System Information (MIB / SIB1 / SIBs) in 5G NR

1. Why System Information Works the Way It Does

Every other configuration mechanism in NR is dedicated: the network knows which UE it is talking to, sends an RRCReconfiguration on a signalling bearer, and gets an acknowledgement. System Information has none of that. It has to reach a UE that has just powered on, has no C-RNTI, no security context and no uplink timing, and it has to do so without knowing that the UE exists. That single constraint explains every design decision in this document:

• It is broadcast, not signalled. There is no acknowledgement, so there is no retransmission and no HARQ. SI is sent on DL-SCH with no HARQ feedback loop at all; robustness comes from repetition inside a window and from conservative link adaptation, not from ARQ.
• Its schedule is derived, not configured. The position of an SI message's window follows from its index in schedulingInfoList and from si-WindowLength -- there is no per-message offset field in Rel-15. Both sides compute the same answer from the same two numbers (§5).
• It is split into a bootstrap and a body. The MIB is 24 bits, fits in a fixed-size transport block on PBCH, and exists purely to make SIB1 findable. Everything else waits for SIB1.
• Changes are announced in advance and applied on a boundary. A UE acting on a parameter that changed underneath it would be worse than a UE acting on a parameter it knows to be one modification period stale (§9).
• Rarely-needed parts are optional to transmit. Broadcasting SIB12 for sidelink in a cell where no UE uses sidelink is pure wasted capacity, so si-BroadcastStatus lets the network hold it back until asked (§6).

A UE reads System Information at power-on, but that is the least of the occasions. It also re-reads after every cell reselection into a cell whose stored SI it does not hold, on return from out-of-coverage, when the network signals a change, when a stored SIB passes its three-hour validity limit, and when an application needs a SIB the UE never bothered to acquire before. §11 covers all of them.

2. The Acquisition Chain: PSS/SSS to Other SI

Table 1. The four stages. Note that SIB1 and Other SI use the same RNTI and differ only in search space and in how their occasions are computed TS 38.213 cl. 10.1.

2.1 Why MIB gets no window and SIB1 does

MIB rides BCH, a transport channel with a fixed transport block size of 24 bits, a fixed coding chain, and no PDCCH. It is not scheduled -- it is simply part of the SSB, which means the UE that just found the SSB has already received it. There is nothing to search for and nothing to miss beyond the SSB itself.

SIB1 and every other SIB ride DL-SCH, which means a PDCCH grant, a modulation and coding scheme, a transport block size, and therefore a search space in which to look for the grant and a time window in which the grant may appear. That machinery is what §4, §5 and §9 are about, and the whole of it exists because SI content outgrew 24 bits.

3. The SIB Catalogue: What Each One Carries

Table 2. The SIB catalogue TS 38.331 cl. 5.2.1 and 6.3.1. Positioning SIBs (posSibType1-1 onward) are scheduled separately through posSI-SchedulingInfo and are not part of this list. A decoder built against an earlier release reports an unknown SIB type rather than failing, so always check which release your analyser was built for before concluding a SIB is malformed.

4. si-SchedulingInfo: How Other SI Is Advertised

si-SchedulingInfo sits inside SIB1 and is the whole schedule for Other SI. It has exactly four things in it that matter: an ordered list of SI messages, a single window length shared by all of them, an optional request configuration, and an optional area identifier.

schedulingInfoList

si-WindowLength

si-RequestConfig

si-RequestConfigSUL

systemInformationAreaID

Table 3. SI-SchedulingInfo fields TS 38.331. Note that si-WindowLength is a single cell-wide value in slots, not milliseconds -- its duration in time therefore halves each time the subcarrier spacing doubles.

4.1 Per-message fields

si-BroadcastStatus

ENUMERATED {broadcasting, notBroadcasting}

si-Periodicity

sib-MappingInfo

SIB-TypeInfo.type

ENUMERATED {sibType2, sibType3, ...}

SIB-TypeInfo.valueTag

SIB-TypeInfo.areaScope

Table 4. SchedulingInfo and SIB-TypeInfo fields. Bundling is the single most consequential design choice here: it trades BCCH capacity against acquisition latency for the SIBs in the bundle.

