WO2019245598A1

WO2019245598A1 - Filter modelling for pim cancellation

Info

Publication number: WO2019245598A1
Application number: PCT/US2018/062443
Authority: WO
Inventors: Shavantha Kularatna; Ramasamy Gopalan; Teddy DAVIS; Imke HUISMANN; Marko FLEISCHER
Original assignee: Nokia Solutions And Networks Oy; Nokia Usa Inc.
Priority date: 2018-06-19
Filing date: 2018-11-26
Publication date: 2019-12-26
Also published as: EP3827522A1; WO2019245533A1

Abstract

An apparatus, method and computer program product is disclosed. The apparatus comprises means for sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband and means for transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter. The apparatus also comprises a means for providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

Description

FILTER MODELLING FOR PIM CANCELLATION

Field

This specification relates to an apparatus, method and computer program product relating to filter modelling, for example for passive intermodulation (PIM) cancellation.

Background

Passive InterModulation (PIM) is a well-known telecommunications issue. It is caused if plural signals are transmitted through a non-linear system. A non-linear system may be a system comprising active components, but it may also occur in passive

components, e.g. due to aging antennas, corroded or loose connectors and passive duplex filters etc. Due to PIM, intermodulation products occur at frequencies/ corresponding to k_af_a + k_bft, + k_cf_c + ..., wherein f_a, fi, f_,... are the frequencies of the plural signals, and k_a, k_b, kc,... are integer coefficients (positive, negative, or o). The sum k_a+k_b+k_c+_,... is denoted as the order of the intermodulation product, denoted as IMP3,

IMP5, IMP7 etc. for IMP of 3^rd, 5^th, and 7^th order, respectively. The amplitude of the IMPs decreases with increasing order of the IMPs. IMP3 is typically most relevant because it is located close to the input signal and has relatively high amplitude. Issues may occur when PIM products line up with received signals in a receiver, as if the PIM power is greater than the received signal itself, the receiver decoding process may fail due to negative signal to noise ratio.

Summary

According to one aspect, there is provided an apparatus, comprising: means for sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband; means for transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and means for providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

The means for providing the time domain complex coefficients may comprise means for transforming or solving an equation that includes frequency domain samples. The means for sampling the frequency domain response may be configured to sample S21 forward complex gain parameters produced by the filter. The means for sampling the frequency domain response may be configured to sample at a resolution of substantially too kHz or less. The apparatus may further comprise means for modifying the window of frequency domain response samples, prior to transforming, to have a sample frequency fc, such that: fc =fs/n, where fs is the sample frequency of a subsequent passive intermodulation cancellation means, and wherein n = 2, 3, 4 etc.

The apparatus may further comprise means for over-sampling the filter model after transformation to match the sample frequency (fe ) of the subsequent passive intermodulation cancellation means.

The means for transforming the samples may comprise means for taking the inverse discrete Fourier transform (IDFT) of the frequency domain samples.

The apparatus may further comprise means for inputting noise to the provided filter model and means for performing linear regression to fit the input noise to the output of the filter model to produce an updated filter model comprising updated time domain complex coefficients for replacing the provided filter model and for use in subsequent passive intermodulation cancellation.

The noise may be additive white Gaussian noise (AWGN).

The means for providing the filter model may be configured to use only a subset of the time domain complex coefficients for the model.

The subset of time domain complex coefficients may comprises substantially 35 coefficients or fewer. The apparatus may further comprising means for performing passive intermodulation cancellation on a received signal based on the rate-matched filter model incorporated into non-linear terms. The apparatus may further comprise means for filtering the derived PIM model prior to performing passive intermodulation to reject duplicate frequency components outside of the filter passband.

The cancellation means maybe configured to remove, at a sample frequency (fs), passive intermodulation components from the received signal by subtracting from the received signal a cancelling signal determined using the filter model incorporated into non-linear terms.

The cancellation means may provide the cancelling signal by:

using the filter model incorporated into non-linear terms to produce a non-linear vector ( v );

determining cancellation coefficients (W_cancellation coefficients) based on the non-linear vector ( v );

determining the cancelling signal as the product of the non-linear vector 0)and the cancellation coefficients (W_cancellation coefficients)·

The cancellation means may determine the cancellation components by:

where Cross Correlation Vector = v^H. R_x,

Autocorrelation Matrix = v . v^H

v is the nonlinear vector, H is the Hermitian operation and Rx is the received signal. The apparatus may be provided in a cellular base station.

