Thanks for visiting! Welcome to a new way to research case law. You are viewing a free summary from Descrybe.ai. For citation and good law / bad law checking, legal issue analysis, and other advanced tools, explore our Legal Research Toolkit — not free, but close.
Carnegie Mellon University v. Marvell Technology Group, Ltd.
Citations: 986 F. Supp. 2d 574; 2013 U.S. Dist. LEXIS 135208; 2013 WL 5332108Docket: Civil Action No. 09-290
Court: District Court, W.D. Pennsylvania; September 23, 2013; Federal District Court
Carnegie Mellon University (CMU) filed a patent infringement lawsuit against Marvell Technology Group and Marvell Semiconductor, claiming infringement of two patents related to high-density magnetic recording sequence detection (U.S. Patent Nos. 6,201,839 and 6,438,180). CMU alleges Marvell's infringement was willful, while Marvell argues the patents are invalid. The case was tried for four weeks, concluding with a jury verdict on December 26, 2012, in favor of CMU, affirming the patents' validity and willfulness, and awarding damages of $1,169,140,271. Post-trial, both parties filed motions including renewed requests for judgment as a matter of law, a new trial, and various other equitable remedies. The court has now addressed the pre-verdict JMOL motions and will rule on the post-trial motions. The patents pertain to methods for sequence detection in hard disk drives (HDD), where data is stored on magnetic tracks, and issues related to discrepancies, or "noise," during the read-back process are acknowledged.
A Viterbi-like read channel detector in hard disk drives (HDD) processes read-back signal samples to identify the sequence of symbols recorded on the disk through a method known as "sequence detection." The detector utilizes a trellis structure to represent bit sequences, with four possible two-bit states (01, 11, 00, 10) interconnected by branches. Each trellis section encompasses all potential bit sequences, allowing representation of sequences of varying lengths. The detector identifies the "best path" through the trellis using branch metric values derived from the read samples. The sequence with the lowest cumulative branch metric is determined as the detected output.
The document highlights that noise in the bit regions of the disk is not uniform. Instead, these regions consist of small magnetic grains, leading to areas of magnetization that do not conform to ideal bit patterns, particularly as recording density increases. With higher density, fewer grains per bit region can result in islands of grains where accurate data reading becomes challenging. Increased recording density also correlates with heightened noise levels and uncertainty in the signal, influenced by the specific sequence of bits written on the track.
Historically, noise was presumed to be white (uniform) across time and branches, which was effective in the low-density recording environments of the 1970s and 1980s. Over time, the industry advanced to consider correlated noise with structure invariant to the symbol sequence, and now recognizes correlated signal-dependent noise, which varies based on the actual sequence of symbols recorded, reflecting the complexity of modern storage media.
The CMU Patents focus on a method for detecting sequences by maximizing a likelihood function influenced by all read back signals and written symbols from a disk. Given the vast number of symbols, the likelihood function is divided into smaller per sample functions. The patents outline a process that starts with observed signal samples, hypothesizes a sequence of written symbols, and employs mathematical manipulations to derive a quotient of the likelihood function. This leads to various embodiments, including a correlation matrices embodiment and a Finite Impulse Response (FIR) embodiment. The FIR embodiment uses specific filter coefficients for each sequence of written symbols to address signal-dependent noise. When applied, this allows for the use of a Viterbi-like detector, which improves accuracy as recording density increases. The CMU Patents claim to provide an "optimal" detection solution, particularly as media noise has become a critical issue in accurately reading data, constituting 90% of the noise in read channels. Research efforts have been initiated by both academia and industry to tackle the media noise challenge at theoretical and practical levels.
Carnegie Mellon University (CMU) is a prominent research institution in Pittsburgh, recognized for its engineering, information technology, and computer science programs. Dr. Jared Cohon has served as President since 1997. In response to a decline in the U.S. storage industry, CMU established the Magnetics Technology Center (MTC) in 1983, which evolved into the Data Storage Systems Center (DSSC) in 1990 with support from the National Science Foundation. The DSSC is an interdisciplinary entity that funds long-term research via federal grants, university investments, and corporate sponsorship. Associate members, who pay an annual fee of $250,000, include companies like IBM, Seagate, and Western Digital, and receive disclosures of inventions and a royalty-free license for those developments. The patents in question originated within the DSSC, but Marvell was not an associate member.
