5G NR Physical Downlink Shared Channel (PDSCH)

This is a new episode in our series, "5G Explained." In this video, we'll discuss downlink data transmission in 5G New Radio. We will look at the downlink shared channel chain, which includes LDPC coding, the physical downlink shared channel chain, our resource elements are allocated for PDSCH transmission, the different types of PDSCH mapping, and conclude with a quick word on transport block sizes.

The downlink shared channel, or DL-SCH, is the channel that carries user data. It also carries other pieces of information, such as the different types of system information blocks, or SIB. The coding chain includes the usual steps, such as CRC, code block segmentation, rate matching, and concatenation--all steps we're familiar with from LTE.

The main difference with LTE is the use of LDPC coding. The output of a coding chain is a codeword. 5G supports transmission of up to 8 layers to a single user on the downlink. This means that there can be up to eight streams transmitted in parallel. Those streams are coming from one or two codewords--one codeword if there are fewer than four layers, and two codewords if there are more. The coded data is then mapped to the physical downlink shared channel or PDSCH.

Here you can see how every stage of the downlink shared channel processing is mapped to functions in MathWorks 5G Toolbox. You can recognize CRC encoding, code block segmentation, LDPC coding, and rate matching.

The physical downlink shared channel is highly configurable, much more so than in LTE. And we will see some of the detail on the next slide. It is configured both by downlink control information, which can change from slot to slot, and radio resource control, which can set up some parameters as well. There isn't much of a surprise here compared to LTE.

We find scrambling, modulation, layer mapping, precoding for MIMO processing, and resource mapping. While those are all known blocks, there are a few differences worth pointing out. Chiefly, the precoding step is not specified explicitly in the standard, although it is fully expected to be present. The detailed precoding will be addressed in another episode of this video series.

Here we can see code for complete downlink data processing in MathWorks 5G Toolbox. The PDSCH processing stage is highlighted, but you can also see the DL-SCH stage, as well as the insertion of demodulation reference signals, or DM-RS. 5G NR uses the exact same list of modulations on the downlink as LTE, from QPSK through 256QAM. The NR symbol modulate function implements modulation for 5G NR in one simple call.

Layer mapping is the operation that maps one or two codewords to up to eight layers. This operation is somewhat simplified compared to LTE, where you could see one or two codewords for a given number of layers. In 5G NR, as mentioned earlier, anything up to four layers uses a single codeword. Anything beyond four layers uses a second codeword.

The mapping is pretty straightforward--direct for one layer, alternatively for two. Similarly, for three and four layers, each group of three or four input bits is mapped to a set of three or four layers. For five through eight layers, the two codewords are split, as shown here, between the different layers.

After one or two codewords got mapped to between one and eight layers, the layers undergo precoding, which, interestingly, is not specified in the standard for downlink. Precoding is the operation that maps the layers to as many or more antenna boards using a matrix multiplication with a precoder. The special case of precoding is mapping one layer to multiple antennas, which enables beam forming. For line-of-sight transmission, this would likely mean targeting a particular direction.

Another case of precoding is mapping several layers to multiple antennas. This more general case is sometimes referred to as spatial multiplexing. One key aspect of precoding in 5G is that the associated demodulation reference signals, or DM-RS, must undergo the same precoding. As a result, the UE doesn't need to be made aware of the precoder, as the effect of a precoder is included in channel estimation. This is why the exact precoder the G node b is to use is not specified in the standard.

The precoder output is then mapped to physical resource blocks, either directly or indirectly, as we will see on the next two slides. Downlink channels and signals, including the PDSCH and associated DM-RS, share the UI DM grid.

PDSCH symbols are first mapped to virtual resource blocks. When mapped to the grid, PDSCH symbols avoid locations reserved for other purposes. This includes all physical signals, DM-RS, channel state information reference signals, or CSIRS, and phase-tracking reference signals, or PTRS. This also includes any resource block that is fully or even partially used by synchronization signal blocks, or SSB. SSBs are explained in detail in another episode of this video series.

