▲ 166 points
back
91 comments
It's interesting to see the huge channel bandwidths they're using to attain this. Advances like this are largely driven by higher sample rate ADCs and DACs becoming more viable in recent years.
Edit for clarity: channel bandwidths in the datasheet are up to 2 GHz. Need to close the link at 32 QAM to hit 10 gb before error correction overhead at that bandwidth, which is certainly doable. Also, it's interesting to note that they quote 7 gbps throughput @ 1 GHz bandwidth, and 10 gbps at 2 GHz. It implies that they can run at 256 QAM at 1 GHz, but only 32 QAM at 2 GHz, which also makes sense.
Having worked in the satellite communications industry, where channel sizes used to be limited to 72 MHz (because of hardware limitations of the spacecraft), getting modems designed to operate on larger channels is no small task. I would love to learn more about the internal architecture of these radios to understand exactly what's going on - are they using interleaved ADCs? Is the modulation/demodulation being done in an FPGA, or an ASIC?
If there are terrestrial microwave engineers on here, I'd love to hear your thoughts on this!
terrestrial microwave engineer here: "80 GHz" is actually 71-86 GHz FDD. The original FCC band plan allowed for 5000 MHz wide channels each direction and basically OOK or BPSK level modulation. Newer radios use 250, 500 or 1000 MHZ wide FDD channels and QPSK or better. Incredibly wide channels can be used because it falls off in the atmosphere so rapidly after a few km, and the antennas are all very narrow parabolic reflectors with less than 1 degre beamwidth.
Basically no colocation interference issues are possible unless two companies try to shoot from the same rooftop, to also the same rooftop, using the same channel AND the same linear polarity.
Thanks for jumping in! Sounds like it's really nice for urban small-cell deployments, campus networks, etc. Do you experience a lot of issues with keeping links aligned at those tiny beamwidths?
The specs sound very similar to what the BCM8500 can do. Add an analog frontend and an antenna, and that's basically it.
https://www.broadcom.com/products/Microwave-%26-Mobile-Backh...
You can pack many bits into a constellation point. Think 1024 QAM.
Most new licensed band 6, 11, 18, 23 GHz radios these days are 1024QAM capable. For 80 GHz the hot new thing is radios capable of 16/64/256QAM at varying code rates.
That's a pretty narrow ledge to stand on @ 800 GHz. Wonder what the FEC regime is?
> Advances like this are largely driven by higher sample rate ADCs and DACs becoming more viable in recent years.
I'm looking forward to what these will do for SDR. I salivate over the thought of an SDR using something like the TI ADC12J4000 [1] which has a 4 GSPS sampling rate.
The digital hardware to handle the output of a 4 GSPS ADC is beefy as hell. This part has integrated DDC's if you're happy with downsampling, but if you want full nyquist data coming out of this, you need to handle 6 Gbyte/Sec.
>higher sample rate ADCs and DACs becoming more viable
Question, on optical networks doesn't this make dispersion more of a problem?
80 GHz isn't quite optical yet. It's a couple octaves down from infrared. SFAIK, it's still very much radio.
What sort of latency would you expect with a system like this (I mean the link in this submission, not satellites)?
FWIW, I've got several similar point-to-point microwave links in my network, including the following setup:
h1 - sw1 - ptp1a -/- ptp1z - sw2 - ptp2a -/- ptp2z - sw3 - r1 - h2
h1 : Sun Fire server
sw1 : Cisco 3560G switch
ptp1a : 6 GHz point-to-point microwave transceiver ("A" end)
ptp1z : 6 GHz point-to-point microwave transceiver ("Z" end)
sw2 : Cisco 3560G switch
ptp2a : 6 GHz point-to-point microwave transceiver ("A" end)
ptp2z : 6 GHz point-to-point microwave transceiver ("Z" end)
sw3 : Cisco 3560G switch
r1 : Cisco 3945
h2 : Sun Fire server
ptp1a to ptp1z : 12.80 miles (20.60 km)
ptp2a to ptp2z : 05.01 miles (08.06 km)
--- x.x.x.x ping statistics ---
20 packets transmitted, 20 received, 0% packet loss, time 19305ms
rtt min/avg/max/mdev = 0.955/0.995/1.157/0.049 ms
Point to point wireless links like these usually have low latency, some under a millisecond.
• Secure communication due inability to intercept the laser-like beam transmission at free air
I would not trust that as a security layer.
It's definitely not something to trust in, but intercepting (at a layer 1 level) a PTP 80 GHz link is actually harder than tapping fiber. You'd have to have Rx equipment either directly in the path or directly behind both ends of the radio link.
As compared to the effort required to cut an aerial or underground singlemode cable and fusion splice in place a passive prism split tap (basically the same thing as inserting a split in a GPON FTTH network). A practiced outside plant fiber crew of 2 persons and a bucket truck could do this with less than 5 minutes of downtime on a router-to-router optical interface, short enough time to clear any NMS alerts and prevent a repair team truck roll. Assuming we're talking about only two strands.
Either way actual security is accomplished through standard based crypto, not obfuscation or preventing people from messing with the layer-1.
I haven't read the datasheet on this particular radio, but a lot of them do optionally support AES encryption (sometimes as an extra, paid feature).
This would get really bad rain fade at 80GHz or even on a humid day the speed would back off a lot. You need to run a lower frequency backup link in parallel.
That's the nature of 80 GHz, design the links for your climate and don't try to go more than 2-3km. You can achieve five nines. And yes, run a 5.x GHz backup path in parallel.
> You can achieve five nines.
As long as nobody pilots a drone in the path.
Crap, now I need to go buy a couple of these and lobby the nearest data center for some rooftop space.
