RSSI accuracy

[Stanislav Silnitskiy]

Hello!

As per documentation, RSSI is measured during the time, when particular packet is being received. In particular, nRF51822 Product Specification v3.1 states RSSIperiod as 8.8 us at minimum. Does it mean, that it can differ for packets with different length?

If so, does it mean, that overall accuracy of RSSI can be improved, implementing a sophisticated filter over greater number of RSSI samples for shorter packets, compared to averaging for longer packets, as it is implemented in hardware?

[Kenneth]

The RSSI measurement is done one time for each packet received, there is a radio short 'ADDRESS_RSSISTART' that can be enabled to do one RSSI sample after each radio address match.

[Stanislav Silnitskiy]

As a sidenote, I have implemented the feature with NRF_RADIO->MODE = RADIO_MODE_MODE_Nrf_2Mbit << RADIO_MODE_MODE_Pos and got the following observation: For the packet with payload 120 bytes long I got 39 RSSI samples. If my calculations correct, every RSSI sampling iteration consumes 12+ us, that corresponds to ref. manual of nrf51 (8.8 us as minimum for RSSI sampling). This result allowed to calculate mean variance of received packet's RSSI and I was pleased to see live float calculation of a such for EVERY packet (it must be pretty tough for float point emulation, I suppose). Experiment was done with four identical beacons, emitting 120 byte packets every 10 ms. No idea, either it could be described as a high load, I noticed something about 5% packets with CRC errors. For completeness: 2500 MHz frequency was used.