LoRa Range Testing in San Francisco(blog.beepnetworks.com)
blog.beepnetworks.com
LoRa Range Testing in San Francisco
http://blog.beepnetworks.com/2016/09/loras-wireless-range-is-bananas-a-first-look-at-cellular-for-iot-in-san-francisco/
64 comments
Shawn from Beep here. It's definitely an interesting area for us. We've created a tracker demo but don't have any information on our site about it yet.
I love the idea of inductively charging the dog tracker when the dog is in bed (though we're unlikely to pull that off anytime soon given the state of wireless charging).
You can do a lot with battery life if you only turn the GPS radio on when the user wants to locate the item. This type of product requires really good network coverage. We could make one last for 4-5 months on a single AAA battery.
If you want it to check in every 10 minutes with GPS data it would only last for 3.5 days. If you get smart with an accelerometer you can push that out to 1-4 weeks or more depending on how much the thing you're tracking moves around.
It's a fun product area with lots of parameters to tweak!
I love the idea of inductively charging the dog tracker when the dog is in bed (though we're unlikely to pull that off anytime soon given the state of wireless charging).
You can do a lot with battery life if you only turn the GPS radio on when the user wants to locate the item. This type of product requires really good network coverage. We could make one last for 4-5 months on a single AAA battery.
If you want it to check in every 10 minutes with GPS data it would only last for 3.5 days. If you get smart with an accelerometer you can push that out to 1-4 weeks or more depending on how much the thing you're tracking moves around.
It's a fun product area with lots of parameters to tweak!
Seems like the police could bait bikes with this technology to make a serious dent on bike theft in their city.
They could already do that with APRS if they wanted to.
FWIW this already exists, just over cellular networks instead. Works well enough for the pet use case though: http://www.whistle.com/ is one, I think there are a couple others also.
Shameless plug: Have a look at my GNURadio OOT transceiver if you're interested in LoRa and SDR. https://github.com/BastilleResearch/gr-lora
Here is a summary of my black box/blind signal analysis that informed its design. https://github.com/matt-knight/research/tree/master/2016_08_...
Here is a summary of my black box/blind signal analysis that informed its design. https://github.com/matt-knight/research/tree/master/2016_08_...
I'd like to see how it does in rural areas. There's a gap in the market for GPS horse trackers because you can't count on any particular communications network, unless you use licensed amateur radio APRS.
Rural areas would generally be much better due to the lower number of objects in the way. Now, that will give you less options for bouncing signals, but supposedly, you can achieve ~13 miles with line of sight.
My rural area (the northern frontier of Appalachian mountain range) doesn't have long lines of sight. Around here the high points are around 2000 ft above sea level and most of the population is in a network of valleys that dissect what would otherwise be a plane. Hilltops are wooded too (even on the many hills named "Bald Hill") so you have to add another 40 ft to your tower just to look into two adjacent valleys.
If, on the other hand, this hardware mounted on one of my buildings can cover my 70 acre farm and a bit of the valley around it, I could find a use for it.
If, on the other hand, this hardware mounted on one of my buildings can cover my 70 acre farm and a bit of the valley around it, I could find a use for it.
Ideally these would be mesh networking and cheap enough to place in several spots on a ranch or farm. This way you could place half a dozen and get full coverage easily. Alas, I've read too many of the IoT radios of late and am not sure if the beep kits are mesh enabled.
I am highly skeptical of mesh networks.
Maintaining one radio is one thing, maintaining ten of them is another. Each one has a battery to worry about as well, can go bad, can be misconfigured, etc. Whereas if I have one radio on the roof my house or barn there is less to go wrong.
There is also no promise adding more nodes to a mesh network will improve performance since the nodes can interfere with each other. There is a huge literature on mesh networks because the area gets huge amount of military funding, and maybe the military really can use them, but I think they are a solution looking for a problem in an environment where people act as if wires were ritually unclean.
Maintaining one radio is one thing, maintaining ten of them is another. Each one has a battery to worry about as well, can go bad, can be misconfigured, etc. Whereas if I have one radio on the roof my house or barn there is less to go wrong.
There is also no promise adding more nodes to a mesh network will improve performance since the nodes can interfere with each other. There is a huge literature on mesh networks because the area gets huge amount of military funding, and maybe the military really can use them, but I think they are a solution looking for a problem in an environment where people act as if wires were ritually unclean.
Your skepticism around deploying lots of hardware vs one central transmitter is valid. But I'm kind of confused by your comment about wires. It sounds like you're implying that a wired network will somehow make the equipment attached to it more reliable.
