Hacker News new | past | comments | ask | show | jobs | submit login
Wi-Fi and LTE in Unlicensed Spectrum Bands (googlepublicpolicy.blogspot.com)
102 points by sinak on Aug 5, 2015 | hide | past | favorite | 68 comments



The linked paper is a great read. The punch line is on 6-7 (LTE's duty cycling co-existence mechanism reduces wifi throughout because it doesn't listen before transmitting and invariably stomps on a WiFi transmission mid-frame causing errors and possibly renegotiation to a lower link speed).

Rant: It's sad that we are fighting to put more users in existing unlicensed bands instead of expanding the amount of unlicensed spectrum available: http://www2.itif.org/2015-coase-wifi.pdf; http://wireless.fcc.gov/outreach/2004broadbandforum/comments....

Our current approach to unlicensed leaves enormous performance on the table, and it doesn't have to be that way. Basically the solution is two-fold. First, there is a lot of spectrum that could be reallocated to unlicensed. Second, we could get a lot more use out of the band if we imposed least cost avoider type principles on use of the band (i.e. devices have an affirmative duty to exercise reasonable measures to be resistant to interference).

If you're interested in the subject, read Bill Lehr's work (at MIT). He's one of the few people writing about the issue who knows what he's talking about (he's an engineer who went on to get an Econ PhD).


Just to remark, there is much more than Wifi and Bluetooth in the 2.4 GHz and 5 GHz going.

There other users like ZigBee and many industrial users as well. Especially the industrial users have problems. Because there is no licensed band for them, they have to work in the unlicensed as well. Industrial users have stronger real time and reliability requirements than Wifi can offer (response times in ns).

Technologies like LTE will hurt innovation for industrial usage.


Can you expand on "have to use unlicensed"? Is that really saying "don't want to deal with global regulatory systems"?


Nope. At least in Europe, there are working on to get their own frequency. There are also first contacts made to regulating bodies. But we learned, it is very complicated with lots of intransparent groups and clubs of those regulating bodies on national, Europe and international level. That means it will take at least 10 years and than additional years to then transform those international agreements into national regulations.

So, to be honest nobody believes that it will work out. The industrial industry is very diverse compared to telcos. There are only 3 different telcos in Germany. There 3 or 4 telcos for the whole US. So they have a different market position.


Wouldn't devices for industry be certified under part 18 as ISM devices rather than as part 15 devices? Part 18 allows much higher power than part 15, which should allow the ISM users to deal with interference.


As soon as the article brought up plans of carriers to expand LTE into the unlicensed 5GHz band, I immediately thought about the implications for current users of WiFi. Specifically that high-powered LTE would overwhelm the signal output by home WiFi routers and in other typical scenarios.

The article comes essentially to to the same conclusion, and it will be unfortunate if new FCC regulations have to be created. It seems to be human nature that sooner or later powerful interests stop "playing nice", and start acting like bullies until regulators are forced to step in.

Of course it doesn't have to go that way, but what are the odds that it won't?


Yes, my thoughts flew to the same thing. This is a recipe for regular users getting stomped.

Allowing organized and coordinated commercial exploitation of unlicensed bands is a total violation of the intention of those bands. If you want to run your home router, fine, but when you turn it into a business that has the potential to cause a meaningful disruption of other users you need to either get a band of your own or contract/integrate with someone who does (like a cell carrier).


The proposal in the LTE standard is not to use high-power transmissions, like those used for macrocells in licensed spectrum, in the unlicnensed band. This wouldn't be allowed anyway by existing regulations. The proposal is to use a form of LTE with low-powered tranmissions (in fact subject to exactly the same regulatory limits as WiFi) in the unlicensed band.

All unlicensed band technologies are subject to the same regulations on power levels. While of course these may be changed I don't think there is a realistic chance that, even with heavy lobying, one particular technology would ever be given different rules from all the others.

Basically all unlicensed technologies have to coexist as best they can. WiFi interfers with other WiFi users. Bluetooth interferes with WiFi etc. The questions are: 1) whether LTE-Us coexistence strategy is reasonable 2) whether the economic power of the carriers means that LTE-U might be deployed so densely that other users of unlicensed spectrum suffer degraded service


I would think that everyone operating in unlicenced parts of the spectrum was subject to the same regulations.


They are-- until you lobby the FCC to change those regulations. Part 15 is very clear on this.


