Recently (as of Sep 2015) you might have seen two obvious directions of LTE evolution. One direction is 'higher and higher throughput' based on Carrier Aggregation (2 CC, 3CC and even 4CC) and the other direction is 'lower and lower throughput' with 'Low Cost' and 'Lower Energy Consumption'. The second direction (Lower throughput with Low Cost/Energy) is called 'MTC'. MTC has a certain set of criteria to meet (Refer to MTC page for these criteria and technical issues). Probably the first category that might be classified as an MTC device would be Category 1 device. But this Category 1 is just desinged as low performance LTE and not so much optimized to be called as 'efficient MTC' device. Probably the first LTE device that is specifically designed for MTC would be Category 0 device. (Refer to Category 0 page to see why Category 0 is so special). Category M (or LTE-M) is the device that is even more optimized towards even lower power consumption, even simpler complexity leading to even lower device cost.

What dose 'M' stands for ? Not sure... probably 'MTC' or 'Minus (lower than Cat 0 -:))

• How did LTE get a low-end branch?
• What changes in the radio at 1.4 MHz?
• What changes again at 200 kHz?
• LTE-M1 and M2
• Reference

How did LTE get a low-end branch?

LTE started at the top of the market and worked downward. Every release from Rel-11 onward added something for devices that need less, and the two tables below track that from different angles. One lists what each release contributed. The other puts the resulting device profiles side by side, so you can see what was given up to reach each one.

Following table (Reference 1 ) would show you the general trands of the evolution of LTE-M devices .

Three releases of low-end work, from Reference 1. Rel-11 only tuned how the network treats a device. Rel-12 created a device category for the purpose. Rel-13 is where coverage enhancement joins the list, and that is the release which produced the categories this page is about.

• Rel-11 (2012) : UE power preference indication and RAN overload control. Neither one changes the device, only how the network treats it.
• Rel-12 (2014) : the first low-cost UE category, which is Cat-0, plus power saving mode and UE assistance information for eNB parameter tuning.
• Rel-13 (expected 2016) : another low-cost UE category, coverage enhancement and power saving enhancement. The year is written as a guess because the table was drawn before Rel-13 froze.

Following table( Reference 2 ) would show you some of key technical charateristics of each LTE based MTC devices in comparison to normal LTE device (Cat 4).

Five device profiles from Reference 2, with modem complexity on the bottom row as the number the whole exercise aims at. Cat-0 keeps a 20 MHz receiver and cuts only the data rate, while both Rel-13 columns cut the receive bandwidth itself. That is the step which forces the radio changes in the next section.

• The top two rows are peak rate : 150 Mbps down for Cat 4 and 10 Mbps for Cat 1. Cat-0 and the 1.4 MHz column both sit at 1 Mbps, and the 200 kHz column at 200 kbps.
• Receive bandwidth is the row that matters : Cat 4, Cat 1 and Cat-0 all keep 20 MHz. Only the two Rel-13 columns drop it, to 1.4 MHz and to 200 kHz.
• Antennas and duplex go earlier : the second receive antenna disappears at Cat-0, and so does full duplex. Both stay gone in the Rel-13 columns.
• Transmit power does not fall steadily : the 1.4 MHz column drops to 20 dBm while the 200 kHz column stays at 23 dBm. Less bandwidth does not automatically mean a smaller amplifier.
• Modem complexity is the target : 100% for Cat 4, then 80%, 40%, 20% and under 15%. Each column buys that reduction by giving up a different thing.
• The two right-hand headers sit in quotation marks : "Cat. 1.4MHz" and "Cat. 200kHz" were working labels in 2015 rather than category names. The last section gives the names 3GPP settled on.

• Cost is the design target, not throughput : the bottom row of the second table is the one every other row serves. A lower peak rate is a means to it rather than a goal.
• Cat-0 and Cat-M save in different ways : Cat-0 keeps the wideband receiver and cuts the rate. Cat-M cuts the receiver itself, which is why it needs new physical channels.
• Both tables predate Release 13 : read their right-hand columns as intent rather than as final numbers, and check the last section on this page for what was settled.

What changes in the radio at 1.4 MHz?

Cutting the receive bandwidth to 1.4 MHz sounds like a small step, and it is the one that breaks the most. Six resource blocks is all such a UE can see, so it cannot read channels spread across the whole carrier. Several Release 8 channels therefore stop being usable for it, and everything in the list below follows from that.

