CMD_PROP_CS¶

The carrier sense command monitors the RF activity on a channel and decides, whether a channel is occupied or free. It can be used to implement a CSMA scheme where a TX command only executes when nothing else occupies the channel. Channel detection is based upon two criteria: RSSI measurement and Preamble Quality (PQ).

Execution¶

The RF channel can be in several states:

Invalid:	This is the default state at CS start.
Idle:	No signal was found.
Busy:	A signal was detected (RSSI above threshold or correlation succeeded).

scale 0.8

[*] -> Invalid
Invalid -> Idle
Invalid -> Busy
Idle --> Busy
Idle -> Idle
Busy -> Idle
Busy -> Busy

Idle --> [*]
Busy --> [*]
Invalid --> [*]

Figure 28. Channel states and possible transitions.¶

All possible transitions between these states are illustrated in Figure 28.. Carrier sense always starts in Invalid state. Depending on the measured RSSI or the preamble quality, the channel goes either into Idle or into Busy state. The operation ends or may continue depending on the configuration. In addition, a timeout may occur which ends the measurement process immediately. After that, a status code is written to the radio operation and an execution result is generated for the evaluation phase.

A detailed sequence description of the carrier sense execution phase is shown in Figure 29.. In total, four procedures run in parallel:

watchChannelState():
watchRssi():	Optional. Constantly measures the RSSI value and updates the channel state.
watchCorr():	Optional. Correlates the RF signal with the expected preamble sequence and updates the channel state. The preamble settings are configured via CMD_PROP_RADIO_SETUP or CMD_PROP_RADIO_DIV_SETUP.
	Watches and evaluates the channel state changes caused by `watchRssi()` or `watchCorr()`.
waitForCsEndTrigger():
	Optional. Waits for an end trigger (timeout) and evaluates the channel state when triggered.

Figure 29. The execution phase of CMD_CS as a SDL diagram. Four procedures run in parallel.¶

$scale 0.8 start :idleCount = 0;\nbusyCount = 0;\nrssiState = INVALID;] repeat :RSSI Sample available< note right See section 'Timing' below. end note :rssi = readRssi();] if (rssi < cmd.rssiThr) then (yes) :idleCount++;\nbusyCount = 0;] if (idleCount >= cmd.numRssiIdle) then (yes) :rssiState = IDLE;] :updateChannelState(rssiState);| :RSSI state changed> endif else (no) :busyCount++;\nidleCount = 0;] if (busyCount >= cmd.numRssiBusy) then (yes) :rssiState = BUSY;] :updateChannelState(rssiState);| :RSSI state changed> endif endif repeat while () legend right Command parameters: - cmd.numRssiBusy - cmd.numRssiIdle - cmd.rssiThr endlegend$

Figure 30. The watchRssi() function from Figure 29.. For each change of rssiState, the resulting state is updated ccording to Table 14..¶

$scale 0.8 skinparam ranksep 120 state "Corr INVALID" as invalid state "Corr IDLE" as idle state "Corr BUSY" as busy [*] -> invalid invalid --> idle : \n\n[no tops within cmd.CorrPeriod] idle --> invalid : [tops >= cmd.corrConfig.numCorrInv within corrPeriod]\n && [cmd.corrConfig.numCorrBusy > 0] invalid --> busy : \n[tops >= cmd.corrConfig.numCorrBusy\n within cmd.corrPeriod] idle --> busy : [tops >= cmd.corrConfig.numCorrInv within corrPeriod]\n && [cmd.corrConfig.numCorrBusy == 0] busy --> idle : [no tops within cmd.corrPeriod]$

Figure 31. The watchCorr() function from Figure 29. as a state chart. For each state change, the correlation state is updated according to Table 14..¶

Both procedures, watchRssi() and watchCorr() result in an internal channel state that is one of INVALID, IDLE or BUSY. Depending on cmd.bEnaRssi and cmd.bEnaRssi only one or both functions are active. If only function is active, the resulting channel state is equal to the internal channel state. When both functions are active, the cmd.csConf.operation flag specifies, how the results from these functions are combined:

0: The channel is BUSY if either watchRssi() OR watchCorr() indicate BUSY.
1: The channel is BUSY if both watchRssi() AND watchCorr() indicate BUSY.

A complete mapping can be found in Table 14.. The resulting channel state is then fed into watchChannelState() (Figure 32.) and waitForCsEndTrigger() (Figure 33.).

Table 14. Truth table that combines the output of `watchRssi()` and `watchCorr()` depending on `cmd.csConf.operation`. The resulting channel state is used as input for `watchChannelState()` in Figure 32. and `waitForCsEndTrigger()` in Figure 33..¶
RSSI OR Correlation `csConf.operation = 0`				RSSI AND Correlation `csConf.operation = 1`
	Correlation state
RSSI State	INVALID	IDLE	BUSY	INVALID	IDLE	BUSY
INVALID	INVALID	INVALID	BUSY	INVALID	IDLE	INVALID
IDLE	INVALID	IDLE	BUSY	IDLE	IDLE	IDLE
BUSY	BUSY	BUSY	BUSY	INVALID	IDLE	BUSY

$scale 0.8 start :finished = false;] repeat :channelState changed< :channelState = lookup(rssiState, corrState);| note right See lookup table above endnote if (channelState) then (Idle) if (cmd.csConf.idleOp) then (1) :finished = true;\nresult = FALSE;\ncmd.status = PROP_DONE_IDLE;] else (0) endif elseif (channelState) then (Busy) if (cmd.csConf.busyOp) then (1) :finished = true;\nresult = TRUE;\ncmd.status = PROP_DONE_BUSY;] else (0) endif else (Invalid) endif repeat while (finished?) is (false) stop legend right Command parameters: - cmd.csConf.busyOp - cmd.csConf.idleOp endlegend$

Figure 32. The watchChannelState() function from Figure 29..¶

$scale 0.8 start :csEndTrigger fired< if (channelState) then (Idle) :cmd.status = PROP_DONE_IDLE;\nresult = FALSE;] elseif (channelState) then (Busy) :cmd.status = PROP_DONE_BUSY;\nresult = TRUE;] else (Invalid) if (cmd.csConf.timeOutRes) then (1) :cmd.status = PROP_DONE_IDLETIMEOUT;\nresult = FALSE;] else (0) :cmd.status = PROP_DONE_BUSYTIMEOUT;\nresult = TRUE;] endif endif stop legend right Command parameters: - cmd.csConf.timeOutRes endlegend$

Figure 33. The waitForCsEndTrigger() function from Figure 29. in an SDL diagram.¶

Timing¶

The CMD_PROP_CS provides two timing values:

cmd.csEndTime is the minimum execution time in RAT ticks, after which channel state evaluation is forced. Optional.
cmd.corrPeriod specifies the number of RAT ticks between each correlation evaluation.

How to chose a minimum value for cmd.csEndTime so that at least either cmd.numRssiIdle or cmd.numRssiBusy RSSI samples are measured before a final decision is forced? The following drawing illustrates the CS command execution phase.

../../_images/aafig-baabab8541a06985c3fc7c39aaa5973303f55b3e.png

After the obligatory parsing delay of approximatly 50 µs, command execution starts and the end trigger for cmd.csEndTime is armed. After a fixed time $Rssi_0$ , the first RSSI sample is available and the RSSI available signal in Figure 30. is fired for the first time. Successive RSSI samples occur with a period $Rssi_n$ . The execution may either end up due to channel state evaluation in Figure 32. or a timeout of cmd.csEndTime, but has an additional evaluation delay of 5 µs.

Please note, that the RSSI sample period might be affected by the Automatic Gain Correction (AGC) and the AGC is always enabled in RX mode. In case of a very strong signal, the AGC might increase the first sample period $Rssi_1$ up to 4 times. If the signal strength is changing very fast, the AGC re-adjusts every RSSI sample period. However, in reality it is safe to assume that only the first period $Rssi_1$ is delayed by the AGC. The value $Rssi_0$ in contrast, is not affected, but the value might be wrong if the signal strength is above -50 dBm. For CSMA applications, this is expected to be uncritical.

A minimum value for cmd.csEndTime can be estimated with formula (1).

(1)¶ $csEndTime_{\text{min}}^{\text{rssi}} = \begin{cases} Rssi_0 + Rssi_n \cdot cmd.numRssiIdle + \SI{5}{\us} \\ Rssi_0 + Rssi_n \cdot cmd.numRssiBusy + \SI{5}{\us} \end{cases}$

Table 15. provides measurement results for $Rssi_0$ and $Rssi_n$ . Both depend on the configured RX bandwidth (cmd.rxBw), specified either in CMD_PROP_RADIO_SETUP or CMD_PROP_RADIO_DIV_SETUP.

Table 15. Timing `watchRssi()` and `watchCorr()`¶
RX bandwidth	Value for `rxBw`		First RSSI sample	RSSI sample period
[kHz]	CC13xx	CC26xx	$Rssi_0$ [µs]	$Rssi_n$ [µs]
45/55/66	32/33/34	n/a	348	136
88/110/130	35/36/37	1/2/3	214	68
180/220/260	38/39/40	4/5/6	149	34
350/430/530	41/42/43	7/8/9	114	17
700/870/1060	44/45/46	10/11/12	98	10
1410/1740/2120	47/48/49	13/14/15	89	10