Physical channel processing is where a transport block stops being bits and becomes a signal on an antenna. 36.211 draws the whole of it as six boxes in a row, and the picture looks calmer than the process is. Two of those boxes behave completely differently depending on how many antennas the cell transmits from.

That is the reason this page splits into cases. The chain never changes and the boxes never move. What changes is how much work the layer mapper and the precoder do when the data reaches them.

• Physical Channel Processing
• SISO
• Tx Diversity
• 2 x 2 MIMO
• The Three Cases Side by Side
• Reference

Physical Channel Processing

This is the process that convert the bit stream coming out of channel encoding process into radio frame data to be transmitted by each antenna.

Overall flow is as shown below and I put the 3GPP specification for each steps. This process is also the one that would be very difficult to understand in very details unless you follow each steps of the procedure written in the specification. But this process is described in a very short mathematical format and it is hard to describe in other way without losing the clarification.

In this section, I would like you to grasp the basic idea of each step.

At lower part of the diagram, I labeled the data format of each step. If you follow those labels, you would know at which steps bit stream is being handled and at which steps I/Q data is handled etc.

36.211 Figure 6.3-1. The whole chain in six boxes, with the clause number for each one written above it and the data format written below it.

• The clause numbers along the top read 36.211 6.3.1 for Scrambling and 6.3.2 for the Modulation mapper. They continue with 6.3.3 for the Layer mapper, 6.3.4 for Precoding and 6.3.5 for the Resource element mapper.
• The last box, OFDM signal generation, carries no clause number. It sits at 36.211 6.12 rather than inside the 6.3 run, so the labelling stops one box early.
• Three vocabulary labels sit above the chain: code words on the left, layers in the middle and antenna ports on the right. Each names what travels at that point, and the three counts need not agree.
• The labels underneath are the data format. Bits at the input and after scrambling, then I/Q after the modulation mapper, after the layer mapper and after precoding.
• The two pictures below the chain are what the resource element mapper produces. The left one is a grid of resource elements. The right one is a subframe running from RB=0 to RB=99, with DC in the middle and Slot #0 and Slot #1 inside 1 ms.

Each box has one job, and naming them in order is enough to see the shape of the chain. Scrambling multiplies the coded bits by a sequence tied to the cell and the UE, so two neighbouring cells do not interfere in a correlated way. The modulation mapper turns groups of bits into complex symbols. The layer mapper decides how those symbols are shared among layers, and precoding decides how the layers reach the antenna ports. The resource element mapper places the result on the grid, and OFDM signal generation turns the grid into a waveform.

Two of those six carry all the variation. Scrambling, modulation, resource element mapping and signal generation work the same way whatever the antenna configuration is. The layer mapper and the precoder are the pair that 36.211 breaks into sub-clauses, and the sections below follow those sub-clauses one case at a time.

Three counts matter here, and they are not the same count. A code word is a transport block after channel coding. A layer is a stream inside the processing chain. An antenna port is something the receiver can estimate a channel for. The cases below differ only in what those three numbers are.

One good way for you to study this procedure would be to classify the procedure into several specific cases like SISO, Diversity, 2x2 MIMO etc and mark the path for each case as shown below.

• SISO
• Tx Diversity
• 2x2 MIMO

Each of those cases redraws the same overview with its own path marked, so the three pictures below can be read against each other. The points to carry into them are these.

• The chain is fixed and the blocks are not : every case walks the same six boxes, and only the layer mapper and the precoder change behaviour.
• Bits become symbols at the modulation mapper : everything before it is bits and everything after it is I/Q, which the labels under Figure 6.3-1 mark.
• Code words, layers and antenna ports are three counts : the drawing labels all three, and reading them as one number is the usual mistake.
• The sub-clause numbers are the map : 6.3.3.x picks the layer mapping and 6.3.4.x picks the precoding, and each case below names its own pair.

SISO

As shown below, marked in red/blue arrow, only one codeword (transport block) comes into the process and directly goes to the antenna. In this case, layer mapper and Precoding steps do almost nothing.

SISO. One red path along the top of the chain, and two sub-clause labels that make the middle of it trivial.

• The red line runs the full width of the drawing on the top row only. The lower row of blocks is drawn in grey and carries nothing.
• Inside the layer mapper and the precoder the line becomes a blue dotted arrow. It stays one straight line through both, so nothing is split and nothing is combined.
• The labels underneath name the pair. 6.3.3.1 is the layer mapping and 6.3.4.1 is the precoding, and 36.211 titles both of them for transmission on a single antenna port.
• One code word enters, one layer runs through the middle and one antenna port leaves. All three counts are 1, which happens in no other case on this page.

The two blocks do almost nothing, and the specification says exactly how little. 36.211 6.3.3.1 maps the single code word onto one layer, so the layer count equals the code word count and no reordering happens. 6.3.4.1 then sets each antenna port symbol equal to the layer symbol. Both are identity operations, and the specification still writes them out as equations.

The two blocks stay in the chain because the chain is one chain. A receiver on a single antenna port runs the same six steps as a receiver on four ports, and only the equations inside two of them differ. That is what lets a cell change transmission mode without touching anything else in the chain.

• Nothing here is identity by accident : 6.3.3.1 and 6.3.4.1 are real clauses, and each defines the trivial mapping explicitly rather than leaving it implied.
• All three counts are one : one code word, one layer and one antenna port, which is the definition of this case rather than a consequence of it.
• The lower path is drawn but idle : the second chain of blocks appears in grey, so the same drawing serves the cases that do use it.

Tx Diversity

As shown below, marked in red/blue arrow, only one codeword (one transport blocks) comes into the process and split into two stream (layers) by the layer mapper and finally go out through the two antenna.

