1 2 3 4 UNITED STATES DISTRICT COURT 5 NORTHERN DISTRICT OF CALIFORNIA 6 7 CARNEGIE MELLON UNIVERSITY, Case No. 18-cv-04571-JD

8 Plaintiff,

CLAIM CONSTRUCTION

v. 9

10 LSI CORPORATION, et al., Defendants. 11

12 13 The parties in this long-running patent infringement action, plaintiff Carnegie Mellon 14 University (CMU) and defendants LSI Corp. and Avago Technologies (together LSI), have asked 15 for construction of seven terms or phrases in U.S. Patent No. 6,201,839 (the ’839 patent) and U.S. 16 Patent No. 6,438,180 (the ’180 patent). The parties’ familiarity with the record is assumed. 17 Construction of two disputed terms -- “Viterbi-like algorithm” and “each sample corresponds to a 18 different sampling time instant” -- is deferred to summary judgment because LSI raises 19 indefiniteness challenges. The other five disputed terms are construed. 20 LEGAL STANDARD 21 “Claim construction must begin and remain centered on the claim language itself, for that 22 is the language the patentee has chosen to particularly point out and distinctly claim the subject 23 matter which the patentee regards as his invention.” Hybrid Audio, LLC v. Asus Comp. Int’l Inc., 24 No. 17-cv-05947-JD, 2022 WL 3348594, at *1 (N.D. Cal. Aug. 12, 2022) (cleaned up) (quoting 25 Source Vagabond Sys. Ltd. v. Hydrapak, Inc., 753 F.3d 1291, 1299 (Fed. Cir. 2014)). Claim terms 26 are given their “ordinary and customary meaning,” which is “the meaning that the term would 27 have to a person of ordinary skill in the art in question at the time of the invention.” Broadcom 1 Corp. v. Netflix Inc., No. 20-cv-04677-JD, 2022 WL 1619151, at *1 (N.D. Cal. May 23, 2022) 2 (quoting Phillips v. AWH Corp., 415 F.3d 1303, 1312-13 (Fed. Cir. 2005) (en banc)). 3 The Federal Circuit has emphasized that the “only meaning that matters in claim 4 construction is the meaning in the context of the patent.” Trustees of Columbia Univ. v. Symantec 5 Corp., 811 F.3d 1359, 1363 (Fed. Cir. 2016). “The presumption in favor of giving terms their 6 plain and ordinary meaning may be overcome by a patentee’s express definition of a term, or 7 express disavowal of the scope of the claim.” Broadcom, 2022 WL 1619151, at *2. “A term may 8 be redefined ‘by implication’ when given a meaning that is ascertainable from a reading of the 9 specification or the patent documents.” Hybrid Audio, 2022 WL 3348594, at *2 (citation 10 omitted). “The ordinary meaning of a claim term is not the meaning of the term in the abstract,” 11 but the term’s “meaning to the ordinary artisan after reading the entire patent.” Astra Zeneca AB 12 v. Mylan Pharm. Inc., 19 F.4th 1325, 1330 (Fed. Cir. 2021) (quotations and citations omitted). 13 “A claim and its constituent words and phrases are interpreted in light of the intrinsic 14 evidence. The touchstones are the claims themselves, the specification, and, if in evidence, the 15 prosecution history.” Hybrid Audio, 2022 WL 3348594, at *2. “This intrinsic evidence is the 16 most significant source of the legally operative meaning of disputed claim language.” Id. (citing 17 Vitronics Corp. v. Conceptronic, Inc., 90 F.2d 1576, 1582 (Fed. Cir. 1996)). Although “reading a 18 limitation from the written description into the claims is one of the cardinal sins of patent law,” 19 Broadcom Corp. v. Netflix Inc., 762 F. Supp. 3d 878, 884 (N.D. Cal. 2025) (cleaned up) (quoting 20 Phillips, 415 F.3d at 1320), “the specification is ‘the single best guide to the meaning of a disputed 21 term,’” Broadcom, 2022 WL 1619151, at *2 (citation omitted); see also Merck & Co., Inc. v. Teva 22 Pharms. USA, Inc., 347 F.3d 1367, 1370 (Fed. Cir. 2003) (“[C]laims must be construed so as to be 23 consistent with the specification . . . .”). “The Court may also use extrinsic evidence (e.g., 24 dictionaries, treatises) to resolve the scope and meaning of a claim as circumstances warrant.” 25 Broadcom, 2022 WL 1619151, at *2. 26 CLAIM CONSTRUCTION 27 The parties request construction of terms in Claim 4 of the ’839 patent and Claim 2 of the 1 Claim 4, ’839 patent:

