`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`PATENT TRIAL AND APPEAL BOARD
`
`
`APPLE INC.,
`Petitioner
`
`v.
`
`TELEFONAKTIEBOLAGET LM ERICSSON
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Patent Owner
`
`
`
`Case No. IPR2022-00464
`Patent No. 10,193,600
`
`
`SUPPLEMENTAL DECLARATION OF DR. MURIEL MÉDARD, Sc.D
`IN SUPPORT OF PATENT OWNER’S RESPONSE
`
`
`
`
`
`
`
`
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 1
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`TABLE OF CONTENTS
`
`I.
`II.
`III.
`
`INTRODUCTION ........................................................................................ 3
`Additional materials Considered ................................................................ 4
`Supplemental Opinions regarding the Petitioner’s Proposed
`Application of the Prior art to the Challenged Claims ............................. 5
`A. Novlan’s paragraph 54 does not teach rank-agnostic signaling. ............... 6
`B. Novlan and 36.213 do not teach “codebook subset restriction
`signaling that, for each of one or more groups of precoders, jointly
`restricts the precoders in the group by restricting a certain
`component that the precoders have in common.” .................................... 11
`a. Novlan’s Fourth Method of General CSR Does Not Disclose
`These Limitations. ............................................................................... 12
`b. Novlan’s Sampling Based Subset Restriction Is A Distinct
`Embodiment from Novlan’s Fourth Method of General CSR,
`and Does Not Cure Its Deficiencies. ................................................... 15
`C. The Petition’s Proposed Combination Requires Numerous
`Hindsight Modifications to Novlan Which Are Not Taught And
`Would Not Have Been Obvious to a POSITA. ....................................... 22
`IV. CONCLUSION ........................................................................................... 31
`
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 2
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`I.
`
`INTRODUCTION
`
`1. My name is Dr. Muriel Médard. I make this declaration based upon my
`
`own personal knowledge and, if called upon to testify, would testify competently to
`
`the matters contained herein.
`
`2.
`
`I have been asked to provide technical assistance in the inter partes
`
`review of U.S. Patent No. 10,193,600, which I may abbreviate as “the ’600 Patent.”
`
`Ex. 1001. I previously provided a statement of my opinions related to the
`
`patentability of claims of the ‘600 Patent in a declaration dated May 9, 2022 (Ex.
`
`2001), which I incorporate by reference here. This supplemental declaration is a
`
`statement of certain additional opinions on issues related to the unpatentability of
`
`claims of the ’600 Patent. I am being compensated at my normal rate of $600 per
`
`hour for my analysis, plus reimbursement for expenses. My compensation does not
`
`depend on the content of my opinions or the outcome of this proceeding.
`
`3.
`
`I make this declaration based on my personal knowledge. I am over the
`
`age of 18 and competent to make this declaration. The statements herein include my
`
`opinions and the bases for those opinions, which relate to at least the Petition and
`
`associated exhibits. I have reviewed the Board’s Institution Decision in detail, and
`
`I have also reviewed Patent Owner’s Response (POR). I agree with the analyses and
`
`conclusions set forth in the POR.
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 3
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`4.
`
`For the reasons set forth in Ex. 2003, and for the additional reasons set
`
`forth below, it is my opinion that the Petition fails to show that the challenged claims
`
`are unpatentable.
`
`II. ADDITIONAL MATERIALS CONSIDERED
`
`5.
`
`In addition to the materials listed in Section IV of my May 9, 2022
`
`declaration (Ex. 2001), I have now additionally reviewed the Institution Decision as
`
`well as the deposition transcript of Dr. Apostolos K. Kakaes. A complete list of the
`
`materials I have considered is set forth below.
