Unwired Planet International Ltd v Huawei Technologies Co, Ltd & Ors

This is another judgment in a mobile telecommunications patent case involving Unwired Planet as claimant and Samsung and Huawei as defendants. It arises from the third of six trials scheduled to take place in 2015 and 2016. All six trials relate to Unwired Planet’s patent portfolio, most of which it obtained from Ericsson. This trial is called “Trial C”. The first five trials relate to the validity, infringement and essentiality of the patents themselves and have been called technical trials, the sixth trial relates to FRAND and competition law. The judgments in the first two technical trials (A and B) are [2015] EWHC 3366 (Pat) and [2016] EWHC 94 (Pat). The issues in these various technical trials are distinct.

Trial C concerns a European patent, EP (UK) 1 230 818 entitled “Method for improving handovers between mobile communication systems”. The patent claims priority from a US filing on 17th November 1999. The patent was granted on 12^th November 2008.

The patent is about handover between different “RATs” or Radio Access Technologies. GSM (2G), UMTS (3G) and LTE (4G) are different RATs because the ways in which the radio signals are used to make the communication link between a mobile phone and a base station are different. GSM is an FDMA/TDMA system. When a phone call is in progress the phone transmits digital signals to the base station at a particular frequency which has been allocated to it and in one of eight time slots at that frequency. Each time slot is 577 microseconds long. The uplink signals from each phone in a cell are on a different combination of frequency and time slot. From the point of view of the receiver, all the radio signal energy at that frequency at a particular time is signal from just one phone. Everything else in that time and frequency is noise. The downlink from the base station to the phones works the same way. UMTS is different. It is a CDMA system. All the phones in a cell transmit in the same broad frequency band at the same time. What distinguishes one signal from another is a scrambling code. Each transmitter has a different code. The codes are orthogonal which means that (in theory) they can all be uniquely distinguished from one another. From the point of view of the receiver, the radio signal energy in a frequency band at all times is made up of all the transmitters “talking” at once. If the receiver wants to “listen” to a particular transmitter it uses the right code to pull out from that overall radio energy the signal sent by that transmitter. Everything else, including all the other transmitters, is effectively noise. The base stations in UMTS (called Node Bs) also broadcast something called the CPICH. This is a “pilot” channel which allows phones to find the signals from that base station using the primary scrambling code for the cell. LTE works in a different way again. Its radio access technology uses OFDMA and SC-FDMA. It is not necessary to know how this works or what these acronyms mean.

In a cellular telecommunications network handover is crucial in order to maintain connections (phone calls or data). The quality of the channel between the phone and the base station in one cell may start to deteriorate but there may be another cell available which can provide a better connection. The phone is handed over from one base station to another. This can happen seamlessly in the middle of the phone call or data connection so that the user does not notice. It may need to happen very quickly. There are different kinds of handover but this case is about handover known as network controlled mobile assisted handover. The decision to handover is made by the network not the phone, but the phones assist the process by using their radio receivers to make measurements of neighbouring cells and then reporting these results to the network so that the network can make a decision.

When the GSM system began, it was generally the only digital mobile telecommunications system in use in a given place. The only handovers in that context were GSM to GSM, in other words intra-RAT handovers. However by 1999 the development of what was to be the next generation of mobile system (UMTS) was well underway. It was clear that phones would be available which were capable of working in GSM or UMTS and there would be a need for inter-RAT handovers in which a phone could be handed over from a GSM cell to a UMTS cell and vice versa.

The case the patent is most concerned with is handover from GSM to UMTS (but it is not limited to that). The claimed invention works in this way. The phone has a GSM connection with a serving GSM base station. The phone makes measurements of local GSM neighbour cells and also local UMTS neighbour cells. The GSM measurements measure different things from the UMTS measurements because GSM and UMTS work in different ways. The invention requires the phone to “convert” the UMTS measurements into GSM measurements, compare the converted measurements to a threshold and if they pass to send converted measurement(s) to the GSM base station. That way the network has the information needed to make inter-RAT handover decisions. For present purposes I am ignoring significant arguments about claim construction.

Unwired Planet contends that the patent is essential to the GSM standard in the context of both GSM-UMTS handover and GSM-LTE handover. Unwired Planet contends that UMTS (and LTE) measurements are converted in the phones and reported to the GSM base stations based on thresholds. Therefore the patent is essential.

The defendants do not agree. They argue that the claimed “conversion” is meaningless. The claims are uninfringeable and the ambiguity is of a kind and severity that the claims are invalid for insufficiency. They deny that the method mandated by the standard falls within the claims but if the way in which the reported measurements are handled in the standard is “conversion” then the claims are obvious over common general knowledge alone and over a prior art document T Doc 1145/99. The obviousness argument is advanced as a kind of Gillette defence in that what is said to be obvious is the way the standards operate.

In this judgment I use the word phone to refer to what is variously called a mobile phone, mobile station (MS), mobile terminal and user equipment (UE).

There are parallel German cases between Unwired Planet, Huawei and Samsung based on the German designation of this patent. They came before the Landgericht Düsseldorf, which gave judgment on 19^th January 2016. The judgments are 4b O 49/14 and 4b O 120/14, for Huawei and Samsung respectively. The German court held in favour of Unwired Planet. The patent was found to be infringed and the court held that there was no cause to stay the infringement proceedings due to concerns about validity. The court rejected the defendants’ arguments of lack of novelty, lack of inventive step and insufficiency. The arguments before the court in Germany are similar to but not identical with the arguments in this case. The evidence is distinct. To try and compare and contrast the German case with this one would just lengthen the judgment even further and I will not attempt to do so.

The claims alleged to be independently valid are method claims 1, 5, 8 and corresponding product claims 9, 13 and 16. Claim 1 (suitably labelled) is in this form:

Claim 9 is to a mobile terminal with means for doing what is set out in claim 1. There is a difference of wording between claim 9 and claim 1 which gives rise to an added matter attack. Claims 5 and 13 require the plurality of downlink measurement values to be RSCP values. Claims 8 and 16 require the control channel to be the SACCH. These terms will be explained below.

Unwired Planet called Dr Howard Thomas as an expert. Dr Thomas is employed by JDSU Limited as a principal research engineer and also is a consultant with Hillebrand Consulting Engineers. He has a PhD in mobile radio propagation from UCL. After his PhD, he worked at Communication Research Laboratory in Japan (1992-1994), Motorola (1994-2011), Nokia (2011-2013) and JDSU Ltd (2013 to date). He has been involved in the 2G, 3G and 4G systems, in particular in relation to handover.

The defendants called a single expert, Mr Richard Townend. Mr Townend is an electrical engineer with a Masters degree in Electrical and Electronic Engineering with Management from Imperial College, London. He joined BT in September 1997 as a mobility systems engineer. In March 1998 Mr Townend started working on UMTS, specialising in particular in handover and radio architecture. From May 1998 he was BT’s delegate to ETSI SMG2 Architecture Expert Group. This was part of the team which created the UMTS standard. The group became RAN WG3 and Mr Townend was the technical secretary of WG3 from March 1999. In 2003 Mr Townend joined TeleworX Group Inc, a specialist telecoms consultancy. Since 2008 he has worked for his own consultancy in this field.

Unwired Planet submitted that Mr Townend’s evidence needed to be treated with caution for a number of reasons. Only one is worthy of consideration. The point which concerns me is that Mr Townend’s reports gave the impression that T Doc 1145/99 would be viewed as credible and of practical interest to a skilled person when in fact the cross-examination showed that his view was that the main proposal in the document was “off the wall”. Mr Townend should have explained that in his report.

The identity of the skilled person was common ground. They would be a telecommunications engineer with about five years’ experience of measurement reporting and handover technologies.

Arnold J summarised the law on common general knowledge in KCI Licensing v Smith & Nephew [2010] EWHC 1487 (Pat) at paragraph 105-115. This summary was approved by the Court of Appeal ([2010] EWCA Civ 1260, see paragraph 6).

An explanation of the technical background of GSM, UMTS and LTE can be found in my earlier judgments in this overall dispute. This section will only mention the points necessary to understand what comes later and will take the abbreviations as they are without attempting to explain them unless it is necessary. Other points of common general knowledge are addressed as they arise later.

The technical content of paragraphs 3, 4 and 5 above are part of the common general knowledge.

The structure of the land side of the GSM network is described below. The term land side conveys the idea of the fixed part of the network, as opposed to the phone, which itself is mobile and which the land side communicates with over the air interface. The land side of GSM consisted of essentially three kinds of boxes: the MSC, the BSC and the BTS. The BTS is the base station. The BSCs control a number of BTSs and an MSC in turn sits above a number of BSCs forming a hierarchy. The distinction between circuit switching and packet switching and everything to do with GPRS does not need to be explored. The BSC and BTS together can be referred to as the BSS. The GSM network is sometimes called GERAN.

When measurements are made for handover purposes the phone can measure the received signal level of a cell such as a neighbour. This is called RXLEV. RXLEV is a measure of power and its units are dBm, which is explained below. The quality of the received signal can be represented by a value called RXQUAL. RXQUAL is a kind of signal to noise ratio. As a ratio its units are dB. Generally a phone can measure RXLEV for both the serving cell to which it is connected and neighbour cells while it only measures or reports RXQUAL for the serving cell.

In GSM there are two available control channels on the uplink, SACCH and FACCH. These letters stand for Slow Associated Control Channel and Fast Associated Control Channel. The SACCH worked as follows. Together the eight time slots of 577 µsec each form one single frame of about 4.6 msec. A multiframe amounts to 26 frames and takes up about 120 msecs. In the normal case 24 of those 26 frames in a multiframe are used to carry traffic channels. The 26^th frame is empty for various reasons. The 13^th frame is used to send control data. One SACCH message requires four bursts, in other words four of these 13^th frames. So to send one SACCH message takes about half a second (4 x 120 msecs = 480 msecs).

The FACCH works in a different way. Traffic channel frames are used to send FACCH messages. That means those traffic frames cannot be used to send data such as voice data. The traffic frames are said to be “stolen”. A FACCH message needs four frames worth of data sent over five successive frames. So it is much faster than the SACCH. Moreover, owing to the error correction methods which are used it is possible to steal a small number of traffic data bits without sacrificing connection quality at all. However for various reasons the error correction may not be able to compensate for a FACCH message, in which case using the FACCH frequently may degrade connection quality.

The structure of the land side of the UMTS network was also going to consist of essentially three kinds of boxes: the MSC, the RNC and the Node B. The Node B is the base station. The RNCs control a number of Node Bs and an MSC in turn sits above a number of RNCs forming a hierarchy. The UMTS network is sometimes called UTRAN. UMTS works in two ways, FDD and TDD. They are different. The argument has focussed on FDD (aka UTRAN FDD) because that is the predominant system used in the UK but I believe it was common ground that the conclusions applicable to FDD also apply to TDD and so it does not need to be considered separately.

In UMTS measurements can be made of the strength and quality of a received signal. The value RSCP is a measure of signal strength for the cell. The letters stand for Received Signal Code Power. The reason this measure is concerned with code power is because UMTS is a CDMA system and what distinguishes each cell from its neighbour is the code (i.e. the scrambling code). Strictly the value is CPICH RSCP, i.e. RSCP after despreading on the pilot channel. The units of RSCP are dBm.

In UMTS signal quality can be expressed as the ratio E_c/I₀. This is the ratio of code energy to interference. Another measure is the ratio between code energy and noise power spectral density, written as Ec/No. These two are not the same but for the purposes of this case they can be treated interchangeably. Both quantities represent a measure of signal to noise ratio and hence quality. In the judgment I will refer to them both at different places in the discussion to reflect the underlying documents. Strictly Ec refers to energy per chip. A chip can be thought of as a single bit of a given scrambling code. The units in which E_c/I₀ and Ec/No are expressed are dB.

LTE was not common general knowledge at the relevant date. Just as background, in LTE the hierarchy has only two kinds of box instead of three. The base stations are called E-Node Bs. The network is sometimes called E-UTRAN.

The priority date is 17^th November 1999.

The patent starts by explaining that the technical field relates to handover. Paragraphs [0002] – [0003] describe the background under the heading “related art”. Paragraph [0002] explains that GSM has been established for a number of years and UMTS is under development. In order to offer global coverage for UMTS when it starts, it will need to be possible to carry out handovers between UMTS and GSM. In this way, GSM networks will be able to provide coverage in those regions where UMTS network coverage has not, or has not yet, been provided. So when a phone sets up a call in a region where there is only GSM, it will set up a GSM connection but then if the phone moves into an area with UMTS coverage it will be “handed back” to a UMTS network.

Paragraph [0003] states that a basic problem to be resolved for handovers from GSM to UMTS is working out how to transport UMTS measurement information from a phone in GSM to the BSC. The patent explains that currently, the GSM standard provides no spare signalling capacity on the uplink in the relevant circumstances. “Consequently”, states the patent, the transport of UMTS measurement information from a phone to the BSC will have to be performed at the expense of other information.

At paragraph [0004] the patent describes a proposed solution and the drawback of that proposal. The proposed solution is to send the measurement information on the FACCH. However “a significant problem” with the FACCH is that it operates in stealing mode, exchanging speech data for signalling information required for the handover. This would “severely reduce the quality of the speech information being conveyed”. The paragraph ends by stating that the invention “successfully resolves this pressing handover problem and other related problems”.

Paragraph [0005] explains that the “preferred embodiment” (which the reader would understand in this context means what is claimed) provides a method of conveying measurement information in a non-stealing mode. For example, UMTS measurement information can be conveyed from a phone to a GSM BSC, in a GSM message on the SACCH.

Paragraphs [0006] – [0008] describe what the patent calls “important technical advantages” of the invention, namely (i) that measurement information can be conveyed effectively between different mobile communication systems without sacrificing the quality of speech information being conveyed, (ii) that handovers between different types of mobile communication systems can be effectively performed and (iii) that coverage for a UMTS network can be expanded using GSM network coverage.

The patent contains two figures. Figure 1 depicts a phone (22) in a car. On the left of the figure is a GSM network (10) with a central VLR and MSC, a BSC (14) which controls the cell, and a base station (16). On the right of the dotted line 18 is a UMTS cell (20). The Node B for that UMTS cell is not shown. Figure 1 is:

Figure 2 is an exemplary flow diagram:

The detailed description runs from paragraph [0010] to [0017]. The phone is connected to the GSM network. It can make measurements and transmit them to the GSM base station, which then conveys them on to the BSC (paragraph [0017]). So far this is a conventional GSM system.