2 A method of determining branch metric values for branches of a trellis for a Viterbi-like detector, comprising: 3 selecting a branch metric function for each of the branches at a certain time index from a set of signal-dependent branch 4 metric functions; and applying each of said selected functions to a plurality of signal 5 samples to determine the metric value corresponding to the branch for which the applied branch metric function was 6 selected, wherein each sample corresponds to a different sampling time instant. 7 8 Claims 1 and 2, ’180 patent1: 9 A method of determining branch metric values in a detector, 10 comprising: receiving a plurality of time variant signal samples, the signal 11 samples having one of signal-dependent noise, correlated noise, and both signal dependent and correlated noise 12 associated therewith; selecting a branch metric function at a certain time index; and 13 applying the selected function to the signal samples to determine the metric values. 14 The method of claim 1, wherein the branch metric function is selected 15 from a set of signal-dependent branch metric functions. 16 The parties agree the relevant person of ordinary skill in the art (POSITA) here is an 17 individual who, as of 1998, held a Master’s degree in electrical engineering and had at least two 18 years of experience in the industry, and who specialized or had experience in data-detection 19 techniques or technologies. Dkt. Nos. 144-4 (McLaughlin Decl.)2 ¶¶ 57-58; 144-5 (Soljanin 20 Rebuttal Decl.) ¶¶ 23-24; see In re GPAC Inc., 57 F.3d 1573, 1579 (Fed. Cir. 1995) (“The person 21 of ordinary skill in the art is a hypothetical person who is presumed to know the relevant prior 22 art.”). The ensuing constructions draw on this framing of the POSITA. 23 24 25

26 1 The Court is construing a term that appears only in Claim 2, but Claim 1 is included for context.