From 1995 to 1998, Professor Jose Moura and doctoral student Aleksandar Kavcic developed a method to address high-density media noise issues, submitting an invention disclosure in March 1997. A provisional patent application was filed in May 1997, followed by a final application in April 1998, resulting in the '839 Patent granted on March 13, 2001. A second patent, the '180 Patent, was filed on March 1, 1999, and granted on August 20, 2002. The inventors published relevant research papers in 1998 and 2000. Kavcic left CMU for Harvard University after earning his Ph.D. in 1998, while Moura remains at CMU. CMU's policy stipulates that half of any proceeds from these patents, including from the current lawsuit, are shared between the inventors.
CMU's Center for Technology Transfer and Enterprise Creation, led by Robert Wooldridge, oversees the management of the University's intellectual property. In August 2003, Carl Mahler, a subordinate, sent licensing inquiries regarding the '180 and '839 Patents to fourteen companies, including Marvell and Toshiba, though responses were limited, with two companies declining to license. In September 2004, CMU established a subscription agreement with Intel for a yearly fee of $75,000, allowing licensing from a rotating patent pool. Despite inventors' interest, the '180 Patent was not included or licensed by Intel.
Marvell, a significant fabless semiconductor company founded in 1995 by Dr. Sehat Sutardja and others, specializes in integrated circuits for storage products. The company, headquartered in Santa Clara, California, has expanded from seven to nearly 7,000 employees and is publicly traded. Key personnel include Dr. Sutardja (President and CEO), Weili Dai (Vice President), and Dr. Pantas Sutardja (CTO). Together, they own 19% of Marvell. Dr. Alan Armstrong serves as Vice President of Marketing, while Bill Brennan was previously Vice President of Sales. Other notable engineering staff include Dr. Zining Wu, Gregory Burd, and Toai Doan, with Doan having been a principal engineer and later Vice President before leaving in 2009. Dr. Nersi Nazari was a key figure in the early 2000s, acting as a manager and contact for Dr. Kavcic, who sought employment opportunities at Marvell.
Marvell's team, including Dr. Wu and Mr. Burd, focused on enhancing digital signal processing and read channel technologies, which are crucial for interfacing between a disk drive's controller and its analog recording head. As data density increases on disks, errors from adjacent tracks also rise, which limits data storage capacity. To combat this, Marvell aimed to improve the signal-to-noise ratio (SNR) of their read channel chips, addressing various noise sources. An early effort from 1999 to 2001 involved implementing iterative coding to enhance SNR, a concept also central to Dr. Wu's Ph.D. thesis. However, this approach initially failed, with some team members dubbing the resulting chips “coffee warmer” chips due to their high power consumption.
After the iterative coding setback, Mr. Burd explored alternative solutions and, in 2001, began developing a method inspired by Dr. Kavcic and Dr. Moura's work. CMU alleges that Marvell copied this method, as claimed in patents '180 and '839. In contrast, Marvell argued that Mr. Burd's solution was independently developed, labeling it "MNP" in 2003, and later incorporating it into their EMNP and NLD technologies for read channel and system-on-chip (SOC) applications. Marvell engineers created simulators to test these chips before production, running their designs against the benchmark simulator KavcicViterbi. The technologies and simulators involved are collectively referred to as the "Accused Technology" in the ongoing litigation.
Marvell's read channel chips were predominant until around 2004, when the industry shifted to SOC-type chips, which integrate multiple blocks, including the read channel block, to enhance performance and reduce costs. Currently, Marvell employs approximately 800 personnel for SOC chip development. CMU alleges infringement concerning Marvell's MNP, EMNP-type, and NLD type chips, including both read channel and SOC chips, all custom-made per customer specifications. These chips undergo a rigorous sales cycle, starting with a 3-6 month qualification phase followed by a 12-18 month development period, and then a 3-6 month timeframe before volume production begins. This cycle involves significant investments from Marvell with high risks of customer withdrawal. Marvell's engineers support customers throughout the process, with most activities occurring in Santa Clara, California. Accused simulators are used during product development to verify chip designs, and simulation code is provided to customers for evaluation. Major clients include Fujitsu, Hitachi/IBM, and Seagate. From March 2003 to July 2012, Marvell sold 2.34 billion Accused Chips, with average revenue per chip at $4.42 and average operating profit at $2.16. The chips are manufactured in Taiwan by TSMC and are integrated into hard disk drives (HDDs) primarily for laptops, some of which are subsequently imported back to the United States. CMU estimates that between 329 million and 556 million of the Accused Chips were imported into the U.S.
