Broadband Across The Congo

If you live in much of the world today, high-speed Internet is a solved problem. But there are still places where getting connected presents unique challenges. Alphabet, the company that formed from Google, details their experience piping an optical network across the Congo. The project derived from an earlier program — project Loon — that used balloons to replace traditional infrastructure.

Laying cables along the twisting and turning river raises costs significantly, so a wireless approach makes sense. Connecting Brazzaville to Kinshasa using optical techniques isn’t perfect — fog, birds, and other obstructions don’t help. They still managed to pipe 700 terabytes of data in 20 days with over 99.9% reliability.

This appears to be one of those problems that seem simple until you dig into it. Even though the link is only around 3 miles in distance, to get the performance required, the company claims:

Imagine pointing a light beam the width of a chopstick accurately enough to hit a 5-centimeter target (about the size of a US quarter) that’s 10 kilometers away;

The project requires atmospheric sensing, control of mirrors, and motion detection. In a pilot in India, the same system had to even deal with monkeys shaking the equipment.

We couldn’t help but wonder how this all stacks against alternative technology, for example wireless and low Earth orbit systems would seem to offer stiff competition with different tradeoffs.

Of course, if you drop your expectations, you can get by with less hardware. Hams have been known to do quite a number with common laser gear.

34 thoughts on “Broadband Across The Congo

      1. This historically has been the issue with all lite infrastructure projects. A corporation needs access to a market and builds out a system with the largest short term profits. There is simply no substitute for hard infrastructure because with current technology the internet can provide neither clean water nor physical goods. If pipelines, plumbing and roads already existed, the internet becomes an issue of digging a new ditch.

  1. Would love more details on this. Are the two endpoints fixed in place, or are there intermediate repeaters? How is power handled? If they’re fixed sites, why not a traditional microwave link?

    1. There are FSO (Free space optics) links in the market, long before Ronja. Nowadays you can find 10Gbps links, and even more by aggregation. This Google article is just ridiculous marketing.

    1. I honestly don’t know, but I’m willing to speculate:

      The 60 GHz ISM band exists (just like the 2.4 GHz one) because the absorption by water vapor is quite high in that band.
      This, over strongly heating riverbeds (daily **means** are above 22 °C / 73 Freedom degrees every month) together with 14 raindays in November might make the attenuation at the much lower 60 GHz worse than it is for the optical frequencies.

      Fig. 4 in patent US10333617B2 (assigned and filed: X Development LLC, formerly Google X, who also own all the Loon patents) suggest that what they’re expecting to get at the receiver is more than – 5 dBm in received power – that’s humongous, immense, crazy much if you’re coming from an ISM or cellular downlink background. In that case, the key ingredient might simply be Power (with a capital P), as your ISM network isn’t allowed to do enough gain × transmit power to ever come close to that, whereas regulations for optical transmitters I guess are mostly based on “can I through positioning make sure nobody manages to look into the beam with their remaining good eye”. Local regulations apply – but never underestimate how much rules can influence the technical direction you have to take.

      1. At 60GHz molecular Oxygen would be the dominate peak, attenuating at about 1dB/km
        (ref: https://en.wikipedia.org/wiki/Extremely_high_frequency#Propagation )

        12 km would be ~12dB of attenuation in addition to the normal “free space path loss” at 60GHz which would be about 149.6 dB (for 24km it would be 155.6 dB FSPL + ~24dB attenuation from molecular Oxygen), which is why high gain antennas are used in the ~$400 device.

        So in the Congo where high levels of photosynthesis is producing pesky oxygen during daylight hours would increase the attenuation at 60GHz.

      2. Absorption by water at 2.4 GHz isn’t particularly high. It was just a suitable frequency for making microwave ovens smaller, since the oven is a cavity resonator that needs to be sized and shaped according to the excitation wavelength. Larger industrial ovens operate at 915 MHz which is also an ISM frequency.

        The peak dielectric loss (absorption) for water is actually between 10 -100 GHz depending on temperature.

      3. >X Development LLC, formerly Google X, who also own all the Loon patents

        yeah, about that. The reason Loon was terminated was lawsuit from the real inventors (Space Data Corp.) who allowed Google engineers to “inspect” their company pre purchase under NDA, purchase that never happened because Google decided to outright steal everything.

        https://www.wired.com/story/the-lawsuit-that-could-pop-alphabets-project-loon-balloons/

        https://9to5google.com/2017/07/10/project-loon-patents/

        https://www.law.com/therecorder/2019/07/29/google-settles-ip-case-over-high-altitude-balloons/

    2. 60GHz link availability is quite low above 2km. However in some circumstances you can get dozens of km using big enough dish antennas. But when humidity arises, no antenna gain can compensate the high loss and the link goes down. Just do the math

  2. wonder why they dont do arrays of these things. as it would be very unlikely that multiple beams would be broken simultaneously. you would also need significantly lower specs from the unit as they could operate in parallel.

  3. > In a pilot in India, the same system had to even deal with monkeys shaking the equipment.

    After ip over avian carriers in africa ( https://hackaday.com/2009/09/12/homing-pigeon-email/ ), we have physical realisation of chaos monkeys ( https://en.wikipedia.org/wiki/Chaos_engineering#Chaos_Monkey ). Android systems already have proper functions to check for monkeys ( https://developer.android.com/reference/android/app/ActivityManager.html#isUserAMonkey() )

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.