LTE-M1 System Bandwidth is always 1.4 Mhz, but legacy LTE System BW with which LTE-M1 operate can be wider than 1.4. When LTE-M1 operate in a wider legacy LTE band, it split the wider band into multiples of 1.4 Mhz band slot and use any one of those slots. These 1.4 slot for LTE-M1 is called 'Narrowband'.

• How many Narrowbands are there for each System Bandwidth ?
• Mapping between PRB and Narrowband
• How to figure out which Narrowband to be used ?
• Does the downlink use the same narrowbands ?
• What is a wideband
• Reference

How many Narrowbands are there for each System Bandwidth ?

Two questions have to be answered before a narrowband index means anything. How many narrowbands does a given system bandwidth hold, and which resource blocks does each one contain ? 36.211 5.2.4 answers both with one short definition and two formulas, and the picture below is that clause drawn out.

The way in which narrowbands are devided and indexed is specified in 36.211-5.2.4. Based on 36.211-5.2.4, following table can be created.

The narrowband count and the PRB mapping in one picture. The top table is per system bandwidth. The three cases underneath are the mapping formula, which branches on whether the resource block count is even or odd.

• The count is a floor division : the number of narrowbands is the resource block count divided by six and rounded down, which is the formula at the upper left.
• Six times the narrowband count rarely equals the resource block count : 25 resource blocks give 4 narrowbands and use only 24 of them, and 75 give 12 narrowbands and use 72.
• i0 absorbs the leftovers : it is the offset that pushes the whole block of narrowbands towards the middle of the carrier. The row in the table gives its value for each bandwidth.
• The third branch adds one : when the resource block count is odd, every narrowband in the upper half is shifted up by one resource block, which is what steps over the middle.
• i runs 0 to 5 : that is the position inside a narrowband, and n_NB runs from 0 to one less than the narrowband count.

The third branch is the part worth slowing down on. A carrier with an odd number of resource blocks has one resource block sitting exactly in the middle, and the DC subcarrier sits inside it. Rather than place a narrowband across that block, 36.211 steps over it, so every narrowband above the midpoint moves up by one.

A carrier with an even count has no middle block to avoid, and there the offset alone does the work. The narrowbands sit as a solid group with the spare resource blocks split between the two edges.

Figure 1. The same rule drawn for two bandwidths. At 5 MHz the odd resource block count leaves the centre block unused; at 10 MHz the even count leaves one block spare at each edge. Either way the narrowbands themselves stay six blocks wide and contiguous.

Putting the numbers in one place makes the pattern easier to check against the picture above.

The first three columns are the picture above. The last two are six times the narrowband count, and the difference. The left over column is how many resource blocks a BL UE can never be scheduled on.

• A BL UE cannot use every resource block on the carrier : between one and four blocks fall outside every narrowband, depending on the bandwidth.
• Only 1.4 MHz wastes nothing : six resource blocks make exactly one narrowband, which is the case LTE-M1 was designed around.
• The wasted blocks are not lost to the cell : a legacy LTE UE on the same carrier can still be scheduled on them, so the capacity goes somewhere.
• Narrowbands never straddle the centre : on an odd carrier the middle resource block is stepped over, which keeps every narrowband six contiguous blocks wide.

Mapping between PRB and Narrowband

Based on the formula and table shown above, you can figure out each Narrowband and their index are assigned as shown below. (The PRB marked in grey are the RBs that are not used)

Reading the table is worth a moment, because a blank cell and a zero mean different things. A number in a cell is the index of the narrowband that resource block belongs to at that system bandwidth. A blank means one of two things: either the carrier does not have that many resource blocks, or the block exists and belongs to no narrowband at all.

Three columns show the second case clearly. At 5 MHz the run stops at resource block 11, skips 12, and picks up again at 13, which is the centre block being stepped over. At 10 MHz the numbers start at resource block 1 rather than 0 and stop at 48, leaving one spare block at each edge. At 20 MHz the same thing happens with two blocks at each edge.

