Orcasound's orcastream

This software contains audio tools and scripts for capturing, reformatting, transcoding and uploading audio for Orcasound. The directory structure reflects that we have developed a base set of tools and a couple of specific projects, orcanode and orcamseed (in the node and mseed directories). Orcasound hydrophone nodes stream by running the node code on Intel (amd64) or Raspberry Pi (arm32v7) platforms using a soundcard. While any soundcard should work, the most common one in use is the Pisound board on either a Raspberry Pi 3B+ or 4. The other project (in the mseed directory) is for converting mseed format data to be streamed via Orcanode through the Orcasound human & machine detection pipeline. This is mainly used for streaming audio data from the OOI (NSF-funded Ocean Observatory Initiative) hydrophones off the coast of Oregon. See the README in each of those directories for more info.

Background & motivation

This code was developed for source nodes on the Orcasound hydrophone network (WA, USA) -- thus the repository names begin with "orca"! Our primary motivation is to make it easy for lots of folks to listen for whales using their favorite device/OS/browser.

We also aspire to use open source software as much as possible. We rely heavily on FFmpeg. One of our long-term goals is to stream lossless FLAC-encoded data within DASH segments to a player that works optimally on as many listening devices as possible.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See the deployment section (below) for notes on how to deploy the project on a live system like live.orcasound.net.

If you want to set up your hardware to host a hydrophone within the Orcasound network, take a look at how to join Orcasound and our prototype built from a Raspberry Pi3b with the Pisound Hat.

The general scheme is to acquire audio data from a sound card within a Docker container via ALSA or Jack and FFmpeg, and then stream the audio data with minimal latency to cloud-based storage (as of Oct 2021, we use AWS S3 buckets). Errors/etc are logged to LogDNA via a separate Docker container.

Prerequisites

An ARM or X86 device with a sound card (or other audio input devices) connected to the Internet (via wireless network or ethernet cable) that has Docker-compose installed and an AWS account with some S3 buckets set up.

Installing

Create a base docker image for your architecture by running the script in /base/rpi or /base/amd64 as appropriate. You will need to create a .env file as appropriate for your projects. Here is an example of an .env file (tested/working as of June, 2021)...

AWS_METADATA_SERVICE_TIMEOUT=5
AWS_METADATA_SERVICE_NUM_ATTEMPTS=0
REGION=us-west-2
BUCKET_TYPE=dev
NODE_TYPE=hls-only
NODE_NAME=rpi_YOURNODENAME_test
NODE_LOOPBACK=true
SAMPLE_RATE=48000
AUDIO_HW_ID=pisound
CHANNELS=1
FLAC_DURATION=30
SEGMENT_DURATION=10
LC_ALL=C.UTF-8

... except that the following fields are excised and will need to be added if you are integrating with the audio and logging systems of Orcasound:

AWSACCESSKEYID=YourAWSaccessKey
AWSSECRETACCESSKEY=YourAWSsecretAccessKey
 
SYSLOG_URL=syslog+tls://syslog-a.logdna.com:YourLogDNAPort
SYSLOG_STRUCTURED_DATA='logdna@YourLogDNAnumber key="YourLogDNAKey" tag="docker"

(You can request keys via the #hydrophone-nodes channel in the Orcasound Slack. As of October, 2021, we are continuing to use AWS S3 for storage and LogDNA for live-logging and troubleshooting.)

Here are explanations of some of the .env fields:

• NODE_NAME should indicate your device and it's location, ideally in the form device_location (e.g. we call our Raspberry Pi staging device in Seattle rpi_seattle.
• NODE_TYPE determines what audio data formats will be generated and transferred to their respective AWS buckets.
• AUDIO_HW_ID is the card, device providing the audio data. Note: you can find your sound device by using the command "arecord -l". It's preferred to use the logical name i.e. pisound, USB, etc, instead of the "0,0" or "1,0" format which can change on reboots.
• CHANNELS indicates the number of audio channels to expect (1 or 2).
• FLAC_DURATION is the amount of seconds you want in each archived lossless file.
• SEGMENT_DURATION is the amount of seconds you want in each streamed lossy segment.

Running local tests

At the root of the repository directory (where you also put your .env file) first copy the compose file you want to docker-compose.yml. For example, if you have a Raspberry Pi and you want to use the prebuilt image, then copy docker-compose.rpi-pull.yml to docker-compose.yml. Then run docker-compose up -d. Watch what happens using htop. If you want to verify files are being written to /tmp or /mnt directories, get the name of your streaming service using docker-compose ps (in this case orcanode_streaming_1) and then do docker exec -it orcanode_streaming_1 /bin/bash to get a bash shell within the running container.

Running an end-to-end test

Once you've verified files are making it to your S3 bucket (with public read access), you can test the stream using a browser-based reference player. For example, with Bitmovin HLS/MPEG/DASH player you can use select HLS and then paste the URL for your current S3-based manifest (.m3u8 file) to listen to the stream (and observe buffer levels and bitrate in real-time).

Your URL should look something like this:

https://s3-us-west-2.amazonaws.com/dev-streaming-orcasound-net/rpi_seattle/hls/1526661120/live.m3u8

For end-to-end tests of Orcasound nodes, this schematic describes how sources map to the dev, beta, and live subdomains of orcasound.net --

-- and you can monitor your development stream via the web-app using this URL structure:

dev.orcasound.net/dynamic/node_name

For example, with node_name = rpi_orcasound_lab the test URL would be dev.orcasound.net/dynamic/rpi_orcasound_lab.

Deployment

If you would like to add a node to the Orcasound hydrophone network, contact [email protected] for guidance on how to participate.

Built With

• FFmpeg - Uses ALSA to acquire audio data, then generates lossy streams and/or lossless archive files
• rsync - Transfers files locally from /tmp to /mnt directories
• s3fs - Used to transfer audio data from local device to S3 bucket(s)

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests.

Authors

• Steve Hicks - Raspberry Pi expert - Steve on Github
• Paul Cretu - Lead developer - Paul on Github
• Scott Veirs - Project manager - Scott on Github
• Val Veirs - Hydrophone expert - Val on Github

See also the list of orcanode contributors who have helped this project and the [Orcasound Hacker Hall of Fame] who have advanced both Orcasound open source code and the hydrophone network in the habitat of the endangered Southern Resident killer whales.

License

This project is licensed under the GNU Affero General Public License v3.0 - see the LICENSE.md file for details

Acknowledgments

• Thanks to the backers of the 2017 Kickstarter that funded the development of this open source code.
• Thanks to the makers of the Raspberry Pi and the Pisound HAT.
• Thanks to the many friends and backers who helped improve maintain nodes and improve the Orcasound app.