Skip to content

AirStack 2.0.0

 

23 December 2024

 

The Deepwave Digital team is happy to announce the release of AirStack 2.0. This is a major release that includes significant new features to enable novel applications on the AIR-T as well as an update to the underlying operating system on the AIR-T. AirStack 2.0.0 provides support for new [AIR-T Embedded Series](https://deepwavedigital.com/hardware-products/air-t-embedded-series, which utilizes the NVIDIA Orin GPU.

AirStack 2.0.0 incorporates NVIDIA JetPack 6.0. NVIDIA does not support the TX2 or TX2i in Jetpack 6.0, therefore Deepwave AIR-T models AIR7101, AIR7201, and AIR8201 should continue to use AirStack 1.0.

New Features

  • NVIDIA Orin 16GB Support - We have implemented and tested our support for the NVIDIA Orin 16GB GPU module allowing for improved computation capability and the latest AI software packages.
  • JetPack 6.0 - NVIDIA JetPack 6.0 is a comprehensive SDK for building AI applications on Jetson modules, featuring an updated Linux kernel, Ubuntu 22.04, and the latest AI compute stack, including CUDA 12.2 and TensorRT 8.6. Full highlights of these changes can be found here.
  • Linux For Tegra 36.3 - NVIDIA Jetson Linux 36.3 includes Linux Kernel 5.15.
  • Ubuntu 22.04 LTS - The rootfs is derived from Ubuntu 22.04.5.
  • CUDA 12.2 - NVIDIA's CUDA has been upgraded to 12.2.1.
  • cuDNN 8.9 - NVIDIA's cuDNN has been upgraded to 8.9.4.
  • TensorRT 8.6 - NVIDIA's TensorRT has been upgraded to 8.6.2.
  • Triton Inference Server Containers - Performing AI inference using NVIDIA Triton is easy now using the Triton Docker Containers.
  • CuPy v13 - CuPy is a Python library for GPU-accelerated numerical computing, offering a NumPy-compatible interface for fast array manipulations and mathematical operations, making it ideal for signal processing tasks like filtering, Fourier transforms, and convolution. The cupyx.scipy.signal toolbox provides GPU-optimized implementations of common signal processing routines, enabling efficient execution of advanced techniques such as digital filter design, signal transformations, and spectral analysis directly on the GPU. See our updated GPU Signal Processing Tutorial for a great demonstration.
  • Python 3.10 - AirStack 2.0 has native support for Python 3.10.
  • Native Rust Support - AirStack 2.0 has native support for Rust. The Rust programming language is powerful for embedded applications because it combines low-level control, memory safety, and zero-cost abstractions, enabling developers to write efficient, reliable, and safe code for resource-constrained environments.
  • RF Signal Path API - The AIR7310-B RF daughtercard has multiple signal paths available. Included in AirStack 2.0.0 is an API to control these paths through SoapyAIRT for both RX and TX. See the updated Tutorials for more info. See the receiver and transmitter sections of the AIR7310-B user guide for information on the signal paths:

Tutorials

  • The Deepwave team has worked hard to provide a comprehensive set of updated tutorials for AirStack 2.0.0. They may be found here. These include tutorials for how to leverage the new features and APIs found in AirStack 2.0.0.

Availability

All new AIR7310-B AIR-Ts ship with AirStack 2.0 pre-installed.