`
`Exhibit
`1001
`
`Description
`U.S. Patent No. 10,193,600 (“the ’600 Patent”)
`
`1002
`
`1003
`
`1004
`1005
`1006
`1007
`1008
`1009
`
`1010
`
`1011
`1012
`
`1013
`
`
`
`Certified File History of U.S. Patent No. 10,193,600
`
`Declaration of Dr. Apostolos K. Kakaes for Inter Partes Review of
`U.S. Patent No. 10,193,600
`Curriculum Vitae of Dr. Apostolos K. Kakaes
`U.S. Patent Application Publication No. 2014/0016549 (“Novlan”)
`3GPP TS 36.213, v12.3.0 (“36.213”)
`3GPP TS 36.213, v10.1.0
`U.S. Patent No. 8,891,676
`Declaration of Friedhelm Rodermund in Support of Petition for Inter
`Partes Review of U.S. Patent No. 10,193,600
`U.S. Provisional Patent Application No. 62/103,101 (“the ’600
`Patent Provisional”)
`U.S. Patent Application Publication No. 2013/0163687 (“Jing”)
`U.S. Provisional Patent Application No. 61/670,936 (the “Novlan
`Provisional”)
`Dahlman et al., 4G – LTE / LTE-Advanced for Mobile Broadband
`(Academic Press 2011) (“Dahlman”)
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 4
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`Description
`Declaration of James L. Mullins in Support of Petition for Inter Partes
`Review of U.S. Patent No. 10,193,600
`U.S. Patent Application Publication No. 2008/0051091
`Sesia, et al., LTE - The UMTS Long Term Evolution From Theory
`to Practice (Wiley 2d. ed. 2011) (“Sesia”)
`Declaration of Jacob Robert Munford in Support of Petition for Inter
`Partes Review of U.S. Patent No. 10,193,600
`Declaration Of Dr. Muriel Médard, Sc.D In Support Of Patent Owner’s
`Preliminary Response
`Curriculum Vitae of Dr. Muriel Médard
`Deposition Transcript of Dr. Apostolos K. Kakaes (Nov. 11, 2022)
`
`Exhibit
`1014
`
`1015
`1016
`
`1017
`
`2001
`
`2002
`2003
`
`
`III. SUPPLEMENTAL OPINIONS REGARDING THE PETITIONER’S
`PROPOSED APPLICATION OF THE PRIOR ART TO THE
`CHALLENGED CLAIMS
`
`6.
`
`This supplemental declaration provides further support for my opinion
`
`that neither Novlan alone (Ex. 1005), nor Novlan taken in view of 36.213 (Ex. 1006)
`
`discloses, teaches, or suggests each and every limitation of the Challenged Claims.
`
`My May 22 declaration focused on Novlan’s failure to disclose “the codebook subset
`
`restriction signaling is rank-agnostic signaling that jointly restricts the precoders in
`
`a group without regard to the precoders’ transmission rank.” This supplemental
`
`declaration further addresses this argument, particularly regarding paragraph 54 of
`
`Novlan’s failure to teach “rank-agnostic signaling.” I also provide my opinions with
`
`respect to two additional flaws in the petition. These arguments are organized in my
`
`declaration as follows:
`
`
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 5
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`A) Novlan’s paragraph 54 does not teach rank-agnostic signaling;
`
`B) Novlan and 36.213 do not teach “codebook subset restriction signaling
`
`that, for each of one or more groups of precoders, jointly restricts the
`
`precoders in the group by restricting a certain component that the precoders
`
`have in common.”
`
`C) The Petition’s Proposed Combination Requires Numerous Hindsight
`
`Modifications to Novlan Which Would Not Have Been Obvious to a
`
`POSITA.
`
`A. NOVLAN’S PARAGRAPH 54 DOES NOT TEACH RANK-AGNOSTIC
`SIGNALING.
`
`7.
`
`Novlan’s paragraph 54 states the following:
`
`[0054] As shown in FIG. 7, angle θF , which represents the elevation
`angle range experienced by the outdoor UEs, is much smaller than
`θC,which represents the elevation angle range experienced by all the
`UEs. Codebooks may be designed to sample the entire spatial domain
`(based on a sampled DFT for example). Thus, one method for
`improving the efficiency of vertical PMI selection and reducing CQI
`computation complexity is to restrict the UE to searching through the
`codebook only over those precoders that correspond to relevant spatial
`domain.
`
`8.
`
`The POSITA would have understood this paragraph to be generally
`
`introducing the idea of codebook restriction, not describing an innovative way to
`
`signal codebook restrictions using “rank-agnostic signaling.” The first two
`
`sentences of this paragraph generally describe the concept that while the UE may
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 6
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`have access to a codebook that has precoders that collectively cover the entire spatial
`
`domain, there may be individual precoders not relevant to that UE’s spatial location.
`
`Figure 7 is illustrative of that idea, where the UE’s elevation angle may make certain
`
`precoders not relevant to that UE (i.e., particular UEs may be located at elevation
`
`angles (θF) that are much smaller than the possible elevation angles (θC) of all the
`
`UEs):
`
`
`
`9.
