The Future of the Internet of Things

When buying anything, you’re going to have a choice: good, fast, or cheap. Pick any two. A plumber will fix a drain good and fast, but it won’t be cheap. The skeezy guy you can call will fix a drain fast and cheap, but it won’t be good.

Such it is with radios. You can have long-range (good), high bandwidth (fast), or a low price (cheap). Pick any two. The Internet of Things demands a cheap, long-range radio module, but until now this really hasn’t existed. At Electronica last week, Microchip demoed their IoT solution, the LoRa. This module has a 15km (rural) or ~3km (heavy urban) range, works for a year on two AAA batteries, and is very cheap. Bandwidth? That’s crap, but you’re not streaming videos to your shoe.

LoRaThe LoRa uses a bit of frequency somewhere below 1GHz to communicate with a base station. This base station serves as a bridge for these chips and devices to the Internet. You don’t have the range of WiFi or Bluetooth, either: For the duration of Electronica, the entire eastern half of Munich was covered by exactly six base stations. One base station handled the city center.

This isn’t a first of its kind device – a few folks in Cambridge are working on Weightless, a similar small & cheap radio with terrible bandwidth that communicates with base stations. Weightless is the only protocol so far that isn’t proprietary, but TI, Sigfox, Semtech, and a whole lot of other companies are pouring money into devices like this. This is a huge market, and you simply would not believe the amount of time, engineering, and money that is going into these systems.

Is Microchip’s LoRa the future of the Internet of Things? That’s something for the market to decide. However, GSM will never be low power or cheap, WiFi will never be low power enough, and Bluetooth will never have the range. In a few years, something like LoRa will be found in all those Things for Internet we haven’t been able to build until now.

Here’s a link to the devices in Munich that are still around a base station. My device is 1A 37 64. Find me.

81 thoughts on “The Future of the Internet of Things

    1. yeah the rest of the internet thinks that SSL and fuzzing and code analysis are good things, but the IOT folks have their heads wedged in the same mindset that brought us ActiveX and other gaping sores

        1. STM32M4 has hardware SSL support

          How much does it cost to clean up after a security break-in? This is the actual cost of security. If that “low cost” computer has access to “high value” information assets then its “low cost” is irrelevant.

          1. With a 1200 bps half-duplex wireless WAN, SSL just isn’t going to fly. (the rate that LoRa uses in its range advertising is actually closer to 300 bps). IP isn’t going to fly, nor is 6LoWPAN. So it is up to the implementer to build a good MAC-layer encryption. EAX is a pretty good option for this, because it uses AES and it doesn’t require much overhead.

            The fact of the matter is that the distributed, decentralized IoT that we all want is still a decade away. We can call that IoT 2.0. For the mean time, the sort of build-outs that might use LoRa are going to use a tight, low-overhead protocol stack, MAC-layer encryption, and a network model that sandboxes the endpoints from the internet (i.e. a proxy).

      1. I have spent a lot of time in network and internet security and to me the current situation is a farce.

        When you design a system of operation like the internet without having security as a prerequisite then you are inadvertently pre-selecting the “reject” model of security when at a later stage you wish to “add” security.

        In the “reject” model you have to anticipate every action that may be used maliciously and this is impossible.

        If however, you started with a “permit” model then only those actions that cannot be used maliciously exist in the system to start with.

        Designing insecure internet applications and protocols has an extremely large economy of scale to the extent that secure applications and protocols are considerably more expensive.

        Most clients will want security to be a focus of their product but when they learn of the associated costs they adjust their expectations at the expense of security.

        There is opportunity here with the IoT to get things right from the start, as best is possible given that some of the core protocols already exist.

        When I am asked what SSL, security certificates and “secure websites” do, I say – they will increase your turnover and conversion rate by giving your customers the false perception that your website is greatly more or completely secure.

        1. SSL has pretty good security, but not as it’s configured in the standard browsers that come with lots of certificates pre-installed.

          If you:

          – use your own certificates and require their use,

          – have a mechanism to quickly build and install new copies of libraries as they are released upstream,

          -have a well-thought out workflow system to make sure the security of the keys is maintained and that certificate revocations can be handled swiftly,

          then it is possible to rely on SSL for secure communications.

          But people who need to pay outside consultants for security advice are probably going to ignore most of it anyway.

          1. I acknowledge everything you have said here and agree that with good end to end encryption the weakest point comes down to the security of the private key(s).