According to another aspect, there may be provided a method, comprising: sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband; transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

Sampling the frequency domain response may comprise sampling S21 forward complex gain parameters produced by the filter.

Sampling the frequency domain response may be at a resolution of substantially too kHz or less.

The method may further comprise modifying the window of frequency domain response samples, prior to transforming, to have a sample frequency fc, such that: fc =fs/n, where fs is the sample frequency of a subsequent passive intermodulation cancellation operation, and wherein n = 2, 3, 4 etc.

The method may further comprise over-sampling the filter model after transformation to match the sample frequency (fs) of the subsequent passive intermodulation cancellation operation.

Transforming the samples may comprise taking the inverse discrete Fourier transform (IDFT) of the frequency domain samples.

The method may further comprise inputting noise to the provided filter model and performing linear regression to fit the input noise to the output of the filter model to produce an updated filter model comprising updated time domain complex coefficients for replacing the provided filter model and for use in a subsequent passive

intermodulation cancellation operation.

The noise may be additive white Gaussian noise (AWGN).

Providing the filter model may use only a subset of the time domain complex coefficients for the model. The subset of time domain complex coefficients may comprise substantially 35

coefficients or fewer.

The method may further comprise performing passive intermodulation cancellation on a received signal based on the rate-matched filter model incorporated into non-linear terms.

The method may further comprise filtering the derived PIM model prior to performing passive intermodulation to reject duplicate frequency components outside of the filter passband.

The cancellation operation may remove, at a sample frequency (fs), passive

intermodulation components from the received signal by subtracting from the received signal a cancelling signal determined using the filter model incorporated into non- linear terms.

The cancellation operation may provide the cancelling signal by:

determining cancellation coefficients (W_{cancellation coe}ffi_cients based on the non-linear vector ( v );

determining the cancelling signal as the product of the non-linear vector 0)and the cancellation coefficients (W_cancellation coefficien_ts)· The cancellation operation may determine the cancellation components by:

where Cross Correlation Vector = v^H. R_x,

Autocorrelation Matrix = v . v^H

v is the nonlinear vector, H is the Hermitian operation and Rx is the received signal.

The method may be performed at a cellular base station. According to another aspect, there is provided an apparatus comprising at least one processor, at least one memory directly connected to the at least one processor, the at least one memory including computer program code, and the at least one processor, with the at least one memory and the computer program code being arranged to perform the method of any preceding method definition.

According to another aspect, there is provided a computer program product comprising a set of instructions which, when executed on an apparatus, is configured to cause the apparatus to carry out the method of any preceding method definition.

According to another aspect, there is provided a non-transitory computer readable medium comprising program instructions stored thereon for performing a method, comprising: sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband; transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation. According to another aspect, there is provided an apparatus comprising: at least one processor; and at least one memory including computer program code which, when executed by the at least one processor, causes the apparatus: to sample a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband; to transform the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and to provide at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

Drawings

Embodiments will now be described in detail with reference to the accompanying drawings, in which:

FIG. 1 is a schematic circuit diagram of a radio system comprising a duplexer and a passive intermodulation (PIM) cancellation system according to example

embodiments;

FIG. 2 is a flow diagram indicating processing operations performed in generating a filter model according to example embodiments; FIG. 3 is a flow diagram indicating processing operations performed in generating a filter model according to other example embodiments;

FIG. 4 is a graph indicative of a receive duplex filter impulse response, providing an example model as produced in accordance with example embodiments;

FIG. 5 is a flow diagram indicating processing operations performed in passive intermodulation (PIM) cancellation, using the filter model, according to example embodiments;

FIG. 6 is a graph indicative of a filter model when oversampled, in accordance with example embodiments;

FIG. 7 is a graph indicative of the frequency responses from a known passive intermodulation (PIM) cancellation operation;

FIG. 8 is a graph indicative of the frequency responses from a passive intermodulation (PIM) cancellation operation, in accordance with example embodiments;

Fig. 9 shows hardware modules according to some embodiments; and

Fig. to shows a non-volatile media according to some embodiments.