In terms of eNB lower layer point of view, there wouldn't be any modification required until Cat 1 and Cat 0. (In case of Cat 0, there will be some modification required at higher layer protocol like PSM, but it is nothing to do with physical layer in eNB). Most of the modification would be on UE side (overall strategy of those modification is described in MTC page).

However, there should be several modifications on both side (UE and eNB) for Cat 1.4 Mhz or Cat 200 KHz. Some of these modifications are as follows :

• Cat-M UEs only monitor 6-RB bandwidth signal even when the system BW support wider BW for normal device.
• The control region signals (= PDCCH, PHICH, PCFICH) shall not be used for Cat-M UEs : Why ? All of these channels are spreaded across the whole system bandwidth (Refer to PDCCH, PHICH, PCFICH page and check resource element locations for these channels). But Cat M UE can monitor only 6 RB. It means these UE cannot decode every resource elements for these channel. So these channel is not useful for Cat M UE.
• DCI is carried by MPDCCH (this is almost the same signal as EPDCCH) : Why ? Since PDCCH, PCFICH cannot be decoded by Cat M UE, we need another mechanism to send control information to UE. It is to use the area that is normally used for PDSCH. It is similar motivation for EPDCCH.
• DCI(MPDCCH) and DL-SCH(PDSCH) can be transmitted in different subframes : First, Cat-M UE use only 6 RB, there wouldn't be much space in one subframe if some of PDSCH area is used for MPDCCH. Also, by allocating separate subframe after MDCCH, the UE can have enough time to decode DCI to prepare for the reception of following PDSCH. This is called “cross-subframe scheduling”. However, it is still possible to allocate MPDCCH and PDSCH in same subframe. In this case, two RBs are allocated for MPDCCH and the remaining 4 RBs are allocated for PDSCH.
• Almost all the channel data (PRACH, PBCH, PUSCH PUCCH, MPDCCH and PDSCH shall be transmitted with repetition )shall be transmitted with repetition : Why ? It is assumed that Hardware performance of Cat-M device is not as good as normal device (like Smartphone) because Cat-M device is supposed to be designed for low cost. This kind of repetition would help even such a low cost device to decode the channels and even in harsh channel condition.

• Six resource blocks explain the whole list : the control region goes, MPDCCH replaces PDCCH, and cross-subframe scheduling appears. None of those is an independent choice.
• This is where the eNB has to change too : Cat 1 and Cat-0 need nothing new in the eNB physical layer. From 1.4 MHz onward both ends change.
• Repetition pays for the narrow receiver : a narrow receiver lowers cost and loses sensitivity at the same time. Repeating every channel recovers what it lost.

What changes again at 200 kHz?

One resource block is where the LTE frame structure stops being reusable. At 1.4 MHz a Cat-M device still lives inside an ordinary LTE carrier and borrows its numerology. At 200 kHz that is no longer a given, and the study had to decide where the carrier sits before it could decide what the carrier looks like.

In case of 200 Khz (NB Cat M), there should be even more modification in terms of protocol because it is only 1RB and 1 RB System BW has never been tried before in LTE.  The overall feature of NB IoT is now outlined in 3GPP RP-151621 as follows.

3 Mode of Operation

The three options below differ in where the 200 kHz carrier is placed, not in what it carries. That placement decides what an operator has to find: new spectrum, spare guard band, or nothing beyond a spare resource block inside a carrier already running.

1.‘Stand-alone operation’ utilizing for example the spectrum currently being used by GERAN systems as a replacement of one or more GSM carriers

2.‘Guard band operation’ utilizing the unused resource blocks within a LTE carrier’s guard-band

3.‘In-band operation’ utilizing resource blocks within a normal LTE carrier

Physical Waveform

At this width the waveform itself was still open when the study was written. The list below therefore records options under consideration rather than a settled design, and the GMSK uplink candidate is not the one that reached the specification.

