Radio Probe
Measures how well two phones can talk to each other through their speakers and microphones: signal strength in each 1 kHz band from 1 to 21 kHz, real error rates for multi-tone DBPSK/DQPSK, the room's echo, each phone's own harmonics at its own microphone, clock drift and audio latency.
This phone
Distance between the phones
Audio settings
Microphone processing
Microphone while sending
Off sends from the loudspeaker without opening the microphone, to compare iPhone output routing.
iOS audio session
Sets navigator.audioSession.type before the microphone opens. Auto uses play-and-record with the microphone and playback without.
Output level
Digital level of the probe. Set the phone's own volume to about 75%: most phones distort above that.
Runs
No runs yet.
How to test
- Open this page on both phones. Set both to about 75% volume. On an iPhone, turn the silent switch off.
- Lay the phones flat on a table 30 cm apart, bottom edges facing. Pick the distance above on both.
- On phone A tap Listen, then on phone B tap Send probe. Keep quiet for 15 seconds.
- Swap: B listens, A sends. Each phone now has a send run and a listen run, which unlocks the full-duplex check.
- Repeat at 10 cm and 60 cm, and once with people talking or music playing.
- On an iPhone, repeat its send with Microphone while sending: off, and its send and listen with Microphone processing: Voice.
- Run the Latency test on each phone.
- Tap Copy all results on each phone and paste them into the chat. WAV downloads keep the raw audio for offline modem work.
What the probe plays
A 250 ms chirp (1.2–7.5 kHz), half a second of quiet for the noise floor, then twenty 1 kHz bands from 1 to 21 kHz. Each band is 16 OFDM symbols (21.3 ms plus a 5.3 ms cyclic prefix, tones every 46.875 Hz) carrying known DQPSK data, so the listener counts real decision errors. A second chirp ends the sequence; the gap between the two gives the clock offset between the phones. The sender records itself, which shows its own harmonics at its own microphone and its loopback latency. Error rates are raw, before any error correction: a rate-½ convolutional code with soft decoding copes with a few percent.