That is the short way of saying I had no idea you could get antenna this effective for wireless data transmission. I'd seen the 5mbps ones but nothing close to a gigabit much less 10 gigabits. Time to draw a 10km radius circle around my home address :-)
I've seen up to about 400 Mbps on my Webpass [1] connection for a couple years, and I think that's pretty run-of-mill tech at this point [2]
There are solutions that are lower end than this one, but still gigabit, at relatively affordable prices.
This Ubiquiti setup is $999 per side: https://www.ubnt.com/airfiber/airfiber5/
The NSA Bluffdale facility might just squeak in that line for me. Think they'd hook me up?
Metrolinq has a 1GE product using unlicensed 60 Ghz which is like $500 a pop.
What kind of licensing (FCC or otherwise) would one need to operate such a radio bridge in the US? or is it public spectrum?
80 GHz is "light licensed" in the US, the paperwork requirements are not onerous. It's less costly and complicated than a regular part 101 licensed microwave link.
The word "Ethernet" in the title is misleading. That word does not appear in TFA, which appears to discuss a radio transceiver.
[EDIT:] I stand corrected; thanks!
It has a 10GbE SFP+ optical interface and functions as a layer 2 ethernet bridge. From an ethernet port perspective same as 99% of the other PTP microwave and millimeter wave radios on the market. What's new is the 10Gb (vs existing radios with 1Gb SFP).
Read the datasheet linked at the bottom of the page.
They're transparent Ethernet bridges. They have Ethernet interfaces, you push an Ethernet frame in one side and the same Ethernet frame comes out the interface on the other side.
I have line of sight to the building where I work, I would love something like this so I could get gigabit internet at home w/o paying through the nose :)
Check out Mimosa's offerings.. For a few hundred dollars you can get ~750mbps via unlicensed spectrum up to ~5km distance.
https://www.mimosa.co/Products/Backhaul/backhaul-specs/B5-Li...
I know Ubiquti has similar products as well but I'm more familiar with Mimosa's.
Most employers would never let you hook that up - too much risk for zero gain (for them).
Hell mine won't even let me bring my own screen to work. (It lacks some enterprise cert apparently).
it's worth noting that this is just the first one publicly announced and not under NDA.. It's from a russian radar/microwave/millimeter wave manufacturer. All of the other much larger players such as Bridgewave also have 10 Gbps radios coming.
Those BW 80 GHz links are pricy$$$, but like anything the cost will come down in a few years.
Its actually bit curious, they don't mention being Russian anywhere on their site and even their "Contacts" page refers the company being registered in Sweden.
Ooh, I really want two of these. I'm stuck on an island with poor last mile service providers. But, I have line of sight to a number of spots with good quality fiber.
If you've got line of sight then current (cheap) tech should do just fine. Sure not 10x gigabit but you can def get some decent internet...
I love Ubiquiti's products. They're easy to setup and cheap too. From experience, they're also quite reliable.
The company's community forums can be quite helpful too.
What kind of link are you after? (Speed, distance, how far over water, are both sites powered, etc)
Interesting.
There has been a huge effort in the UK for mobile carriers to add fiber to as many cell towers as possible.
Do people think this would undo this trend? I'm sure that 10gbit would be more than enough to carry the backhaul of 3G+4G with plenty of room to spare?
not really, fiber is still greatly preferred (For example: it's impossible to do 40Gb by microwave/millimeter wave, but a CWDM 4-channel 4x10GbE passive mux/demux on two strands of singlemode is trivial and very very cheap), or just a pair of 40Gb QSFP 10km reach optics using a single 1550nm wavelength between two routers or metro-E switches. But wireless can reach small cells, rooftops and towers that might be a very costly underground fiber build at $400-900/meter total construction cost to dig up streets in urban cores. It very much depends on the location.
What is the price for this?
Do these products require a clear line of sight to work?
Yes.
According to Nielsen's law of bandwidth, consumer gigabit last mile should be common by now and 10G would be a logical next step development, becoming ubiquitous in 2020 or so.
There's a huge number of houses with idle fiber installed 10+ years ago. Gigabit ethernet was introduced 17 years ago and carries 5 km over fiber. Cable is just waiting for providers to switch on 10 Gbps since years ago. Phone line copper has similar story..
Maybe something like this could jump-start the stalled development of last-mile consumer internet.
> There's a huge number of houses with idle fiber installed 10+ years ago.
Do you have a source for that? Verizon may have passed 18 million homes, but only the homes which have placed an order for Fios were ever connected. So, no or very little idle fiber.
> Gigabit ethernet was introduced 17 years ago and carries 5 km over fiber.
Wrong. You can buy off the shelf SFPs with 200 km reach. Use amps if you want longer reach.
> Cable is just waiting for providers to switch on 10 Gbps since years ago.
You don't just switch on 10 Gbps on cable. First your vendor needs to release DOCSIS 3.1 equipment and you need to test it. Then you need to upgrade your CMTS to DOCSIS 3.1 and swap out any cable modems that don't support 3.1 and dedicate spectrum to DOCSIS 3.1 downstream channels.
And it's not since years ago. DOCSIS 3.1 was released a bit over two years ago. Comcast will start rolling out commercial DOCSIS 3.1 service this year.
> Phone line copper has similar story...
What?! G.fast can theoretically give you 1 Gbps, but only if you already have fiber to your driveway. There aren't even any commercial deployments yet and all vendors don't even have products yet.
You can put the savings into better bandwidth for consumers or lower cost for ISPs. With no competitive pressure for the former in the majority of the country, providers have happily pocketed the 3+ order of magnitude improvements while providing the same old 1990 speed.
And here I am with 12MbS down/800KbS up