`
`Novlan’s selection of which precoders to restrict is based on the
`
`observation that certain UEs may be located at an elevation angle that makes certain
`
`precoders irrelevant. To restrict based on UE elevation (i.e., in the vertical direction)
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 7
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`Novlan teaches to restrict the UE to searching only over vertical PMIs that
`
`correspond to the UE’s relevant elevation angle (i.e., spatial domain).
`
`10. The concept of Novlan to “restrict the UE to searching through the
`
`codebook only over those precoders that correspond to relevant spatial domain” is
`
`hardly innovative. The POSITA would have understood that restricting precoders
`
`to a particular spatial domain was a conventional way to choose a subset of precoders
`
`for codebook restrictions. The POSITA would have understood Novlan’s merely
`
`proposing in this paragraph that one way to choose the relevant spatial domain is to
`
`restrict the vertical precoders to the relevant elevation angle, as discussed above.
`
`The mere silence of this paragraph on rank would not be assumed by the POSITA
`
`to suggest rank-agnostic signaling.
`
`11. Even if it were wrongly assumed that paragraph 54 teaches a “rank-
`
`agnostic restriction,” such a teaching is insufficient to disclose the invention claimed
`
`by the ’600 Patent because the claims of the ’600 Patent require more than a rank-
`
`agnostic restriction in the abstract. The Challenged Claims require “rank-agnostic
`
`signaling” and the mere disclosure of a “rank-agnostic restriction,” without more,
`
`fails to disclose “the codebook subset restriction signaling is rank-agnostic signaling
`
`that jointly restricts the precoders in a group without regard to the precoders’
`
`transmission rank.”
`
`
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 8
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`12. Paragraph 54 is not describing some non-conventional way to signal
`
`the restriction to the UE (other than extending the restriction to the vertical). The
`
`POSITA would have understood that this paragraph was not proposing to change
`
`how restrictions could be signaled, only to how vertical precoders could be selected
`
`for restriction based on elevation angle. Paragraph 54’s silence on rank would not
`
`have been understood by the POSITA to teach a rank-agnostic signaling; the
`
`paragraph merely suggests that the UE should be restricted to searching “only over
`
`those precoders that correspond to the relevant spatial domain.” Paragraph 54 does
`
`not teach how any restriction should be signaled to the UE.
`
`13. Choosing the appropriate vertical PMIs must be followed by
`
`appropriate CSR signaling to inform the UE of the restricted codebooks. Paragraph
`
`54 comes before the section of Novlan that specifically describes methods for
`
`signaling subset restrictions to the UE. See, e.g., Ex. 1005 ¶¶[0055]-[0064]. To
`
`understand how to signal the restricted vertical PMIs, one must look to one or more
`
`of the various alternative embodiments disclosed by Novlan for doing so.
`
`14. And as I explained in my May 22 declaration, all of Novlan’s signaling
`
`methods for CSR expressly rely on precoder and rank-specific signaling. Ex. 2001,
`
`¶¶33-34. I also explained that the POSITA would have understood that Novlan’s
`
`disclosure of a restriction based on elevation angle does not necessarily mean that
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 9
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`its restriction is based solely on elevation angle to the exclusion of all else; the
`
`restriction signaling based on Novlan is also specific to rank. Ex. 2001, ¶¶33-34.
`
`All of the evidence I have seen shows that Novlan’s CSR signaling methods are
`
`specific to rank; there is no teaching of a rank-agnostic signaling method in Novlan.
`
`15. Novlan and TS 36.213, whether alone or in combination with one
`
`another, do not teach rank-agnostic signaling, and the codebook subset restrictions
`
`do not restrict precoders in a group without regard to the precoders’ transmission
`
`rank.