27 2 Dkt. No. 144-4 contains both an affirmative declaration and rebuttal declaration from Dr. Steven 1 A. “branch metric function” (Claim 2 of the ’180 patent & Claim 4 of the ’839 patent) 2 3 CMU’s Proposed LSI’s Proposed Court’s Construction 4 construction Construction 5 A function for determining a A mathematical relation that A mathematical function for 6 branch metric value for a uniquely associates signal determining a branch metric 7 branch, where the first set of samples with branch metric value for a branch, where a 8 the function comprises one or values. “function” is a mathematical 9 more signal samples and one relation that uniquely 10 or more target values, and the associates members of a first 11 second set comprises branch set with members of a second 12 metric values. set. 13 Although not binding, the Court “would be remiss to overlook another district court’s 14 construction of the same claim terms in the same patent.” Finisar Corp. v. DirecTV Grp., Inc., 15 523 F.3d 1323, 1329 (Fed. Cir. 2008). In a prior litigation, Carnegie Mellon University v. Marvell 16 Technology Group, Ltd. (Marvell), the District Court for the Western District of Pennsylvania 17 conducted extensive claim construction proceedings related to the patents at issue involving 18 multiple rounds of briefing and a court-appointed technical advisor. 19 In Marvell, the parties agreed, and the district court so construed, “branch metric function” 20 to mean “a mathematical function for determining a ‘branch metric value’ for a ‘branch.’” 21 Carnegie Mellon Univ. v. Marvell Tech. Grp., Ltd., No. 09-cv-00290-NBF, 2011 WL 4527353, at 22 *7 (W.D. Pa. Sept. 28, 2011). The district court construed “function” to mean “a mathematical 23 relation that uniquely associates members of a first set with members of a second set,” because 24 that was the term’s ordinary meaning “in the pertinent field” and the patent’s full context revealed 25 that “function” was being “used in its ordinary sense.” Carnegie Mellon Univ. v. Marvell Tech. 26 Grp., Ltd., No. 09-cv-00290-NBF, 2012 WL 1203353, at *6 (W.D. Pa. Apr. 10, 2012) (citations 27 omitted). Here, CMU and LSI again agree with those constructions. See Dkt. Nos. 148 at 6; 158 1 at 4; see also Dkt. No. 250 at 13:8-15, 21:22-24. The Court adopts them based on the parties’ 2 agreement and for the reasons set forth in Marvell. 3 The parties disagree whether “function” means a one- or two-input function, see Dkt. Nos. 4 148 at 6-9; 158 at 4-8, but the question need not be answered at this juncture. The issue of inputs 5 versus parameters was addressed in Marvell solely in connection with the contention that the 6 patents were invalid for lack of written description, see 35 U.S.C. § 112, because their 7 specifications did not disclose a set of functions. See Marvell, 2012 WL 1203353, at *8-9. It was 8 in that context that the district court characterized certain elements as parameters versus inputs, 9 and even so, it spoke solely with respect to what Equation 13 in the specification did or did not 10 disclose. See id. at *9 (“[I]t follows that Equation 13 must represent a set of functions.”). The 11 parties did not explain how that narrow question is relevant to what the claim language means. It 12 bears mention that, even after the summary judgment ruling, the district court neither changed its 13 prior constructions nor instructed the jury with the additional language about which the parties 14 presently fight. See Dkt. No. 158-2 at ECF 19 (jury instruction about claim construction from 15 Marvell); see also Dkt. No. 250 at 15:2-10. 16 B. “signal-dependent branch metric function” (Claim 2 of the ’180 patent & Claim 4 of 17 the ’839 patent) 18 19 CMU’s Proposed LSI’s Proposed Court’s Construction 20 Construction Construction 21 A “branch metric function” A branch metric function that A branch metric function that 22 that accounts for the signal- accounts for the noise accounts for the signal- 23 dependent structure of the attributable to the specific dependent nature of noise 24 media noise symbol sequence associated attributable to a specific 25 with one branch symbol sequence 26 The patents’ specifications describe an invention that “[c]ombat[s] media noise and its 27 signal dependence” by “tak[ing] into consideration the correlation between noise samples in the 1 2:35-38. The parties agree that a POSITA would understand “signal-dependent branch metric 2 functions” in the context of the patents to refer to branch metric functions which in some way 3 account for the noise that is associated with, or dependent on, a particular signal. See Dkt. Nos. 4 148 at 9; 158 at 12-13. 5 The points of disagreement are discrete and relate to: (1) whether the signal-dependent 6 noise for which the function accounts is limited to “media noise”; (2) whether the function is 7 accounting for “noise” or “noise structure”; and (3) whether there must be different functions for 8 each branch. The first two disputes are identical to the proper construction of “signal-dependent 9 noise.” Consequently, the Court resolves the disagreements for both disputed terms. 10 1. MEDIA NOISE 11 As a matter of plain language, neither claim references media noise or magnetic recording 12 in connection with signal-dependent branch metric functions or signal-dependent noise.3 It also 13 bears mention that Claims 11, 19, 23, and 24 of the ’839 patent, see Dkt. No. 1-3 at 15:2-17, 14 15:51-67, 16:22-17:4, “demonstrate that the patentee knew how to use verbiage denoting” 15 magnetic recording technology and related issues. Broadcom, 762 F. Supp. 3d at 885. 16 Extrinsic evidence establishes that a POSITA would understand there to exist other types 17 of noise that are signal-dependent but are not “media noise,” as the term is used in the 18 specifications. LSI’s expert, Dr. Emina Soljanin, stated that a POSITA would recognize that 19 “[s]ignal-dependent noise arises in various applications such as . . . wireless communication 20 systems where the transmitted signal can be distorted by nonlinearities in the transmission 21 equipment that depend on the signal strength and bandwidth.” Soljanin Decl. ¶ 92; see also id. 22 ¶ 92 n.18 (citing U.S. Patent No. 6,151,370). Dr. Soljanin also stated that “an optical signal 23 transmitted through fiber can have ‘signal-dependent noise’ because of its interactions with the 24 optical medium.” Id. ¶ 93. One of the inventors, Aleksander Kavcic, testified that signal- 25 dependent noise in the context of his patent would include both “media noise” and “electronics 26

