Supersedes 493's predecessor — use this one. Build 492 shipped with an embedded web client built minutes before the last fix, so the app's own controls and the page it serves disagreed about one indicator. Identical otherwise; 493 embeds VibeServer 5.6.53.
For Onfliner and tgcfabian — read the deviation note before comparing against MPX Tool.
The FM deviation meter now reports a peak and an average
We measured the right quantity and published the wrong statistic. The reading has always been the full MPX, taken before de-emphasis and correctly scaled — but what we displayed was a 1.5 second average of 50 ms window peaks, labelled as peak deviation.
On heavily processed programme the modulation is nearly continuous, so the average is almost the peak and the reading was right — within 1 kHz of MPX Tool, as tgcfabian measured. On jazz, classical and speech the peaks are sparse, most windows contain none, and a 75 kHz peak read about 38 — as Onfliner reported. Both reports were correct. Only the averaging explains both.
Three figures from one measurement, as a PIRA analyser does:
- Peak — a true peak meter, instant attack. The digits hold it flat for 3 seconds then step, so it can be read; anything higher still appears immediately.
- Average — the 1.5 s figure the meter used to show. This is the number validated against MPX Tool and it has not changed.
- The bar follows the fast peak, with a quiet mark at the average and the hold tick above it.
★ Comparing with earlier notes: match your old readings against the new avg, not the new peak. On speech the peak reads far higher than before — BBC Radio 4 measured peak 51 · avg 34 kHz, Classic FM peak 55 · avg 41. Both numbers are right; they answer different questions.
★ Known soft spot: on a weak signal the peak may read high. The noise-subtraction constant was fitted for the averaged figure and still is, so it errs towards showing a reading as noisy rather than inventing silence. Re-fitting it is a bench job.
Also fixed in the measurement: the percentile discard scaled with the sample rate and threw sparse transients away whole, and the post-retune blanking is cut from 1.5 s to 0.4 s.
Airspy
- Tuner bandwidth no longer resets to ±100k when the phone is rotated. The per-mode default table was re-applied on every mode-bearing message, including re-assertions of the mode you were already in.
- The free gain mode is remembered. It was never persisted, and re-opening the radio built a fresh source defaulting to Linear — so the app was faithfully painting what the radio honestly reported.
DAB
- A padlock on the frequency entry in DAB mode, explaining why, instead of accepting a number the server then silently refuses. There is no VFO inside a multiplex.
- An EXIT DAB button in the decoder box header. There is no close button in DAB — it used to strand you in the mode with nothing on screen.
- Leaving DAB restores your last non-DAB frequency and demodulator, or falls back to AM at the block.
Elsewhere
- NR/NB/AN indicators no longer appear on broadcast FM, where they are on by default for everyone and so say nothing. They remain on MW, HF and narrow FM, where they are an audible choice you made.
- Auto squelch, the server health pill, and the storm meter's immunity to AGC steps.
Sideload as usual; this is a pre-release for testing.