`
`16. Dr. Kakaes conceded at his deposition that there are methods of CSR
`
`taught by Novlan that are not rank-agnostic. Ex. 2003 at 96:2-97:14; 128:15-129:3
`
`(conceding the embodiments of [0077] and [0102] are not rank-agnostic). Paragraph
`
`77 is, of course, describing the same embodiment as [0078], which is relied on by
`
`Dr. Kakaes. Dr. Kakaes’ concession is important, because it shows that a CSR may
`
`restrict to the relevant spatial domain using signaling that is not rank-agnostic, which
`
`undercuts Dr. Kakaes’ view that [0054] necessarily teaches a rank-agnostic
`
`restriction. Indeed, he does not show any rank-agnostic method of CSR signaling
`
`taught by either Novlan or 36.213. Even assuming a POSITA would have been
`
`motivated to extend the teachings of [0064] to Rank 2 V-PMIs, Table 1 at best
`
`teaches the POSITA to send higher level signaling that gives the UE the precise
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 10
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`mapping of PMI indication field bits to a group of specific desired precoders; it does
`
`not teach to signal a restriction of some common component between the Rank 1
`
`and Rank 2 V-PMI regardless of rank. Ex. 2003, at 117:17-118:19. Therefore, even
`
`if the POSITA were to try and extend Table 1 to Rank-2 PMIs, the POSITA would
`
`have done so using groups of specifically identified Rank-2 V-PMIs (which does not
`
`disclose a rank-agnostic signaling).
`
`B. NOVLAN AND 36.213 DO NOT TEACH “CODEBOOK SUBSET
`RESTRICTION SIGNALING THAT, FOR EACH OF ONE OR MORE GROUPS
`OF PRECODERS, JOINTLY RESTRICTS THE PRECODERS IN THE GROUP
`BY RESTRICTING A CERTAIN COMPONENT THAT THE PRECODERS
`HAVE IN COMMON.”
`
`17. Novlan and 36.213 only teach signaling restrictions that identify
`
`specific precoders (or groups of precoders) to restrict, which fails to teach any
`
`“codebook subset restriction signaling that, for each of one or more groups of
`
`precoders, jointly restricts the precoders in the group by restricting a certain
`
`component that the precoders have in common.” Novlan and 36.213 both signal
`
`restrictions using bitmaps of restricted precoders (or groups of precoders as
`
`suggested in Novlan’s Table 1), and contain no suggestion to use νm for CSR
`
`signaling.
`
`18. Dr. Kakaes’ rationale for this limitation based on the observation that
`
`each of 36.213’s restricted codebooks may have a common component (νm) across
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 11
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`Rank 1 and 2. Dr. Kakaes confuses νm as a component that has some common
`
`quantity or “value,” apparently based on a common nomenclature. As I will explain
`
`in Section III.C below, this is not correct at least because νm is a vector that takes on
`
`different values according to rank. In other words, the same PMI values do not index
`
`to the same νm vectors regardless of rank. Regardless, neither reference has any
`
`teaching, suggestion, or disclosure to use νm to signal a restriction. Rather, each
`
`embodiment of both references teaches to signal CSRs using a rank-specific bitmap
`
`to identify restricted precoder(s). The suggestion to signal restrictions of νm runs
`
`counter to the teachings of each reference to signal CSRs using rank-specific
`
`bitmaps. It is my opinion that Dr. Kakaes’ rationale is only made with the benefit
`
`of hindsight reasoning; the only evidence I’ve seen to restrict certain components of
`
`precoders, as opposed to the precoders themselves, comes from the teachings of the
`
`’600 Patent.
`
`a.
`
`Novlan’s Fourth Method of General CSR Does Not Disclose
`These Limitations.
`
`19. As teaching “jointly restricting the precoders in the group,” the Petition
`
`relies on ¶[0064] and Table 1 of Novlan. Pet. 25. But this embodiment does not
`
`restrict a group of precoders by restricting a certain component that the precoders
`
`have in common. This point was conceded by Dr. Kakaes, who explained that
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 12
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`Novlan’s fourth general CSR method is made “irrespective of what the structure of
`
`the precoders . . . is.” Ex. 2003, at 123:15-124:18.
`
`20.
`
`In Novlan’s fourth method of general CSR, the “M subset restricted
`
`matrices are configured by the eNB” using higher-level signaling. Ex. 1005 ¶[0064].
`
`The eNB sends a mapping of restricted PMIs shown in Table 1 using higher-level
`
`signaling, which provides a mapping of 2-bit values to groups of desired PMI
`
`indices.
`
`
`
`21. Ex. 1005 ¶[0064]. In this embodiment, once the UE has received the
`
`mapping, the eNB may restrict the UE to the desired PMI group using a single, two
`
`bit “PMI indication field.” See also Fig. 8. As shown in Figure 8, the UE is then
`
`restricted to 2-bit PMI reporting, which a POSITA would have understood would be
`
`used to identify one of the four desired PMIs (all others being restricted). This is
`
`consistent with Dr. Kakaes’s understanding of Table 1 and Figure 8. Ex. 2003, at
`
`119:18-123:7 (explaining Novlan’s fourth method of general CSR in the context of
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 13
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`Figure 8).
