Wifi: increase the sample rate of TX and RX

[henrychang81]

@bastibl Hello, I'm using your example code "wifi_tx.grc" and "wifi_rx.grc", and have modified them into the following 2 images:

• This is the TX flow graph:
  

• This is the RX flow graph:
  

• Objective: Our main goal is to use WiFi CSI data for object sensing. Increasing the sample rate is crucial for our experiments. We have made some initial attempts but are still not sure how to improve the sample rate. Could you please provide some suggestions?

• We conducted the following experiments:

  • TX at 10MHz and RX at 10MHz:
    This setting operates normally
    

  • TX at 20MHz and RX at 20MHz:
    With this configuration, there seems to be some delay on the RX side
    

  • TX at 30MHz and RX at 30MHz:
    There are errors on the RX side - the dsp_rate cannot be evenly divided by samp_rate, and the waveform is completely stuck
    

  • TX at 40MHz and RX at 40MHz:
    The other scenario produces the same error
    

  • TX at 50MHz and RX at 50MHz:
    The waveform on the RX side is completely stuck, and no reason is given.
    

• My hardware configuration:

  • Two USRP X310 units (one for TX, one for RX)
    • Both USRPs connected to the same computer
  • Network card:
    • Intel E810-XXV
      • Features two ports, Both ports support 25GbE
  • Motherboard:
    • Gigabyte X670 AORUS ELITE AX
  • CPU:
    • AMD R7 7700

• My software configuration:

  • MTU configuration:
    • Set to 9000
  • Socket buffer configuration:
    • rmem_max=33554432
    • wmem_max=33554432
  • Ring Buffer configuration:
    • rx=8160
    • tx=8160
  • Network card features:
    • gro: enabled
    • gso: enabled
    • tso: enabled

• USRP stream_args configuration:

  • recv_frame_size=8000
  • num_recv_frames=512
  • recv_buff_size=33554432

Any advice you could offer would be greatly appreciated.
Thank you for your time and assistance.

Henry

[bastibl]

I assume you're PC is overloaded. Maybe not all cores but there might be one core (e.g., divide or sync short) that will max out a CPU core. So you will drop most of the samples and not decode the data.