5. SI Windows: The Arithmetic

This is the part that is most often misdescribed, so here it is exactly. There is no configurable window offset in Rel-15. The SI-windows are consecutive, non-overlapping, all the same length, and allocated to the entries of schedulingInfoList in list order. A UE computes the position of the window for SI message n like this TS 38.331 cl. 5.2.2.3.2:

x = (n - 1) x w

  the SI-window starts at slot #a of the radio frame satisfying

        a = x mod N
        SFN mod T = FLOOR( x / N )

  n   order of the entry in schedulingInfoList, counting from 1
  w   si-WindowLength, in slots
  N   number of slots in a radio frame  (10 x 2^mu)
  T   si-Periodicity of *this* SI message, in radio frames

  mu = 0 (15 kHz)  -> N = 10      mu = 2 ( 60 kHz) -> N = 40
  mu = 1 (30 kHz)  -> N = 20      mu = 3 (120 kHz) -> N = 80

Within the window the UE monitors the Type0A-PDCCH common search space for a DCI with CRC scrambled by the SI-RNTI, and it may assume the SI message is transmitted at least once somewhere inside the window. The network is free to place it anywhere in the window and to repeat it, which is how it gets robustness without HARQ. Slots used for SSB and for SIB1 transmission are not available for Other SI.

5.1 Worked example: locating one SI message

That gives the worst case directly. A UE that decides it needs SIB4 at t = 95 ms into the cycle has missed this cycle's window, which closed at 60 ms, and must wait for the next one at 160 + 40 = 200 ms:

5.2 Worked example: how many SI messages fit

Because the windows are consecutive and derived, there is a hard limit on how many SI messages a cell can schedule at a given window length and periodicity. It is not stated as a rule anywhere obvious -- it falls out of the arithmetic:

5.3 Filling the windows efficiently

One more consequence of derived windows: at 30 kHz with si-WindowLength = s10, two windows fit in every radio frame, so an eight-message list occupies four radio frames out of however many the periodicity spans. Everything else is free BCCH capacity. Reordering schedulingInfoList therefore changes when every message after the reordered entry is transmitted -- which is why a list reorder must be treated as an SI change and get a valueTag bump like any other (§9).

6. Broadcasting, notBroadcasting and On-Demand SI

si-BroadcastStatus is one bit per SI message with a large consequence. broadcasting means the message is on air in its window right now, and the UE should simply go and read it. notBroadcasting means the network has the message configured but is not transmitting it, and a UE that needs it must ask.

Table 5. Five situations, three of which produce no network-side evidence whatsoever. This is the single most important thing to internalise about debugging SI problems.

7. Requesting SI: the MSG1 and MSG3 Paths

7.1 The MSG1 path

si-RequestConfig carves out dedicated PRACH resources for SI requests. Each entry of si-RequestResources corresponds, in order, to one SI message with si-BroadcastStatus = notBroadcasting, and gives a ra-PreambleStartIndex. The gNB therefore learns which SI message is wanted from the preamble index alone -- no uplink data is needed at all, and the acknowledgement is a RAR containing a RAPID subheader with no MAC RAR payload behind it TS 38.321 cl. 6.1.5. The companion 03 Random Access document covers the payload-free RAR case.

rach-OccasionsSI.rach-ConfigSI

RACH-ConfigGeneric

rach-OccasionsSI.ssb-perRACH-Occasion

ENUMERATED {oneEighth ... sixteen}

si-RequestPeriod

si-RequestResources[i].ra-PreambleStartIndex

INTEGER (0..63)

si-RequestResources[i].ra-AssociationPeriodIndex

si-RequestResources[i].ra-ssb-OccasionMaskIndex

Table 6. SI-RequestConfig TS 38.331. The preamble indices here come out of the same 64-preamble space as everything else in the cell.

7.2 The MSG3 path

With si-RequestConfig absent, the UE performs ordinary contention-based Random Access and puts RRCSystemInfoRequest on CCCH in MSG3. The message is a bitmap:

RRCSystemInfoRequest-IEs ::= SEQUENCE {
    requested-SI-List   BIT STRING (SIZE (maxSI-Message)),   -- 32 bits
    spare               BIT STRING (SIZE (12))
}

  bit 1 (leftmost) -> the FIRST SI message in schedulingInfoList whose
                      si-BroadcastStatus is notBroadcasting
  bit 2            -> the SECOND such SI message
  ...

  Note: the bit index counts only notBroadcasting messages, NOT the raw
  schedulingInfoList index. A cell where messages #2 and #5 are
  notBroadcasting maps bit 1 -> message #2 and bit 2 -> message #5.