@whoa generally wired networks have been preferred in most industrial-scale sensor networks to ensure reliable connectivity. Until very recently I wouldn't have even considered doing an active control network over a wireless network. But recent IoT hardware and mesh networks have gotten surprisingly good!
@PaulHoule those are some good points. I would've agreed up until recently, but the newer IoT mesh radios have gotten surprisingly robust. The frequency modulators and collision detection baked into the last few iterations of lower cost IoT chips (like DigiMesh/Nordic chips) really have improved. We've been using DigiMesh 900 MHz units in a greenhouse setting with great success (especially compared to 2.4GHz which drops packets with abandon), with no measured packet loss to date. Pretty sweet!
To wrap that up, low-throughput but hi-scalability mesh networks combined with better firmware, network configuration, tuned mesh protocols, and proper software support to make the "throw a few of these down and forget them" scenario a real possibility. Check out the Beep networks blog for a longer explanation: http://blog.beepnetworks.com/2016/09/messy-networks-for-the-...
@PaulHoule those are some good points. I would've agreed up until recently, but the newer IoT mesh radios have gotten surprisingly robust. The frequency modulators and collision detection baked into the last few iterations of lower cost IoT chips (like DigiMesh/Nordic chips) really have improved. We've been using DigiMesh 900 MHz units in a greenhouse setting with great success (especially compared to 2.4GHz which drops packets with abandon), with no measured packet loss to date. Pretty sweet!
To wrap that up, low-throughput but hi-scalability mesh networks combined with better firmware, network configuration, tuned mesh protocols, and proper software support to make the "throw a few of these down and forget them" scenario a real possibility. Check out the Beep networks blog for a longer explanation: http://blog.beepnetworks.com/2016/09/messy-networks-for-the-...
Sounds like an interesting challenge. Hit us up if you'd like to try out Beep's kit -- daniel at beepnetworks dot com.
I'd be interested in near-shore(a few miles out and cell service is unusable/you don't have a high thoroughput requirement). There are plenty of "nearish" ocean applications that could benefit from an extended range without requiring iridium(where packet costs can be a killer).
What ocean application are you talking about?
Shawn from Beep here. These use-cases sound really interesting. Send me an email? shawn at beepnetworks dot com
Hey what are your thoughts on sigfox? They got an investment from Samsung who are apparently integrating into their devices. Will this play out like vhs vs betamax?
the overall verdict from Lora alliance members on Sigfox is that they are big on words but small on revenue or customers. add grain of salt
~~Disclaimer : Sigfox employee ~~
Big on words, but not speaking about others ;)
I'd suggest anyone interested in LPWAN (be it Sigfox, LoraWAN or else) to get its own idea by field testing. Talk is cheap. Sigfox coverage is available in 25 countries, including some large US cities (SF, NYC, Chicago, Atlanta, ..) http://www.sigfox.com/coverage
Get in touch if interested by more info about Sigfox, solutions for developers, etc.. http://makers.sigfox.com
Big on words, but not speaking about others ;)
I'd suggest anyone interested in LPWAN (be it Sigfox, LoraWAN or else) to get its own idea by field testing. Talk is cheap. Sigfox coverage is available in 25 countries, including some large US cities (SF, NYC, Chicago, Atlanta, ..) http://www.sigfox.com/coverage
Get in touch if interested by more info about Sigfox, solutions for developers, etc.. http://makers.sigfox.com
To see more coverage tests with heterogenous devices on the things network (https://www.thethingsnetwork.org/), see http://ttnmapper.org/.
The most interesting place to see is in the Netherlands, followed by the Zürich area in Switzerland.
In my experience it's easy to get miles of range when you have line of sight. Just look at the measurement from nearby Lucerne to Zürich (Switzerland). That's around 50-60km. If you have buildings in the way, things get tougher. Putting up gateways high on the roof helps.
The Things Network Forum (https://www.thethingsnetwork.org/forum/) is always a good place to talk about these things.
The most interesting place to see is in the Netherlands, followed by the Zürich area in Switzerland.
In my experience it's easy to get miles of range when you have line of sight. Just look at the measurement from nearby Lucerne to Zürich (Switzerland). That's around 50-60km. If you have buildings in the way, things get tougher. Putting up gateways high on the roof helps.
The Things Network Forum (https://www.thethingsnetwork.org/forum/) is always a good place to talk about these things.
I'm rural and have a lot of things sending data to a central node for collection. I found LoRa interesting, but sadly none of the modules offered native encryption capabilities. Ended up using RFM69HW modules with AES128 in the same band and still get pretty good range, up to a mile+ NLoS (but not through 300m hills either..)
