We have a Python application that connects to 10 BLE devices at a time and performs a manufacturing test.
Here is the work flow of the test:
Connects to the first of 10 devices.
Reads and writes to a few characteristics one of which reboots the peripheral device
Reconnects to device 1, enables notifications on 1 characteristic, and listens for ~170-200k of data
Disconnects from device 1, turns device 1 off, and repeats step 1-3 on the next device
As we progress through the devices the data transfer for each device slows gradually, the first device is fast, the second device slightly slower etc, even though we are disconnecting from each device after performing the test. We assume that there may be something related to our usage that is causing this slowdown, however after reviewing the relatively small amount of code in the project we haven't been able to explain anything in the Python code that could be causing the slowdown.
We've tried turning each device off as we test to ensure low interference, removing cached devices from the Central, and restarting the bluetooth service in between tests. After testing these changes, the sudo service bluetooth restart is the only solution that will mitigate the delay growth during data transmission.
We are looking for any ideas on solutions to the described situation or debug techniques.
Thank you in advance!
Versions used:
Ubuntu 20.04, Bluezero 0.4.0, bluez 5.53" (edited)
Related
I'm currently working on a research project which revolves around me getting to know the transfer speeds of BLE in a simple setting. To be specific, I'll be working with an Arduino Nano 33 BLE board. I'm well aware that BLE v5 is capable of reaching speeds of up to 1Mb/s (Mega-bits/s) but is unrealistic in real-world applications. Are there any resources that I can get the transfer speeds of BLE? If not, I'm guessing I will have to work with an experimental setup for finding the speeds for my specific use-case. Thank you in advance!
The Bluetooth Radio on the Arduino Nano 33 BLE is built into the NRF52840 MCU.
More information on the Radio is found here:
https://infocenter.nordicsemi.com/topic/ps_nrf52840/radio.html?cp=4_0_0_5_19
There are definitely ways to determine real throughput of this peripheral, you just need a packet sniffer (link below) or have a transmitter/receiver board-to-board setup.
Packet Sniffer:
https://www.nordicsemi.com/Products/Development-tools/nRF-Sniffer-for-Bluetooth-LE
The board-to-board setup will have the relay serially send the number of packets received to a PC for analysis. You could configure the relay board to send the number of packets received within a certain interval (using the Timers in the NRF52840 with <microsecond precision).
If your project changes and you no longer need to use BLE and you would like to increase throughput to ~1.5Mbps, I developed and tested a working configuration of the NRF52840's Radio on the Arduino Nano 33 BLE. Link below.
https://forum.arduino.cc/t/stream-binary-data-from-arduino-nano-33-ble-to-pc-via-ble/917206
I am using BLE v4.1 for my application where I am able to transfer 200 packets of length 20 bytes each from peripheral to central side.
I want to transfer data via BLE and play some music while the data is being transferred. The music file is available locally and will be played on earphones(can be wired or wireless BlueTooth earphones)
I am facing an issue of packet drops around 15-20% when I am playing audio through wireless earphones while data is being transferred via BLE and audio is played on phone Bluetooth earphones(tried on Android and iOS, both) simultaneously.
With wired earphones connected; there is no issue, I can transfer data via BLE and play audio smoothly without any issue but things are different with wireless Bluetooth earphones case. I tried it with different Android phones(Android 11, different manufacturers) and different BlueTooth earphones, the issue still exists.
What could be a cause for the same and how can I debug it?
Bluetooth and Bluetooth Low Energy has only one radio that can only do one thing at a time. The connection events are scheduled and cannot overlap. This means the throughput is shared among all connected devices. You will get a higher throughput per device with only one connection than with two connections.
You will have to experiment with connection interval and connection event length to find parameters that seems to work, and test with multiple Bluetooth controllers (phones), since all have different scheduling algorithms.
I am able to use bleak and get data from all 5 BLE sensors. But the problem is that I am unable to identify which data is from which device. I mean i need a string representing the address of the device alongwith the data itself.
I was able to get data simultaneously from all the BLE modules using bleak in Windows, as well as on raspberry pi.
The only problem with Windows is that if you use Bluetooth 4.2 adapter & above. It will be much better for the high speed data rate, and proper devices connection handling. As with the 4.0 adapter, I have always gone through one or two exceptions each time i started the script, and the maximum number of connections I got was 3.
When i tried this script on Raspberry pi 3b+, it has on-board chip of Bluetooth 4.2. It was able to give high speed data rate, and my 5 sensors were connected to it simultaneously.
Also, The two_devices examples in bleak source code on Github, is very good example for starting with further coding.
And if you run the script and found the data on console, but you were unable to get that which data is from device. Then you need to use functools, (what it does is it will inject the client you're connected to at present, with the callback function, and it will make the work much easier).
Check this out on Github: https://github.com/hbldh/bleak/issues/601
I am trying to evaluate the suitability of some different wireless interfaces for our project on 2xRaspberry Pi 4 and currently I’m evaluating Bluetooth Low Energy. Therefore I have written an Central and Peripheral device application with the Qt framework (5.15). In my case the latency time between messages is important, because of some security aspects. The message size of each command is around 80-100 Bytes. In one of my tests I have sent 80 Bytes commands every 80ms. Ideally the messages should be received on the other device in 80ms interval as well. For the LAN (TCP) interface this test works well.
For Bluetooth Low Energy I observed that messages, which are sent from Peripheral to Central work quite good and I measured no big delay. Different results I got for the Central to Peripheral direction. Here, I have received the messages in the interval of 100ms to 150ms really exactly. It seems that there couldn’t be a very big magic behind it, so is there any plausible explanation for this? I tested it with a Python script as well and I observed the same results. So it seems that the Qt implementation shouldn’t be the problem.
During research I found out, that the connection interval may influence this, but in Qt the QLowEnergyConnectionParameterRequest (QLowEnergyConnectionParameters Class | Qt Bluetooth 5.15.4) doesn’t work for me. Is there any command, where I can set the connection interval for test purposes at the command line on Linux?
Kind regards,
BenFR
It is possible that your code is slower from central to peripheral because WRITE is used instead of WRITE WITHOUT RESPONSE. The difference is that WRITE waits for an acknowledgement, therefore slowing the communication down, while WRITE WITHOUT RESPONSE is very much like how notifications/indications work in that there's no ACK at the ATT layer. You can change this by changing the write mode of your application and ensuring that the peripheral's characteristic supports WriteNoResponse.
Regarding changing the connection interval, the change needs to be accepted from the remote side in order for it to take effect. In other words, if you are requesting the connection parameter change from the peripheral, then the central needs to have code to receive this connection parameter change request and accept it.
Have a look at the links below for more information:-
How does BLE parameter negotiation work
Understand BLE connection intervals and events
The different types of BLE write
I'm writing a Bluetooth Low Energy library. For now, it will only run on Linux (and with Bluez 5.41+).
I'd like to have tests that can be run on any computer with BLE support and test the entire stack (application, host and controller), without requiring extra hardware (another BLE-capable device). Requiring extra hardware is a great way of making sure the tests won't be run more than once in a blue moon.
For that end, I'd like to use my computer to connect to itself (as both a peripheral and central) during the tests. It seems like this should be possible, since I can be connected at the same time as peripheral and central to other devices. But advertisements from my own computer never show up on (my own computer's) BLE scans.
Is it all possible to have the computer connect to itself in this way?
No, not if you only have one Bluetooth controller.
The reason is that radio peripherals can only transmit OR receive in a given time moment.