NRF52840 dongle sensitivity

[Libertas]

I am testing my own mesh protocol based on 802.15.4 using the NRF52840 dongle.

The reported RSSI value from nrf_radio_rssi_sample_get() never goes below -89dbm.

On the other side, there are over 60 devices running on TI CC2538, their reported RSSI can do down to -97dbm, in line with the datasheet.

The NRF52840 specs say that the sensitivity shall be 100dbm.

So what could be the possible cause? MCU? Is the antenna on the dongle? or software?

[Bendik Heiskel]

Hi,

RSSI is not a good measure of the receiver sensitivity, as a ADC is used to meaure the voltage in the receiver before the demodulator. That means that if there is any other devices transmitting on the 2.4GHz the RSSI value will be affected.

In what type of environment is the RSSI measurements being done? Is the devices isolated from any external interference?

The performance of the antenna on the dongle is dependent on the devices its connected to. The ground plane on the PCB is very small and the antenna is affected by the increased ground plane when connecting to a device.

Are you seeing the same RSSI values when testing using a nRF52840 DK?

 

Best regards,

Bendik

[Libertas]

Hello Bendik

Thank you for your response. The NRF52840 dongle is acting as a transceiver to my Hub. Currently, the dongle is connected to a PC. The measurement is pretty consistent.

Apart from the Hub, I have over 60 devices, all based on TI CC2538. Both the dongle my devices are using an output power of 7dBm. Between the dongle and my other devices, there is a consistent difference of 4-5 dBm in the RSSI reading.

Below is the reported neighbor table of the Hub(NRF52840 dongle) and a device called "Central Window"(TI CC2538).

As you can see, from the Hub, the RSSI of "Central Window" is -87dBm. From the "Central Window", the RSSI of the Hub is -91dBm. The readings are consistent between the Hub and other devices. Since all devices are set to output at 7dBm, the antenna of the dongle is performing about 4dBm worse. It is understandable because the dongle's antenna is significantly smaller and my devices are optimized to the maximum. Probably I won't be able to get that level of optimization in the future.

The problem I see is that, while all other devices can receive RSSI down to -97dBm (noise floor of CC2538), the RSSI of the dongle's receiver never goes below -90dBm. In other words, the dongle sees a lot fewer devices than other devices.

The 60 devices spread across a 3-floor building. The Hub is placed on the edge. My testing environment is far from perfect. But the result is so consistent that I believe the result is highly significant.

The whole thing is part of a bigger project. The Hub for Raspberry Pi can be downloaded from GitHub. I will release the complete source for Nordic MCUs very soon.

https://github.com/LibertasIoT/libertas-rpi-img

[Bendik Heiskel]

The range for the RSSI are -90 to -20dBm: https://infocenter.nordicsemi.com/topic/ps_nrf52840/radio.html?cp=4_0_0_5_19_14_8#unique_1927333262

For lower than -90dBm the RSSI is inaccurate, meaning that it would not report the correct signal strength.

We haven't characterized the nRF52840 dongle antenna, so we don't have any numbers on antenna gain and radiation pattern. The antenna may not have a uniform radiation pattern leading to less antenna gain in certain directions. Based on the tables it seems that the hub and the "central window" device report similar RSSI for a couple of devices, but for others there is a large gap in the reported RSSI. This could be a result of a non uniform radiation pattern from the dongle antenna, assuming the hub and the "central window" device is located close to each other.

[Libertas]

Hi Bendik, Thanks a lot for the clarification. I missed that statement.

That explains why I never received any packet with reported RSSI below -90dBm.

BTW, the "Hub" and "Central Window" are far away, one floor apart. You can tell from the RSSI value from one another. I was not trying to compare the antenna performance in any way. I was just confused about the reported RSSI always above -90dBm.

One potential problem I am facing is that I use RSSI (actually an exponential moving average) to estimate the link quality. So the link quality depends on the RSSI  above the noise floor. With NRF52840 I will have to assume the noise floor is -90, with a 10 dBm gap from the -100 dBm sensitivity from the datasheet.