Following illustration is the 5G/NR channel mapping from MAC layer through PHY layer based on 38.321 and 38.211. In this page, I would try to give you only the big picture of the whole channel structure in NR, I would not get into the details of each type of channel processing separately. It is too big topics to be described in single page. I would write separate pages for channel mapping and channel processing for each different channels.

• How to read the big picture
• Every physical channel and signal in the diagram
• The dotted lines, one by one
• Channel Mapping at MAC Layer
• The three 38.321 tables, written out
• Following one message all the way down

If you give it just a brief glance, you would think it is same as LTE. In terms of MAC layer mapping (MAC to Transport mapping), I also would say it is same as LTE. However, if you take a little bit of closer look at PHY layer, you would notice some difference from LTE PHY channel and PHY signal. Followings are the list of difference of NR PHY channel and LTE PHY channel.

• NR does not use (has no) CRS (Cell Specific Reference Signal)
• NR PDSCH requires DMRS whereas LTE PDSCH does not use DMRS. This is understandable because NR PDSCH would require its own reference signal (DMRS) since there is no CRS.

NOTE : one small refinement on the second bullet. LTE PDSCH does have a UE specific DMRS, but only in the later transmission modes (TM7 to TM10, antenna ports 7 and up). In the ordinary modes that most LTE deployments used, PDSCH really was demodulated from CRS. The difference in NR is that DMRS is not one option among several - it is the only way, always present, for every physical channel that carries information.

Those two bullets are really one decision and its consequence, and it is worth saying out loud because almost every other difference on this page follows from it. LTE sprayed CRS across every resource block of every subframe, whether anybody was being scheduled or not. It was always there, so LTE could use it for everything - demodulation, channel measurement, even just finding the cell. NR deleted it.

The reason is that an always-on signal is an always-on cost. It burns power at the base station and it radiates interference into every neighbour cell around the clock, even at three in the morning with nobody attached. NR is built to go quiet when there is nothing to send, and CRS made that impossible. So NR replaced one always-on signal with several on-demand ones, each sent only when its particular job needs doing :

How to read the big picture

There is a lot in the illustration shown above, so here is the order I would read it in.

1. The three bands on the right. MAC at the top, Transport in the middle, PHY at the bottom. Everything above the first horizontal line is logical channels and MAC internals ; everything between the two lines is transport channels ; everything below is physical. A channel changes its name every time it crosses one of those boundaries, and that renaming is the single thing this diagram exists to explain.

2. The colours. Red is downlink and blue is uplink, and the arrows at the very bottom repeat it - red pointing down, blue pointing up. Scan that bottom row alone and you can see the whole duplex split at a glance : PSS, SSS, PBCH, CSI-RS, PDCCH, PDSCH are all red ; SRS, PUSCH, PUCCH, PRACH are all blue.

3. What is inside the MAC box. The four blocks are not decoration - they are the actual work MAC does between a logical channel and a transport channel :

4. What is not connected to anything above it. This is the detail most people scroll past, and it is the most informative thing in the picture. Look at PSS, SSS, CSI-RS and SRS. Every other item at the bottom has a line running up into the transport layer. These four have nothing.

That is not an omission in the drawing. Those four carry no information from any higher layer at all - they are generated in the physical layer, from the cell identity or from a configured sequence, and their entire purpose is to be measured. There is no data in them to map. PSS and SSS let you find the cell, CSI-RS lets the network learn the downlink channel, SRS lets it learn the uplink channel. A signal, not a channel.

Inside the SSB group this contrast is very sharp : PSS and SSS float free, but PBCH right next to them has a line going up to BCH, because PBCH does carry something from above - the MIB.

Every physical channel and signal in the diagram

Reading the bottom row of the illustration across, box by box :

The PT-RS column is worth staring at. It is ticked for exactly two channels - PDSCH and PUSCH - and for nothing else. That is not arbitrary. PT-RS tracks phase noise, which becomes a problem at high carrier frequencies and at high modulation orders, where the constellation points sit close together and a slow rotation of the whole constellation destroys them. Those two conditions only ever apply to the data channels : PDCCH and PUCCH use robust low-order modulation by design, so a little phase drift does them no harm and a PT-RS would be pure overhead. This is a genuinely new signal in NR with no LTE ancestor, and the diagram shows you its scope precisely.

NOTE : As mentioned above, the dotted line does not indicate any direct mapping between channels. It indicate 'there is some relationship between the two channels connected by the dotted line, but it is not a direct physical mapping. Followings are more detailed description of each dotted lines.

