefully they will be able to announce the performance in 2017."

[2] http://arxiv.org/pdf/cs/0701055.pdf

[3] http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=tru...

I don't even get 4G speeds almost ever. Bandwidth is always massively oversold. Often I can't even successfully load a random web page when I have 4+ bars of "4G".

What part of the country are you located in? I can get full 4G LTE speeds in the middle of nowhere Appalachia where my brother lives, via AT&T. Streaming HD video on my phone there is a non-issue.

San Francisco, both AT&T and Verizon. It's terrible here.

With T-Mobile LTE in San Francisco, I get 20Mbps+ in a lot of locations, including about half my residence. (Unfortunately there are also a bunch of very-weak spots with low bandwidth and multi-second pings, and far more dead zones along highways in rural areas.)

Also super happy with T-Mobile's speed- it's faster (bandwidth-wise) than my stupid U-verse. But the whole text message issue they have is driving me nuts to the point that I'll probably switch to another carrier with less available bandwidth just to get sms that works.

What's the SMS issue?

Rural and T-Mobile just don't get along right now. A year ago, the entire I-5 corridor as I approached Oregon and through crater lake was invisible for data or calls (with a small band of

I am, however, hopeful that T-Mobile has this in their sights [1]. They have done more to improve mobile user experience in the past 2 years than the rest of the networks for the past 10.

San Francisco has notoriously bad infrastructure due to all the NIMBYs blocking any kind of development. I don't think it's very representative.

Just tested my T-mobile. 8 Mbps down, 9 Mbps up one block from 1 Post Plaza.

I've got 15 down and 10 up in Rockport, MA

Streaming Netflix in Central London isn't an issue (I'm on Three); do you have a choice for a better provider?

http://www.speedtest.net/my-result/i/1154366377.

Just posted 22/9 40ms ping in DC Union Station. It's 7pm and the place is packed with people playing on their cell phones while waiting. Maybe blame SF NIMBYs for not letting the cell companies put up enough towers or run enough fiber backhaul?

To be fair, I've seen ISPs give special treatment for packets coming from speedtest.net. I'm looking at you Comcast.

Doesn't speedtest use lots of mirrors that have different IPs and hostnames?

Yes I've noticed the same thing. I ended up rolling my own test just to verify. I wish I was surprised comcast does this, but they totally do.

Actually, transportation hubs are usually outfitted with specialized microcells so the throughput is higher there even though the user density is large.

https://en.wikipedia.org/wiki/Microcell

Okay, just 13/13 in my large apartment complex in downtown Baltimore. http://www.speedtest.net/my-result/i/1154413210

What phone are you using? Some phones pawn off HSPA+ as "4G".

I wonder if we'll still be stuck with our puny data caps when this arrives. Blowing through your monthly quota in a minute might be astounding from a technical point of view, but it only takes one broken page with an ajax infinite loop to get you.

If you look out to 2020, and assume 5G will be ready, what you'll get in production is nowhere near 1tbps. You might see 500mbps - 1gbps commonly deployed across the US and Europe during the 2020-2025 time frame. Deploying the initial infrastructure was very expensive, but now it's in place. That's part of the reason why the US was able to transition relatively quickly to 4G LTE, there was no need to build 200,000 towers again.

Data caps will rise substantially over the next five years. By 2020x, they'll be commonly at several hundred gigabytes at 5G speeds across the US and Europe, with unlimited data at lower speeds as a universal feature. No doubt numerous carriers will offer variations of unlimited data.

You're entirely optimistic, I fear:

"One of the new AT&T plans will cost $25 per month and offer two gigabytes of data per month, which AT&T says will be enough for 98 percent of its smart phone customers. Additional gigabytes will cost $10 each."

This is in 2010.

Today that $25 will actually get you HALF the data it did in 2010, five years ago.

This is why I buy phones outright to keep my unlimited data plan from verizon. I tether it for my home internet as well, usually use between 80-100gb/month.

I'm waiting for the day they just take it away.

[Citation needed]

Data caps are a form of price discrimination, and will evolve with supply and demand. It's like Apple's upcharge for each jump in flash or SSD size--it's a way to push a higher proportion of costs onto those who value the product the most. The brackets will keep going up and be priced such that lots of people will pay it.

Yes very true. But i think mobile data has to be cheaper as increased data speeds are achieved and users are promoted to make more use of the excess speed on portable devices.

Here in the North we have insane data caps on our regular DSL and Cable internet.

On a couple of the packages, it's possible to blow through your entire monthly cap in 6 hours - and I've seen people do it.

Reply on news.ycombinator.com