In January 2002, Mr. Burd informed Mr. Doan via two emails that the Kavcic method was patented and assigned to Carnegie Mellon University (CMU). In 2003, Carl Mahler from the CMU Technology Transfer Office sent fourteen letters to various technology companies, including two directed to Marvell personnel, Dr. Pantas Sutardja and General Counsel Matthew Gloss, inviting them to discuss licensing the '180 and '839 Patents. Marvell did not respond to these communications. Additionally, in 2004, Fujitsu, a customer of Marvell, inquired about the '839 Patent and its relation to Marvell's technology, but there was also no response from Marvell to Fujitsu's inquiry.
CMU alleged that Marvell infringed Claim 4 of the '839 Patent and Claim 2 of the '180 Patent through its chips and simulators. Claim 4 of the '839 Patent details a method for determining branch metric values in a Viterbi-like detector, which involves selecting and applying branch metric functions to signal samples. Claim 2 of the '180 Patent, which incorporates Claim 1, describes a similar method that also involves receiving time-variant signal samples with associated noise and selecting a branch metric function from a set of signal-dependent functions.
CMU specifically claimed that Marvell's MNP and NLD Chips, along with various simulators (MNP-Type Simulators, NLD Simulator, and KavcicViterbi Simulator), infringed these patent claims. A stipulation prepared by both parties confirmed that certain circuits in Marvell’s products were accurate representations of the patented technology, and it was submitted as evidence of infringement to the jury.
CMU's expert, Dr. Steven McLaughlin, provided detailed testimony over two days, utilizing a presentation with over 130 slides to illustrate how Marvell's products met each element of the claims. He analyzed the stipulation, simulator code, and Marvell's technical documents to support his findings, emphasizing that infringement occurs only when the patented methods are executed on the products.
In contrast, Marvell's expert, Dr. Richard Blahut, argued against infringement, claiming that Marvell's use of the Viterbi algorithm did not involve the required selection and application steps as specified in the patents since it employed a uniform branch metric function across all branches.
Marvell argued that CMU's patents were invalid due to anticipation and obviousness, presenting U.S. Patent No. 6,282,251 ("Worstell Patent") as prior art, filed on March 21, 1995, before the CMU patents. Marvell also cited two 1992 IEEE articles as further evidence. Expert testimony included Dr. John Proakis supporting Marvell's invalidity claims, while CMU's Dr. McLaughlin rebutted this, asserting the patents were valid. CMU introduced a 1997 email from Glen Worstell indicating that the CMU patents were innovative compared to his own previous work.
In terms of damages, CMU sought a $0.50 per chip royalty for Marvell's sales of Accused Chips from March 2003 onward, arguing that the technology was essential and industry-standard. Industry expert Dr. Christopher Bajorek testified to the technology's necessity and its use by major companies, with Marvell not countering his testimony. CMU's damages expert, Catherine Lawton, analyzed Marvell's sales data, resulting in a reasonable royalty estimate of $0.50 per chip based on various benchmarks. CMU also provided internal communications and deposition testimonies to support its claim. Marvell countered with its damages expert, Creighton Hoffman, who suggested a one-time payment of $250,000 based on different licensing agreements and CMU's alleged lack of marketing efforts, without presenting additional evidence on licensing.
CMU presented evidence to support its claim that Marvell's infringement of its patents was willful. This included internal communications from Marvell, emails from Mr. Burd, letters from CMU and Fujitsu, and deposition testimonies from several individuals, including Dr. Wu and Mr. Burd. CMU also cited expert testimony from Dr. McLaughlin, who asserted that Marvell's MNP was derived from CMU's patents. In response, Marvell argued against willful infringement, providing its own internal communications, patents that referenced CMU's patents, and testimonies asserting they did not copy CMU's patents, including Dr. Sutardja's statement that Marvell's employees were "not stealers."
