Hacker News new | ask | show | jobs
by bri3d 1929 days ago
Worth noting that the maximum bitrate of the base LoRa encoding is not going to replace your cellphone anytime soon, even for a fantasy re-hash of the text-based Internet that this article suggests.

I believe the maximum speed of LoRaWAN on 900Mhz spectrum is a blazing 27 kbps (that's bits), so the cited 80Kb/s in the linked article for Sidewalk-to-IP communication is several orders of magnitude higher and must contain a lot of (unsurprising) overhead.

LoRa is good for applications where it used, like meter monitoring, control systems (oilfield etc.), and RC airplane control (R9/Crossfire/Ghost). It could certainly be used for the proposed motion detection and lighting use cases. With modern codecs, you could maybe complete 1-2 voice calls at a time over it, maybe. But my guess is that Amazon's play here is "smart home without the WiFi configuration," not "replace your cell phone."

It's not going to replace your cell phone data plan.

11 comments

But it's sufficient for my next TV to report fingerprints of my viewing habits back to HQ, even when I intentionally leave the TV disconnected from my LAN. Yay.

It was inevitable we'd reach this point—if not because of mesh networking, then because of ever cheaper cellular radios. And modern cars are already here. But we're fast losing control and visibility over which of our consumer electronic devices are allowed to talk to the outside world.

I tried (and failed) to build a GoTenna competitor using LoRa and what you say here is 100% correct. Large scale mesh networks are incredibly difficult to build and often end up requiring extreme optimizations for specific use cases. We ultimately abandoned mesh networking in favor of a TDMA approach with base stations.

“smart home without the WiFi configuration” is exactly what Amazon’s network is for, but it won’t be anything more than that. The bandwidth and latencies required for content rich applications is simply not there. Sidewalk is cool enough without trying to sell the magic mesh network pipe dream.

> 27 kbps (that's bits), so the cited 80Kb/s in the linked article for Sidewalk-to-IP communication is several orders of magnitude higher and must contain a lot of (unsurprising) overhead.

That’s not several orders of magnitude, that’s only 3×. Both figures are kilobits per second.

I would also mention that the Amazon Sidewalk thing is for a hybrid of Bluetooth Low Energy and 900 MHz, and it’s quite plausible that that 80 Kbps could only be achieved over the close-range Bluetooth and not in the long-range 900 MHz frequency. As an outsider to the industry with no specific knowledge of what actually caps LoRaWAN’s speed, I’m going to wildly guess that this 900 MHz band, in whatever guise, may be more likely to yield 10–20 Kbps speeds in good conditions.

Hey, that's more than an order of magnitude in binary! Great way to make a doubling (or greater) sound bigger than it is.
Still not several. :D
"Order of magnitude" is one of my linguistic pet peeves. I've seen it used referring to base 2, base 10, base 1000, and base 1024. What does it all mean? In base 2^(1/100), trebling is many, many orders of magnitude increase!
“Order of magnitude” is context-dependent, like a great many things in natural language. In the absence of any contrary context, it’ll mean a decimal order of magnitude in English. It will do you no good to rail against the inclusion of cultural context in resolving the meaning of language (and even its structural parsing!), because it’s so very widespread in English and I presume in every other natural language (though logical languages could potentially theoretically evade it).

I’ve never encountered non-integral bases in real life, but https://en.wikipedia.org/wiki/Non-integer_base_of_numeration tells me they are sometimes used. Fun stuff!

Agreed, railing against language and cultural context is literally futile. Factoradic is the one true base. There, "order of magnitude" would depend not only upon (my very unique) cultural context but also the absolute magnitude involved (at the low end, "double" is an order of magnitude; then "triple", then "quadruple", etc).
The only logical base to use here is 2.71828182.
I prefer to say “several times larger/greater” and it is even fewer syllables! “Order of magnitude” works better for convincing people to do something your way, though.
Exponential too, which as far as I can tell just means non-linear, at least in media.
It doesn't even always mean that. It often just means "a lot bigger"
Pretty sure they're getting the 80Kb/s over LoRa. I've used some Pycom devices [0] to try the protocol out and for mesh [1] and it's rather forgiving. Given the number Ring devices people seem to have these days I'm sure Amazon is going to have some relatively large mesh networks at their disposal.

[0] https://pycom.io/products/supported-networks/#lora [1] https://docs.pycom.io/pymesh/

The fastest LoRaWAN data rate specified in the US is Data Rate 13 LoRa: SF7 / 500 kHz: 21900kbit/s.