Mapping of virtual to physical resource blocks can be interleaved or non-interleaved. Non-interleaved mapping consists in directly mapping each virtual block to the same position in the physical resource grid. Interleaved mapping provides frequency diversity by distributing the virtual blocks over the whole bandwidth part. The interleaver granularity is two or four resource blocks. This scheme that's assigned consecutive virtual resource blocks to a PDSCH, a pattern that is easy to signal--simply with the starting resource block and number of resource blocks, while still getting frequency diversity.

Here we see two examples of PDSCH resource allocation in time. The PDSCH may span the whole slot, as shown at the bottom of the grid. It may also use part of a slot. This is sometimes referred to as partial slot allocation, and it is a new capability in 5G New Radio, compared to LTE. As you may remember. LTE always allocates a full slot frame of 1 millisecond for PDSCH.

Let us explore some of those allocation options interactively with a user interface that uses MathWorks 5G Toolbox. Here, we are looking at 10 subframes with 30 kilohertz subcarrier spacing, which means a total of 20 slots. The PDSCH is shown in teal or light blue, and this is what I wanted you to look at here. We will focus on the rest of our resource grid in detail in another episode of this series, "5G Explained."

Resource allocation does not have to be contiguous, although it is easier to signal when it is. Let's make it contiguous from 0 to 20. We can see PDSCH transmission in the first 10 slots, followed by five empty slots. This is because we allocated slot 0 through 9 with a periodicity of 15 slots.

Let's change the allocation to something different. Now, slots 6, 7, and 9 do not have PDSCH transmission. Finally, notice that within each slot, the PDSCH only uses symbols 2 through 10. This is called partial slot allocation. You can choose to allocate the full slot, in which case there is no break between PDSCH transmission.

Reference signals for PDSCH are shown in yellow. Those locations are not available for PDSCH mapping. Reference signals will be addressed in detail in another episode of the series "5G Explained."

Here we can see some of the parameters available in MathWorks 5G Toolbox to specify one or multiple PDSCH. We have just seen how these parameters impact the PDSCH chain in the live example.

As we have just seen, a PDSCH slot allocation may start at the beginning of a slot, or midway through a slot. This corresponds to two different mapping types: mappings type A and B. Strictly speaking, the PDSCH mapping type only affects the location of the demodulation reference signals.

For mapping type A, DM-RS assigned symbol 2 or 3 of the slot, while for mapping type B, DM-RS are in the first symbol of the PDSCH allocation. Both mapping types A and B support full and partial slot allocation. In practice, however, mapping type B is the preferred option for partial slot allocation, and especially for transmissions that do not start at the beginning of the slot.

Having DM-RS at the beginning of the allocation enables to reduce processing latency, which is critical for low-latency communications. We will cover DM-RS allocation in more detail in another episode of this "5G Explained" series.

To conclude this section, we want to turn to the receiver and explore how transport block sizes are communicated to the receiver. The receiver needs to be able to figure out the transfer block size in order to perform inverse-rate-matching LDPC decoding. The concept is similar to LTE.

In another episode of this "5G Explained" video series, we will explain how downlink control information is transmitted. But it is important to know that the transfer block size itself is not signaled. Instead, a few different pieces of information are signaled.

They include the modulation encoding scheme, MCS, as well as the resource allocation, which is which resource blocks are assigned to the PDSCH, and the duration of the assignment in OI-DM symbols. 5G NR uses a formula-based approach to compute the transport block size, where LTE uses a number of tables.

Because of the way the formula is defined, including quantization, there are several slightly different configurations of all those parameters that result in the same transport block size. Far from being an issue, this gives the operator more flexibility in selecting different parameters for the retransmission of a packet that didn't go through the first time.

This concludes this episode of the "5G Explained" video series on downlink data transmission.