On December 21, 2012, the jury was tasked with determining issues of infringement, validity, damages, and willfulness. After nearly two days of deliberation, on December 26, 2012, the jury found that CMU proved by a preponderance of evidence that Marvell's products infringed specific claims of the '839 and '180 Patents. They also concluded that Marvell induced and contributed to infringement by its customers. The jury ruled against Marvell's invalidity claims, stating it did not demonstrate by clear and convincing evidence that the patents were anticipated by prior art or obvious. As a result, CMU was awarded $1,169,140,271 for the patent infringement. The jury found that Marvell had actual knowledge of the patents before the lawsuit and did not possess a reasonable defense against CMU's claim, ultimately concluding that Marvell knew or should have known its actions infringed the patents.
CMU filed its complaint on March 6, 2009, and the case underwent extensive discovery and various procedural motions, including a Motion to Transfer and Claim Construction.
Following the Court’s Summary Judgment and Daubert rulings, a two-day hearing occurred on October 17-18, 2012, addressing nineteen Motions in limine—five from CMU and fourteen from Marvell. Rulings were issued shortly thereafter. On November 9, 2012, the parties submitted responses to objections regarding exhibits and deposition designations, along with joint stipulations and proposed jury instructions. A two-day Pretrial Conference was conducted on November 14-15, 2012, where further objections and trial issues were discussed. Trial briefs were also submitted by the parties addressing legal issues for the trial.
Jury selection took place on November 26, 2012, with the trial commencing on November 28, 2012. Throughout the trial, the Court addressed motions related to witnesses, exhibits, and legal points. CMU agreed not to rest its case until three defense witnesses testified. After CMU rested, Marvell moved for judgment as a matter of law on non-infringement, damages, and willful infringement, which the Court denied. Following CMU's rebuttal, Marvell renewed its motions regarding invalidity and other defenses.
The jury was charged on December 21, 2012, and returned a verdict on December 26, 2012, finding in favor of CMU on infringement, validity, and willfulness, awarding damages of $1,169,140,271. The Court entered a joint proposed judgment on January 14, 2013. Subsequently, on February 11, 2013, Marvell filed several motions, including for judgment as a matter of law and a new trial on non-damages issues, while CMU sought a permanent injunction, post-judgment royalties, enhanced damages for willful infringement, and attorneys' fees. These matters were fully briefed, and oral arguments were heard from May 1 to May 2, 2013.
The Court previously denied CMU's request for attorneys’ fees and Marvell's motion for a new trial based on alleged CMU misconduct. It is now considering the parties' motions for judgment as a matter of law (JMOL), motions for a new trial, and a motion for remittitur.
For JMOL, the standard requires that the motion be granted only if the evidence, viewed favorably for the non-moving party, reveals no material factual questions, making any other verdict erroneous. The Court must identify any genuine issues of material fact that could allow a reasonable jury to find for the non-moving party, without weighing evidence or assessing witness credibility. A directed verdict is warranted when the facts and law support only one conclusion.
Regarding motions for a new trial under Federal Rule of Civil Procedure 59, the Court may grant such a motion for various reasons, including when the jury's verdict contradicts the evidence, new evidence emerges, improper conduct affects the verdict, or the verdict appears inconsistent. The Court's discretion varies based on the nature of the alleged errors.
When a motion for a new trial argues that the verdict contradicts the clear weight of the evidence, the Court's discretion is restricted. A verdict can only be overturned if it is contrary to the great weight of the evidence, signifying a miscarriage of justice if allowed to stand. A plausible basis for the verdict prevents it from being set aside. The Court is not to replace the jury's evaluation of facts and witness credibility. If the motion cites court errors, such as in jury instructions or evidence rulings, the court must first ascertain whether an error occurred, ensuring the jury was properly informed on the issues and law. If an error is found, the court assesses its prejudicial impact on the trial's fairness. Generally, new trials are not granted for evidentiary objections not raised during the original trial, and courts typically view post-judgment issues with skepticism.
In the context of the ongoing case, substantial evidence supports a reasonable jury's decision regarding claims made by CMU against Marvell for patent infringement. CMU asserts that Marvell’s chips and simulators infringe specific claims of its patents, backed by expert testimony from Dr. McLaughlin. Marvell has filed for judgment as a matter of law (JMOL) and, alternatively, a new trial, arguing insufficient evidence of infringement. Direct patent infringement occurs when a party makes, uses, sells, or offers to sell a patented invention without authorization, and method claims require performance of all steps in the claimed method for infringement to be established. In this case, Marvell would literally infringe the patents only if its products perform all steps of the specified claims.