MSG4 in this path carries a UE Contention Resolution Identity MAC CE and no RRCSetup. The UE stays in RRC_IDLE throughout -- it has used the Random Access procedure purely as a one-bit-per-SI-message uplink signalling channel, and returns to camping when it is done.

7.3 Dedicated SI delivery in RRC_CONNECTED

Neither path is available to a UE that is already in RRC_CONNECTED, which was a real gap in Rel-15: a connected UE that needed a notBroadcasting SIB had no way to ask. Rel-16 added DedicatedSIBRequest on SRB1, carrying requestedSIB-List and requestedPosSIB-List; the network then delivers the SIB inside a DLInformationTransfer as dedicated signalling rather than putting it on air. The delivered content is identical, but the stored-SI bookkeeping is not -- see §8.

8. SI Validity, valueTag and Area Scope

A UE stores acquired SIBs so it does not have to re-read them on every reselection. Three independent conditions decide whether a stored copy is still usable TS 38.331 cl. 5.2.2.2.1.

Table 7. Stored-SI validity. All three must hold. The valueTag test is the one the network controls, and therefore the one that goes wrong.

8.1 Area scope, and what it buys

areaScope in SIB-TypeInfo says "this SIB's content is the same across a whole SI area". systemInformationAreaID in si-SchedulingInfo names the area. A UE that acquires an area-scoped SIB in one cell may reuse it in every other cell advertising the same area ID and the same valueTag, which for a UE moving through a dense cluster of cells eliminates almost all of its BCCH reading.

Table 8. Area scoping outcomes. The last row is the one to watch for: everything works, just with several times the BCCH load and several times the reselection latency it should have.

9. SI Change Notification and the BCCH Modification Period

System Information changes at modification period boundaries and nowhere else. The boundary is defined in SFN terms: the modification period is m radio frames long, where

m = modificationPeriodCoeff x defaultPagingCycle      [radio frames]

  modificationPeriodCoeff   ENUMERATED {n2, n4, n8, n16}
                            in SIB1 -> servingCellConfigCommon
                                    -> downlinkConfigCommon
                                    -> bcch-Config
  defaultPagingCycle        ENUMERATED {rf32, rf64, rf128, rf256}
                            in SIB1 -> servingCellConfigCommon
                                    -> downlinkConfigCommon
                                    -> pcch-Config

  boundaries occur at every SFN for which   SFN mod m = 0

9.1 The notification itself

During the modification period preceding the change, the network includes a Short Message with the systemInfoModification bit set in DCI format 1_0 with CRC scrambled by the P-RNTI, at every paging occasion TS 38.331 cl. 6.5. There is no paging message and no PDSCH involved -- the notification is carried in the DCI itself, which is why it costs almost nothing and why an idle UE monitoring its paging occasions receives it without waking any further.

systemInfoModification

etwsAndCmasIndication

Table 9. Short Message bits TS 38.331 cl. 6.5. Bits 1 and 2 differ fundamentally: bit 1 is advance notice of a scheduled change, bit 2 is an instruction to act now.

10. Emergency SI: ETWS and CMAS

SIB6, SIB7 and SIB8 are exempt from almost everything in §8 and §9, and for an obvious reason: an earthquake warning that arrives at the next 5.12-second boundary is a warning that arrives too late.

valueTag

valueTag

Table 10. Emergency SI versus ordinary SI. The messageIdentifier / serialNumber pair is the ETWS/CMAS equivalent of a valueTag, and it lives in the SIB rather than in SIB1's mapping info.

11. Re-Acquisition Outside Initial Access

Power-on is the least interesting SI acquisition. These are the others, roughly in order of how often they happen:

etwsAndCmasIndication

Table 11. Eight re-acquisition occasions. Only the third and fourth are network-driven; the rest are the UE deciding on its own, which is why BCCH load does not correlate with anything the gNB does.

12. Parameter Reference

Almost every parameter here trades broadcast capacity against acquisition latency, and si-Periodicity trades it most directly:

si-WindowLength

si-Periodicity

si-BroadcastStatus

systemInformationAreaID

valueTag

areaScope

modificationPeriodCoeff

defaultPagingCycle

si-RequestPeriod

ssb-periodicityServingCell

Table 12. SI-related parameters and the direction each one trades in. Almost every row trades BCCH capacity against acquisition latency; there is no setting that improves both.

