On the Performance of Active Analog Self-Interference Cancellation Techniques for Beyond 5G Systems

Haifeng Luo,Mark Holm,Tharmalingam Ratnarajah

1 Institute for Digital Communications,School of Engineering,The University of Edinburgh,Edinburgh,UK

2 Huawei Technologies(Sweden)AB,Gothenburg,Sweden

Abstract:The in-band full-duplex(IBFD)mechanism is of interest in beyond 5G systems due to its potential to enhance spectral efficiency and reduce delay.To achieve the maximum gain of IBFD systems,the significant self-interference(SI)must be efficiently suppressed.The challenges of wideband selfinterference cancellation(SIC)lie in the radio frequency(RF)domain,where the performance will be limited by the hardware.This paper reviews current RF cancellation mechanisms and investigates an efficient mechanism for future wideband systems with minimum complexity.The working principle and implementation details of multi-tap cancellers are first introduced,then an optical domain-based RF canceller is reviewed,and a novel low-cost design is proposed.To minimize the cost and complexity of the canceller,the minimum required number of taps are analyzed.Simulation results show that with the commonly used 12-bits analog-to-digital converter(ADC)at the receiver,the novel optical domain-based canceller can enable efficient SIC in the 3GPP LTE specifications compatible system within 400 MHz bandwidth.

Keywords:full-duplex;beyond 5G system;selfinterference cancellation;optical domain-based canceller

I.INTRODUCTION

The in-band full-duplex(IBFD)mechanism invalidates a long-held assumption that transmission and reception cannot happen simultaneously within the same frequency band.It is supposed to significantly enhance the spectral efficiency and reduce the delay to meet the beyond 5G services[1—3].The key factor to enable the IBFD systems is to suppress the significant self-interference(SI)efficiently[2]since the received SI signal is about 100 dB stronger than the signal-of-interest(SOI)[1,3],which causes the SOI to be drowned in the significant interference.There have been several methods proposed to suppress the SI in the antenna,analog,and digital domains,which are summarized in[1,4].Previous studies have experimentally implemented full-duplex systems with efficient self-interference cancellation(SIC)and demonstrated the benefits of full-duplex(FD)over standard half-duplex(HD)[1,5].

Theoretically,self-interference can be cancelled in the digital domain with low complexity since it is caused by a fully known signal and the selfinterference channel can be precisely estimated.However,due to the limited dynamic range of the receiver chain,the self-interference has to be suppressed in the analog domain before the receiver chain to avoid the low-noise amplifier(LNA)and analog-to-digital converter(ADC)saturation[3,5].In the analog domain,the self-interference can be first mitigated via deploying transmit(TX)and receive(RX)antennas in a spatially separated way to achieve natural passive isolation in the propagation domain.Further researches propose to achieve the passive cancellation through absorptive shielding,cross-polarization,and a combination of the three mechanisms,which can provide up to 45.9 dB of isolation in the reflective room and 73.8 of isolation in the anechoic chamber[6].This impressive performance is experimentally achieved and may be degraded in practice.Then some active cancellation schemes are applied to take the residual selfinterference(RSI)to the receiver noise floor.The active cancellation follows a subtractive idea that a nulling signal,which has the same magnitude and inverse phase as the received SI,is generated and injected into the receiver at appropriate locations to subtract the SI out.Depending on the operation domain,it is categorized as RF cancellation and digital cancellation.The digital cancellation has low implementation complexity since it is operated in the digital domain,and its performance strongly depends on how accurate the effective baseband SI channel can be estimated and how the transceiver nonlinearity is modelled[7].In contrast,the RF cancellation is always a limiting factor of SIC performance due to hardware imperfections.Depending on the domain where the reference signal is tapped from,the RF cancellation can be categorized as baseband-tapping,i.e.,Rice architecture,and RF-tapping,i.e.,Stanford architecture[4,8].For the baseband-tapping Rice architecture,the digital pre-distortion and an auxiliary RF chain with a feedback loop are required since the reference baseband signal does not include the effects of transmitter nonlinearity,and the generated baseband nulling signal has to be converted into the RF domain by the auxiliary chain.Its performance is limited by the RF chain noise and nonlinearity,e.g.,phase noise and power amplifier nonlinearity,although the digital processing significantly improves the flexibility of such canceller[4],e.g.,it is easy to enlarge the operational bandwidth of such canceller.The RF-tapping Stanford architecture generates the nulling signal directly in the RF domain.Benefitting from the fact that the reference signal includes the effects of transmitter nonlinearity and distortions,the canceller only needs to mimic the linear wireless multi-path SI channel,which significantly reduces the signal processing complexity for the canceller.However,it is demonstrated that the operational bandwidth and the performance of such canceller is usually limited by the hardware[9,10].So,to enlarge the operational bandwidth of such cancellers is of our research interest to enable efficient SIC with low complexity in beyond 5G FD systems where the communication bandwidth will go to hundreds of MHz or even in the order of GHz while utilizing mmWave.To this end,we will first theoretically analyze the limiting factor of the Stanford architecture-based multi-tap canceller and then explore alternative solutions to break the limitations.