Marvell's direct infringement of a patent claim is contingent on its methods, as implemented in its chips and simulators, containing all steps of the claim. Each patent claim must be evaluated individually for literal infringement. Indirect infringement encompasses two types: inducing infringement and contributory infringement, governed by 35 U.S.C. § 271(b) and (c), respectively. To establish either form of indirect infringement, there must first be proof of direct infringement. For inducement, CMU must demonstrate that Marvell knowingly encouraged another party's infringement, which necessitates active inducement and awareness of the patent. Contributory infringement requires showing that Marvell sold or imported a component specifically made for use in infringing the patent, with knowledge of the patent's existence and no substantial non-infringing uses. The burden of proof lies with the patentee to demonstrate the lack of substantial non-infringing uses for the accused products. While expert testimony is not mandatory to prove infringement, its use is frequently endorsed in complex technology cases, particularly when the accused infringer presents expert testimony against infringement.
The Court considered comprehensive arguments from all parties, including those presented during a motion hearing on May 1 and 2, 2013, and evaluated the full trial record and recent submissions. CMU, as the plaintiff alleging infringement of method claims, must prove that Marvell implements every step of the claimed method. Expert Dr. Steven McLaughlin testified regarding CMU's patents and their infringement by Marvell's MNP/EMNP chips and related technologies. Marvell contends that a reasonable jury could not find infringement, asserting that the accused chips do not select branch metric functions for trellis branches at specific time indices, apply these functions to signal samples, or determine branch metric values for a trellis. In contrast, CMU argues that Dr. McLaughlin's analysis effectively maps the claims onto Marvell's technology.
Specifically, CMU claims that Marvell's MNP Chips infringe Claim 4 of the '839 Patent, which details a method for determining branch metric values for a Viterbi-like detector. Dr. McLaughlin deconstructed this claim into three components, demonstrating how the MNP functions as a detector that computes branch metric values and operates as a Viterbi-like detector using circuit evidence from the Chip Stip. Supporting internal documents from Marvell, including specifications and presentations, corroborated Dr. McLaughlin's findings, illustrating the MNP's role in adaptive detection and media noise processing. Furthermore, he identified aspects of the MNP technology that align with the claim's requirement for selecting branch metric functions.
Dr. McLaughlin utilized Exhibit A from the Chip Stip to illustrate how the Marvell MNP technology applied a method that processes multiple signal samples to determine corresponding branch metric values, with each sample linked to different time instants. He confirmed that the MNP technology met all elements of Claim 4 of the '839 Patent, concluding it infringed this patent. Additionally, CMU alleged that the MNP technology also infringed Claim 2 of the '180 Patent, which includes Claim 1. Claims 1 and 2 describe a method for determining branch metric values involving receiving time-variant signal samples with specific types of noise, selecting a branch metric function at a designated time index, and applying this function to the signal samples.
Dr. McLaughlin segmented these claims into five elements and demonstrated how they corresponded to the MNP circuits, highlighting similarities between the language of the '180 Patent and the '839 Patent, which allowed him to forgo revisiting the circuit schematics. He confirmed that the MNP technology performed the necessary steps outlined in Claim 1, including receiving signal samples that exhibit the described noise characteristics, thus showing that all elements of Claim 2 were used in the MNP technology and concluding it infringed this claim as well.
In response, Marvell argued against the infringement claim related to the NLD Chips, asserting that these chips do not select branch metric functions for each trellis branch at specific time points, do not apply these functions to multiple signal samples, and do not determine branch metric values for trellis branches. However, Dr. McLaughlin began his analysis of the NLD Chips by referencing the '839 Patent and evaluating its first element in relation to the chips.
