16.1.5.1 The Primary Synchronization Sequence (PSS)

The PSS is the first signal that a device entering the system will search for. At that stage, the device has no knowledge of the system timing. Furthermore, even though the device searches for a cell at a given carrier frequency, there may, due to inaccuracy of the device internal frequency reference, be a relatively large deviation between the device and network carrier frequency. The PSS has been designed to be detectable despite these uncertainties.

Once the device has found the PSS, it has found synchronization up to the periodicity of the PSS. It can then also use transmissions from the network as a reference for its internal frequency generation, thereby to a large extent eliminating any frequency deviation between the device and the network.

As described above, the PSS extends over 127 resource elements onto which a PSS sequence $\left\{x_n\right\}=x_n\left(0\right),x_n\left(1\right),\dots ,x_n\left(126\right)$ is mapped (see Fig. 16.4).

There are three different PSS sequences $\left\{x_0\right\}$, $\left\{x_1\right\}$, and $\left\{x_2\right\}$, derived as different cyclic shifts of a basic length-127 M-sequence [70] $\left\{x\right\}=x\left(0\right),\mathit{x}\left(1\right),\dots ,\mathit{x}\left(126\right)$ generated according to the recursive formula (see also Fig. 16.5):

$x\left(n\right)=x\left(n-7\right)⨁x\left(n-3\right)$

By applying different cyclic shifts to the basic M-sequence $x\left(n\right)$, three different PSS sequences $x_0\left(n\right)$, $x_1\left(n\right)$, and $x_2\left(n\right)$ can be generated according to:

$x_0\left(n\right)=x\left(n\right)\mathrm{;}{x}_1\left(n\right)=x\left(n+43\mathit{mod}127\right)\mathrm{;}{x}_2\left(n\right)=x\left(n+86\mathit{mod}127\right)$

Which of the three PSS sequences to use in a certain cell is determined by the physical cell identity (PCI) of the cell. When searching for new cells, a device thus must search for all three PSSs.

16.1.5.2 The Secondary Synchronization Sequence (SSS)

Once a device has detected a PSS it knows the transmission timing of the SSS. By detecting the SSS, the device can determine the PCI of the detected cell. There are 1008 different PCIs. However, already from the PSS detection the device has reduced the set of candidate PCIs by a factor 3. There are thus 336 different SSSs, that together with the already-detected PSS provides the full PCI. Note that, since the timing of the SSS is known to the device, the per-sequence search complexity is reduced compared to the PSS, enabling the larger number of SSS sequences.

The basic structure of the SSS is the same as that of the PSS (Fig. 16.4), that is, the SSS consists of 127 subcarriers to which an SSS sequence is applied.

On an even more detailed level, each SSS is derived from two basic M-sequences generated according to the recursive formulas

$x\left(n\right)=x\left(n-7\right)⨁x\left(n-3\right)$

$y\left(n\right)=y\left(n-7\right)⨁y\left(n-6\right)$

The actual SSS sequence is then derived by adding the two M sequences together, with different shifts being applied to the two sequences.

$x_{m_1,m_2}\left(n\right)=x\left(n+m_1\right)+y\left(n+m_2\right)$

16.1.5.3 PBCH

While the PSS and SSS are physical signals with specific structures, the PBCH is a more conventional physical channel on which explicit channel-coded information is transmitted. The PBCH carries the master information block (MIB), which contains a small amount of information that the device needs in order to be able to acquire the remaining system information broadcast by the network.⁶

Table 16.2 lists the information carried within the PBCH. Note that the information differs slightly depending on if the carrier is operating in lower-frequency bands (FR1) or higher-frequency bands (FR2).

As already mentioned, the SS-block time index identifies the SS-block location within an SS burst set. As described in Section 16.1.4, each SS block has a well-defined position within an SS burst set which, in turn, is contained within the first or second half of a 5 ms frame. From the SS-block time index, in combination with the half-frame bit (see below), the device can thus determine the frame boundary.

The SS-block time index is provided to the device as two parts:

Eight different scrambling patterns can be used for the PBCH, allowing for the implicit indication of up to eight different SS-block time indices. This is sufficient for operation below 6 GHz (FR1) where there can be at most eight SS blocks within an SS burst set.⁷

For operation in the higher NR frequency range (FR2) there can be up to 64 SS blocks within an SS burst set, implying the need for three additional bits to indicate the SS-block time index. These three bits, which are thus only needed for operation above 10 GHz, are included as explicit information within the PBCH payload.

The CellBarred flag consist of two bits:

If detecting that a cell is barred and that access to other cells on the same frequency is not permitted, a device can and should immediately re-initiate cell search on a different carrier frequency.

