Trouble decoding AFSK/FSK signals with SX1278 and Soundmodem

[k42um]

Hello administrators and community members,

I’m attempting to perform AFSK and FSK communication using the RadioLib library with an SX1278-based module, but I’m having trouble getting Soundmodem (the Windows software TNC) to successfully decode the signal.

According to the official discussion threads, Soundmodem is recommended for decoding, so I suspect there may be an issue with my parameter settings, but I haven’t been able to resolve it yet.

---

Environment

• Transceiver module: Ra-01 (SX1278)
• Microcontroller: Raspberry Pi Pico (via Arduino IDE)
• Receiver: Nano 3 SDR
• SDR software: SDR Console
• Software TNC: Soundmodem and High-Speed Soundmodem

---

Programs Used

I’m using the official sample programs for both modes —

• For FSK: AX25_Transmit.ino
• For AFSK: AX25_Transmit_AFSK.ino

I’ve only made minor modifications, which are described below.

Common Modifications

Changed GPIO pin assignments for use with the Raspberry Pi Pico:

SX1278 radio = new Module(27, 21, 20, 26);

FSK Modifications

Changed the frequency and bitrate:

int state = radio.beginFSK(436.5, 9.6);

AFSK Modifications

Changed the frequency and bitrate:

int state = radio.beginFSK(436.5, 1.2);

---

Soundmodem Settings

Soundmodem is configured as follows:

• AFSK
  • AFSK AX.25 1200bd
• FSK
  • FSK G3RUH 9600bd

---

During Operation

From the serial monitor output, everything appears to be working correctly. The SDR also shows what seems to be a valid signal.
However, neither AFSK nor FSK transmissions are successfully decoded.
I’ve attached two screenshots below for reference.

：AFSK ：FSK

Ultimately, I would like to use GMSK at 9600 baud.

---

I would greatly appreciate any help or suggestions for resolving this issue.

Thank you in advance for your support.

---

P.S. Please let me know if it would be better to move this to an Issue.

[K4KDR]

The first thing you might want to address is testing with a signal that is 60db above your noise floor. You also see that the waterfall in SoundModem is extremely saturated. So, both your received RF level and possibly your audio level to SoundModem would produce better results if they were more realistic.

[k42um]

Thank you very much for your kind reply.

According to the official Soundmodem guide, it says: “if using color waterfalls, you should set for green with some yellow patterning in receive state and red when data is detected,” so I adjusted it with that in mind.
As you advised, I tested with a signal about 60 dB above the noise floor, but neither case produced good results.
What could be the cause?

[K4KDR]

If you would like to share an audio or I/Q file, that would allow others to do some testing

[k42um]

Thank you, these ones.
iqdatas.zip

[jgromes]

I think there's a few distinct problems here:

1. In the FSK case, you are demodulating in FM mode - that won't work, since the signal is not audio-modulated. You have to demodulate in a sideband mode, LSB or USB. You have also set 9.6 kbps bitrate at the transmitter, but as far as I can tell, soundmodem only supports AX.25 in 300 - 2400 bps rates, so I would not expect 9.6 kbps to work. I'm a bit confused as to which screenshot is for which transmission type, but it doesn't make sense either way (one is 9k6 G3RUH, second is 1k2 AX25, you are transmitting 9k6 AX25).
2. In the AFSK case, I would check how your DIO2 pin is connected, since you you have not posted anything about it. In this case, you will demodulated in FM mode and soundmodem AFSK AX.25 1200 baud setting will work (I checked on my setup).
3. RadioLib does not currently suport G3RUH scrambling, so soundmodem will most likely not be able to decode the packets in this mode.

[k42um]

Thank you very much for your prompt response!
As for AFSK 1200 baud, it seems that the main cause was that I had forgotten to connect DIO2. After connecting DIO2 and adjusting the settings in Soundmodem, I was able to successfully decode the signal.
Regarding FSK 9600 baud, I am planning to decode it using the high-speed Soundmodem. I understand that even if it doesn’t specifically support G3RUH, it should still be compatible as long as it’s a standard AX.25 frame. However, since I haven’t managed to get it working yet, I can’t say this with certainty.

[k42um]

It doesn’t necessarily have to be 9600 bps, but I’d like to know how I can verify that a normal FSK signal is being decoded correctly. I need 9k6 GMSK transmit and receive for satellite communication.

[jgromes]

This is not about the bit rate, it's about the scrambler. AX.25 frames from RadioLib are sent without any additional scrambling, only bit stuffing described by the AX.25 specification. So if your receiver assumes the scrambling is there, it will always try to unscramble the input, producting garbage for unscrambled data.

You will have to either find soundmodem configuration that does not include the descrambling step (I don't see one like that in the available ones), or implement the scrambler in RadioLib prior to transmission.

[k42um]

So, does that mean that RadioLib does not support FSK communication at 9600 bps (and that it’s not possible to verify successful transmission through decoding)?
Thank you very much for your kind and thoughtful response.

[jgromes]

I'm not sure I know what you mean by "FSK communication at 9600 bps" - if you mean sending plain FSK-modulated packets (not AX.25) with a preamble, sync word, optionally the payload length, some user data and a optionally a checksum, then yes, this is fundamentally supported by the SX1278 hardware. You can do it simply by calling beginFSK with the appriopriate arguments and then transmit.

But are you sure you will be able to decode that with soundmodem? What packet structure does it expect? And what about the scrambling?

[jgromes]

Since I cannot actually help myself, I had a go at implementing a (de)scrambler here just for fun, from scratch. This is now in master branch, however, I wasn't able to get it to work with soundmodem yet. I think the scrambler itself works, because if I scramble some bytes and the de-scramble them, I get the original bytes back. But there's probably something I'm missing in the full AX.25 encoding process, since soundmodem does not recognize the packets.

I'm especially unsure when exactly to do the scrambling (before or after NRZI encoding? or instead of it?), whether I am using the correct polynomial and whether the HDLC preamble of the AX.25 frame is also supposed to be scrambled (seems a bit weird but online resources do suggest it). Some testing vectors (i.e. data before and after scrambling) would be really helpful if anyone has access to something like that.