27 3 For present purposes, “media noise” denotes a phenomenon tied to magnetic recording 1 noise,” but that “[s]ome people” would understand “that the conglomerate of the two noises is 2 media noise.” Dkt. No. 157-14 at 53:13-25; see also Dkt. No. 157-8 at 43:13-45:22. 3 Despite good evidence that the disputed claim language is drawn broadly, CMU says that 4 “[t]he ’839 Patent’s specification repeatedly and exclusively references the signal-dependent 5 nature of media noise, making no mention of any signal-dependent noise that is not media noise.” 6 Dkt. No. 148 at 9 (emphasis in original). CMU contends that a POSITA reading the claims would 7 understand media noise to be the claimed signal-dependent noise of which the claimed branch 8 metric functions are taking account. See id. at 9, 13-14. 9 It is true the ’839 patent specification repeatedly connects the concepts of signal 10 dependence and media noise, specifically and exclusively, see, e.g., Dkt. No. 1-3 at 1:39-43, 2:18- 11 20, 4:24-27, 10:18-21, 12:51-54, as does the ’180 patent, though to a somewhat lesser extent, see, 12 e.g., Dkt. No. 1-4 at 1:39-43, 4:39-42, 10:52-53. It also is true the specification is often “the 13 single best guide to the meaning of a disputed term.” Broadcom, 2022 WL 1619151, at *2 14 (citation omitted). But “reading a limitation from the written description into the claims is one of 15 the cardinal sins of patent law.” Broadcom, 762 F. Supp. 3d at 884 (quotations omitted) (quoting 16 Phillips, 415 F.3d at 1320); see SuperGuide Corp. v. DirecTV Enters., Inc., 358 F.3d 870, 875 17 (Fed. Cir. 2004) (“Specifications teach. Claims claim.” (citation omitted)). Other language in the 18 specifications and unasserted claim language, which will be discussed shortly, undermine the 19 notion that “signal-dependent noise” and “signal-dependent branch metric function” were defined 20 “by implication” to relate solely to media noise. Trustees of Columbia Univ., 811 F.3d at 1364. 21 The ’180 patent specification is instructive. To start, the background to the claimed 22 invention(s) is not limited to addressing the issues of correlated and signal-dependent noise in 23 magnetic recording, see Dkt. No. 1-4 at 1:20-67, as the ’839 patent was, see Dkt. No. 1-3 at 1:20- 24 67. The ’180 patent also discusses the “need” for detectors “designed for channels which have 25 correlated and/or signal-dependent noise.” Id. at 2:19-23. When the specification begins to 26 describe those detectors, it makes clear “the teachings of the present invention, in which the 27 branch metric computations are performed assuming the channel has memory, i.e. the noise is 1 device or algorithm in which branch metrics must be computed.” Id. at 14:16-21 (emphasis 2 added). It teaches that the detector in figure 15 “can work on a trellis, tree, finite-state machine, 3 graph, or any other structure with branches for which the detector [ ] has a component that must 4 compute branch metrics.” Id. at 14:39-42. The specification describes a particular embodiment 5 featuring “the BCJR algorithm”, see id. at 14:48-15:22, before expressly stating that “[t]he 6 generalization of the case described above for the BCJR algorithm can be made for any other soft 7 output or hard output algorithm defined on a trellis or a graph of any communications (or other 8 dynamic) system. Id. at 15:23-26 (emphasis added). 9 No mention is made of media noise or magnetic recording technology. The focus is on 10 detector use in communication channels or systems, see Dkt. No. 1-4 at 14:9-47, and the 11 specification twice emphasizes that the teachings may apply to “any” such system, see id. at 12 14:20-21, 15:25-26. Nothing suggests a necessary connection between media noise and 13 communication channels, or that the relevant communication channels are limited to those 14 involving magnetic recordings or media noise. The concept of media noise appears to completely 15 fall out of the picture. CMU did not identify extrinsic evidence suggesting a POSITA would 16 understand otherwise. 