            However the reality is that end to end encryption only prevents one specific type of Man In The Middle attack and this is an uncommon attack vector anyway.

            I see many sites that offer ssl only to have credit card details sent to a mail box in plain text where there accessed via POP or IMAP instead of POPS or IMAPS so there goes the end to end straight away. Furthermore bob’s mail access credentials may be username:bob Password:password.

            Most people want and expect a high level of security and with the current methodologies very few people are willing to put in the effort it takes for the security they want. Some way has been made with two factor authentication because it is little effort for the end user.

            What we need to do is offer end users an easy way to securely keep a complex private key and at the same time take away from them the ability to choose dumb credentials like “password”.

            Perhaps this can play out in new WiFi protocols for these devices as they’re new and fresh.

        2. This is just silly, if you want a device to send some slow data you simply do a freaking sanity check on the input and bob’s your uncle.

          You know you don’t HAVE to follow the microsoft model of not doing sanity checks when you don’t work for them right? /S

    2. Interference is also a problem.

      Imagine if the IoT scenarios actually happen, and there are tens of thousands of these devices in everything from flower pots to shoes within the same square kilometer.

      That’s why it won’t happen.

      1. ob disc: I work at a large networking co that is doing IoT stuff. one approach is to reduce data by filtering at the edge and sending ‘up’ only data that fits the filter. this is how IoT can scale and. at least its one co’s approach (called ‘data in motion’)

        1. That doesn’t really solve the issue with people in the area constantly sending commands and requests to some device.

          A three mile radius in a city can fit a million people. If every fridge in, say a big skyscraper, are fitted with transmitters, it’s going to be a hell of a chatter come dinner time.

          1. ROFL, yes agreed about the chatter.

            This is not a new problem, it even existed on very large cable based LAN’s. Even a cable has a limit to it’s bandwidth.

            See Token ring – http://en.wikipedia.org/wiki/Token_ring for an example of how this problem was approached in the past.

            I expect will see similar protocols for WiFi as the bandwidth starts choking up.

          2. “Empty information frames are continuously circulated on the ring.”

            Yeah. No.

            That would require that the transmitters are continuously chattering empty frames in the air instead of sitting passively and waiting, which means any low-power sensor that is supposed to work for a decade on a battery is right out of the question.

          3. Simple.
            1. Set a maximum data rate. The device can only send x bytes per y time.
            2. Devices should most of the time wait to be polled but also have an alert mode. Say a temperature sensor records a temp greater than Y. It should transmit an alert wait for an ack and if no ack retrans the alert.
            3. Keep the data short. For example a door sensor might have just a single byte.
            4. All devices should be polled for things like battery status.

      1. The inAir9 SX1276 from Modtronix is USD15.96 (10+ qty). See http://modtronix.com/inair9-bare.html
        It uses the SX1276 chip from Semtech. Is also available mounted on STM32 CPU board. The Microchip LoRa chip documentation seems not to be available yet. But, looking at previous wireless technologies Microchip licensed from Semtech, the Microchip version of the chip was pretty much identical! Also be careful of some of these cheap, 2 layer LoRa modules! We evaluated some of them, and they didn’t even come close to advertised range!

    1. It would work well, but the reason for that is more because the Semtech/CSEM receiver is very sensitive and has pretty easy-to-use configuration options for using error-correction coding, rather than because of the spread-spectrum modulation it uses (LoRa). In line-of-sight conditions, spread-spectrum modulation doesn’t really help.

  1. Wow, LoRa (for Long Range) is a Semtech technology and SX1272 is on sales for nearly 2 years now. Microchip became a partner to promote LoRa networks but is not at all the owner or the creator of this radio modulation.

    Moreover, Nemeus (a french company also partner of Semtech) proposes also a SX1272-based module with a STM32 as a controller: http://www.nemeus.fr/en/

    You’re also free to implement the LoRaMAC on any controller of your choice:
    https://github.com/Lora-net/LoRaMac-node

      1. Second. It seems like everyone wants to hope on the bandwagon all the while forgetting that there’s a lot of other more pressing stuff that needs to be done. “Love science?!?! Step that step and hope on the bandwagon. It’s okay, everyone else will figure it out sooner or later.” SEP field activate!!

