I think if you consider ttn gateways most of them are bought by people without any business case or outlook of ROI at all. Besides the same that goes for the gateway prices could be argued for the price of the relatively expensive nodes. Appr 7 or 8 euro for an RFM is pretty expensive compared to alternatives.
Wether the gateway costs 200 or 300 euros doesn't really matter to me, because you really do need only one (and maybe a handfull more from friends or sponsored ones in the next 5-10 kilometers to cover your town). I would argue (like @Sylvain_M and @TijnOnlijn before me) that the price per node is way to high atm. You can get a esp8266 for 3-4 euros in china and there is the network stuff already on it, while the rfm95w still costs 7+ euros (without a microcontroller). And trying to buy lora stuff in europe results in node prices starting around 50 euros - if you build your own and try to get a arduino pro mini + rfm95w + suitable pcb together for under 30 euros you are hard pressed. And i/we could use dotzends/hundreds of nodes per gateway.
If node chips were cheaper, say $2 like other transceivers it would be worth buying 38 of them to build a true multichannel gateway (6 spreading factors * 6 frequencies)
You could throw in an ESP8266 on each do Mesh between them and still come out cheaper than most commercial ones.
For those of us looking for lower cost gateways - all others please look away - I found a new SX1301 board just under $100 (single unit price!) on Taobao:
Sadly it's only 433Mhz so we'd need a new band plan for these in TTN (I understand it's coming) and it'd be incompatible with 868Mhz nodes.
I'm noticing this trend where 433Mhz transceivers and now gateways are much cheaper than 868/915Mhz counterparts, nodes in particular are almost half the price (E.g. SX1278 modules from AiThinker are currently sold online for $3 in china)
It certainly hasn't, also I wan't talking about only 8 modules but 38..
Then please indulge me how you are going to circumvent the multiple SF/range and RF interface problem?
With 38 modules you can assign one to each frequency and SF combination so the range issue doesn't apply.
The RF interface problem? As long as you make sure all your modules only either receive or transmit - and ideally only one transmits - and never have both happening at the same time it should be fine.
I mean you only really need one antenna to receive - though it will take some effort maintaining the correct impedance - and as in real SX1301 gateways you can then have an RF switch when you need to do a transmission.
So yeah it's not easy, but debunked? Far from it.
Technically everything is possible, but going for such an exotic option just to save a few bucks of the gateway costs seems rather far stretched to me.
It wouldn't just be about saving a few bucks, but - at equivalent cost - having a fully open source gateway (no closed, binary only, firmware to load like on the SX1301) unencumbered by NDAs and as far as I can tell legal since it would use Semtech's own silicon.
Also would be an extremely cool project in my book and would look incredible
The chips on a concentrator board are somewhere in the 50 to 70$ range. Filters, powersupply and stuff maybe another 15, Antenna and stuff another 10. So were in the range of 75 to 100$ now. Without any margin and only for the concentrator. Now you need the processor, LAN / Wifi controller and stuff. And finally an enclosure. So were at let's say 120$ now. Ad some margin, initial costs for the production and stuff. And your at 200$, without earning that much. Somebody needs to pay for your released magic blue smoke, wrong parts, prototypes on the way there.
Thanks, nice explanation. If only those chips for the concentrator boards would become cheaper.
Power supply and filter should be possible for $10. For antenna and (SMA) connector $3.50 is more realistic than $10 (not sure what you mean with 'and stuff'). Processor, memory, LAN and WiFi is available for as low as $7 (Orange Pi Zero), let's make that $10. All prices mentioned are based on existing Chinese products (which already include production costs and margin). Add some $10 for a plastic enclosure. I round off your '$50 to $70 for concentrator board chips' to $65. That makes a total of around $98.50. My assumption is that much of the required software is probably already available. Based on this I think a consumer grade indoor LoRa gateway for around $125 (excl. VAT) can be possible.
As long as you sponsor my (non-commercial use) gateway I have no problems with costs of gateways.
no way (you should never start your own business)
In the hardware world, if do it to get food on your and your employees tables, you need to sell it three times the price you pay for the parts. Because you iterations costs, you have initial costs, you start with small expensive batches....
Right! That's why I don't have my own hardware business!
For makers and enthusiasts we should have a cheaper Open Source LoRa gateway (not targeted at commercial use). I'm convinced the community will then drive its further development. Arduino, ESP8266 and Raspberry Pi are nice examples of such communities.
The TTN community efforts for implementing (Multi-)Single Channel Packet Forwarder type LoRa gateways (wannabees) on different MCU platforms already are a good example.
I would not mind having to compose my LoRa gateway from separate readily available components and construct my own custom enclosure if I can do that for some $100.
Right, but mine is an optimistic buyer's view, not the seller's view.
I do try to stay realistic and share what I think is possible (under certain conditions).
I'm not referring to equipment for the professional market, nor am I referring to low(er) quantity custom electronic equipment development. I'm referring more to 'common off-the-shelf' hardware (components) stuff, consumer grade equipment for the 'consumer' market in high (enough) quantities. Other consumer electronics pricing and Chinese electronics supermarkets serve as my reference.
(Fyi I'm not trying to make fun of electronics engineering business here.)
From the people I know who are 'experimenting' with LoRa in their spare time, some 50% would very possibly have bought a LoRa gateway already if they could do so for not more than $100. Currently only one of them has privately bought a TTN gateway (and one commercial gateway was bought by a company). The others are still using Single Channel Packet Forwarder variants as workaround wishing it were a true gateway.
This isn't "hardware business", it's the custom PC assembly business.
For a gateway you're buying three parts - 2 off the shelf and 1 which is just PCB - and putting them together with an enclosure and then installing the same software over and over.
Trust me no one who assembles PCs for a living sells for 3x the cost.