Hi Dialog,
I am developing a pair of BLE device, both server and slave, base on DA14580. After connected to each other, even I leave these processor to be idle(no BLE read/write/notify, no i2c, gpio... active ) but they still disconnect randomly. It looks like one of them corrupts in link layer, because after disconnect they could not reconnect again. Below are the parameters I set for this connection and some wireshark captures.
msg->nb_peers = 1;
memcpy((void *) &msg->peers[0].addr, (void *)peer_addr, BD_ADDR_LEN);
msg->peers[0].addr_type = peer_addr_type;
msg->con_intv_max = 10;
msg->con_intv_min = 10;
msg->ce_len_max = 0x20;
msg->ce_len_min = 0x20;
msg->con_latency = 0;
msg->superv_to = 100;
msg->scan_interval = 0x180;
msg->scan_window = 0x160;
msg->op.code = GAPM_CONNECTION_DIRECT;
wireshark capture:
9796 61.273854000 Master Slave BLE Data 26 Empty Data PDU
9797 61.275295000 Slave Master BLE Data 26 Empty Data PDU
9798 61.286603000 Master Slave BLE Data 26 Empty Data PDU
9799 61.287569000 Slave Master BLE Data 26 Empty Data PDU
9800 61.298855000 Master Slave BLE Data 26 Empty Data PDU
9801 61.299789000 Slave Master BLE Data 26 Empty Data PDU
9802年62.320227000 ab_mat5 BLE ADV 48 ADV_IND
9803 62.321788000 ab_mat5 BLE ADV 48 ADV_IND
Server and client keep the connection for 60s and could not connect to each other anymore.
I could still see slave side advertising, but even use an Iphone with LightBlue application, I could not connect to the slave anymore.
Wireshark capture:
25 6.273182000 ab_mat4 BLE ADV 48 ADV_IND
26 6.274736000 50:0b:f8:18:ac:29 80:ea:ca:00:00:41 BLE ADV 60 CONNECT_REQ
27 6.276461000 Master Slave BLE Data 26 Empty Data PDU
28 6.303327000 Master Slave BLE Data 26 Empty Data PDU
29 6.333342000 Master Slave BLE Data 26 Empty Data PDU
30 6.363329000 Master Slave BLE Data 26 Empty Data PDU
31 6.393384000 Master Slave BLE Data 26 Empty Data PDU
32 6.423327000 Master Slave BLE Data 26 Empty Data PDU
33 6.453367000 Master Slave BLE Data 26 Empty Data PDU
34 6.483385000 Master Slave BLE Data 26 Empty Data PDU
35 6.513421000 Master Slave BLE Data 26 Empty Data PDU
36 6.543378000 Master Slave BLE Data 26 Empty Data PDU
37 6.573448000 Master Slave BLE Data 26 Empty Data PDU
38 6.603361000 Master Slave BLE Data 26 Empty Data PDU
39 6.633361000 Master Slave BLE Data 26 Empty Data PDU
40 7.640482000 ab_mat4 BLE ADV 48 ADV_IND
41 7.642505000 ab_mat4 BLE ADV 48 ADV_IND
In this test, I set the connection interval to be 12.5 ms. In this case, sever and client will disconnect randomly after 60 s to 45 mins they connect.
In my latter test, I change the connection interval to be 30ms. They could keep the connection much longer(more than 10 hours), but still they disconnect.
I would assume if I leave the processor in idle state, everything should handle by BLE stack automatically and sending empty to keep the connection should have nothing to do with my application code. Also, I not have any application activity at the time they disconnect.
Please help me to find the reason why the disconnect and how to improve the stability of the connection.
Thanks a lot.
Some update,
From my overnight test, it looks like the ble timer is not working appropriately.
Firstly, it works okey for about 3 hours(probably more). We could see master sends an empty PDU to slave every 30ms as the value I set.
726216 10926.699285000 Master Slave BLE Data 26 Empty Data PDU
726217 10926.699861000 Slave Master BLE Data 26 Empty Data PDU
726218 10926.729275000 Master Slave BLE Data 26 Empty Data PDU
726219 10926.729815000 Slave Master BLE Data 26 Empty Data PDU
726220 10926.759274000 Master Slave BLE Data 26 Empty Data PDU
726221 10926.760253000 Slave Master BLE Data 26 Empty Data PDU
726222 10926.789250000 Master Slave BLE Data 26 Empty Data PDU
726223 10926.790219000 Slave Master BLE Data 26 Empty Data PDU
726224 10926.818524000 Master Slave BLE Data 26 Empty Data PDU
I guess the sniffer lost one capture here..
726225 10926.848719000 Master Slave BLE Data 26 Empty Data PDU
726226 10926.850223000 Slave Master BLE Data 26 Empty Data PDU
To some reason, I lost the capture for about 10 hours. Luckily, just before they disconnect, the sniffer get the signal again. But at this time, master side seems to behave crazy, it sends a empty PDU to client every about 2ms or maybe as soon as it receives an empty PDU from the slave.
726227 47816.865216000 Master Slave BLE Data 26 Empty Data PDU
726228 47816.866244000 Slave Master BLE Data 26 Empty Data PDU
726229 47816.867143000 Master Slave BLE Data 26 Empty Data PDU
726230 47816.868015000 Slave Master BLE Data 26 Empty Data PDU
726231 47816.868887000 Master Slave BLE Data 26 Empty Data PDU
726232 47816.869742000 Slave Master BLE Data 26 Empty Data PDU
726233 47816.870613000 Master Slave BLE Data 26 Empty Data PDU
726234 47816.871492000 Slave Master BLE Data 26 Empty Data PDU
.......