• 180 kHz UE RF bandwidth for both downlink and uplink
• According to TR 45.820, there can be muliple types of physical waveforms and they are labelled a little bit differently depending on their implementation (e.g, EC-GSM, NB-MTM,NB-LTE).
• NB-LTE waveform as an example is defined as
• OFDMA on the downlink
• Two numerology options will be considered for inclusion: 15 kHz sub-carrier spacing (with normal or extended CP) and 3.75 kHz sub-carrier spacing.
• For the uplink, two options will be considered: FDMA with GMSK modulation (as described in 3GPP TR 45.820 section 7.3), and SC-FDMA (including single-tone transmission as a special case of SC-FDMA)

• Placement stayed an operator decision : stand-alone, guard band and in-band all survived into the final design, so one deployment can differ from another at the spectrum level.
• The 3.75 kHz option is the real break from LTE : every other LTE carrier uses 15 kHz. A second numerology is what makes this a different radio rather than a narrow one.
• Read this section as a snapshot : it lists candidates from late 2015, not conclusions. The next section says which of them became the specification.

LTE-M1 and M2

Now we have other names which seems to be coined just to confuse us :). Recently (around Feb 2016), I start hearing other names called LTE-M1 and LTE-M2.  I haven't seen these names in 3GPP, but LTE-M1 seems to indicate LTE 1.4Mhz and LTE-M2 seems to indicate LTE 200 Khz. I think the 3GPP terms for LTE-M2 is NB-LTE according to TR 45.820 ( Ref [4] )

That guess was half right, and the half that was wrong is worth knowing, because both names are still used loosely today.

The 1.4 MHz device did become Category M1. 3GPP froze it in Release 13, and a UE reports it in ue-CategoryDL and ue-CategoryUL, the same Release 12 fields Cat-0 uses. Release 14 then added Category M2. That one is a wider eMTC device rather than a narrower one, so M2 is not the 200 kHz category.

The 200 kHz device became NB-IoT instead, and it has a capability field of its own. 36.306 defines NB-IoT as access to network services via E-UTRA with a channel bandwidth limited to 200 kHz, corresponding to one PRB. Its categories are NB1 from Release 13 and NB2 from Release 14, and a UE reports them in ue-Category-NB. So the two branches are not two values of one field. They are two different fields.

The name NB-LTE came from the TR 45.820 study, where it was one of several candidate technologies rather than a decision. What went into Release 13 was named NB-IoT, and that is the name the specifications use.

The four categories sit side by side in the table below. Every number in it is from 36.306, and the columns are the ones that decide what the modem has to be able to hold.

The alternatives in that table are not free choices. In the downlink M1 reaches 1736 bits only when the UE indicates ce-PDSCH-MaxTBS-r17, and NB2 reaches 4968 only with npdsch-16QAM-r17. In the uplink M1 reaches 2984 only with ce-PUSCH-NB-MaxTBS-r14. Without those capability fields the UE is held to the smaller number in each pair.

Two more rules tie the pairs together. 36.331 requires a UE indicating category m2 to also indicate category m1, and 36.306 requires a UE indicating Category NB2 to also indicate Category NB1. A log showing M2 alone, or NB2 alone, is therefore incomplete rather than unusual.

• M1 is the 1.4 MHz device, so that half of the guess was right : Category M1 is Release 13 eMTC, reported in ue-CategoryDL and ue-CategoryUL.
• M2 is not the 200 kHz device : Category M2 is the Release 14 wider eMTC category. The 200 kHz branch became NB-IoT, with categories NB1 and NB2.
• The two branches use different capability fields : ue-CategoryDL and ue-CategoryUL carry M1 and M2, while ue-Category-NB carries NB1 and NB2. A decoder reading only the first pair sees nothing at all for an NB-IoT device.
• An "or" in the table depends on a capability : check for ce-PDSCH-MaxTBS-r17, ce-PUSCH-NB-MaxTBS-r14 or npdsch-16QAM-r17 before reading the larger transport block size.

Reference

[1] Recent Advancements in M2M Communications in 4G Networks and Evolution Towards 5G

Rapeepat Ratasuk‡, Athul Prasad†, Zexian Li†, Amitava Ghosh‡, and Mikko A. Uusitalo†

‡Nokia Networks, Arlington Heights, IL, USA.

†Nokia Technologies, 00045 Nokia Group, P.O. Box 226, Espoo, Finland.

[2] (Nokia Whitepaper) LTE-M – Optimizing LTE for the Internet of Things

[3] 3GPP RP-151621

[4] 3GPP TR 45.820 V13.1.0 (2015-11) : Cellular system support for ultra-low complexity and low throughput Internet of Things (CIoT) (Release 13)