13. Failure Modes and What Each One Means

Table 13. Fourteen SI failure modes. Note the second column: in eleven of the fourteen, nobody detects the fault. System Information has no acknowledgement, so it has no error reporting either.

14. ASN.1: si-SchedulingInfo and the Request Messages

-- TS 38.331, abridged; ... marks omitted members and extension markers

SI-SchedulingInfo ::= SEQUENCE {
    schedulingInfoList        SEQUENCE (SIZE (1..maxSI-Message)) OF
                                  SchedulingInfo,
    si-WindowLength           ENUMERATED {s5, s10, s20, s40, s80, s160,
                                          s320, s640, s1280},
    si-RequestConfig          SI-RequestConfig      OPTIONAL, -- Cond MSG-1
    si-RequestConfigSUL       SI-RequestConfig      OPTIONAL, -- Cond SUL-MSG-1
    systemInformationAreaID   BIT STRING (SIZE (24)) OPTIONAL,
    ...
}

SchedulingInfo ::= SEQUENCE {
    si-BroadcastStatus        ENUMERATED {broadcasting, notBroadcasting},
    si-Periodicity            ENUMERATED {rf8, rf16, rf32, rf64, rf128,
                                          rf256, rf512},
    sib-MappingInfo           SIB-Mapping
}

SIB-Mapping ::= SEQUENCE (SIZE (1..maxSIB)) OF SIB-TypeInfo

SIB-TypeInfo ::= SEQUENCE {
    type                      ENUMERATED {sibType2, sibType3, sibType4,
                                          sibType5, sibType6, sibType7,
                                          sibType8, sibType9, ...,
                                          sibType10-v1610, sibType11-v1610,
                                          sibType12-v1610, sibType13-v1610,
                                          sibType14-v1610, ...},
    valueTag                  INTEGER (0..31)        OPTIONAL, -- Cond SIB-Type
    areaScope                 ENUMERATED {true}      OPTIONAL
}

SI-RequestConfig ::= SEQUENCE {
    rach-OccasionsSI SEQUENCE {
        rach-ConfigSI         RACH-ConfigGeneric,
        ssb-perRACH-Occasion  ENUMERATED {oneEighth, oneFourth, oneHalf,
                                          one, two, four, eight, sixteen}
    }                                                OPTIONAL,
    si-RequestPeriod          ENUMERATED {one, two, four, six, eight,
                                          ten, twelve, sixteen} OPTIONAL,
    si-RequestResources       SEQUENCE (SIZE (1..maxSI-Message)) OF
                                  SI-RequestResources
}

SI-RequestResources ::= SEQUENCE {
    ra-PreambleStartIndex     INTEGER (0..63),
    ra-AssociationPeriodIndex INTEGER (0..15)         OPTIONAL,
    ra-ssb-OccasionMaskIndex  INTEGER (0..15)         OPTIONAL
}

Listing 1. SI-SchedulingInfo and its children. maxSI-Message = 32 and maxSIB = 32; the 32 in maxSI-Message is what sizes requested-SI-List (§7.2).

-- the two SI-related messages, TS 38.331

UL-CCCH-MessageType ::= CHOICE {
    c1 CHOICE {
        rrcSetupRequest             RRCSetupRequest,
        rrcResumeRequest            RRCResumeRequest,
        rrcReestablishmentRequest   RRCReestablishmentRequest,
        rrcSystemInfoRequest        RRCSystemInfoRequest
    },
    messageClassExtension SEQUENCE {}
}

RRCSystemInfoRequest ::= SEQUENCE {
    criticalExtensions CHOICE {
        rrcSystemInfoRequest        RRCSystemInfoRequest-IEs,
        criticalExtensionsFuture-r16 CHOICE {
            rrcPosSystemInfoRequest-r16 RRCPosSystemInfoRequest-r16-IEs,
            criticalExtensionsFuture    SEQUENCE {}
        }
    }
}

RRCSystemInfoRequest-IEs ::= SEQUENCE {
    requested-SI-List           BIT STRING (SIZE (maxSI-Message)),
    spare                       BIT STRING (SIZE (12))
}