Dr. McLaughlin detailed that the Nonlinear Detector (NLD) technology involves a method for calculating branch metric values within a trellis framework for a Viterbi-like detector, supported by his expertise and Marvell documentation, including Dr. Hongxin Song's application note. He asserted that the NLD selects branch metric functions based on signal dependency and applies these to multiple signal samples to derive corresponding metric values, thereby infringing Claim 4 of the '839 Patent. Additionally, for Claim 2 of the '180 Patent, McLaughlin identified that the NLD employs a method for determining branch metric values, selects functions, and processes time-variant signals. He concluded that the NLD infringes this claim as well. In contrast, Marvell argued that its Accused Simulators do not infringe either patent, stating that the '839 Patent pertains to detectors that handle actual signal samples, which the simulators do not.
Marvell references the case Harris Corp. v. Ericsson Inc. to argue that running a simulator does not equate to actually performing the claimed methods in a signal processing detector. In Harris, the Federal Circuit emphasized that direct infringement requires actual use of a system, not merely the creation or sale of its components. The plaintiff's failure to demonstrate that the claimed method was executed rather than simulated was pivotal to the court's decision. However, it did not establish a strict rule against simulation programs.
In contrast, CMU has presented sufficient evidence to convince the fact-finder that the claimed method is indeed executed on Marvell’s simulators. Dr. McLaughlin testified that Marvell's simulators infringed the asserted claims after examining the code of five specific simulators: KavcicPP, MNP, EMNP, NLD, and KaveicViterbi. The first four simulators are linked to specific chips, while the KaveicViterbi serves as a benchmark. Dr. McLaughlin described their use in chip research and development, hardware design verification, and providing simulation code to customers for evaluating chip performance.
Supporting this, Marvell's corporate designee, Greg Burd, explained that the C code used in simulations serves as a "golden source" for verifying chip designs. Dr. McLaughlin detailed how he compared the simulator code to the physical chip circuits, showing a direct correlation between the code and the circuitry calculations. He concluded that if the chips infringed the claims, the simulators must also infringe. His analysis of the KaveicViterbi Simulator included insights from Marvell engineers, who described its role in benchmarking algorithms and understanding prior developments in the field.
The document outlines the findings related to a legal dispute involving Marvell and CMU regarding patent infringement. Key points include:
1. **Research Foundation**: The excerpt references a class file (KavcicViterbi.cpp) developed by Marvell engineers, which allegedly implements intellectual property (IP) described in Professor Kavcic’s papers and patents.
2. **Simulation Controversy**: Dr. McLaughlin asserted that Marvell’s simulators process both synthetic and real signal samples, contradicting Marvell's claim that only synthetic samples are used. He cited two Marvell documents to support his position: one detailing test results for a Toshiba waveform and another regarding data from a spin stand.
3. **Infringement Claims**: Dr. McLaughlin concluded that the computer code and chip technology infringe all elements of CMU’s patents. Marvell contested these claims, and both parties presented arguments to the jury regarding the nature and function of the simulators.
4. **Jury's Role**: The excerpt emphasizes that the determination of infringement and the credibility of expert witnesses, such as Dr. McLaughlin, is within the jury's purview. The court notes that Marvell had the opportunity to counter Dr. McLaughlin’s conclusions with their own expert, Dr. Blahut, who claimed non-infringement.
5. **Evidence for Verdict**: The court concluded that CMU provided sufficient evidence, through Dr. McLaughlin and Marvell’s documents, to support a potential jury verdict in CMU's favor regarding infringement of specific patent claims. Consequently, Marvell's motions for judgment as a matter of law (JMOL) and for a new trial were denied.
Overall, the court found adequate grounds for a reasonable jury to rule in favor of CMU based on the presented evidence and arguments.
Direct infringement by the Accused Chips and Accused Simulators was appropriately addressed by the jury. Regarding indirect infringement through inducement, Marvell contended that it deserved a judgment as a matter of law or a new trial, claiming that CMU failed to prove three essential elements: 1) direct infringement, 2) Marvell's actual knowledge of the patents, and 3) Marvell's specific intent for others to infringe. The Chip Stip identified chip models linked to circuits analyzed by Dr. McLaughlin, whose expert testimony supported CMU's claims of direct infringement of Marvell's chips.
For CMU to succeed in its induced infringement claim, it needed to demonstrate Marvell's knowledge of the patents or deliberate actions to avoid confirming infringement. CMU presented substantial evidence indicating Marvell's awareness of the patents, including emails from Dr. Kavcic to Marvell in 1998 and from Greg Burd in 2002, highlighting the patented nature of the Kavcic detection scheme. This evidence suggested that Marvell's engineers were aware of the patents and the potential for infringement.