The phone operating in a GSM cell is capable of making UMTS measurements related to the UMTS cell. It conveys them in measurement reports to the GSM BSC on the SACCH for handover decisions (paragraph [0013]). The patent states (p3 ln32-33): “Notably the UMTS measurement information being reported is different from the typical GSM measurement information being reported.” In paragraph [0014] the patent explains this in more detail:

“For example, a typical GSM MS continuously measures and reports (on the [ uplink ]) signal strength (dBm) and quality (Bit Error Rate or BER) of its own cell, and signal strength of the Broadcast Control Channel (BCCH) carriers of the neighboring cells. On the other hand, instead of measuring signal strength in a UMTS cell, a UMTS MS measures and reports (on the [ uplink ]) the Code Energy-to-Interference Ratio (E _c /I ₀ ) or Received Signal Code Power (RSCP in dBm) of the UMTS cell.”

Consequently in order for a phone to convey UMTS measurement information to a GSM BSC, the UMTS measurement information is “converted to an appropriate GSM measurement format”. The text goes on to state that an example of this will be given in relation to Figure 2.

The description of Figure 2 starts at paragraph [0015]. The large rectangle represents what is going on in a micro-computer in the phone. The phone’s computer retrieves from memory the UMTS measurements the phone has already made. This is shown in box 102a by reference to E_c/I₀ but the text makes clear it could be some other value such as RSCP. At step 102b the phone’s computer retrieves GSM signal strength information. At step 104, the phone’s computer “converts the retrieved UMTS E _c /I ₀ or RSCP measurement information to appropriate GSM signal strength information, which can be used by the GSM BSC 14 for making handover decisions.” (p3 ln48-49). An example of the sort of “appropriate GSM signal strength information” referred to is RXLEV.

An equation is set out at line 50-57 but the equation is so generic that it does not add anything to the teaching. Nothing more is said in any detail about how the conversion is to be accomplished.

At paragraph [0016] the patent describes two embodiments. The first embodiment is described as follows:

“At step 106, the MS 22 compares the converted UMTS signal strength values (derived from step 104) with the set of stored GSM signal strength values (derived from step 102b). The MS 22 then retrieves a predetermined number (e.g., 6, or the maximum number of neighboring cells included in a GSM measurement report) of the "best" measurement values from step 106 to be reported in a measurement report to the GSM BSC 14. At step 108, for this exemplary embodiment, the MS 22 sends a GSM-type measurement report for receipt at the BSC 14 (via BTS 16) on the GSM SACCH over the radio air interface 23. The GSM-type measurement report can include signal strength information about UMTS neighboring cells.”

The text in box 106 is relevant to both embodiments. It states “Compare converted UMTS values with GSM values and send the best values in the standard GSM measurement result message”.

The first embodiment works in this way. The phone has a set of converted UMTS signal strength values for various UMTS cells and a set of best measurement values for various GSM cells. Normally in GSM the scheme is to report the best GSM values, and normally in the GSM standard at the time there was space in the measurement report to report values for a total of six cells. The phone compares the converted UMTS values against the GSM values. The cells to be reported are the “best” cells in the comparison of the converted UMTS values against GSM values. In the specification, although not in the figure, the word best appears in inverted commas. The skilled reader would understand the writer to be doing this to indicate that best is being used in a figurative sense. This is addressed below on the issue of conversion.

The measurement report which is sent is a GSM-type measurement report. The six cells reported will include both UMTS cells and/or GSM cells depending on the comparison.

The second embodiment is described as follows:

“Alternatively, at step 106, instead of comparing the converted UMTS measurement values with GSM measurement values, the MS can convert each stored UMTS measurement value to an appropriate GSM signal strength value, and select each converted UMTS measurement value that exceeds a predetermined signal strength threshold value, for reporting to the GSM BSC 14.”

In the second embodiment the comparison of the converted UMTS values is against a signal strength threshold. This threshold is a GSM threshold (that is clear, see e.g. the text in box 106). So instead of comparing converted UMTS measurements against actual GSM measurements, a notional standard GSM threshold is used. Counsel for Unwired Planet called this a “notional GSM cell” but Mr Townend was not happy with that description. I will not use it.

The description of the second embodiment in paragraph [0016] does not say so in terms but the reader would understand that the values are then reported in a standard GSM measurement result message just as in the first embodiment (see e.g. the text in box 106).

Paragraph [0017] finishes by explaining that using the method:

“ […] the UMTS measurement information from the MS 22 can be converted to a GSM measurement format and sent to the GSM BSC 14 on the SACCH for use in making handover decisions. Consequently, since the SACCH does not operate in a stealing mode, the quality of the speech and on-line data being conveyed between the MS 22 and the GSM network 10 will not be diminished due to the use of speech frames for measurement signalling.”

The reader would understand that this result happens for two reasons. First, by converting the UMTS measurements into a GSM measurement format, the reporting can be carried out using the existing GSM measurement result message. That message is sent on the SACCH and so the reporting of UMTS measurements can take place on the SACCH. Second, by converting the UMTS measurements into what is described as “appropriate GSM signal strength information” (para [0015] ln49) and “some adequate GSM signal strength” (box 102a of Fig 2) the values can be compared against GSM values. This can be done in the BSC when making handover decisions (para [0015] ln49) and is done in the phone in order to work out which UMTS cells to include in the standard GSM measurement report instead of GSM cells. When the phone does the comparison it may be done using real GSM measurements (first embodiment) or a GSM threshold (second embodiment). Although not stated in the patent expressly for the second embodiment, the reader would understand that both embodiments could result in a measurement report containing both UMTS and GSM cells. If (say) only three converted UMTS measurements passed the threshold test and were to be reported in a standard GSM message report with room for six cells, the phone could report the three best GSM cells in the available spaces.

Counsel for Unwired Planet described this as the invention enabling the phone to “flexibly and intelligently” allocate the limited signalling capacity available on the SACCH such that it was likely that the most useful cells are reported. Making some allowance for hyperbole, this is correct to the extent that starting with a standard GSM measurement report which can include six cells, this does not enable more cells to be reported but it does allow the phone to select from the pool of both GSM and UMTS cells, choose the six “best”, and flexibly allocate space to those six. It is flexible in the sense that next time a report is sent, a different mix of UMTS and GSM cells may be sent.

Lord Hoffmann summarised the law on claim construction in Kirin-Amgen [2005] R.P.C. 9. He stated at paragraph 32 that construction of patent claims “is concerned with what a reasonable person to whom the utterance was addressed would have understood the author to be using the words to mean.” Jacob LJ summarised the effects of the judgment in Kirin-Amgen and the principles to be applied in Virgin Atlantic Airways v. Premium Aircraft Interiors [2010] R.P.C. 8 at paragraph 5.

Claim 1 presents three particular issues of construction which are addressed below. In general, it is worth noting that the method of claim 1 corresponds to the second embodiment described in the specification (comparison with a threshold). The claim relates to the second embodiment rather than the first (comparison with measured values), nevertheless the first embodiment is not irrelevant to construction since it helps the reader understand what is described and claimed.

The claim refers to a first and a second communication system. In that respect it is general in nature but in order to make sense of it, it is easier to think of the case of handover from GSM to UMTS, in other words the first communication system is UMTS and the second is GSM. In the remainder of this section on construction I will consider that case, being aware that the claim is not so limited.

The claim covers a method for conveying UMTS measurements to a GSM system. It has three steps. The first is converting the UMTS measurements to measurement values for GSM (feature [b]). The measurements are of the downlink, in other words the signals sent by the base station down towards the phone. The second step is comparing the converted values with a threshold measurement value (feature [c]). In the UMTS/GSM example under consideration the threshold measurement value referred to in feature [c] is a GSM value. The third step (feature [d]) involves sending the converted values to the GSM network on a control channel (which will be the SACCH in claim 8). Claim 1 refers to sending to a control node, which is probably best seen as the BSC but need not be. The sending of values in feature [d] is subject to a condition that the measurement value exceeds the threshold. The particular issues on construction are concerned with what conversion means, the nature of the condition in feature [d] and what must be sent at feature [d].

At times in the argument about obviousness it seemed to the defendants that Unwired Planet was relying on limitations not present in the claim. Unwired Planet denied this and the point will come up again in the obviousness section. Since the issue did arise, it is worth addressing briefly. The claims do not require the use of an existing GSM measurement report nor do they require the phone to “flexibly and intelligently” allocate the limited signalling capacity on the SACCH.

This is a key debate in the case and a lot of time was spent on it. It is not a term of art. Both sides submitted repeatedly that the construction of this feature was common ground. That was true in a sense but it tended to obscure the issues rather than illuminate them. It was common ground that one thing conversion was to accomplish was to make the converted measurement values comparable with measurement values from the other system. It was also common ground that this comparability should have the result that a higher number will indicate a better connection for the phone in that cell. Strictly it could be reversed with lower numbers being better but nothing turns on that. The point is that once UMTS values are converted so that they are directly comparable with GSM values, one can rank all the GSM and UMTS cells in a single numerical ranking and read off the ones with the best connection. However from this common ground the parties then part company.

As another preliminary point, it is worth noting that often arguments on claim construction are better addressed in the context of infringement or validity since it is at the point the claim is compared to something else that the issue emerges. There is an element of that in this case but I think it will be clearer to start by considering conversion in the claim without an eye on those debates.

Although one thing conversion is to accomplish is comparability, it is clear that conversion has another requirement as well – formatting. The specification is clear that one purpose of conversion of a UMTS value is to convert it into a GSM measurement format. A format is a scheme for encoding information in a given communications protocol. When a phone measures the received level of signal strength from a cell in GSM that produces a value. In order to convey that value to another entity in the network the two entities have to agree about the format in which the value is to be conveyed. In GSM that value (RXLEV) is encoded or “mapped” into a 6 bit format in the standard GSM measurement report. A 6 bit binary number provides 64 integer values expressed in decimal numbers (0-63). The scheme works as follows. The received power level for a cell can be expressed as a value in watts. A value in watts can also be expressed in dBm. dBm is a logarithmic scale which allows a range of powers which differ by several orders of magnitude to be expressed in a convenient way. For example -90 dBm is 10^-12 watts or 1 picowatt and -30 dBm is 10^-6 watts or 1 microwatt, a power level six orders of magnitude greater. In GSM the default encoding scheme is that values less than -110 dBm are all mapped to zero in the 6 bit format. Values within the range -110 to -48 dBm are mapped to the integers 1 to 62 in 1dBm increments. Values greater than -48 dBm are all mapped to the integer 63. It will be seen that there are 62 steps of 1 dBm each between -110 and -48. So if the phone measures a received power level for a GSM cell of 56 picowatts (which is 5.6 x 10^-11 watts), expressed in dBm that is -72.5 dBm. That value would be encoded in the 6 bit format as 100110 (or 38). The GSM scheme has further aspects (such as a SCALE parameter) but that is not relevant at this stage. Note that this encoding is not arbitrary because, within the range a higher encoded value will represent a higher power and the steps are the same size in dBm (although not in watts). That is commonly done and for good reason but it is not a necessary part of encoding. An encoding scheme could be entirely arbitrary provided both the sender and the receiver know what the rules are. The term RXLEV sometimes means the measured value and sometimes means the encoded integer. Usually there is no problem with this and the skilled person understands, but nevertheless they are two distinct entities. All of this understanding of formats and the GSM encoding of RXLEV was part of the common general knowledge of the skilled person.

There is no reason why the encoding of one value in one communications protocol (say RXLEV in GSM measurement reports) has to correspond to the encoding of a different value in a different protocol (say RSCP in UMTS measurement reports). In fact they are not identical although whether the actual encoding format used for RSCP in UMTS measurement reports was known at the priority date is an issue addressed below. At this stage the point is simply that formats can differ.

It also follows that encoding does not have any necessary connection with comparability. This would be understood and Unwired Planet illustrated the point with an example using Fahrenheit and Celsius temperature scales (bundle X/4).

In the patent, once the UMTS measurement is converted into the GSM format, it can be sent in the normal GSM measurement report on the SACCH. The skilled person would understand that this is part of the formatting objective of conversion.

Turning to comparability, one of the debates was about whether conversion is supposed to render comparable values which are not inherently comparable. Both sides made submissions along these lines but they tend to confuse the issue because it all depends what you mean by comparable in the first place. A convenient place to start is with handover itself. The handover decisions in issue are made by the network. The algorithms are not standardised, they are left to the operators. Very many factors are in play. All other things being equal one might expect that a network would hand a phone over to a new cell if the connection quality in the existing cell was poor and the new cell offered a better connection quality. This is one, common, occurrence. It requires the network to take a view about the likely relative connection quality of the new cell. The information sent in measurement reports is a key part of that assessment. One reason why RXLEV for neighbour cells is reported by phones to the network in GSM is because that value has a utility in making an assessment of likely connection quality. Similarly one reason why values like RSCP and E_c/I₀ described in the patent might be reported to the network in UMTS is because those values have a utility in making an assessment of likely connection quality. This is all common general knowledge.

The relationship between any of these measured values and connection quality is not direct or exact, it is statistical. The skilled person knows that it is not true to say that for all time and in all circumstances, a handover of a phone in GSM from a cell with a given low RXLEV to a cell for which the most recently reported RXLEV is much higher will inevitably provide better connection quality. It is likely to be true but it is by no means necessarily so. For one thing, as Mr Townend explained, the handover decisions often take into account the history of a measured value, distinguishing between a high value which has come from nowhere and a cell with a consistently high value over an appreciable time. There are other confounding factors too. Nevertheless handover decisions have to be made. The decisions are made by computers running deterministic algorithms but the settings and calibration of those algorithms are matters of judgment, involving masses of statistical data and fine tuning. This is all common general knowledge.

Connection quality is not the only consideration either, there are load balancing and traffic volume factors to consider. Furthermore, Mr Townend explained that network operators can classify their cells into conceptual layers for the purposes of handover decisions. So in the same geographical area one could have some large cells which make a contiguous collection and also numerous small cells. Phones travelling at speed (e.g. in a car) can be accommodated by being kept within the layer of large cells whereas slow moving or stationary phones may be handed over amongst the layer of small cells. Again this is all common general knowledge.

Another complicating factor is the different RATs. The skilled person reading the patent knew that once UMTS was up and running, operators may prefer to put phones into the UMTS network and keep them there if possible. The fact that a GSM cell might offer better connection quality at a given instant may not be a sufficient reason to handover a phone with a UMTS connection to a GSM cell. Similarly a phone may be “handed back” to UMTS as soon as a UMTS cell is available which offers connection quality above a given threshold irrespective of the fact that the GSM connection quality is good and maybe even irrespective of the fact that the connection quality might suffer somewhat by the handover.

Despite all these complications however, the skilled person was familiar with the concept of connection quality for a cell. It could be expressed in the statistical models by reference to different metrics. Examples mentioned in the case are bit error rate, frame erasure rate, probability of dropped calls, quality of speech and probability of handover being required. Different people might decide to measure the quality of the cells by using any one or a combination of these values. They tell the skilled person similar but actually different things about the likely experience of a user of the system.