    1. it’s the natural progression of computers to weasel themselves into every application where they become cost effective

      for the sake of my paycheck I better not get sick of better and more capable computers

    2. Not at all. Although, giving it a second thought, there might be some good uses for IoT, and given the fact different persons have different views, there might be something about that. I’m using a small central at home that sends exact time and outside temp via RF12B, and other gadgets (mostly clocks) at my home can adjust their internal time… but still, why a flower pot should be connected to Internet is beyond my comprehension.

      1. Because the answer is not why, but why not. When these ICs cost $2, incorporating them into a nice new Terra Cotta planter falls within the profit margin of that product. Now your pot broadcasts when it needs water and complains to twitter that you don’t give it enough light.

  2. In all seriousness, just how long range does an IoT device have to be? My “internet enabled refrigerator” doesn’t need to report to the local market, just my phone so I can create a shopping list. The radio in my smart electrical meter doesn’t need to reach all the way back to the utility, just to the next node in the mesh network. At what point does perfection become the enemy of good enough?

  3. We’ve seen an attempt of IoT to rise out of low power radio, and it all failed because of lack of collaboration. And this: http://xkcd.com/927/
    Now these kind of technologies try something different but will probably still fail. It’s not that they are bad, just that they want to add new HW where there is no need for new one.

    I’ll just stick to Wifi and Bluetooth, something we all seem to have an abundance of around….

    1. why do you think these things have “failed”? Because you don’t own them? Many of these devices have successful applications right now in manufacturing and warehousing and medical applications. Just because you don’t see an application for your own living room doesn’t mean that these devices are not successful.

  4. Sparkfun don’t seem to get a mention much here on HAD.
    Pity . They have set up a data server free for use to store whatever you want to send with a round robin database system which is going to be very useful for developing IoT ideas and educating the futur “Smarts”
    data.sparkfun.com –Well done Sparkfun

  5. If you try to connect to the device via the usb connector, it displays: CDC Mode Command ( or CDC Command Mode ) an a virtual com port will be installed. But nothing seems to happen if you try to send various commands on different baudrates except that the devices seems to echo the input. After the disconnect the device is trying to: “Joining Network” in an endless loop. After a short time it tells you: Lora Modem Time Out. So be carefull if you want to connect to the usb.

  6. Sooo… if someone implements mesh networking and voip on these things, all it requires is a node for each car/building and we show a middle finger to mobile phone companies, at least in urban areas?

    1. Won’t happen. Mobile companies pay a fortune for bandwidth. Bandwidth is like real estate, the more you buy the less there is left so what you buy is more valuable. Mobile companies are like property investors in real estate except they buy huge proportions of the market. I would be nice though.

      1. the problem with your story is that “mobile companies” are not the lord god above, they do not reign supreme. they dance with the devil but companies like google and apple are always there to push back because google and apple want you to have lots of cheap bandwidth to their devices.

    2. You’re not going to get VOIP over this – the bandwidth is described as “100s” of bits/second. That’s plenty of speed for typing*, if you want to build some kind of community text net, but remember that you’re also limited to the bandwidth of the head end divided by the number of simultaneous users (and that’ll need to cope with protocol overhead to handle multiple simultaneous connections, user ids, security, etc.)

      (* Words per minute of typing speed is about the same as bits/second, and most people type well under 100wpm. Teletypes used to run at 75 baud or sometimes 110.)

      1. Ouch… completely missed the low bandwidth rate, I was thinking of a sort of upgrade of the ESPsomething chip. Then yes, no way to get VOIP or anything else than text messaging. Even enclosing data into hypertext would kill its performance, ie no html: pure binary data plus some hand made marshalling. I wonder how good will be their mesh networking at these conditions.

  7. CC430 with Dash7 is pretty nice for IoT. You can do hub and spoke and avoid mesh/gossip and your gateway can just uart with the CC430. The range is def not quite as large but the power is on par (actually much better) than this chip and smaller range, say the size of a house, apartment complex or office is adequate for many applications where the standard zigbee/wifi/ble crowd doesn’t work.

    https://github.com/CoSys-Lab/dash7-ap-open-source-stack

  8. Something that I think is often overlooked with these IOT projects is the finite nature of the radio spectrum. Even if 10,000 of these devices in a given area can get along with each other on the 900 MHz band, they are unlikely to play nicely with the many other devices also using that band, including ham radio operators who, despite being “secondary” users on that band, would very much prefer not hearing 10,000 IOT trinkets spamming the airwaves with useless information that is being gathered “just because”.