Detailed Description

Certain example embodiments are described in detail with reference to the

accompanying drawings, wherein the features of the embodiments can be freely combined with each other unless otherwise described. However, it is to be expressly understood that the description of certain embodiments is given by way of example only, and that it is by no way intended to be understood as limiting to the disclosed details. Moreover, it is to be understood that the apparatus is configured to perform the corresponding method, although in some cases only the apparatus or only the method are described. The operations of the method maybe embodied in a computer program product on, for example, a non-transitory medium. Certain abbreviations will be used herein, which are set out below for ease of reference.

Abbreviations

2G/3G/4G/5G 2^nd/3^rd/4^th/5^th Generation

3GPP 3^rd Generation Partnership Project

CD Compact Disc

CFR Crest Factor Reduction DCS Digital Cellular System

DPD Digital Predistorsion

DVD Digital Versatile Disk

eNB, NB evolved NodeB

FDD Frequency Division Duplex

IMP / IM Intermodulation Product

IMP3 / IMP5 / ... IMP of 3^rd order / 5^th order / ...

LMS Least Mean Square

MIMO Multiple Input - Multiple Output

MMS Minimum Mean Square

MSE Mean Squared Error

NL Non-linear

PIM Passive Intermodulation

RDF Receiver Duplexer Filter

RMS Recursive Mean Square

RF Radio Frequency

RX Receive

TX Transmit

UE User Equipment

UI User Interface

USB Universal Serial Bus

WiFi Wireless Fidelity

Cellular base stations may de-sense their own uplink owing to PIM products, for example introduced by passive components such as duplexers, cables, connector interfaces, antennas etc. If PIM is not mitigated, e.g. reduced or cancelled, it may not be possible to decode received signals. Operators may PIM cancellation algorithms to improve uplink signal quality. Passive InterModulation (PIM) is a natural process where transmit signals generate intermodulation products in passive devices. PIM products maybe generated at very low power levels, for example due to the aging of antennas, corroded or loose connectors and duplex filters that are passive. Imperfections of cables, combiners and attenuators may also generate PIM. PIM generation with transmit signals is generally harmless due to its low level. However, when PIM products line up with receive signals, issues can arise. Although the level of PIM in a typical radios can range from -nodBc to -150 dBc (w.r.t to the transmit signal) it can cause the receiver to desensitize. As an example, a transmit signal that is 49dBm of power causes PIM levels that are -8idBm to -loidBm. Hence, on some occasions, PIM signals can be higher than the receive signals. When PIM is higher than the receive signal, the receiver decoding process will fail due to negative signal to noise ratio. This may cause a significant throughput loss in the uplink direction (mobile to base station).

Some radios or associated equipment are designed to mitigate, i.e. reduce or avoid, such PIM effects with PIM cancellation algorithms. In this respect, the term PIM cancellation may also mean mitigate. A PIM cancellation algorithm estimates PIM by comparing, or correlating, the transmit (Tx) and the receive (Rx) signal path. The PIM cancellation algorithm may then build up a model that attempts to cancel, or at least reduce, the PIM products on the receiver (Rx) bandwidth.. Example embodiments relate to PIM cancellation which can cater for situations where one or more PIM signals are close to, span, or outside of the band edge of the receiver (Rx). In such cases, conventional PIM cancellations algorithms may be sub-optimal.

Example embodiments herein enable to estimate accurately a PIM model which is used as part of a PIM cancellation algorithm. The PIM model takes into account the impact of a radio duplexer component which may be provided in part of a telecommunication system, for example in a cellular base station or eNB, for any existing or future cellular system (e.g. 2G/3G/4G/5G). As will be known, a duplexer may be considered as a three-port filtering device which allows transmitters (Tx) and receivers (Rx) operating at different frequencies to share the same antenna. A duplexer typically comprises two bandpass filters in parallel, one providing the path between the transmitter (Tx) and the antenna, and the other providing the path between the antenna and the receiver (Rx). No path between the transmitter and receiver should exist. The group delay characteristics of a receiver (Rx) duplexer filter (RDF) may cause a non-linear spreading effect on PIM signals such that PIM signals may exist close to or outside of the band edge of the RDF. PIM corrections may degrade by as much as 3 dB, depending on the group delay characteristic, this being due to inherent characteristics of PIM signals. For example, normal radios consist of two passive duplexer filters (i.e. one for the transmitter (Tx) and one for the receiver (Rx)) that remove the outer band spectrum leakage. They both exhibit frequency attenuation just outside the band edge frequencies. Most duplex filters begin this attenuation process even 2MHz inwards from the band edge. This is to make space for the lOodB attenuation that is to be expected outside the band. Such rapid frequency responses create a group delay rise in the transition bandwidths. Although both the transmitter (Tx) and PIM signals are affected by the transmitter (Tx) filter group delay and the receiver (Rx) filter group delay respectively, the performance impact is severe with PIM signals where they are mainly impacted by the receiver (Rx) duplexer filter. The reason is that transmitter (Tx) signals are narrower when compared to the PIM signals. Just considering only the 3^rd order component of a PIM signal (neglecting the 5^th order component) a typical 5MHz transmitter (Tx) signal combination will create a non-linear spreading effect on PIM signals to cause a threefold bandwidth enhancement up to 15MHz.

