Difference between revisions of "TTNHABBridge"

From RevSpace
Jump to navigation Jump to search
m
Line 12: Line 12:
 
Next steps:
 
Next steps:
 
* get some example JSON from the TTN application MQTT stream
 
* get some example JSON from the TTN application MQTT stream
* get an MQTT client working and see if we get data
+
* implement the MQTT client and see if we get data
* <del>implement the REST access towards habitat using Java jersey</del>
+
* implement the listener info and listener telemetry upload + cache
** get it clear how the station latitude/longitude/altitude is uploaded to habitat
+
* testing if it works
 +
** Register an app, program a Sodaq One with the proper keys
 +
** Pick a binary telemetry format (reverse engineer the default sodaq software) and implement a decoder in Java.
 +
** Drive around in range of a TTN gateway and see if it works (balloon visible on the map)!
  
 
== Introduction ==
 
== Introduction ==
Line 101: Line 104:
 
* a PUT is done with a "listener_information" doc to /habitat/<the first UUID>
 
* a PUT is done with a "listener_information" doc to /habitat/<the first UUID>
 
* a PUT is done with a "listener_telemetry" doc to /habitat/<the second UUID>
 
* a PUT is done with a "listener_telemetry" doc to /habitat/<the second UUID>
 +
 +
Perhaps implement this with an expiring cache, so listener information/telemetry is uploaded regularly (say every 20 minutes) but not with *every* payload telemetry.
  
 
==== Payload encoder/decoder ====
 
==== Payload encoder/decoder ====
Line 114: Line 119:
 
Libraries to use:
 
Libraries to use:
 
* all application settings are kept in a .properties file
 
* all application settings are kept in a .properties file
* slf4j as the logging interface
+
* slf4j as the logging interface, log4j as the logging implementation
* log4j as the logging implementation
 
 
* junit/mockito for unit testing
 
* junit/mockito for unit testing
 
* jetty/jersey/jackson for the REST interface towards habitat
 
* jetty/jersey/jackson for the REST interface towards habitat

Revision as of 11:13, 17 August 2017

Project TTNHABBridge
Ttnhabbridge.png
A software bridge between TTN and the HAB network
Status In progress
Contact bertrik
Last Update 2017-08-17

Status

Investigating various pieces of software. I plan to implement this in Java, setting it up as a gradle project.

Next steps:

  • get some example JSON from the TTN application MQTT stream
  • implement the MQTT client and see if we get data
  • implement the listener info and listener telemetry upload + cache
  • testing if it works
    • Register an app, program a Sodaq One with the proper keys
    • Pick a binary telemetry format (reverse engineer the default sodaq software) and implement a decoder in Java.
    • Drive around in range of a TTN gateway and see if it works (balloon visible on the map)!

Introduction

This idea is about using the-things-network as a receiver for amateur balloon telemetry.

Receiving telemetry from amateur balloons is currently typically done on the 434 MHz band using RTTY modulation, with dedicated receivers listening for RTTY-modulated ASCII strings. The operator of each receiver has to prepare his radio setup for receiving the telemetry, by tuning to the correct frequency at the correct time, setting up a dedicated software client that decodes the RTTY modulation and forwards the data to a central system over the internet. The central system accepts data from many such receivers, performs deduplication, keeps track of who received what and updates a nice graphical map of where each balloon is and where the receivers are.

A network like the-things-network can help a lot, it has a lot of gateways already (in the Netherlands at least..), already performs deduplication. It uses LoRa as a modulation scheme which is much more sensitive and much less susceptible to slight tuning errors than RTTY.

In short, the idea is:

  • you attach a LoRaWAN transmitter to the balloon
  • the LoRaWAN transmitter is pre-configured with a set of keys generated by the TTN
  • the balloon broadcasts its telemetry every once in a while (say a few times per minute) and this is picked up by one or more TTN gateways
  • the bridge software listens for packets received by the TTN and decodes the payload data into an id, latitude, longitude, altitude of the balloon
  • for each packet, we know which gateways received it and where they are. So we can "fake" a client for each gateway and construct an ASCII sentence according to the HAB server conventions
  • the HAB server still sees the same messages like it would if there were many traditional receivers, so doesn't need any modification!

This way, the entire things network can be used to receive balloon telemetry! There is no longer a need for radio operators to be present at their receiver at the exact time the balloon is launched, making manual adjustments, etc. The Netherlands is already covered by many TTN gateways, greatly increasing the chance the balloon telemetry will be picked up.

Software

Source code

Source code is available at: https://github.com/bertrik/ttnhabbridge

It's nowhere near ready yet. The README explains the tool chain setup.

Tasks

Stuff to do:

  • come up with a simple but flexible way to encode telemetry in a binary packet, to be transmitted over TTN. Perhaps borrow some ideas from how it's done with [1]
  • figure out how to receive data from TTN, this is an MQTT stream -> pick an easy-to-use Java MQTT client library. Perhaps this one: https://github.com/fusesource/mqtt-client ?
  • figure out the protocol between dl-fldigi and the HAB server -> look into https://github.com/ukhas/habitat-cpp-connector , reverse engineer it and create a Java version. It looks like a couchdb-specific database connection, sending JSON messages! -> it seems we can access this interface as if it were a REST service!
  • implement this (in Java for example) and publish it on github!

Modules

The software consists of the following modules.