-- Rel-16, for a UE already in RRC_CONNECTED (UL-DCCH, SRB1)
DedicatedSIBRequest-r16-IEs ::= SEQUENCE {
    requestedSIB-List-r16       SEQUENCE (SIZE (1..maxOtherSIB-r16)) OF
                                    SIB-ReqInfo-r16   OPTIONAL,
    requestedPosSIB-List-r16    SEQUENCE (SIZE (1..maxAssistDataSIB-Element-r16))
                                    OF PosSIB-ReqInfo-r16 OPTIONAL,
    lateNonCriticalExtension    OCTET STRING            OPTIONAL,
    nonCriticalExtension        SEQUENCE {}             OPTIONAL
}

-- and the SI container itself, on BCCH-DL-SCH
SystemInformation-IEs ::= SEQUENCE {
    sib-TypeAndInfo SEQUENCE (SIZE (1..maxSIB)) OF CHOICE {
        sib2  SIB2,   sib3  SIB3,   sib4  SIB4,   sib5  SIB5,
        sib6  SIB6,   sib7  SIB7,   sib8  SIB8,   sib9  SIB9,
        ...
    },
    lateNonCriticalExtension    OCTET STRING            OPTIONAL,
    nonCriticalExtension        SystemInformation-v1700-IEs OPTIONAL
}

Listing 2. The request messages and the SI container. Note that SystemInformation-IEs carries a list of SIBs -- this is the bundling of §4.1 made concrete.

15. Illustrative Message Traces

One cell runs through all five traces: PCI 217, PLMN 310-410, 30 kHz subcarrier spacing, si-WindowLength = s40, four SI messages, SI area 0x0A1B2C.

15.1 The schedule, as broadcast in SIB1

SIB1
 si-SchedulingInfo
  si-WindowLength .................. s40         -- 40 slots = 20 ms @30 kHz
  systemInformationAreaID .......... 0x0A1B2C
  si-RequestConfig ................. (absent)    -- so MSG3 path (S7.2)
  schedulingInfoList
   [1] si-BroadcastStatus .......... broadcasting
       si-Periodicity .............. rf16        -- 160 ms
       sib-MappingInfo
        {type sibType2, valueTag 3, areaScope true}
        {type sibType3, valueTag 1, areaScope true}
   [2] si-BroadcastStatus .......... broadcasting
       si-Periodicity .............. rf32        -- 320 ms
       sib-MappingInfo
        {type sibType4, valueTag 7, areaScope true}
        {type sibType5, valueTag 2, areaScope true}
   [3] si-BroadcastStatus .......... notBroadcasting
       si-Periodicity .............. rf16
       sib-MappingInfo
        {type sibType12-v1610}                   -- NR sidelink, on demand
   [4] si-BroadcastStatus .......... broadcasting
       si-Periodicity .............. rf64        -- 640 ms
       sib-MappingInfo
        {type sibType9, areaScope (absent)}      -- no valueTag: time info

 servingCellConfigCommon
  downlinkConfigCommon
   bcch-Config
    modificationPeriodCoeff ........ n4
   pcch-Config
    defaultPagingCycle ............. rf128
                                                 -- m = 4 x 128 = 512 rf
                                                 --   = 5.12 s

  -- Derived window positions, N = 20 slots per frame, w = 40:
  --   n=1  x=  0  a= 0  FLOOR(  0/20)=0  -> slot  0, SFN mod 16 = 0
  --   n=2  x= 40  a= 0  FLOOR( 40/20)=2  -> slot  0, SFN mod 32 = 2
  --   n=3  x= 80  a= 0  FLOOR( 80/20)=4  -> slot  0, SFN mod 16 = 4
  --   n=4  x=120  a= 0  FLOOR(120/20)=6  -> slot  0, SFN mod 64 = 6
  -- Total window span 4 x 40 = 160 slots = 8 frames, comfortably inside
  -- the shortest periodicity (rf16 = 320 slots). Configuration is sound.

Listing 3. si-SchedulingInfo with the window arithmetic done alongside. Doing this calculation once per cell configuration catches the §5.2 failure mode before any UE ever sees it.

