METHOD AND SYSTEM FOR LOCATING

A PIM FAULT IN A PHASED ARRAY ANTENNA

TECHNICAL FIELD

[0001] The present invention relates to the identification of a passive intermodulation (PIM) fault in a phased array antenna. In particular, although not exclusively, the invention relates to the determination of the location of a PIM fault in a phased array antenna under test.

BACKGROUND ART

[0002] Phased array antennas are important building blocks in many radio communications systems. By coherently receiving or transmitting signals from multiple radiating elements at the same time, this type of antenna can achieve a much more directional radiation pattern than would be possible with one radiating element alone.

[0003] The internal construction of phased array antennas can be very complex, especially when the antenna contains a large number of radiating elements. Modern antennas can contain literally hundreds of interconnects, cables and components. For these reasons, phased array antennas can be challenging to manufacture, and even more difficult to troubleshoot when faults occur.

[0004] One of the faults that can arise in phased array antennas is passive intermodulation distortion (PIM), a common source of interference in radio communications systems. PIM problems typically occur in cases where high-power transmit signals and low-power receive signals propagate through shared radiofrequency (RF) infrastructure. In the event that the infrastructure contains a component or interconnect with a nonlinear transfer characteristic, the transmit signals can mix together in the nonlinearity and produce spurious PIM products across a wide range of frequencies. If a PIM product happens to land in the receive band of the base station, the receiver can become desensitized, leading to dropped calls and reduced data rates. Common sources of PIM include loose connectors, cracked solder joints, damaged plating, swarf, dirt and other contaminants in the RF path.

[0005] Much of the difficulty in pinpointing PIM faults in phased array antennas stems from the branched architecture employed by this type of antenna. Traditional fault-finding techniques like Distance-to-PIM testing can determine the distance from the antenna’s RF port to the PIM fault, but cannot determine which branch the fault resides in. As a result, manufacturing staff often have no choice but to manually inspect and re-work every component and interconnect in every branch of the antenna. This process can be labour-intensive, slow and inefficient, and is typically ill-suited to a volume manufacturing environment, where cycle time is a critical parameter.

[0006] Another difficulty in pinpointing PIM faults is that PIM faults may not necessarily manifest in a consistent manner. As an illustrative example, some PIM faults may manifest only under certain conditions, e.g. when moved or when force is applied to a part of the antenna.

[0007] As such, there is clearly a need for improved identification of PIM faults in a phased array antenna.

[0008] It will be clearly understood that, if a prior art publication is referred to herein, this reference does not constitute an admission that the publication forms part of the common general knowledge in the art in Australia or in any other country.

SUMMARY OF INVENTION

[0009] The present invention is directed to methods and systems for locating PIM faults in a phased array antenna, which may at least partially overcome at least one of the abovementioned disadvantages or provide the consumer with a useful or commercial choice.

[0010] With the foregoing in view, the present invention in one form, resides broadly in a system for locating a passive intermodulation (PIM) fault in an antenna under test (AUT), the AUT comprising an input port and a plurality of branches, each branch ending in at least one radiating element, the system comprising: a fixed array of over-the-air probes, each over-the- air probe configured to receive PIM signals emitted from at least one of the plurality of radiating elements; a PIM analyser, connected to the AUT and the fixed array of over-the-air probes, for: applying a plurality of RF stimulus signals to the input port of the AUT; and measuring one or more characteristics of the Forward PIM signals received by the fixed array of over-the-air probes to locate the PIM fault in the AUT.

[001 1 ] According to another embodiment of the invention, there is provided a method of locating a passive intermodulation (PIM) fault in an antenna under test (AUT), the AUT comprising an input port and a plurality of branches, each branch ending in at least one radiating element, the method comprising: applying a plurality of RF stimulus signals to the input port of the AUT; receiving, by a fixed array of over-the-air probes, the Forward PIM signals emitted from at least one of the plurality of radiating elements; measuring, by a PIM analyser, one or more characteristics of the Forward PIM signals received by the fixed array of over-the-air probes; and analysing the one or more characteristics of the Forward PIM signals to locate the PIM fault in the AUT. [0012] Advantageously, the system and method enable PIM faults to be easily located in an AUT, which is particularly useful when manufacturing phased array antennas.

[0013] The antenna under test may be a phased array antenna, an omni-band antenna or a small cell antenna. The antenna under test may be situated in an anechoic chamber. The antenna under test may include a power divider, which splits an input signal into multiple branches. Each branch may terminate in at least one radiating element.

[0014] Locating a PIM fault may comprise determining which individual branch or set of branches of the antenna under test contains the PIM fault.

[0015] Each over-the-air probes may be configured to receive signals from two or more of the radiating elements.

[0016] Beam forming may be used to isolate signals from each of the radiating elements of the AUT received at the over-the-air probes.

[0017] The OTA probes may be positioned at least 1 m from (e.g. above) the radiating elements. The OTA probes may be at least about 50cm from the radiating elements. Two or more of the OTA probes may be at least about 50cm from the radiating elements.

[0018] The OTA probes may be arranged in a curved arrangement. The OTA probes may be arranged in a curved arrangement in a direction perpendicular to the AUT (e.g. in the vertical direction when tested from above). The OTA probes may be arranged such that they are furthest from the radiating elements at a centre of the antenna, and closest to the radiating elements at edges of the antenna. The OTA probes may form a dome-shape above the antenna.

[0019] The OTA probes may be positioned to lay on an elliptical curve in the vertical direction. The OTA probes may be equidistantly placed on the elliptical curve.

[0020] The elliptical curve may be defined according to the following parameterisation:

where f is from 0 to TT, W is the width of the curve, h is the height of the upper most point of the curve above the AUT, and b is the height of the curve from vertically, from its upper most point to its lowermost point.

[0021 ] The OTA probes may be linearly arranged. The OTA probes may be

approximately parallel to the radiating elements. [0022] The OTA probes may comprise less than 30 probes. The OTA probes may comprise about 16 probes. The skilled addressee will, however, readily appreciate that as many probes as are needed to separate the radiating elements.

[0023] The OTA probes may be configured such that amplitude response functions of the elements of the antenna have at least about 10dB target isolation.

[0024] Alternatively, each over-the-air probes may be configured to receive signals from a single radiating element only. The probes may be configured to achieve a strong coupling with the radiating elements of the AUT. Each of the radiating elements of the antenna under test may be coupled to one of the plurality of over-the-air (OTA) probes. The angular orientation of the of the probe may be adjusted to achieve a strong coupling.

[0025] Each of the OTA probes may be centred over its counterpart radiating element in the AUT. The probes may be positioned within about 30mm of the radiating elements of the AUT.

[0026] The OTA probes may be mounted on a gantry which is height-adjustable to allow the vertical separation between the OTA probes and the radiating elements to be configurable. The vertical separation between each OTA probe and its corresponding radiating element may be less than 100mm. The vertical separation may be in the range 10-100mm.

[0027] The system may include a controller, connected to the PIM analyser, for analysing the one or more characteristics of the Forward PIM signals to locate the PIM fault in the AUT.

[0028] The at least one of the RF stimulus signals may be swept over a range of frequencies.

[0029] Preferably, the PIM analyser may be configured to measure the one or more characteristics of the Forward PIM signals to identify static PIM faults and dynamic PIM faults.

[0030] The AUT may be selectively vibrated to identify dynamic PIM faults.

[0031] The system may include a vibration table, for vibrating the AUT to identify dynamic

PIM faults.

[0032] A controller may be configured to activate vibration of the vibration table during first measurement of the Forward PIM signals, and not activate vibration of the vibration table during second measurement of the Forward PIM signals.

[0033] The vibration table may be provided within the anechoic chamber. One or more actuators may be configured to provide vibration to the vibration table. The actuators may be provided in the anechoic chamber. The actuators may be screened for PIM. The actuators may be screened to limit conducted and radiated emissions.

[0034] The radiated screening may be performed by enclosing the actuator in an aluminium housing. The conducted screening may be performed by one or more low pass filters in line with power cables of the actuators. For example, each actuator may include a low pass filter on each of its positive and negative power cables.

[0035] The vibration table may include a support surface, for supporting the AUT. The actuators may be configured to vibrate the support surface directly.

[0036] The actuators may be configured to vibrate the table at a resonance frequency of the table and an AUT. The actuators may be configured to magnify the force applied to the AUT at a discrete set of resonance frequencies.

[0037] The vibration table may include one or more sensors to measure a vibration of the table. Such sensors may be used to provide closed loop control of the vibration of the vibration table and AUT together with the actuators.

[0038] The apparatus may comprise an RF switchbox. The RF switchbox may be a solid- state device built from electronic switches. The RF switchbox may be constructed from a network of electromechanical relays.

[0039] The apparatus may comprise a bandpass filter configured between a common port of the RF switchbox and a port of a PIM analyser. The bandpass filter may pass frequencies in the receive band of the PIM analyser and reject frequencies in the transmit band.

[0040] The PIM analyser may be a single-receiver PIM analyser capable of measuring both Forward and Reverse PIM across a range of frequencies. The PIM analyser may be a multi-receiver PIM analyser capable of measuring both Forward and Reverse PIM across a range of frequencies. The multi-receiver PIM analyser may be capable of measuring the Forward PIM emitted from all over-the-air probes simultaneously. The multi- receiver PIM analyser may include an internal diplexer bank containing a plurality of receive diplexers, one for each of the plurality of over-the-air probes. The PIM analyser may contain a transmit module. The transmit module may adjust the frequency and amplitude of the carrier signals to perform a swept frequency PIM measurement.

[0041] Forward PIM may be measured across a range of frequencies within the passband of the AUT. [0042] A Range-to-Fault (RTF) module may be used to determine the distance to the PIM fault on a branch of the AUT. The Range-to-Fault module may measure Reverse PIM. Reverse PIM may be measured across a range of frequencies within the passband of the AUT.

[0043] The PIM analyser may be configured to measure the phase characteristics of Forward PIM signals received by the fixed array of over-the-air probes. The PIM analyser may include a coherent receiver bank to measure the phase characteristics of Forward PIM signals received by the fixed array of over-the-air probes.

[0044] The system may be configured to locate the PIM fault in the AUT using rank ordering to classify received forward PIM signals into fault classes, each class corresponding to at least one potential failure point. Detection of a fault condition comprises finding which class the data corresponds to, which leads to the position (location) of the fault.

[0045] Training data may be used to locate the PIM fault in the AUT, the training data comprising data corresponding to a PIM source in each of a plurality of possible fault positions and a corresponding signal received of the array of over-the-air probes.

[0046] The training data may be used in a method similar to Massive MIMO, wherein spatial separation of PIM sources is provided in a similar manner to separation of mobile phone users in Massive MIMO, wherein the training data provides analogous information to the uplink pilots in Massive MIMO.

[0047] Locating the PIM fault may include machine learning to map received PIM signals. In such case, a mapping is built from collected data to map data from the many sources to classes of possible failure.

[0048] The method may include analysing the characteristics of the PIM signals to determine the branch of one or more PIM faults. The characteristics of the PIM signals may include amplitude, or phase or amplitude and phase. The method may include measuring the group delay response of the PIM signals emitted by the radiating elements of the AUT. The group delay is defined as the gradient of the measured PIM phase response with respect to frequency.

[0049] The method may further comprise the prior step of identifying PIM faults on the input port of the AUT by performing a distance to PIM (DTP) test. The method may further comprise repairing a PIM fault on the input port of the AUT before measuring the Forward PIM response of the AUT.

[0050] The method may further comprise repeating the steps of receiving and measuring the characteristics of the PIM signals a plurality of times, adding a low-PIM mismatch to a different branch of the AUT at each iteration. The at least one of the RF stimulus signals may be swept over a range of frequencies.

[0051] The method may further comprise setting the AUT’s down-tilt to a preconfigured value. The method may further comprise introducing an impedance mismatch into one or more branches of the AUT and observing the change in PIM response. The method may further comprise measuring the scalar Forward PIM response of each branch and using heuristic techniques to infer the location of the PIM fault.

[0052] A PIM sweep may be performed comprising a stepped-frequency sweep, in which PIM is measured at a set of discrete frequency points within the receive band of the PIM analyser. The PIM sweep may be performed repetitively at multiple down-tilt settings.

[0053] Measuring the characteristics of the PIM signals received by the fixed array of over-the air probes may comprise measuring PIM phase as well as amplitude. Analysing the characteristics of the PIM signals may comprise calculating the PIM group delay. Analysing the characteristics of the PIM signals may comprise fitting a parametric model to the

measurements of the PIM signals using an L1 -norm minimization procedure.

[0054] According to another embodiment of the invention, there is provided a system for testing an antenna under test (AUT), the AUT comprising an input port and a plurality of branches, each branch ending in at least one radiating element, the system comprising: an input for receiving a plurality of RF stimulus signals and providing the plurality of RF stimulus signals to the input port of the AUT; a fixed array of over-the-air probes, for receiving Forward PIM signals emitted from at least one of the plurality of radiating elements; and an output for providing the Forward PIM signals emitted from at least one of the plurality of radiating elements. The at least one of the plurality of RF stimulus signals may be swept over a range of frequencies.

[0055] According to another embodiment of the invention, there is provided a vibration table for inducing dynamic passive intermodulation (PIM) faults in an antenna under test (AUT), the vibration table including one or more actuators, coupled to a support surface of the vibration table, and configured to selectively vibrate support surface, and thereby an AUT thereon to induce dynamic passive intermodulation (PIM) faults in the AUT, wherein the actuators are RF screened.

[0056] Suitably, the screening includes radiated screening. Suitably, the screening includes conducted screening. [0057] The radiating screening may comprise one or more housings, for housing the one or more actuators.

[0058] The conducted screening may comprise a low pass filter.

[0059] The actuators may be configured to vibrate the support surface and AUT at one or more resonance frequencies.

[0060] Advantageously, such use of actuators may be simpler and less expensive than pneumatic actuators.

[0061] The vibration table may include one or more sensors to measure a vibration of the table. Such sensors may be used to provide closed loop control of the vibration of the vibration table and AUT together with the actuators.

[0062] Any of the features described herein can be combined in any combination with any one or more of the other features described herein within the scope of the invention.

[0063] The reference to any prior art in this specification is not, and should not be taken as an acknowledgement or any form of suggestion that the prior art forms part of the common general knowledge.