The reported values referred to in the patent (RXLEV in GSM and RSCP and E_c/I₀ in UMTS) are and are understood to be proxies for these metrics. Dr Thomas was criticised in cross-examination for not producing examples of actual statistical models which do this. I do not accept that criticism. What is clear, and I find, is that the skilled person knew that those concerned with designing real handover algorithms could and did relate these proxies to metrics of connection quality by statistical modelling techniques.

Connection quality in general and metrics like probability of dropped calls in particular, are characteristics which apply to cells irrespective of the access technology (GSM, UMTS or LTE). I find that the skilled person knew that in the context of inter-RAT handovers, one of the ways in which a handover decision could be made was by trying to identify the cell with the likely best connection quality, as between a GSM cell and a UMTS cell. Mr Townend accepted that this was one of the ways an inter-RAT handover decision could be done. The algorithm would typically be looking at both sets of measurements, in other words RXLEV for GSM and some measurement from UMTS such as RSCP, E_c/I₀ or both if they were available.

This is important because it supports the inference that although GSM measurements like RXLEV and UMTS measurements like RSCP are different things, they are and can be evaluated by the same algorithm in order to choose between a GSM cell and a UMTS cell. I find that it is meaningful to say that for a given RXLEV in a GSM cell, a given value of RSCP in a candidate UMTS cell is statistically likely, all other things being equal, to produce a better connection quality. Real handover decisions can include decisions made that way and the skilled person reading the patent knew this. After all at the end of the day that is why these values are being reported in measurement reports in the first place. In this sense RXLEV and RSCP can be compared with one another. Both sides at times submitted that the RXLEV and RSCP were not comparable. The defendants submitted they were inherently not comparable and no conversion could make them comparable. Unwired Planet submitted in closing that they were not comparable until conversion takes place. Neither party is correct in the sense I have described.

There was a point about whether either of RSCP or E_c/I₀ alone is a meaningful value for connection quality. It is clear that for a given level of one, the likely connection quality will vary with the other. For example for a cell with high RSCP, the cell may be very nearly full and E_c/I₀ may therefore be low. Likely connection quality in that case will be lower than the likely connection quality would be of a cell with the same high RSCP and also a high E_c/I₀. However in practice handover decisions can be and are made using one of these values alone. I reject the suggestion, if made, that one of these quantities on its own cannot be used as a proxy for likely connection quality. It can.

This inherent but inexact comparability of values in GSM and values in UMTS explains what the patent is taking about. In the patent specification UMTS measured values are converted into GSM measured values in the phone so that they can be directly compared in the phone in order to report the “best” cells. This figurative use of the word best is understood by the skilled person with all the background common general knowledge I have explained. The skilled person understands that these values are proxies for statistical likelihoods of connection quality. Nevertheless although the inherent comparability is inexact, the patent is teaching the skilled person to rank cells from different RATs by likely connection quality using these values as proxies by putting the values onto a common single scale from one system (GSM in the examples). This allows for a direct comparison and the skilled person understands that this is meaningful, albeit that the correlation will never be perfect and different people will have different views about where to put a given UMTS value against a given GSM value. Neither party used the term “directly comparable” but I believe I am using that term in essentially the same way that Unwired Planet used the term “comparable”.

The ranking only correlates to a probabilistic likelihood of better connection quality. The patent does not pretend that this reporting decision in the phone based on the simple ranking of values is the actual handover decision and the reader would not think that it did. The reader would understand that this conversion to the same scale could be used as part of the handover decision in the network in appropriate circumstances but need not be.

For the reasons explained already, these UMTS and GSM values are inherently comparable in the sense described above and the network does not need to convert UMTS values into GSM values in order to perform the handover algorithms, although it could do so. However in the phone, the direct comparability which is accomplished by the conversion of UMTS values to GSM values has particular utility. The UMTS value is expressed as if it was a GSM measured value, even though it is not. This ranks cells from both RATs on a common scale and allows them to be directly compared. That is what the patent is talking about.

The probabilistic or statistical nature of the likelihood of better connection quality relates to what Dr Thomas was referring to as making the values “broadly” comparable and what was also referred to in argument sometimes as a “rough and ready” comparison. The defendants submitted that there was no evidence that, for example, a UMTS cell with an RSCP of -95.5 dBm will provide a better connection quality than a GSM cell with an RXLEV of -92.5 dBm either statistically or at all. It is correct that Dr Thomas did not attempt to give evidence of that specific nature, but there was evidence from which to infer, and I find, that those operating handover algorithms can use these measurements as statistical proxies for likely call quality and that the skilled person would understand that. Taking the example in the defendants’ submission, such a model could work on the basis that handover will be made to UMTS from GSM if there is a 60% chance of better call quality and could attribute values to the chance that, all other things being equal, a UMTS cell with an RSCP of X will provide a better connect quality than a GSM cell with an RXLEV of Y. For the sake of argument the algorithm could attribute a value of 70% if the UMTS cell has an RSCP of -95.5 dBm and the GSM cell has an RXLEV of -92.5 dBm but only 50% if the RSCP has dropped to -98.5 dBm. A skilled person using such an algorithm could therefore convert RSCPs into GSM values in such a way that an RSCP of -95.5 dBm has a higher value than an RXLEV of -92.5 dBm whereas an RSCP of -98.5 dBm has a lower value. This is just an example and the figures are invented but having heard the evidence in this case, it is fair.

My conclusions on conversion are as follows. In the claims conversion of measurement values associated with the first communication system to measurement values for a second communication system requires two things to occur. The conversion must render the measured values into an appropriate format of the second communication system. That format is the one used for reporting the relevant measurement values in the second communication system. The conversion must also make the converted values directly comparable with measurement values in the second communication system. To make them directly comparable requires the UMTS value to be expressed as if it was a GSM measured value, even though it is not, and thereby allow ranking of the cells from both GSM and UMTS on a common scale.

The words in claim 1 (and claim 9) refer to values exceeding a threshold. The issue is about infringement. The standards to which Unwired Planet contends the patent is essential (i.e. which infringe) set out a test which is satisfied if the measured value is “equal or greater than XXX_REPORTING_THRESHOLD”. One example of this test is at 3GPP TS 45.008 v5.22.0 paragraph 8.4.8.1. There are others. The field XXX_REPORTING_THRESHOLD is a defined threshold in the standard. Each RAT has a different XXX and so can have a different value for the threshold. In mathematical notation the point is that the standards require “≥” whereas the patent claim uses “>”. The defendants submit that the claim cannot properly be construed to cover the standard. Mr Townend supported the defendant’s submission.

Unwired Planet submits that in the context of a set of discrete values, to say that a discrete value is greater than or equal to a threshold is to say the same thing as saying that the discrete value is greater than a threshold one unit lower. This would not be so for continuous variables but Unwired Planet’s point is that it is true for any digital, i.e. discrete, values; which is what the standard is concerned with. For example if one lists integers greater than 7, they are 8, 9, 10 etc.. That is the same as listing integers greater than or equal to 8. They too are 8, 9, 10 etc.. Dr Thomas said the two tests were technically equivalent.

The words in the claim are indeed clear as the defendants point out, nevertheless I find that the skilled reader would understand the proper construction of the claim to include a test such as the one applied by the standard. Art 2 of the Protocol on the Interpretation of Art 69 EPC provides additional support for that conclusion.

The defendants submit that feature [d] of claim 1 is only satisfied if the phone reports all the values which exceed the threshold. The significance of this argument is again concerned with infringement. There are two points. First, in the reporting schemes in the standards which are alleged to infringe there is always a limit on the number of values which can be sent. The precise limit varies and can vary from circumstance to circumstance for a given phone but there is no doubt there will always be a limit and it is not in dispute that a phone could find more measurements that exceed the threshold than could be reported. Second, in the standard there are other criteria which are or can be applied before a value is sent. So a GSM value is only sent even if it passes the threshold if the BSIC (the number identifying the cell) is a valid BSIC. For UMTS values for UTRAN FDD the two values measured are RSCP and E_c/I₀ but only one value is reported. The GSM standard provides for a second threshold applicable to the value which is not reported. The second threshold could be effectively turned off by being set to zero, but if it is in use, then in the case in which the reported value is RSCP and the RSCP exceeds the relevant threshold, it will not be reported unless the corresponding E_c/I₀ also exceeds the second threshold.

The defendants submit the meaning of the claim is perfectly straightforward and is bolstered by the description of the second embodiment. The point on the claim language is the word “said” in the phrase “sending said at least one …” in feature [d]. This refers back to the value which exceeds the threshold and so means that every value which passes the test is to be sent. The key point on the second embodiment is that the sentence at the end of paragraph [0016] refers to selecting each (my emphasis) converted value that exceeds the threshold, for reporting to the GSM BSC. So, the defendants submit, on the face of it every value exceeding the threshold is sent and while that is not true of the first embodiment, this does not help Unwired Planet because the first embodiment is not what is claimed. They submit that Unwired Planet is seeking to change the meaning of the claim so that it covers a system in which (i) you don’t have to send all the values which exceed the threshold – you can cap the number and (ii) you don’t have to send even the capped number of values which exceed the threshold because you can set another threshold which prevents their being sent.

Unwired Planet does not agree. It submits that it is unreal to read the two embodiments in a different way. The skilled person would be well aware that there would be practical limits on the number of cells that could be reported and if there were more cells above the threshold than there was space for in the report, the skilled person would expect that only some would be sent.

Mr Townend supported the defendants’ case and Dr Thomas supported Unwired Planet’s case. Dr Thomas explained that excluding cells for reasons such as there being too many to fit into the reporting message was completely standard. I accept his evidence.

Although it is possible to read the words in claim 1 as referring to sending every value which exceeds the threshold, it is also possible to read the claim as being intended to cover a system which, if it finds at least one value which satisfies the test, will send that value. Read that way the claim is indifferent to what happens if more values satisfy the test.

I prefer the alternative construction of the claim and reject the defendants’ submission, for the following reasons. The skilled person would think it odd that the patentee might be using these words to exclude an arrangement in which a limited set of values which satisfied the test was sent in a case in which many passed the test. Part of the purpose of what is described in the specification is a concern about limited space in which to send reports and part of the point of what is described is to be able to send the reports on the SACCH and to use the standard GSM message. This applies to both embodiments. In these contexts space is limited. The words in the sentence in paragraph [0014] do contemplate sending all the values which pass the test but that is not what this part of the teaching is really concerned with. It is really drawing the contrast with the first embodiment that it is converted values which pass the predetermined threshold test which are reported, rather than values compared to actual GSM measurements.

Unwired Planet contends that the Patent is essential to the GSM standard. Two versions of the standard are relied on:

The standard addresses the process of handover, including handover between GSM and other RATs. Section 3 of the April 2006 version contains an overview of the handover process. This explains (3.1) that measurements will be made by phones and reported to the BSS for assessment. The BSS will also measure the uplink performance for the phone being served and assess signal level interference on its idle traffic channels. Handover strategy (3.4) will be based on reported measurement results and various parameters set for each cell. An example of a basic algorithm is given in the standard. It is an example because the actual algorithms are not standardised. Section 8 of the standard deals with radio link measurements to be used in the handover process. Section 8.1 deals with signal level. Sections 8.1.2 – 8.1.4 deal with GSM and section 8.1.5 deals with other RATs. In GSM the RMS received signal level is measured at the receiver over the range -110 dBm to -38 dBm. The accuracy of the measurement is specified. It is different in different circumstances and at different measured power levels (the specified accuracy between -110 to -70 dBm is different from the specified accuracy between -48 to -38 dBm). In GSM the signal level is RXLEV. Section 8.2 deals with signal quality. In GSM received signal quality is RXQUAL. Section 8.4 deals with measurement reporting. There are two types of measurement reports - “normal” and “enhanced”. Section 9 Table 2 sets out the control parameters used in handover.

The standard requires that the phone measures both signal level and signal quality on the serving cell and signal level on neighbouring GSM cells.

In the context of a potential handover from GSM to UMTS, the serving cell will be a GSM cell and so the phone will be measuring RXLEV and RXQUAL on the GSM serving cell. The neighbouring cells may be GSM or UMTS cells. In the case of neighbouring GSM cells, the phone will be measuring RXLEV and, in the case of neighbouring UMTS FDD cells, the phone will measure RSCP and Ec/No.

The encoding of measured RXLEV values into the 6 bit format is set out in paragraph 8.1.4. I have described the particular encoding scheme above. The standard also provides that an offset, known as “SCALE”, may be used in the RXLEV mapping process in the enhanced measurement report message, otherwise it is set to zero. The value of SCALE may be 0dB or +10dB. The phone is told by the network which offset value to use by the parameter SCALE_ORD. The default value for SCALE-ORD is zero (i.e. off). The purpose of the SCALE parameter is to cause the phone to record RXLEV measurements at the higher end of the dB range with greater fidelity in certain situations.

The standard specifies two kinds of measurement reports for so called normal and enhanced reporting. The reports are sent on the SACCH. The structure of the reports is different as between normal and enhanced reporting. Unwired Planet’s infringement case relating to the April 2006 standard was based on enhanced reporting and not normal reporting but for clarity’s sake it is more convenient to start here with the normal GSM reporting. The structure of a normal GSM measurement report is:

The report consists of seventeen octets and this map shows what each octet is used for and within each octet what information is conveyed by each bit. The shading was added by Dr Thomas for clarity. After a header, the data for the serving cell comes next. This includes RXLEV and RXQUAL for that cell. The distinction between FULL and SUB is not relevant. The map shows that RXLEV is reported using 6 bits. After the serving cell and a few other irrelevant bits, the report deals with six neighbour cells. The data for a given neighbour cell consists of three fields. There is a 6 bit field for RXLEV followed by a 5 bit field for the BCCH frequency for the neighbour cell and followed by a 6 bit field for the BSIC for the neighbour cell. In the map, when the field straddles two horizontal lines (two octets) the field is divided into a high part and a low part.

The enhanced measurement report is organised very differently. There is a neighbour cell list which places the neighbouring cells in order. The enhanced measurement report uses that order. Mr Townend illustrated how the enhancement measurement report works with the diagram below:

The first bit being set to 1 indicates that neighbour cell measurements are present. The next bit relates to the first neighbour in the neighbour cell list. If that bit is set to 0 then there is no reported measurement for that cell and the next bit will relate to the next cell in the neighbour cell list. That is what is shown in the example in the diagram. In the diagram the third bit from the left therefore relates to the second cell in the neighbour cell list. It is set to 1. That means that the next following six bits will be the encoded 6 bit neighbour cell measurement, i.e. RXLEV for a GSM cell. So if the second neighbour was a GSM cell, the value 011011 (or 27) would be the encoded RXLEV which represents -83 to -84 dBm (assuming SCALE is off or zero). Then the next bit represents the next cell in the neighbour cell list. In the example it is set to 0, for no measurement, and so on.