Part 15 of what?


For US hams it's an extremely often cited regulation, touching on many aspects of the hobby. There's a good chance 'KB1JWQ' is a ham :)

If you or others care: https://en.wikipedia.org/wiki/Title_47_CFR_Part_15

I'm not based in the States, but it pops up enough everywhere ham-related for general familiarity.


Part 97 is the amateur radio rules in the U.S. Part 15 is the catch all of FCC rules. If it's not define in some set of rules, then it's likely covered in 15.


Part 15 says (amongst many other things) that if a service causes (or suffers) interference it must go of the air.

Nothing to do with Amateur Radio.


RFI doesn't affect amateur radio? Don't take my word for it, the ARRL have a few thoughts on Part 15:

http://www.arrl.org/part-15-radio-frequency-devices


I think it's interesting (and scary) to consider this in the context of how smartphone vendors seem to be really dragging their heels on 5GHz WiFi on all but the high-end phones. If we're going to be broadening use of 5GHz, we should be doing it in a way that is interoperable with the existing deployments. At the very least, nobody should be allowed to ship a device that supports 5GHz LTE but doesn't support 802.11ac.


It's also called LTE-U. There was a lot debate around this in the recent IEEE conference. I mean yelling at each other kind of debate!


> the recent IEEE conference

IEEE has a lot of conferences. Which one are you referring to? I'm genuinely curious.


Any more details? What were the two sides of that debate?


One side was (is) AT&T and Vrizon and other side was Google and other members


If AT&T and Verizon are on the same side, I'd probably be on the other side.


Its American mobile companies so id agree with you :-)


IEEE develops the WiFi standards, while LTE is standardized at 3GPP/ETSI/ITU. There tends to be some rivalry, and doing similar things in two different ways.


I thought radios in the unlicensed bands were capped to really weak TX power levels which would make them useless for broadcasting to large areas.


Not true in most countries (including USA). There is a whole service provider industry (WISP) with sizable equipment vendors (Miktotik and Ubiquity) that depends on using the unlicensed bands over wide areas and long range.


In the absence of a license, aren't those long range and wide area systems still subject to the same power caps as consumer gear?


Yes, but a) those caps are not so small as to be useless as stated by the parent, and b) there are sub-bands of the 5GHz unlicensed band that allow higher txpower in the US only when deployed "outdoor". This page provides a decent summary: http://www.wisp-router.com/page.php?11

fwiw fear of overwhelming interference from well-funded entities such as carriers is not new. In fact it was very common until the extra capacity in 5GHz was opened up 10 years ago or so. Prior to that you could find a cell carrier in your area decided to backhaul a tower with a T-1 on 2.4GHz, lighting up the entire band with impenetrable noise. Since that time the carriers' backhaul capacity needs increased significantly, driving them to fiber-feed their sites or use 23GHz licensed. At the same time the available channels in 5GHz expanded (through re-use of space previously allocated for military use). So things have been reasonably quiet (sic) on the interference front until this new challenge.

Disclosure: I built and run a very small WISP due to owning a home in a place with no Internet service, in conjunction with making my living writing software (oops!)


There's a big difference in how they use the available power.

WISP-type equipment generally uses very directional antennas so the available power is used more efficiently.

A laptop and a home WiFi router are like people talking in a crowded room. Long-range systems are like talking into a unpowered megaphone while the other side is listening with a megaphone on its ear.


If what the carriers are talking about is highly directional links for backhaul then I don't see what the fear is. Directional is not going to interfere with nearby WiFi.


They're talking about handset to base station, split band -- using licensed frequencies for downlink and unlicensed for uplink. So, not backhaul.


I think that new 5GHz carrier cells may have some very powerful directional transmitters aimed at them if they interfere with wifi.


Depends on your receiver sensitivity and desired throughput. Smart Grids operating at 1 watt dBm with receivers at -98 RSSI routinely have mesh-hops of 5km+ @ 100 kbits.


What does "1 watt dBm" mean. Do you mean 1 dBm (1.3mW) or 1 Watt (30 dBm)? Presumably the former?


Up to a point "professionals" can go over the 200mw limit but this is for PTP links using wifi tech.


It's spectrum licensing that's obsolete. We know how to make smart radios. It's time to enable people to use as much spectrum as they need, not as much as we are willing to pay to rent.