Even dropping outside of the highly restrictive LoRaWAN specification and going with the fastest physical parameters available for LoRa on the 900Mhz bands (on-time restrictions aside), BW500 (500kHz bandwidth), Spreading Factor 6, Coding Rate 4:5, you end up with 35700 baud.

You can drop these modems into 2FSK mode and increase speed, but at that point the range isn't particularly appealing.

Regardless, the point stands: not a mesh network WAP internet redux, an IoT tool.

> 21900kbit/s

Correction here, you meant 21900 bits per second (21.9 kibibits per second in SI terms), not kilobits per second.

My fault, I thought the figures in the article were kilobytes (and even then, I suppose it's just a single order of magnitude in base-10!). I don't think this mistake affects my point.
The universal convention is: b = bits, B = bytes. I see errors only rarely, and they’re almost always when people write Mb or Gb instead of MB or GB (just MB, not MB per second or such). It also fascinates me how we conventionally write Mbps (with a p), but MB/s (with a /).
It also has a proprietary PHY protocol, which always struck me as an major downside to something whose adoption is closing to making it the next de facto standard. DASH7 [1] is an interesting alternative in this regard, good for urban areas, but not quite as long range for very sparse nodes in a rural environment. It does not the same duty cycle limitations that LoRa has and is actually used to complement LoRa even on the same device in some interesting case studies [2] which come from Semtech themselves (the patent holders on the LoRa PHY).

[1] https://dash7-alliance.org/ [2] https://tech-journal.semtech.com/making-the-most-of-the-unli...

A network for Amazon devices is certainly one use case. Another is selling network access to other IoT devices.

I would expect most residential broadband TOS would explicitly prevent reselling their network bandwidth/access. That's what you are doing with networks like Helium, even if it is in the form of a token instead of dollars. Amazon has gotten around this by just not paying. You buy the Amazon device, you provide the network access, Amazon gets the revenue.

That makes sense, I had wondered about the crypto angle on Helium. Then again, even if you were relaying a lot of messages from sensors etc. I wouldn't expect it would actually add up to a very big percentage of your total usage (I guess it'd be 24/7, unlike your Netflix/Zoom consumption). So it seems a little implausible your ISP would care (or notice), unless they wanted to get into that business themselves?
Yes, Amazon already offer a Sidewalk SDK as part of AWS IoT offerings. I didn't even think about the cost angle, which is a really interesting point (although, I think ISPs have an argument against the Amazon devices here still as it's effectively connection sharing, which they also usually ban in ToS). My consideration was just for the customer sales pitch, which is "your IoT devices Just Work magically."
Also, with LoRaWAN, devices have duty cycle limitations. They can't hog all the bandwidth for a extended period of time. It is really intended just for short, bursty, infrequent transmissions like you describe.
That limitation is not specific to LoRa but to anything transmitting in ISM band e.g. ~868 MHz in EU, ~900 MHz in US. It limits single transmitters air time to 1% so that one can build radio communication with any modulation, any protocol and limits probability of collision with different devices in range.

I wonder if ISM band will provide dedicated spectrum for LoRa with unlimited airtime.

I'm pretty sure that's not true for 2.4 ghz part 15 devices in the U.S. (i.e. 802.11 wireless).

(I had thought that ISM referred specifically to the 2.4 ghz band, but I guess there is actually more than one band. I do find it funny that that one of the most heavily used spectrum bands is the one where we put all the unintentional radiators like microwave ovens and industrial and medical devices, and then the FCC decided that we might as well let people do unlicensed transmission on that "junk band" because it wouldn't be interfering with anything "important". It's sort of the policy equivalent of the common phenomenon where over time the most important services eventually often end up running on the oldest, slowest computer.)

https://en.wikipedia.org/wiki/ISM_radio_band

I don't think that applies to 900MHz in the US. It definitely does to 433 and I think 315, though.
433 and multiples are harmonics.
Most of the ISM bands are harmonically rated (they are referred to as junk bands for obvious reasons)
I think this depends on the region. In the US you shouldn't exceed a dwell time limit on a single channel.
I don't often see FPV related knowledge on HN. I've wondered if others were aware of these implementations and it's cool to see them mentioned here.
What about super basic internet and sms like applications ?

I thinking mostly on stuff build on scruttlebutt or Gemini.

The first one is build for on/off crappy connection. The second is basically gopher on steroid.

Any input would be welcome

Would be enough for reading Hacker News, about 3000 characters per second.

HN frontpage seems to be about 50 kB so it would take about 3 s to load on a 128 kbit link.

Maybe Amazon is looking for a way to go back to the old Free Internet Kindles without having to pay carriers.
So basically for IoT.