BRIEF DESCRIPTION OF DRAWINGS

[0064] Various embodiments of the invention will be described with reference to the following drawings, in which:

[0065] Figure 1 is a block diagram of a phased array antenna with one power divider, in accordance with the prior art;

[0066] Figure 2 is a block diagram of a phased array antenna with 3 power dividers, in accordance with the prior art;

[0067] Figure 3 is a block diagram of an embodiment of an aspect of the invention, which includes an electronic RF switchbox, diplexer bank and Rx bandpass filter;

[0068] Figure 4 is a block diagram of an embodiment of an aspect of the invention, which includes a relay-based RF switchbox;

[0069] Figure 5 is a block diagram of an embodiment of an aspect of the invention, which includes a PIM analyser with one transmitter and multiple receivers; [0070] Figure 6 is a flowchart depicting the key steps in fault-finding Method 1 ;

[0071 ] Figure 7 is a graph depicting the simulation of PIM emitted by radiating elements in a 7-branch phased array antenna during a Forward PIM measurement sweep using Method 1 ;

[0072] Figure 8 is a graph depicting example simulated results from Method 1 , in which the PIM fault is located 20mm from a splitter port;

[0073] Figure 9 is a graph depicting example simulated results from Method 1 , in which the PIM fault is located 71 mm from a splitter port;

[0074] Figure 10 is a flowchart depicting the key steps in fault-finding Method 2;

[0075] Figure 1 1 is a graph depicting simulated results obtained using Method 2;

[0076] Figure 12 is a flowchart depicting the key steps in fault-finding Method 3;

[0077] Figure 13 is a graph depicting simulated results obtained using Method 4;

[0078] Figure 14 is a set of graphs depicting simulated sweep data from Method 5;

[0079] Figure 15 is a graph depicting simulated results of PIM fault-finding analysis on a branch antenna using Method 5;

[0080] Figure 16 is a block diagram of a phased array antenna with two radiating elements per branch, in accordance with the prior art;

[0081 ] Figure 17 is a block diagram of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry.

[0082] Figure 18 illustrates an exemplary plot of an amplitude response function of near field focusing for a single element;

[0083] Figure 19a illustrates a flat OTA probe arrangement of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry;

[0084] Figure 19b illustrates amplitude response functions of near field focusing for each of the elements of the arrangement of Figure 19a.

[0085] Figure 20a illustrates a curved OTA probe arrangement of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry;

[0086] Figure 20b illustrates amplitude response functions of near field focusing for each of the elements of the arrangement of Figure 20a.

[0087] Figure 21 a illustrates a curved OTA probe arrangement of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry;

[0088] Figure 21 b illustrates amplitude response functions of near field focusing for each of the elements of the arrangement of Figure 21 a.

[0089] Figure 22a illustrates a further curved OTA probe arrangement of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry;

[0090] Figure 22b illustrates amplitude response functions of near field focusing for each of the elements of the arrangement of Figure 22a.

[0091] Figure 23a illustrates yet a further curved OTA probe arrangement of a system of an embodiment of an aspect of the invention, which includes a near field focusing gantry;

[0092] Figure 23b illustrates amplitude response functions of near field focusing for each of the elements of the arrangement of Figure 23a.

[0093] Figure 24 illustrates an example system including a single PIM source;

[0094] Figure 25 illustrates a plot of a simulation of the example of Figure 24 in a first configuration; and

[0095] Figure 26 illustrates a plot of a simulation of the example of Figure 24 in a second configuration.

[0096] Preferred features, embodiments and variations of the invention may be discerned from the following Detailed Description which provides sufficient information for those skilled in the art to perform the invention. The Detailed Description is not to be regarded as limiting the scope of the preceding Summary of the Invention in any way.

DESCRIPTION OF EMBODIMENTS

[0097] A phased array antenna is one in which two or more radiating elements are connected together in such a way as to produce an antenna with a more directional radiation pattern than would be possible with one of the radiating elements alone.

[0098] Phased arrays may be configured according to a variety of designs, from low-cost assemblies made entirely of passive RF components, to advanced systems in which each radiating element is driven by a dedicated digital transceiver. Radiating elements can be arranged in a single row, forming a so-called linear array, or into a two-dimensional grid known as a planar array. In some phased array antennas, the radiation pattern can be steered in different directions by adjusting the relative phasing of the signals travelling to/from each radiating element.

[0099] One class of phased array antenna that is of relevance to aspects of the current invention is depicted in the schematic diagram in Figure 1 . In this design of a phased array antenna 100, the antenna’s RF input port 102 is connected to a power divider (also known as a “splitter”) 104, which splits the signal into multiple branches comprising carefully chosen lengths of coaxial cable or waveguide. Each branch terminates in its own radiating element 106a-n. Adjustable phase shifters 108a-n are usually included in each branch in order to steer the radiation pattern in the desired direction.

[00100] In alternative antenna designs the power divider may comprise a number of separate stages, as shown in Figure 2. This is often done for reasons of mechanical convenience. Functionally, however, the antenna in Figure 2 is substantially similar to that of Figure 1 .

[00101 ] In yet another antenna design 1600, as shown in Figure 16, the antenna may contain multiple radiating elements per branch. In this arrangement, each branch is terminated by a small splitter module (for example, 1602), wherein the splitter module’s common port (for example, 1604) is connected to the end of the branch, and the splitter module’s output ports (for example, 1606a and 1606b) are connected to the radiating elements (for example, 1608a and 1608b). A possible antenna configuration may use two radiating elements per branch, although more than two elements per branch may be used. Despite the additional complexity of this configuration compared to the antenna configurations depicted in Figures 1 and 2, the antenna 1600 depicted in Figure 16 is functionally similar to the antennas depicted in Figure 1 and Figure 2.

[00102] Returning to Figure 1 , the power divider 104 may be designed to deliver less power to radiating elements at the ends of the array than to the elements in the middle of the array, a strategy referred to as“tapering” or“windowing”. By an appropriate choice of tapering function, it may be possible to trade off the antenna’s beamwidth against its peak sidelobe level.

[00103] The phased array antennas shown in Figure 1 , Figure 2 and Figure 16 may also incorporate one or more RF filters into their assemblies in order to reject noise and unwanted interference from other radio transmitters.

[00104] In some applications several independent phased arrays, each having a different operating frequency range and/or polarization to the others, may be installed into the same enclosure, often with their radiating elements interleaved with each other to minimize the overall size of the unit.

[00105] The phased array antenna topology in Figure 1 (and related variants such as those shown in Figure 2 and Figure 16) may be used in the cellular mobile telecommunications industry. An application for phased array antenna topologies, as shown in Figure 1 , Figure 2 and Figure 16, is in sector antennas, whose purpose is to restrict the radio coverage of a base station to a small number (typically three) of limited geographical areas known as“sectors”. This type of antenna design may provide the following advantages in this application.

[00106] Firstly, this phased array antenna topology may lend itself to a cheap and lightweight construction, which may utilise mechanically simple components. Secondly, it may be capable of operation at high RF power levels without damage. Finally, if assembled carefully using good quality materials, it may generate very low levels of PIM.

[00107] The disadvantages of the type of antenna shown in Figure 1 , Figure 2 and Figure 16 is that it may have poor branch-to-branch isolation and low output return loss. This undesirable behaviour is predominantly caused by the power divider module, which is usually based on a reactive power divider design that integrates both the splitting and phase shifting functionality into a single device of relatively low cost and complexity. Unfortunately, this convenience comes at the expense of isolation and output return loss performance. For reasons to be discussed subsequently, these characteristics can significantly increase the difficulty of troubleshooting PIM faults in this type of antenna.

[00108] The description, herein, describes phased array antennas of the type used in cellular telecommunications networks; however, it shall be understood that embodiments of the invention may be used with phased array antennas that share the following architectural features in common with Figure 1 , Figure 2 and Figure 16: a single RF port; a power divider network that splits the feed network from the RF port into two or more branches; an array of radiating elements; and a set of transmission lines connecting the output ports of the power divider to the radiating elements. Note that these architectural features may also apply to pseudo- omnidirectional antennas, which are not strictly phased array antennas, but which nevertheless share a similar architecture to the classical phased array antenna topology depicted in Figure 1 .

[00109] PIM Response of Phased Array Antennas

[001 10] Without being limited by theory, antennas are reciprocal devices, meaning that their radiation patterns are the same regardless of whether they are transmitting or receiving. In a cellular base station, antenna reciprocity allows the same antenna to be used for both transmitting and receiving, resulting in a simpler, cheaper hardware configuration than one which employs dedicated transmit and receive antennas; however, one disadvantage of this approach is that it may result in downlink and uplink signals sharing a common RF path. In the event that the antenna develops a PIM fault, the high-power downlink signals may generate a PIM signal that falls in the uplink band of the system, which then propagates unattenuated back into the base station’s receiver.

[001 1 1 ] This type of fault can be very difficult to troubleshoot. Modern phased array antennas may contain 16 branches, each comprising multiple cables, components and interconnects, any of which could potentially develop a PIM fault. To complicate matters, the branch-to-branch isolation in many phased array antennas can be extremely low, sometimes as little as 2-3 dB. As a result, when a PIM signal is generated in one branch, it can leak strongly into neighbouring branches, so much so that the amplitudes of the leaked signals can be comparable to that of the original. In such cases, if the amount of PIM emitted by each radiating element in the antenna was measured, it may give the impression that the antenna contains PIM faults in multiple branches, when actually most of the PIM seen is simply leakage from a single faulty branch.

[001 12] Exacerbating the confusion caused by low interbranch isolation is the poor output return loss of the power dividers used in many phased array antennas. Output return losses of 1-2 dB are not uncommon in many commercial antennas. This can be problematic due to the way in which PIM signals are generated, namely, that when a PIM signal is generated in a faulty branch, half of the signal travels in the forward direction, towards the radiating element, while the other half of the signal travels back towards the splitter. If the splitter has low output return loss, most of the backward-travelling PIM signal reflects off the splitter and travels back towards the radiating element, where it adds coherently with the forward-travelling component of the PIM signal. Depending on the location of the PIM fault in the branch and the phase of the splitter’s output reflection coefficient, the two PIM signals can interfere destructively at the input to the radiating element. As a result, it is entirely possible for a faulty branch to emit less PIM than its neighbours, despite the fact that the neighbouring branches are simply emitting PIM that has leaked across from the faulty branch.

[001 13] Yet another complication that can hamper the PIM troubleshooting process is that the PIM response of a phased array antenna can vary depending on the settings of its phase shifters (which in a cellular antenna are collectively referred to as the“down-tilt” setting of the antenna). It is not unusual for an antenna to meet the manufacturer’s PIM specification at one down-tilt setting, but fail it at another; however, this does not necessarily mean that the phase shifters are faulty. It can also be the result of signals from multiple PIM faults, or even the forwardtravelling and backward-travelling components of a single PIM fault, following different pathways through the antenna assembly before coming together at a common output port, like the antenna’s RF port or one of its radiating elements. In such cases, changing the settings of the phase shifters can alter the phase relationship between the PIM signals, resulting in constructive interference at some down-tilt settings and destructive interference at others.

[001 14] A related phenomenon can arise in which the antenna’s interbranch isolation varies with down-tilt. This can produce similar effects to the one just described, in which the relative amplitudes of two or more PIM signals vary with down-tilt, resulting in different amounts of coherent interference from one down-tilt setting to the next.

[001 15] In summary, phased array antennas can have very complex and counterintuitive PIM responses. This is due to their mechanical complexity, poor interbranch isolation, low power divider output return loss and sensitivity to down-tilt. These factors have historically hindered the development of effective PIM troubleshooting tools and techniques forthis category of antennas.

[001 16] PIM Troubleshooting Procedures

[001 17] Aspects of the invention disclosed herein are primarily, but may not exclusively be, directed to the location of PIM faults in cellular sector antennas during manufacture.

[001 18] Without being limited by theory, a brief summary of the PIM testing and troubleshooting procedure that may be used by cellular antenna manufacturers is presented below.

[001 19] Cellular sector antennas may be manufactured in a high-volume, low-cost environment, often by production staff with little or no technical training. Cycle time is a key parameter, which means that it is important for defects to be detected and remedied as quickly and as early in the manufacturing process as possible. This is especially important in multi-port, multi-band antennas, wherein several independent phased arrays may be installed in the same housing, usually in a tightly integrated mechanical assembly. With this type of antenna, late detection of a faulty component can require extensive disassembly of several of the arrays inside the antenna in order to access the site of the fault.

[00120] For this reason, PIM testing may be performed progressively during the assembly process. One approach is to PIM test each array as soon as it is installed into the antenna housing. This approach entails moving the partially-constructed antenna unit from the production line to an anechoic chamber, and measuring the Reverse PIM response of the array in accordance with the IEC 62037 standard.

[00121 ] If the array passes PIM testing, it may be returned to the production line, where it moves onto the next stage in the assembly process. In the event that the array fails PIM testing, it may be taken to a repair station where a technician begins the troubleshooting process.

[00122] A first step in the troubleshooting process is to search for obvious defects, such as loose connectors, broken solder joints and cables protruding into areas with high levels of RF radiation. If no obvious assembly defects are found, the next step may be to rework every interconnect in every branch of the array. This includes reflowing all solder joints and performing back-torque checks on all coaxial connectors and mechanical fasteners.

[00123] Upon completion of the rework procedure, the antenna may be taken back to the anechoic chamber for a second round of PIM testing. If the array fails the test a second time, it is returned to the repair station, where the troubleshooting process may be repeated. This cycle of testing and rework continues until the array meets the target PIM specification.

[00124] It may be necessary to perform the PIM testing and troubleshooting procedure, as described above, on every array in the antenna unit. Accordingly, it may take hours or even days for a multi-array sector antenna to pass PIM testing on all ports and in all bands. Even worse, in an effort to identify PIM faults more rapidly, manufacturing staff may be tempted to adopt unsafe testing practices, such as tapping or perturbing parts of the antenna assembly during PIM testing. This requires a staff member to be inside the anechoic chamber while the PIM test is running, thereby exposing themselves to potentially harmful levels of RF radiation.

[00125] Aspects of the invention may not be exclusively applicable to cellular antennas only. As will be apparent to a person skilled in the art, aspects of the invention may be adapted to suit a variety of phased array antenna designs in any frequency band.

[00126] Identifying the Faulty Branch(es)

[00127] When locating a PIM fault within an antenna comprising a plurality of branches, it is desirable to determine which branch, or subset of branches, of the antenna the PIM fault resides in. If the faulty branch, or branches may be identified, then additional benefit may be obtained by determining the distance to the PIM fault using Distance-to-PIM (DTP) capability. For example, in cases where it is found that only one branch is faulty, a DTP sweep could pinpoint the exact location of the fault(s) within that branch. Similarly, DTP testing could be used to reveal whether any PIM faults exist in the input section of the antenna under test, between the RF port and the input to the power divider.