It may seem strange to deploy a cell and then prevent devices from accessing it. Historically this kind of functionality has been used to temporarily prevent access to a certain cell during maintenance. However, the functionality has additional usage within NR due to the possibility for non-standalone NR deployments for which devices should access the network via the corresponding LTE carrier. By setting the CellBarred flag for the NR carrier in an NSA deployment, the network prevents NR devices from trying to access the system via the NR carrier.

The 1st PDSCH DMRS position indicates the time-domain position of the first DMRS symbol assuming DMRS Mapping Type A (see Section 9.11).

The SIB1 numerology provides information about the subcarrier spacing used for the transmission of the so-called SIB1, which is part of the system information (see Section 16.1.6). The same numerology is also used for the downlink Message 2 and Message 4 that are part of the random-access procedure (see Section 16.2). Although NR supports four different numerologies (15 kHz, 30 kHz, 60 kHz, and 120 kHz) for data transmission, for a given frequency band there are only two possible numerologies. Thus, one bit is sufficient to signal the SIB1 numerology.

The SIB1 configuration provides information about the search space, corresponding CORESET, and other PDCCH-related parameters that a device needs in order to monitor for scheduling of SIB1.

The CRB grid offset provides information about the frequency offset between the SS block and the common resource block grid. As discussed in Section 16.1.2, the frequency-domain position of the SS block relative to the carrier is flexible and does not even have to be aligned with the carrier CRB grid. However, for SIB1 reception, the device needs to know the CRB grid. Thus, information about the frequency offset between the SS block and the CRB grid must be provided within the PBCH in order to be available to devices prior to SIB1 reception.

Note that the CRB grid offset only provides the offset between the SS block and the CRB grid. Information about the absolute position of the SS block within the overall carrier is then provided within SIB1.

The half-frame bit indicates if the SS block is located in the first or second 5 ms part of a 10 ms frame. As mentioned above, the half-frame bit, together with the SS-block time index, allows for a device to determine the cell frame boundary.

All information above, including the CRC, is jointly channel coded and rate-matched to fit the PBCH payload of an SS block.

Although all the information above is carried within the PBCH and is jointly channel coded and CRC-protected, some of the information is strictly speaking not part of the MIB. The MIB is assumed to be the same over an 80 ms time interval (eight subframes) as well as for all SS blocks within an SS burst set. Thus, the SS-block time index, which is inherently different for different SS blocks within an SS burst set, the half-frame bit and the four least significant bits of the SFN are PBCH information carried outside of the MIB.⁸

16.2.1.1 Characteristics of Preamble Transmission

Several factors impact the structure of the preamble transmission.

As described in Section 15.2, the transmission timing of NR uplink transmissions is typically controlled by the network by means of regularly provided time-adjustment commands (“closed-loop timing control”).

Prior to preamble transmission, there is no such closed-loop timing control in operation. Rather, the device must base the preamble transmission timing on the received timing of some downlink signal, in practice the received timing of the acquired SS block. Consequently, there will be an uncertainty in the preamble reception timing of at least two times the maximum propagation delay within the cell. For cell sizes in the order of a few hundred meters, this uncertainty will be in the order of few microseconds. However, for large cells the uncertainty could be in the order of 100 μs or even more.

In general, it is up to the base-station scheduler to ensure that there are no other transmissions in the uplink resources in which preamble transmissions may take place. When doing this, the network needs to take the uncertainty in the preamble reception timing into account. In practice the scheduler needs to provide an extra guard time that captures this uncertainty (see Fig. 16.7).

Note that the presence of the guard time is not part of the NR specifications but just a result of scheduling restrictions. Consequently, different guard times can easily be provided to match different uncertainty in the preamble reception timing, for example, due to different cell sizes.

In addition to the lack of closed-loop timing control, there is also no closed-loop power control in operation prior to preamble transmission. Rather, similar to the transmission timing, the device must determine its transmit power based on the received power of some downlink signal, in practice the received power of the acquired SS block. The lack of closed-loop power control may lead to a relatively large uncertainty in the received preamble power for several reasons:

Finally, while normal uplink transmissions are typically based on explicit scheduling grants, thereby enabling contention-free access, initial random access is inherently contention-based, implying that multiple devices may initiate preamble transmission simultaneously. The preamble should preferably be able to handle such a situation and as much as possible allow for correct preamble reception when such “collisions” occur.

16.2.1.2 RACH Resources

Within a cell, preamble transmission can take place within a configurable subset of slots (the RACH slots) that repeats itself every RACH configuration period (see Fig. 16.8).⁹

Furthermore, within these “RACH slots”, there may be multiple frequency-domain RACH occasions jointly covering K·M consecutive resource blocks where M is the preamble bandwidth measured in number of resource blocks and K is the number of frequency-domain RACH occasions.