This explanation deals with how the enhanced measurement report is understood by the receiver but does not address how the report is created in the first place. It is created as follows. For reporting purposes the standard contains a scheme for prioritising the cells to be reported. This will be addressed in more detail below. At this stage all it is necessary to know is that all the cells which might be reported are ranked in order of priority. The enhanced measurement report starts as a series of zeros (save for the first bit). The position in the neighbour cell list of the cell with the highest reporting priority is identified and thus the corresponding bit in the enhanced measurement report can be found. That bit is set to one and the six bit reported value is inserted after the bit which is now 1. In effect this process shifts the zero bits downstream to the right. Then the cell with the next highest reporting priority is considered. Its place in the neighbour cell list is identified and again this identifies its place in the enhanced report. The same setting of the relevant bit and downstream insertion of the 6 bit value then takes place. This process continues, inserting values for the cells in order of reporting priority. There is a limit to the number of octets which an enhanced measurement report can consist of and this limits the number of cells which can be reported. Mr Townend explained this in paragraph 116 of his first report as follows:

“using normal measurement reporting, measurement values for a maximum of 6 neighbouring cells can be reported. With enhanced measurement reporting, a maximum of 18 neighbouring cells can be reported, based on a 21 octet measurement report. As a result of the indexing, however, unless all these 18 neighbouring cells are amongst the first cells in the indexed list, the additional bits used for each non-reported cell (as for the 1st and 3rd neighbours in the example above) will reduce the number of bits available for measurement values. It is more usual, therefore, for the enhanced measurement report to include measurements for between 12 and 15 neighbouring cells.”

At paragraph 8.1.5 the standard refers to inter-RAT measurements. For the relevant kind of UMTS (UTRAN FDD) it requires the measurement of RSCP and Ec/No for the neighbour cells. Although two values are measured only one is reported. Which one is to be reported is set by a parameter called FDD_REP_QUANT. The standard states (8.1.5) that the measured value which is to be reported “shall replace RXLEV in the measurement reports”. So in the six bit fields for RXLEV described above, if a cell being reported is a UMTS cell the six bits will be used to report a value of RSCP or Ec/No as the case may be.

To be reported in the six bit field, the values of RSCP or Ec/No are encoded into six bit integers as follows:

There are a number of points to note. There is no SCALE parameter. For Ec/No the integer codes 50 to 63 are unused. Note the difference in units between RSCP and Ec/No. For RSCP, the encoding is similar to but not the same as the encoding of RXLEV. The increments are the same (1dBm). For RXLEV (without SCALE) the integer zero represents -110 dBm whereas for RSCP it represents -115 dBm. So in a sense one scale is offset by 5dBm relative to the other. There is also a relationship between the way the RSCP and Ec/No values are encoded in the GSM standard and the way they are encoded in the UMTS standard for reporting to a UMTS network but that is best dealt with in the context of obviousness.

Normal measurement reporting under the April 2006 standard is not alleged to infringe. There is no need to explain it in any detail. For a GSM only phone working in a single frequency band, the phone reports the six neighbour cells with the highest RXLEV integer values (8.4.1). In multi-band GSM and in multi-RAT reporting, things are more complicated but broadly speaking the cells with the highest values to be reported are reported. In UMTS there can be a relevant second threshold applied to the non-reported value but that is best covered in more detail below. The claim requires reporting of values which exceed a threshold. That does not happen in normal measurement reporting in accordance with the April 2006 standard.

Enhanced measurement reporting in accordance with the April 2006 standard is dealt with in section 8.4.8. With enhanced measurement reporting, the phone is not limited to sending back measurement information for 6 neighbouring cells, but may send back information for a greater number of cells. As explained already, in order to populate the enhanced measurement report, the various cells have to be ranked in order of reporting priority. That is done as follows:

For each of the four priority levels used in enhanced reporting, two additional rules apply:

Thus within priority levels 1 to 3, the phone selects cells within each RAT entirely independently of the measurement values obtained from other RATs. The selection is done from a RAT-specific pool of cells (either GSM serving band; GSM non-serving band(s) or UMTS). Thresholds are applied to the reported values. Only values above the appropriate threshold are reported. Within the levels the ranking is in order of reported value.

At priority level 4 the pool of potential candidate cells is mixed as between GSM and UMTS (and other GSM bands and other RATs). In order to choose which cells to report at priority level 4 the encoded measured values are compared with each other. So the 6 bit integer encoded RXLEV value for a GSM cell is compared to the 6 bit integer encoded RSCP value for a UMTS cell. The higher integer gets a higher priority. The comparison can be affected by offsets, which can be set separately for each kind of cell. At that priority level only candidates that meet the RAT-specific entry requirements can enter the pool – i.e. for UMTS the second threshold can be applied. Even if Ec/No is the value which is going to be reported, in order to assess priority it is the encoded RSCP value which is used in the comparison with other candidates.

There was no dispute about how the standard works as explained above. The disputes are about whether this arrangement falls within the claims. Unwired Planet contends that the operation of priority levels 1 to 3 in enhanced measurement reporting amounts to the use of a method as claimed in claim 1 and a phone which can work this way is within claim 9. The issue can be considered in relation to claim 1 only. If claim 1 is infringed, so too are the other allegedly independently valid claims 5, 8, 9, 13 and 16.

It is clear that feature [a] of claim 1 is satisfied. The method is a method for conveying measurement information from a terminal in a first communication system to a second communication system. Feature [b] is conversion. It is addressed below. Feature [c] is satisfied, subject to conversion. The method compares a plurality of encoded downlink measurement values with at least one threshold measurement value. For feature [d], as I have construed the claim, this is satisfied subject to conversion. The test specified in the standard is that the value is reported if it is greater than or equal to the specified threshold whereas the word in the claim is “exceeds” but as construed, this is an infringement. The space in the report is limited and the method will not report every cell which exceeds the threshold if there is not enough space but as construed above, that satisfies feature [d]. The second threshold test would not avoid the claim when it is used for the same reason. The second threshold is just a qualifying condition like the requirement for a valid cell.

The outstanding issue is whether the method specified in the standard amounts to converting a plurality of downlink measurement values associated with UMTS to a plurality of down link measurement values for GSM. As I have construed the term, conversion has two aspects: formatting and direct comparability. I find that the formatting requirement is satisfied by the encoding of the RSCP and Ec/No measured values into the specified 6 bit format. By being encoded as 6 bits the value can be put into the GSM report in the same field as would be used by the GSM reported value RXLEV.

Turning to comparability, the defendants submitted, supported by Mr Townend, that this was not satisfied while Unwired Planet submitted, supported by Dr Thomas, that it was.

Unwired Planet’s case was based entirely on what takes place at priority level 4. Unwired Planet draws a link between priority level 4 and the first embodiment in the patent. It argues that these are the same. I accept that they are similar but I do not propose to focus on that because the claimed method is the method of the second embodiment not the first.

Unwired Planet did not contend that carrying out priority level 4 fell within the claim as a series of method steps, its case was that the comparison between encoded integers representing values from different RATs at priority level 4 demonstrates that the RSCP measured value has been rendered directly comparable with the encoded RXLEV value by converting the UMTS measured value into a measurement value for the GSM system. The priority assessment at level 4 works by ranking UMTS and GSM cells on the same numerical scale and choosing the highest ranked values to report. So, it is submitted, priority level 4 proves that the encoding of RSCP in the GSM standard satisfies all the requirements for the conversion required by claim 1. Unwired Planet produced a table of the encoded values putting RXLEV, RSCP and Ec/No side by side. It is set out below. Note that the comparison conducted at level 4 is always between RXLEV and RSCP even if Ec/No is to be reported. The table is:

In the table the encoded integer is in the left hand column. The table only applies when SCALE is off and when no offsets are applied. SCALE and the offsets will shift the columns relative to one another. So if SCALE is set to 10 dBm, when it is encoding RXLEV the integer 0 will represent < -100 dBm and the integer 63 will represent >-39 dBm. The offsets work in an analogous way albeit they shift the values in the opposite sense.

For example consider priority level 4 and assume SCALE, the offsets and the second UMTS threshold are all set at zero. Assume also that there are two candidate cells each of which has its reported value above the relevant threshold: a GSM cell with RXLEV of -64.5 dBm and a UMTS cell with RSCP of -70.5 dBm. The encoding in the standard renders these measured values into integer 46 for the GSM cell and 45 for the UMTS cell. At priority level 4 these numbers are directly compared and the GSM cell is given a higher priority. If there is only one more space in the enhanced report then the GSM cell will be reported with its value and not the UMTS cell.

Unwired Planet contends that priority level 4 is a scheme for reporting the best cells chosen by directly comparing converted measurement values. Thus the claimed conversion is undertaken and the method infringes.

The defendants, supported by Mr Townend, profoundly disagree with all this. Summarising the defendants’ submissions may lose the tenor in which they are expressed. The defendants’ case can be seen from the relevant part of their written closing. It is:

“46.

The following points (at least) need to be borne in mind here:

(a)

The so-called ‘offset’ is of course simply an artefact of two different mapping processes. We have seen how the decision on the UMTS mapping process, and in particular on the matters which give rise to the ‘offset’ (i.e. the value reported as 0 and the resolution) were arrived at independently for technical reasons in UMTS reporting, in particular the decision as to which values might be detectable and useful for the network to know. [Townend 1 196-8, not disputed by Dr Thomas and which he accepted in XX T/1/p112-116]. They had nothing to do with the GSM mapping or values. We do not know how or why the decision on the GSM values was taken – it seems to be lost in the mists of time [T/1/p109-110] – but obviously it had nothing to do with UMTS either. On the face of it, it would be astonishing if this process, arrived at for completely different reasons (which could easily have given rise to a different ‘offset’), had thrown up, serendipitously, the mystery ‘equation’ of the Patent [p3 line 55]. Dr Thomas would only say it was ‘not entirely astonishing’ [XX T/1/p122] but his reasons made little sense.

(b)

Mr Townend’s evidence was very clear in his first report (para 211): ‘the encoding of UMTS measurement values described by the Standard…constitutes an independent encoding of RSCP values to a 0-63 range of positive integers which bear no relationship of functional equivalence to any GSM measurement values…the fact that the same integer value may represent (in each separate scheme) signal strength values in dBm which are, in some circumstances, ‘offset’ by 5dBm is coincidental and not the result of a conversion’. He did not resile from any of that in XX.

(c)

Neither Unwired nor Dr Thomas ever attempted to establish by technical evidence, experiment, published or unpublished data, articles or reports, that the mapped values in RXLEV had any functional equivalence to the mapped values in RSCP. This despite the fact that Dr Thomas had claimed that a ‘statistical model’ could be created which would establish how RSCP and RXLEV measurements could be usefully related such that a comparison could be made in terms of connection quality of rival cells [¶35 Reply Report]. In XX Dr Thomas claimed that he chose not to look for the kind of information which might have been used to test his proposition because he was not directed to do so [XX T/1/p93 line 13]. He accepted that some modelling could have been carried out by Universities and other organisations to seek to prove the case [XX T/1/p94 lines 2-8], but that had apparently not been done either.

(d)

Neither did Unwired or Dr Thomas ever seek to establish that any comparability that might exist in any part of the range of measured values was caused or created by the process of mapping which is said to be the act of ‘conversion’ (as opposed to being inherent in the measured numbers themselves). This would have required some form of test indicating that the ‘offset’ of 5dB made the measured values comparable, or at least more comparable than they were in the first place. It was plain from Dr Thomas’ evidence in XX that he did not believe for a moment that this was the case: at no point in the XX on this subject did he suggest that he believed that 5dB made the numbers any more comparable than they were in the first place. See XX T/1/p124 lines 18-24, T/1/p126 lines 16-24, T/1/p127 line 17 – 128 line 2.

(e)

It would have been utterly inconsistent for Dr Thomas to claim any significance for the 5dB offset, given his evidence about the effect of the SCALE parameter which was that mapped values of RXLEV 10dB apart were still ‘good enough’ to be treated as comparable with the same mapped value of RSCP – Reply Report para 81.

(f)

The point in (d) and (e) above is all the more damaging to Unwired’s case because they aver that the ‘raw’ numbers were entirely non-comparable before the mapping process. See XX T2/234 lines 5-8. Apparently, just by adding 5dBm to one number, they become comparable. Adding another 10dBm (see below on SCALE) however does not make them non-comparable again, rather the comparison is ‘ simply thrown a bit off’ [Thomas Reply Report para 81].

(g)

At some points it seemed to be part of Unwired’s case that by fixing lowest detectable measurement as zero in each mapping process, it may be assumed that at least your very lowest mapped values are in some way comparable. This does not work either, whether as a matter of logic or on the evidence in this case:

(i)

As a matter of logic, even if 0 did in each case represent the lowest detectable signal, that does not mean that there is any comparability between connection quality of signals at the lower end of the scale. Even assuming that the 0s were in some way ‘comparable’ in that both represented essentially no detectable signal, we have completely different technologies here. There is no basis in the evidence for assuming that a very low level signal in GSM (say), reported 1, 2 or 3 would represent the same connection quality as a very low signal reported as the same numbers in UMTS (where it could be overwhelmed by noise).

(ii)

In fact, 0 does not represent the lowest measurable RSCP in UMTS. It will be recalled that the lowest value of RSCP being reported to the UMTS network is in fact -120 (mapped for reporting to UMTS as -5).

(iii)

There is of course no evidence as to how the lowest reported number in GSM was arrived at. We therefore cannot assume that it actually represents the lowest detectable measurement anyway.”

Unwired Planet’s case is that the complete answer to all these submissions is the argument over priority level 4. With that in mind I will deal with the submissions as part of consideration of that argument.

So turning to the arguments on priority level 4, the defendants rightly describe it as the “alpha and omega” of Unwired Planet’s case. The defendants’ main written submissions on this are:

“47.

… They [Unwired Planet] say, in effect, simply ‘there is a form of comparison here, therefore the mapped values must be comparable in the sense of the Patent.’ No other point was put to Mr Townend at all on this topic, and his technical evidence on the actual non-comparability of the values was essentially unchallenged.

48.

The case based on Priority Level 4 (which is not part of the standard said to infringe, but only to be evidence that a conversion has taken place) is entirely flawed:

(a)

We have no idea, let alone any evidence, of what thought process the setters of the Standards actually went through when agreeing Priority Level 4.

(b)

The ranking between RXLEV and RSCP in Priority Level 4 is not in fact a straight comparison, because the values are treated differently. You do not pick simply on the basis of the higher measured value. RSCP can only be reported ahead of RXLEV if the Ec/Io value in the cell is above a threshold set by the network. No such second threshold is applied to the RXLEV value. This is clearly inconsistent with the idea that the Standards setters considered a comparison to be inherently indicative of relative connection quality.