Correspondingly, a 20MHz transmitter signal combination will generate a 60MHz wide PIM signal. Extending the argument to 5^th order non-linearity, the PIM signal will be fivefold larger when compared to the transmitter (Tx) signals.

With increased bandwidths observed in the PIM signals, the chances are that not all of the PIM spectrum will be contained within the receiver (Rx) band. A possible scenario is that only 20% to 70% of the PIM spectrum will be within the Rx band (i.e. 80% to 30% will be outside the Rx band). This may create an uneven delay distribution within the PIM signal. PIM spectrum portions within the receiver (Rx) bandwidth will experience benign group delay differences. However, PIM spectrum portions outside of the Rx bandwidth will experience severe group delay differences, in the order of hundreds of nanoseconds. Some PIM cancellation algorithms attempt to model the PIM spectrum with one specific delay. This delay is generally pivoted to a single delay point within the group delay distribution of the receiver (Rx) duplexer filter (RDF). A non-linear regression model will attempt to find the optimum point to accurately model PIM with: a) amplitude b) phase and a c) single suitable delay. Since the actual delay is distributive, the mathematical model will find this environment to be insurmountable, resulting in degradation in PIM cancellation performance.

FIG. 1 is a schematic diagram of an example transceiver system 10 that involves PIM cancellation according to example embodiments. The transceiver system 10 comprises a duplexer 12 connected to a common antenna 14; the duplexer 12 comprises first and second duplexer filters, a first being a transmitter (Tx) duplexer filter having a passband providing a signal path between a transmitter (Tx) 16 and the antenna, and the second being a receiver (Rx) duplexer filter providing a signal path between the antenna and a receiver (Rx) 18. No path between the transmitter 16 and the receiver 18 should exist.

A PIM cancellation module 20 may be provided between the input and output paths of the transmitter 16 and receiver 18 respectively. The PIM cancellation module 20 may be implemented in hardware, software or a combination thereof. The PIM cancellation module 20 operates using an algorithm which is similar to known PIM cancellation algorithms. For example, a filter model 29 corresponding to the receive duplexer filter may be determined in accordance with embodiments to be explained below. The net effect of the algorithm employed by the PIM cancellation module is to provide improved or optimal PIM cancellation at the received signal. FIG. 1 also indicates example spectra including a transmission band 24, transmission and receive bands 24, 26 including PIM components 28 at the duplexer 12, the receive band 26 and PIM components 28 at the receiver 18, and the“cleaned” receive band 26A after PIM cancellation by the PIM cancellation module 20.

In example embodiments, the distributive nature of the receiver duplexer filter’s group delay is negated by removing the characteristic from the PIM cancellation performed at the PIM cancellation module 20. With this, a PIM model is generated, for subtraction from the signal received at the receiver 18 (which includes PIM), the PIM modelling being restricted to amplitude, phase and a single, suitably determined delay. The PIM model takes account of the receiver duplexer filter’s characteristics to provide a set of non-linear coefficients that negate or exclude said characteristics. This is found to provide an improved, if not optimal, cancellation solution.

Two aspects will be explained below, namely, receiver duplexer filter modelling, hereafter simply“filter modelling” and PIM cancellation.

Filter Modelling

Modelling a practical receiver duplexer filter is not a rudimentary operation. At 307.2 Mega samples per second (MSPS) or at 491.52 MSPS, a typical receive duplexer impulse response spans up to 500 samples. It may be onerous to dedicate 500 complex coefficients to model a receive duplexer filter. Hence, it would be useful to model a typical duplexer filter without having to dedicate hundreds of multipliers. In examples described herein, it is shown how to model even challenging duplex filters with relatively few complex coefficients, for example as little as 30-35 complex coefficients. Modelling may take place at a reduced sampling rate of/s/n, e.g.fi /2, fs /3,fc/4 or /s/6 etc., where fs is the sampling rate of the PIM cancellation algorithm to be described later on. The reduced sampling rates discussed above are for illustrative purposes only. However, it should be appreciated that the reduced sampling rate relative to the PIM cancellation algorithm sampling rate is not essential to the broad concept.

