NRF52810 RSSI accuracy in long time testing

RE: NRF52810 RSSI accuracy in long time testing

xulin — Fri, 04 Jan 2019 02:15:42 GMT

Thanks for your reply.

RE: NRF52810 RSSI accuracy in long time testing

Kenneth — Wed, 02 Jan 2019 15:50:41 GMT

Have in mind that the accuracy is +-2dB TYPICAL (not MAX), that in mind I would say that your measurements are as expected.

Best regards,
Kenneth

RE: NRF52810 RSSI accuracy in long time testing

xulin — Wed, 26 Dec 2018 05:05:11 GMT

As you mention that power receiver get is a little high, I changed the signal source’s transmit power to -30dBm, and send 408 packets every second. After 36 hours , the statistic result is below:

Data statistic from 2018-12-21 12h~23h

Frequency	-74 dBm	-75 dBm	-76 dBm	-77 dBm	-78 dBm	-79 dBm
Freq1	357,840	17,291,877	46,108	/	/	/
Freq2	/	3,233,995	14,459,380	2,397	/	/
Freq3	/	/	95,107	16,101,572	1,439,553	22
Freq4	/	/	2,199,524	15,484,449	11,056	/

Data statistic from 2018-12-22 00h~23h

Frequency	-74 dBm	-75 dBm	-76 dBm	-77 dBm	-78 dBm	-79 dBm
Freq1	2,243,034	33,121,204	27,350	/	/	/
Freq2	/	11,276,635	24,113,425	1,413	/	/
Freq3	/	/	971,500	33,454,642	881,368	2
Freq4	/	4	9,826,233	25,559,532	4,597	/

Receiver get -74~-79dBm power, and the RSSImax - RSSImin <=4. In the datasheet, it says the “RSSI Accuracy Valid Range” is -90dBm~-20dBm. In my first test, the RSSI is about -30dBm which is in the range, but the RSSI distribution when got about -30dBm is not in the accuracy range +/-2dB.

The RF attenuator, mine is quite cheap, and it has fixed attenuation during test. We test the attenuator using net analyzer, the frequency response is stable under the fixed frequency.

Maybe our test environment is not so perfect, it’s tested only in a shield case other than in anechoic chamber. We expected that when all the component are placed quietly in shield case (most external radio should be blocked, the internal radio bounce maybe exist) and connected by cable, the RSSI should have little deviation. The test result above is still a little more than our expection.

RE: NRF52810 RSSI accuracy in long time testing

awneil — Thu, 20 Dec 2018 22:01:30 GMT

[quote userid="55634" url="~/f/nordic-q-a/41810/nrf52810-rssi-accuracy-in-long-time-testing"]By calculate signal source RSSIs got from different receiver we can get the approximate position of signal source[/quote]

This comes up again and again with monotonous regularity.

The word "approximate" is key - and its a very rough approximation indeed!

RE: NRF52810 RSSI accuracy in long time testing

AmbystomaLabs — Thu, 20 Dec 2018 15:37:04 GMT

I hope someone from Nordic chimes in but in terms of radio architecture, most RSSI's are calculated using the received SN ratio and the estimated noise figure of the front end to figure out what the RSSI should be. It is rare they actually do it by measuring the actual power. Since this is not the actual received signal strength it is influenced by interferers such as WiFi (especially the synth spurs and IMD noise that sit at the edge of OFDM waveforms).

Then to further confound the data, any gain control done by receiver can cause cyclical changes in the SN and thus the RSSI.

Finally, radios and antennae are never flat across their passband. So power at chan 1 won't be the same as at chan 4, etc. Radio that have closed loop control can calibrate out these errors, but a simple bluetooth radio has no closed loop power control and thus you won't even have the same power on all channels and thus won't have the same RSSI.

Also rooms never have flat response. 2.4GHz easily bounce off of stuff in the room, metal work may act as reactive loads and all the reflected signals will either be constructive or destructive as you move around the room and the phases relationships change.

On your end easiest thing would be to characterize your system and cal out the errors in your software. Then once you start with flat response your work might get easier. But all your numbers will be nuts anyway as soon as someone walks around the room or moves the furniture.

On your test setup above doing conductive measurements all the same is true. The radio will have non-flat 50ohm performance, micro coax cables are garbage and have non-flat performance along with the small hirose connectors are rarely close to 50ohm which only makes the radio response worse.

And who made your attenuator??? Good RF attenuators are about $1,000 and are calibrated regularly. A high quality step attenuator might be $5,000. A simple 30dB pad is maybe +/- 3dB and that's if you actually spent a few bucks on it.

You should also do the test at a lower level. You will never see -30dBm in real life and I doubt Nordic optimized the gain control for that range.