A Python experiment with packet modems using AGWPE.
Maybe good enough to get an idea how to use AGWPE.
- Direwolf 1.7.0 with my patch to enable flow control from non AX.25 packets
- Python 3.x
By default, when sending a 'y' AWGPE packet Direwolf only counts queues data that is a valid AX.25. Because we're sending raw packets, Direwolf always returns 0 packets queued.
For testing I've modified line 987 of src/tq.c:
From: if (ax25_get_num_addr(pp) >= AX25_MIN_ADDRS) {
To: if (TRUE || ax25_get_num_addr(pp) >= AX25_MIN_ADDRS) {
I'm sure it breaks the source and dest filtering. For now, I'm just seeing if this entire idea will work, later on I can talk to the Direwolf devs about it.
For local testing, need one modem for receiving and one for sending. For testing, I'm using SoundModem_HS by UZ7HO for receiving and Direwolf for sending.
Set the AGWPE ports in test_recv.py and test_send.py
Run test_recv.py in one console, and test_send.py in another.
When the data is all received, ctrl+c test_recv.py to see speed/etc
Potential message protocol
PACPAL Magic
VERSION 0x01
CALLSIGN 10 bytes Right padded with 0x00 (I think 7 is the longest, but people like SSID too so...)
TRANSFER# 16bit int One transfer, multiple files (random number probably fine)
FILE 8bit int Which file are the blocks/data associated with.
0 = metadata about transfer (names, comments, etc)
BLOCK 16bit int Which block is the data for
MAX_BLOCK 16bit int How many blocks in total
This can't be in the metadata, because the metadata might be >1 block long
CRC32 32bit int
DATA Bytes[]
38 byte header, eww...... I think we actually need to make sure it's at 36 bytes anyway for Direwolf to consider it a valid frame?
256 byte messages
Msec taken 9149
Bytes transferred 8192
Speed 895.3984041971801 byte/s
512 byte messages
Msec taken 7895
Bytes transferred 8192
Speed 1037.6187460417987 byte/s
1024 byte messages
Msec taken 7841
Bytes transferred 8192
Speed 1044.7646983803086 byte/s
Hmm, if you ignore the startup and setup times it's actually even better. Though maybe I've stopped counting the time the first message takes to transfer, so it's 10% slower?
256 byte messages
Msec taken 7399
Bytes transferred 8192
Speed 1107.1766454926342 byte/s
512 byte messages
Msec taken 6760
Bytes transferred 8192
Speed 1211.8343195266273 byte/s
1024 byte messages
Msec taken 6161
Bytes transferred 8192
Speed 1329.6542769031003 byte/s