The article is organized as follows.Section II models the wireless SI channel and gives implementation details of the multi-tap RF canceller.In Section III,the limiting factor of wideband RF cancellation is theoretically analyzed at first,and then some novel optical domain-based cancellers are introduced to break the limitations.In Section IV,the minimum required number of taps is explored to minimize the cost and complexity of the canceller,and simulation results are demonstrated to support the statements.Then,how to extend the design to MIMO scenarios is discussed in Section V.Finally,conclusions are drawn in Section VI.

II.PRELIMINARIES

In a full-duplex node,the received signal consists of the self-interference signal imposed by its own transmitter in addition to the signal-of-interest from the intended node.The co-channel interference(CCI)from other nodes is assumed to be sufficiently mitigated by interference alignment as in[11]or appropriate resource allocation as in[12]that will be ignored in this paper.

2.1 SI Channel

The wireless SI channel consists of the direct path,i.e.,line-of-sight path,and multiple reflection paths in nature,where each path introduces specific delay and attenuation to the transmit signal.Except for the direct path,these reflection paths are unpredictable and depend on the geometry and location of the reflectors in the surrounding environment.For the canceller design,the most critical indicator is the maximum excess delay(MED)and the RMS delay spread of the SI channel[13].The MED determines the delay line length since the canceller has to cover the MED of the SI channel,while the RMS delay spread determines the coherence bandwidth,i.e.,frequency selectivity of the SI channel that significantly affects the performance of the RF canceller with limited taps.The passive antenna isolation mainly decreases the direct path strength,so it transforms the high-power line-ofsight dominated SI channel to a low-power multi-path channel and increases the frequency selectivity as it increases the RMS delay spread[6].The SI channel has been measured with passive antenna isolation in various environment and is demonstrated that the coherence bandwidth varies from 1 MHz to 4 MHz and the MED is below 200 ns for FR1(0.45 GHz—6 GHz)communications[13,14].

2.2 Multi-tap RF Canceller

Based on the multi-path nature of the SI channel,the optimal RF canceller design should consist of multiple tunable delay lines,i.e.,taps,to delay and attenuate the transmit SI signal as the wireless SI channel does.It is demonstrated that a single-tap canceller limits the operational bandwidth from the view of the frequency domain[10]or the cancellation performance from the view of the time domain[15].From the view of the frequency domain,the single-tap canceller has frequency-flat magnitude and phase that can only emulate the SI channel at a single frequency component that limits the operational bandwidth.From the view of the time domain,a single-tap canceller can be used to cancel the direct-path components,which is efficient only if there are not significant reflection paths.

For the multi-tap canceller design,there are usually two options available:1)Set the delays and attenuation of the taps to match the first significant paths of the SI channel;2)Set the delays of the taps uniformly distributed within the minimum and maximum excess delay of the SI channel and tune the weights to match the SI channel in the band of interest.Actually,the first one is not adaptive to environment changes surrounding the FD node since the true delay lines delay the signal through natural propagation that the delay depends on the untunable length of physical lines.Once the delays of significant paths change,the canceller cannot be tuned to cancel the components due to these paths.In contrast,the second choice gives the canceller much more flexibility that allows the canceller to be effective with random SI channel conditions.Besides,we want to highlight that it indeed theoretically makes sense to tune the canceller via minimizing the differences of impulse responses of the canceller and the SI channel.However,the Fourier transform suggests that any error in the impulse response distorts all frequency components.This implies that optimization in the time domain optimized the canceller over the entire frequency band.A factor that needs to be noted is that only the communication band is of our interest rather than the entire band.With a limit number of taps,it is for sure that optimizing within the band of interest outperforms optimizing over the entire band.So,we demonstrate that optimizing the canceller in the time domain yields a suboptimal solution,while the optimal solution should be found in the frequency domain over the band of interest.

