HamTransfer

A file transfer program for amateur radio using modern techniques.

Hamtransfer is not a modem, but the protocol on top.

Status

Currently just has proof of concept code. But speeds are still pretty good. E.g. better than ZModem over standard 1200/9600bps packet.

Quick start

This is not the final form of hamtransfer, obviously, but here's a step by step to testing the current code.

1. Build ax25ms.
2. Start ax25ms's ./serial -p /dev/rfcomm0 -l '[::]:12001' (both on client and server)
3. Build hamtransfer: cargo build (binaries end up in target/debug/ by default)
4. Start uploader, using the uploading ax25ms serial port:

   uploader \
       --router http://localhost:12001 \
       --parser http://localhost:12001 \
       --source M0XXX-1 \
       --input testdata   # directory with data
   

   You'll see a file listing, with checksums.
5. Start downloader, using the downloading ax25ms serial port:

   downloader \
       --router http://localhost:12001 \
       --parser http://localhost:12001 \
       --source M0XXX-1
       --output test.out
       checksum-from-the-uploader-file-listing
   

Overall architecture

Because AF_AX25 sockets are only available on Linux, and are extremely buggy there too, hamtransfer uses ax25ms to transmit and receive packets.

The interface to ax25ms is gRPC.

If you have a TNC over serial or Bluetooth (E.g. the Kenwood TH-74), then it's:

TNC <bluetooth> ax25ms serial <gRPC> hamtransfer

If you don't have a TNC then you'll need to emulate one in software, for example by using Direwolf.

Radio <audio cables> Direwolf <loopback serial> ax25ms serial <gRPC> hamtransfer

Performance

As of 2023-05-28, before the protocol has been fixed to remove a roundtrip, a 10kB file can go at 5576 bps.

I don't really even understand how it can be that fast, because that was with 239 byte payloads, which takes ~200ms airtime, and with the Kenwood D74 set to 200ms TX Delay, it should top out at 4800bps.

With an SDR it looks more like ~10ms of carrier before modulation starts, so maybe it's 200ms with 1200bps, but only 25ms with 9600?

Reasons why it's not faster, and misc rambling notes:

• One needless roundtrip. Protocol will be fixed to avoid this.
  • This should get the speed up to 6040bps.
• Packet size is not big. Bigger tends to make the D74 crash.
  • Overhead without repeaters is 18 bytes, so with 200 byte payload that's 8.25% overhead. 8807bps theoretical max just from that.
  • Time before transmitting:
    • Maybe T107?
    • T102 slot time (randomized) before key-up.
    • TXDELAY is time between key-up and packet.
      • 300ms by default with direwolf.
      • 20ms is good. There are ones with 1ms
      • Speed formula: (8size)/(sec+8size/9600).
    • AXDELAY, if using repeaters. Not used?
    • Optionally plus T105, remote receiver sync.
  • Time after transmitting:
    • TXTail (default 30ms on D74, though I see 11ms with SDR)
    • T108 is time after packet before re-keying.
• Looks like there are gaps between packets when sent via D74.

2023-05-29 experiment with 1200 byte packets for 10kB and removed extra roundtrip give best run 8024bps. 83% of theoretical max. Most of that is probably due to the fact that there's always that one roundtrip.

Blog post with example video

https://blog.habets.se/2023/07/Multichannel-AX.25.html

Protocol

The protocol is being designed, so no point yet to describe it here.