Timer N channel fix #9276
Timer N channel fix #9276
Conversation
|
Thanks for the fix, We just did a test on bench. motor3 works when using "Master" slider in output tab or using transmitter after arming. I uploaded a video for your refference. |
|
You compile it with |
without Dshot burst. I tested it with normal DMA |
|
I think we should switch this board to burst DMA in INAV 7 |
|
Just tested it again with #define USE_DSHOT_DMAR
|
|
Added |
|
@shirase how about extracting Burst DMA into a separate PR? I mean those are independent changes and looks like it does not work like expected. I will try to test it with a different board that has N timer channel only |
This reverts commit 46b1ece.
|
@DzikuVx agree. It must to work without dma burst. |
|
@shirase I mean we can switch this target to DMA Burst. But not as part of the fix for N channel timers |
|
@DzikuVx Can you make PR for DMA burst targets? I do not have boards for testing. And i don`t known lists of problem boards. |
|
I will do that |
|
As I posted in #9277, even with those changes, the output from S1 to S4 on the Matek F405 Mini-TE board looks like this on an oscilloscope: What draws the attention is S3 (scope Ch3, magenta line) being "inverted", a mirror of the other outputs. This is not unexpected, once ST's "N" timers are complementary and have inverted output levels from the "normal" channel. But this is probably the source of the DSHOT errors. I don't fully recall the DSHOT spec, but I don't think it is tolerant to inverted logic levels. |
|
Ok, so I realized that both #9265 and #9277 didn't include the fix in this PR. I manually added this PR to my testing code, flashed to the Matek F405 mini-TE board and checked S1 through S4 on the scope (DSHOT 300). This is what I got, at 0% and 50% throttle:
All 4 lines look corrected, without the mirroring (inverted signal) on S3. Yay!! Next up will be installing this board in a quad and test flying. But, initially, I believe that this trio of PR's (#9265, #9276 and #9277) fixed the DSHOT issues on the miniTE board. |
|
Has anyone tried Dshot beeper after these changes? It seems not working. |
Yes, I can confirm Dshot beeper is not working, even in targets where DMA burst mode is not used/enabled. However, reverting #9265 makes it work again. Some change within that PR broke it for all targets. Also, my test flight on the Matek F405miniTE with #9265, #9276 and #9277 didn't go well. The quad simply dropped less than 30 seconds after take-off, apparently with the motors being cut. Due to an error on my part, I didn't have blackbox logging enabled, but I do have DVR recording that shows the current dropping to less than 1A before impact and no flight mode changes (so not a failsafe, which was set to RTH or land if closer than 20m from home). Here's that DVR: I am still not sure if it could be directly related to the DMA code changes or a hardware issue or another issue altogether, but it does cast suspicion and doubt. |
|
@shota3527 @rmaia3d try this fix #9321. Can you check it on dshot600 and dshot300? |
|
@shirase Is this still relevant? And working? |
|
I beleive @shota3527 got some reports it was working for one of the siwi engineers. I believe that was with this patch. |
|
@DzikuVx I don't have board with N channel. Need feedback. It works or not. |
As can be seen from the scope traces I posted a few comments above, this did indeed work to bring channel 3 to the "non-inverted" state. An inverted (0 being logic HIGH) DSHOT signal is how the ESC knows if the FC wants bi-directional comms or not. So, without this fix, 1 ESC (ch3) will work in bi-dir mode and the other 3 ESCs will work in normal mode. Not ideal. All ESCs should work in the same mode, and by the scope traces above, this PR ensures that. |
I'm away from home and wasn't able to test the changes after the dshot beeper fix, but before leaving I did further investigation and tests and was able to trace the motor stopping issue to the ESC itself. It "overheats" and then stops the motors (actually it looks like it's a bug in the temperature cutoff protection or a faulty temp sensor). So it has nothing to do with the FC or Inav (I was able to reproduce the issue both on the bench and in-flight while running betaflight too). Good news! |
|
Here's someone with a similar issue and another mention of the possible temperature cutoff bug. The ESC I'm using is the same, just the "mini" version. https://discuss.ardupilot.org/t/no-throttle-with-props-on-hexa-680-13/80847/5 |
|
Sounds like the fix works and can be merged. Nice shirase, and thanks for testing Rodrigo![]s,Marcelo Bezerra ***@***.***>On 11 Oct 2023, at 00:40, Rodrigo Maia ***@***.***> wrote:
Here's someone with a similar issue and another mention of the possible temperature cutoff bug. The ESC I'm using is the same, just the "mini" version.
https://discuss.ardupilot.org/t/no-throttle-with-props-on-hexa-680-13/80847/5
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>
|
Maybe it fix problem with N timer channel for MATEKF405TE. I don`t have that board. Can someone test.
#8089