17 This is salient because Claim 8 of the ’180 patent, which is related to those portions of the 18 written description just described, claims a “method of detecting a sequence that exploits a 19 correlation between adjacent signal samples for adaptively detecting a sequence of symbols 20 through a communications channel having intersymbol interference” “wherein the channel has 21 nonstationary signal dependent noise.” Id. at 15:60-63, 16:12-13 (emphasis added). If CMU were 22 right in saying that “signal-dependent noise” (and by extension “signal-dependent branch metric 23 functions”) in both patents relates solely to media noise, then Claim 8 must, by implication, only 24 cover instances in which there was “media noise” in the channel. But that conclusion runs 25 contrary to the specification’s express contemplation of embodiments for “any communications 26 (or other dynamic) system,” id. at 15:25-26; see also id. at 14:16-47, and the Court does not see 27 anything in the claim language or the relevant passages of the specification from which a POSITA 1 would conclude the claims is to be understood as so limited. CMU certainly did not proffer 2 extrinsic evidence to that effect. 3 Consequently, use of the term “signal-dependent noise” in Claim 8 undermines CMU’s 4 position because its proposed construction could only be correct if “signal-dependent noise” 5 means different things in different claims.4 See, e.g., Phillips, 415 F.3d at 1314 (“Other claims . . . 6 can also be valuable sources of enlightenment as to the meaning of a claim term.”). The Court 7 “presume[s], unless otherwise compelled, that the same claim term in the same patent or related 8 patents carries the same construed meaning.” Omega Eng’g, Inc. v. Raytek Corp., 334 F.3d 1314, 9 1334 (Fed. Cir. 2003); see also Paragon Sols., LLC v. Timex Corp., 566 F.3d 1075, 1087 (Fed. 10 Cir. 2009). There are certainly circumstances in which that presumption is properly overcome, see 11 Aventis Pharms. Inc. v. Amino Chems. Ltd., 715 F.3d 1363, 1374 (Fed. Cir. 2013), but CMU did 12 not demonstrate they are present here. Notably, CMU’s arguments and materials at the claim 13 construction hearing never responded to LSI’s identification of this point of incongruence about 14 communication channels. See Dkt. No. 158 at 18:13-15; see also Dkt. No. 250 at 22:7-23:21, 15 46:24-47:4. 16 CMU’s arguments to the contrary are unpersuasive. It urges the Court to follow Marvell, 17 but the Court respectfully disagrees for the reasons stated. That decision focused almost 18 exclusively on the ’839 patent specification and did not address the aspects of the ’180 patent 19 discussed here. See Marvell, 2010 WL 3937157, at *17-20. CMU says LSI’s construction should 20 be disregarded because it would mean the patents “do not disclose any set of signal-dependent 21 branch metric functions.” Dkt. No. 148 at 10 (emphasis in original). The comment is 22 undeveloped, and CMU did not explain how its cited deposition testimony is germane or supports 23 that characterization. 24 25 4 The adjective “nonstationary” in Claim 8 does not support concluding the terms in the two 26 claims should have different meanings. The patents teach that “[t]he non-stationarity of the media noise results from its signal dependent nature.” Dkt. Nos. 1-3 at 36-41; 1-4 at 1:39-41. In other 27 words, the “nonstationary” quality is a product of signal dependence, and no intrinsic or extrinsic 1 Two additional points bear mention. Although “signal-dependent branch metric function” 2 and “signal-dependent noise” are different terms, with Claim 8 of the ’180 patent only using the 3 latter, CMU’s evidence and arguments for both are identical. Compare Dkt. No. 148 at 9-10, with 4 id. at 13-14; see also Marvell, 2010 WL 3937157, at *17-20 (construing both terms identically 5 based on the same evidence and arguments), and there is no suggestion the two could or should be 6 construed differently in this regard. CMU also made no argument the terms should be construed 7 differently across the two patents and instead has proceeded as though they are coterminous. 8 2. NOISE STRUCTURE 9 Reading the patents as a whole, it is clear the claimed branch metric functions are 10 accounting for the way in which the noise is dependent on a specific sequence of symbols. The 11 branch metric functions are accounting for the signal-dependent nature of the noise. For that 12 reason, too, “signal-dependent noise” refers to noise the nature of which is attributable to a 13 specific sequence of symbols. 