2022-03-02 notes

[etihwnad]

A goal is to determine good LoRa modulation settings (bitrate, spreading factor, etc.) that are expected to work well from LEO to a ground station.

WRT Jay's m17 experiment (writeup https://wiki.m17.link/sx1276)

• Maybe a slower MCU could still do FM voice by updating RegFdevLsb at a slower rate?

Frequency offsets should be worked out. How does this offset affect error rates over the various mod settings? Need to map out:

• Individual module frequency uncertainty.
• Doppler shift from LEO at 70cm.

[etihwnad]

• Maybe a slower MCU could still do FM voice by updating RegFdevLsb at a slower rate?

According to datasheet sx1276_77_78_79.pdf p.93, the frequency deviation is:

Fdev = Fstep * {RegFdevMsb, RegFdevLsb}

Default value is 0x52 and Fstep is nominally FXOSC / 2^19 or 61 Hz. This lines up with the "default 5 kHz" datasheet note.

The typical deviation for FM voice in the 70cm band is ±5 kHz -- about 164 Fsteps then for full deviation, which corresponds to a little more than 7 bits of audio signal quantization. Such quantization implies a ~40 dB SNR.

For a 3 kHz audio bandwidth we need a sample rate of more than 6 kHz. Faster is better to push away the alias terms.

Minimal write to the RegFdevLsb register is an 8-bit address + 8-bit data + (~2 SCK periods for start/stop). This then requires an SPI SCK frequency of more than (8+8+2) * 6 kHz = 108 kHz.

The fastest SCK on the ATmega328p is Fosc / 2. The current ORI-AmbaSat 'mega has a 4 MHz clock. This spread in rates seems to be plenty to allow time overhead needed to fetch and update the SPI peripheral from internal flash storage.

Conclusion: on paper, it seems possible to send FM voice by updating the SX1276's RegFdevLsb register.

[robojay]

• Verify which (maybe all?) LoRa modulation settings would be acceptable from a regulatory standpoint for use in space on 70cm. If time constrained, could focus on testing only acceptable options. If not time constrained, would be good to know what we're missing out on.