2.3 Canceller Control

Including the hardware insertion and propagation losses,the delays,and the tuneable attenuation introduced by these taps,the canceller can be mathematically modelled in both the time and frequency domain(see in[9]for details).Assume the channel state information(CSI)of the SI channel is available,which can be precisely estimated in practice through some off-the-shelf methods as in[16],the optimisation problem can be cast as minimising the error between the frequency components of the canceller and SI channel within the band of interest through tuning the tuneable weights.This problem can be solved by the well-known Wiener solution or the“CVX”toolbox in MATLAB[9],which achieves the optimal performance.This requires an additional RF chain at the feedback loop of the canceller to obtain the CSI of the SI channel through channel estimation.An alternative way is to tune the attenuators through the gradient descent search method with the residual selfinterference signal strength,which is a function of the tunable weights[5,17].The feedback loop only consists of a residual signal strength indicator(RSSI),an ADC,and a field programmable gate array(FPGA)for this method.However,the perfect performance may not be achieved since the problem is not convex,and it is usually stuck in the local minimum instead of the global minimum.

III.WIDEBAND RF CANCELLER

3.1 Limiting Factors

For the aforementioned canceller,the number of taps actually determines the available degrees of freedom for the optimisation problem as it is achieved by tuning the attenuators at associated taps.Thus,the cancellation ability strongly depends on the number of taps,which is proved in our previous work in[3].Enlarging the operational bandwidth means more frequency components to be optimized so that more degrees of freedom,i.e.,taps,are required.However,it is demonstrated that generating a large number of true delay lines in the RF domain is challenging due to the lossy nature of RF components[9,10].It may come into mind to compensate for the significant loss by the amplifier,but this will introduce additional nonlinearity to the received signal,which is not related to the transmitter and is challenging to be mitigated.Besides,the RF components usually have poor frequency flatness over a wide band,which makes accurate wideband processing challenging[10].In summary,the RF components limit the operational bandwidth of the canceller from the following two aspects:

·inability to construct sufficient taps,

·lack of wideband processing properties.

3.2 FBG-Based Canceller

The optical components,which have smaller losses,high accuracy,and wideband RF processing properties,are promising to break the bandwidth limitation[9,10,15],so it is motivated to modulate the reference signal onto optical carriers to be processed in the optical domain for efficient wideband operations by exploring the benefits of optical components.In the optical domain,the fiber Bragg grating(FBG)is a perfect component to cause multiple delay lines in a single fiber,so a FBG-based RF canceller is proposed in[9]as shown in figure 1(a).The RF reference signal is first converted into the optical domain by being modulated ontoMoptical carriers with linearly increasing wavelengths.Taking advantage of the fact that the grating can reflect signals at specific wavelengths propagating through the core,theMgratings,i.e.,taps,can reflect signals at a wavelength fromλB,1toλB,Mcorresponding to the wavelengths of theMoptical carriers.According to the distribution of associated gratings,the signal modulated onto the optical carrier with wavelengthλB,mwill be delayed while propagating through the fiber.Then,these reflected delayed signals are down-converted back to the RF domain by photodiodes while the signals are weighted according to the signal power,which can be adjusted through the variable optical attenuators(VOAs).It is experimentally demonstrated that such design is able to provide about 20 dB of RF cancellation over 1 GHz bandwidth centered at 2.5 GHz[9].

Besides,it should be noted that the complex response is represented through a superposition of 4 phase-shifted signals.In other words,four branches identical to the one shown in figure 1(a)are utilized to represent the complex response with phase-shifted RF reference signal input.If the optimization problem is directly solved in the complex domain,perfect phase shifters have actually been assumed.The obtained weights will not be optimal in practice as long as the phase shifters have phase noise.So a novel robust algorithm is proposed,which divides the response of the canceller into four parts to get the optimal weights separately for each branch[9].Actually,such phase noise usually occurs if an analog control is used but may not be a big problem with digital control.Therefore,we proposed to tune the weights in the complex domain and do the assignment to save the implementation cost in[3].

