WARP Project Forums - Wireless Open-Access Research Platform

No, the WARPnet framework isn't currently compatible with the 802.11 Reference Design, but we are prioritizing remote management and data collection for the 1.0 release of the 802.11 Reference Design.

Our first control/testing framework for the 802.11 Reference Design will be based on WARPLab 7. I know that sounds odd- WARPLab has always been intended for raw waveform I/O. But the architecture of WARPLab 7 is actually very similar to the original WARPnet. The WARPLab ref design happens to move I/Q samples, but we designed the WARPLab 7 framework to be much more general. Each node running the 802.11 ref design will be a WARPLab node, controllable from an m script. Each node will use ETH A for wired/wireless bridging and ETH B for WARPLab I/O. Critically, WARPLab will not be required to use the 802.11 design, but will only enable extra functionality.

We think it will provide the same utility as the original WARPnet framework, without the hassle of installing/maintaining the warpnet_server. Your (and other researchers) feedback will be very useful once it's posted.

As for the 1.0 release, we're still working on a few things:
-Adding 48/54Mbps rates (the 64-QAM rates) to the PHY
-Adding hooks for the control/measurement framework
-Refactoring the CPU_High code to separate AP-specific code from more reusable code
-Implementing top-level code for a station (STA)

100% of Mango's engineers (i.e. all 3 of us) are working on this. We hope to have the 1.0 design finalized in the next few weeks.

Can Warplab 7 perform gather data in real-time for analysis like Warpnet?

Yes. Well, we think so; early experiments show it's fast enough to keep up with sending 1 wired packet per wireless reception, which is sufficient to support real-time experiments that previously required warpnet co-processors (like ber_processor).

I know WARPLab has a well-deserved reputation for being fairly slow. However we've made huge progress in the past few releases to significantly speed things up, with the explicit goal of supporting new applications beyond the simple waveform Tx/Rx capabilities of the original WARPLab ref design. Take a look at the latest benchmarks; when reading waveforms the WL7.3 design moves data at ~489Mb/s (32k samples/burst * 4 bytes/sample * 8 bits/byte * 466 bursts/sec). This includes all of MATLAB's overhead (which accounts for nearly all of the end-to-end overhead) for moving retrieved samples into the workspace for processing. These benchmarks used raw I/Q for payloads, but the WL7 framework doesn't care what payload it's moving between nodes and the controller.