[00128] The location of PIM faults inside a phased array antenna may be located by applying a plurality of RF stimulus signals, for example two high-power carriers at frequencies F1 and F2, to the input port of the antenna under test (AUT), while measuring the Forward PIM signals emitted from each of the antenna’s radiating elements using a fixed array of over-the-air (OTA) probes. The high-power carriers F1 and F2 are typically unmodulated sinusoidal continuous wave (CW) signals, with frequencies chosen such that the PIM product of interest lands within the receive bands of both the PIM analyser and the AUT. In an alternative approach, one or both of carriers F1 and F2 may be modulated using any one of a variety of schemes, including direct sequence spread spectrum (DSSS), orthogonal frequency division multiplexing (OFDM) or frequency-modulated continuous wave (FMCW), to name just a few. In yet another approach, the RF stimulus signals, carriers F1 and F2, may be replaced by a single ultra-wideband modulated signal.

[00129] Any of the above variants of F1 and F2 may be used in practice, provided that the resulting PIM product lies within the receive bands of the PIM analyser and the AUT, and that the PIM product’s amplitude is within the dynamic range of the PIM analyser’s receiver. By analysing the amplitude and/or phase characteristics of the radiated Forward PIM signals, it may be possible to identify which branch or set of branches of the antenna the PIM fault resides in. Depending on the configuration of the test apparatus, which fault finding procedure is used, the number and location of PIM faults in the AUT and the configuration of the AUT, embodiments of aspects of the invention may be capable of detecting a PIM fault in one branch, or may be capable of detecting multiple PIM faults in multiple branches.

[00130] The accuracy, precision and robustness of the results produced by an embodiment may be improved by measuring the Forward PIM across a range of frequencies within the passband of the antenna. Further improvements in the quality of the results can be achieved by measuring the Forward PIM at multiple down-tilt settings, and/or by deliberately introducing large impedance mismatches into one or more branches of the antenna in order to perturb the antenna’s PIM response in a controlled fashion.

[00131 ] Unlike embodiments that employ a single probe mounted on a movable gantry, the embodiments of aspects of the current invention may use a fixed array of probes to measure the Forward PIM signal emitted by each antenna element. This ensures that all antenna elements are presented with a near-constant impedance during the PIM measurement process, thereby minimising return loss-induced measurement uncertainty.

[00132] Additionally, using a fixed probe array rather than a single movable probe may allow PIM measurements to be performed more rapidly, as no time is lost in waiting for the moveable probe to physically move from element to element. This effect may be especially evident in embodiments in which each probe in the fixed probe array is connected to its own receiver, thereby allowing the Forward PIM signals from all branches to be measured simultaneously.

[00133] Advantageously, embodiments of the invention may be reconfigured for different antenna models, because there is no need to specify the exact X-Y-Z coordinates of the individual radiating elements in every antenna model, something that may be required in systems based on a single probe mounted on a moveable gantry.

[00134] As will be apparent to a person skilled in the art, a test apparatus configured in accordance with aspects of the invention may be implemented in a variety of hardware configurations, each offering different strengths and limitations.

[00135] OTA Probe Array with Filter Bank and Solid-State RF Switchbox

[00136] An example system 300 which embodies an aspect of the invention is depicted in Figure 3. The antenna under test (AUT) 302 is placed on a testbench (not shown) inside an anechoic chamber 304, with its radiating elements 320a to 320n facing towards the ceiling. It is preferred that the testbench and the anechoic chamber 304 both be constructed from low-PIM materials, so that they do not generate spurious PIM when illuminated by the transmit carriers F1 and F2 that radiate from the AUT 302 during testing.

[00137] A two-port PIM analyser 306 is installed outside the anechoic chamber 304. Any make and model of PIM analyser may be used, including those that are capable of making both Forward and Reverse PIM measurements according to or exceeding the IEC 62037 standard.

[00138] It is preferred that the PIM analyser 306 be capable of performing swept-frequency PIM measurements within the passband of the AUT 302.

[00139] It may also be advantageous for the PIM analyser 306 to be capable of measuring the phase of the Forward and/or Reverse PIM signals as well as their amplitude. This capability is denoted by the RTF Module 308 in Figure 3. The RTF Module 308 may comprise a companion product for PIM analysers manufactured by Kaelus Pty Ltd of Brisbane, Australia that confers on the analyser the ability to measure the phase of a PIM signal. PIM analysers from other manufacturers usually have the necessary phase detection circuitry integrated into their transceivers, and therefore do not require an RTF module in order to measure PIM phase.

[00140] Port 1 of the PIM analyser 306 is connected to the RF port of the AUT 302 by a length of low-PIM coaxial cable, illustrated by line 310. It may be preferred that the insertion loss of the coaxial cable be as low as possible, in order to maximise the amount of RF power that is delivered to the AUT 302. [00141] Port 2 of the PIM analyser 306 is connected to the output port of a filter bank 312 by a length of coaxial cable, illustrated by line 314. It may not be necessary that the coaxial cable 314 have superior PIM performance, but it is desirable for the coaxial cable 314 to have low insertion loss in order to maximise the signal-to-noise ratio (SNR) of the PIM measurement.

[00142] It is desirable that the PIM analyser 306 contains an AISG modem 316, so that the down-tilt of the AUT 302 can be adjusted automatically during the PIM fault finding process. If the PIM analyser’s AISG interface takes the form of a dedicated RS-485 connector, this port should be connected to the Remote Electrical Tilt (RET) port of the AUT 302 (if available) by a length of shielded digital communications cable, illustrated by line 322.

[00143] The Forward PIM signals emitted by the radiating elements 320a to 320n in the AUT 302 are detected by an array of over-the-air (OTA) probes 318a to 318n suspended above the AUT 302.

[00144] Each OTA probe 318a to 318n in the array is coupled to one radiating element 320a to 320n in the AUT 302. In order to maximise the coupling between the two structures, the vertical separation between the OTA probe and the corresponding radiating element is typically in the range 10-100 mm. Each OTA probe 318a to 318n is centred over its counterpart radiating element 320a to 320n in the AUT 302, and the angular orientation of the probe adjusted to achieve strongest coupling.

[00145] The OTA probes 318a to 318n are mounted on a gantry 322 constructed from low- PIM materials that are preferably transparent to RF radiation, such as timber, Delrin or PTFE. The gantry is height-adjustable to allow the optimal vertical separation between the probe array 318a to 318n and the AUT 302 to be attained, and to facilitate the easy placement and removal of the AUT 302 on the testbench (not shown).

[00146] For preferred operation of the embodiment as shown in Figure 3, the individual OTA probes 318a to 318n should have the following performance characteristics.

[00147] Tight coupling to AUT 302: Each probe 318a to 318n should couple as strongly as possible to the radiating element 320a to 320n to which it has been assigned, in order to maximise the SNR of the PIM measurements. A coupling loss of approximately 5-10 dB is recommended for cellular antenna testing, although weaker coupling levels may be acceptable in other applications, depending on the range of PIM levels to be measured and the noise floor of the PIM analyser’s receiver.

[00148] High isolation from adjacent elements: Each probe 318a to 318n should respond predominantly to the PIM signal emitted by the radiating element 320a to 320n to which it has been assigned, while ignoring PIM signals from neighbouring elements in the AUT 302.

[00149] High power handling: In addition to the PIM signal from the AUT 302, the radiating elements 320a to 320n in the AUT 302 will also emit the two high-power carriers F1 and F2. These two carriers will couple into each OTA probe 318a to 318n at power levels that may exceed 8 watts rms. It is therefore important for the OTA probes 318a to 318n to be capable of passing this amount of RF power without suffering damage due to RF heating or dielectric breakdown.

[00150] Low PIM: Due to the substantial amount of carrier power that can couple into the OTA probes 318a to 318n, it is important for the probes to have a low-PIM construction, so that the residual PIM is well below the minimum PIM signal of interest from the AUT 302. This requirement also applies to the RF cables that connect the probe output ports to the input ports of the filter bank 312.

[00151 ] No disruption to input return loss of radiating elements 320a to 320n in AUT 302: When an OTA probe couples to one of the elements in the AUT 302, there is a risk that the input return loss of the element will degrade due to the load impedance presented by the OTA probe. A poor load match can dramatically perturb the amplitude of the PIM generated inside an RF device, thereby increasing the uncertainty in the PIM measurement. In order to avoid this, it is preferred that the OTA probes be designed in such a way that they do not degrade the input return loss of the radiating elements 320a to 320n in the AUT 302.

[00152] Bandwidth: It is desirable that the OTA probes 318a to 318n satisfy the above criteria at all frequencies in the transmit and receive bands of the PIM analyser. Furthermore, it is desirable for the OTA probes 318a to 318n to satisfy the above criteria across multiple cellular frequency bands, so that PIM troubleshooting can be performed in different frequency bands if desired, without requiring a different OTA probe array to be used in each band.

[00153] Circular polarization: It is desirable for the OTA probes 318a to 318n to be relatively insensitive to the angular orientation of each probe relative to its counterpart in the AUT. One way of achieving this is to base the OTA probe design on a circularly polarized antenna. This has the advantage of minimizing the sensitivity of the test apparatus to small differences in probe orientation from one element to the next. It also allows the OTA probes 318a to 318n to measure the PIM from both arrays in a cross-polar antenna array without requiring the probes’ angular orientations to be changed from one array to the next, thereby allowing the same equipment setup to be used regardless of which half of the cross-polar array is being tested. [00154] The output ports of the OTA probe array 318a to 318n are connected to the common ports of a bank of diplexers 324a to 324n, which separate the measured PIM signals from the leaked F1 and F2 carriers from the AUT. One diplexer is provided for every OTA probe. The PIM signal from each OTA probe passes through the corresponding diplexer’s receive (Rx) filter, and continues on to an RF switchbox 328. The F1 and F2 carriers from the OTA probe pass through the diplexer’s transmit (Tx) filter and into a high-power matched load 326a to 326n.

[00155] The diplexer bank 324a to 324n is necessary due to the relatively large amount of RF power in the F1 and F2 carriers that couple from the AUT 302 to the OTA probes 318a to 318n. As mentioned previously, this power can exceed 8 watts rms per probe in extreme cases. The risks posed by the F1 and F2 carriers are two-fold, and both are related to the RF switchbox 328. Firstly, the RF switchbox 328 may be damaged by the power in F1 and F2 carriers, and secondly the F1 and F2 carriers may generate spurious intermodulation products in the active electronic circuitry inside the RF switchbox 328. In order to avoid these risks and maximise the accuracy of the test apparatus, it may be preferred that each diplexer in the diplexer bank 324a to 324n satisfies the following criteria.

[00156] Rx filter attenuation in Tx band: The diplexer’s Rx filter should provide sufficient attenuation in the transmit band of the PIM analyser 306 to protect the RF switchbox 328 from damage by the leaked F1 and F2 carriers, and to ensure that any spurious intermodulation products generated in the RF switchbox 328 by F1 and F2 are well below the minimum PIM level of interest from the AUT 302.

[00157] Tx filter attenuation in Rx band: The diplexer’s Tx filter should provide sufficient attenuation in the receive band of the PIM analyser 306 to ensure that any spurious PIM products that are generated in the high-power load resistor 326a to 326n at the output of the Tx filter are well below the minimum PIM level of interest from the AUT 302 by the time they reach the common port of the diplexer.

[00158] Common port return loss: The input return loss of the diplexer at its common port should be as high as possible in both the transmit and receive bands of the PIM analyser 306, in order to minimize the PIM measurement uncertainty of the test apparatus due to impedance mismatches at the F1 , F2 and PIM frequencies.

[00159] The diplexer bank 324a to 324n is followed by an RF switchbox 328, whose purpose is to multiplex the filtered PIM signals from the diplexer bank 324a to 324n onto a single output port. Each input port on the switchbox 328 is connected to a corresponding Rx output port in the diplexer bank 324a to 324n. The active input port on the RF switchbox 328 is set by sending a software command from the System Controller 330 via a USB communications link. [00160] In the embodiment depicted in Figure 3, the RF switchbox 328 is preferably a solid- state device built from electronic switches. Solid-state RF switches may offer benefits, including low cost, low power consumption, small size and high speed compared to other RF switches.

[00161 ] The limitations of solid-state RF switches are low power handling, poor linearity and high insertion loss. For example, commercially available RF switches can usually tolerate only a couple of hundred milliwatts of RF power without damage, and have third-order intercept points of no more than +50 dBm. As for insertion loss, this can exceed 6 dB in an RF switchbox with a large number of ports.

[00162] The former two limitations - i.e. power handling and linearity - are the reasons why the diplexer bank 324a to 324n may be needed. While the low-level PIM signals from the OTA probes 318a to 318n pose no risk to the RF switchbox 328, the leaked F1 and F2 carriers have the potential to either cause permanent damage, or generate spurious intermodulation products that can interfere with the measurement of the PIM product of interest from the AUT 302.

[00163] Regarding the insertion loss limitation of solid-state RF switches, if the insertion loss is so high that it seriously impairs the sensitivity of the system, the options are to: increase the power in the F1 and F2 carriers; insert a low-noise amplifier (LNA) before each of the input ports on the RF switchbox 328 to limit the degradation in SNR; use a lower-loss switch technology such as electromechanical relays; or reduce the noise floor in the PIM analyser’s receiver.

[00164] In the embodiment depicted in Figure 3, it is preferred that the RF switchbox 328 meets the following RF performance criteria.

[00165] Return loss: The return loss on all ports of the switchbox should be sufficiently high to ensure that the PIM measurement uncertainty of the test apparatus due to impedance mismatches at the F1 , F2 and PIM frequencies is as low as possible.

[00166] Insertion loss: The insertion loss of the RF pathway between the active input port and the common port of the switchbox should be as low as possible, in order to preserve the SNR of the measured PIM signal. If this is not possible, the insertion loss mitigation strategies described above may be considered.

[00167] Isolation: The isolation between the common port of the switchbox and its inactive input ports should be as high as possible, to ensure that the test apparatus only measures the PIM signal from one OTA probe at a time.

[00168] Power handling: The 1 dB compression point of the switchbox should be sufficiently high to allow accurate measurement of the maximum PIM level of interest from the AUT 302. [00169] The embodiment depicted in Figure 3 includes a bandpass filter 332 between the common port of the RF switchbox 328 and Port 2 of the PIM analyser 306. The bandpass filter 332 passes frequencies in the receive band of the PIM analyser 306, while rejecting frequencies in the transmit band.

[00170] The purpose of the bandpass filter 332 is to protect the RF switchbox 328 from damage in the event that the two high-power carriers F1 and F2 are accidentally transmitted from Port 2 of the PIM analyser 306 instead of Port 1 as intended. This could occur as the result of an error in the instrument’s software settings, or if the test port cables on Ports 1 and 2 of the PIM analyser 306 were to be interchanged by mistake when the test apparatus is being assembled. In either case, the RF switchbox 328 may suffer irreparable damage. With the bandpass filter 332 in place, the F1 and F2 carriers are reflected back into the PIM analyser 306, where they are absorbed by the high-power isolators in the instrument’s transmit module (not shown). There is no risk of damage to the PIM analyser 306 in this scenario, as the isolators (not shown) are designed to dissipate this amount of power if necessary.

