I'm sure if I did a search I'd find an axiom that says something about version 1.0 of any software needing to be followed with another version.  I shorted some testing in my rush to get the 1.0 firmware release out (because I was going to soon be busy with other projects) and of course it came back to bite me.  Another weeBell user reached out because he was having echo problems and this started me looking at the code again in my off hours.  Turns out that there were two problems, one of which was mine (more about the other one further down).  I still hadn't gotten the audio code interface with the I2S driver so locked down that the delay between outgoing and incoming audio would always remain inside the bounds the OSLEC echo cancellation routines could handle.

I also built an isolated telephone interface for two reasons.  I wanted something that could output audio from phone calls I could record on my computer for an upcoming video I want to make about weeBell, and I also wanted something to feed a caller ID decoder using an ESP32 and spandsp for testing and for display of incoming numbers on my own Ooma VOIP landline.

This little gadget let me look at the Caller ID waveform - and compare it with the one generated by the Ooma.  This lead to finding a few bugs related with Caller ID generation and race conditions with users picking up before it was transmitted or the call ending early.  It, and more review of various standards, also led to a change in releasing the ring voltage between rings.

So a bit more testing and code changes later there is version 1.1 of the firmware (available in the repo or through the gCore Serial Programmer desktop software for easy updates).

The full release notes follow.

Bug Fixes

1. Restructure audio_task and improve interface to I2S driver to ensure consistent synchronization between RX and TX paths so we can count on a static alignment of TX/RX data into OSLEC. Increase OSLEC tail to 32 mSec.
2. Fixed a bug where the Caller ID transmission wasn't cancelled if the user picked early and it ran after call hung up.
3. Fixed a bug where a race condition between picking up the phone after the first ring could re-trigger Caller ID transmission. Increased flush buffer to 50 mSec.
4. Fixed a bug in the Caller ID logic where the final long ring of a cadence pair could re-trigger Caller ID transmission.

Functionality

1. Slight mods to code so building with optimization turned on won't cause compiler warnings/errors.
2. Detect more cases where the cellphone might route audio to weeBell outside of a phone call and have weeBell display the right state if the user picks up the phone.
3. Modify ring waveform to return to normal levels between rings to match expected behavior.
4. Set full volume output for Caller ID transmissions and Off-hook tones independent of volume setting.
5. Set Caller ID Mark pre-amble bits correctly for Bellcore and European FSK (spandsp defaults were shorter).
6. Modify bt_task so the local volume takes precedence over the volume sent by the phone when a call is established for outgoing calls.
7. Modify screen dump to be triggered by power button (when compiled in).

Strange issue with older Motorola Android phone

The person who reached out was kind enough to do a bunch of experiments for me with his Android phones including collecting audio data for review.  weeBell works perfectly with his Samsung phone but not his Motorola phone.  The strange thing is that OSLEC doesn't seem to be effective when running with the Motorola phone even though the audio synchronization appears to be working and well within OSLECs echo cancellation tail.  More experiments are in progress and hopefully we can get to the bottom of this.  It may result in another release but I didn't want to hold this one up since it fixed some real issues.