(c)

Some predictable form of arbitration between neighbouring cells in different RATs plainly had to be used at this stage in the process [Thomas XX T/1/151/lines 5-10] [Townend XX T/3/p394-401]. This particular method would appear to be easy to implement, and has the benefit that it will always pick the best remaining cells from within at least one and probably both RATs. See Thomas XX 151-153. Since the best cells will be expected to have been reported already at Priority Levels 1, 2 and 3, it may well be that the setters of the Standards are unlikely to have been too bothered about the potential remaining slots.

(d)

As Mr Townend made clear on a number of occasions, the purpose of reporting the lower level cells in the remaining slots is not to provide a genuine ‘contender’ for handover, but rather to provide further information and ‘context’ to the BSC which will inform its strategy both for the MS in question and other mobiles. It is therefore false to assume that the ranking process must be driven by a belief that one cell has a ‘better’ connection quality than another. See e.g. T/3/388-390.

(e)

Very importantly, the scheme in Priority Level 4 cannot tell us anything about element (b) of the definition of ‘conversion’ discussed above – i.e. the causative element. We know that the ‘raw’ data prior to mapping already happened to comprise a similar range of numbers in RXLEV and RSCP. The ‘mapping’ process was also very similar. Even on Unwired’s best case, all that Priority Level 4 could be said to indicate is that the mapped numbers can for some purposes be treated as comparable ‘on a rough and ready basis’ for taking decisions as to reportability. But it cannot indicate that this comparability has been brought about or even improved by the process of mapping. The original numbers may have been just as comparable, or even more comparable.

(f)

The system at Priority Level 4 involves the use of ‘XXX_REPORTING OFFSET’ (see our opening skeleton at 145). If one looks at the mechanism as a whole, it is providing for the network to set up its own bespoke algorithm for favouring UMTS or GSM cells. See Townend XX 3/398. It is not depending on or assuming any kind of fundamental comparability. Unwired seem to be contending that the existence of the offset somehow depends on or proves an underlying ‘comparability’ between the numbers, but this is plainly not so, either as a matter of logic or reality. See Thomas XX T/1/p144-145. See also Townend XX T/3/p395 to 411.

(g)

Priority Level 4 does not strip out the effect of SCALE which would be necessary if it was intending a proper comparison.

(h)

There is no evidence in any T-doc or anywhere else that Priority Level 4 was based on an assumption that the numbers were comparable. It is notable that in Priority Level 4, the words ‘best’ and ‘strongest’ which appear in relation to the comparisons taking place at Priority Levels 1, 2 and 3 (where the values are genuinely comparable as within a RAT) are not used. See Townend XX T/3/p398 lines 14-22.

(i)

In the Later Standard any idea of a direct mathematic comparison of the reported values was dropped altogether, which suggests that it was never thought to be useful. Instead the comparison was made between the extent to which the cells had beaten the individual threshold set for their own RAT.

49.

We submit that there is no evidential basis for finding that the ‘mapping’ of RSCP in the Standards amounts to a conversion within the meaning of the claim.”

Dealing with the main points raised by these submissions: first, a number of the defendants’ submissions involve considering what the thought processes or motives of standard setters in the GSM and UMTS standards may have been. I did not find that a useful approach and I will not embark on considering it. This claim is not concerned with motives. It is either infringed or it is not based on the objective characteristics of the entities and processes concerned. If the UMTS values are being converted as a result of the encoding required by the GSM standard, that is taking place or not taking place irrespective of the motives of the standard setters.

Second, I reject the submission in 48(e) of the closing that the priority level 4 scheme cannot tell us anything about a causative element in the definition of conversion. At priority level 4 a direct comparison between 6 bit integers which represent encoded GSM and RSCP values takes place. Contrary to the defendants’ case, it seems to me that what priority level 4 shows is that the encoding of RSCP into a format which uses a numerical scale which is the same as the scale into which RXLEV has been encoded is indeed what makes that direct comparison possible. It is because these two different values are encoded in the same integers that they can be directly compared.

Third, in my judgment SCALE and the offsets do not assist the defendants. It is correct (48(f)) that the ability to set offsets gives the network operators the ability to adjust the comparison to some extent. The evidence of the experts was clear that the network operator might well use this facility to bias the result of the comparison, for example so as to favour reporting of UMTS cells. The SCALE parameter is analogous although it works in the opposite sense. The reason why none of this helps in my judgment is that if the comparison in issue selects those cells which are most likely to be useful in the handover process as Unwired Planet submits then that is not falsified by the fact that different operators can choose to bias the results one way or another. Given the complexity and proprietary nature of real handover algorithms, it is not inconsistent with utility in handover decision making that SCALE and offsets can be applied.

Fourth, the defendants argue that all the alleged conversion amounts to is a 5 dBm difference between the two encoded scales for RXLEV and RSCP as shown in the table prepared by Unwired Planet which is set out above, and that this is “simply an artefact” of two different mappings created for different reasons. I do not agree. This argument does not face up to the fact that at priority level 4 the two different mappings are directly compared. The real issue is what inference should be drawn from the existence of that comparison. That is addressed below.

Fifth, part of the defendants’ argument is that values like RXLEV and RSCP are not inherently comparable in any sense at all and so cannot be made comparable by conversion. I have already rejected that submission in the construction section. I accept Mr Townend’s evidence that RSCP bears no functional equivalence to GSM measurement values in the sense that they are two different radio technologies and work in different ways. But I have already addressed the point that handover decisions can and do involve making a comparative assessment of likely connection quality in GSM cells as compared to UMTS cells and do this taking into account the RXLEV for the GSM cells and RSCP for UMTS cells. I have also already taken into account that Dr Thomas did not produce an example of a particular statistical model. It was not necessary to do so. Again if Unwired Planet is correct about the inference to be drawn from the comparison at priority level 4 then that represents a complete answer to this point as well.

Sixth, the defendants argue that there is no basis in the evidence to show that very low signal levels at the bottom end of the ranges represent the same connection quality. Moreover, they submit that if Unwired Planet contends that the bottom of the ranges represent the lowest measurable levels that does not prove anything. In any case integer 0 is not the lowest measurable RSCP in UMTS. The lowest mapped level for RSCP in the encoding of RSCP in UMTS standard is -120 dBm. In my judgment these arguments do not add weight to the defendant’s case. Yet again, if Unwired Planet is right about priority level 4 then that answers the first sentence in this paragraph. If the first sentence is dealt with then the others do not advance the arguments.

Unwired Planet contended that the issue on which the case turns is the point in 48(c). The defendants contend the comparison is just there to ensure that there is a repeatable set of rules which produces a predictable result. That was Mr Townend’s evidence. Unwired Planet contends that comparison selects those cells which are most likely to be useful in the handover process.

The answer to this is clear. Priority level 4 is a scheme for selecting those cells to report which are most likely to be useful in the handover process. I reach this conclusion for the following reasons.

I have already addressed the defendants’ submission that there was no evidence that, for example, a UMTS cell with an RSCP of -95.5 dBm will provide a better connection quality than a GSM cell with an RXLEV of -92.5 dBm either statistically or at all. This example can also be used to look at the way the encoding works. Without SCALE and offsets, the RSCP values of -95.5 dBm and -98.5 dBm would be encoded as integers 20 and 17 respectively whereas the RXLEV value of -92.5 dBm would be encoded as 18. So at priority level 4 the GSM cell will be ranked higher than the UMTS cell if the UMTS cell’s RSCP is -98.5 dBm but the UMTS cell will rank above the GSM cell if the RSCP is -95.5 dBm. So the encoding allows the phone at priority level 4 to choose which cell to report by simply comparing the integers.

It is true that cells reported at priority level 4 will not be better than cells in the same RAT reported at one of the higher priority levels. Mr Townend’s evidence was that the cells reported at this level were not genuine contenders for handover but were reported to provide context to the BSC. That requires explanation. I accept that it is unlikely that a cell reported in level 4 will be the cell a phone is handed over to immediately after the report is sent. That is because there will be a better one reported at level 1, 2 or 3 as long as the number to be reported at that higher level is not zero. However the behaviour of candidate cells over time is useful contextual information. A cell reported in level 4 in one report may, some time later, have moved up to become the best cell in its RAT and be reported in level 1. Having the measured value reported at level 4 allows the BSC to have access to this information for the cells most likely to be useful.

It is clear that for a given RAT or GSM band, the rank order of the measured value is a measure of likely connection quality and would be understood as such by the skilled person. The standard itself uses the term “best” expressly in this context at priority level 3 (8.4.8.1). This approach is carried into level 4. Mr Townend’s view about this scheme as just a repeatable scheme was based on part on his opinion that RSCP cannot be meaningfully compared to RXLEV. I have already dealt with that evidence. Moreover the fact that UMTS cells are reported in level 3 based on ranking RSCP alone falsifies any suggestion that one cannot use RSCP alone or Ec/No as a proxy for connection quality of a cell in UMTS. If the second threshold is set to zero, RSCP alone (or Ec/No) will be the only measured value for a cell made available to the BSC making a handover decision to UMTS. Even if the second threshold is set, that does not allow the second value to be reported.

Priority level 4 works in the way it does because the encoding schemes have made the values directly comparable. As Unwired Planet explained with the temperature example, encoding does not necessarily render values comparable. The RXLEV and RSCP values have been encoded in such a way that this comparison can be done. The 5 dBm shift between the two scales is not the only important feature. The resolution of the scale (1 dBm per step) and the range covered by the scale also play a part.

Although it is repetitive to say so, in my judgment priority level 4 is a scheme for selecting those cells to report which are most likely to be useful in the handover process. I find that the encoding of RSCP in the April 2006 standard is a conversion as called for by the claim. Accordingly claim 1 is infringed by carrying out enhanced reporting in accordance with this standard and the patent is essential to the 3GPP TS 45.008 V5.22.0.

The September 2011 version of the standard is a later version of the same standard considered already. It was updated to take into account the development of LTE technology. The Sept. 2011 standard is similar to the April 2006 standard and the same general sections are important.

The signal level/quality measured by the phone of GSM and UMTS neighbouring cells remains the same as under the April 2006 standard. For LTE neighbouring cells, the signal level to be measured is RSRP, which is a power measurement in units of dBm. In addition, the phone should also measure RSRQ, which is a measurement of signal quality in dB.

In the Sept. 2011 standard RSRP and RSRQ are mapped to a 6 bit format just as the other measured values for GSM and UMTS are. The detailed mapping scheme is as follows:

Note that the mapping scheme for RSRP does not use the same increments across the whole range. The ends of the ranges use a 2 dBm increment and the middle range uses a higher resolution 1 dBm increment.

In the same way as the April 2006 standard, the Sept 2011 standard states that the measured values for LTE shall replace RXLEV in the measurement reports. No SCALE parameter is used when mapping RSRP or RSRQ measurement values.

The Sept 2011 standard uses both normal and enhanced reporting but this time Unwired Planet contends that both kinds infringe. That is because a threshold has been introduced into normal reporting.

Under the normal reporting scheme, the following logic applies to a multi-RAT phone:

Thus the logic requires selection of the ‘best’ cell within each non-GSM RAT, with the ‘best’ cell being determined by reference to the highest reported number within that particular non-GSM RAT. There is no comparison of reported values across different RATs.

The logic for multi-RAT reporting in the enhanced scheme is as follows:

If there is not enough space in the report for all valid cells then, for each of the priority levels, the cells are reported for which the result of the following calculation: “reported value – XXX_REPORTING_THRESHOLD + XXX_REPORTING_OFFSET” is highest.

Whilst there are some differences in the implementation of the reporting scheme set out in the Sept 2011 standard, as compared with the April 2006 standard, the basic principles are similar.

Unwired Planet contends that the method of the Sept 2011 standard infringes just as the April 2006 standard did. Dr Thomas produced a table showing the relationship between the encoding of RSRP and RSRQ and RXLEV. It also shows where the changes between 1 dBm and 2dBm offsets occur. The table is:

Having decided the issue of infringement relating to the April 2006 standard, it is only necessary to focus on the differences in the Sept 2011 standard and see if they make any difference.

For normal reporting, the Sept 2011 standard now involves a threshold. That means that normal reporting will also infringe all the claims subject to the issue of conversion.

The differences between the two standards (April 2006 and Sept 2011) which may have a bearing on conversion are that the RSRP encoding scheme uses different increments, the priority ranking comparison in level four may now compare RXLEV to Ec/No or RSRQ and not just RSCP (or RSRP), and the comparison test itself is different.

I am not persuaded that any of these differences take the Sept 2011 standard out of infringement if the April 2006 standard is within the claim. Fundamentally, the level 4 comparison, although it now compares the differences between the encoded level for a cell and the threshold for a given RAT against the equivalent differences for other cells, is still selecting those cells to report which are most likely to be useful in the handover process. The same reasoning on this issue applies as it did before.

The fact that different increments are used for RSRP within the encoding scheme does not make any difference to this conclusion nor does the fact that the comparison may now be between RXLEV and Ec/No or RSRQ. The measured values of these latter two are in different units from measured RXLEV (dB as opposed to dBm) but they are still statistical proxies for connection quality. At level 4 the encoded numbers are directly compared. Also the fact that the highest integer used in the encoded ranges for Ec/No and RSRQ are 49 and 34 respectively whereas RXLEV runs up to 63 does not undermine the conclusion that the exercise is reporting cells most likely to be useful in the handover process.

The defendants pointed out that for a given pair of cells, say a GSM and UMTS cell, if one was to compare the integers encoding RXLEV against RSCP that might favour the GSM cell whereas if on the same occasion one compared the integers encoding RXLEV against Ec/No, that might favour the UMTS cell. So the two comparisons can give different results. If the encoding of both RSCP and Ec/No are both conversions within the claim, how, asks the defendants, can the comparability test be met? Unwired Planet’s answer is that all this example illustrates is the statistical nature of the comparison. I accept that the fact the comparison is only based on likelihoods answers this point to a large degree, but I think another important aspect is that it is up to the network operators to choose which kind of values the phones will report. There is no reason why the statistic model used by a network operator who chooses to require reporting of Ec/No has to give the same results as the statistical model used if RSCP is the reported value.

I find that the encoding of each of the four relevant measured values (RSCP and Ec/No for UMTS and RSRP and RSRQ in LTE) in the Sept 2011 GSM standard is a conversion as called for by the claim. Claim 1 and the other allegedly independently valid claims are infringed by carrying out normal and enhanced reporting in accordance with this standard and the patent is essential to standard 3GPP TS 45.008 V8.12.0.

I turn to consider validity. I will start with the insufficiency issue since it is closely related to the arguments on claim construction and infringement. After that I will address obviousness and added matter.

The defendants contend the claims are insufficient because “conversion” is ambiguous. To address this it is necessary to consider the law with some care first and then apply it to the facts.

