Channel Sounding Performance Metrics#

Accuracy#

Silicon Labs has performed real environment testing in an indoor office setting to best represent the performance of the Channel Sounding distance estimation solution using BRD4198A boards. The images below show the floor map of the office with 7 static devices and one mobile device. The mobile device (H) and one of the static devices (1) are mounted on a ceiling rail robot. The rest of the devices (2, 3, 4, 5, 7, 8) are either mounted on the ceiling (antenna pointing downward) or on the side walls (antenna bent to point upward) at varying heights. Listed below are the heights at which the devices are placed in the office environment.

Device H: 2.144 m
Device 1: 2.140 m
Device 2: 2.080 m
Device 3: 2.100 m
Device 4: 2.440 m
Device 5: 2.150 m
Device 7: 2.700 m
Device 8: 2.125 m

The table below provides the performance metrics of PBR and RTT distance estimation based on measurements taken at multiple distances between 0.5 m and 34 meters, and in different measurement configurations:

Conducted, line-of-sight (LOS), and non-line-of-sight (NLOS) environments
- In conducted environment, the antenna paths of 2 BRD4198A boards are connected together using a power splitter to have 4 different cable lengths -> 2.54 m , 6.4 m, 15 m and 17 m.
- In LOS environment, the distance is estimated at multiple positions between devices H and 1 ranging from 0.5 to 19.5 meters using 72 channels.
- In NLOS environment, the distance is estimated at 21 different poistions between the devices mounted in the office (2-5, 2-4, 4-7, etc.).
SL_RTL_CS_ALGO_MODE_REAL_TIME_BASIC and SL_RTL_CS_ALGO_MODE_STATIC_HIGH_ACCURACY algorithm modes referenced by Real-time and Static respectively.
- In SL_RTL_CS_ALGO_MODE_REAL_TIME_BASIC mode, the distance is estimated 200 times at each position (i.e. 200 CS prodedures).
- In SL_RTL_CS_ALGO_MODE_STATIC_HIGH_ACCURACY mode, the distance is estimated 4 times at each position (i.e. 400 CS prodedures as 100 CS procedures = 1 distance estimation).
High, Medium, and Low Channel Map settings
- High: 72 channels are used for CS, 1 MHz channel spacing
- Medium: 37 channels, 2 MHz channel spacing
- Low: 20 channels, 1 MHz channel spacing

The statistical values in the table are derived from the combined data set of all distance estimations in each configuration.


Measurement configuration				90th percentile of abs error (m)	95th percentile of abs error (m)	Std of the signed error (m)	Mean Signed Error (m)
Measurement mode	Environment	RTL algorithm mode	Channel Map	90th percentile of abs error (m)	95th percentile of abs error (m)	Std of the signed error (m)	Mean Signed Error (m)
PBR	Conducted	Static	High	0.4	0.4	0.26	-0.04
		Real-time	High	0.4	0.4	0.26	-0.04
			Medium	0.4	0.4	0.26	-0.04
			Low	0.43	0.43	0.29	0.03
	LOS	Static	High	0.89	1.21	0.47	-0.1
		Real-time	High	0.44	0.48	0.25	0.15
			Medium	0.41	0.45	0.21	0.13
			Low	1.45	2.11	0.71	0.67
	NLOS	Static	High	3.05	3.29	1.82	-0.39
		Real-time	High	3.75	4.79	1.61	0.97
			Medium	4.04	4.63	1.71	1.45
			Low	9.3	9.5	4.15	2.02
RTT	Conducted	Real-time	High	2.97	3.06	0.32	-2.51
	LOS			4.21	5.15	1.78	2.4
	NLOS			8.95	10.15	3.49	4.31

Note: Increasing the channel spacing results in a reduced maximum range. This relationship is mathematically expressed as , where represents the channel spacing. For further details, refer to the Fundamentals section. Additionally, as the number of channels decreases, the accuracy of distance estimation also diminishes.

The following Cumulative Distributive function (CDF) plots show the cumulative probability of the absolute distance errors in each measurement configuration for PBR and RTT measurement modes. The absolute errors are calculated for each distance estimated per measurement configuration.

X-axis is the absolute error
Y-axis is the cumulative distribution function

PBR#

Conducted Environment

LOS Environment

NLOS Environment

RTT#

Note that Channel Sounding is still in the development phase and software improvements are made in each software release to improve the performance of both PBR and RTT modes. See Known Issues and Limitation section for more information.

Update Rate#

The measurement update rate of channel sounding is affected by the chosen channel sounding procedure interval, number of connections, connection interval, number of antenna paths, PHY, and the number of channels used for channel sounding.

The table below provides the performance metrics of PBR update rate using the out of the box configurations of the SoC Initiator and SoC Reflector examples, for a single connection.


Reference board	Number of Antenna Paths (NAP)	Number of channels	Update rate (Hz)	Calculation time (ms)
BRD4198A	1	72	~6	~46
		37	~8.9	~10.5
		20	~10.2	~3.8

Note that, by default, the examples use 2M PHY for connection, a 7.5 ms connection interval and the initiator is run at 80 Mhz. Additionally, the maximum CS procedure count is set to 1, meaning that only one CS procedure is scheduled by the controller at a time, and the application restarts a new procedure sequentially. While this approach provides better flow control and stability, it is not optimal for measurement update frequency. By setting the max CS procedure count to 0, CS procedure interval to 12, it is possible to achieve measurement frequency upto 10 Hz using single antenna path and 72 channels.

Optimizing Performance#

The antenna of a device plays a significant role in achieving accurate distance estimations based on its characteristics such as group delay, radiation pattern, and others. To learn more about the antenna design guidelines and considerations to improve the accuracy of distance measurement using Channel Sounding, see AN1493.