Hello.
We are developed 8 channel SX1301 based gateway many years ago. For now we are soldering them and using in our hardware solutions utilizing LoRa technology.
But for now we got request to develop solution that have two key difficulties:
Expected traffic from end nodes to be 100x - 1000x higher that our usual solutions.
Difficulties with mounting and providing electricity to gateways.
Continuously sending traffic from Gateway to nodes on one of the channel (RTK amendments).
So we have to achieve higher capacity of network with lower amount of gateways (ideally with just a single gateway).
So we googled and found that two different solutions exists:
LR-FHSS
16-/64- channel gateways
Problem is that both thing doesn’t have any existing open source solutions in Internet.
LR-FHSS gateway reference design is not publicly available. Request to Semtech support was rejected by them with suggestion to buy LoRaWAN gateway from another manufacturers.
But we don’t want do that cuz we are using our own hardware. And even if we will decide to buy such gateways seems there are no any stock LR-FHSS gateway exists for now.
Completely unclear how to implement 16/64 channel gateway. It is not explained in Semtech’s website at all so we are googling information bit by bit but it is quite often contradictory. Someone says that specific and very expencive FPGA acquired directly from Semtech need to be used for that purpoces.
We found few comments here at TTN forum: 1, 2 that advices workaround to use two SX1302 on different channels or spread factors.
So we decided to move in that way and developed experimental gateway that looks like on the photo:
two slots for pcie modules of pin to pin compatible SX1301/SX1302/SX1303 based RF cards. Each of which will recide on separate SPI (separate NSS pin).
But something still not clear for us so we would like to ask for any advices from people ho has experience of solving similar problems - significant LoRa network capacity encreasing using stock and chip ICs (SX1301/SX1302/SX1303).
Does SX1301 is applicable for that purpoces (pseudo multi channeling) as well as SX1302/SX1303?
What is preferable - splitting by channels or by spread factors?
Does it possible to mix SX1301 (that uses packet forwarder) cards with SX1302/SX1303 (that uses SX1302_HAL) somehow ?
Is it reasonable to use more than two cards same time (three or four SX130x cards) ?
Not clear do same packet-forwarder/SX1302_hal can/should drive all cards (e.g. splitting in configuration file) or each card should be driven by separate packet-forwarder/SX1302_hal ?
Could someone please share working configuration files for such 16+ channel packet-forwarder/SX1302_hal solutions?
Depending on your usual solution that can be easy to extremely hard. With short transmissions and sufficient channels it can be done. However given your ‘less gateways’ statement there will be larger distances to cover resulting in increased spreading factors and longer transmissions requiring even more space in the frequency/time continuum.
Continuously sending is just not an option for any spread spectrum technology. You will hit either issues with LBT, dwell time or transmission time limits because you need to send on one well defined frequency and spreading factor for any chance of nodes being able to receive the transmissions. And even then, that won’t be possible with LoRaWAN as it interferes with the uplink/downlink cycle of your node. (A node can’t continuously receive multicast and uplink as well because of the downlink window after an uplink)
Of course when continuously transmitting you will need a different antenna and transmitter for the purpose (a simple node level chip should work) because otherwise the receivers won’t get any signals (Gateways don’t receive while transmitting because the RF paths can’t be used for both at the same time as a transmission would damage the receivers. So when transmitting the receiver circuit is isolated from the antenna to prevent damage)
What do you mean by this? By definition gateways receive multiple spreading factors on the same frequency. The issue when using them will be that the SX chips can only decode a limited number of parallel transmissions. So three nodes transmitting on the same frequency with different spreading factors is possible, however it will use 3 of the SXs decoders resulting in less being available to decode other transmissions. I am not sure you can instruct the SX chips to receive and decode only a subset of the spreading factors on a certain frequency. So having two of them listen at the same frequency will probably result in both of them using decoder resources to decode the same transmissions.
Every card requires to be driven by a packet forwarder. The HAL is just an hardware abstraction layer and does not receive and forward data. You could potentially adapt an existing packet forwarder to drive multiple cards or use multiple packet forwarders in parallel if your hardware has two totally separate SPI channels and not just a different chip select for each card.
You are pushing the limits so I guess there won’t be many people able to answer your questions. Specifically on the TTN forum where most participants do not design and build gateways.
I’ve worked on a packet forwarder some years ago and don’t see an obvious way to make your requirements work. Keep in mind the LNS needs to know about the special features of your gateway as well. I’m not sure the current TTS stack can handle > 16 channel gateways let alone setups where channels are split on spreading factor.