    I’m not saying that there aren’t legitimate uses for this kind of technology, but seriously, if we need to know if the fridge door has been left open, we can get up off of our lazy butts and go look.

      1. Does not work indoors: why not? You mean that there is not country-wide indoor coverage yet? Or do you see specific reasons why it would not work indoors? At least there is already a network in the air in several countries offering country-wide outdoor coverage. Adding indoor coverage means adding more base stations, which only makes sense for dense areas.

        Sigfox is not full duplex: true, but neither is LoRa? Both solutions offer half duplex bidirectional communication. Enough to send data from your device to the cloud and receive commands from the backend.

          1. Guillaume,

            Sigfox can send data bidirectionally. The Telit ME51-868S sends officially already supports this.

            I’m interested in the private outdoor LoRa network you are talking about. Do you have some information on the details of this? What base stations are you using?

  9. Why the fuss with radios when these days one gets free 200m/month data from a t mobile sim? I’ve been running a car tracker with a sim I got for ten bucks over the counter months ago just to verify their free data claim and so far it hasn’t stopped working lol

    1. Because there are regions without GSM coverage, costs (GSM modules are expensive), weight and dimension and power consumption.

      If you think about a high altitude balloon GSM can be very expensive if leafing the lift off country.

      1. What do you mean expensive? Your GSM model just changed this month, no? There is no longer roaming in EU. Same rate all over. So in your example let’s say the balloon DOES lift off at a high altitude where it can tap into your neighboring country tower — no fuss; same rate as the home country GSM “operator” as you call them. I fail to see the expense? Correct me if I am wrong of course!

  10. But… but… this is not going to be any useful for real products unless there is a nationwide or continent-wide deployed and supported network. And this costs money to build, maintain… it seems to me this doesn’t exist yet.

    I’ve heard of traditionnal GSM operators building their own network with technologies similar than sigfox or lora or weightless, but they are going to price it a premium so as not to reduce their revenue…

    It seems to me that the IoT is not there yet, because LoRa doesn’t provide access to the “I”.

  11. Our switzerland based company miromico helped the last 2 years build Team Black Sheep’s Crossfire R/C Model Long Range Link using Semtech’s SX1272 LoRa transceivers. Most other Long Range UHF systems (ImersionRC and lots others) are 433 MHz Narrow Band FSK based. The crossfire uses the 868 or 915 UHF bands. The 2x higher frequency results in a theoretical /2 range disadvantage.
    Some guys made videos comparing the most popular FSK based long range system to the crossfire. The FSK was set to 500mW power and is only unidirectional. The TBS Crossfire was set to 25mW. So to the 6dB frequecy disadvantage comes a 13 dB power difference resulting in a 19 dB stronger signal reaching the distant receiver. The guys placed the transmitter into New York Central stations basement and walked up the stairs and went into the bordering streets. The FSK transmitter using 20 times the power of the LoRa based system lost the battle preaty early. TBS itself flew to 25 km distance using only 25mW. This summer during beta test one of TBS customers controlled his airplane over a record breaking distance of 100km (62mi) only using omnidirectional antennas. Search on youtube for the videos. Transmitters power was 1.2 watts. The link is fully bidirectional with full rate down link telemetry unlike all other systems. And we are not using the 300bps modes cause we have to real time control R/C models. When we only would have to send occasionally some sensor data from a not moving station, we would reach the 100km distance without any external power amplifier and stock antennas.

    So in my view LoRa is outstanding. The IoT LoRaWAN is on a good way as well. Orange, KPN, Swisscom and other providers have rolled out IoT networks in larger cities. The dutch startup TheThingsNetwork wants to build a crowd sourced world wide open LoRaWAN backbone. No roaming and tying to aprovider. Jost power on your device somethere on the world and your cloud app will connect to it. The radio chip is about 1-2 USD in higher quantities. A base station chip can handle 60.000 nodes simultaneous if every node sends a message every hour.

    I think I’ve seen the future of IoT.
    Regards Marcel

  12. After revisiting the LoRa videos I have to add some details to my infos above:
    In the Grand Central Station test the FSK transmitter was set to 600 mW and the LoRa based to 500 mW. Another guy compared range of ImmersionRC’s EZUhf set to 600 mW and LoRa based Crossfire set to 10mW in San Francisco. No misstake: 10 mW vs. 600 mW. Guess what: the 10mW Crossfire won the battle. TBS itself flew a 20 km distance using the Crossfire set to 10mW.
    Here are the videos:

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s