15.2 Acquiring an always-broadcast SI message

14:22:08.140  [RRC] SIB1 acquired, PCI 217, SFN 3
              -> need SIB4 (inter-frequency neighbours) for reselection
              -> SIB4 is in SI message #2
              -> window: slot 0, frames where SFN mod 32 = 2
              -> next such frame: SFN 34

14:22:08.450  [PHY] SFN 34 slot 0 -- SI-window for message #2 opens
              monitoring Type0A-PDCCH CSS, SI-RNTI 0xFFFF
14:22:08.452  [PHY] SFN 34 slot 4  PDCCH DCI 1_0, CRC = SI-RNTI
              -> PDSCH  24 PRB, MCS 5, TBS 704 bits
14:22:08.453  [RRC-DL-BCCH] SystemInformation
               sib-TypeAndInfo
                [1] sib4
                     interFreqCarrierFreqList
                      [1] dl-CarrierFreq ......... 632628
                          subcarrierSpacing ...... kHz30
                          q-RxLevMin ............. -62      -- = -124 dBm
                          cellReselectionPriority  5
                          threshX-HighP .......... 10       -- = 20 dB
                          threshX-LowP ........... 4        -- = 8 dB
                [2] sib5
                     carrierFreqListEUTRA
                      [1] carrierFreq ............ 1850
                          cellReselectionPriority  3
14:22:08.453  [RRC] stored: SIB4 valueTag 7 areaScope 0x0A1B2C
                            SIB5 valueTag 2 areaScope 0x0A1B2C
              -> SI-window closed early: message received, stop monitoring

  -- Elapsed from 'need SIB4' to 'have SIB4': 313 ms, almost all of it
  -- waiting for SFN 34. Nothing was retransmitted and nothing failed.
  -- SIB5 arrived free, because it is bundled in the same SI message.

Listing 4. A completely normal acquisition. The 313 ms is window wait, not decode time -- which is why tuning MCS or bandwidth does nothing for SI latency.

15.3 On-demand acquisition via MSG3

14:31:02.006  [APP] sidelink service started -> RRC needs SIB12
14:31:02.006  [RRC] SIB12 is in SI message #3
              si-BroadcastStatus = notBroadcasting
              si-RequestConfig absent  -> use the MSG3 request path
              SI messages with notBroadcasting, in list order:
                 message #3   -> requested-SI-List bit 1
              -> set bit 1

14:31:02.010  [MAC-UL] MSG1  ra-PreambleIndex 19 (ordinary CBRA pool)
                       PRACH SFN 84 slot 19 sym 0, RA-RNTI 1121
14:31:02.017  [MAC-DL] MSG2  RAR: RAPID 19, TA cmd 118, TC-RNTI 0x39C4
14:31:02.022  [MAC-UL] MSG3  CCCH, 6 octets, TC-RNTI 0x39C4
              UL-CCCH-Message
               c1: rrcSystemInfoRequest
                criticalExtensions: rrcSystemInfoRequest
                 requested-SI-List .... 1000 0000 0000 0000
                                        0000 0000 0000 0000
                 spare ................ 0000 0000 0000
14:31:02.031  [MAC-DL] MSG4  PDSCH to TC-RNTI 0x39C4
              LCID 62  UE Contention Resolution Identity, 48 bits
              -- no RRCSetup: the UE is not being connected
14:31:02.031  [RRC] SI request acknowledged; remain in RRC_IDLE
                    discard TC-RNTI 0x39C4

14:31:05.120  [RRC] SIB1 re-read at the modification period boundary
              schedulingInfoList[3] si-BroadcastStatus .. broadcasting
14:31:05.760  [RRC-DL-BCCH] SystemInformation  sib12-v1610  (SFN 4 slot 0)

  -- 3.1 s from request to delivery, of which 3.09 s is the wait for the
  -- next BCCH modification period boundary. The RACH exchange cost 25 ms.
  -- On-demand SI trades a long, quantised delay for BCCH capacity.

Listing 5. The MSG3 request path end to end. Note that the dominant cost is not the request but the modification period boundary at which the network is allowed to start broadcasting.

15.4 SI change notification and re-acquisition

15:04:41.200  [PHY] paging occasion, SFN 620 -- DCI 1_0, CRC = P-RNTI 0xFFFE
              Short Messages Indicator ....... 10   -- short message only
              Short Messages ................. 1000 0000
                bit 1 systemInfoModification . 1
                bit 2 etwsAndCmasIndication .. 0
                bits 3-8 ..................... 0
15:04:41.200  [RRC] SI change notified. Current modification period is
                    SFN 512..1023 (m = 512). Change applies at SFN 0.
              -> continue using stored SI until the boundary
              -> do NOT re-acquire now