In generating the filter model, an example method is now described with reference to the flow diagram of FIG. 2. The flow diagram shows processing operations that may be performed by a processing means, for example as shown in FIG. 9, which may or may not be at the PIM cancellation module 20, whether by means of hardware, software or a combination thereof. Additional or less operations maybe performed in some embodiments.

Broadly speaking, the processing comprises sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband, transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter, and providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

A first operation 2.1 may comprise frequency domain sampling of the receiver duplexer filter response to provide a window of samples which cover the passband of the filter and also part of the stopband, i.e. a part of the stopband adjacent the passband.

A second operation 2.2 may comprise applying an inverse discrete Fourier transform (IDFT) to the window of samples. A third operation 2.3 may comprise getting a plurality of time-domain complex coefficients as a result of the second operation 2.2, representing a filter model.

A fourth operation 2.4 may comprise performing PIM cancellation using the filter model. FIG. 3 is a more detailed flow diagram shows processing operations that may be performed by a processing means, which may or may not be at the PIM cancellation module 20, whether by means of hardware, software or a combination thereof.

Additional or less operations may be performed in some embodiments.

A first operation 3.1 may comprise frequency domain sampling of the receiver duplexer filter response to provide a window of samples which cover the passband of the filter and also part of the stopband, i.e. a part of the stopband adjacent the passband. In this example, the response may be generated by using a vector network analyser (VNA) applied to the duplexer, and sampling of the produced forward complex gain scatter parameters (S21). A sample resolution of too kHz may be used for useful results, but is not essential.

A second operation 3.2 may comprise preparing the samples obtained from the first operation 3.1 to match a selected sample rate, less than the sample rate that will be used by the PIM cancellation algorithm subsequently. For example, reduced sampling rate of/s/3,/s/4 or/s/6 etc. maybe used, where fs is the sampling rate of the PIM cancellation algorithm. We will assume a sampling rate of/s/3 and a window size of 2048 samples, for illustration. The window maybe referred to as W_s3.

A third operation 3.3 may comprise applying an ID FT to the window W_s3 to generate time-domain samples, referred to as Tap_s3.

A fourth operation 3.4 may comprise getting a plurality of time-domain complex coefficients as a result of the third operation 3.3, representing a filter model.

A fifth operation 3.5 may comprise selecting the first N complex coefficients to represent the model, as opposed to all. For example, it has been found that the first 35 complex coefficients gives a very good representation.

A sixth operation 3.6 may comprise performing PIM cancellation using the filter model.

The above-described operations for filter modelling are found to provide improvements to the subsequent PIM cancellation algorithm. However, further operations may provide further improvement, to be described now. A seventh operation 3.7 may comprise inputting noise, such as additive white Gaussian noise (N_awgn) to the complete filter model (Tap_s3) from the fourth operation 3.4 (e.g. 2048 complex coefficients) to generate an output referred to as Filt_out_s3. An eighth operation 3.8 may comprise performing linear regression to best-fit the input noise (N_awgn) to the output (Filt_out_s3) from the seventh operation 3.7 to produce, in a ninth operation 3.9, modified time domain complex coefficients representing a more optimised filter, referred to as W_opt. The process may return to the fifth operation 3.5 whereby the selected first N coefficients (which can be 30 or 35 in the current example) represent W_opt for subsequent PIM cancellation.

An alternative to the FIG. 3 example is to determine a number of complex coefficients that are needed to dedicate to a duplexer model, e.g. a frequency domain vector describing the coefficients, which vector is referred to as Wf. Another operation may comprise generating an IDFT matrix suitable for Fs/ N, which we will call IDFT_m.

Another operation maybe to build a linear equation as: (IDFT_m)^H(IDFT(S2l)) = (IDFT_m)^H(IDFT_m) * Wf (1)

Another operation may be to solve the above linear equation to maximise the number of S21 points, and then taking the IFFT of Wf to generate time domain complex coefficients Wt.

FIG. 4 is a graph showing in this case the first 30 time domain complex coefficients, which may model the impulse response of the receiver duplexer filter for subsequent PIM cancellation.