I imagine the low throughput of LoRa would make encryption a greater challenge than with other modules, but maybe there are other reasons for that.
I imagine the low throughput of LoRa would make encryption a greater challenge than with other modules, but maybe there are other reasons for that.
Some of the modules have dedicated native encryption coprocessors such as the Laird RM186 - it's still easy enough to shoot your toe off though. I would argue there is negligible benefit for an encryption coprocessor (it's usually just AES ECB anyway) - you're either better off using a software implementation of ChaCha20/Poly1305 or a dedicated TPM.
LoRaWAN requires AES128: https://www.lora-alliance.org/portals/0/documents/whitepaper...
The radio chip itself doesn't perform the encryption though, it's done on the host microcontroller.
What kind of things are you monitoring?
The radio chip itself doesn't perform the encryption though, it's done on the host microcontroller.
What kind of things are you monitoring?
If you are curious about tropospheric ducting this[1] site shows openings determined by looking at what packets are coming through APRS igateways. Some of the LoRa stuff runs on UHF so it could be subject to similar openings(like the huge one over the midwestern US right now.
[1] http://aprs.mountainlake.k12.mn.us/
[1] http://aprs.mountainlake.k12.mn.us/
Troop is not going to be so strong as it is in 2 meter (140-144 MHz) band but 900 can bounce strongly off meteor trails, the moon, etc. Reflections off airplanes could be a big deal too.
Sporadic long range channels could be a bug and not a feature for this kind of system because if you are not a ham who thinks it is a blast to talk to someone 500 miles away on a a handheld, you want a radio system to work predictably and not be interfered with by distant stations.
Sporadic long range channels could be a bug and not a feature for this kind of system because if you are not a ham who thinks it is a blast to talk to someone 500 miles away on a a handheld, you want a radio system to work predictably and not be interfered with by distant stations.
Is it still the case that LoRa is a single vendor proprietary technology? When I looked into it a bit it seemed like only Semtech produced LoRa chips and that it wasn't an open standard like 802.11, so it is unlikely that we will see offers from other vendors.
I believe Microchip licensed the IP from Semtech and has made their own chips: http://www.microchip.com/design-centers/wireless-connectivit...
Shameless plug.
I am working as consultant for a startup building a backend for LoRa devices. If you have any question from the backend software point of view feel free to write me :)
email in the profile.
I am working as consultant for a startup building a backend for LoRa devices. If you have any question from the backend software point of view feel free to write me :)
email in the profile.
How does this sort of tech handle large scale use? Living in an apartment where all but one corner is in range of 60+ traditional access points has been a huge pain for my connectivity. I can only imagine it would be worse when you're in range of everybody in miles.
Honestly I am deeper in the stack, in mg understanding however, is the gateway (the thing that receive the radio message) to filter out message belonging to your network. (please note that I may be wrong)
The gateway is extremely simple it receive messages and send them to a bigger server (usually via Ethernet).
Of course the scale of your problem is dependent of the legislation... In Europe (where I work) there is a duty cycle of 1% (if you talk for 1 second you need to don't talk for the next 100 seconds).
Also keep in mind that device usually trasmit at 14 dBm WAY lower than the 27 dBm in the article. Every 3 dBm half the actual power...
The gateway is extremely simple it receive messages and send them to a bigger server (usually via Ethernet).
Of course the scale of your problem is dependent of the legislation... In Europe (where I work) there is a duty cycle of 1% (if you talk for 1 second you need to don't talk for the next 100 seconds).
Also keep in mind that device usually trasmit at 14 dBm WAY lower than the 27 dBm in the article. Every 3 dBm half the actual power...
It's a great point. A smart network needs to scale down the transmit power of its nodes and gateways as the network density increases.
Hummm actually is more about the distance (in the radio topology) between the nodes and the gateway. Closer are the gateway to the nodes less power you need.
Yes, so if you have a uniform distribution of nodes and gateways over space, a node's nearest gateway gets closer as network density increases.
Definitely, but the philosophy that I usually see is that once your gateway cover an area you may want to add a second one (to cover in case of failure) but no more.
So the distribution is not uniform... On the other way, once an area is covered, more and more nodes are added.
So the distribution is not uniform... On the other way, once an area is covered, more and more nodes are added.
Anyone remember Metricom?
900mhz, miles of range, dial-up modem speeds.
900mhz, miles of range, dial-up modem speeds.