14 This is confirmed by the intrinsic evidence. The ’839 patent specification repeatedly refers 15 to “[t]he non-stationarity of the media noise results from its signal dependent nature.” Dkt. No. 16 1-3 at 1:39-42 (emphasis added); see also, e.g., id. at 1:41-51, 4:24-27. Several of its teachings 17 make plain the connection between the asserted innovation and accounting for the way in which 18 noise is signal dependent. See, e.g., id. at 10:18-20 (“It is important to point out that, due to the 19 signal-dependent character of the media noise, there will be a different covariance matrix to track 20 for each branch.”). The ’180 patent specification is in accord. See, e.g., Dkt. No. 1-4 at 4:39-42 21 (“Due to the signal dependent nature of media noise in magnetic recording, the functional form of 22 joint conditional pdf [equation omitted] in (1) is different for different symbol sequences [equation 23 omitted].”). Both patents’ specifications also teach that the symbol sequences relevantly comprise 24 the signal. See, e.g., Dkt. No. 1-4 at 3:21-39. Consequently, a POSITA reading the claims in light 25 of the written descriptions would readily understand “signal-dependent branch metric functions” 26 to refer to functions that account for the signal-dependent nature of noise and that “signal- 27 dependent noise” is noise the nature of which is signal dependent. 1 CMU’s proposed construction is nearly identical in this regard but proposes the term 2 “structure” in lieu of “nature.” Dkt. No. 244 at 3, 6. “Structure” is only used once in the ’839 3 patent, see Dkt. No. 1-3 at 2:17-20, and not at all in the ’180 patent. Its usage in the ’839 patent is 4 cursory and undefined, and CMU did not establish with extrinsic evidence that “noise structure” 5 has an accepted meaning in the field that a POSITA would recognize. Consequently, there is no 6 reason to substitute “structure” in construing the terms when the specifications themselves 7 repeatedly refer to the noise’s signal-dependent nature. See, e.g., Dkt. Nos. 1-3 at 4:24-27, 12:51- 8 54; 1-4 at 1:39-41. LSI’s proposed construction would refer simply to “noise,” but its arguments 9 demonstrate that it understands the term as the Court presently construes it. See Dkt. No. 158 at 10 19 (“[C]onsistent with the claim language, the specification also describes the noise itself as 11 signal-dependent.”). 12 3. ASSOCIATED WITH EACH BRANCH 13 The parties disagree about the construction of “signal-dependent branch metric functions,” 14 specifically about whether there must be a different function for each branch tailored to that 15 branch’s noise statistics. Claim 4 of the ’839 patent requires only that a signal-dependent branch 16 metric function be selected for each branch. See Dkt. No. 1-3 at 14:10-14. As a matter of plain 17 reading, the language of the claim does not necessarily require the limitation that the signal- 18 dependent function be unique or tailored to the branch for which it is selected. Unasserted claims 19 suggest the omission of language to that effect was intentional. Claim 6 relates to “[a] method of 20 generating a signal-dependent branch weight for branches of a trellis for a Viterbi-like detector,” 21 which includes “calculating a second value representing a branch dependent joint probability 22 density function of said signal samples.” Dkt. No. 1-3 at 14:24-32 (emphasis added). The 23 patentee thus knew how to say that a function was tailored to a particular branch if it wanted to. 24 See Broadcom, 762 F. Supp. 3d at 885. 25 In support of a narrower construction, LSI reads the specification to denote a 1:1 26 relationship between branch and branch metric function. See Dkt. No. 158 at 12-13; see also Dkt. 27 No. 1-3 at 2:15-19, 5:53-55, 7:12-13. LSI also points to submissions from CMU during 1 in a way that solves the problem the inventors addressed with their invention as discerned from the 2 specification. That includes using different branch metric functions for each specific symbol 3 sequence (e.g., trellis branch).” Dkt. No. 144-19 at ECF 54. 4 “[T]here is sometimes ‘a fine line between reading a claim in light of the specification, and 5 reading a limitation into the claim from the specification.’” Bell Atl. Network Servs., Inc. v. Covad 6 Comms. Grp., Inc., 262 F.3d 1258, 1270 (Fed. Cir. 2001) (quoting Comark Comms., Inc. v. Harris 7 Corp., 156 F.3d 1182, 1186 (Fed. Cir. 1998)). The intrinsic evidence to which LSI points does not 8 suffice to show disclaimer, see Omega Eng’g, 334 F.3d at 1323-24, and the Court cannot conclude 9 at this juncture that a POSITA would understand the claim to be limited to requiring a different 10 branch metric function for each branch, rather than covering such embodiments and others. 11 To start, the reexamination proceeding is ambiguous because the relevant language gave 12 “trellis branch” as an example of “each specific symbol sequence.” Dkt. No. 144-19 at ECF 54. 