[00171 ] In the embodiment depicted in Figure 3, it is preferred that the Rx bandpass filter 332 meet the following performance requirements:

[00172] Attenuation in Tx band: The bandpass filter 332 should provide sufficient attenuation in the transmit band of the PIM analyser 306 to protect the RF switchbox 328 from damage in the event that the F1 and F2 carriers are inadvertently transmitted from Port 2 of the PIM analyser 306.

[00173] Power handling: The bandpass filter 332 should be capable of withstanding the F1 and F2 carriers being applied to its output port at full power, without incurring any damage or significant changes in its frequency response due to RF heating effects or voltage breakdown.

[00174] The system controller 330 is a programmable device, for example a personal computer or microcontroller, that runs the control software for the test apparatus. It also provides a user interface to enable the operator to configure and initiate a measurement sweep and view the results upon completion of the PIM fault-finding process.

[00175] The system controller 330, included in the embodiment depicted in Figure 3, includes the following core functions.

[00176] The system controller 330 provides a user interface for configuring the instrument’s sweep settings and entering the details of the AUT 302, such as model, serial number and number of radiating elements. The system controller 330 automates the operation of the PIM analyser 306, RTF module 308 and RF switchbox 328 during a PIM measurement sweep. The system controller 330 adjusts the down-tilt of the AUT 302 by sending AISG commands to the AUT’s RET controller (not shown) via the AISG modem 316 in the PIM analyser 306. The system controller 330 post-processes the raw sweep data from each OTA probe 318a to 318n at the end of a PIM measurement sweep, and estimates the location(s) of any PIM fault(s) inside the AUT 302. The system controller 330 reports the estimated PIM fault location(s) to the user. The system controller 330 saves the results to a non-volatile storage medium (not shown) for future reference and statistical analysis.

[00177] OTA Probe Array with Relay-Based RF Switchbox

[00178] Another embodiment of an aspect of the invention is depicted in Figure 4. In this embodiment, a hardware configuration substantially similar to the embodiment depicted in Figure 3 is used, except that the diplexer bank 324a to 324n and Rx bandpass filter 332 are eliminated, and the solid-state RF switchbox 328 is replaced by an RF switchbox 402 constructed from a network of electromechanical relays. Additionally, in contrast to the embodiment depicted in Figure 3, the coaxial cable 408 that connects the RF switchbox 402 to Port 2 of the PIM analyser 404 must now be a low-PIM cable.

[00179] The motivation for using a relay-based RF switchbox 402 in the embodiment depicted in Figure 4 is that relays can handle much higher RF power levels than electronic switches, and generate much lower levels of spurious intermodulation distortion (also known as “residual PIM”). It is these features that allow the diplexer bank 324a to 324n and Rx bandpass filter 332 (of the embodiment depicted in Figure 3) to be omitted: The relay-based RF switchbox 402 does not require protection from the power in the F1 and F2 carriers (either from the OTA probes 406a to 406n or from Port 2 of the PIM analyser 404), and it does not require the leaked F1 and F2 carriers from the OTA probes 406a to 406n to be attenuated in order to avoid generating spurious PIM in the relay network inside the switchbox 402.

[00180] The use of a relay-based RF switchbox 402 may offer advantages; namely, the test apparatus is simplified greatly by the removal of the diplexers, bandpass filter and associated cabling. Furthermore, RF relays have much lower insertion loss than solid-state switches, which increases the signal to noise ratio (SNR) of the measured PIM signal significantly compared to the embodiment depicted in Figure 3. Additionally, RF relays may operate over a relatively wide frequency range. With the removal of the diplexers 324a to 324n and Rx bandpass filter 332, this means that the embodiment depicted in Figure 4 is capable of wideband operation, being limited only by the bandwidth of the OTA probes 406a to 406n and the PIM analyser 404.

[00181 ] Without being limited by theory, the disadvantages of relay-based RF switches include their large size and slow speed relative to solid-state switches. Furthermore, while relay- based RF switches with good residual PIM performance are commercially available, they tend to be prohibitively expensive, and their residual PIM performance can degrade significantly after a certain number of switching operations.

[00182] If the latter two limitations of high cost and degradation of residual PIM over time could be overcome, then the system 400 as depicted in Figure 4 would be preferred over the configuration as depicted in Figure 3, under some circumstances, despite the larger size and slower switching speed of this solution. However, the configuration as depicted in Figure 3 may be far more economical and easy to implement than the configuration as depicted in Figure 4, and may therefore be the preferred configuration, under certain circumstances.

[00183] OTA Probe Array and PIM Analyser with Multiple Receivers

[00184] In the system 500 depicted in Figure 5, the two-port PIM analysers 306, 404 depicted in Figure 3 and Figure 4 respectively, are replaced by a PIM analyser 502 which includes as many receivers 510a to 51 On as there are OTA probes 506a to 506n, plus an additional receiver 509 that is used to measure the Reverse PIM response of the AUT 512.

[00185] This multi-receiver PIM analyser 502 can measure the PIM signals from all OTA probes 504a to 504n simultaneously, greatly speeding up the data acquisition phase of the PIM fault finding process. It also allows the RF switchboxes 328, 402 depicted in Figure 3 and Figure 4 respectively, to be eliminated completely, while the diplexer bank 324a to 324n and Rx bandpass filter 332 depicted in Figure 3 are replaced by an internal diplexer bank 506a to 506n. The OTA probe array 504a to 504n may be substantially identical to the OTA probe arrays included in the configurations depicted in Figure 3 and Figure 4. The coaxial cables that connect the OTA probes 504a to 504n to PIM analyser’s internal diplexer bank 506a to 506n should all be low-PIM cables.

[00186] Another advantage of the multi-receiver PIM analyser 502 is that it may be better adapted to detecting unstable or transient PIM sources, which may be referred to as“dynamic” PIM sources. This type of PIM source can exhibit sudden and dramatic changes in PIM level from moment to moment, usually as the result of vibration, percussive stimulus or thermal expansion and contraction of the AUT. When this occurs, the Forward PIM signals emitted from all OTA probes may track the change in the level of the dynamic PIM source within a few nanoseconds of each other. Using the multi-receiver PIM analyser, this near-simultaneous change in PIM amplitude across all elements may be recorded, thereby enabling the relative PIM levels emitted by each element to be accurately measured, even as the absolute PIM level varies from instant to instant. [00187] The implementation of a multi-receiver PIM analyser as depicted in Figure 5 is designed to measure the Forward PIM in one direction only, from Port 1 to any of the other N ports, while the Reverse PIM can be measured at Port 1 only.

[00188] The configuration as depicted in Figure 5 is one possible implementation of a multireceiver PIM analyser. A person skilled in the art would understand that in principle, more elaborate implementations of a multi-receiver PIM analyser are possible that would allow the Forward and Reverse PIM to be measured in any direction through any combination of ports. Some of the implementations of a multi-receiver PIM analyser could offer significant benefits, such as the ability to perform Distance-to-PIM testing into each radiating element individually. Such an instrument would be capable of unambiguously identifying PIM-affected branches, as it could pinpoint the exact location of PIM faults located between the radiating element and the splitter output port in that branch; however, this capability may come at the expense of significantly increased hardware complexity.

[00189] The multi-receiver PIM analyser 502 will be described in more depth with reference to the block diagram as depicted in Figure 5.

[00190] The multi-receiver PIM analyser 502 contains a transmit module 508 that generates two high-power carriers F1 and F2, similar to a conventional one-port or two-port PIM analyser. The transmit module 508 allows the frequency and amplitude of carriers F1 and F2 to be adjusted if desired, for example to perform a swept-frequency PIM measurement.

[00191 ] The multi-receiver PIM-analyser 502 contains an internal diplexer bank 505, 506a to 506n in its RF front end. One of these diplexers 505 is a Tx/Rx diplexer that connects the transmit module 508 and Receiver 1 509 in the receiver bank 509, 510a to 51 On to Port 1 of the PIM analyser 502.

[00192] During PIM testing, carriers F1 and F2 pass through the transmit filter of the Tx/Rx diplexer, which removes any spurious intermodulation products that may have been generated in the transmitter. At the same time, Reverse PIM signals from the AUT 512 enter Port 1 of the PIM analyser 502 and pass through the receive filter of the Tx/Rx diplexer 505 and into Receiver 1 509 in the receiver bank.

[00193] Note that in the configuration as depicted in Figure 5, the Tx/Rx diplexer 505 is essential to the PIM analyser’s Reverse PIM measurement capability. This enables the operator to perform a standard factory PIM test before and after reworking the AUT 512, in order to determine whether the repair process has been successful. [00194] The internal diplexer bank 505, 506a to 506n in the multi-receiver PIM analyser 502 also contains a set of receive diplexers 506a to 506n, one for every probe in the OTA probe array 504a to 504n. The OTA probe outputs are connected to the common ports of the receive diplexers, labelled Ports 2 to N+1 in Figure 5.

[00195] The receive diplexers 506a to 506n serve a substantially identical function to the external diplexer bank 324a to 324n as depicted in Figure 3, namely to protect the electronic circuitry inside the instrument from damage by the leaked F1 and F2 carriers from the AUT 512, as well as attenuating the F1 and F2 carriers sufficiently to avoid generating spurious intermodulation products inside the instrument that could interfere with the PIM measurement process.

[00196] As was the case for the external diplexer bank 324a to 324n in Figure 3, the input return loss at the common port of each receive diplexer 506a to 506n in the internal diplexer bank should be as high as possible in both the transmit and receive bands of the PIM analyser 502, in order to minimize the PIM measurement uncertainty of the test apparatus due to impedance mismatches at the F1 , F2 and PIM frequencies.

[00197] The PIM signals from the OTA probes 504a to 504n pass through the receive filters in the diplexer bank 506a to 506n and into a bank of tuned radio receivers 510a to 51 On. One receiver is provided for each diplexer in the diplexer bank 505, 506a to 506n, including the Tx/Rx diplexer 505, which as mentioned previously is required for making Reverse PIM measurements.

[00198] During a Forward PIM measurement sweep, the incoming PIM signals from the OTA probes 504a to 504n are down converted in the receiver bank 510a to 51 On, before being digitized and transmitted to the system controller 516 for post-processing. This operation can be performed simultaneously by all receivers in the receiver bank, enabling Forward PIM sweeps to be carried out more rapidly in the configuration as depicted in Figure 5, compared to the configurations depicted in Figures 3 and 4.

[00199] If necessary, the receivers in the receiver bank 509, 510a to 51 On can be designed to measure PIM phase as well as amplitude. There are numerous existing techniques for realizing this functionality, which will be well-known to one skilled in the art. Three such techniques are described in United States patent 8,995,517 (“Method and apparatus for locating faults in communications networks”).

[00200] It is preferred that each receiver in the receiver bank meet the following criteria:

[00201 ] Sensitivity: The receiver should be capable of detecting the minimum Forward PIM signal of interest from the AUT 512, taking into account the coupling loss of the OTA probes 504a to 504n and the insertion loss of the cables and receive diplexers 506a to 506n.

[00202] Noise figure: The noise figure should be sufficiently low that the SNR of the measured PIM signal is not unduly affected by noise generated within the receiver itself

[00203] Input return loss: The input return loss should be as good as possible in order to minimize PIM measurement uncertainty due to poor load match at frequency of PIM signal

[00204] Compression point: The receiver should be capable of detecting the maximum PIM signal of interest from the AUT 512 without exceeding the receiver’s 1 dB compression point

[00205] Over-power protection: In the event that the PIM signal from the AUT 512 exceeds the 1 dB compression point of the receiver, it is prudent to provide an RF-limiting device at the input to the receiver as a protective measure.

[00206] In principle the multi-receiver PIM analyser 502 implemented in the configuration as depicted in Figure 5 could be adapted to support an RTF module of Kaelus Pty Ltd. This would offer an alternative way of measuring the phase of PIM signals from the AUT 512.

[00207] However, for the configuration as depicted in Figure 5, the use of an RTF module may not be recommended, as an RTF module cannot measure the phase of the PIM signals from all OTA probes simultaneously. Instead, the PIM signals would have to be measured one by one, using an RF switch matrix to multiplex the PIM signals onto a common port connected to the RTF module’s AUX port. This approach may have the effect of moderating an advantage of the configuration as depicted in Figure 5 compared to the configurations as depicted in Figures 3 and 4, namely the superior measurement speed provided by the configuration depicted in Figure 5.

[00208] A preferable approach to measuring PIM phase may be to integrate the phase detection circuitry into the receiver bank. This circuitry could leverage some of the design principles used in an RTF module, although other approaches are also possible as described above.

[00209] In yet another approach, a multiport version of an RTF module could be developed based on the inline PIM and scattering parametertest set described in International (PCT) Patent Publication WO/2019/000034 entitled“System and Apparatus for Identifying Faults in a Radio Frequency Device or System”.

[00210] Similar to the apparatus configurations as depicted in Figure 3 and Figure 4, it may be desirable for the PIM analyser 502 to contain an AISG modem 514, so that the down-tilt of the AUT 512 can be adjusted during the PIM fault finding process. If the PIM analyser 502 is equipped with an AISG modem 514 that is connected to a dedicated RS-485 front panel connector, this port should be connected to the Remote Electrical Tilt (RET) port of the AUT 512 by a length of shielded digital communications cable 518.

[0021 1 ] The system controller 516, depicted in Figure 5, may be functionally the same as the system controllers 330, 410 of Figures 3 and 4, respectively, except that it now collects PIM readings from multiple receivers 509, 510a to 510n, rather than from one receiver and an RF switchbox 328, 402.

[00212] PIM Fault-finding Methods

[00213] There is a diverse range of methods by which a test instrument, embodying aspects of the invention, can identify PIM fault locations in a phased array antenna. Without limiting the intended scope of the invention, available methods to identify PIM fault locations in a phased array antenna include both scalar and vector techniques, using Forward PIM only or a combination of Forward and Reverse PIM measurements. Some methods may involve making PIM measurements at multiple antenna down-tilt settings, while others may entail deliberately introducing an impedance mismatch into one or more branches of the AUT and observing the change in PIM response. Methods are available that treat the AUT as a“black box”, in which the fault location is inferred from the measured PIM data using a set of empirical rules. Other methods rely on theoretical models of the AUT’s PIM response, which in turn depend on detailed knowledge of the internal construction of the AUT and the RF characteristics of every component therein. Some methods are only applicable to scenarios in which the AUT contains a single, dominant PIM fault, whereas other methods can detect multiple PIM faults in multiple branches.

[00214] The choice of which method to use may be influenced by a trade-off considering accuracy, complexity and robustness. That is, the more accurate the method, the more complicated the test apparatus and measurement procedure may be, and the less tolerant the results may be to noise and other imperfections in the measured data.