15:04:45.240  [PHY] SFN 0 -- modification period boundary
15:04:45.243  [RRC] SIB1 re-acquired
              schedulingInfoList[1] sib-MappingInfo
                {type sibType2, valueTag 4, areaScope true}  -- was 3
                {type sibType3, valueTag 1, areaScope true}  -- unchanged
              schedulingInfoList[2] sib-MappingInfo
                {type sibType4, valueTag 7, areaScope true}  -- unchanged
                {type sibType5, valueTag 2, areaScope true}  -- unchanged
15:04:45.243  [RRC] SIB2 valueTag 3 -> 4 : stored copy invalid
                    SIB3, SIB4, SIB5      : stored copies still valid
              -> acquire SI message #1 only
15:04:45.250  [RRC-DL-BCCH] SystemInformation  sib2 + sib3  (SFN 0 slot 0)
              -- SIB3 arrives again anyway: it is bundled with SIB2
              sib2  cellReselectionServingFreqInfo
                     q-Hyst ................. dB4     -- was dB2

  -- Only one of four SIBs actually changed, and only one of four SI
  -- messages was re-read. That selectivity is the entire purpose of
  -- per-SIB valueTags.

Listing 6. A well-behaved change: one valueTag moves, one SI message is re-read. Compare the failure modes in §13 where the tag does not move, or moves for everything.

15.5 Failure path: a window that can never occur

-- Configuration pushed to the cell:
   si-WindowLength ................. s20
   subcarrierSpacing ............... kHz15    -> N = 10 slots per frame
   schedulingInfoList: 5 entries, all si-Periodicity = rf8  -> T = 8

-- gNB configuration validation: PASSED (no rule checks this)
-- SIB1 on air advertises all five SI messages as 'broadcasting'

16:12:33.010  [RRC] need SIB12, mapped in SI message #5
              x = (5 - 1) x 20 = 80
              a = 80 mod 10 = 0
              FLOOR(80 / 10) = 8
              -> window condition: slot 0 of frames where SFN mod 8 = 8
              -> SFN mod 8 is in {0..7}; the condition is unsatisfiable
16:12:33.010  [RRC] no SI-window computed for SI message #5
              -- and si-BroadcastStatus = broadcasting, so no request is
              -- sent either: the UE believes the SIB is already on air

-- Observable evidence, network side:
     SI-request preambles for message #5 ......... 0
     RRCSystemInfoRequest with that bit set ...... 0
     BCCH transport blocks for message #5 ........ 0
     error counters, alarms, KPIs ................ none

-- The only trace of the fault anywhere is the arithmetic above, done
-- by hand from SIB1. 5 x 20 = 100 slots of window span against a
-- periodicity of 8 x 10 = 80 slots.

Listing 7. The §5.2 failure mode as it actually presents: a SIB that no UE ever receives, with nothing anywhere reporting a problem. Compare the sound configuration in §15.1, where the span is 160 slots against 320.

16. Release Deltas: Rel-15 to Rel-18

Table 14. SI-related changes by release. The Rel-17 si-WindowPosition addition is the one that changes how you read a trace, because it invalidates an assumption that held for two releases.

17. Reading System Information in Logs: A Checklist

1. Do the window arithmetic before anything else. For every entry in schedulingInfoList, compute x, a and FLOOR(x/N), and check that FLOOR(x/N) is strictly less than that message's si-Periodicity in radio frames. One unsatisfiable condition means one SIB that no UE will ever receive, with no error reported (§5.2).

2. Check the total window span against the shortest periodicity. Number of SI messages x si-WindowLength must fit inside the shortest si-Periodicity expressed in slots. This is the same check as step 1, expressed as one multiplication.

3. Confirm the subcarrier spacing before converting anything to milliseconds. si-WindowLength is in slots. s40 is 20 ms at 30 kHz and 40 ms at 15 kHz. Half the arithmetic errors in SI analysis are this.

4. Match every `areaScope` to a `systemInformationAreaID`. An area-scoped SIB in a cell with no area ID is re-acquired in every cell, which is the opposite of what was intended.

5. Compare `valueTag` values across the cells of an SI area. Same area ID and same tag must mean identical content. A partial parameter push is invisible in any single cell's capture (§8.1).

6. Read `si-BroadcastStatus` for every message, and then ask whether the matching request path exists. notBroadcasting with si-RequestConfig present means MSG1 requests; absent means MSG3. If si-RequestConfig is present, verify its preamble indices sit outside totalNumberOfRA-Preambles and outside the CFRA range.