Referring to FIG. 5, further operations maybe performed depending on the filter modelling employed.

Taking the example mentioned above with reference to FIG. 4, a duplexer model W_opt is obtained at a sampling rate oifs/N, with N being 3 in the example. However, PIM cancellation generally takes place at a higher sampling rate (i.e./s) than the low sampled duplexer filter model. Hence, to match the rate of cancellation, the low rate duplexer model may be oversampled in a first operation 5.1, or, broadly speaking, rate matching the filter model to the correction sample rate. If the model is obtained using fs/3, then the oversampled model at fs can be generated by inserting two zeros after every complex coefficient (as in [coef₀ o o coefi o o... etc]).

As with any oversampling technique, oversampling the W_opt coefficients by three times will create three frequency duplicates within the spectrum offs. This is shown in FIG. 6.

A further, optional operation 5.2 may comprise applying a mask filter. More specifically, in some embodiments, PIM cancellation techniques are related to operations on duplicate spectrums, and hence the duplicate spectrums of the duplexer may be removed by a mask filter, or anti-aliasing filter. With mask filtering, all spectrum duplicates will be removed, thus equating it to the actual bandwidth of the receiver duplexer filter. Without the mask filter, all frequency duplicates will be visible in model extraction. PIM cancellation without the mask filter can be pursued as a less complex solution, for the reason that the duplicate frequencies outside the bandwidth of interest may absorb interference into the model domain. A further operation 5.3 may comprise performing PIM cancellation using the filter model.

PIM Cancellation

Linear regression yields sub optimum results due to the fact that delay is distributive when the PIM signal is located at receiver band edge. In addition to the delay the duplexer introduces a non-flat magnitude response on the PIM signal. To overcome these deformities, the complex receive duplexer filter model described above, is incorporated into the linear regression solution. It can be described by an equation as shown in (2). Note that the above-mentioned mask filter is ignored in this equation (2), as is any memory of the non-linear terms, which are ignored for simplicity. An actual implementation may comprise memory taps.

where x is a reference signal used to compute the PIM model.

Note that each of the non-linear terms are filtered by the rate matched duplexer model f duplex In this way, the non-linear terms do not have to care for the magnitude and the distributive delay of the Rx duplexer. Hence, the PIM cancellation model is free to model the PIM only, giving an improved result.

For even more improved PIM cancellation, the above-mentioned mask filter may be included on top of t e f _duplex filter model to remove any duplicate spectrum. The cancellation technique, including the mask filter, can be illustrated as in equation (3). No transmission leakage or outer band noise will interfere in this case.

In all PIM cancellation techniques, described in equations (2) and (3) above, PIM model coefficients are extracted by formulating the non-linear vector to an auto correlation matrix as well as a cross correlation vector. Equation (4) shows this this formulation:

Aauto * W_Canceiiation_coefs = Cross Correlation (4)

Aauto = V * V^H,

Cross Correlation = v^H * Rx where v is the non-linear vector, H is the Hermitian operation, and Rx is received signal that includes both PIM and the actual receiver signal.

PIM will be removed from the Rx path as shown in equation (5).

Clean Rx signal = Rx - v * W_{Cancellationcoefs} (5).

In simulations and tests, the above systems and methods demonstrated improved PIM cancellation by at least 3 - 5 dB. The following table indicates measured gains in an actual radio product. Note also the significant gains observed when the receive duplexer model was increased from 30 to 35 coefficients.

PiM [dBc] -130

Table 1: Measured Results of PIM Cancellation

Observations from table 1 include that a complex filter model gives better performance over a real filter, that the use of 35 coefficients gives better performance over 30 coefficients, in both cases, especially at the band edges.

FIG. 9 shows an apparatus according to an embodiment. The apparatus maybe configured to perform the operations described herein, for example operations described with reference to any of FIGS. 2, 3 and 5. The apparatus comprises at least one processor 420 and at least one memory 410 directly or closely connected to the processor. The memory 410 includes at least one random access memory (RAM) 410b and at least one read-only memory (ROM) 410a. Computer program code (software) 415 is stored in the ROM 410a. The apparatus may be connected to a TX path and a RX path of a base station in order to obtain respective signals. However, in some embodiments, the TX signals and RX signals are input as data streams into the apparatus. The apparatus may be connected with a user interface UI for instructing the apparatus and/or for outputting results. However, instead of by a UI, the instructions maybe input e.g. from a batch file, and the output maybe stored in a non-volatile memory. The at least one processor 420, with the at least one memory 410 and the computer program code 415 are arranged to cause the apparatus to at least perform at least the method according to FIGS. 2, 3 and 5

Fig. 10 shows a non-transitory media 130 according to some embodiments. The non- transitory media 130 is a computer readable storage medium. It may be e.g. a CD, a DVD, a USB stick, a blue ray disk, etc. The non-transitory media 130 stores computer program code causing an apparatus to perform the method of any of FIGS. 2, 3 and 5, when executed by a processor such as processor 420 of FIG. 9. Names of network elements, protocols, and methods are based on current standards. In other versions or other technologies, the names of these network elements and/ or protocols and/ or methods may be different, as long as they provide a corresponding functionality. For example, embodiments maybe deployed in 2G/3G/4G/5G networks and further generations of 3GPP but also in non-3GPP radio networks such as WiFi. Accordingly, a base station may be a BTS, a NodeB, a eNodeB, a WiFi access point etc.

A memory may be volatile or non-volatile. It may be e.g. a RAM, a sram, a flash memory, a FPGA block ram, a DCD, a CD, a USB stick, and a blue ray disk.

If not otherwise stated or otherwise made clear from the context, the statement that two entities are different means that they perform different functions. It does not necessarily mean that they are based on different hardware. That is, each of the entities described in the present description may be based on a different hardware, or some or all of the entities may be based on the same hardware. It does not necessarily mean that they are based on different software. That is, each of the entities described in the present description maybe based on different software, or some or all of the entities may be based on the same software. Each of the entities described in the present description may be embodied in the cloud.

According to the above description, it should thus be apparent that example

embodiments provide, for example, a delay estimation device for PIM cancellation, or a component thereof, an apparatus embodying the same, a method for controlling and/or operating the same, and computer program(s) controlling and/or operating the same as well as mediums carrying such computer program(s) and forming computer program product(s). Such a delay estimation device for PIM cancellation maybe incorporated e.g. in a Nokia Airframe expandable base station.

Implementations of any of the above described blocks, apparatuses, systems, techniques or methods include, as non-limiting examples, implementations as hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof. Some embodiments may be implemented in the cloud. It is to be understood that what is described above is what is presently considered the preferred embodiments. However, it should be noted that the description of the preferred embodiments is given by way of example only and that various modifications may be made without departing from the scope as defined by the appended claims.

Claims

1. Apparatus, comprising:

means for sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband;

means for transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and

means for providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

2. The apparatus of claim l, wherein the means for sampling the frequency domain response is configured to sample S21 forward complex gain parameters produced by the filter.

3. The apparatus of claim 1 or claim 2, wherein the means for sampling the frequency domain response is configured to sample at a resolution of substantially too kHz or less. 4. The apparatus of any preceding claim, further comprising means for modifying the window of frequency domain response samples, prior to transforming, to have a sample frequency/c, such that: fc =fs/n, where fs is the sample frequency of a subsequent passive intermodulation cancellation means, and wherein n = 2, 3,

4 etc.

5. The apparatus of claim 4, further comprising means for over-sampling the filter model after transformation to match the sample frequency (fs) of the subsequent passive intermodulation cancellation means.

6. The apparatus of any preceding claim, wherein the means for transforming the samples comprises means for taking the inverse discrete Fourier transform (IDFT) of the frequency domain samples.

7. The apparatus of any preceding claim, further comprising means for inputting noise to the provided filter model and means for performing linear regression to fit the input noise to the output of the filter model to produce an updated filter model comprising updated time domain complex coefficients for replacing the provided filter model and for use in subsequent passive intermodulation cancellation.

8. The apparatus of claim 7, wherein the noise is additive white Gaussian noise (AWGN).

9. The apparatus of any preceding claim, wherein the means for providing the filter model is configured to use only a subset of the time domain complex coefficients for the model.

10. The apparatus of claim 9, wherein the subset of time domain complex coefficients comprises substantially 35 coefficients or fewer.

11. The apparatus of any preceding claim, further comprising means for performing passive intermodulation cancellation on a received signal based on the rate-matched filter model incorporated into non-linear terms.

12. The apparatus of claim 11, further comprising means for filtering the derived PIM model prior to performing passive intermodulation to reject duplicate frequency components outside of the filter passband.