For a given preamble type, corresponding to a certain preamble bandwidth, the overall available time/frequency RACH resource within a cell can thus be described by:

16.2.1.3 Basic Preamble Structure

Fig. 16.9 illustrates the basic structure for generating NR random-access preambles. A preamble is generated based on a length-L preamble sequence $p_0,p_1,\dots ,p_{L-1}$ which is DFT precoded before being applied to a conventional OFDM modulator. The preamble can thus be seen as a DFTS-OFDM signal. It should be noted though that one could equally well see the preamble as a conventional OFDM signal based on a frequency-domain sequence $P_0,P_1,\dots ,{\mathit{P}}_{L-1}$ being the discrete Fourier transform of the sequence $p_0,p_1,\dots ,p_{L-1}$.

The output of the OFDM modulator is then repeated N times, after which a cyclic prefix is inserted. For the preamble, the cyclic prefix is thus not inserted per OFDM symbol but only once for the block of N repeated symbols.

Different preamble sequences can be used for the NR preambles. Similar to, for example, uplink SRS, the preamble sequences are based on Zadoff–Chu sequences [25]. As described in Section 8.3.1, for prime-length ZC sequences, which is the case for the sequences used as a basis for the NR preamble sequences, there are L–1 different sequences, with each sequence corresponding to a unique root index.

Different preamble sequences can be generated from different Zadoff–Chu sequences corresponding to different root indices. However, different preamble sequences can also be generated from different cyclic shifts of the same root sequence. As described in Section 8.3.1, such sequences are inherently orthogonal to each other. However, this orthogonality is retained at the receiver side only if the relative cyclic shift between two sequences is larger than any difference in their respective receive timing. Thus, in practice only a subset of the cyclic shifts can be used to generate different preambles, where the number of available shifts depends on the maximum timing uncertainty which, in turn, depends on, for example, the cell size. For small cell sizes a relatively large number of cyclic shifts can often be used. For larger cells, a smaller number of cyclic shifts will typically be available.

The set of cyclic shifts that can be used within a cell is given by the so-called zero-correlation zone parameter which is part of the cell random-access configuration provided within SIB1. In practice, the zero-correlation zone parameter points to a table that indicates the set of cyclic shifts available in the cell. The name “zero-correlation zone” comes from the fact that the different tables indicated by the zero-correlation-zone parameter have different distances between the cyclic shifts, thus providing larger or smaller “zones” in terms of timing error for which orthogonality (=zero correlation) is retained.

16.2.1.4 Long vs Short Preambles

NR defines two types of preambles, referred to as long preambles and short preambles, respectively. As the name suggests, the two preamble types differ in terms of the length of the preamble sequence. They also differ in the numerology (subcarrier spacing) used for the preamble transmission. The type of preamble is part of the cell random-access configuration, that is, within a cell only one type of preamble can be used for initial access.

Long preambles are based on a sequence length L=839 and a subcarrier spacing of either 1.25 kHz or 5 kHz. The long preambles thus use a numerology different from any other NR transmissions. The long preambles partly originate from the preambles used for LTE random-access [28]. Long preambles can only be used for frequency bands below 6 GHz (FR1).

As illustrated in Table 16.3 there are four different formats for the long preamble where each format corresponds to a specific numerology (1.25 kHz or 5 kHz), a specific number of repetitions (the parameter N in Fig. 16.9), and a specific length of the cyclic prefix. The preamble format is also part of the cell random-access configuration, that is, each cell is limited to a single preamble format. It could be noted that the two first formats of Table 16.3 are identical to the LTE preamble formats 0 and 2 [14].

In the previous section it was described how the overall RACH resource consists of a set of slots and resource blocks in the time-domain and frequency-domain, respectively. For long preambles, which use a numerology that is different from other NR transmissions, the slot and resource block should be seen from a 15 kHz numerology point of view. In the context of long preambles, a slot thus has a length of 1 ms, while a resource-block has a bandwidth of 180 kHz. A long preamble with 1.25 kHz numerology thus occupies six resource blocks in the frequency domain, while a preamble with 5 kHz numerology occupies 24 resource blocks.

It can be observed that preamble format 1 and preamble format 2 in Table 16.3 correspond to a preamble length that exceeds a slot. This may appear to contradict the assumption of preamble transmissions taking place in RACH slots of length 1 ms as discussed in Section 16.2.1.2. However, the RACH slots only indicate the possible starting positions for preamble transmission. If a preamble transmission extends into a subsequent slot, this only implies that the scheduler needs to ensure that no other transmissions take place within the corresponding frequency-domain resources within that slot.