A patent specification must disclose the invention clearly and completely enough for it to be performed by a person skilled in the art (s72(1)(c) Patents Act 1977, Art 83 EPC). Different kinds of insufficiency have been identified. One is sometimes called classical insufficiency. There the problem is that the skilled person just cannot make the invention work either at all or without undue effort. Another type is sometimes called Biogen insufficiency. There the problem is that the claim is too wide and is not commensurate with the inventor’s technical contribution. This case is concerned with a third type, ambiguity. Some forms of ambiguity pose no legal difficulty. An ambiguity in a descriptive passage in the specification which causes the skilled person to be unable to make anything at all is a kind of classical insufficiency which falls squarely within the section. However other cases are about ambiguity in the way the invention is defined in the claim. There is no doubt that ambiguities of this latter kind can be a basis for a finding of insufficiency but to put the cases into context it is worth briefly considering the legal background.

Before the EPC and the 1977 Act being passed to give effect to it, UK law included quite an extensive suite of separate grounds for revocation of patents. In the 1949 Patents Act, as well as lack of novelty and obviousness, the grounds for revocation included “insufficiency”, failure to disclose the best method, lack of fair basis, claim ambiguity, inutility, and false suggestion (see respectively s32(1)(h), (h), (i), (i), (g), and (j)). The word “insufficiency” is in inverted commas because its definition in the 1949 Act is not co-extensive with the definition of insufficiency in the modern law. I have used the expression “claim ambiguity” to make clear that the ground under s32(1)(i) 1949 Act was about the claims. That term was not used at the time. Some of the leading cases were of some antiquity and need to be read bearing in mind that the common law of validity was only completely replaced by legislation in the 1949 Act. The cases include Edison and Swan v Holland (1889) 6 RPC 243 on sufficiency, Mullard v Philco (1936) 53 RPC 323 on fair basis, British Thomson-Houston v Corona (1921) 39 RPC 49 on ambiguity, and Lane-Fox v Kensington and Knightsbridge Electric Lighting (1892) 9 RPC 413 on inutility. Although claim ambiguity was a ground of invalidity, it rarely succeeded.

Today what I have called classical insufficiency corresponds broadly to 1949 Act “insufficiency” together with inutility, Biogen insufficiency corresponds to lack of fair basis, while failure to disclosure best method and false suggestion have been abolished completely. The question is then, what happened to claim ambiguity.

To the extent that these pre-1977 Act objections remain part of modern European patent law they are to found in Articles 83 and 84 EPC. The corresponding provisions in the 1977 Act are s72(1)(c) and s14(5) respectively. I will deal with the EPC articles because the sections in the Act mean the same thing. The articles provide:

Article 83

Disclosure of the invention

The European patent application shall disclose the invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art.

Article 84

Claims

The claims shall define the matter for which protection is sought. They shall be clear and concise and be supported by the description.

Both articles use the word “clear” but, at least at first sight, the legislative requirement that a claim must not be ambiguous is in Art 84 (s14(5)) rather than Art 83. Similarly fair basis, now called support, is also in Art 84 and not Art 83. But critically while Art 84 (and s14(5)) are grounds for refusal by the relevant patent office pre-grant, they are not grounds for revocation (Art 138 EPC and s72). This was not an accident, it was a deliberate policy decision. It was thought that patent offices would be best place to be able to regulate the issues which fall into Art 84 and so there was no need to provide for post-grant revocation on that ground. However it is commonplace that difficulties often only emerge post-grant when the patent comes to be litigated. Ever since the EPC, the EPO and the courts have wrestled with the problem this division created.

An important case is Scanvaegt v Pelcombe [1998] FSR 786. Here the Court of Appeal held that the claim was avoidably obscure and contrary to s14(5) of the 1977 Act. Aldous LJ referred to P&G v Peaudouce [1989] FSR 180 in which a patent had been revoked on the 1949 Act ground of claim ambiguity and approved the judgment of Jacob J (as he then was) in Milliken v Walk Off Mats [1996] FSR 292 in which the judge held that there could be no infringement of an ambiguous claim. Aldous LJ said this at p797:

“Ambiguity is no longer a ground of invalidity. That being so, there is a heavy onus upon the relevant Patent Office to ensure that obscure claims are not allowed. Failure to carry out that duty results in an infringement of a fundamental right of the public – to know what they are prevented from doing by a monopoly granted by the State.

I believe that Jacob J is right and that despite the fact that lack of clarity if no longer a matter that results in a patent being invalid, it can result in the patentee being unable to establish infringement. If you cannot define the invention claimed, you cannot conclude that it is being used.”

Scanvaegt was decided in March 1998. As that court will have been well aware, a significant step forward relating to the fair basis aspect of the Art 83/84 problem had been taken in the UK about 18 months earlier in Biogen v Medeva [1997] RPC 1 when the House of Lords, following EPO decisions which had been wrestling with the relationship between Art 83 and Art 84, brought support (i.e. fair basis) into the consideration of validity post-grant. In Biogen the requirement that a wide claim be fairly based on the disclosure (i.e. supported) was held to form part of the requirements for a sufficient disclosure under s72(1)(c). However the House of Lords in Biogen said nothing about ambiguity and Scanvaegt represented an orthodox approach to Arts 83 and 84.

As I mentioned already, if the directions in the specification are so ambiguous that the invention cannot be put into practice, then it is insufficient. Plimpton v Malcomson (1876) 3 Ch.D. 531, which was alluded to by counsel for the defendants for another purpose, was cited in old editions of Terrell for the proposition that it was sometimes impossible to separate ambiguity and insufficiency. This is just as true today as it ever was and so if an ambiguity makes it sufficiently hard to practice what is taught, then the invention is not sufficiently disclosed contrary to Art 83. However claim ambiguity is different. It relates to the scope of the monopoly and can involve the point which the parties referred to in argument as “puzzles at the edge of the claim” or “fuzzy boundaries”. The specification may be a sufficient teaching in the sense that the patentee can make the product or carry out the process and it may be perfectly clear that certain things, including the alleged infringement, are within the claim. All the same the claim boundary may not be a sharp one so that the infringer can think of difficult cases near the boundary. The fact that these difficult cases may or may not infringe does not justify a finding of non-infringement by the clear cases and would not, before the 1977 Act, justify revocation. That was part of the point in British Thomson-Houston v Corona . The claim required the light bulb to have a “large” diameter filament so that the filament could be raised to high temperature without shortening its life. Although “large” is inherently imprecise, the claim did not fail for vagueness (p90 ln12). Skilled people seeking to make the invention work would be able to make it large enough to work.

I can now turn to Kirin-Amgen . Here Neuberger J (as he then was) had held that the claim to a material (rEPO) which was defined as being rEPO with a higher molecular weight than another material (uEPO) was ambiguous because the skilled person did not know which uEPO to use as the standard against which the test the rEPO. In effect the claim assumed all uEPOs had effectively the same molecular weight but this had been shown not to be the case. The trial judge held that the claim was incapable of being infringed and also meant that the invention had not been disclosed clearly enough for it to be performed by a skilled person and was therefore invalid for insufficiency. The Court of Appeal overturned the finding of insufficiency on the basis that this was “lack of clarity dressed up as insufficiency”. The House of Lords restored the judge’s finding. In paragraph 126 Lord Hoffmann disagreed with the reasoning of the Court of Appeal, holding:

“The lack of clarity does not merely create a fuzzy boundary between that which will work and that which will not. It makes it impossible to work the invention at all until one has found out what ingredient is needed.”

Lord Hoffmann went on to address a further point taken in the Court of Appeal. The Court of Appeal held that the claim was sufficiently enabled because a skilled person seeking success would be able to make it work, in a similar way to that described by Lord Shaw in British Thomson-Houston v Corona . Lord Hoffmann explained that the problem with the rEPO claim was of a different kind and therefore what Lord Shaw was addressing was a different situation. At paragraph 129 Lord Hoffmann said:

“In the present case, however, the choice of uEPO has nothing to do with making the invention work. It is simply a criterion against which one tests whether the rEPO falls within the claims. The very concepts of ‘success’ or ‘failure’ seems irrelevant to the choice of uEPO. What counts as ‘success’? Ex hypothesi the skilled person does not know in advance whether any given uEPO will bring his rEPO within the claim or not. From the point of view of success or failure, one is as good as another. All the skilled man can do is try to guess which uEPO the patentee had in mind and if the specification does not tell him, then it is insufficient.”

Thus Kirin-Amgen establishes that an ambiguity in a claim can be a cause of insufficiency. That was because of the nature of the claim in question and the role uEPO played in that claim as a standard against which to judge whether a product was inside its scope or not. Lord Hoffmann was not finding that all ambiguities in claim language necessarily make the claim invalid for insufficiency. He did not address Scanvaegt at all (it was cited, along with Milliken , see [2005] All ER 667 at 671). Lord Hoffmann found that the problem in Kirin-Amgen was not a “puzzle at the edge of the claim”, it was a failure to disclose the invention clearly enough for it to be performed at all because the skilled person could never know if they were within it or not. So, unlike Biogen , Kirin-Amgen did not simply bring an aspect of Art 84 EPC into the test for sufficiency.

A month after Kirin-Amgen the Court of Appeal gave judgment in SmithKline Beecham v Apotex [2004] EWCA Civ 1568. The claim was held not to be infringed but Jacob LJ (with whom Arden and Ward LJJ agreed) went on and considered the issue of ambiguity and insufficiency. This was expressly obiter (see paragraph 114) but clearly highly persuasive. Jacob LJ explained that Scanvaegt and Milliken remained good law but that if “one just does not know what the patentee meant at all”, then a claim would be “truly ambiguous and thus insufficient” on the principles laid down in Kirin-Amgen .

In Sandvik v Kennametal [2011] EWHC 3311 (Pat) Arnold J referred to Kirin-Amgen and SmithKline Beecham to show that ambiguity was an aspect of insufficiency. On the facts, the patentee conceded that the claim was insufficient if it was unclear to the skilled person which “PDF card” to use in the test needed to work out if the product was within the claim or not. The judge held that the case was not merely a one of a fuzzy boundary, the concession was rightly made and the claim was insufficient (paras 162-165). On its facts Sandvik therefore is very like Kirin-Amgen . A technical test has to be carried out to find out if the product is within the claim but the skilled person is uncertain what the correct test is.

In Generics v Yeda [2012] EWHC 1848 (Pat) Arnold said as follows:

“… it is necessary to distinguish between claims that are difficult to construe or that have a "fuzzy boundary" (in the words of Lord Hoffmann in Kirin-Amgen Inc v Hoechst Marion Roussel Ltd [2004] UKHL 46, [2005] RPC 9 at [126]) on the one hand from claims that are truly ambiguous on the other. It is regrettably common for claims to be difficult to construe, but the court will nevertheless strive to give such claims a sensible meaning having regard to the inventor's purpose. It is also common for claims to have a fuzzy boundary, because an integer of the claim involves some question of degree or an imprecise functional limitation. It is well established that is not itself objectionable. If a claim is truly ambiguous, so that it is unclear what is the correct test to determine whether or not a product or process infringes, however, then the claim is insufficient, ….”

[In this case ambiguity was rejected on the facts.]

Pulling all this together, I agree that claims can often be difficult to construe. Sometimes those difficulties are due to avoidable obscurity for which the patentee should get no sympathy, but it can be because trying fairly to describe an invention in words is not always an easy task. I also agree with Arnold J that the existence of a fuzzy boundary in a claim is not objectionable. The contrast is between that and a claim which is truly ambiguous. The factual circumstances in which such a truly ambiguous claim has been identified so far in the modern law ( Kirin-Amgen and Sandvik ) are ones which depend on carrying out a technical test to find out if a product or process is within the claim or not. If the skilled person cannot know whether they are carrying out the right test, then the claim is truly ambiguous and therefore insufficient. That makes sense. However, while the principle cannot be limited just to technical tests, after all SmithKline Beecham was not that sort of case, nevertheless it does not apply simply because one can imagine difficult cases to judge at the edge of a claim. When a defendant has been found to infringe, demonstrating that the claim’s scope is at least clear enough to work that out, an argument that the claim should be regarded as truly ambiguous is likely to be met with scepticism.

Huawei’s and Samsung’s Grounds of Invalidity put the case this way:

“The specification does not contain sufficient directions as to the meaning of “conversion”. There is no sufficient explanation of what constitutes a “conversion” or how any such “conversion” should be carried out or the scope or meaning of “converting” and “converted” as used in the claims. As a result, the skilled reader would be unable to implement the invention, or determine whether they were working the invention, without undue effort or at all.” [Huawei para 3(a), Samsung para 5.1]

The defendants contend that “conversion” is truly ambiguous in the relevant sense. Their case is that the skilled person cannot know whether or not they have carried out a “conversion” that falls within the meaning of the claims. Unwired Planet does not agree. It contends that the skilled person knows what a conversion is and would be able to carry it out. Unwired Planet submits that all the objection amounts to in practice is that there may be difficult cases close to the line, but that is not insufficiency.

The problem relied on by the defendants is that it is clear different network operators will use different algorithms and that the comparison between cells is always statistical in nature. The defendants pointed out that Dr Thomas used the term “broadly speaking” to describe the comparison. The defendants ask rhetorically, how does the skilled person know whether any operation they have carried out on their UMTS measurement values is such that by performing that operation they have arrived in a position where they can compare the figures such that “broadly speaking” cells with a higher number will indicate a better connection quality? The defendants submit that “even Dr Thomas” accepted that whether or not the test was satisfied depended on what was meant by “broadly”.

The defendants argued that for a skilled person to work out whether they had created a conversion within the claim they would need to do the following. First decide how to measure connection quality, effectively choosing a metric such as bit error rate or probability of dropped calls (there are others); second run tests by reference to that parameter to work out what the connection quality would be in different GSM cells at different values for RXLEV; third run the same sort of tests on UMTS cells at different RSCP levels, and fourth deal with the need to take into account E _c /I ₀ in UMTS as well. This would still only produce data for the characteristics of particular cells. To extrapolate to other cells loses accuracy and makes the exercise rough and ready. The best the skilled person could hope for would be some kind of statistical guidance. For example, the statistics might be able to deduce that in the case of a 20dBm difference between mapped RXLEV and mapped RSCP then there would be a 70% likelihood of suffering fewer dropped calls on GSM compared with UMTS. Whereas the statistics might simultaneously show that if mapped RXLEV were only 5dBm higher than mapped RSCP, there was a 60% chance of suffering more dropped calls on GSM compared with UMTS. Subject to the next two paragraphs, I accept the whole of this submission. It is based on the cross-examination of Dr Thomas.