Marvell's defense centered on the assertion that its technology was merely a “sub-optimal” version of Kavcic’s work, yet CMU countered that the method used was indeed based on Kavcic’s patented algorithm. Evidence presented included emails from Dr. Wu discussing the similarity of their approach to Kavcic's original proposal. Ultimately, CMU provided enough evidence for a jury to potentially find that Marvell was willfully blind to the high probability of infringement.
Mr. Doan testified that he had no engagement with the patents in question, neither reviewing them nor directing others to do so, and he did not consult Marvell's legal department. However, he reported to Nersi Nazari on January 14, 2002, about ongoing work on a non-linear detector based on Kavcic’s model. In August 2003, CMU’s Technology Transfer representative Carl Mahler informed Marvell’s CTO and General Counsel about CMU’s patents related to correlation-sensitive adaptive sequence detectors, attaching the patents and inviting further discussion, but received no response. Additionally, in November 2004, Junya Suwanai from Fujitsu reached out to Marvell regarding a license offer for CMU’s patents, but there was no documented response from Marvell.
CMU needed to demonstrate Marvell's specific intent to encourage infringement, which could be shown if Marvell actively aided or encouraged such conduct. Evidence presented included Marvell producing chips with the accused methods and instructing customers to use them in infringing ways. CMU provided emails, firmware, and programming instructions indicating that Marvell directed customers, such as Western Digital, Samsung, and Toshiba, to utilize the chips in infringing modes. Testimony from Dr. McLaughlin highlighted Marvell's engineers assisting customers in integrating these chips, with Western Digital's designee confirming that the chips were enabled for infringing use and that they typically followed Marvell's technical recommendations.
Expert testimony from Dr. Bajorek indicated that the accused technology became an industry standard. Favoring CMU's perspective and drawing reasonable inferences from the evidence, the trial established sufficient grounds to support the claim that Marvell developed technology that incorporated the accused methods, either with actual knowledge or willful blindness regarding potential patent infringement, and sold these products to its customers.
CMU provided sufficient evidence demonstrating that Marvell actively induced at least one customer to utilize a method covered by Claim 4 of the '839 Patent or Claim 2 of the '180 Patent, leading to the denial of Marvell's motion for judgment as a matter of law (JMOL) on this matter. The jury's verdict of induced infringement was upheld as it was supported by the evidence. On the issue of contributory infringement, Marvell sought JMOL or a new trial, arguing that CMU failed to prove direct infringement, Marvell's knowledge of the infringement, and that its components had no substantial non-infringing uses. The Court, however, found that CMU presented ample evidence of direct infringement by Marvell's customers and that Marvell had the requisite knowledge of the patents during the infringement period. Expert testimony indicated that the Accused Chips were specifically designed for infringing use, lacking substantial non-infringing applications. Thus, the Court ruled that CMU provided sufficient evidence for a reasonable jury to conclude that Marvell engaged in contributory infringement of the specified claims. Marvell's JMOL regarding contributory infringement was denied, and the jury's finding was deemed consistent with the evidence. Both parties filed cross-motions for JMOL concerning the validity of the patents, which the Court ultimately denied.
Marvell has renewed its motion for judgment as a matter of law (JMOL) regarding the validity of a patent, seeking either a ruling of invalidity or a new trial. In contrast, Carnegie Mellon University (CMU) did not renew its JMOL on validity due to the jury's favorable verdict.
The legal standard for proving anticipation of a patent claim requires clear and convincing evidence, as established by 35 U.S.C. § 282. A patent is presumed valid, and the burden of proving invalidity lies with the party asserting it. Even if the prior art was not considered by the Patent and Trademark Office (PTO), the standard of proof remains unchanged. A claim is deemed invalid for anticipation if a single prior art reference discloses every element of the claim, including the arrangement of those elements.
The obviousness standard under 35 U.S.C. § 103(a) is determined by factual analyses that assess whether the invention would have been obvious to a person skilled in the art at the time of its creation. This involves evaluating the prior art's scope, the ordinary skill level in the art, differences between the claimed invention and prior art, and any secondary factors that may indicate non-obviousness. Secondary considerations must be taken into account when assessing obviousness, and merely presenting separate references that cover each claim limitation is insufficient to establish obviousness.