Yes, yes indeed. Changed my life in 1998 when I no longer had to be in the office when on Pager Duty. Just velcroed a ricochet modem to my laptop and I was good to go from anywhere. We kind of take that convenience for granted these days.
I find this technology incredibly exciting. I'm waiting until cell phones use it in conjunction with mesh networks to create wide adhoc networks in places without cell service. There's some fun app ideas there.
Unfortunately the spectrum that's available for this type of thing varies significantly country-by-country, so it's unlikely to be added to a major phone maker.
Unfortunately the spectrum that's available for this type of thing varies significantly country-by-country, so it's unlikely to be added to a major phone maker.
As somebody working in the IoT space (specifically telemedicine), Lora is the technology that excites me the most. It really does sound like a magic bullet that really actually does solve a lot of real connectivity problems.
Go beep!
BTW, so, uh...hey beep, wanna sponsor a hackathon for this stuff this spring?
Go beep!
BTW, so, uh...hey beep, wanna sponsor a hackathon for this stuff this spring?
Thanks!
You mean in 9 months? We're hoping to do one earlier than that :). Hit me up: shawn at beepnetworks dot com
You mean in 9 months? We're hoping to do one earlier than that :). Hit me up: shawn at beepnetworks dot com
There are limitation to do with the Duty Cycles in LoRa networks which we've found inhibits certain real time applications. It effectively limits the number of "messages" you can send during a specific time frame.
Shawn from Beep here. Are you referring to LoRaWAN networks specifically? I'm always down to talk wireless schemes. Hit me up (email in my profile).
We go beyond LoRaWAN with our own protocol for some use cases.
For those following along: LoRa is a wireless encoding technique (PHY layer), LoRaWAN is a protocol for how a large group of devices should structure their packets (MAC layer).
We go beyond LoRaWAN with our own protocol for some use cases.
For those following along: LoRa is a wireless encoding technique (PHY layer), LoRaWAN is a protocol for how a large group of devices should structure their packets (MAC layer).
what SF and bandwidth were you using for the range tests? what packet lengths?
True in Europe (CE) but it works differently in the US (FCC).
The FCC limits time spent transmitting in a single channel ("time on air"), but there's no duty cycle limit if you follow the channel hopping rules.
The FCC limits time spent transmitting in a single channel ("time on air"), but there's no duty cycle limit if you follow the channel hopping rules.
I wonder what the frequency of operation is -- 315, 433 or 869 MHz?
868 MHz band in Europe
915 MHz in the US.
EDIT: Actually I suppose since it isn't LoRaWAN, it could be 433 MHz.
EDIT: Actually I suppose since it isn't LoRaWAN, it could be 433 MHz.
433MHz isn't unlicensed spectrum in the US, its in an amateur radio band. Wouldn't be legal for commercial use like this.
Shawn from Beep here. We use the 902-928MHz band in the US.
Directionality. If you have a single gateway with tons of power for tx, is it ultra sensitive for RX and what the requirements on the end devices from a power perspective?
Yes -137dBm sensitivity: http://www.semtech.com/wireless-rf/rf-transceivers/sx1272/
In the blog post we were using 27dBm transmissions. At that level the radio and power amplifier draw about 1.5W. If a transmission takes .3seconds that's .5Watt-seconds. So with a single AA battery (about 5Wh) you can do 40000 transmissions.
Transmissions can be much shorter in time depending on radio parameters, by a factor of 32 at most, but range can be affected.
In the blog post we were using 27dBm transmissions. At that level the radio and power amplifier draw about 1.5W. If a transmission takes .3seconds that's .5Watt-seconds. So with a single AA battery (about 5Wh) you can do 40000 transmissions.
Transmissions can be much shorter in time depending on radio parameters, by a factor of 32 at most, but range can be affected.
What's the throughput like for LoRa?
Extremely low. Useless for any kind of demanding applications like web browsing or media, but very serviceable for simple data packets or even a basic TTY terminal.
Think a few bytes per second
Some are working on decoding the LoRa protocol with an RTL-SDR: https://revspace.nl/DecodingLora .
I wonder if Beep has done any testing with Mavlink. I think 50kbps is pushing it unless you turn off packet repeats.
Bernal heights is labeled incorrectly. Its north of 280, not south of it.
Ah you mean on the topo map, not the Google map at the top. Thanks.
If the transmitter could be inductively charged in a dog crate or where I park my bike -- it would be very convenient too. Ping time could be much slower, even once a minute would be plenty fast. Appears you could get 24 hours transmit and gps without too much trouble in a pretty small package.