13. The apparatus of claim 11 or claim 12, wherein the cancellation means is configured to remove, at a sample frequency (fs), passive intermodulation components from the received signal by subtracting from the received signal a cancelling signal determined using the filter model incorporated into non-linear terms.

14. The apparatus of claim 13, wherein the cancellation means provides the cancelling signal by:

determining cancellation coefficients (W_cancellation coefficients) based on the non-linear vector ( v ) ; determining the cancelling signal as the product of the non-linear vector 0)and the cancellation coefficients (W_{cancellation coe}ffi_cients)·

15. The apparatus of claim 14, wherein the cancellation means determines the cancellation components by:

^_{cancellation coe}ffi_cients ⁼ inverse(AutoCorr elation Matrix) * Cross Correlation Vector where Cross Correlation Vector = v^H. R_x,

Autocorrelation Matrix = v . v^H

16. The apparatus of any preceding claim, provided in a cellular base station.

17. A method, comprising:

sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband;

transforming the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and

providing at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.

18. The method of claim 17, wherein sampling the frequency domain response comprises sampling S21 forward complex gain parameters produced by the filter.

19. The method of claim 17 or claim 18, wherein sampling the frequency domain response is at a resolution of substantially too kHz or less.

20. The method of any of claims 17 to 19, further comprising modifying the window of frequency domain response samples, prior to transforming, to have a sample frequency fc, such that: fc =fs/n, where fs is the sample frequency of a subsequent passive intermodulation cancellation operation, and wherein n = 2, 3, 4 etc.

21. The method of claim 20, further comprising over-sampling the filter model after transformation to match the sample frequency (fe ) of the subsequent passive intermodulation cancellation operation.

22. The method of any of claims 17 to 21, wherein transforming the samples comprises taking the inverse discrete Fourier transform (IDFT) of the frequency domain samples.

23. The method of any of claims 17 to 22, further comprising inputting noise to the provided filter model and performing linear regression to fit the input noise to the output of the filter model to produce an updated filter model comprising updated time domain complex coefficients for replacing the provided filter model and for use in a subsequent passive intermodulation cancellation operation.

24. The method of claim 23, wherein the noise is additive white Gaussian noise (AWGN).

25. The method of any of claims 17 to 24, wherein providing the filter model uses only a subset of the time domain complex coefficients for the model.

26. The method of claim 25, wherein the subset of time domain complex

coefficients comprises substantially 35 coefficients or fewer.

27. The method of any of claims 17 to 26, further comprising performing passive intermodulation cancellation on a received signal based on the rate-matched filter model incorporated into non-linear terms.

28. The method of claim 27, further comprising filtering the derived PIM model prior to performing passive intermodulation to reject duplicate frequency components outside of the filter passband.

29. The method of claim 27 or claim 28, wherein the cancellation operation removes, at a sample frequency (fs), passive intermodulation components from the received signal by subtracting from the received signal a cancelling signal determined using the filter model incorporated into non-linear terms.

30. The method of claim 29, wherein the cancellation operation provides the

cancelling signal by:

determining the cancelling signal as the product of the non-linear vector

0)and the cancellation coefficients (W_cancellation coefficients)·

31. The method of claim 30, wherein the cancellation operation determines the cancellation components by:

where Cross Correlation Vector = v^H. R_x,

Autocorrelation Matrix = v . v^H

32. The method of any of claims 17 to 31, performed at a cellular base station.

33. An apparatus comprising at least one processor, at least one memory directly connected to the at least one processor, the at least one memory including computer program code, and the at least one processor, with the at least one memory and the computer program code being arranged to perform the method of any of claims 17 to

32.

34. A computer program product comprising a set of instructions which, when

executed on an apparatus, is configured to cause the apparatus to carry out the method of any of claims 17 to 32.

35. A non-transitory computer readable medium comprising program instructions stored thereon for performing a method, comprising: sampling a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband;

36. An apparatus comprising: at least one processor; and at least one memory including computer program code which, when executed by the at least one processor, causes the apparatus:

to sample a frequency domain response of a receiver duplex filter to provide a window of samples inside and outside of the filter passband;

to transform the frequency samples into a plurality of time domain complex coefficients, representing an impulse response of the filter; and

to provide at least some of the time domain complex coefficients as a filter model representing the filter for a subsequent means for passive intermodulation cancellation.