[00215] A selection of PIM fault-finding methods that may be compatible with the configurations as depicted in Figures 3, 4 and 5 are presented in the following sections.

[00216] First Step: Distance-to-PIM Sweep into RF Port of AUT

[00217] Types of phased array antenna configurations, with relevance to aspects of the invention, are depicted in Figures 1 , 2 and 16. These phased array antenna configurations all contain a common RF path between the RF port of the AUT and the input to the power divider network, as also shown in Figure 1 , Figure 2 and Figure 16.

[00218] If a PIM fault arises in the RF path between the RF port of the AUT and the input to the power divider network of the AUT, the resulting Forward PIM signal may propagate through the power divider and into all branches of the AUT, eventually radiating from every element. This can create the erroneous impression that the AUT contains PIM faults in every branch, when, in reality, only a single PIM fault may exist, located in the input section of the AUT. For this reason, it may be preferable that PIM faults in this part of the AUT be identified and eliminated first.

[00219] A convenient and reliable way of detecting PIM faults in the input section of the AUT may be by performing a Distance-to-PIM (DTP) sweep into the AUT’s RF port. The RTF module of Kaelus Pty Ltd, when used in conjunction with a compatible PIM analyser, is an example an instrument which is capable of performing a DTP sweep. The RTF module is supported by the configurations as depicted in Figures 3 and 4. The configuration as depicted in Figure 5 could be designed to support the RTF module, but as mentioned previously it may be preferable to integrate the phase detection functionality upon which DTP analysis depends into the transceiver of the multi-receiver PIM analyser.

[00220] In the event that a PIM fault is detected in the common branch of the AUT, being the RF path between the RF port of the AUT and the input to the power divider network of the AUT, it is preferable that the PIM fault be repaired immediately. After repairing the PIM fault in the common branch of the AUT, the AUT may be PIM tested according to the normal factory test procedure. If the AUT’s Reverse PIM meets the factory test limit, it can be returned to the main production line. Further PIM troubleshooting may be required.

[00221 ] All of the troubleshooting methods described in the following sections assume that the input section of the AUT is free from PIM faults. Indeed, the flowcharts for each method, shown in Figure 6, Figure 10 and Figure 12, all include a DTP testing step at the beginning of the process. If the input section of the AUT cannot be assumed to be free from PIM faults, e.g. because the test apparatus does not have a DTP testing capability, then the methods in the following sections may still be used; however, the interpretation of the measured PIM data may require more complicated heuristics, or a more elaborate model of the AUT, in order to reliably identify the faulty branches.

[00222] Scalar Fault-finding Methods

[00223] The faulty branch in an AUT may be identified by measuring the scalar Forward PIM response of each branch, and using heuristic techniques, such as pattern recognition based on empirical knowledge of the AUT, to infer which branch is the faulty one. The term“scalar” in this context is intended to mean that the test apparatus only measures the power (or amplitude) of the Forward PIM signal, not its phase.

[00224] Three methods for identifying a faulty branch in the AUT are presented in the following sections. These methods may be best suited to scenarios in which the AUT contains a single dominant PIM fault in one branch only. Under these conditions, scalar methods can greatly speed up the troubleshooting process.

[00225] In cases involving multiple PIM faults spread across multiple branches, scalar methods can still be used, but may require the use of more sophisticated post-processing algorithms, such as Machine Learning, the CLEAN algorithm or parametric estimation.

[00226] Method 1: Scalar PIM Measurements at a Fixed Down-tilt Setting

[00227] A scalar fault finding method (hereafter referred to as “Method 1”) involves performing a swept Forward PIM measurement across the widest possible frequency range, subject to the frequency limits of the PIM analyser and the bandwidth of the AUT, and measuring the power of the PIM signals emerging from the radiating elements.

[00228] Figure 6 shows an exemplary set of steps to perform Method 1 in accordance with an aspect of the invention. As will be apparent to a person skilled in the art, a scalar fault-finding method may be performed through a set of steps which may differ from the steps shown in Figure 6, whilst still falling within the scope of the present invention.

[00229] The first step 604 in Method 1 600 is to place the AUT in an anechoic chamber and connect the AUT to the test apparatus.

[00230] The second step 606 is to configure the PIM analyser’s frequency sweep settings. In most cases the lowest-order PIM product in the receive band of the PIM analyser is preferred, due to the fact that low-order PIM products like IM3 usually have larger amplitudes than higher order PIM products like IM5 and IM7, and are therefore more likely to be responsible for the AUT failing its factory PIM test.

[00231 ] The third step 608 is to set the AUT down-tilt to a known value. This may be done by sending a command to the AUT’s RET controller via the PIM analyser’s AISG port. The down-tilt value should preferably be the same as it was when the AUT failed the original factory PIM test. Once the down-tilt of the AUT has been set, it will remain unchanged for the remainder of the fault-finding process. [00232] The fourth step 610 is to perform a DTP sweep into the RF port of the AUT and look for PIM faults as step 612 in the input section of the AUT, between the RF port and the input to the power divider network. Any PIM faults found in this part of the AUT should be repaired immediately, preferably through re-working the input branch in step 614. Once the input section of the AUT has been declared PIM-free, a standard factory PIM test may be performed as step 616. If the AUT passes this test, as determined by step 618, it can be returned to the main production line to continue to the next stage in the assembly process, thus ending the process shown in Figure 6. Otherwise, the PIM troubleshooting procedure moves on to step 620.

[00233] The next steps 620 to 632 in Method 1 involve performing a Forward PIM sweep into the RF port of the AUT, and measuring the PIM levels emitted by its radiating elements. The preferred sweep type is a stepped-frequency sweep, in which the PIM is measured at a set of discrete frequency points within the receive band of the PIM analyser. In general, the either the F1 or F2 carriers may be swept, usually with little difference in the results. In some cases, it may be advantageous to perform both an F1 sweep and an F2 sweep, and average the results in order to increase the reliability of the fault location estimate. Alternatively, the F1 and F2 carriers could be swept simultaneously. This latter approach may provide more reliable results when the AUT contains impedance mismatches in one or more of its branches.

[00234] Even more elaborate stimulus methods may be utilised, in accordance with an aspect of the invention, such as applying Frequency Modulated Continuous Wave (FMCW) modulation, spread-spectrum modulation or multicarrier modulation to either or both of F1 and F2. These stimulus methods may be compatible with Method 1 , although the complexity of the transceiver hardware may be considerably higher with these approaches than with the stepped- frequency approach described above.

[00235] Note that in principle there is no reason why Method 1 could not be performed at a fixed frequency, without sweeping carriers F1 and F2 at all. In general, however, it is preferable to measure the PIM response of the AUT across a range of frequencies, as the information contained in the additional frequency points tends to increase the immunity of the results to noise and other imperfections in the measured data.

[00236] A typical set of swept PIM traces is presented in Figure 7, which shows a simulation of PIM emitted by radiating elements in a 7-branch phased array antenna during a Forward PIM measurement sweep using Method 1 . The results were taken from a Microwave Office™ (registered mark of AWR Corporation) simulation of an 1800 MHz phased array antenna having a topology similar to that of Figure 1 . In this example the AUT has 7 branches, with a single PIM fault in Branch 2 at a distance of 20 mm from the splitter output port in that branch. [00237] Upon completion of a swept PIM measurement on the AUT, the raw PIM data from each radiating element is post-processed, in step 630, by the system controller, which then estimates which branch is the faulty one and reports the result to the user. A variety of postprocessing algorithms are possible to identify the faulty branch. One approach is to simply calculate the average power emitted by each radiating element across the measured frequency range, and plot the results as a bar graph. The highest bar on the bar graph will correspond to the faulty branch. An example of this is shown in Figure 8, which was computed using the simulated PIM sweep data in Figure 7. Figure 8 depicts the simulated results of Method 1 , showing average PIM emitted by radiating elements in a 7-branch phased array antenna. The true location of the PIM fault is Branch 2 of AUT, at a distance of 20mm from the splitter output port. Using Method 1 , the faulty branch is correctly identified in the case depicted in Figure 8.

[00238] Method 1 may be modified slightly for antennas having a configuration similar to the configuration shown in Figure 16, due to the fact that each branch, of the antenna 1600 shown in Figure 16, is connected to multiple radiating elements (for example, 1608a and 1608b). Accordingly, for antenna configurations in which one or more branches is connected to a plurality of radiating elements, it may be advantageous to treat elements that share a common branch as a single logical element for the purpose of applying Method 1 . This may be achieved by summing together the measured PIM levels emitted by such elements, thereby obtaining a list of the total Forward PIM power emitted by each branch of the AUT. From this point onwards, the algorithm to identify the faulty branch proceeds exactly as it would for antennas with architectures similar to those shown in Figure 1 and Figure 2.

[00239] Note that the above described algorithm to identify the faulty branch may identify an alternative branch as the faulty branch. This is because it is possible in some cases for a faulty branch to emit a lower PIM level than a neighbouring branch which contains no faults at all, thereby causing the neighbouring branch to be misidentified as the source of the PIM fault. This results from a fundamental property of the AUT itself, caused by the high output return loss and poor interbranch isolation of the splitter module.

[00240] An example of Method 1 identifying an alternative branch than the faulty branch is shown in Figure 9, which depicts average PIM emitted by radiating elements in a 7-branch phased array antenna. These results were taken from a simulation identical to the previous example depicted in Figure 8, except that the PIM fault is now located 71 mm from the splitter output port rather than 20 mm. It is clear that the largest bar in the bar graph now corresponds erroneously to Branch 1 , even though the true fault location is in Branch 2.

[00241 ] Referring again to the bar graph shown in Figure 9, it can be seen that the PIM levels for Branches 1 , 2 and 3 are all significantly elevated compared to those of Branches 4 to 7, indicating that a PIM fault may be located on one or more of Branches 1 to 3. Accordingly, Branches 1 to 3 may be reworked to ensure that the faulty branch is fixed.

[00242] After the suspected faulty branch or branches in the AUT have been re-worked, as depicted by step 632 in Figure 6, the next step in Method 1 is to carry out a standard factory PIM test on the AUT again 616, to confirm that the repairs have been successful. If the AUT passes this test, it can be returned to the main production line to continue to the next stage in the assembly process. Otherwise, it may be necessary to return to step 620 to repeat the faulty branch identification and rework process until all PIM faults have been eliminated.

[00243] Method 2: Scalar PIM Measurements at Multiple Down-tilt Settings

[00244] It may be possible to improve upon Method 1 by repeating the Forward PIM measurement process at multiple down-tilt settings, and analysing the change in the PIM response of the AUT from one down-tilt setting to the next. A flowchart summarizing the steps involved in this alternative approach (hereafter referred to as“Method 2”) is presented in Figure 10.

[00245] The first four steps in Method 2 are substantially identical to Method 1 , namely: place the AUT in an anechoic chamber and connect the AUT to the test apparatus 1002; configure the PIM analyser’s sweep settings 1004; initialize the AUT’s down-tilt to a known value 1006; and use DTP testing to ensure that the input section of the AUT is free of PIM faults 1008, 1010, 1012.

[00246] Once the input section of the AUT has been declared PIM-free, a standard factory PIM test may be performed as step 1014. If the AUT passes this test, as determined by step 1016, it can be returned to the main production line to continue to the next stage in the assembly process, thus ending the process shown in Figure 10. Otherwise, the PIM troubleshooting procedure moves on to step 1018.

[00247] The Forward PIM sweep through the AUT, starting at step 1018, may also be substantially identical to Method 1 except that the measurement process is repeated at multiple down-tilt settings across the AUT’s full down-tilt range.

[00248] Upon completion of the measurements on the AUT, the raw PIM data from each radiating element is post-processed by the system controller 1034, which then estimates which branch is the faulty one and reports the result to the user. In a similar fashion to Method 1 , when analysing an AUT with an architecture like that of Figure 16, it may be advantageous to treat elements that share a common branch (for example, 1608a and 1608b) as a single logical element. This may be achieved by summing together the measured PIM levels emitted by such elements, thereby obtaining a list of the total Forward PIM power emitted by each branch of the AUT. From this point onwards, the algorithm to identify the faulty branch proceeds exactly as it would for antennas with architectures similar to those shown in Figure 1 and Figure 2.

[00249] The raw PIM data that is collected using Method 2 can be analysed in several different ways. The final choice of which analysis algorithm to use will often be based on practical experience with a given model of AUT, or derived from a detailed computer simulation of the AUT. Some algorithms may return more reliable results for a given AUT model than others, depending on the specific differences in antenna architecture from one model to the next.

[00250] One analysis strategy is to measure the maximum change in the average PIM emitted by each radiating element across the AUT’s full range of down-tilt settings. The branch that experiences the largest variation in PIM with down-tilt is most likely to be the faulty one. An example of this is shown in Figure 1 1 , which was generated from the same Microwave Office™ (registered mark of AWR Corporation) simulation that was used for the example of Method 1 . The bar graph in Figure 1 1 displays the average PIM emitted by each radiating branch in a 7- branch phased array antenna with a single PIM fault in Branch 2, at down-tilt settings ranging from 0° to 10°. It is clear that Branches 1 -3 exhibit large changes in PIM over down-tilt, whereas Branches 4-7 exhibit almost no change at all. The largest change occurs in Branch 2 (the faulty branch), with an 1 1 .6 dB variation in average PIM level across all down-tilts. Thus, Method 2 has correctly identified the faulty branch, where Method 1 has failed to correctly identify the faulty branch.

[00251 ] Another analysis strategy that can be used with Method 2 is to record the maximum value of the average PIM emitted by each radiating element across all downtilt settings.

[00252] After the suspect branch in the AUT has been re-worked in step 1036, the next step 1014 in Method 2 is to carry out a standard factory PIM test on the AUT to confirm that the repairs have been successful. If the AUT passes this test, it can be returned to the main production line to continue to the next stage in the assembly process. Otherwise, the AUT may be returned to step 1018 to repeat the faulty branch identification and rework process until all PIM faults have been eliminated.

[00253] Method 3: Scalar PIM Measurements with a Low-PIM Mismatch added to AUT

[00254] Another way of improving Method 1 may be to deliberately introduce a large impedance mismatch into one or more branches of the AUT. This strategy (hereafter referred to as“Method 3”) exploits the sensitivity of PIM to load match, a well-known phenomenon that has been documented by previous workers in the field [1 ] An impedance mismatch at the end of an antenna branch can dramatically affect the amplitude of the PIM generated in that branch. This effect is further amplified if the source end of the branch is also mismatched, a common feature of many phased array antennas in the cellular industry.

[00255] A flowchart summarizing the steps involved in Method 3 is presented in Figure 12.

