Skip to content

v0.3.5 - Lab Timing

Choose a tag to compare

@musithang musithang released this 28 Jun 14:37
· 909 commits to main since this release
v0.3.5
a159f44

Your radio keeps perfect time. The question is whether your computer can keep up. This release adds the panel that finally answers that, out loud.

A focused release built around one new instrument: Lab Timing, the host-side real-time clock-watcher. Plus the usual round of sharpening and bug-squashing under the hood.

⏱️ Lab Timing (Preset, key 7)

Here is the thing nobody tells you about SDR: the radio is never the flaky part. It samples at a rock-steady rate and ships those samples to your machine in regular USB bursts, one callback at a time. Your computer's only job is to catch every burst on schedule. Miss one and the buffer backs up. Miss a few and samples start hitting the floor. Lab Timing watches that handoff and tells you, honestly, whether your host is holding up its end.

Three zones, one story:

  • Timing Diagnostics (focus t): the numbers. Measured callback period versus the expected period at your sample rate, host clock drift in ppm, jitter, and the per-callback deviation percentiles (p95 / p99 / peak) drawn as bars against a deadline budget that scales with the sample rate. Plus a late-callback count, sample-rate accuracy, throughput, and a plain-language verdict: Excellent, Good, Marginal, or Poor.
  • Callback Interval Strip Chart (center): the picture. Every point is one real callback, plotted as how far its arrival drifted from the expected interval. Late deliveries climb above the line, early ones dip below, and anything past the deadline band gets tagged (▲ late at the top, ▼ early at the bottom). When your host hiccups, you see it happen in real time instead of guessing.
  • Hardware Vitals (focus v): the supporting cast. Sample drops, ADC saturation and CPU / RAM as rolling 60 second trends, USB link utilization against your device's real ceiling, ring-buffer fill with overrun headroom, and session uptime.

Why you would care: if you have ever seen mystery drops at a sample rate your hardware should handle fine, the culprit is almost always the host (a CPU spike, USB contention, an overeager power-saving governor). Lab Timing turns that "probably my laptop" hunch into a reading you can point at. And it is honest about what it measures: everything here is the host's real-time behavior against the monotonic clock, no hidden hardware magic.

🔧 Under the hood

  • Carrier / image detection in Lab IQ now gates on the noise floor, so an idle radio reads "no signal" instead of pinning a random noise bin as a carrier and crying wolf about image rejection.
  • Gains clamp to each device's own model on load, so a config saved on one radio never tries to set an illegal gain on another. The RTL-SDR tuner-AGC indicator stays honest when you nudge the gain manually.
  • The spectrum / waterfall cursor respects each device's real frequency range instead of a hard-coded one.
  • Matching subheadings and consistent breathing room across the lab panels, so the whole bench reads as one instrument.

Get It

Linux with a HackRF One or RTL-SDR. Build with cargo build --release. Press 7 to open Lab Timing, start RX, and watch your computer sweat.