3.3 Fiber Array-Based Canceller

The FBG-based canceller is efficient but costly due to the requirements of expensive optical components,e.g.,a large number of tunable lasers and customized FBGs.To save the cost,we propose a novel fiber array-based design,as shown in figure 1(b).Different from the FBG-based canceller,the received signal is also converted into the optical domain,that the subtraction also happens in the optical domain to guarantee the effective operational bandwidth.And we explore some simple optical components,i.e.,fiber array with different fiber lengths and VOAs,to cause multiple delay lines instead of the elaborate FBG.The benefit is that the reference signal can be modulated onto a single optical carrier and then be splitted toMmodulated signals to be independently processed by the fiber array.This significantly reduces the required tunable lasers fromMto 1 for every single branch.Besides,the fiber array is much cheaper than the FBG,which helps to save the cost of the RF canceller.Compared to the FBG-based design,the main weakness of the fiber array-based design is that the splitter and power combiner have increasing insertion loss with an increasing number of taps,while the FBG-based design has almost constant losses with an increasing number of taps.The large insertion loss is always a limiting factor for conventional RF components constructed cancellers since compensation for the insertion loss by the PA introduces additional nonlinearity that is not related to the transmitter and is challenging to be mitigated.Luckily,the large insertion loss can be compensated by increasing the optical carrier power in our design instead of a PA at the output.Although the electro-optic modulator will yield higher nonlinearity with higher optical carrier power,the later large insertion loss will also decrease it that makes the additional nonlinearity almost constant and at an acceptable level so that it will not limit the number of taps that can be created.It should be noted that the optical combiner must be SM-MM(single mode - multi mode)combiner to avoid coherent beat noise while combing these same-wavelength carriers,and the modulator must work in the amplitude modulation mode.Figure 1(b)also only shows a single branch,while four identical branches with phase-shifted input are required in practice to represent the positive and negative parts of the I-and Q-channels of the complex SI channel.

Figure 1.The Architecture of the optical domain-based RF cancellers.

Figure 2.Power budget of a FD node at different SIC steps.

Figure 3.System capacity of FD radios with ASIC.

Figure 4.Performance evaluation of the analog and digital cancellation.

IV.EFFICIENT RF CANCELLER

4.1 Minimum Efficient Performance

As mentioned above,it is not necessary that the SI is suppressed to the receiver noise floor in the analog domain by the RF canceller from the view of cost.The most efficient receiver should only leverage the RF canceller to suppress the SI to avoid the ADC saturation,and the low-cost digital canceller takes responsibility to bring the RSI to the receiver noise floor.So,the minimum efficient RF cancellation amount should be analyzed to determine the taps in the canceller.The power budget of a FD node is illustrated in figure 2.The essence of RF cancellation is to make the quantization noiseNqto be at or below the receiver thermal noise floorNt.Thus,the minimum efficient RF cancellation level strongly depends on the received SI power,the dynamic range of the ADC,and the receiver noise floor,as figure 2 suggests,so it is affected by the operational bandwidth,effective bits of the ADC,transmit power of the SI,and antenna isolation amount.Actually,the minimum efficient RF cancellation amount is when the quantization noise is identical to the thermal noise floor,and further suppression in the RF domain does not provide benefits.

Start with a simple case,assume the SI is 100 dB larger than the received signal of interest and the RSI and linear distortions can be eliminated through digital cancellation.Figure 3(a)shows the maximum achievable rate of the FD node with specific analog self-interference cancellation(ASIC)level compared to ideal FD and HD nodes under various SINR.The results are achieved with 12-bits ADC used at the receiver chain,and the nonlinear distortion factor of the receiver chain is set as-90 dB.It shows that about 60 dB of ASIC is efficient to achieve the maximum gain of FD under this scenario.Then,we investigate how the hardware of the receiver chain affects the minimum required ASIC that can achieve the maximum spectral efficiency of the FD system.Figure 3(b)demonstrates how much ASIC is essential with various hardware configurations,i.e.,the number of bits of the ADC and the distortions of the receiver chain.The results are achieved assuming transmit SI power of 24 dBm,and the intended node has transmit power of 23 dBm.The intended node is 70 meters away.The parameters of the system and the pathloss model follow the 3GPP LTE specifications[18].Figure 3(b)shows the achievable SINR of the signal-of-interest against the ASIC depths in a 3GPP LTE specifications compatible system.It can be seen that about 60 dB of analog cancellation is sufficient to achieve the maximum efficiency of FD systems with the 12-bits ADC,which corresponds to figure 3(a).Besides,it proves that further ASIC does not provide benefits as the achievable SINR does not increase with deeper ASIC as long as 60 dB of ASIC is achieved.With the help of the antenna isolation in[6],about less than 20 dB of RF cancellation will be efficient,which has been achieved over 1 GHz with 20 taps created by FBGs in[8].