726560 47817.196839000 Master Slave BLE Data 26 Empty Data PDU
726561 47817.197927000 Slave Master BLE Data 26 Empty Data PDU
726562 47817.198787000 Master Slave BLE Data 26 Empty Data PDU
726563 47817.199657000 Slave Master BLE Data 26 Empty Data PDU
726564 47817.201497000 ab_mat5 BLE ADV 48 ADV_IND
726565 47817.202433000 ab_mat5 BLE ADV 48 ADV_IND
我希望这些信息给你们更有意义about this problem, thanks a lot.
Thanks!
Hi zwang308,
Which SDK you are using in your current implementation and what is the development kit you have ? As far as i can tell the connection parameters you are using are the almost same with the DSPS project. Your assumptions are correct, when you are connected with a device the stack will keep the connection alive by exchanging empty packets without demanding from the application to do anything.
From the 1st log i can see that your master stops sending empty packets in order to keep the connection alive so there is no response for the slave side, timeout occurs and the connection is lost when this happens perhaps you should check the state of your host.
From the 2nd log seems the slave doesn't respond to the connection request from the host, check how you are handling the connection request in your peripheral code, and check if the device reaches the connection confirmation.
About the 3rd log, this seems very odd, i cant imagine why the central would start sending empty packets in such a high rate especially if has no data to send, the central itself or the device.
Thanks MT_dialog
I am using SDK 3.0.10.1.
The problem is the connection breaks randomly.
For 12ms connection interval, sometimes the connection goes well for more than an hour; but sometimes they just disconnect after 2 minutes.
After I change the connection interval to 30ms, the average connected time gets much longer. But they still disconnect randomly,
When they get disconnect, there is really no application activity. I assume the BLE stack will take care of the connection, send/ack empty PDU automatically during that time.
Also, even they disconnected, they should be able to reconnect. But once they disconnect, I could not connect to the slave anymore, no matter I use my master application or use some standard device e.g. IPhone.
To me, it looks like part of slave's BLE stack is screwed up: slave could receive the connection request from master(it stops advertising) but it could not send an ack. So, the both side just disconnect after 1s. Afert that, slave begin to advertise again.
These are everything I got from my test.
Thanks a lot for the help!
By the way, I use the same code base for all tests. I am sure the slave side send the connect request confirmation, otherwise they would not connect. Again, the problem is the connect-----keep the connection for a random time---disconnect and never able to reconnect anymore.
Thanks,
Hi zwang308,
I ve implemented a central and a peripheral on the SDK 3.0.10.1 (no profiles included) i ve set the connection parameters to the host identical with your's with a connection interval equal to 12.5 ms between a pro and a basic kit. So far everything seems correct and the connection lasts as long as i could watch (more than two hours) also i am able to reconnect to host, (in extended sleep mode yet). Have you implemented any profile or any other functionallity in order to try and reproduce your issue?
Thanks MT_dialog
Hi MT_dialog,
What I did is pretty much the same as your DSPS project.
I develop an I2C driver and put data into nvram when I receive a package. In app_asynch_trim function, I pull the buffer and send data through BLE.
But in my test, I disable the I2C driver and the pull function in app_asynch_trim. In the test, the two DA14580s are really doing nothing when they disconnect.
From my observation, they keep sending empty PDU for a while and suddenly the master stop to send empty PDU. Do we have method to know why master stop sending empty PDU? Could we get a message from the kernel to know why they disconnect.
To me, it seems like they disconnect randomly. We did lots of test last week, sometimes the connection is good for more than 16 hour; sometimes they disconnect after 2 minutes.
It really confused me.
谢谢!
Hi zwang308,
Can you run your master form keil and check when this issue occurs, what is the state of your master ? Also are your handlers indicate that there is a disconnection either by the master or at least by your slave when he stops receiving data by your host. You can place a break point to your device and check what you get as a disconnection reason. There is no way to check why the master doesn't send any data, an error indication or something like that since this should not happen the stack should always keep the connection.
Thanks MT_dialog
Hi MT_Dialog,
For the first time they disconnect. The error code on peripheral is 0x13(REMOTE USER TERMINATED CONNECTION) while the error code on central side is 0x08(CONNECTION TIMEOUT).
After that, the error code on peripheral side is always 0x08 and central side as 0x3e(CONNECTION FAILED TO BE ESTABLISHED).
I don't think the code actually stops at a point because both side is keep switching between advertising/scanning and connected state.
Do we have any idea why that happens?
Do we have a method to soft reset the chip(run firmware from the beginning without reload the firmware or just reset BLE stack)?
Thanks a lot.
Hi zwang308,
We haven't observed anything similar in tests we performed (i.e. the central to just stop sending packets) although we managed to track a cornercase where the peripheral isn't able to connect under specific circumstances, the patch for the specific cornercase isn't yet released. For this kind of situations we propose either to perform a sw reset to the 580 or reset the stack.
You can try to either issue a reset by using the SW_RESET field of the SYS_CTRL_REG and set the starting address to the sysram (please check the datasheet) although its not a good idea and reboot from the original image is prefered (download the code in the 580 using the bootloader) also you can try to send a GAPM_RESET command in order to reset the stack.
Also is it possible to share some code (central and device) in order to have a look since we couldn't replicate the problem on our side?
Thanks MT_dialog
HI MT_dialog,
Could you please provide me a e-mail address for sending the code? You could reach me in person through e-mail.
Thanks a lot,
Hi zwang308,
Problem was due to an obsolete SDK.
Thanks MT_dialog
Hi, MT_dialog.
I am suffering the same error as Mr zwang308's. I am using the sdk version V3.0.8. You said it was due to an obsolete SDK, please tell me which sdk is ok. Thanks.
Hi menson,
The latest SDK is the 5.0.4 for the 580 device.
Thanks MT_dialog