nRF7002 wifi throughput measurement

[jiang.ronghui]

Hi There:

I used the nRF7002 DK to evaluate its WiFi throughput performance.

The test environment is below:

    1. nRF7002 DK

    2. NCS version: 2.5.0

    3. Sample application: Wifi shell sample

    4. Client: nRF7002, connected to the AP by WiFi, running zperf;  Server: PC, connected to the AP by LAN cable, running iperf.

    5. Test command: Client: 'zperf tcp upload 192.168.0.26 5001 30 1k'; Server: 'iperf -s -i 1'

The results are as follows:

1, Connect to AP with WiFi 6 2.4Ghz 

uart:~$ wifi status
Status: successful
==================
State: COMPLETED
Interface Mode: STATION
Link Mode: WIFI 6 (802.11ax/HE)
SSID: TP-Link_DFE4
BSSID: F0:09:0D:E5:DF:E4
Band: 2.4GHz
Channel: 10
Security: WPA2-PSK
MFP: Optional
RSSI: -30
Beacon Interval: 100
DTIM: 1
TWT: Not supported

uart:~$ zperf tcp upload 192.168.0.26 5001 30 1k
Remote port is 5001
Invalid IPv6 address 192.168.0.26
Connecting to 192.168.0.26
Duration:       30.00 s
Packet size:    1024 bytes
Rate:           10 kbps
Starting...
-
Upload completed!
Duration:       30.00 s
Num packets:    51518
Num errors:     0 (retry or fail)
Rate:           13.41 Mbps

2, Connect to AP with WiFi 4 2.4Ghz 

uart:~$ wifi status
Status: successful
==================
State: COMPLETED
Interface Mode: STATION
Link Mode: WIFI 4 (802.11n/HT)
SSID: TP-Link_DFE4
BSSID: F0:09:0D:E5:DF:E4
Band: 2.4GHz
Channel: 5
Security: WPA2-PSK
MFP: Optional
RSSI: -29
Beacon Interval: 100
DTIM: 1
TWT: Not supported

uart:~$ zperf tcp upload 192.168.0.26 5001 20 1k
Remote port is 5001
Invalid IPv6 address 192.168.0.26
Connecting to 192.168.0.26
Duration:       20.00 s
Packet size:    1024 bytes
Rate:           10 kbps
Starting...
-
Upload completed!
Duration:       20.00 s
Duration:       30.00 s
Num packets:    48582
Num errors:     0 (retry or fail)
Rate:           12.65 Mbps

Regarding the above results:

1. I found that the throughput difference between WiFi 4 (802.11n) and WiFi 6 (802.11ax) is not significant, so I'm afraid that there may have some limitation about nRF7002 in software side.

     2. And the test results do not match the ideal outlined in the Nordic documentation.

             

Therefore, please provide information about the WiFi measurement environment and the examples used so that I can achieve the throughput specified in the datasheet.

[Håkon Alseth]

Hi,

 

The numbers in the PS has a dependency in terms of what type of throughput is measured. Please see the description text:

https://docs.nordicsemi.com/bundle/ps_nrf7002/page/chapters/functional/doc/qspi_timing.html

 

Your TCP TX/RX measured throughput is quite optimal, as shown in the wifi throughput sample documentation which lists a few scenarios:

https://docs.nordicsemi.com/bundle/ncs-latest/page/nrf/samples/wifi/throughput/README.html#results

 

Kind regards,

Håkon

[jiang.ronghui]

Thanks for your reply.
> The numbers in the PS has a dependency in terms of what type of throughput is measured. Please see the description text:
   I got it. The throughput is over QSPI/SPIM, but not WiFi, so it can not be compared with my my measurement. Is my understanding correct?

> Your TCP TX/RX measured throughput is quite optimal, as shown in the wifi throughput sample documentation which lists a few scenarios:
   If you mean that my measurement values are very close to the expected values of the nRF7002?

[Håkon Alseth]

Hi,

 

jiang.ronghui said:

   I got it. The throughput is over QSPI/SPIM, but not WiFi, so it can not be compared with my my measurement. Is my understanding correct?

Yes, this is correct.

jiang.ronghui said:

   If you mean that my measurement values are very close to the expected values of the nRF7002?

Yes, your results are even better than what we list in the examples.

 

Kind regards,

Håkon

[jiang.ronghui]

Thanks for your reply.
And sorry I just want to confirm again.

1. QSPI/SPIM is the communication interface between the nRF7002 and nRF5340, so its throughput is not directly related to the WiFi throughput of the nRF7002 (nRF7002 --> AP), correct?

2. In a situation where there are no limitations of the nRF7002 DK, what will the WiFi performance be?(the maximum throughput that can be achieved in real-world situations)?

[Håkon Alseth]

Hi,

 

jiang.ronghui said:

1. QSPI/SPIM is the communication interface between the nRF7002 and nRF5340, so its throughput is not directly related to the WiFi throughput of the nRF7002 (nRF7002 --> AP), correct?

Correct.

jiang.ronghui said:

2. In a situation where there are no limitations of the nRF7002 DK, what will the WiFi performance be?(the maximum throughput that can be achieved in real-world situations)?

As previously mentioned, your results for TCP throughput is higher than the ones we list in the sample itself:

https://docs.nordicsemi.com/bundle/ncs-latest/page/nrf/samples/wifi/throughput/README.html#results

 

It is not likely that you get any higher throughput than what you've already measured.

 

Kind regards,

Håkon

[jiang.ronghui]

Sorry for another question.

I conducted several more tests to compare the performance differences between Wi-Fi 6 and Wi-Fi 4 in different testing environments.
However, I found that the test results for Wi-Fi 4 and Wi-Fi 6 still did not show significant differences.
Therefore, I would like to ask if the current measurement results in line with the design specifications? If so, what is the significance of using Wi-Fi 6? What are its advantages?

[Håkon Alseth]

Hi,

 

jiang.ronghui said:

I conducted several more tests to compare the performance differences between Wi-Fi 6 and Wi-Fi 4 in different testing environments.
However, I found that the test results for Wi-Fi 4 and Wi-Fi 6 still did not show significant differences.
Therefore, I would like to ask if the current measurement results in line with the design specifications? If so, what is the significance of using Wi-Fi 6? What are its advantages?

The advantages with Wi-Fi 6 is not in speed, but in features like Target Wake Time.

The on-air speed will be more or less equal between Wi-Fi 4 and Wi-Fi 6. This can be seen with the MCS index charts as well:

https://mcsindex.com/

  

Kind regards,

Håkon