Cell selection — The S criterion — the inequality a device must satisfy to camp
The device — identity, state, mobility
Where it sits
What it is
What it is. The test a cell must pass before a device will camp on it. TS 38.304 §5.2.3.2 calls it the cell selection criterion S, and it is one of the few places in this folder where a device's behaviour is fixed by an explicit inequality rather than by a procedure.
The criterion is fulfilled when both quantities are positive:
with the cell suitable only when Srxlev > 0 and Squal > 0.
Where each term comes from:
| Term | Is | From |
|---|---|---|
Qrxlevmeas |
measured RSRP — received power | the device, measuring the SSB |
Qqualmeas |
measured RSRQ — received quality | the device |
Qrxlevmin |
minimum required receive level, dBm | broadcast in SIB1/SIB2/SIB4 |
Qqualmin |
minimum required quality, dB | broadcast |
Qrxlevminoffset |
an offset applied only while searching for a higher-priority PLMN from a visited one | broadcast, per TS 23.122 |
Qoffsettemp |
a temporary penalty on a cell the device has had trouble with | TS 38.331 |
Pcompensation |
how much the device's own transmit power falls short | the device itself |
Pcompensation is the term worth stopping on, because it reaches into the RF front end. TS 38.304
§5.2.3.2 defines it, for FR1, as max(PEMAX1 − PPowerClass, 0) — where PEMAX1 is the maximum uplink
power the cell permits and PPowerClass is "the maximum RF output power of the UE (dBm) according
to the UE power class as defined in TS 38.101-1".
Read what that does. If the cell expects devices to be able to transmit at, say, 26 dBm and this
device's power amplifier can only manage 23, then Pcompensation is 3 dB, and the device raises its
own bar for this cell by 3 dB. It refuses to camp somewhere it could hear perfectly well — because
it knows the cell would not be able to hear it.
This is a downlink measurement being corrected by an uplink hardware limit. The criterion is
asymmetric on purpose: a device that camped on the strongest cell it could receive would sit in cells
it cannot reach, and would discover this only at
random access. Pcompensation moves that discovery to before camping, and it
does it with a number that comes from the power class — a
front-end property. It is one of the cleanest seams in this folder between layer 1 and layer 2.
And it vanishes in FR2: "For FR2, Pcompensation is set to 0." Millimetre-wave links are beamformed at both ends, so a single scalar power class no longer captures the link budget, and the compensation would be meaningless.
Two quantities, not one, and that is deliberate. Srxlev asks is the signal strong enough;
Squal asks is it clean enough. A cell at the centre of a dense deployment can be very strong and
very interfered — high RSRP, poor RSRQ. Requiring both to pass rejects cells that would give a good
bar count and bad service.
After selection comes camping, and then cell reselection takes over as the device moves.
Read on
This concept is read as part of one argument in ref-ue-mobility, alongside the rest of its group.
Before this concept, the hierarchy says to learn the following — the full chain, in order: