SoftRF version mb163

[Moshe Braner]

Now posted version mb163 for the T-Beam.  It includes two changes, both not relevant to the T-Echo, and also not important for most T-Beam users:

(1) The blue LED turns off as soon as a press-and-hold of the button (nearest the USB jack) has triggered the shutdown sequence.  The rest of the shutdown takes a few seconds after that, first the downward beep, then the "OFF" message on the OLED, then the red LED turns off - and the blue LED will turn back on if battery charging is happening.

(2) If a Ublox 7 GNSS module is detected, the PPS timing is moved 100 ms earlier.  This is to work around an apparent problem with the "VK2828" modules.  They supposedly are based on a Ublox 7 chip, but may be a fake, with a bug that apparently pulses the PPS pin 100 ms too late, resulting in about 25% of the packets (in either direction) not being received since the time slots are not synchronized.  This PPS shift could be made optional.  But currently it is done automatically for any module identifying itself as Ublox 7.  The LilyGo T-Beam boards use Ublox 6 or 8 modules.  If anybody is using a Ublox 7 module other than the VK2828 type, let me know.  The VK2828 modules are available, cheap, and work well otherwise, with much better GNSS reception than the built-in Ublox 6 on most T-Beams.

[Moshe Braner]

Now posted version mb165, which replaces that ad-hoc fix for the PPS timing with a better one.  This only affects T-Beams with an added Ublox 7 GNSS module (such as the VK2828 type).  Also fixed a bug that affects the use of an add-on GNSS module on old T-Beam v0.7 boards.

For the curious: after much investigation (thank you Nick for the hardware measurements!) we figured out that the cause of the bad timing with the VK2828 GNSS module is due to noise on the PPS wire at the time it falls back to low, at the end of the PPS pulse.  That noise resulted in the PPS interrupt routine being called multiple times - and 100 ms late.  The better fix was to ignore rising edge interrupts on the PPS wire until at least 200 ms have passed since the last one.  It wouldn't hurt to add some noise filtering to that wire (resistor and capacitor), but this software fix does the job even without any hardware filtering.

[Moshe Braner]

Also made one more small change to mb165 today (without changing the version number), that tries to avoid the occasionally seen behavior where (only) the first transmission sent out after it got a GNSS fix reports a very wrong position (due to the lat/lon still being zero at that moment).  This first position was reported by OGN stations as some strange location many km away (but within about 200 km of the station location, due to the way the location is encoded in the transmission and decoded by the receiver).

I did not recompile the T-Echo version at this time, only the T-Beam version.