`
`
`
`22. For example, as shown above at step 804, the eNB might signal a PMI
`
`indication field of ‘00,’ restricting the UE to PMI indices “1, 5, 9, 13.” Ex. 2003
`
`(Kakaes Dep.) at 121:17-122:5. The UE is then restricted to reporting a 2-bit PMI
`
`(shown directly to the right of step 804), which identifies one of PMIs 1, 5, 9, or 13.
`
`While Table 1 does not require an individual bit for each precoder, each 2-bit value
`
`restricts precoders by identifying a specific group of desired precoders.
`
`23.
`
`In the example used for Table 1, there are N = 16 vertical matrices as
`
`described in ¶[0057], and each 2-bit PMI indicator refers to a group of four PMIs
`
`identified using a numerical value 1..16 in the PMI indices field of Table 1. ¶[0064].
`
`For example, a PMI indication of ‘00’ limits the desired PMIs to PMIs 1, 5, 9, or 13,
`
`as discussed above. This implicitly restricts the UE from reporting on PMIs 2-4, 6-
`
`8, 10-12, or 14-16. But this form of restriction is still restricting a specific set of
`
`precoders as identified in the PMI indices field, and it does not restrict any particular
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 14
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`component of the precoders. Novlan’s fourth method, therefore, does not restrict
`
`precoders in each group by restricting a “common component” to precoders in the
`
`group. To the contrary, it restricts using a bit field that identifies a specific group of
`
`desired precoders, thereby excluding undesired precoders from selection by the UE
`
`as discussed above.
`
`24. The Petition contends this embodiment discloses a “joint restriction”
`
`because “it restricts multiple precoders without requiring a bit associated with each
`
`individual precoder matrix.” Pet. 26. But, as conceded by Dr. Kakaes, the structure
`
`of the precoder is not relevant to this embodiment because the restriction is made
`
`irrespective of what the structure of the individual precoder is. Ex. 2003, at 123:15-
`
`124:18. Accordingly, the “joint restriction” taught by Novlan does not restrict “a
`
`group of precoders by restricting a certain component that the precoders have in
`
`common.”
`
`b.
`
`Novlan’s Sampling Based Subset Restriction Is A Distinct
`Embodiment from Novlan’s Fourth Method of General
`CSR, and Does Not Cure Its Deficiencies.
`
`25. While the Petition relies on Novlan’s fourth method of CSR as teaching
`
`a “joint restriction,” it leaps to a different embodiment as teaching “restricting a
`
`certain component that the precoders in the group have in common.” Pet. 27 (citing
`
`Ex. 1005 ¶[0078]). This embodiment of Novlan relates to methods for sampling-
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 15
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`based subset restriction. Ex. 1005 ¶[0074]-[0080]. Neither embodiment teaches
`
`“codebook subset restriction signaling that, for each of one or more groups of
`
`precoders, jointly restricts the precoders in the group by restricting a certain
`
`component that the precoders have in common.”
`
`26. The Petition relies on a portion of Novlan that reproduces Table 7.2.4-
`
`1 from 36.213. Ex. 1005 ¶[0078]. Dr. Kakaes concedes that Novlan’s Table 1 (relied
`
`on for “joint restriction” and this part of Novlan describe “different embodiments.”
`
`Ex. 2003, at 124:20-125:4. I agree, and I have seen no plausible explanation for how
`
`or why the POSITA would be motivated to combine these embodiments. Dr. Kakaes
`
`confirmed that these two embodiments use a different precoder indexing scheme,
`
`and one could not use the indexing scheme from Table 1 to index Table 7.2.4-1. Ex.
`
`2003, at 124:20-125:4. Moreover, Table 7.2.4-1 and -2 provide a mechanism for the
`
`UE to index and identify a particular codebook for CSI reporting purposes, it does
`
`not even describe a method for CSR restriction at all. Table 7.2.4-1 of Novlan (and
`
`36.213) does not disclose jointly restricting groups of precoders nor discloses doing
`
`so “by restricting a certain component that the precoders have in common” either.
`
`27. Novlan’s sampling-based subset restriction embodiment is based on
`
`Table 7.2.4-1 of 36.213, which is a “double codebook structure” that requires two
`
`prematrix indicators, i1 and i2, to index any particular codebook. ¶¶[0075],[0077]-
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 16
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`[0078]. Table 7.2.4-1 is reproduced below.