[00256] Steps 1202 to 1216 in Method 3 may be substantially identical to Methods 1 and 2, namely: place the AUT in an anechoic chamber and connect the AUT to the test apparatus 1202; configure the PIM analyser’s sweep settings 1204; initialize the AUT’s downtilt to a known value 1206; and use DTP testing to determine whether the input section of the AUT is free of PIM faults 1208, 1210, 1212.

[00257] Step 1218, the Forward PIM sweep through the AUT, may be substantially identical to Method 1 except that the measurement process is repeated multiple times, with a low-PIM mismatch added to a different branch in the AUT each time.

[00258] The low-PIM mismatch could simply be a highly reflective object, like a metal plate or block of high-dielectric constant material, that is placed onto a radiating element in order to reflect the radiated signals from that element back into the AUT. As the name implies, the low PIM mismatch should be made of a non-PIM-generating material, so that it does not pollute the PIM response of the AUT. If a metal plate is used for this purpose, care should be taken to prevent it from coming into contact with any metallic components in the AUT, as this can be a strong source of PIM even if the individual materials are low-PIM.

[00259] Note that when performing a Forward PIM sweep using Method 3, only the radiating elements in branches that do not contain the low-PIM mismatch are measured. The Forward PIM emitted by the radiating element in the branch that contains the low-PIM mismatch cannot usually be measured, as most of the energy in that branch is reflected back into the AUT.

[00260] Upon completion of the measurements on the AUT, the raw PIM data from each radiating element is post-processed by the system controller, in step 1232, which then estimates which branch is the faulty one and reports the result to the user. Note that when introducing a low-PIM mismatch into a phased array antenna, the amplitude of the PIM signal emitted by each radiating element can change in a complicated and sometimes dramatic fashion, depending on where the low-PIM mismatch is placed and what down-tilt setting is used. Consequently, inferring which is the faulty branch based on a set of measured PIM responses is not always straightforward. As was the case for Methods 1 and 2, the interpretation of the PIM responses obtained with Method 3 will often have an empirical aspect to it, usually involving heuristics derived from test data from previous antenna units, or from a detailed computer simulation of the AUT. Alternatively, an adaptive software algorithm could be developed to identify the faulty branch based on the measured PIM responses.

[00261] In an alternative version of Method 3, further insight into the likely location of the PIM fault may be gained by repeating all of the steps in Method 3 at several different down-tilt settings. This may provide additional data from which more effective heuristics - or a more effective adaptive algorithm - could be derived, thereby improving the success rate of this method.

[00262] After the suspect branch in the AUT has been re-worked, in step 1234, the final step in Method 3, step 1214, is to carry out a standard factory PIM test on the AUT to confirm that the repairs have been successful. If the AUT passes this test, it can be returned to the main production line to continue to the next stage in the assembly process. Otherwise, it may be necessary to return to step 1218, to repeat the faulty branch identification and rework process until all PIM faults have been eliminated.

[00263] Vector Fault-finding Methods

[00264] If the test apparatus is capable of performing vector PIM measurements (i.e. measurement of Forward PIM phase as well as amplitude), a wide range of sophisticated postprocessing algorithms may be available for estimating which is the faulty branch. These postprocessing algorithms may offer superior precision over scalar techniques, and may be configurable to detect multiple PIM faults simultaneously.

[00265] Method 4: Comparative Group Delay Measurements

[00266] One type of vector PIM troubleshooting method (hereafter referred to as“Method 4”) involves measuring the group delays of the PIM signals emitted by the radiating elements of the AUT. PIM group delay is defined as the gradient of the measured PIM phase response with respect to frequency, hence the need for a vector PIM measurement apparatus. By comparing the PIM group delays across all branches, the faulty branch can be identified.

[00267] The flowchart for Method 4 is essentially the same as the one for Method 1 (see Figure 6), with differences being that the Forward PIM sweep of Method 4 entails the measurement of PIM phase as well as amplitude, and the post-processing algorithm of Method 4 entails the calculation of PIM group delay rather than its average amplitude over frequency.

[00268] Method 4 may be best suited to cases where the AUT contains only one dominant PIM fault, and the AUT is comprised solely of components with little or no dispersion (which may be a reasonable assumption for many broadband antennas used in the cellular industry). Under these conditions, the PIM group delay will have the following properties.

[00269] In general, the group delay of the PIM signal emitted by the radiating element (or elements, in the case of antennas with architectures similar to the architecture shown in Figure 16) in the faulty branch will exhibit a distinct variation over frequency compared to the other elements, due to the coherent interference between the forward- and backward-travelling components of the PIM signal.

[00270] The exception to the above rule is if the faulty branch is connected to a splitter with high output return loss on that port. In such cases, the PIM group delay will not exhibit any significant frequency variation compared to the other elements, but can still be distinguished by the fact that its value will be the smallest of any element.

[00271] In branches that do not contain a PIM fault, the group delay of the PIM signals emitted from the radiating elements in those branches will be nearly frequency-invariant.

[00272] As an example, the Microwave Office™ (registered mark of AWR Corporation) simulation described in Method 1 may be used to compute the group delays of the PIM signals emitted by a 7-branch phased array antenna with a PIM fault in Branch 2. The output return loss of the splitter in this example is known to be poor on all ports. Using the above guidelines, it may be expected that the PIM group delay at Element 2 exhibits a noticeable variation with frequency, while the PIM group delays from the remaining elements should be nearly constant with frequency. This prediction may be confirmed by the results of the simulation, shown in Figure 13 which depicts group delays of PIM signals emitted by radiating elements of a 7-branch phased array antenna with a PIM fault in Branch 2. The faulty branch is distinguished by its relatively high group delay variation over frequency, and its low average group delay compared to the other branches. The group delay of the PIM signal from Element 2 exhibits approximately 5 ns of variation over the measured frequency range, while the group delays corresponding to the other elements exhibit almost no variation at all.

[00273] In implementing Method 4, it may be preferable to characterize the phase contribution of the test apparatus in advance, in order to eliminate it from the PIM group delay calculation. There are at least two ways in which this could be achieved, both of which may be familiar to one skilled in the art.

[00274] Firstly, create a purpose-built calibration fixture of known electrical length that generates a PIM signal that couples to every OTA probe in the test apparatus. The group delay of the RF pathways through the OTA probes, diplexer bank and RF switchbox (if applicable) are obtained by measuring the PIM group delay through each OTA probe in turn, and subtracting the known group delay of the calibration fixture.

[00275] Secondly, measure the group delays of the relevant RF pathways in the test apparatus, from the OTA probes to the filter bank output port, using a vector network analyser

[00276] Additionally, some AUTs may contain branches whose cables have been lengthened by an integer multiple of a wavelength in order to physically span the distance between the splitter output port and the radiating element. This may increase the group delay through those branches, potentially creating confusing and misleading results if the PIM fault occurs in one of those branches. To compensate for this, it may be desirable to normalize the measured PIM phase in each branch to the known insertion phase of that branch prior to computing the PIM group delay. These insertion phases may need to be characterised in advance using one of the following three methods: 1 . obtain from original engineering design drawings, calculations or simulations of AUT; 2. perform a characterization sweep using a“golden” (i.e. PIM-free) antenna of the same model as the AUT, with a strong, stable PIM source connected to its RF port; or 3. measure insertion phase directly with a vector network analyser.

[00277] Method 5: Parametric Model Fit to Measured PIM Data

[00278] In an alternative vector PIM troubleshooting method, a parametric model of the AUT is fitted to the measured PIM data. Once the model has been fitted, the locations and amplitudes of the PIM faults in the AUT can be obtained from the coefficients of the model. In principle this approach works equally well with AUTs containing one PIM fault or many, although in practice the finite SNR and sweep bandwidth of the measured PIM data will limit the number of PIM faults that can be reliably detected.

[00279] With a suitably chosen parametric model, it may be possible to develop and refine the necessary heuristics for Methods 1 -4 by computer simulation, rather than by protracted laboratory testing and data analysis.

[00280] A wide range of parametric models and associated fitting techniques are described in the literature, and it will be apparent to a person skilled in the art that one or more of these may be adapted for use with the configurations shown in Figures 3, 4 and/or 5.

[00281 ] In accordance with an aspect of the current invention, a fitting technique, referred to as“Method 5”, is herein described. Method 5 involves the use of a mathematical optimisation technique called“L1 -norm minimization” to fit a parametric model to a set of PIM measurements from the test apparatus. Without being limited by theory, an advantage of L1 -norm minimization in this application may be that it tends to favour solutions which are“sparse”, meaning that the algorithm attempts to find the minimum number of PIM sources that explains all of the measured PIM signals from the AUT [2] In the context of phased array antennas, a sparse solution is strongly preferred over a non-sparse one, as it may be far more common for antennas to fail because of one or two PIM faults rather than hundreds.

[00282] The first step in applying Method 5 is to develop a detailed linear circuit model of the AUT, including all cables, filters, splitters, phase shifters and radiating elements. This information can be gleaned from computer simulations created during the engineering design process, or from vector network analyser measurements on a sample antenna of the same model as the AUT.

[00283] The second step is to compile a catalogue of locations within the AUT where PIM faults are likely to arise, such as solder joints and RF connectors. The precise position of each of these potential PIM faults may then be measured relative to a known datum, such as the nearest splitter port.

[00284] The third step is to generate a set of mathematical functions describing the PIM signals that are generated when the PIM sources in Step 2 are energized. Without being limited by theory, these functions should capture the amplitude and phase characteristics of each PIM signal, referenced to the radiating element in each branch, taking into account the transmission and reflection characteristics of all of the devices in the PIM signal’s transmission path. The mathematical functions preferably also take into account the amplitude and phase changes experienced by the transmit carriers F1 and F2 as they travel from the RF port of the AUT to each PIM source location. Additionally, each function preferably includes both a forward- and backward-travelling component, in order to properly model the observed behaviour of real-world PIM signals. Finally, each function preferably models the amplitude of the corresponding PIM source as a real, frequency-invariant multiplicative constant, so that the final system of equations can be expressed in a form that lends itself to solution by L1 -norm minimization.

[00285] The fourth step is to sum up the mathematical functions for the PIM signals as they appear at each radiating element, in order to obtain an expression for the total PIM signal emerging from each element. Each of these complex expressions will typically contain a contribution from every PIM source in every branch in the AUT, with their amplitudes and phases adjusted to reflect the different transmission paths followed by each signal.

[00286] The fifth step is to cast the complex expressions from Step 4 as a linear system of equations with the following form: [00287] f = ^AX (1)

where

[00288] = complex vector of PIM signals emitted by radiating elements in AUT at a given frequency, as measured by test apparatus

[00289] = matrix of complex coefficients derived from linear circuit model of AUT at a given frequency, taking into account transmission and reflection characteristics of all components in AUT, as well as pre-selected list of likely PIM fault locations

[00290] ^:¾' = real vector of PIM fault amplitudes, assumed to be frequency-invariant

[00291] Equation 1 describes the complex PIM response of the AUT as a function of the linear transmission and reflection characteristics of the components in the AUT assembly, and the amplitudes of the PIM faults at the pre-selected list of locations in the AUT. However, it should be emphasized that Equation 1 only applies to a single frequency point. While in some cases this may be sufficient information to allow Equation 1 to be solved for the PIM amplitudes

, in general it is preferable to use PIM measurements from multiple frequency points when fitting a parametric model, as it improves the noise immunity and overall robustness of the solution. In order to extend Equation 1 to accommodate measurements at multiple frequency points, one can simply stack together the measured PIM vectors and coefficient matrices at each frequency point in the sweep into block versions of themselves, thereby creating the following expanded system of equations:

[00293] where

[00294] ^i: = complex vector of PIM signals emitted by radiating elements in AUT at I ^th point in frequency sweep, as measured by test apparatus [00295] = matrix of complex coefficients derived from linear circuit model of AUT at ^ point in frequency sweep, taking into account transmission and reflection characteristics of all components in AUT, as well as pre-selected list of likely PIM fault locations

[00296] ¾ = real vector of PIM fault amplitudes, assumed to be frequency-invariant

[00297] = number of frequency points in sweep

[00298] Further improvements in the quality of the parametric model can be achieved by constructing multiple versions of Equation 2 over a range of down-tilt settings, and stacking each version together into a block system of equations that can be solved for the vector of PIM amplitudes ¾ .

[00299] In a similar fashion, the quality of the parametric model can be further improved by constructing additional versions of Equation 2, in which a low-PIM mismatch is introduced into one or more of the AUT’s branches, and then stacking each version of Equation 2 together into a block system of equations that can be solved for the vector of PIM amplitudes ¾ .

[00300] Yet further improvements in the quality of the parametric model can be achieved by modifying Equation 2 to include the Reverse PIM response of the AUT, as measured at the RF port of the antenna. This would increase the lengths of the observation vectors by one element each, and add one row to each of the coefficient matrices The vector of PIM amplitudes ^ would be unchanged.

[00301] It may not be necessary for the PIM response of all of the radiating elements in the AUT to be measured in order for Method 5 to work. In principle, it may be possible to obtain the same result from measurements on a subset of the radiating elements. Without being limited by theory, the Forward PIM response of one radiating element alone or the Reverse PIM response of the AUT alone may theoretically provide enough information for Method 5 to converge to the correct solution (although in practice this may result in an algorithm with significantly reduced stability). This allows the fault detection process to be optimised, whereby the sampled radiated elements, frequency points and down-tilt settings are selected to offer the best accuracy in the shortest possible measurement time.

[00302] An example of where this approach might be useful is if a repair technician was observing elevated PIM levels in 3 branches of an AUT out of a total of 7. In this case, rather than measure the PIM response from all 7 radiating elements, the technician might choose to measure only the 3 branches with high PIM, thereby cutting the measurement effort by more than half.

[00303] Returning to the solution procedure, the penultimate step before solving Equation 2 for the PIM amplitude vector ¾ is to separate Equation 2 into its real and imaginary parts, thereby creating two systems of purely real equations, and then stacking the two systems of real equations together in block form as shown:

[00306] Equation 3 can now be solved for the vector of PIM amplitudes using L1 -norm minimisation. A wide variety of efficient L1 -norm minimization algorithms have been published in the literature, many of which are available as Open Source software libraries. A non-limiting example of such an algorithm is Iteratively Reweighted Least Squares (IRLS) (with reference to [3]). Other L1 -norm minimization algorithms, such as Orthogonal Matching Pursuit, may also be used.