13 That did not unmistakably create an equivalence between the two or suggest the given example 14 was exhaustive. See Omega Eng’g, Inc., 334 F.3d at 1324 (limiting claim scope inappropriate 15 where “alleged disavowal of claim scope is ambiguous”). The same ambiguity persisted with 16 respect to the specification. Beside those instances to which LSI points in which the written 17 descriptions posits a 1:1 relationship between branch and branch metric function, there are 18 passages where the specification explains that the functions differ “for different symbol 19 sequences.” Dkt. No. 1-3 at 4:24-27; see also id. at 11:1-9 (“The matrix [equation omitted] 20 becomes our estimate for the covariance matrix corresponding to this particular symbol sequence 21 (trellis path).”); McLaughlin Decl. ¶ 28 (explaining that a “path is made up of a series of branches 22 end-to-end”). The specification consequently contemplates functions that account for symbol 23 sequences longer than a trellis branch. LSI’s expert, Dr. Soljanin, takes the written description’s 24 discussion of figures 4 and 5 to demonstrate that the claim is limited by implication, see Soljanin 25 Decl. ¶¶ 165-67, but that passage is best read as providing an example of how the invention would 26 work under certain assumptions, see Dkt. No. 1-3 at 10:26-52. That a specification extensively 27 discusses one embodiment, or one set of embodiments, does not mean the claims are limited to 1 It appears that CMU, in briefs filed with the Federal Circuit in the Marvell litigation, made 2 statements about the inventions that arguably are different from the claim constructions it proposes 3 here. Namely, CMU contended that “[e]ach function in the set accounts for the noise-correlation 4 structure attributable to the specific symbol sequence associated with one branch.” Dkt. No. 5 158-12 at ECF 31-32. Extrinsic evidence refers to “evidence external to the patent and 6 prosecution history, including expert and inventor testimony, dictionaries, and learned treatises,” 7 Phillips, 415 F.3d at 1317 (citation omitted), and LSI did not cite any authority to suggest that 8 attorney argument in a brief constitutes such evidence. Even so, the Court will not close the door 9 to revisiting claim construction on this point as circumstances warrant. 10 C. “a set [of signal-dependent branch metric functions]” (Claim 2 of the ’180 patent & 11 Claim 4 of the ’839 patent) 12 13 CMU’s Proposed LSI’s Proposed Court’s Construction 14 Construction Construction 15 Two of more “signal- No construction necessary; in Plain and ordinary meaning. 16 dependent branch metric the alternative, plain and 17 functions” ordinary meaning. 18 Claim terms are generally given the “ordinary and customary meaning” a POSITA would 19 ascribe to them. Broadcom, 2022 WL 1619151, at *1 (citation omitted). Nothing in the claims or 20 specifications suggests that a POSITA would understand “a set” to mean anything other than the 21 term’s customary meaning in ordinary English. No further construction is required. 22 23 // 24 // 25 // 26 // 27 // 1 D. “signal-dependent noise” (Claim 2 of the ’180 patent) 2 3 CMU’s Proposed LSI’s Proposed Court’s Construction 4 Construction Construction 5 Media noise in the readback Noise in a received signal Noise the nature of which is 6 signal whose noise structure is attributed to the specific attributable to a specific 7 attributable to a specific sequence of symbols sequence of symbols 8 sequence of symbols (e.g., 9 written symbols) 10 The Court’s construction of “signal-dependent branch metric function” applies with equal 11 force to this disputed term, and no further construction is required. 12 E. “correlated noise” (Claim 2 of the ’180 patent) 13 14 CMU’s Proposed LSI’s Proposed Court’s Construction 15 Construction Construction 16 Noise with “correlation” Noise among signal samples Noise among signal samples 17 among “signal samples,” such that tends to vary together. that tends to vary together. 18 as that caused by coloring by 19 front-end equalizers, media 20 noise, media nonlinearities, 21 and magnetoresistive (MR) 22 head nonlinearities. The noise 23 in signal samples is 24 “correlated” when the noise in 25 the signal samples has a 26 tendency to vary together. 27 There is not much daylight between the proposed constructions. CMU accepts that “LSI’s 1 definition of ‘correlated.’” Dkt. No. 148 at 15. CMU does not explain why “non-exclusive 2 || examples” of correlated noise sources are needed to understand the term. Id. The additional 3 language reads more like what a POSITA might associate with the term, but not what a POSITA 4 || would think the term itself means. Neither party suggested there is a material difference between 5 || their preferred prepositions “among” versus “in.” 6 IT IS SO ORDERED. 7 Dated: August 12, 2025 8 9

JAM ONATO

10 Unitga States District Judge 11 a 12

15 16

= 17

Z 18

19 20 21 22 23 24 25 26 27 28