`
`
`
`28. The Petition relies on νm shown above as the “certain component that
`
`precoders in the group have in common.” Pet. 27. The Petition theorizes that a
`
`POSITA would have somehow recognized that the “precoders in the group identified
`
`above (those that do not ‘correspond to [the] relevant spatial domain’) would all
`
`have a certain component in common, νm., that is associated with a particular
`
`elevation angle that is outside of the relevant spatial domain.” Pet. 27. The Petition
`
`goes on to observe that 36.213 (Ex. 1006) includes Table 7.2.4-2, which is a Rank 2
`
`codebook also indexed by i1 and i2. Pet. 29. Table 7.2.4-2 is reproduced below.
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 17
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`
`
`29. The Petition theorizes that the νm “values” would be identical for the
`
`same elevation angle, citing only to paragraph 54 of Novlan and Dr. Kakaes’
`
`declaration at ¶90. Pet. 31. However, Dr. Kakaes conceded under oath that this is
`
`incorrect, and that the elevation angle of any particular νm will vary based on the
`
`base station’s antenna configuration. Ex. 2003, 58:12-59:25. The Petition goes on to
`
`theorize that it would have been obvious to signal a joint restriction using νm as a
`
`common component regardless of rank. Pet. at 31-33.
`
`30.
`
`I disagree with these theories and submit that the POSITA would not
`
`have been motivated to use νm as a common component to restrict across ranks.
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 18
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`31. First, the Petition’s assumption that there is a particular νm associated
`
`with an elevation angle (e.g. 80º) is wrong. As conceded by Dr. Kakaes, the vm is
`
`an array of four complex values that does not, by itself, resolve to any particular
`
`elevation angle. Ex. 2003, at 54:23-58:19; 61:25-62:5; 68:13-19. Dr. Kakaes was
`
`unable to calculate an elevation angle from a particular νm. Id. This is because the
`
`POSITA would need more formulas and information to perform such a calculation
`
`(none of which is taught by Novlan or 36.213 nor explained by the Petition). Id.
`
`32. A POSITA would understand that the elevation angle of any particular
`
`beam created by the 3GPP precoders would be dependent on the antenna
`
`configuration and could vary from base station to base station. The POSITA would
`
`be faced with the challenge of not only coming up with a new way to signal vm than
`
`is taught by the prior art, she would also need to figure a way to associate vms to
`
`elevation angle across varying antenna configurations. In my opinion, this would
`
`not have been an obvious problem for the POSITA to solve.
`
`33. But even assuming that the Petitioner could somehow show that
`
`particular νm “values” can be resolved to particular elevation angles, Novlan and
`
`36.213 still fail to disclose, teach, or suggest signaling νm “values” to restrict
`
`precoders that supposedly would have νm “values” in common. The portions of
`
`Novlan relied on by the Petition and Dr. Kakaes do not describe how to signal CSRs
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 19
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`at all. The 7.2.4 tables define what the precoders are and provide a mechanism for
`
`the UE to identify a precoder (using i1 and i2) in its CSI reporting. These tables do
`
`not describe how to signal a CSR. The POSITA would have been motivated to look
`
`elsewhere in the standard to determine how to signal a CSR according to 3GPP. Ex.
`
`2003, at 48:24-50:13.
`
`34. The section of 36.213 that discusses signaling CSRs does so by
`
`indicating whether particular precoders are restricted. Ex. 2003, at 53:7-54:8.
`
`Section 7.2 of 36.213 discusses the bitmap parameter used to communicate
`
`restrictions and states:
`
`Codebook subset restriction is supported for transmission modes 3, 4,
`5, 6 and for transmission modes 8, 9 and 10 with PMI/RI reporting. The
`resulting number of bits for each transmission mode is given in Table
`7.2-1b. The bitmap forms the bit
`where
`0a is the LSB
`,...,
` , , ,aaaa
`
`
`
`aA
`1
`3
`2
`1
`0
`is the MSB and where a bit value of zero indicates that the
`and
`1cAa
`PMI and RI reporting is not allowed to correspond to precoder(s)
`associated with the bit. Ex. 1006 at 62.
`
`c
`
`35.
`
`In all cases in 36.213, the bitmap yields separate bits that identify by i1
`
`and i2 the precoders that are restricted for each layer. There is no suggestion in
`
`36.213 or the cited portions of Novlan to use a common component to jointly restrict
`
`multiple precoders, rather, each precoder is signaled using the bitmap described
`
`above. Id. at 62-63.