[00307] Note that when applying Method 5 to a real-world antenna under test, Equation 3 may prove to be overdetermined, meaning that it has more equations than unknowns. This is a departure from the classical L1 -norm minimization problems reported in the literature, which usually involve underdetermined systems of equations constructed from a relatively small number of random samples of the signal of interest. Such applications are so common that they have been given special names in the literature, such as“compressed sensing” or“compressive sampling”. However, extensive computer simulations by the inventors have found that L1 -norm minimization appears to work equally well with overdetermined systems of equations based on deterministic (as opposed to random) sampling. The total number of samples used in Method 5 is a function of the number of radiating elements, frequency points, down-tilt settings and low PIM mismatch locations at which the PIM is measured. These parameters can be optimized for a given antenna design. [00308] To demonstrate the use of Method 5, consider a simulated 1800 MHz phased array antenna with 7 branches and a topology similar to Figure 1. For this example, it is assumed that all branches are 2 metres in length from splitter to radiating element. Each branch has 3 interconnects that are known to be prone to developing PIM faults: one at the splitter output, one at the radiating element input, and one at the mid-point of the branch. This gives a total of 21 potential PIM fault locations inside the antenna assembly.

[00309] If the antenna develops PIM faults in 3 interconnects, the antenna will fail its factory PIM test. For example, the PIM faults may be in the following locations: in Branch 2, at the splitter output; in Branch 5, at the input to the radiating element; and in Branch 7, at the midpoint of the branch. For the sake of simplicity, it is assumed that all PIM faults have equal severity, i.e. the same third-order intercept point (TOIP). The noise floor of the PIM analyser’s receiver is assumed to be -130 dBm, while the residual PIM of the test apparatus is assumed to be negligible. The simulated Forward PIM response of the antenna across the frequency range 1730-1785 MHz, obtained by sweeping the F1 carrier, is shown in Figure 14.

[00310] The challenge is now to solve the inverse problem, namely: what are the locations and amplitudes of the PIM fault(s) that produced this PIM response?

[00311] Based on the PIM response in Figure 14 and knowledge of the internal construction of the antenna, a system of equations can be constructed according to Equation 3. Solving for the PIM amplitude vector ^ using the IRLS algorithm, the results in Figure 15 may be obtained, showing the estimated PIM faults as a set of black circles overlaid onto a schematic of the AUT. The diameter of each circle is proportional to the severity of the PIM fault at that location. Referring back to the original simulation, the estimated PIM fault locations align with their true locations in the antenna assembly, while the estimated amplitudes are within ±1 dB of the true values, even in the presence of noise.

[00312] In implementing Method 5 as a practical method, it may be necessary to characterize the vector transmission characteristics of the test apparatus in advance, in order to exclude their contributions from the parametric model. One way to exclude their contributions from the parametric model may be to create a calibration fixture with known RF characteristics, which generates a PIM signal that couples to every OTA probe in the test apparatus. The calibration fixture will typically comprise a purpose-built phased array antenna with an architecture similar to that depicted in Figure 1 , with a strong, stable PIM source connected to its input port. By making a series of vector Forward and Reverse PIM measurements on the calibration fixture, the vector transmission coefficient of each RF pathway through the test apparatus can be computed. [00313] Another way to exclude their contributions from the parametric model may be to measure the vector transmission coefficients of the relevant RF pathways through the test apparatus using a vector network analyser.

[00314] Use of Multiple Methods to Improve PIM Fault Detection Accuracy

[00315] In some circumstances it may be possible to identify PIM faults more accurately by using two or more of the methods described above in combination with each other. For example, the troubleshooting process could start with a scalar PIM sweep using Method 1 , so chosen for its speed and simplicity. If the results produced by Method 1 do not clearly and unambiguously identify the faulty branch, then additional measurement sweeps could be performed at different down-tilt settings according to Method 2, until the faulty branch can be discerned with greater confidence.

[00316] In an alternative approach, a vector PIM sweep could be performed at a fixed down- tilt setting, wherein both the amplitude and phase of the PIM signals are measured. The sweep data could then be analysed according to Methods 1 and 4, and the results compared in order to identify the faulty branch. In this way, a single vector PIM sweep allows two independent fault-finding strategies to be brought to bear on the problem.

[00317] The above examples are just two ways in which multiple measurement and analysis methods could be used to improve the accuracy of the PIM fault-finding process. Other combinations of methods may be utilised. The final choice of which methods to use will in general depend on the architecture of the AUT, the specific failure mode of the AUT, and the hardware embodiment upon which the test apparatus is based.

[00318] Near Field Beam Forming Gantry

[00319] Figure 17 is a block diagram of a system 1700 according to an embodiment of an aspect of the invention, which includes a near field focusing gantry. In particular, the gantries depicted in Figure 3, 4 and 5 are replaced by a gantry 1702 comprising a plurality of OTA probes 1704 arranged in a curved arrangement above an AUT 1706.

[00320] Unlike the gantries of Figures 3, 4 and 5, the gantry 1702 and OTA probes 1704 do not need to be near the AUT 1706, or specifically aligned with the AUT 1706 as the system 1706 utilises beam forming to separate the signals from each of the radiating elements of the AUT. Typically, the gantry 1702 and OTA probes 1704 are at least 1 m from the AUT 1706. This has the advantage that is enables a quick setup and adjustment to any antenna design.

[00321 ] The gantries of Figures 3, 4 and 5 have to be strongly coupled to the AUT, which is generally achieved by placing the gantry OTA probes within about 30mm of a respective antenna element of the AUT. Such configuration takes time to setup as the probe elements must be aligned with the elements of the AUT and appropriately spaced. The gantry 1702 does not require any such alignment, and the AUT 1706 may simply be locked in place in the chamber. The inventors estimate that such configuration may reduce setup time from several minutes to around 10 seconds.

[00322] Furthermore, to achieve the strong coupling using the gantries of Figures 3, 4 and 5, the probes need to interact strongly with the elements of the AUT. This may change the return loss of the AUT and disturb the measurement so much to change the PIM response. As a result, the gantry is generally lifted, to avoid disturbing the antenna, the PIM response is then measured, after which the gantry probes are dropped to generate the strong coupling to detect the fault.

[00323] In some cases, the change of return loss may change the PIM levels so much as to hide the PIM source due to cancellation. In more common, less severe cases, magnification of PIM is induced on other branches, reducing the PIM on the faulty branch to the point where it confuses the detection methods.

[00324] The gantry 1702 may completely avoid such problems, as it removes the strong coupling between the probes and the AUT elements by increasing physical separation of the probe elements 1704 to the AUT 1706 and alleviating the need to match each probe to antenna element.

[00325] Without being limited by theory, a number of gantry and probe arrangements may be used in accordance with embodiments of aspects of the invention.

[00326] Typically, anechoic chambers are about 2m high. The size of a sector array is typically up to about 1 .6m at 1800Mhz with 12 to 16 elements. For a probe array the Rayleigh distance, which identifies the far field from the near field which is around 2D ² ,, is 120m. The same limit for each element as the source is 0.48m. Hence, within a 2m chamber, there is a region of far field for each element, but a near field for the probe array as a phased array. As such, near field focusing methods can be used to separate the sources of the signals.

[00327] Figure 18 illustrates an exemplary plot 1800 of an amplitude response function of near field focusing for a single element, where the y axis is the gain in dB and the x axis is the distance along the array under test.

[00328] The plot includes a beam centre 1802, a side lobe level 1804 and a grating lobe level 1806. The beam centre at point 1802 represents the gain from the point of interest, and thus maximum sensitivity is provided at that point.

[00329] The side lobe level 1804 represents the gain from sources adjacent to the point of interest. Ideally, minimum gain is desirable at the side lobe level 1804, however this is limited by the aperture width of the probe array and how far it is from the surface of AUT.

[00330] The grating lobe level 1806 is due to aliasing, due to under sampling in space. For the same aperture, if more elements are added with smaller spacing, the grating lobe 1806 moves away from the area of interest up to l/2 separation where it would disappear. Using values larger than l/2 allows optimization of the system to use minimum number of components.

[00331 ] It is desirable to position the OTA probes 1704 for best side lobe ratio, beam width i.e. resolution, and grating lobes (spatial Aliasing). Processing can be performed using match filtering and/or Kirchhoff migration for non-zero offset sources.

[00332] Figures 19-22 illustrate various OTA probe arrangements and their associated amplitude responses, in accordance with embodiments of aspects of the present invention.

[00333] Figure 19a illustrates a flat OTA probe arrangement where a plurality of OTA probes 1902 are arranged approximately 1200mm from corresponding elements 1904 of an AUT in a linear arrangement. 16 OTA probes 1902 are spread out over a length of 2000mm, directly above the AUT. The AUT comprises 16 elements 1904 spread out over a length of 1500mm.

[00334] Figure 19b illustrates amplitude response functions of near field focusing for each of the 16 elements 1904 using the OTA probe arrangement of Figure 19a. As can be seen from Figure 19b, there is significant overlap of the amplitude response functions and less than 8dB target isolation is achieved.

[00335] Figure 20a illustrates a curved OTA probe arrangement where a plurality of OTA probes 2002 are arranged approximately 1200mm from corresponding elements 2004 of an AUT at a central portion, and about 630mm at outer edges. 16 OTA probes 2002 are spread out over 2000mm, directly above the AUT, and are spaced equally in the X-direction (i.e. along a length of the AUT). In other words the OTA probe arrangement of Figure 20a is the same as that of Figure 19a, but with the probes 2002 offset in the y-direction. The AUT comprises 16 elements 2004 spread out over 1500mm.

[00336] Figure 20b illustrates amplitude response functions of near field focusing for each of the 16 elements 2004 using the OTA probe arrangement of Figure 20a. As can be seen from Figure 20b, target isolation is improved, but grating lobes and beam centres are close to each other, resulting in less than 8dB target isolation being achieved.

[00337] Figure 21 a illustrates a curved OTA probe arrangement where a plurality of OTA probes 2102 are arranged on an elliptical curve, approximately 1200mm from corresponding elements 2104 of an AUT at a central portion, and about 600mm at outer edges. 16 OTA probes 2102 are spread out equally spaced along the elliptical curve (rather than being equally spaced in the x-direction), and directly above the AUT. The OTA probes are spread out over about 1600mm in the X-direction (i.e. along a length of the AUT). The AUT comprises 16 elements 2104 spread out over 1500mm.

[00338] Figure 21 b illustrates amplitude response functions of near field focusing for each of the 16 elements 2104 using the OTA probe arrangement of Figure 21 a. As can be seen from Figure 21 b, about 10dB target isolation can be achieved with a minimum number of (16) elements.

[00339] The inventors have found that an ellipse having the following parameterisation is particularly suited for a probe array:

where f is from 0 to TT, W is the width of the curve, h is the height of the upper most point of the curve above the AUT, and b is the height of the curve from vertically, from its upper most point to its lowermost point.

[00340] In larger anechoic chambers (e.g. 3m+ high), a larger number of probes may be used in an elliptical arrangement to provide even better target isolation. This is possible as a larger ellipse according to the above parameterisation can physically fit in the chamber.

[00341 ] Figure 22a illustrates a curved OTA probe arrangement where a plurality of OTA probes 2202 are arranged on an elliptical curve, similar to the arrangement of Figure 22a, but with 30 OTA probes 2202. The probes 2202 are spread over the same shaped elliptical curve as the arrangement of Figure 21 a (i.e. according to the parameterisation above), but at a greater size (3000mm in the X-direction).

[00342] Figure 22b illustrates amplitude response functions of near field focusing for each of the 16 elements 2204 using the OTA probe arrangement of Figure 22a. As can be seen from Figure 22b, more than 16dB target isolation can be achieved in such arrangement.

[00343] As can be seen from the above, several configurations of embodiments of the invention are able to provide isolation levels of around 10dB or greater, which enables target isolation of each of the signals from the elements of the antennas using near field beam forming.

[00344] Calibration is also used to correct the phase and amplitude variation of each the receiver channels. Calibration is initially performed using a probe element of the counter clockwise polarization, to couple to the clockwise polarised probes. In addition, this reduces the effect of first reflection multipath in the calibration process. This arrangement is set up in five different positions and a response of the system is measured. A calibration coefficient is found for each element at each frequency point, which is later used to adjust measurements.

[00345] Now turning back to Figure 17, the AUT 1706 is fixed to a vibration bench 1708, which enables vibration of the AUT 1706 during testing, avoid the need to excite dynamic PIM faults manually by a technician.

[00346] Dynamic PIM faults in an AUT often manifest only in the presence of mechanical force or vibration, and otherwise test without fault. As such, mechanical excitation, i.e. mechanical force applied to the AUT, is required to detect such dynamic faults. This may be otherwise performed by a technician that is located inside the anechoic chamber while power is applied. This is, however, highly unsafe for the technician, as the technician may be subjected to up to 40W of RF power.

[00347] The vibration bench 1708 thus alleviates the need for a technician to be inside the chamber during testing, as it enables mechanical excitation of the AUT during testing. The vibration bench 1708 also simplifies separation of static and dynamic PIM, as testing may be automatically performed both with and without vibration. This may in turn give further clues on debugging any antenna faults.

[00348] To provide mechanical vibration to the vibrating bench 1708, a motor may be provided on the bench close to the antenna. A standard DC motor, however, contains ferrite cores in the rotor or the linear moving part in linear case. RF incident on such material generates intermodulation products due to the non-linear hysteresis response of the ferrite cores. This generates large signals and renders the system incapable of detecting PIM by masking any signal generated in the antenna.

[00349] In one embodiment, pneumatic actuators may be used to eliminate such interfering signals. In such case, an electric motor of a pneumatic pump may be placed outside of the anechoic chamber and mechanical excitation may be provided using pressurised gas and actuators near the antenna.

[00350] In another embodiment, motors (actuators) may be provided inside the anechoic chamber, but screened to prevent RF power from reaching the core of the motor. In such case, the screening may consist of two parts: conducted and radiated screening. The radiated screening may be performed by enclosing the motor in an aluminium casing to completely shield the inside ferrite core from the RF power. The motor inside the shield moves a ferrite core with weight attached to it, to transfer force to the table (and thus antenna). This arrangement is most advantageous as the motor is completely enclosed by rigid structure and the casing exert the force.

[00351] For the conducted screening a simple pi (TT) low pass filter was found sufficient to attenuate RF signal leaking in to the motor core and stop any PIM signal coming out. In such case, the pi low pass filter includes a coil with two capacitors on either side, and without ferrite feedthroughs as they will produce PIM as much like the electric motor itself.

[00352] The structure does need ground to be connected to make sure no PIM is generated by dissimilar metals. The grounding is provided by capacitively coupling input and output cylinders isolated with a dielectric sheet.

[00353] Resonant frequency excitation is further used, to enable the use of a smaller motor, which in turn reduces costs and simplifies screening. As a result, the motors excite the antenna in certain vibration resonant modes that provides acceleration magnification to generate enough g-force to test the antenna.

[00354] This configuration including screened motors may be more cost-effective than pneumatic actuators, as large motors are needed to power the pneumatic actuators to overcome power loss in the pressure power transferred to the actuators.