I agree and disagree. Licensing is obsolete for a lot of cases. At the same time, smart radio technology isn't there yet. In fact, this article is a great example of that. Both WiFi and LTE have pretty primitive spectrum sharing technology built in. But it doesn't work that well together.

I worked on spectrum sharing technology as part of the DARPA XG project about 10 years ago. It worked great for many things and had some significant limitations. Not to mention, the necessary hardware is still very expensive (high bandwidth analog frontends, extremely sensitive detectors, substantial CPU and memory to run the algorithms). The simplified version of that sort of technology you have in LTE on unlicensed band is probably the best you can do without blowing your budget. And that's not good enough to replace licensing. The challenge is probably a bit but not tremendously simpler than self-driving cars, and the relevant cost, size, and power budgets are a couple of orders of magnitude tighter.

Besides that, the ideal of smart radios interacting in a totally unregulated world is suboptimal. As Bill Lehr points out, smart radios work way better when you impose certain rules on everyone. See also Kevin Werbach's work on the subject: http://werbach.com/research/supercommons.pdf. The unlicensed band for example isn't totally unregulated. You have strict power output limitations. But that's sub-optimal. If you're a WiFi access point in the country, you can probably use a lot more power without causing interference. At the same time, you don't want to create a shouting match scenario, where each node ups their power output to drown out the others. More sophisticated rules help you get more efficient use of your unlicensed band. For example, you can mandate transmission power control (only enough power to close the link) you can mandate the use of FEC, so you don't have dumb devices that can't handle even a tiny bit of interference. Etc.

But you're right about the big picture. Allocating spectrum is not the way forward. We don't "allocate road" to cars--we just define rules they have to follow to play nicely with others on the road. Licensing has its place in enforcing those rules, and doesn't need to imply fixed allocations.


I wouldn't call current wifi smart any more than I'd call the primitive beamforming in wifi "SDMA." I'd also agree that smart radios need to observe some common set of rules, or at least principles, much in the way that self-driving cars need compatible principles of operation.

I am, however, optimistic about cost and timeline. The amount of hair in your mobile phone's radio to make it work on 2G FDMA/TDMA, WCDMA, and OFDMA, and their respective layer 2 protocols, while it certainly adds cost, doesn't add a multiple of cost compared with, say, a pure wimax 2 radio. Making radios that use different technologies smart enough to share spectrum is a relatively cleaner and less compute-intensive problem. I think it could be done now, and a really huge amount of spectrum could be made available for use cases similar to today's wifi right now.


Uh, no. Leave unlicensed spectrum alone. That belongs to the people.

If licensed spectrum is in finite, constrained supply, let the market take care of that. Raise the price as it becomes more constrained, and wait a couple of years for the next technology change that increases the density.


What does "belongs to the people" mean in this context? Should Starbuck not be able to put up Wifi access points? Only private users? Belongs to the people sounds a lot more like ham radio really.


It means that it's a commons. You, I or Starbucks doesn't need a lobbyist to build, buy, or operate a device on unlicensed spectrum that complies with a few FCC rules. One of the key rules is that you cannot interfere with others.

Licensed telcos shouldn't be able to stomp on the commons because of their inability to manage the spectrum they have. If Verizon & AT&T lack capacity to deliver, they can raise the prices, subsidize replacement of older devices that use more spectrum, or make capital investments in high-density areas.


But... wifi interferes with others all the time.

The actual language:

>The primary operating condition for unlicensed devices is that the operator must accept whatever interference is received and must correct whatever interference it causes. Should harmful interference occur, the operator is required to immediately correct the interference problem or cease operation.

Practically wifi at 2.4 GHz is a mess from the density of wifi which is why that isn't even considered. 5 GHz isn't yet but will be soon enough.

It's the commons, they're not operating outside of the rules unless they're broadcasting beyond the specifications: https://en.wikipedia.org/wiki/U-NII

Qualcomm's propaganda page is interesting: https://www.qualcomm.com/invention/technologies/lte/unlicens...


It means there should be roughly equal ability to access it among people. Someone should not be able to benefit from using the spectrum in a way that actively degrades other people's ability to use it.


The article would have been more complete if it had noted that Google is attempting to be an 'SAS' database provider for the 3.5GHz band. See http://www.fiercewireless.com/tech/story/google-touts-spectr...

So what Google is saying is "LTE doesn't work well with Wi-Fi in the 5GHz band, so you should use the 3.5 GHz band. And since we are not evil, we won't mention that you need to go through yet another Google service to use the 3.5 GHz band."