So far,ideal digital cancellation has been assumed.Then,the nonlinear digital cancellation is utilized to process the RSI to check whether the RF cancellation is efficient in practice.Considering an OFDM system with 64 subcarriers with 3.125 MHz subcarrier space centered at 2.5 GHz,and the thermal noise density is 174 dBm/Hz.The transmit power of the SI,and the signal of interest are 40 dBm(14 dBm/MHz)and 39 dBm(13 dBm/MHz),respectively.In addition to the RSI,the nonlinear digital canceller also processes the power amplifier(PA)nonlinearity here while ignoring other transceiver nonlinearities since the transceiver nonlinearity is dominated by the PA[19].The PA nonlinearity is modelled by the well-known memory polynomial model,which is proved to be effective in describing the practical PA nonlinearity[20].Figure 4 shows the overall SIC performance with the fiber array-based RF canceller and the nonlinear digital canceller.In order to check the cancellation level at each step,we move the RF signal to the baseband for clarity.Figure 4 is achieved using a 17-taps(right)and 8-taps(left)RF canceller,respectively.The 8-taps RF canceller achieves 7.15 dB of RF cancellation,which is not sufficient to avoid ADC saturation and results in large quantization noise.Thus,the nonlinear digital canceller cannot bring the RSI to the level of noise floor of the receiver.In contrast,the 17-taps RF canceller achieves 31.20 dB of RF cancellation,and the nonlinear digital canceller brings the RSI to the level of noise floor,which allows the FD node to achieve maximum efficiency.The results suggest that the RF cancellation is a limiting factor of the efficiency due to the quantization noise,which is unable to be eliminated.The efficient SIC can only be achieved with efficient RF cancellation.

4.2 Number of Taps

It is theoretically analyzed that the number of taps decides the available operational bandwidth and the achievable cancellation performance,and wideband operations need a large number of taps.Here,we will experimentally prove this through the simulation results.Let the sampling interval Δωto be 1 MHz,which is identical to or narrower than the coherence bandwidth of the SI channel as discussed in Section II-A,so it is sufficient to describe the frequency characteristics of the SI channel.The propagation loss is set as 2.976 dB/m and 0.461 dB/m for the microstrip and the coiled fiber(r= 2 cm),respectively.Figure 5 provides the comparison for the conventional RF components(microstrip)constructed canceller and the novel fiber array-based canceller.For simplicity,the nonlinear effects and the noise while converting the signal to and back from the optical domain are ignored here,and we assume perfect CSI of the SI channel is available.This clearly shows the relationship between the operational bandwidth,cancellation performance,and the number of taps and explains why the novel optical domain-based canceller is critical for wideband RF cancellation.It can be seen that no matter how many taps are created using microstrips,desired performance cannot be achieved over wide bands due to a large insertion loss penalty.In contrast,the novel fiber array-based design can achieve efficient wideband RF cancellation,e.g.,more than 30 dB within bandwidth beyond 400 MHz.For a specific FD transceiver,the minimum efficient RF cancellation amount can be calculated according to figure 2 with the knowledge of antenna isolation amount,transmit SI power,the dynamic range of the ADC,and the thermal noise of the receiver.Then,the number of taps can be decided according to figure 5(b)to avoid redundant taps created.

Figure 5.Cancellation performance versus the bandwidth and number of utilised taps.

Figure 6.Cancellation performance of the two transmitter beamformer designs with different antenna array sizes.

V.FURTHER DISCUSSION

Above,we are considering an ideal scenario that the attenuators are controlled without resolution limitations.For a more practical scenario,we should consider limited control resolution.We made MATLAB simulations on how the RF canceller performs with various control resolutions.The simulation results show that 12-bits of control resolution is sufficient to achieve the performance close to the unlimited resolution scenario when there are not too many taps utilized.However,the performance decreases with an increasing number of taps.8-bits of control resolution will introduce about 5 to 10 dB of performance degradation depending on the number of taps,while 4-bits of resolution can hardly achieve efficient RF cancellation.

