muse-lsl
muse-lsl copied to clipboard
alexandrebarachant
Bluetooth disconnecting when PPG streaming (Muse_2 issue)
It seems that Bluetooth is disconnecting when PPG streaming is on after about 10~15 minutes for Muse_2. No problem at all (30+ minutes testes) when any other combination without PPG is set. Any buffering issue? Any help as having pulse information would be nice for ML!
Otherwise, many thanks for a wonderful toolbox!
Are there any error messages printed when the bluetooth disconnects? I'm curious if this is a device issue (the Muse wasn't really built for sessions longer than 30 minutes)
I can partially confirm this, at least enough to keep this issue afloat..
I see frequent disconnection using muse 2018, with or without PPG. No special errors printed ~~other the the usual missing sample .... : ...., before it Disconnected.~~ (Edit: no errors at all)
I'm seeing sessions of 2-20 min. with Muse 2018, where ~15min is avg.
On same setup the Muse 2016 works very stable! I'm using MacOS+BLE112. I need to test on another stack, Windows machine w/ Bluemuse.
Right, I do confirm @oori message, and I have the same problem on MacOS+BLE112 and Win10+BLE112. We have several BLE112, and my collaborators could reproduce the issue with several Muse_2 headsets, so it might be a hardware issue (battery dropping power sometimes causing Bluetooth to break the connection, as it usually tries to reconnect after that).
@jdpigeon - right, but from my experience running 2 hours long streaming sessions with MUSE_2016 is completely no problem with the fabulous muse-lsl software ;)
Below is an example with 6 minutes long streaming (see time stamps on the right) with a suddenly broken connection using MUSE_2:
I've tested with higher number (10, 60) for AUTO_DISCONNECT_DELAY. I did get a long 42min session, but on the other hand, a bunch of short ones as well (9m, etc..). Does not seem to be the issue.
@jdpigeon Any tip how to debug this ? it's a consistent issue, and makes the muse 2018 unusable for some user applications.
See also: https://github.com/alexandrebarachant/muse-lsl/issues/55 and the linked issues.
The previous tests were on MacOS. Now on Raspbian (on rPi 4) with internal bluetooth, we get very clear results:
EGG + ACC + PPG immediately gives missing sample (10-20 times) and then disconnects in a couple of seconds.
EEG + PPG same
EEG + ACC starts giving missing sample errors only after half a ~30 seconds, then disconnects after another ~30 seconds. Overall - x10 times longer than the unusable PPG.
EEG works very good. long sessions. might throw a missing sample or two, but doesn't disconnect !
Note: while working, it does stream data !
So, I read this issue, and test the RPi with BLED112 and BGAPI backend, and the results are extremely better, similar to MacOS.
EGG + ACC + PPG long steady 30+ minutes sessions, but also short ones.
So, with BGAPI, we can get long sessions, but it's not promised. Software restarts can be fast (using saved address), the only problem I see now is that sometimes after Disconnect, the muse is "stuck", as if it's still connected, and doesn't jump back to "discovery mode". So, one has to take off the headset and restart. But luckily, using BGAPI, I see this phenomena much less often as with GATT.
I am having the same problem. The EEG stream is getting disconnected suddenly without giving any error message while I was trying to receive it through lsl using a python script. Any solution for that?
Same experience here with my [Ubuntu laptop or Raspberry Pi 4B] + Muse2018 device. Stream ends after some random time ranging from a few seconds (rarely) to around 90 minutes (with somewhere about an hour being the most frequent case).
Also tried a Nordic PCA10059 bluetooth dongle but faced the same issue. By any chance, is there a clear understanding of the reason for this by now? (Or probably a workaround by any chance, of course).
