Additional LR2021 Research

[sfowlr]

Hello - I'm posting here because Discussions is off on lr2021-sdr.

I wanted to thank you for publishing your findings on the LR2021’s SDR capabilities - that work was the inspiration for my own research into it, as a potential RF frontend for another open-source project I’m developing.

We've since measured a fair amount on a Meshnology W12 (ROM 1.24, 2026-10-23/24), and some of it extends or corrects your findings:

• Rates: under the FSK modem, the rx_bw byte in SetFskModulationParams sets the route's rate when route byte 0 is 0xFF. The rate is 32 MHz / (mode × decimation). Source 1 streamed at 664.8 kS/s with no loss at 16 MHz SPI.
• Sources: source 1 is int8 and source 2 is int16, both after the channel filter. int8 clips at −32/+31, about 20 dB of range, and int16 gives at least 50 dB more.
• Gain: SetAgcGainManual only holds in RX, because SetRx resets it to step 13. Send it after SetRx.
• TX: selector 1 scales as sample × fdev / 64 Hz at every rate we tried (31.25k to 250k), so the 125 kS/s restriction doesn't apply on our board. No reset is needed between runs, or between RX and TX.
• 4-FSK: opcode 0x023B with argument 0x01 turns the FSK modem into a 4-FSK modem. We sent DMR bursts through it, and 32 of 32 decoded with 0 FEC errors.

I’d be happy send the details or check anything on our board that you'd like confirmed.

Is there a better way to reach you than Discussions? I'd also like to hear what you're doing with the LR2021 next.

[SAMS0N1TE]

Hey sfowlr, thanks for taking the time to test this and share what you found. There’s a lot of useful information here, especially the AGC behavior after SetRx and getting 4-FSK working. The DMR results are particularly interesting, and it’s good to see someone exploring this on another LR2021 board.

I’d be interested in seeing more of your setup, including the RX/TX command sequences and how you’re handling the SPI transfers to sustain 664.8 kS/s without losing samples. Any code, logs, or captures you can share would help with comparing your results against what I’ve been seeing. The wider TX sample-rate range and switching between RX and TX without resets would be useful to understand too.

You can reach me directly at support@terminalbay.com. Feel free to send the details there, or keep sharing here so others can follow along. I appreciate the offer to help with testing as well!

[SAMS0N1TE]

Or Discord - https://discord.gg/YdzE9cyHS

[sfowlr]

Thanks for the quick reply - I'll reach out on Discord!

I just open-sourced my LR2021 work today: https://github.com/RadioFoundry/RadioFoundryGateway
The low-level detailed research is not included there; I'll share that with you on Discord.

The much bigger project it supports, RadioFoundry, got open-sourced yesterday as well, under the same github org.

[lyusupov]

@sfowlr

I just open-sourced my LR2021 work today: https://github.com/RadioFoundry/RadioFoundryGateway

Recent update of LilyGO T-Beam 1W (V1.1) with LR2021 radio looks like a better match for your use case rather than Meshnology W12

1. https://lilygo.cc/products/t-beam-1w
2. https://github.com/Xinyuan-LilyGO/LilyGo-LoRa-Series/blob/master/docs/en/t_beam_1w_lr2021/t_beam_1w_lr2021.md

Meshnology W12 (LR2021, L76K GNSS header). Receives/transmits well only in 902–928 MHz: at UHF its filter costs about 46 dB, so it hears only nearby transmitters

According to this
https://github.com/thefatmoop/esp32s3-lr2021-adsb-receiver#hardware-modification-for-best-reception
the W12 needs removal of built-in SAW filter to operate in a wide band.