Besides,to enable the IBFD in beyond 5G wireless communication systems,it is worthwhile to consider the implementation in the multiple-input-andmultiple-output(MIMO)or massive MIMO systems,which are promising techniques to be used in the beyond 5G systems.Typically,for the Stanford architecture,it calls forNT ×NRcancellers to match each transmit and receive antennas pair in MIMO systems,whereNTis the number of transmit antennas andNRis the number of receive antennas.Such a large number of cancellers is always a limiting factor to extend the RF cancellation into MIMO systems[8].While the Rice architecture is more readily extended to MIMO systems,which only requiresNRauxiliary RF chains[8,19].To practically extend the Stanford architecture with such optical domain-based canceller to MIMO systems,a novel hybrid beamforming scheme in[21]should be considered,which significantly reduces the number of required RF cancellers fromNT ×NRtoLT×LR,e.g.,from 128×128 to 4×4,whereLTandLRare the number of RF chains at the transmitter and receiver,respectively.Then,the fiber array-based canceller is promising to be leveraged for efficient wideband RF cancellation in beyond 5G systems with low cost and complexity.

Furthermore,it is supposed to take advantage of the directional beams in MIMO systems to mitigate the SI in the propagation domain by some simple preprocessing at the transmitter to avoid the challenging RF cancellation.A beamforming cancellation(BFC)design is recently proposed in[22],which works based on the null-space projection idea.The proposed BFC scheme projects the precoder into the null-space of the effective SI channel,i.e.,the wireless SI channel and the effect of combiner so that the SI will be nullified by the combiner.Alternatively,we can project the precoder,i.e.,transmitter beamformer,to the null-space of the wireless SI channel without the combiner effects so that the SI can be nullified in the propagation domain before the receiver chain.A leakage optimization can be done if the wireless SI channel is of fullrank that there are no dimensions for the null-space to survive,where the leakage null-space consists of the singular vector associated with the smallest singular value of the SI channel.The SIC amount achieved by this leakage null-space projected transmitter beamformer is shown as ‘NSP’ in figure 6,whereNbsandNuedenote the number of TX and RX antennas at the FD base station and the intended user equipment,respectively.For more practical considerations,we should jointly consider the SIC with beamforming designs in[3]or[22]to take advantage of the beamforming in other aspects,e.g.,interference management and energy efficiency.We tried to add an additional constraint for the SIC by the precoder into the beamforming design in[18]and solved this problem jointly to achieve the precoder gain for SIC in the propagation domain while remaining the system capacity.The results are shown as‘Joint’in figure 6.It shows that the leakage null-space projected transmitter beamformer achieves considerable SIC that the RF cancellation is not essential with 8 or more TX and RX antennas at the FD node.In contrast,the joint design achieves a lower SIC amount.But it does not degrade the system capacity while the leakage null-space significantly degrades it.Thus,the joint design is a good start,and a more advanced beamforming design is encouraged to improve the SIC ability while maintaining the system capacity.

VI.CONCLUSION

We have highlighted two novel optical domain-based RF cancellers above and analytically proved that such optical domain-based cancellers have the potential to enable efficient active analog cancellation over extremely wide bands with simulation results.In order to avoid creating spare taps in the canceller,the minimum efficient RF cancellation level is explored to decide the appropriate number of taps,which depend on:

·Operational bandwidth,

·Transmit power of the SI signal,

·Antenna isolation amount,

·Frequency selectivity of the channel,

·Number of bits of the ADC at the receiver.

The simulation results demonstrate that the novel fiber array-based design is able to enable efficient wideband SIC and is going to help the full-duplex system become a key part of the next generation radios.We further discussed extending the RF cancellation to MIMO systems in practice and motivated to investigate advanced beamforming cancellation schemes.

ACKNOWLEDGEMENT

The work of H.Luo and M.Holm was supported by the research grant from Huawei Technologies(Sweden)AB.The work of T.Ratnarajah was supported by the U.K.Engineering and Physical Sciences Research Council(EPSRC)under Grant EP/P009549/1.