One consequence is worth carrying into a live trace. A narrowband index in a DCI or a SIB is not a resource block number, and converting between them needs this table or the formula behind it. A scheduler reporting narrowband 2 at 5 MHz means resource blocks 13 to 18, not 12 to 17.

• The cell holds a narrowband index, not a count : reading down a column shows six consecutive rows carrying the same number, and that run is one narrowband.
• A blank is either absent or unused : above the resource block count the carrier has no such block, and inside the range a blank is a block no narrowband claims.
• The offset differs per bandwidth : narrowband 0 starts at resource block 0 for 1.4 and 5 MHz. It starts at block 1 for 3, 10 and 15 MHz, and at block 2 for 20 MHz.
• Narrowband index and resource block index are not interchangeable : a log that names a narrowband has to be translated before it can be compared with a resource block allocation.

How to figure out which Narrowband to be used ?

How a UE can figure out which Narrowband to be used for transmission or reception ? The answer varies depending on situations. The brief summary of the case and Narrowband determination is as follows.

i) PDSCH for SI (SIBs other than SIB1) : si-Narrowband in SIB1

ii) PDSCH for user data : Resource Block assignment field in DCI

iii) PUSCH : Resource Block assignment field in DCI

Three more cases complete the picture, and each one is answered somewhere different again.

iv) MPDCCH : mpdcch-Narrowband in the EPDCCH-SetConfig, whose value range in 36.331 runs from 1 to maxAvailNarrowBands and maps onto narrowband indices 0 upwards

v) PDSCH carrying SIB1-BR : derived from schedulingInfoSIB1-BR in the MIB, through the tables in 36.213 7.1.6

vi) Paging : the paging narrowband, which 36.300 23.7a pairs with the paging occasion in the time domain

Reading the six together, one pattern holds. The earlier a UE is in the procedure, the more the narrowband is decided by rule rather than by a message. SIB1-BR comes from a MIB field and a fixed table, because there is nothing else to read yet. User data comes from a DCI, because by then a control channel exists to carry it.

That ordering is also why the MIB field matters so much on a BL UE. It is the only place a narrowband can be named before any table has been received, so everything after it depends on that one value.

• There is no single answer, and that is the point : each channel gets its narrowband from whatever the UE can already read at that stage.
• The MIB starts the chain : schedulingInfoSIB1-BR points at a table in 36.213 that fixes the narrowband for SIB1-BR, and no control channel is involved.
• SIB1 hands over to si-Narrowband : once SIB1-BR has been read, the other system information blocks have a field of their own.
• DCI takes over for user data : both PDSCH and PUSCH get their narrowband from the resource block assignment field, which is what the earlier sections of this page explain.
• MPDCCH is configured rather than indicated : mpdcch-Narrowband is part of the set configuration, so it does not change grant by grant.

Does the downlink use the same narrowbands ?

Everything above quotes 36.211 5.2.4, and clause 5 of 36.211 is the uplink chapter. That raises a fair question, because most of what a BL UE does with a narrowband happens in the downlink. The answer is that the downlink has a clause of its own, and it is almost the same clause.

36.211 6.2.7 opens with the identical sentence. A narrowband is six non-overlapping consecutive physical resource blocks in the frequency domain. The count is the downlink resource block count divided by six and rounded down, and the numbering runs upwards in resource block order. For every bandwidth where the uplink and downlink resource block counts match, which is the normal FDD and TDD case, the two clauses give the same narrowbands.

One thing is in 6.2.7 and not in 5.2.4. The downlink mapping carries an extra shift, and 36.211 sets it to zero unless a UE in CE mode B has ce-PDSCH-FlexibleStartPRB-AllocConfig configured. When that parameter is set, the shift comes from 36.211 Table 6.2.7-1 and it is not the same for every narrowband.

