CSI Acquisition Data Loss

[ZcxToby]

Hello everyone, I encountered some strange issues when using Nexmon to extract CSI data.

I use two AC86Us as the transmitter and receiver respectively. The AC86U acting as the transmitter operates on channel 157 with 80MHz bandwidth in 4x4 MIMO mode. I use iperf3 to transmit packets at a fixed rate between these two devices, approximately 170 packets per second, thereby fixing the sampling frequency at about 6 milliseconds.

The third AC86U as a monitor device, configured with the parameters: /jffs/mcp -c 157/80 -C 15 -N 1 -m, to receive data via four receiving antennas.

The expected result is that when iperf3 transmits 1000 packets, the CSI extractor should capture 4000 pcap packets from the four antennas respectively. In other words, I should obtain data with a dimension of 256X1000X4X1 at a sampling interval of 6 milliseconds. However, I usually can only capture 1000 packets through the CSI extractor, and the average sampling interval is significantly reduced. On very rare occasions, I can capture around 4000 CSI packets, which is very confusing to me.

[zeroby0]

Csi data is copied through ethernet headers from phy device to processor, which takes a while to do. The radio is deafened while this is happening so that the csi registers don't get overwritten on the device. All this means that some samples will be lost because the csi data copy is underway when they're in the air. The free your csi paper linked in the readme is a very recommended read when working with this software.

[zeroby0]

Also, how are you measuring how many packets iperf3 generates? It might not directly correlate with the number of wifi packets.