That's a big of an exaggeration.It's surely not the main motivation here.

Being a SAS databse provider won't be very meaningful business for Google.On the other hand, we know from large efforts and common sense, that Google really wants faster cheaper internet access to all.


Just FYI, here is response by Qualcomm. They say for example "Qualcomm’s demonstrations have proven that adding a neighboring LTE Unlicensed node does not impact an existing Wi-Fi node any more than would adding another Wi-Fi node. In fact, in many cases, replacing a Wi-Fi node with an LTE-U node improves throughput for nearby Wi-Fi users" http://apps.fcc.gov/ecfs/document/view?id=60001104452


That's some sneaky wording. the way wifi is viewed in society today - as a free resource - what matters is the total free capacity for everybody.

And in their example, most definitely , by taking 1 user and converting it to a non-free method, free capacity decreases.


The unlicensed band was never dedicated to WiFi, and WiFi is not necessarily free. Carriers like AT&T has rolled out a lot of WiFi hotspots that their customers devices can connect to.


There's a big difference in competition levels between the wifi(i.e. anybody is able to offer a free or commercial service), and between LTE-U assisted , which depends on access to the carrier network which is far from being competitive.

And that's the reason why carriers added the assisted part to the standard,according to internal documents:

http://www.eetimes.com/document.asp?doc_id=1326570


I'm sure the big operators are more than happy to limit possible competition, but that's not the primary reason why LTE-LAA is assisted.

First, LTE-U with its fixed pattern is a quick n' dirty non standard proposal. It's not even acceptable everywhere, as some countries (Europe for example) do require LBT in the 5 GHz band, which LTE-U does not do. So I'll skip that part.

For the 3GPP standard LTE-LAA, a fundamental reason that it requires a primary channel in licensed bands is that the way LTE works does require an always-on, reserved channel. This is the way LTE do works in licensed bands: the channel is fully reserved for LTE, under full eNB control. Changing this fundamental assumption would be a major breakage, and much much more effort. The result would be very different. Whereas using LTE carrier aggregation, with a primary channel in a licensed band and using the unlicensed 5 GHz band for downlink only (UL has some added challenges that makes it unpractical) secondary channel(s) with LBT is not too difficult. It's an incremental changes, but does not require breaking a fundamental assumption. And in this case, LTE-LAA could perform like other technologies (to be checked, but possible on principle).

So even if it fits with the operator wishes, the assisted part is also the sane way to add 5 GHz support to LTE from a pure technological point of view.


>> Changing this fundamental assumption would be a major breakage, and much much more effort.

That's non of the concern of the unlicensed spectrum.The LTE-LAA guys should do their work and come prepared for what is common in the unlicensed world. Not the other way around.

And open competitiveness is key in the unlicensed band.


Unfortunately, this is not set in law. And really, that's the level where this should be enforced. You can count on firms to pursue their own self interest to the extent allowed by the law. I'm not saying that's good, I'm saying that's what happens BTW (don't shoot the messenger ;). Here, cellular operators are seeing all this "free" spectrum, and cellular equipment makers are building a path to enable them to use it in a practical way for them, and sell their wares in the process. All legally, as the law is today.

BTW, there are precedents. In the IoT/M2M space Sigfox uses an unlicensed band around 900 MHz (from memory), and you can't buy their base stations: they make them for their own use, as their business is to be an operator. If you want to compete with them, you need to develop your own equipment based on a different scheme. WiFi and BT are the best known techs working in unlicensed bands, but there are proprietary systems too.

The legal constraints on the unlicensed bands are varying depending on country. But as far as I know the strictest constraints are:

- limited transmission power (LTE-LAA won't be allowed to shout above others at 5 GHz, at least where this constraint exists);

- requirements to do listen-before-talk (LBT), at least in Europe (but not in the US);

- limit on the amount of consecutive time a given device can hold the medium (5 ms max.).

LTE-LAA (not LTE-U) will be able to comply with all of this. In this sense, strictly limited to legal constraints, it will behave as is common in the 5 GHz band. To put more restrictions would need to happen at a legal level. And not surprisingly, there's intense lobbying on this topic happening right now from all parties (Google paper is part of this). To be followed...


Sorry, this will be off-topic.