`
`36. Novlan’s disclosure of how to signal CSRs is consistent with 36.213.
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 20
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`In the bitmap signaling for codebook subset restriction described by Novlan in
`
`conjunction with its Sampling Based Subset Restriction, each bit is specific to one
`
`and only one precoder. See Ex. 1005 ¶[0100]-¶[0103]. This is shown explicitly in
`
`paragraph 101, which states the following:
`
`[0101] Codebook subset restriction can be signaled to a UE by using a
`bitmap in a UE-specific manner. A bit value of zero in the bitmap
`indicates that the PMI reporting is not allowed to correspond to the
`precoder associated with the bit. The number of bits in the codebook
`subset restriction bitmap N is determined by the number of precoders
`allowed. The bitmap A forms the bit sequence aN-1, . . . , a3, a2, a2, a1,
`a0, where a0 is the least significant bit (LSB) and aN-1 is the most
`significant bit (MSB).
`
`37. Novlan explains that “codebook subset restriction can be signaled …
`
`using a bitmap,” but that each bit in the bitmap is “associated with the precoder for
`
`ν layers …,” where ν corresponds to rank. Id. ¶[0101]-¶[0102]; see also ¶[0077]
`
`(defining “ν” as “equal to the associated RI value”). Accordingly, Novlan teaches
`
`to use a bitmap to identify the restriction for each precoder for each rank (just like
`
`36.213). The POSITA would have understood that the restriction method of
`
`paragraph 101 does not identify a component of the precoder on which to restrict
`
`and therefore restricts regardless of the structure of any component of each precoder.
`
`The POSITA would have been motivated to use signaling like this to signal CSRs,
`
`not use νm as suggested by Dr. Kakaes. I have seen no teaching within Novlan to
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 21
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`restrict precoders by signaling νm as suggested by Dr. Kakaes.
`
`38. The teachings of paragraph 54 do not change this analysis. For
`
`example, paragraph 54 of Novlan contains no teachings of how to signal a CSR at
`
`all, let alone a suggestion to jointly restrict by restricting a certain component the
`
`precoders have in common. Nor would its vague reference to restricting based on
`
`elevation angle motivated the POSITA to make fundamental changes to the
`
`operation of Novlan or 36.213 of the kind suggested by the Petition and Dr. Kakaes.
`
`In short, the use of νm as a common component is not taught by either reference and
`
`runs counter to the methods for signaling CSR actually taught by both references.
`
`C. THE PETITION’S PROPOSED COMBINATION REQUIRES NUMEROUS
`HINDSIGHT MODIFICATIONS TO NOVLAN WHICH ARE NOT TAUGHT
`AND WOULD NOT HAVE BEEN OBVIOUS TO A POSITA.
`
`39.
`
`It is my opinion that the teachings of Novlan and 36.213 would have
`
`motivated the POSITA to use the conventional, rank-specific CSR signaling
`
`techniques actually taught by Novlan and 36.213. The POSITA would not have
`
`attempted to modify Novlan’s teachings to signal CSR using νm across multiple
`
`ranks. The proposed combination of Novlan requires numerous hindsight
`
`modifications to Novlan that are not taught or suggested by the prior art. In essence,
`
`the Petition’s theory would require the POSITA to do the opposite of what is taught
`
`by these references. The only motivation I have located for making these
`
`
`ERICSSON EXHIBIT 2004
`Apple Inc. v. Telefonaktiebolaget LM Ericsson
`IPR2022-00464, Page 22
`
`
`
`
`
`Supplemental Declaration of Dr. Muriel Médard
`
`modifications is using the teachings of the ’600 Patent. As such, it is my opinion
`
`that the invalidity grounds presented in the petition rely on the benefit of hindsight
`
`using the ’600 Patent as a roadmap. I am informed by Counsel that this is improper.
`
`40.
`
` The Petition’s proposed modification include combining aspects of
`
`Novlan’s method for “General Subset Restriction” (i.e., Table 1) with aspects of
`
`Novlan’s method for “Sampling Based Restriction,” (i.e., Table 7.2.4-1) and with
`
`aspects of 36.213 (i.e., Table 7.2.4-2), and then further modifying these methods to
`
`use νm.to restrict precoders across multiple ranks. These combinations are not taught
`
`or suggested