Tx Diversity. One code word arrives and two antenna ports leave, and the drawing shows exactly where the count changes.

• The top red line runs the whole width. The lower red line starts only after the layer mapper box, and that is where one code word becomes two layers.
• The blue dotted arrows cover both rows through the layer mapper and the precoder, and inside the precoder they cross rather than run straight.
• The labels are 6.3.3.3 and 6.3.4.3. 36.211 titles both of them for transmit diversity, and neither is shared with any other case here.
• Two antenna ports carry the transmission and there is still one code word. A layer count of 2 is not a rank of 2.

NOTE : The two layers here follow from the two antenna ports, not from transmit diversity itself. Only one codeword is involved, and 36.211 6.3.3.3 fixes the layer count: “the number of layers is equal to the number of antenna ports used for transmission of the physical channel.” Table 6.3.3.3-1 then allows two rows only, 2 layers with 1 codeword and 4 layers with 1 codeword. So a four port cell running transmit diversity has four layers, and the drawing above is the two port case.

The picture makes a distinction worth keeping. Two layers here do not mean two data streams. The same code word is spread across both, so the receiver combines the two paths to recover one transport block instead of separating two.

That is why 36.211 gives transmit diversity a precoding clause of its own rather than reusing the spatial multiplexing one. 6.3.4.3 builds the antenna port symbols from the layer symbols and their complex conjugates, and that construction is what lets a receiver recover the code word from either path.

The gain is reliability rather than throughput. Two paths to the same receiver rarely fade at the same moment, so the combined signal is steadier than either one alone. Nothing in this case raises the bit rate.

• Two layers, one code word : the split happens at the layer mapper, and it does not create a second transport block.
• The split point is visible in the drawing : the second red line begins at the layer mapper output rather than at the input.
• Diversity buys reliability, not rate : both antenna ports carry the same information, so the throughput is what one port would have given.
• 6.3.4.3 is a separate clause for a reason : transmit diversity precoding uses complex conjugates, and the spatial multiplexing clause has no use for them.

2 x 2 MIMO

As shown below, marked in red/blue arrow, two codewords (two transport blocks) comes into the process and goes through the layer mapping without any modification and finally go out through the two antenna. In this case, data goes through very complicated precoding process. As described in the specification, there are three sub steps of procoding process as specified in 36.211 6.3.4.2.1, 36.211 6.3.4.2.2, 36.211 6.3.4.2.3. Depending on whether the 2 x 2 MIMO is Closed Loop Mode or Open Loop Mode, the different combination of the substeps are applied.

2 x 2 MIMO. Both rows are red from the far left, and that is the difference from the two cases above.

• Both red lines run the full width, starting at the code word inputs. Two transport blocks enter here, and this is the only case on the page where they do.
• Through the layer mapper the two blue dotted arrows stay parallel and never cross. One code word becomes one layer, which is what the paragraph above calls layer mapping without any modification.
• Inside the precoder the dotted arrows cross. Each antenna port then carries a combination of both layers rather than one of them.
• Four labels sit underneath rather than two: 6.3.3.2 for the layer mapping, then 6.3.4.2.1, 6.3.4.2.2 and 6.3.4.2.3 for the precoding.

The three precoding sub-clauses are not three alternatives to pick one from. 36.211 6.3.4.2.1 is precoding without CDD, 6.3.4.2.2 is precoding for large delay CDD, and 6.3.4.2.3 is the codebook that both of them draw their matrices from. Closed loop uses the first with the third, and open loop uses the second with the third.

The codebook matters more than its position in the list suggests. The UE reports a precoding matrix indicator, and that indicator is an index into 6.3.4.2.3 rather than a matrix. Both ends hold the same table, so a few bits on PUCCH pick the matrix for both ends.

The picture settles where the streams stay separate. The two code words never merge into one. They are mixed at the antenna ports and separated again by the receiver, and two transport blocks come out at the far end.

• Two code words is what makes this multiplexing : the drawing is red on both rows from the far left, and neither of the other two cases is.
• The layer mapper is the easy block here : two code words and two layers map one to one, which is why the dotted arrows stay parallel.
• Precoding carries the complexity : three sub-clauses rather than one, and the dotted arrows cross inside the box.
• Open and closed loop differ by sub-clause : closed loop uses 6.3.4.2.1 with the codebook, and open loop uses 6.3.4.2.2 with the same codebook.

The Three Cases Side by Side

The three drawings above share a chain and differ in five places. Putting those five in one table makes the pattern easier to hold than three separate pictures do, and it also shows which column is the one that really moves.

The columns that separate the three cases. Everything outside this table is identical in all three.

Read the code word column downward and the argument of the page appears. It goes 1, 1, 2, while the antenna port column goes 1, 2, 2. Transmit diversity is the case where those two differ, and that difference is what makes it diversity rather than multiplexing.

The two clause columns are the mechanism behind the first three. A case is nothing more than one sub-clause chosen under 6.3.3 and one chosen under 6.3.4. The blocks around them are unchanged by that choice, which is why the same chain can be redrawn for every case on this page.

• One row is not like the others : Tx Diversity has more antenna ports than code words, and that single mismatch is the whole definition of it.
• A case is a pair of sub-clauses : pick one under 6.3.3 and one under 6.3.4, and the rest of the chain follows without further configuration.
• The page stops at two antenna ports : 36.211 also defines transmit diversity on four ports and layer mappings up to eight layers, and none of those are drawn here.

Reference

• TS 36.211 v19.3.0 (Release 19) - E-UTRA Physical channels and modulation. Clause 6.3 General structure for downlink physical channels, which holds Figure 6.3-1 and every 6.3.3 and 6.3.4 sub-clause this page names.