This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

WL1837MOD: Poor Wifi/BT Coexistence Performance

Part Number: WL1837MOD
Other Parts Discussed in Thread: WL1837

Good Morning,

I am trying to bring up a WiFi/BT solution that supports data transfer via TCP sockets on the WLAN while there is an active BT audio (A2DP) sink running. I can operate both WLAN and BT by themselves and see performance on par with what the TI application documents suggest is reasonable. The issue is that when I attempt to do both simultaneously, I observe a major degradation in the performance of the WLAN link.

I am currently testing with Linux on a BeagleBone Black using the Element14 Connectivity cape (WL1837MOD). I've included some collected performance data below. Most of the numbers look reasonable, however, there is a >90% reduction in performance of the 2.4Ghz WLAN when I enable the A2DP stream. When WiFi is operating at 5Ghz there is no degradation, which makes me suspect that something isn't quite right with the coexistence algorithm rather than a bottleneck in the CPU.

Referencing SWRU424A and comparing the collected numbers in section 9.1 with the numbers in 9.4.1 indicates that the drop in performance should be more along the lines of 50%, rather than >90%. This makes me wonder if I've configured something incorrectly or missed a step that is needed to get the coexistence algorithm to work as designed.

Any thoughts or insight on these results would be appreciated. Ultimately I'd like to get some confidence that these numbers reflect real-world expected performance so that we can adjust our design accordingly. If we should be getting better data rates I'd like to understand what we're missing. Please feel free to let me know if any additional information would be helpful.

Thanks!

Test Details:

  • WL1837/BeagleBone running in AP mode (I have also tried client mode using a local router with similar results)
  • BeagleBone is running a custom Yocto image based on Linux 4.19. The bluetooth is managed by our own BT stack (ie. not using bluez or Bluetopia)
  • iperf3 server running on BeagleBone (iperf3 -s)
  • iperf3 running on client device (iperf3 XXX.XXX.X.X -V -t30) (tried from Android phone, iPad, and laptop, all show similar results)
  • Results in the table below are shown from the perspective of the client  device
  • BT is paired/connected for all tests. Have tried variations of streaming from same device as the iperf3 client as well as other devices with no perceptible improvement
  • A2DP stream is activated by simply playing music or voice audio via some Android/iPad media player
AP ConfigA2DP StreamingChannel (width)iperf3 Result TX (Mbps)iperf3 Result RX (Mbps)
2.4GHz WPA PSK Yes 6 (20 MHz) 2.44 0.229
2.09 0.0779
2.44 0.262
2.44 0.328
2.44 0.264
2.4GHz WPA PSK No 6 (20 MHz) 26.5 24.7
26.2 24.2
29.3 27.6
30.1 28.1
24.1 22.1
6 (20 MHz) MIMO 78 58
75 65
78 50
75 46
77 62
5GHz WPA PSK Yes 44 (20MHz) 48.0 47.1
47.4 46.7
45.9 45.6
48.8 48.4
48.1

47.6

157/159(40MHz) 95.4 95.3
94.5 95.2
95.7 95.3
5GHz WPA PSK No 44 (20MHz) 48.4 47.9
50.0 50.0
49.9 49.5
50.1 49.8
49.9 49.5
157/159(40MHz) 92.7 92.5
96.2 96.1
  • We will review the results and get back. Are the above iperf number with AP mode or STA mode?

    Thanks

  • Thanks for taking a look. The results in the table are from the WL1837 in AP mode. I have also run the same set of iperf tests with the WL1837 in STA mode and observed the same behavior.

  • Hi,

    What version of WiLink8 firmware are you running ? Can you pls test it with fwr 85,84,81 and 79 ?

    https://git.ti.com/cgit/wilink8-wlan/wl18xx_fw/log/

    You just need to replace wl18xx-fw-4.bin in /lib/firmware/ti-connectivity/

    Thanks

    Saurabh

  • The original metrics were collected using firmware version 76. I have tried each of the requested versions and each shows similar behavior. I have noticed that the amount of degradation seems to be highly dependent on what the A2DP source device is. See the table below for more details. The main takeaways are:

    The iPad shows the best performance and is closest to the numbers quoted in the WiLink8 application paper. Note that the iPad is the only device tested that supports MIMO on the client side. All devices perform significantly worse when they are both the A2DP source AND the iperf3 client at the same time, with the OnePlus3 device being the worst of the bunch.

    Any ideas why certain clients would be affected worse than others?

    Test Setup: All devices connected to 2.4GHz AP, BT A2DP streaming from listed device, testing using iperf3, average over 30 seconds. Testing in residential neighborhood with multiple other active APs and stations in range.

    Device
    A2DP Source
    WMM Enabled
    MIMO
    Channel Width
    WLAN TX Throughput (TCP)
    WLAN RX Throughput (TCP) (-R flag)
    iPad None Yes Yes 20 80 60
    Android 1 25 24
    Android 3 31 29
    iPad (Self) 13 12
    Android 1 (Fire Tablet) None Yes No 35 31
    Android 1 (Self) 7 3
    Android 2 (Lenovo) None Yes No 32 21
    Android 1 10 9
    Android 3 (OP3) None Yes No 34 31
    Android 1 10 11
    Android 3 (Self) 2 0.9