[00355] The bench surface may be designed to be a weak structure with low damping loss. Such weak structure may have many resonant modes that may be exploited to magnify the applied force by the motors. The antenna will perturb the modes as the mass of the antenna is comparable to the bench top mass.

[00356] An accelerometer sensor may be provided to enable closed loop excitation. In other words, a controller may operate the motors to achieve resonance, and confirm using the accelerometer that resonance is achieved.

[00357] Figure 23 illustrates a schematic of the vibration bench 2300 with two motor actuators 2305 and an accelerometer sensor 2310, according to an embodiment of the present invention. The bench 2300 may be similar or identical to a bench used in the system 1700.

[00358] The bench 2300 comprises an upper support surface 2315, to which an AUT (not shown) is fastened. The upper support surface 2315 is substantially planar, and is support by substantially vertical legs 2320. The legs 2320 are coupled to the upper support surface 2315 by flexible joints 2325 to prevent the legs 2320 from restricting vibration of the upper support surface 2315.

[00359] The motor actuators 2305 are coupled directly to a bottom of the upper support surface 2315, and thereby enable the actuators to vibrate the upper support surface 2315 directly. The accelerometer sensor 2310 comprises a piezoelectric sensor, also coupled directly to the bottom of the upper support surface 2315, to measure the frequency of the vibration for the resonant modes.

[00360] Now turning back to Figure 17, a vibration bench driver 1710 is coupled to the table 1708 (and thus any sensors 2310 and actuators 2305), and is configured to drive the table at one or more resonance frequencies. The driver 1710 may be configured to pick the lowest frequency mode to minimize the applied excitation.

[00361 ] The OTA probes 1704 are coupled to a coherent receiver/ADC bank 1712 by a plurality of Rx band pass filters 1714. A system controller 1716 is coupled to the coherent receiver/ADC bank 1712 and a PIM analyser 1718, which in turn is coupled to the coherent receiver/ADC bank 1712 by an RTF module 1720.

[00362] The skilled addressee will readily appreciate that previously described embodiments, such as the embodiment illustrated with reference to Figure 5, may be modified to include a coherent receiver/ADC bank, similar or identical to the coherent receiver/ADC bank 1712. In such case, the multi-receiver PIM analyser 502 may be modified to instead include or communicate with such coherent receiver/ADC bank.

[00363] As discussed above, the system 1700 enables static and dynamic PIM to be measured. As a result, different methods of identifying faults may be beneficial, as alternatives or to complement the methods discussed above.

[00364] Fault identification based on Rank Ordering

[00365] Dynamic PIM is a time varying phenomenon, and therefore absolute value-based fault detection methods, such as those described above, are not particularly suited to dynamic PIM faults. Fault identification based on rank ordering may be used in such cases, as a statistical test to identify the dominant fault. Although it has the limitation of providing little information, it generally provides robust results.

[00366] In such case, the order of strength is used to classify the response into fault classes. The classes in this case are single fault at each possible failure point, in addition to the multiple fault as they can be classified a new type of dual source faults.

[00367] Without being limited by theory, rank ordering is scalar technique for detecting PIM fault locations. The method is based on classification techniques based on statistical analysis. Each fault condition is a class to be identified, i.e. a failure at a point or multiple of points is a separate case, and as such, detection of a fault condition comprises finding which class the data corresponds to, which leads to the position (location) of the fault.

[00368] At each frequency, tilt position or swept carrier, the levels at each probe are measured as well as the reflected PIM. The ports are ordered from strongest to weakest, which we call rank. This ranked data is given a suitable weight function to get a vector of numbers to use in detection of fault location. This vector is called in the pattern recognition area a Feature. Hence, we can have N frequency points providing N feature sets. In addition, for each antenna under test tilt position, for total of M tilt positions, another feature set is measured. Also, for every one of three sweeps mode. Hence, we will have 3 NxM feature sets to classify the system and detect the fault positions.

[00369] As outlined above, the absolute level of PIM is not constant in time and cannot be used to compare between elements, however the rank order based on power level is scalar information that can be robustly used with application of a suitable weight function on the rank to produce estimator of faulty branch.

[00370] The rank ordering method is a generalisation of Methods 1 and 2 discussed above with different weighting functions. In particular, Method 1 described above is equivalent to the rank ordering method with the following weighting function: o is max

[00371 ]

elsewhere

[00372] Such configuration basically finds the port of the highest PIM power emitted.

[00373] Method 2 described above is equivalent to the rank ordering method with the following weighting function:

[00374] V (o) = Var(o\ tilt)

[00375] Such configuration basically finds the port of the highest variation in PIM power emitted as a function of tilt.

[00376] The isolation of the ports in the antennas under test is consistent between different antennas under test. Using the rank rather than just the largest element will use the fact that isolation between ports is different depending on the ports, but consistent from antenna to antenna. For example, in Figure 1 the isolation between element 1 and element 2 could be low but the isolation to element 3 or Element N+2 much higher. Hence, if there is a PIM source in branch 1 , there should be higher leakage to element 2 and lower leakage to elements like 3 and N+2. In the same way Element 2 could have low isolation to element 1 and element 3, therefore if elements 1 , 2 and 3 have top rank positions in any order most likely element 2 is the faulty one even if it is not the highest. While element 1 and 2 only are in the top positions and not 3 then you would expect element 1 to be the faulty one. This technique allows the exploitation of even far more complex patterns of this sort.

[00377] The method described above enables a multiple faults case to be identified as single class. This is useful in the most typical multiple faults cases, as the number of combinations increases exponentially, making it impossible to cover all multiple faults situations.

[00378] Finally, based on the result of the method, Voronoi areas can be used to identify the probability for the current measurement likelihood to be in each class. This can be used to identify new fault positions and increate the class set to identify.

[00379] Massive MIMO-like Techniques with Zero Forcing

[00380] Without being limited by theory, in phased array beam forming, the antenna has multiple elements with enough spatial separation to provide separate independent channels to provide the aperture. The signals from the different channels are combined with the proper phasing coefficients to maximise signal strength in certain direction or area in space. Such configuration works well in direct line of sight with little alternative reflectors (or“clutter” as it is referred to in the radar industry).

[00381] Multi path propagation was a problem in such systems, however, developments in the 5G and Massive MIMO has allowed the use of this diversity of reflectors and environment surrounding the transmitter to improve the signal integrity rather than reduce it. This is possible as in such situations the angle of arrival is not important and avoiding channel fading effects and increasing data throughput is most desirable.

[00382] In use, the massive MIMO antenna generates a separate channel to each of the devices (e.g. phones) to which it is communicating. It performs this using uplink pilots and the phasing of the received signal sets up a bespoke beam coefficient for each device to maximize SNR to the phase and nulls in the positions of all other devices, using pseudo inverse of the received coefficients. The condition for this procedure to function reasonably is that the number of antennas in the massive MIMO system is larger than the number of phones simultaneously connected. In addition, there is enough diversity in the system to have enough independent channels.

[00383] In embodiments of aspects of the present invention, a Massive MIMO-like approach is adopted, where the pilot signal is replaced by a training procedure where a PIM source is placed in each possible fault position and the signal stored, and zero-forcing (nulling) is used for other PIM sources. For an N-antenna system, the maximum simultaneous sources it can detect is N. A hybrid algorithm based on a scalar method and rank ordering is used to narrow the possible sources and then detect the levels of each source independently.

[00384] Embodiments of aspects of the present invention may provide similar spatial separation to PIM sources as Massive MIMO offers for separating devices (e.g. mobile phones). A simple training method of placing a PIM source and testing the phase of the received signal replaces the pilot signal.

[00385] The method uses all the vector information measured by the system and provides a much more accurate and reliable result.

[00386] The output of this method is the strength of the PIM source at all identified possible sources. Such output may be easier to interpret than amplitude with distance and branch, as provided by other methods.

[00387] The method is ideal when the number of PIM sources is equal to or less than the number of probes. This means it can have nulls for all other sources. For more sources than probes as typically the case in antenna testing, the method is approximate in the sense that it no longer has enough degrees of freedom to create nulls for all other source points when estimating the signal, a certain source point.

[00388] In a system including a 16-probe antenna, only separate 16 sources can be separated at the same time. There can be thousands of source points in the system but there can only be 16 active source points. Embodiments of the present invention include further establishing a set of possible sources using other methods, such as the methods described above, to 16 or less possible sources, after which this technique to resolve the fault from the 15 possibilities.

[00389] The method may in effect be equivalent to phase array focusing with nulls set to all other sources.

[00390] Calibration may be the same as the near field focusing method, outlined above, the main goal being to optimize the isolation between the probes to maximize the numerical stability of the pseudo inverse step.

[00391 ] In certain embodiments, calibration may involve the following steps.

1) Pick a golden antenna of the type to be debugged. The golden antenna needs to be very clean and without PIM sources in it, i.e. the response should be noise floor (or very close to it).

2) Place a PIM source at each possible source point and add label to that response to reflect that position.

3) Measure PIM traces as would be during the test for all frequencies and tilt positions.

4) Repeat steps 2 and 3 for all possible source points.

[00392] Data from the calibration may then be applied to actual test data.

[00393] Machine learning to isolate ghosting problem due to nonlinearity

[00394] Additional complexity to the PIM fault finding is added due to the interaction of the nonlinear generating source with multiple reflections due to return loss inside the antenna under test. This creates the appearance of multiple sources of PIM inside the system even in the simplest scenarios. Machine learning may, however, be used to identify and classify these into an actual source set (or not), to eliminate false detections. Manufacturing tolerances and variability of one antenna to the other is overcome by training on multiple antennas, to train a neural network using supervised learning.

[00395] Without being limited by theory, a brief example 2400 illustrating such ghosting is presented below with reference to Figure 24, in a system including a single PIM source 2405. In this case there are at least two paths that both carriers take to get to the PIM source as shown by first and second paths 2410a, 2410b. Hence the transmit carriers contain the following components.

[00396]

[00397] The first term in each equation represent the first path 2410a and the second term in each equation is the second path 2410b. With a perfect return loss G the second term disappears. For the third product frequency w _g = 2 w ₁ — w ₂ . The PIM generated and measured back at the source point is [00398] where the first term is the return to the source point, the Tx power as based on the order and (*) is the conjugate, giving

[00399] Since the frequencies are close to each other we can approximate coi to be co _r as the separation is usually small compared to carrier frequency

[00400] We can see that there appears to be five sources at

{A-B, A, A+B, A+2B, A+3B}

[00401 ] Figure 25 illustrates a plot of a simulation of the example 2400 of Figure 24, with A = 3000mm and B=1000mm when both tones are swept to obtain the PIM product. As can be seen in the plot, peaks are present at 2000mm (A-B), 3000mm (A), 4000mm (A+B), 5000mm (A+2B), 6000mm (A+3B).

[00402] For cases where coi is held constant while co ₂ is swept all terms in the equation above with coi are constant and do not contribute to ghost source.

{A, A+B/2, A+B, A+3/2 B}

[00403] Figure 26 illustrates a plot of a simulation of the example 2400 of Figure 24, with A = 3000mm and B=1000mm when co l is held constant while co2 is swept to obtain the PIM product. As can be seen in the plot, peaks are present at 3000mm (A), 3500mm (A+B/2), 4000mm (A+B) and 4500mm (A+3/2 B).

[00404] This method is an alternative to Method 5, discussed above. In Method 5, the user needs to provide a complete module simulation response for each of the antenna building block, whereas this method is a black box method in the sense that it obtains this information for the antenna from the training sequence, meaning that the user is not required to know all the antenna details.

[00405] The calibration and training method are the same as for the MIMO method, described above, however requires the user to measure with different sweep arrangements. In particular, the calibration includes the following steps:

1. Sweep carrier 1 and carrier 2 for each point.

2. Sweep carrier 1 and fix carrier 2

3. Sweep carrier 2 and fix carrier 1 .

[00406] As will be readily appreciated by the skilled addressee, while not explicitly mentioned, the above described methods and systems may be used for single or multiband PIM testing.

[00407] Advantageously, the systems and methods described herein enable PIM to be easily located in an AUT, which is particularly useful when manufacturing phased array antennas.

[00408] In the present specification and claims (if any), the word ‘comprising’ and its derivatives including‘comprises’ and‘comprise’ include each of the stated integers but does not exclude the inclusion of one or more further integers.

[00409] Reference throughout this specification to‘one embodiment’ or‘an embodiment’ means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearance of the phrases ‘in one embodiment’ or‘in an embodiment’ in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more combinations.

[00410] In compliance with the statute, the invention has been described in language more or less specific to structural or methodical features. It is to be understood that the invention is not limited to specific features shown or described since the means herein described comprises preferred forms of putting the invention into effect. The invention is, therefore, claimed in any of its forms or modifications within the proper scope of the appended claims (if any) appropriately interpreted by those skilled in the art.

[00411] The following summary of references is for the reader’s convenience, only. It does not form part of the patent document. Even though care has been taken in compiling the references, errors or omissions cannot entirely be excluded. [1] S. Hienonen and A. V. Raisanen, "Effect of load impedance on passive intermodulation measurements," Electronics Letters, vol. 40, no. 4, pp. 245-247, 2004.

[2] E. J. Candes and M. B. Wakin, "An Introduction To Compressive Sampling," IEEE Signal Processing Magazine, vol. 25, no. 2, pp. 21 -30, 2008.

[3] I. Daubechies, R. DeVore, M. Fornasier and C. S. Giintiirk, "Iteratively reweighted least squares minimization for sparse recovery," Communications on Pure and Applied Mathematics, vol. 63, no. 1 , pp. 1 -38, 2010.

[4] Dennis Long staff, Mostafa Abushaaban, "Pattern Recognition by Data Association Distribution Free Method using Rank Ordering," in Proceedings of Defense Applications of Signal Processing (DASP) Workshop, 2011.

[5] Thomas L. Marzetta, Erik G. Larsson, Hong Yang, Hien Quoc Ngo, in Fundamentals of Massive MIMO, Cambridge University Press; 1 edition, 31 October 2016.

[6] M. Soumekh, in Synthetic Aperture Radar Signal Processing with MATLAB Algorithms 1st Edition, Wiley-lnterscience, April 13, 1999.

[7] N. Bleistein, J.K. Cohen, J.W. Stockwell, Jr., "Interdiscplinary applied mathematics," in Mathematics of multidimensional Seismic Imaging Migration and Inversion, Springer, 2001.

[8] Riccardo Massa , Treviglio ( IT ) Vito, Caggiano , Brugherio ( IT ) Gianluca Ratti , Carnate ( IT ), in APPARATUS AND METHODS FOR DYNAMIC PASSIVE INTERMODULATION DISTORTION TESTING, US Patent 9863987-B2, Jan. 9, 2018.

[9] F. Strachan, B. Horsfield, G. Delforce, I. Davies in System and apparatus for locating faults in a cable network, US Patent 9225441