7. Look for RACH procedures that end at MSG4 with no `RRCSetup`. Those are RRCSystemInfoRequest procedures, not failed connection attempts, and counting them as access failures will distort every accessibility KPI you have.

8. Look for preambles answered with a payload-free RAR. Those are MSG1 SI requests. A payload-free RAR answering a handover preamble is a preamble range overlap (§7.1).

9. Compute the modification period and compare it to your change window. m = modificationPeriodCoeff x defaultPagingCycle radio frames. A parameter push cannot reach camped UEs faster than that. If m reaches or exceeds 1024 the boundary occurs only at SFN 0, so the effective period saturates at 10.24 s no matter what the product says (§9).

10. Distinguish `systemInfoModification` from `etwsAndCmasIndication` in any captured Short Message. Bit 1 is advance notice; bit 2 is act now. Bit 2 set with no SIB6/7/8 on air is a live power-drain fault (§10).

11. For a connected UE behaving on old parameters, check whether a dedicated SI change indication was ever sent. Connected UEs do not monitor paging and therefore never see the Short Message.

12. When nothing explains the behaviour, capture from the UE side. SI has no acknowledgement and therefore no failure reporting; eleven of the fourteen failure modes in §13 are detected by nobody at all.

18. Glossary

si-WindowLength

si-Periodicity

valueTag

19. References

• 3GPP TS 38.331 -- RRC protocol specification. Clause 5.2.1 (introduction and SIB catalogue), 5.2.2.1 (SI acquisition and the cell-barred-on-SIB1-failure rule), 5.2.2.2.1 (SI validity), 5.2.2.2.2 (SI modification and the BCCH modification period), 5.2.2.3.1 (SIB1 acquisition), 5.2.2.3.2 (SI message acquisition and the window formula), 5.2.2.3.3 (SI request), 5.2.2.3.5 (actions on receiving SI), 6.2.2 (message definitions: SystemInformation, RRCSystemInfoRequest, DedicatedSIBRequest), 6.3.1 (SIB definitions), 6.4 (SI-SchedulingInfo, SchedulingInfo, SIB-TypeInfo, SI-RequestConfig), 6.5 (Short Message).
• 3GPP TS 38.300 -- NR overall description. Clause 7.3 (system information delivery, RMSI and OSI terminology), 9.2.1 (cell selection and the SI a UE needs first).
• 3GPP TS 38.213 -- Physical layer procedures for control. Clause 10.1 (search space sets, including Type0-PDCCH and Type0A-PDCCH common search spaces and the SI-RNTI), clause 13 (CORESET#0 and searchSpaceZero for SIB1).
• 3GPP TS 38.212 -- Multiplexing and channel coding. Clause 7.1 (PBCH payload generation), 7.3.1.2.1 (DCI format 1_0 including the Short Messages Indicator and Short Messages fields).
• 3GPP TS 38.321 -- NR MAC protocol specification. Clause 5.1 (random access, including the SI-request case), 6.1.5 (RAR MAC PDU and the payload-free RAPID subheader).
• 3GPP TS 38.304 -- UE procedures in idle mode and RRC inactive state. Clause 5.2.4 (reselection, and the defaults applied when SIB2 to SIB5 are absent).
• 3GPP TS 23.041 -- Technical realization of Cell Broadcast Service, for the ETWS and CMAS message structures that SIB6, SIB7 and SIB8 carry.

Companion documents in this set

• 18 MIB and SIB1 IEs -- field-by-field on the two messages this document treats as a starting point, including the CORESET#0 and searchSpaceZero lookups behind §2.
• 02 Radio Frame Structure -- SFN, radio frames, slots per frame, and the numerology that turns si-WindowLength in slots into a duration.
• 03 Random Access -- the RACH procedure both SI request paths use, including the preamble partitioning of §7.1 and the payload-free RAR.
• 06 Channel Mapping -- BCCH, BCH and DL-SCH, the mapping §2 depends on.
• 19 Paging -- paging occasions, defaultPagingCycle and the DCI that carries the Short Message of §9.
• 20 Measurements and Events -- the reselection parameters SIB2 to SIB5 actually carry.
• 01 Registration Process -- where SI acquisition sits in the power-on latency budget, and what the UE does with SIB1 once it has it.
• 13 RRC Overview / 14 RRC States -- the state model that decides which SI request path is even available.