As part of these submissions the defendants contended that Dr Thomas had accepted that the choice of metric, such as bit error rate or probability of dropped calls, could well affect the results obtained in any particular scenario and thus for the same cells you would get a different ranking depending on the parameter chosen. Unwired Planet submitted that the latter point had not been put to Dr Thomas and was not correct. It submitted that Dr Thomas had explained that the process would not involve looking at specific instances but was a statistical process of modelling. The metrics are telling you similar but slightly different things about the likely experience of the user of the system. It was not established that choosing different metrics would give a different outcome.

In my judgment the facts are these. I find that the modelling process would always be statistical. I am not persuaded that the outcome is entirely independent of the metric chosen. In other words I find that there will be circumstances in which the same cells will be ranked differently depending on the metric chosen. However this will only be a problem at the margins, when the likely connection qualities of the cells being compared are similar anyway. If one cell has a much higher likely connection quality than the other, then the choice of metric will not produce a different outcome. This is the point I understand Dr Thomas to have been referring to in his evidence on this about statistical modelling and about getting a good match in the majority of cases.

The defendants argued that the skilled person has no obvious basis upon which to decide: (i) which test for connection quality was the right one to apply; (ii) how reliable to expect the ‘conversion’ to be (is it good enough to be within the claim if you cannot draw any conclusions where the numbers are up to 5, 10 or 20 dBm apart?); (iii) what statistical boundary to apply (is anything over 50% reliability enough, or does it need to be more?). They argued that there was no objective test of “what works” here to apply to the claims.

I reject the defendants’ first argument. The most that can be said is that different statistical models can be constructed which will produce different answers in marginal cases and that the skilled person will know that this is so. However from the point of view of the person using the models, all of them involve the sort of comparison required by the claims. The fact that they produce different answers in some cases is irrelevant. The claims are not defined by reference to the outcome of the comparison itself. How the skilled person chooses to set and adjust the fine details of the comparison is up to them but in all these cases the skilled person would have no doubt that a relevant comparison was being made.

As for the defendants’ second point, the claim leaves it to the skilled person to choose how reliable the conversion has to be. The fact that some schemes will be capable of drawing a conclusion if the numbers are as little as 5 dBm apart whereas other less complex schemes may only resolve cells which differ by 20 dBm is not relevant. Both schemes would amount to conversion.

The third point is similar to the second. The claim leaves it to the skilled person to choose. The fact that some schemes will choose to handover based on a 50% likelihood of better connection quality while others may only handover based on a 70% likelihood is not relevant. Both such schemes would amount to conversion.

I agree with the defendants that this makes the claim of wide scope. That is so in the sense that it is not concerned with trying to arbitrate between different handover algorithms which the skilled person may choose as long as they compare connection quality in the required way. There is nothing surprising about that since handover algorithms were and remain complicated and multifactorial and were not standardised. In practice the multifactorial nature of real algorithms makes no difference. The particular choice of percentage likelihood of better connection quality which is chosen to apply in the algorithm, all other things being equal, is irrelevant.

What matters is that in my judgment the skilled person can implement a scheme which makes this sort of comparison without undue effort. The actual schemes are not at all trivial but they can be and are done. Based on such a scheme there is nothing difficult about then converting UMTS measured values into values for GSM in such a way that the two values can be directly compared in accordance with whatever scheme and for whatever circumstances the skilled person has chosen. The claim does not require the skilled person to actually perform the direct comparison referred to but nothing turns on that. The values can be encoded in the manner required.

In Kirin-Amgen the claim required an SDS-PAGE test to be applied to two proteins in order to decide which had the higher molecular weight. The fact that the test was understood did not remove the ambiguity, namely knowing which uEPO should be used in the test. The defendants submitted that the position was the same here. I do not agree. The claim in this case is different from the one in Kirin-Amgen . This claim does not set its scope based on the outcome of a particular run of a comparison test. It is concerned with a conversion to allow the skilled person to carry out a direct comparison test in accordance with whatever comparison scheme the skilled person has chosen. What the outcome of the individual instances of undertaking the comparison actually is does not matter as far as the claim scope is concerned. The fact that a network operator who converts values to make them directly comparable will produce a scheme which differs in its outcome for a given case from a scheme implemented by another operator is not relevant, nor is the fact that the operators have the facility to alter the way the comparison works.

In my judgment the claim is not ambiguous. The invention in this patent is disclosed sufficiently clearly and completely for it to be performed by a person skilled in the art. This ground of attack is rejected.

A patented invention must involve an inventive step, which means that it must not be obvious to a skilled person having regard to the state of the art (s1(1)(b) and s3 of the 1977 Act, Art 56 EPC). The structured approach to assessing obviousness was set out by the Court of Appeal in Pozzoli v BDMO [2007] EWCA Civ 588. In Medimmune v Novartis [2012] EWCA Civ 1234 the Court of Appeal, Kitchin LJ said at paragraph 93 that the court's task was ultimately “to evaluate all the relevant circumstances in order to answer a single and relatively simple question of fact”; see also Lewison LJ paragraphs 117 — 186.

The following statement of Kitchin J in Generics v Lundbeck [2007] RPC 32 was approved by Lord Hoffmann (with whom the others of their Lordships agreed) in Conor v Angiotech [2008] UKHL 49, [2008] RPC 28:

“The question of obviousness must be considered on the facts of each case. The court must consider the weight to be attached to any particular factor in the light of all the relevant circumstances. These may include such matters as the motive to find a solution to the problem the patent addresses, the number and extent of the possible avenues of research, the effort involved in pursuing them and the expectation of success.”

A number of points of principle emerged in argument. The defendants criticised some of Unwired Planet’s arguments as placing emphasis on limitations which are not in the claim. It is clearly correct that the scope of what is claimed has to be defined correctly when considering obviousness. Laddie J referred to this in Brugger v Medicaid [1996] RPC 635 at p656 ln15-p657 ln8. The question is not whether an embodiment disclosed in the patent is obvious, the question concerns the claim. If a thing or method is obvious and if it falls inside the claim, then that claim is invalid. It is no answer to say that that thing or method does not have some feature found only in an embodiment if that feature is not in the claim.

However there is another aspect of this debate. The defendants’ obviousness case is directed to the standard alleged to infringe. The obvious scheme, contends the defendants, will involve conversion if conversion covers the standards. Unwired Planet says this is an argument like the one in Leo v Sandoz [2009] EWCA 1189 at paragraph 9-10 in which the skilled person is said to “stumble across” the invention by accident. Part of Unwired Planet’s answer to this argument is to highlight factors which, it contends, show why the skilled person would not stumble across a result which happens to end up within the scope of the claim. As a line of argument, in my judgment it is legitimate. The fact that these factors are not limitations in the claim does not matter.

The defendants also relied on Brugger v Medicaid for the proposition that “the mere existence of alternatives is not in itself an argument against obviousness”. I agree that the mere existence of obvious alternatives is not in itself an argument against obviousness. Before Brugger was decided there was sometimes a tendency to place too much emphasis in argument about alternatives as if they always established inventiveness, on the other hand undue focus on Brugger itself can mislead too, discounting alternatives altogether. The correct proposition about alternatives is the statement of Kitchin J in Lundbeck approved in Conor which I have cited above. It is an important statement because it is a balanced one.

To understand both obviousness arguments it helps to see where they are going. The defendants’ submission is that, expressed at the level of generality of the claim, the process of measurement reporting adopted by the standards which is said to infringe was obvious at the priority date. The defendants contend that the process adopted by the standards which is said to fall within the claim can be characterised as comprising three steps and that if it was obvious to adopt this three step process then the patent is obvious if it is infringed. The three steps are:

The first step is what the defendants contend is the step alleged by Unwired Planet to be the ‘conversion’ step. Their case is that the obvious encoding scheme is the one alleged to infringe. The defendants do not say that it would be chosen because the skilled person would think of converting UMTS measurement values into a scale which is intended to make them directly comparable with GSM measurement values. Rather the defendants say it would be chosen for its own merits and since I have held that the encoding used in the standard infringes, the requirement for conversion will be satisfied.

The second step relates to the threshold and the third to sending the report. The argument implicitly involves using RSCP as the UMTS measurement value, although that is not stated above. Unwired Planet maintained that claims 5 and 13, which were limited to RSCP were independently non-obvious. That is a hopeless argument. If all the other elements are obvious, so too is RSCP.

The foundation of the argument over common general knowledge alone was as set out in Mr Townend’s second report and summarised in paragraph 154 of the defendants’ opening skeleton, as follows:

In his reports Mr Townend relied on various documents which had been before the standardisation committees in the period up to the priority date to support his opinion that these things were common general knowledge. In closing the defendants submitted that the evidence and cross-examination bore out these points.

From here the defendants submitted that “there is absolutely nothing inventive in any of the three steps that are said to constitute infringement of this patent.” They contend it is “a classic example of the standards-setting body taking the obvious approach.” Dealing with the three steps in particular:

The main elements of Unwired Planet’s case in response were these. Sub-paragraphs (a) to (g) above do not correctly reflect the common general knowledge and are based on hindsight. The committee documents relied on did not support Mr Townend’s opinion, they pointed the other way. The way the common general knowledge was expressed by the defendants was based on hindsight both in the level of generality with which the points were expressed and the way they were combined. As for the three steps: encoding of some kind was natural but there was no reason to use an encoding scheme which fell within the claim; thresholds were known and applicable in some circumstances but it was not obvious to apply them in the relevant circumstances which arise in the defendants’ argument; it was not obvious to use the SACCH in those circumstances either. Unwired Planet relied on Dr Thomas’s evidence and submitted he had not accepted a number of the points the defendants contended he had.

Mr Townend relied on five committee documents in support of his opinion about common general knowledge. Nevertheless the primary evidence was his own opinion and so it does not follow that if a given document does not support a particular point then I am bound to reject Mr Townend’s opinion. Nevertheless the documents do need to be considered. In chronological order they are:

The dates relate to the various meetings at which these documents were presented along with numerous other documents addressing the various issues those working on the standards were considering. Broadly, the two most relevant committees at the time were SMG 2 in GSM and RAN 2 in UMTS. The two Vodafone documents relate to one another, the second one is in effect resubmitting the first one. The two Vodafone and the Nokia documents were presented to committee SMG2. The Siemens document was presented to a 3GPP workshop on handover and cell selection. The workshop was a joint GSM/ UMTS meeting. The Liaison statement was presented at the TSG RAN 2 meeting in Cheju, Korea. The Liaison statement was concerned with liaison between RAN 2 and SMG 2.

Other documents which have a bearing on these issues are T Doc 1145, which is relied on as a distinct starting point for an obviousness attack by the defendants and T Doc e64, which had been relied on as a distinct starting point for obviousness but was dropped before trial. T Doc 1145 was presented at the Bordeaux SMG 2 meeting in September 1999, which is the same meeting at which the Nokia and second Vodafone documents were presented. T Doc e64 was also presented to TSG RAN 2 at the Cheju meeting in November 1999.

Sub-paragraph (a) of the list of matters of common general knowledge contains two elements. The first is that it well known at the priority date that measurement information corresponding to UMTS cells reported by the mobile phone would be needed in order to make handover decisions. That is correct. The second is that this was to enable a GSM network to direct handover to a neighbouring UMTS cell. That is not so simple. While it was one obvious way forward to have a handover decision from GSM to UMTS made by the GSM network that was not the only option which was under discussion at the time. Another option (see T Doc 1145) was to have the decision made by the UMTS network. The defendants submitted that Dr Thomas accepted sub-paragraph (a) in cross-examination. That is not correct. He accepted the first element I have mentioned but as to the second he said there were discussions about where the handover decision was going to be made. Mr Townend regarded the idea of the GSM-UMTS handover decision being made in UMTS as a poor one. However, there were ostensible reasons why it might be worth considering, such as future proofing, and I will return to this below when considering the case over T Doc 1145. As a matter of common general knowledge, in my judgment the skilled person would have been aware that both approaches were being considered, albeit they would favour the GSM approach.

Sub-paragraph (b) is correct in that it is self-evident that measuring UMTS cells would generate more measurements than were generated in a GSM only system. The defendants correctly point out that the patent does not provide any more signalling space.

Relative to sub-paragraph (c) the defendants submitted that as UMTS parameters, RSCP and E_c/I₀ were the clearest candidates for reporting. I agree. To the extent sub-paragraph (c) raises again the issue of whether the handover decision is made by the GSM network or UMTS, the defendants submitted that the claim merely requires reporting to the GSM “control node” and does not require the decision to be made there. I accept that.

The first element in sub-paragraph (d) is about the desirability of making use of existing GSM message structures. The point is to support the defendants’ case that it was obvious to reuse the existing GSM measurement report. At the relevant time the normal GSM measurement report was in the GSM standard. What is now called enhanced measurement reporting in GSM did not exist and was not common general knowledge.

The Nokia document has a bearing on this point. It supports sub-paragraph (b) (need for more measurements to be reported) and puts it into context. In his second report (para 32) Mr Townend said that this document showed that Nokia appreciated that measurements for two GSM bands and UMTS could (and would) be accommodated in the (existing) single GSM measurement report encoded in 6 bit format. In cross-examination he described using the existing structure as a “done deal”. I do not accept Mr Townend’s evidence on this document. The Nokia document is concerned about the likely need in future to report more cells than the six available in a standard GSM report. The document contemplates partitioning the existing GSM measurement report to accommodate multiband and UMTS, referring to reporting a maximum of two cells per band (either two cells each from GSM bands 450, 900, and 1800; or two each from GSM 900, GSM 1800 and UMTS). It concludes that six reported cells would not be appropriate in future and recommends considering the possibility of a new measurement report message. In fact looking back from today it is probable that the Nokia document was part of what lead to enhanced measurement reporting, but that is not relevant.

Another document relevant to sub-paragraph (d) is the Liaison statement. Mr Townend accepted that the proposal here would require a new message. Mr Townend thought that the kind of new message proposed by the Liaison statement (a transparent container for UMTS messages) would not be useful for future proofing.

Another document which supports the idea of a new message and transparent messages for UMTS messages is T Doc e64. Dr Thomas gave evidence about that.

The second Vodafone document was relied on for this issue by Mr Townend because the document capitalised “Measurement Report”. This was a weak point. I do not accept that it bears the inference Mr Townend sought to draw from it.

Dr Thomas was asked about re-using the existing GSM message. It was put to him that using the existing measurement report was the most obvious thing to do. He did not agree. That does not mean his evidence was that it was indeed an obvious approach but just not the most obvious one, he was responding to the way the questions were put. Overall Dr Thomas’s evidence here was that there was a concern about the limited amount of available space. He agreed that one option was to use the resources of an existing measurement report, by which I understood him to mean not using the message structure but using the overall number of bits in the report, divided up in some suitable fashion.

One reason advanced by the defendants why it would be desirable to use existing GSM message structures was that it was easier to make changes to an existing message rather than create a new message. That was not a significant consideration.

