Mirrors_Framework/PX4-Autopilot
1
0
Fork 0
mirror of https://github.com/PX4/PX4-Autopilot.git synced 2026-05-19 19:04:33 +08:00
Code Issues Actions 17 Packages Projects Releases Wiki Activity

rc/sbus: restart parser after sucessful decoding & increase time limit

Browse Source 
Instead of directly passing the next packet to the parser after successful
parsing, switch the state to SBUS2_DECODE_STATE_SBUS_START and search for
the start byte again.

The timeout is increased as the IO main loop also takes a bit of time
(max ~0.7ms).

Tested on v5x with Futaba R7008SB (60Hz update rate) and FrSky X8R (111Hz
update rate).

Background:
When using the Futaba R7008SB, I noticed there's additional bytes added in
between packets. Often it's a null byte, but sometimes more. There's some
consistency but I did not find any documentation for it.
Sample data:
a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 34 00 0f 05 ec
1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 04 00 c0 8b
00 0f 04 ec 1f a8 fb 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00
14 00 0f 04 ec 1f a8 fb 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80
00 24 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10
80 00 34 00 0f 05 ec 1f 30 60 bf 1c bd 07 16 5b 81 05 d4 a0 06 20 00 01 08
40 00 02 10 80 00 34 00 0f 07 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01
08 40 00 02 10 80 00 04 00 03 c0 31 00 0f 05 fc 1f a8 fb 07 16 5b 81 05 d4
a0 06 20 00 01 08 40 00 02 10 80 00 14 00 0f 05 fc 1f a8 fb 07 16 5b 81 05
d4 a0 06 20 00 01 08 40 00 02 10 80 00 24 00 0f 04 ec 1f a8 fd 07 16 5b 81
05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 34 00 0f 04 ec 1f a8 fd 07 16 5b
81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 04 00 03 c4 00 00 0f 04 ec 1f
a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 14 00 0f 04 ec
1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 24 00 0f 04
ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 34 00 0f
05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 04 00
03 c0 31 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02
10 80 00 14 00 0f 05 fc 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00
02 10 80 00 24 00 0f 05 fc 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40
00 02 10 80 00 34 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08
40 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80
00 04 00 03 c0 31 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08
40 00 02 10 80 00 14 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01
08 40 00 02 10 80 00 24 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00
01 08 40 00 02 10 80 00 34 00 0f 05 f4 1f a8 fb 07 16 5b 81 05 d4 a0 06 20
00 01 08 40 00 02 10 80 00 04 00 03 c4 00 00 b0 60 7f 1c bd 07 16 5b 81 05
d4 a0 06 20 00 01 08 40 00 02 10 80 00 14 00 0f 05 ec 1f a8 fd 07 16 5b 81
05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 24 00 0f 05 ec 1f a8 fd 07 16 5b
81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 34 00 0f 05 ec 1f a8 fb 07 16
5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 04 00 03 c0 31 00 b0 60 bf
1c bd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 14 00 0f 04 f4
1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 24 00 30 70
7f 1c bd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 34 00 0f 05
f4 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10 80 00 04 00 03
c4 00 00 0f 05 fc 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01 08 40 00 02 10
80 00 14 00 0f 05 ec 1f b0 60 bf 1c bd 07 16 5b 81 05 d4 a0 06 20 00 01 08
40 00 02 10 80 00 14 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00 01
08 40 00 02 10 80 00 24 00 0f 05 ec 1f a8 fd 07 16 5b 81 05 d4 a0 06 20 00

This was causing the parser to skip entire packets resulting in an update
rate of ~31Hz on the FMU side.
With this patch the update rate increases to 42-48Hz.

The investigation was triggered by an RC glitch with a packet containing
random channel data. It's likely, although not completely verified that
the frequent desync randomly happend to pass the CRC check with garbage
data.
This commit is contained in:
Beat Küng
2022-08-18 16:29:25 +02:00
committed by Daniel Agar
parent b58f70726f
commit bae275898b
1 changed files with 9 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/lib/rc/sbus.cpp
+9 -11
View File
@@ -299,7 +299,7 @@ sbus_input(int sbus_fd, uint16_t *values, uint16_t *num_values, bool *sbus_fails
* provides a degree of protection. Of course, it would be better
* if we didn't drop bytes...
*/
if (now - last_rx_time > 3_ms) {
if (now - last_rx_time > 4_ms) {
if (partial_frame_count > 0) {
partial_frame_count = 0;
sbus_decode_state = SBUS2_DECODE_STATE_DESYNC;
@@ -309,13 +309,6 @@ sbus_input(int sbus_fd, uint16_t *values, uint16_t *num_values, bool *sbus_fails
}
}
if (partial_frame_count == 0 && buf[0] != SBUS_START_SYMBOL) {
/* don't bother going through the buffer if we don't get the
* expected start symbol as a first byte */
sbus_decode_state = SBUS2_DECODE_STATE_DESYNC;
return false;
}
#endif /* __PX4_NUTTX */
/*
@@ -372,17 +365,18 @@ sbus_parse(uint64_t now, uint8_t *frame, unsigned len, uint16_t *values,
switch (sbus_decode_state) {
case SBUS2_DECODE_STATE_DESYNC:
/* fall through */
case SBUS2_DECODE_STATE_SBUS_START:
/* we are de-synced and only interested in the frame marker */
if (frame[d] == SBUS_START_SYMBOL) {
sbus_decode_state = SBUS2_DECODE_STATE_SBUS_START;
sbus_decode_state = SBUS2_DECODE_STATE_SBUS2_SYNC;
partial_frame_count = 0;
sbus_frame[partial_frame_count++] = frame[d];
}
break;
/* fall through */
case SBUS2_DECODE_STATE_SBUS_START:
case SBUS2_DECODE_STATE_SBUS1_SYNC:
/* fall through */
@@ -438,6 +432,10 @@ sbus_parse(uint64_t now, uint8_t *frame, unsigned len, uint16_t *values,
partial_frame_count = 0;
}
if (decode_ret) {
sbus_decode_state = SBUS2_DECODE_STATE_SBUS_START;
}
}
break;