WARP Project Forums - Wireless Open-Access Research Platform

The Rx power calculation in the 802.11 design uses the MAX2829 RSSI signal as digitized by the WARP v3 RF interfaces. We implemented the 2.4GHz RSSI -> Rx power mapping using the RSSI vs Rx power figure on pg. 16 of the MAX2829 datasheet. The comparable figure for 5GHz is on pg. 19.

yeowl wrote:

How you come out with the RSSI_OFFSET_LNA_LOW ... ?

We used the "RX RSSI Output vs. Input Power" graph on page 16 of the MAX2829 datasheet and overlaid lines for a best fit linear approximation. The graph in the pdf of that data sheet is really small, so it's helpful to blow it up an overlay a grid on top of it. See what we made here as an example. In that figure, RSSI_A is the analog RSSI output in Volts and RSSI_D is the digital RSSI output after going through the ADC.

The mapping of RSSI_A to RSSI_D in the linked pdf is a little cryptic and is purely a function of the design of the WARP v3 board. The ADC that is fed by the RSSI signal maps the following:

RSSI_A     RSSI_D
2.65 V       1023
   .65 V            0

That mapping will be exactly the same for the 5GHz line, so you should be able to re-use the "RSSI_A = (2/1023)*RSSI_D +3/5" mapping.

Long story short, you need to make a similar approximation to the "RX RSSI Output vs. Input Power" on page 19 of the MAX2829 datasheet.

yeowl wrote:

Is the RSSI_SLOPE_BITSHIFT going to change for 5GHz?

The input to the calculate_rx_power function isn't actually the raw RSSI value out of the ADC (i.e. RSSI_D). It's a rolling sum version of it to help smooth over noisy samples. The length of that sum is defined as 8 samples in the design. More generally, it is defined as PHY_RX_RSSI_SUM_LEN in wlan_phy_util.h. The RSSI value passed to the calculate_rx_power function needs to be divided by 8 in order for it to be in the range of [0,1023] for the mapping of RSSI_D to RSSI_A to make sense. That's what the RSSI_SLOPE_BITSHIFT = 3 is referring to (log2(8) = 3). Shifting the value to the right by 3 bits is just a fast way of dividing by 8. You shouldn't change RSSI_SLOPE_BITSHIFT in the move to 5GHz since you will still be getting the 8 sampled-summed RSSI values.

You can implement whatever function you want to best fit the data in that graph. Eyeballing the graphs, it looks like a piecewise linear function might be a good fit. Each of the three curves seems to have an inflection point. To the left or two the right of that inflection point the curves appear mostly linear but different slopes and y offsets.