36.211 Table 6.2.7-1, the shift applied to downlink narrowbands for PDSCH resource allocation in CE mode B when a flexible starting PRB is configured. Every shift is zero or negative, so the narrowbands move down in frequency rather than up.

The direction of those shifts is the useful detail. Each one is zero or negative, so a shifted narrowband starts on a lower resource block than the unshifted one. At 50 and 100 resource blocks the shift is the same for every narrowband, which simply slides the whole set down by one or two blocks. At 15, 25 and 75 it differs between the lower and upper halves, which is the odd carrier rule reappearing in another form.

• The two clauses define the same thing : six consecutive resource blocks, counted by a floor division, numbered upwards, in both 36.211 5.2.4 and 6.2.7.
• Only the downlink has a shift : the uplink clause has no counterpart to 36.211 Table 6.2.7-1.
• The shift is off by default : 36.211 sets it to zero unless a CE mode B UE has ce-PDSCH-FlexibleStartPRB-AllocConfig configured. The table above therefore describes a configured case rather than the usual one.
• Shifts never move a narrowband upwards : every value in the table is zero or negative.

What is a wideband

Both clauses this page relies on have gained a second half since Release 13, and their titles say so. 36.211 5.2.4 and 6.2.7 are now called narrowbands and widebands. If you only know the narrowband half, you will meet the other one in any CE mode A log from a Release 14 network onwards.

The definition is short. A wideband is four non-overlapping narrowbands in the frequency domain, numbered in order of increasing narrowband number. Four narrowbands is 24 resource blocks. That is the 5 MHz channel a Category M2 UE can be scheduled on, so the wideband is the unit that wider channel is built from.

Small carriers get a special case rather than a fraction. Where there are fewer than four narrowbands to work with, 36.211 says there is a single wideband, and that wideband is composed of whatever non-overlapping narrowbands exist. A 3 MHz carrier therefore has one wideband of two narrowbands rather than half a wideband.

Derived by applying the four narrowbands per wideband rule, and the fewer than four exception, to the narrowband counts earlier on this page. The first three rows are the exception and the last three are the ordinary case.

Whether a UE works in narrowbands or widebands is a configuration rather than a property of the carrier. 36.211 5.2.5 and 6.2.8 both cover this. When ce-pusch-maxBandwidth-config or ce-pdsch-maxBandwidth-config is set to 5 MHz, the retuning rules stop applying between narrowbands and apply between widebands instead. The same guard period rules then measure a different unit.

That is the practical reason to know the term. A Release 13 UE never leaves the narrowband world, and everything earlier on this page describes it completely. A Category M2 UE on a Release 14 network is scheduled in widebands, and the resource block mapping has to be read one level up.

• A wideband is exactly four narrowbands : 24 resource blocks, which is the 5 MHz maximum channel of Category M2.
• Fewer than four narrowbands gives one short wideband : a 3 MHz carrier has a single wideband holding its two narrowbands, rather than a partial one.
• Widebands are numbered the same way : in order of increasing narrowband number, so wideband 0 holds the lowest four narrowbands.
• The unit follows the configuration : a 5 MHz maximum bandwidth switches the retuning rules of 36.211 5.2.5 and 6.2.8 from narrowbands to widebands.
• Nothing above becomes wrong : narrowbands are still six resource blocks and still numbered the same way, and the wideband is a grouping on top of them.

Reference

The clauses the definitions on this page come from, and the versions they were checked against.

[1] 3GPP TS 36.211 V13.2.0 (2016-06)

[2]

[3] 3GPP TS 36.211 v19.3.0 - clause 5.2.4 and clause 6.2.7, narrowbands and widebands, with Table 6.2.7-1

[4] 3GPP TS 36.331 v19.3.0 - the mpdcch-Narrowband and schedulingInfoSIB1-BR fields

[5] 3GPP TS 36.300 v19.2.0 - clause 23.7a, paging narrowbands for BL UEs