Short preambles are based on a sequence length L=139 and use a subcarrier spacing aligned with the normal NR subcarrier spacing. More specifically, short preambles use a subcarrier spacing of:

In the case of short preambles, the RACH resource described in Section 16.2.1.2 is based on the same numerology as the preamble. A short preamble thus always occupies 12 resource blocks in the frequency domain regardless of the preamble numerology.

Table 16.4 lists the preamble formats available for short preambles. The labels for the different preamble formats originate from the 3GPP standardization discussions during which an even larger set of preamble formats were discussed. The table assumes a preamble subcarrier spacing of 15 kHz. For other numerologies, the length of the preamble as well as the length of the cyclic prefix scale correspondingly, that is, with the inverse of the subcarrier spacing.

The short preambles are, in general, shorter than the long preambles and often span only a few OFDM symbols. In most cases it is therefore possible to have multiple preamble transmissions multiplexed in time within a single RACH slot. In other words, for short preambles there may not only be multiple RACH occasions in the frequency domain but also in the time domain within a single RACH slot (see Table 16.5).

It can be noted that Table 16.5 includes additional formats A1/B1, A2/B2, and A3/B3. These formats correspond to the use of a mix of the “A” and “B” formats of Table 16.4, where the A format is used for all except the last RACH occasion within a RACH slot. Note that the A and B preamble formats are identical except for a somewhat shorter cyclic prefix for the B formats.

For the same reason there are no explicit formats B2 and B3 in Table 16.5 as these formats are always used in combination with the corresponding A formats (A2 and A3) according to the above.

16.2.1.5 Beam Establishment During Initial Access

A key feature of the NR initial access is the possibility to establish a suitable beam pair already during the initial-access phase and to apply receiver-side analog beam-sweeping for the preamble reception.

This is enabled by the possibility of associating different SS-block time indices with different RACH time/frequency occasions and/or different preamble sequences. As different SS-block time indices in practice correspond to SS-block transmissions in different downlink beams, this means that the network, based on the received preamble, will be able to determine the downlink beam in which the corresponding device is located. This beam can then be used as an initial beam for subsequent downlink transmissions to the device.

Furthermore, if the association between SS-block time index and RACH occasion is such that a given time-domain RACH occasion corresponds to one specific SS-block time index, the network will know when, in time, preamble transmission from devices within a specific downlink beam will take place. Assuming beam correspondence, the network can then focus the uplink receiver beam in the corresponding direction for beam-formed preamble reception. In practice this implies that the receiver beam will be swept over the coverage area synchronized with the corresponding downlink beam sweep for the SS-block transmission.

Note that beam-sweeping for preamble transmission is only relevant when analog beam-forming is applied at the receiver side. If digital beam-forming is applied, beam-formed preamble reception can be done from multiple directions simultaneously.

To associate a certain SS-block time index with a specific random-access occasion and a specific set of preambles, the random-access configuration of the cell specifies the number of SS-block time indices per RACH time/frequency occasion. This number can be larger than one, indicating that multiple SS-block time indices correspond to a single RACH time/frequency occasion. However, it can also be smaller than one, indicating that one single SS-block time index corresponds to multiple RACH time/frequency occasions.

SS-block time indices are then associated with RACH occasions in the following order:

Fig. 16.10 exemplifies the association between SS-block time indices and RACH occasions under the following assumptions:

16.2.1.6 Preamble Power Control and Power Ramping

As discussed above, preamble transmission will take place with a relatively large uncertainty in the required preamble transmit power. Preamble transmission therefore includes a power-ramping mechanism where the preamble may be repeatedly transmitted with a transmit power that is increased between each transmission.

The device selects the initial preamble transmit power based on estimates of the downlink path loss in combination with a target received preamble power configured by the network. The path loss should be estimated based on the received power of the SS block that the device has acquired and from which it has determined the RACH resource to use for the preamble transmission. This is aligned with an assumption that if the preamble transmission is received by means of beam-forming the corresponding SS block is transmitted with a corresponding beam-shaper. If no random-access response (see below) is received within a predetermined window, the device can assume that the preamble was not correctly received by the network, most likely due to the fact that the preamble was transmitted with too low power. If this happens, the device repeats the preamble transmission with the preamble transmit power increased by a certain configurable offset. This power ramping continues until a random-access response has been received or until a configurable maximum number of retransmissions has been carried out, alternatively a configurable maximum preamble transmit power has been reached. In the two latter cases, the random-access attempt is declared as a failure.