Main process

The main process of the bridge is something like this:

  • listen to the MQTT stream of the HAB application
  • once we get data:
    • decode the payload into latitude/longitude/altitude, and encode it into a habhub ASCII sentence with correct CRC, see https://ukhas.org.uk/communication:protocol
    • for each gateway that received the data:
      • look up the gateway name/lat/long/alt, through some TTN API (and cache it?)
      • fake a HAB receiver and make it send the ASCII sentence to the habhub server

MQTT listener

Probably implemented using fusesource MQTT client or Eclipse paho MQTT client. We can listen on a certain application and catch *all* traffic for all nodes registered to that application.

Habitat uploader

payload upload

This uses a HTTP interface, performing a PUT to a certain URL with a certain content:

  • the URL is http://habitat.habhub.org/habitat/_design/payload_telemetry/_update/add_listener/<doc_id>
  • the <doc_id> is created from the telemetry sentence by doing ASCII to BASE64 conversion, then hashing using SHA-256 and encoding as ASCII-hex (as LOWER case!).
  • the HTTP method is a PUT with the following headers:
    • "Accept: application/json"
    • "Content-Type: application/json"
    • "charsets: utf-8"
  • The contents of the PUT is the following JSON structure
{
  "data": {
    "_raw": "[base64 of telemetry sentence]"
  },
  "receivers": {
    "[receiver_id]": {
      "time_created": "[timestamp]",
      "time_uploaded": "[timestamp]"
    }
  }
}

The "[receiver_id]" part can be repeated as many times as there are gateways that received the data. This is probably also the trickiest part to implement in Jackson, because it uses variable field names.

listener upload

To figure out how to upload information about the receiver station, I captured some traffic between dl-fldigi and habitat, it seems the following happens:

  • a GET is done on http://habitat.habhub.org/_uuids?count=100, this returns a list of 100 UUIDs, where a UUID is the lower-case ascii hex representation of 16 bytes.
  • a PUT is done with a "listener_information" doc to /habitat/<the first UUID>
  • a PUT is done with a "listener_telemetry" doc to /habitat/<the second UUID>

Perhaps implement this with an expiring cache, so listener information/telemetry is uploaded regularly (say every 20 minutes) but not with *every* payload telemetry.

Payload encoder/decoder

This part decodes the binary payload received from the TTN into a standard habitat sentence, like

 $$hadie,181,10:42:10,54.422829,-6.741293,27799.3,1:10*002A

Gateway cache

We don't actually need this, it seems the required gateway data (id/lat/lon/alt) is already provided by the MQTT stream.

Implementation

The bridge will be written in Java.

Libraries to use:

  • all application settings are kept in a .properties file
  • slf4j as the logging interface, log4j as the logging implementation
  • junit/mockito for unit testing
  • jetty/jersey/jackson for the REST interface towards habitat
  • zip/tar-file for easy installation using the gradle application plugin
  • for the MQTT interface: not sure, either eclipse paho or fusesource

Helpful links

From a conversation on #highaltitude:

20:24 < adamgreig> there's a) a python library that's a lot easier to read
20:24 < adamgreig> but b) basically the gist is you just PUT to http://habitat.habhub.org/habitat/_design/payload_telemetry/_update/add_listener/<id> with some stuff
20:24 < adamgreig> http://habitat.readthedocs.io/en/latest/habitat/habitat/habitat/habitat.views.payload_telemetry.html#module-habitat.views.payload_telemetry
20:24 < adamgreig> so you have a new string, you take the sha256 hex digest of the base64 encoded raw data
20:25 < adamgreig> you PUT to that URL with that ID
20:25 < adamgreig> and you include that JSON struture with your callsign/details

C implementation of the interface between the client and the habitat server

list of habitat JSON schemas

I think/hope the interface above can be approached as a REST service, for which is very easy to create a client (and test server) in Java, using a combination of Jetty/Jersey/Jackson libraries.

Settings

Settings likely needed by the bridge application:

  • TTN gateway API settings
    • URL of the TTN gateway API
  • TTN MQTT API settings
    • URL of the TTN MQTT API
    • application id of the TTN 'hab' application
    • application username
    • application password
  • habitat database settings
    • URL of the habitat server
    • ...

JSON examples

TTN console gateway traffic

In the TTN console, under gateway traffic, the 'Event data' looks like this:

{
  "gw_id": "eui-008000000000b8b6",
  "payload": "QONzlgaABAAB8XtDve5FPT4jnzvJ",
  "f_cnt": 4,
  "lora": {
    "spreading_factor": 7,
    "bandwidth": 125,
    "air_time": 56576000
  },
  "coding_rate": "4/5",
  "timestamp": "2017-08-14T12:12:52.614Z",
  "rssi": -103,
  "snr": 3.8,
  "dev_addr": "069673E3",
  "frequency": 868100000
}

TTN console application data

In the TTN console, under Applications, the 'Application data' looks like this:

TODO

I *think* this data contains also the gateways with their id, latitude, longitude, altitude.

MQTT data

Over MQTT, it looks like this (this is a pretty old sample!):

{
  "payload": "eyJIZWxsbyI6ImJlcnRyaWtAc2lra2VuLm5sIn0=",
  "port": 1,
  "counter": 3,
  "dev_eui": "0000000019800501",
  "metadata": [
    {
      "frequency": 867.3,
      "datarate": "SF12BW125",
      "codingrate": "4/5",
      "gateway_timestamp": 1954388012,
      "gateway_time": "2016-08-06T19:06:05.891844Z",
      "channel": 4,
      "server_time": "2016-08-06T19:06:05.926706718Z",
      "rssi": -118,
      "lsnr": -10.5,
      "rfchain": 0,
      "crc": 1,
      "modulation": "LORA",
      "gateway_eui": "AA555A00080605B7",
      "altitude": 40,
      "longitude": 5.16738,
      "latitude": 52.08393
    }
  ]
}