Let's say I am a run of the mill IT guy and I wanted to work my way up into understanding and maybe operating/admin work on cellular networks, how are guys in the US and abroad getting into this so I understand implementation and later policy/regulation stuff like this more intimately?

Years ago I considered after college crap work with ATT and Verzion starting on the bottom ladder just to learn, and every time there is a life crisis I wonder if I can have a just fuck-it moment and try getting into it.

Any tips?


You would probably need a good overview of the technologies in use. I can very much recommend the book "From GSM to LTE"[0]. It goes through all the networks back from GSM and up to LTE, which are needed, since all are in fact still in use.

After that you could probably keep an eye on telecom news sites to follow the ongoing things and regulations.

That should at least get you going.

[0] https://books.google.dk/books/about/From_GSM_to_LTE.html?id=...


Very cool stuff, thanks for the tip. Will find a copy and buy it.


Bunch of college classmates went to work at telecom operators, because, after all, we were majoring in Telecomm Engineering. It was basically a mix of EE and software engineering, with specific telecom classes thrown in. First one was called "Transmission and propagation", very low-level, radio and cable layer, heavy on physics and math. It would help you if you went to work for an equipment manufacturer or to choose radio tower sites. Next ones were higher level, one for how a GSM network works: BTS, BSC, MTS and all that stuff. Then a 3G (brand new at the time) data network overview.

This would prepare you enough, apparently, to work for a mobile operator as an engineer.

Also, most telecoms have a lot of IT/developper people, because they operate and develop a lot of services. You could easily get in on that.


Do you know of places where I can get no-name Telecom Engineering degrees online? I would not even care about the rep.

I had friends in Egypt actually majoring in that, and stupidly it was long before I cared. OR I definitely would have gone to make copies of books at local copy shops like we did for the rest of our classes.


I am amazed commercial exploitation of the unlicensed bands is even on regulators radar. If you are providing commercial use of the spectrum, you have to get off that band.

It seems that "citizens band" radio was much better "branded" than wifi. Perhaps we should call it "citizens wifi"


Unlicensed means just that - anybody can use the band for any commercial or non-commercial purpose using any radio technology (not just WiFi) provided they conform to the regulated power output etc. There are already plenty of commercial users of unlicensed spectrum including subscription-based WiFi.

Now, where you are right IMHO is that there is a kind-of hypocracy by the carriers here. For many years they opposed unlicensed spectrum because they thought it devalued their licensed spctrum asset. Now they have done a complete U-turn and want to colonise unlicensed spectrum instead.


  If you are providing commercial use of the 
  spectrum, you have to get off that band.
Goodbye workplace, cafe and hotel wifi?


I think it is different. The usage of the spectrum is what should be free or not - I can sell anything I like over the spectrum, but the access to the spectrum should be citizens wifi - free and unfettered, except to the extent others try to use it (tragedy of commons is coming of course)

So it's fine to sell internet access over that spectrum ... i think I am turning myself round here.


I wouldn't mind if wifi had to be at no additional charge or free with service.


Well the wifi should be certainly. The issue is the internet backhaul can be chargeable. Which to all intents and purposes says LTE providers can come along and offer high powered radio, and effectively block out any low power wifi routers.

So, if anything this seems like "denial of service" problem more than citizens band wifi.

But I still say the branding is needed.


Usually workplace and cafe wifi are offered for free.


So are most Google services, yet they're still commercial. A cafe and a workplace offers Wifi as part of their commercial activities (to attract customers in the first case, and as a working tool in the second).


I've thought about trying to do long reach WISP stuff in the 900 MHz band for similar reasons - its too hard/expensive to get affordable licensed spectrum. I'd like to see more unlicensed spectrum out there to make this easier, ideally lower than 2.4 GHz


The 900MHz band is great if you can use it, since for the same power 900MHz goes substantially farther than 2.4 or worse 5GHz.

The old inverse square law is 'in a vacuum' Once you your signal is going through air, drywall, masonry, or my favorite steel re-enforced concrete, you're looking at exponents of 2.5 (just air) to 3.5 or worse. This only get more sad the high frequency you go.


When you start lowering the frequency and going further, you can have a path loss exponent of up to 4, so you have to be wary of reflections from the atmosphere. [1]

Radio engineering is really interesting.

[1] https://en.wikipedia.org/wiki/Path_loss#Loss_exponent


multipath is a pain in the ass - so are atmospheric inversions (which effects much higher frequencies).




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: