gr-ntnisac: An Open GNU Radio Testbed for NTN-ISAC

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: Today's paper: "gr-ntnisac: An Open GNU Radio Testbed for NTN-ISAC".

Dev: The gist The paper presents gr-ntnisac, an open GNU Radio testbed that applies an emulated LEO channel to a standard 5G New Radio downlink,

Rosa: First, who's behind it and why it matters.

Title and authors: Rosa: So we're looking at this paper today, "gr-ntnisac: An Open GNU Radio Testbed for NTN-ISAC." Basically, it’s presenting this open testbed that lets you take an emulated LEO channel and run it against a standard 5G New Radio downlink <ref:2610.11307#pg1,a standard 5G New Radio downlink>.

Dev: It’s interesting because they let you send the signal through two cabled USRPs and then score the receiver based on the delay and Doppler commanded for every single path.

Taro: What that means for us is that we can test our receiver code against a repeatable truth, which is what you need before you even try testing it on a real satellite.

Rosa: Exactly, Taro. It’s about getting the correct answer known before you try it on actual hardware where things get messy.

Dev: The paper highlights that the same receiver code runs in three places: simulation, on recorded captures, and even on hardware itself to see how it performs.

The paper's summary: Rosa: They’re showing this setup where the channel is modeled digitally before it even hits the DAC. That means they apply the satellite channel information right there on the transmit side.

Dev: The receiver part of this testbed doesn't generate any New Radio sequences; you can plug in any standards-compliant source or even a recorded IQ file instead.

Taro: They model the channel so that range is the only input for each path, described by its range history and gain over time from a trajectory table.

Rosa: And they have these specific path types listed in a YAML scene file, like direct paths, static paths, rotating paths for micro-Doppler, and even custom ones.

Dev: The receiver processes the OFDM symbol by doing two FFTs to build this delay–Doppler map. The main goal is to hold the direct path at zero delay and Doppler in every symbol.

The paper's improvements: Rosa: One major improvement they point out is their response-aware canceller, which involves removing the direct path’s carrier phase to get rid of its Doppler and Doppler rate echoes.

Dev: They also build an ideal pointspread function for a hypothesized path, find and refine the direct-path peak using least squares fitting, and then subtract it with Högbom’s CLEAN Detection.

Taro: They apply an ephemeris-derived clutter mask, cell-averaging CFAR, density clustering to help clean up the signal further, and they reject the fifty Hz sidelobes <ref:2610.11307#pg2>.

Rosa: The hardware results show that the direct-path peak-to-noise ratio was eighty-two point four nine plus or minus zero point seventy dB over twenty-eight dwells.

Conclusion: Dev: On hardware, they found a cancellation depth of thirty-four point eight dB, but in a matched simulation without radios, it hit thirty-seven point six dB <ref:2610.11307#pg1,in a matched simulation without radios>. So the radios themselves aren't the main limitation here.

Rosa: Controlled simulation showed that the emulator’s eight-tap fractional-delay filter and a simulated rotating blade echo were what limited them in the computer model.

Taro: The purpose of this whole tool is to make sure you can see if simulation and hardware differences are visible, repeatable, and attributable to something specific.

Dev: This setup lets an experimenter test an NTN sensing receiver before a satellite trial because they can score it against commanded path truth with the same code all the way through.

Rosa: The payload-aided example is also telling; it showed gains of up to ten point nine dB in peak-to-noise ratio when using known reference signals.