• PSS, SSS, PBCH : As in LTE, PSS,SSS,PBCH are used and PSS/SSS/PBCH are bundled into a specific area in the downlink resource grid called SSB(SS Block)
• DL-SCH and PDSCH/DMRS : This implies PDSCH/DMRS is required for PHY/MAC scheduling for DL-SCH(PDSCH/PDSCH-DMRS).
• UL-SCH and PUSCH/DMRS : This implies PUSCH/DMRS is required for PHY/MAC scheduling for UL-SCH(PUSCH/PUSCH-DMRS)
• DL-SCH and PUCCH/DMRS : This implies PUCCH/DMRS is required for HARQ Response(ACK/NACK) for DL-SCH(PDSCH/DMRS)

The dotted lines, one by one

The reason the dotted lines have to exist at all is that PDCCH and PUCCH carry no transport channel. Look back at the table above : every solid line in the diagram connects a transport channel to the physical channel that carries it, and PDCCH and PUCCH have nothing in that column. Their contents - DCI and UCI - are created in the physical layer itself, not handed down from MAC.

But they are obviously not unrelated to DL-SCH and UL-SCH, because without them no data would move at all. A dotted line is how the diagram says "these two need each other, but nothing flows from one into the other". Reading them in the order the radio actually uses them :

So the three dotted lines are, in order : here is your data (PDCCH), here is your permission to send (PDCCH), and here is my answer (PUCCH). Every scheduled transmission in NR uses at least two of them.

Channel Mapping at MAC Layer

The illustration shown above may show you a little bit detailed picture of MAC process, but it may not be so clear about the channel mapping unless you follow through each lines very carefully. In terms of channel mapping, the tables in 38.321 would be clearer and simple to understand and my illustration to the right would be even more clear and intuitive :).

The three 38.321 tables, written out

The tables in that picture are small, so here they are again as text you can search and copy.

38.321 Table 4.5.2-1 - what kind of channel each one is

Four control channels and exactly one traffic channel. Everything you actually care about as a user - the web page, the video, the call - travels on that single DTCH row. The other four exist to make it possible.

38.321 Table 4.5.3.1-1 - uplink, logical to transport

Look at the RACH column. It is completely empty, and that is the single most interesting thing in these three tables. RACH is listed as a transport channel and no logical channel maps to it at all.

This is the same fact the first illustration showed in a different way : RACH is fed from the Random Access Control block inside MAC, not from any channel above it. A random access preamble is not data that came down from RRC - it is a signature MAC picks from a configured set and asks PHY to transmit. There is nothing to map, so the column stays blank. Once you have seen it here, the empty column stops looking like a printing error.

38.321 Table 4.5.3.2-1 - downlink, logical to transport

As you see, most of channels from Logical channel to Transport channel is one-to-one or many-to-one relation, but BCCH case it maps to BCH and DL-SCH.

What does this mean ? Does this mean that a BCCH message maps both to BCH and DL-SCH simultaneously ?

No. It means some BCCH data maps to BCH and some BCCH data maps to DL-SCH.  If you are familiar with LTE, you would know there are largely two types of BCCH in LTE. One is MIB and the others are SIBs. MIB goes through BCCH-BCH path and SIBs go through BCCH-DL SCH path. NR would use the same pattern of channel mapping.

It is worth adding why the split exists, because the two paths are not just different pipes - they suit two genuinely different jobs :

That last row is the whole reason for the split. The MIB has to be found by a UE that has just powered on and knows nothing at all, so it cannot be scheduled - being scheduled would require reading a PDCCH, and finding the PDCCH requires information that only the MIB can give. So it gets its own fixed, unscheduled, dedicated channel. Once the MIB is decoded that deadlock is broken, and everything after it can go the normal shared route.

Following one message all the way down

Putting the two illustrations together, here is what the renaming actually looks like for four everyday messages. Every arrow is one boundary crossing in the first diagram.

  MIB                       SIB1                      an RRC Reconfiguration
   |                         |                         |
  BCCH   logical            BCCH   logical            DCCH   logical
   |                         |                         |
  BCH    transport          DL-SCH transport          DL-SCH transport
   |                         |                         |
  PBCH   physical           PDSCH  physical           PDSCH  physical
                             ^                         ^
                    scheduled by a DCI        scheduled by a DCI
                    on PDCCH (SI-RNTI)        on PDCCH (C-RNTI)


  your video stream        a HARQ ACK                a RACH preamble
   |                         |                         |
  DTCH   logical            (no logical channel)      (no logical channel)
   |                         |                         |
  DL-SCH transport          (no transport ch.)        RACH   transport
   |                         |                         |
  PDSCH  physical           PUCCH  physical           PRACH  physical

The bottom row is the interesting one. A HARQ ACK never had a logical channel or a transport channel - it is created in the physical layer, which is exactly why PUCCH reaches DL-SCH only by a dotted line. A RACH preamble skips the logical layer but does have a transport channel, because MAC's Random Access Control genuinely hands something to PHY. And your video is the only one of the six that travels the full ordinary path from a traffic channel all the way down.

Reference

[1]

• 3GPP TS 38.321 - MAC protocol specification (clause 4.5 : channel mapping)
• 3GPP TS 38.211 - Physical channels and modulation
• 3GPP TS 38.300 - Overall description, clause 6 : layer 2 architecture