I find that the common general knowledge was as follows. The skilled person thought that more measurements needed to be sent and was concerned that resources were limited. Using the space taken up by the existing measurement report was one option but so was creating an entirely new message. It was not obvious which (or a mixture) of these routes would be satisfactory. The idea that a skilled person would seek to minimise changes to the structure inside the existing GSM measurement report was not a significant factor.

The second element in point (d) is that “whilst some changes might be required, it was sensible to seek to minimise the impact on GSM infrastructure”. That was common general knowledge. It relates to the overall GSM infrastructure, not to the detailed structure of the particular GSM measurement report.

Sub-paragraph (e) relates to GSM multiband. As stated in isolation the point is true but the juxtaposition of this statement with arguments about reporting UMTS measurements needs to be treated with care. Multiband involves reporting measurements of the same kind to the GSM network. GSM to UMTS handovers involvement measurements of a different kind and separate RATs. The Nokia document considers both issues together but other documents do not.

Sub-paragraph (f) is correct. RXLEV was sent as a 6 bit number to the BSS on the SACCH.

Sub-paragraph (g) is concerned with thresholds. Mr Townend said comparing a parameter to a threshold is a basic engineering concept. To the extent that Dr Thomas’s evidence was that the idea of applying a threshold to a measurement in order to decide if it is worth reporting was not common general knowledge, I was not persuaded. It is true it was not used in GSM but that is not determinative. I find that the common general knowledge included the concept that one of the ways of determining whether or not to report to the network a measurement made by a phone was by using a threshold applied to that measurement.

Thresholds could be used in other ways too, for example a comparison of one value with a threshold might trigger reporting of a different value or trigger some other activity altogether. This was also common general knowledge.

Also relevant is that in GSM reporting at the priority date the strongest neighbour GSM signals were reported. No threshold was involved. One reason why not is because reporting the six best makes the most use of the available bandwidth.

Having addressed the defendants’ points (a) to (g) I turn to consider what would be obvious for a skilled person to do in the light of this knowledge. It is not easy to apply the structured Pozzoli approach to this way of putting the case but it is not a mandatory approach so I will not try. I will focus on the three steps.

It was obvious that UMTS measurement information, such as RSCP, would need to be encoded in order to be reported. There was one and only one obvious way to do this at the priority date without hindsight. That was to start from the same encoding as whatever encoding was to be used inside UMTS. The skilled person thinking about reporting UMTS measurements would report the UMTS measurements in the UMTS encoding scheme. That was Dr Thomas’s evidence and I accept it. The skilled person would do this unless they had a reason to do something different. That would be true whether they were reporting the measured value in the UMTS network or the GSM network. It was not obvious for the skilled person to create a special different encoding scheme for reporting UMTS measurements in GSM without a good reason.

If the skilled person was going to define a new measurement report for use in GSM, it would comprise its own fields and there would be no reason to do anything other than define fields relating to UMTS values in the same way as in UMTS. Nor would there be any reason to do anything else if the skilled person was thinking of using the existing octet resources of the existing GSM measurement report. That is because the bits needed for a report from a UMTS cell are different from those needed for a GSM cells. The fields for a GSM neighbour cell in the existing GSM report include a 6 bit RXLEV field, a 5 bit BCCH frequency and 6 bit BSIC. However UMTS cells do not have a BSIC or a BCCH frequency. The nearest equivalent for the UMTS cell is the primary scrambling code but that is defined as a 9 bit number. So there is no reason to choose to use 6 bits to encode a UMTS measurement unless that is how it was to be done in UMTS.

Even if the idea that the skilled person wanted to minimise disruption to the GSM message structure has more weight than I have given it, it does not help. The fields to report UMTS measurements and the bits they might require were distinct from what was required to report GSM. There is no obvious reason to set out to convert UMTS measurements into a GSM format nor to make them directly comparable with GSM measurements.

The skilled person would make it their business to find out how RSCP was to be encoded for reporting in UMTS. The encoding scheme used in UMTS today for RSCP is a 7 bit scheme ranging from -120 dBm to -25 dBm in 1 dBm steps. Unwired Planet and Dr Thomas, I think, assumed that this encoding scheme was established by the priority date but in fact it seems to have been established sometime later. However there was no suggestion that inside UMTS itself it was obvious to use any other encoding scheme and I find that that 7 bit scheme actually used is the scheme which would be specified as to be used inside UMTS by the skilled person (or team) at the priority date. Dr Thomas’s evidence (in the context of T Doc 1145/99) was that it was an obvious mapping scheme.

A notable difference between this UMTS scheme and the one used in the GSM standard for reporting the same measurement is that in UMTS the range of power levels reported is wider. The defendants emphasise that the 6 bit encoding of RSCP in GSM is effectively a subset of the 7 bit scheme. This is true. The major difference is that RSCP power levels greater than -53dBm cannot be distinguished in GSM whereas they can be distinguished in the range up to -25 dBm in UMTS. I am not satisfied it was obvious to limit the UMTS range in this way when reporting in GSM.

Although Mr Townend’s opinion was that 6 bit encoding for the UMTS values to be reported in GSM was obvious, I do not accept it. The fact that in GSM RXLEV was encoded as 6 bits does not make it obvious to encode UMTS measurements in 6 bits.

I find that the obvious encoding scheme for reporting RSCP measurements in GSM at the priority date was that 7 bit scheme. That is the only obvious scheme. This is not a GSM format and is therefore not within the claims on any view.

It was entirely obvious to use some form of discrimination to decide which cells were worth reporting. The question is: how? The way GSM worked was to send the six best in a fixed report on the SACCH. One obvious approach if the skilled person was going to use the existing GSM report would be to partition the space in the report as proposed by the Nokia T Doc, reporting a total of six cells with the two best cells for each of two GSM bands and the two best cells in UMTS. This does not involve a threshold of the relevant kind.

I agree that an approach based on thresholds was an obvious one but it inevitably produces an unpredictable, variable, number of cells to be reported. The obvious way to do that for the skilled person is to use a new measurement report and send it on the FACCH. The FACCH allows for variable sized messages whereas the space on the SACCH, the channel used for the normal GSM report, is of a fixed size.

To a skilled person who was going to stay with the existing measurement report sent on the SACCH, using a threshold in the manner claimed in the patent together with a report of fixed length (such as using partitions) would not be obvious. That is because if cells do not satisfy the threshold then space in the report remains empty. For such a scheme to be satisfactory, the skilled person would need to decide how to resolve that problem. The obvious thing to do is not to use thresholds at all. In his cross-examination Mr Townend I think suggested it would be obvious to have a mixed approach with a threshold applied to UMTS measurements, filling up any remaining spaces with the best GSM cells. That did not strike me as the thinking of the uninventive skilled person operating without hindsight.

In summary it was not obvious to combine reporting on the SACCH with the use of a threshold as claimed.

I have in effect already covered this. The SACCH was an obvious channel to use but not in combination with a threshold.

In addition to the detailed points on the three steps addressed already, I bear in mind that real skilled people were well aware of the problem of how to send UMTS measurements to GSM in the context of GSM to UMTS handover and had been aware of it for what, given the intensity of the work which was being done on UMTS at the time, was a considerable period. Inter-RAT handover in general and this aspect of it were not problems with easy solutions. This cannot be taken too far because the invention is not the complete solution to UMTS reporting for inter-RAT handover, but it does play a part in it. The invention is a neat way of allowing the phones to report the neighbour cells most likely to be useful in making handover decisions when signalling space is limited.

I find that the claims are not obvious over common general knowledge alone.

T Doc 1145/99 (also referred to at trial as “Bordeaux”) was presented to SMG 2 by Ericsson for the Bordeaux meeting in September 1999. Its title is “GSM to UMTS handover”. The document makes a specific proposal for a way of doing this which it called “a concept”. The concept is divided into four parts: downloading measurement orders, the UE measurements, the UMTS measurement reporting and the handover execution. Measurement orders are the instructions from the network to the phone to make measurements and report and what sort of measurements should be made. The other three parts are fairly self-explanatory.

The document proposes that GSM to UMTS handover decisions be made by the UMTS network. Unwired Planet called this the focus of the document. Although Mr Townend did not accept that, I find that that is what the document is proposing and is what the reader would understand it to be focussed upon.

Mr Townend described the idea in T Doc 1145/99 of making the handover decision in the UMTS network as a terrible idea. The defendants submitted that even if the main idea proposed in the prior art is something a skilled person would reject (e.g. as a terrible idea), it did not preclude a finding of obviousness as a result of the skilled person taking forward other things in the prior art. I accept that as a matter of principle but I am bound to say it is not a good start.

The document mentions both the SACCH and the FACCH as channels on which measurement reports may be sent. The document expresses a clear preference for using the FACCH but the defendants’ case was built on the reference to the SACCH. Again the defendants submitted that even if the main idea proposed in the prior art is something different, it did not preclude a finding of obviousness as a result of the skilled person taking forward another thing referred to in the prior art. I accept that as a matter of principle too but with a similar caveat.

Unwired Planet submitted that if one was going to use the SACCH, the obvious thing to do would be to create a new message and that Mr Townend agreed with that. Counsel for the defendants said that this latter point was not an accurate reflection of Mr Townend’s testimony. I agree. The point is that Mr Townend maintained his view in cross-examination that it was obvious to encode the measured UMTS values in six bits. He did accept that if one was going to use a new message then you could split the space in a different way and report RSCP in 7 bits. He did not agree that on the SACCH a new message was the obvious way forward. His view was that the skilled person would use the GSM measurement report on the SACCH.

In the context of common general knowledge alone, I have not accepted Mr Townend’s view about encoding UMTS values in 6 bits as representative of the thinking of the skilled person. The issue of encoding is the same over T Doc 1145/99. 6 bits would not be obvious. That means there would be no conversion as required by the claims. The defendants’ case is not based on suggesting that the ability to make a direct comparison is obvious.

T Doc 1145/99 also mentions the idea of thresholds but it does not associate them specifically with reporting on the SACCH any more than the common general knowledge did. That issue is therefore the same here as it was above.

I have not delved into every point on obviousness over T Doc 1145/99. It is not necessary to do so. I reject the obviousness case. A skilled person given this document at the priority date and reading it with interest would in all probability just put it to one side as a terrible idea. I doubt they would be motivated to go any further. If they were then they would pursue the idea of sending the UMTS reports on the FACCH, which does not lead to the claimed invention. They would be aware of the known disadvantage of the FACCH that it operated by stealing frames from the traffic channel but that would not deter them. They might well use a threshold if the report was to be sent on the FACCH but the document does not make the combination of such a threshold and reporting on the SACCH obvious. In any event the skilled person would not encode UMTS measurements in a GSM format.

I have addressed the obviousness arguments on their merits. However in my judgment the argument based on common general knowledge alone contained a good number of the familiar flaws in arguments of this kind which were discussed by Floyd J in Ratiopharm v Napp [2008] EWHC 3070 (Pat) at paragraph 155-159 and by me in Accord v Medac [2016] EWHC 24 at paragraphs 119-124. The argument in this case had not been properly pleaded. That is not the defendants’ fault because the parties agreed to treat Mr Townend’s first report as a statement of the case. However the case then shifted very close to trial, which demonstrated why it should have been pleaded properly in the first place. Furthermore the argument presented a combination of common general knowledge features which had been created with hindsight knowledge of the patent. It was presented in a way which lacked inconvenient details which were found when the same ideas appeared in the committee documents and it presented points of common general knowledge at a level of generality which itself was crafted with hindsight.

In opening the defendants submitted that criticisms in Accord of an argument based on common general knowledge alone did not really work in a mobile phone standards case. One reason was because the problem solved by the invention did not exist prior to or independently of the creation of the standard in the course of which the invention was made. They submitted there was no long standing problem. The invention may have been made the first and only time the problem was considered. I do not agree with this. It ignores the intensity of the effort which went into creating the UMTS standard and the evidence that the problem in this case was a well-known problem. The various committee documents demonstrate that it is wrong to suggest that invention was made the first and only time the problem was considered.

Another reason advanced by the defendants why the problems identified in Ratiopharm and Accord are said not to apply is because the standards are arrived at iteratively and there is no room for concrete prior art (a term I used in Accord ). It is true that the process is iterative but the latter submission is wrong. There were plenty of concrete documentary proposals which exposed the thinking of the skilled people before the priority date on the very issues with which the case is concerned and from which the defendants could and did develop obviousness arguments.

Third, the defendants submitted that there may only have been one company tasked with a particular issue in the committee and one does not know what their strategy was. This is all very well in theory but the documents relating to inter-RAT handover which have been considered derived from different independent organisations. This is not a substantial criticism.

Fourth, the defendants argue that it is inevitable that many of the ideas adopted in a standard were simply the obvious and straightforward solutions to problems and did not require a scintilla of lateral thinking or inventive step. I agree but it does not help. No doubt many are, but that does not mean others are not.

In the judgment for trial B, I stated the law on added matter briefly as follows. A patent is invalid if the matter disclosed in the specification extends beyond that disclosed in the application as filed. Amendments which have this effect are prohibited by s76(2) of the 1977 Act but if they have occurred, the patent will be invalid (s72(1)(d)) unless a further permissible amendment cures the difficulty. The provisions of the 1977 Act have the same effect as Art 123(2) of the EPC. The basic approach to be followed is that explained by Aldous J as he then was in Bonzel v Intervention [1991] RPC 553. Added matter has been considered by the Court of Appeal a number of times. Two significant cases are Vector v Glatt [2007] EWCA Civ 805 and AP Racing v Alcon [2014] EWCA Civ 40.

The objection is concerned with the word “said”. The relevant said is one of the saids in claim 1. Feature [d] of claim 1, refers to “if at least one of said converted plurality of downlink measurement values exceeds a predetermined threshold measurement value, sending said at least one of said converted plurality of down link measurement values on a control channel …”. (my emphasis) By contrast the corresponding words of claim 9 lack this said. The feature of claim 9 is “means for sending at least one …”. Unwired Planet argued that even if claim 1 could be read as requiring the sending of all values which exceeded the threshold, claim 9 could not be read that way as a result of the missing said. The defendants said that if that was the effect of the missing said, it was added matter because the word had been removed in prosecution and was present in the application as filed.

The defendants are correct that the version of claim 9 in the application as filed includes the said after the word sending, which was then deleted before grant. However even assuming, which I have not found, that removing the said broadens the scope of the claim as compared to the claim in the application, that does not produce added matter (see the AC Edwards v Acme line of cases including AP Racing ). Even if the claim covers more things as a result of the amendment it does not disclose any new information to the skilled person. The things they would do based on reading the patent are the same as what they would do based on the application as filed. I reject the submission of added matter.

Patent EP (UK) 1 230 818 is valid, infringed by the defendants and essential to the relevant standards.