Obviousness requires that a person of ordinary skill at the time of the invention would have selected and combined prior art elements in the normal course of research to produce the claimed invention. Marvell, alleging the patents' invalidity, bore the burden of proving its defenses by clear and convincing evidence. During the trial, Marvell presented Dr. John Proakis as an expert witness, while CMU called Dr. Steven McLaughlin to rebut Marvell's claims regarding the patents and prior art.
CMU sought judgment as a matter of law on Marvell's invalidity defenses, particularly the anticipation claim, which the court had previously denied due to time constraints. CMU argued that Marvell failed to provide sufficient evidence for its anticipation defense, highlighting that Dr. Proakis acknowledged the Worstell Patent did not disclose all elements of the CMU Patents and that he relied on an incorrect claim construction, rendering his opinion inadmissible.
Dr. Proakis asserted that prior articles, including those by Weining Zeng and Inkyu Lee, demonstrated that Dr. Kavcic and Dr. Moura were not the first to disclose a specific method involving branch metric functions. He also claimed that the Worstell Patent, filed before the CMU Patents, disclosed relevant methods, including the consideration of signal-dependent noise in its equation. Dr. Proakis maintained that the Worstell Patent described a modified Viterbi detector, which aligned with part of the claims in the CMU Patent.
Dr. Proakis concluded that Claim 4 of the '839 Patent was anticipated by the Worstell Patent, citing relevant equations and elements from both patents, supported by deposition testimony from Dr. McLaughlin. He then addressed Claim 2 of the '180 Patent, asserting that four elements previously identified in the Worstell Patent were also applicable, particularly focusing on the receiving step. Dr. Proakis indicated that the Worstell branch metric equation accounted for both correlated and signal-dependent noise, leading him to conclude that Claim 2 was similarly anticipated by the Worstell Patent.
CMU challenged Dr. Proakis' opinions, arguing they were based on an incorrect interpretation of "function" and "signal dependent branch metric function," rendering them insufficient for a legal finding of validity. The Court determined that it was unclear if Dr. Proakis had provided a contradictory claim construction that would breach expert disclosure rules under Rule 26. Previous case law indicates that an expert's new opinions contradicting their report may be stricken, but elaborations consistent with prior opinions are not subject to such sanctions. The Court noted that Dr. Proakis' terminology use was not influenced by the Court's clarifications, and any perceived change in his opinion was not adequately substantiated to warrant striking his testimony. CMU's claims regarding Dr. Proakis' terminology were deemed more semantic than legal, and discrepancies in terminology were addressed during cross-examination, impacting the weight of his opinion rather than its admissibility. The jury was guided by the Court's claim constructions throughout the trial, and traditional methods of cross-examination and evidence presentation were deemed sufficient to challenge the credibility of Dr. Proakis' testimony.
CMU asserts that during cross-examination, Dr. Proakis acknowledged that the Worstell Patent did not disclose all elements of the CMU Patents, specifically mentioning “zero” and “one” branches without any multiplier on the “zero” branches. He claimed that this omission would be "totally obvious" to someone skilled in the art. The standard for anticipation requires that every element of the claimed invention be described in prior art, which can be inherent if a skilled person could practice the invention without undue experimentation. The Court noted that Dr. Proakis, as a technical expert, does not have the authority to make legal determinations, and his statement alone does not conclusively establish that the Worstell Patent does not anticipate the patents in question. The credibility of Dr. Proakis's opinion is for the jury to decide, not the Court. After reviewing the evidence favorably for Marvell, the Court concluded that there was sufficient evidence for a jury to find that Claims 4 of the '839 Patent and 2 of the '180 Patent were anticipated, thus denying CMU's motion for judgment as a matter of law on this issue.
On the matter of obviousness, CMU argued for judgment as a matter of law due to Marvell's insufficient evidence. This was based on three points: Dr. Proakis's admission regarding the Worstell Patent's incompleteness, reliance on incorrect claim construction, and the conclusory nature of his opinions. Dr. Proakis contended that even if the Worstell Patent did not anticipate the claims, it demonstrated that the claims would be obvious to someone skilled in the art, asserting that such an individual would recognize the need for sigma squared to vary from branch to branch, thus rendering Claims 2 and 4 obvious. His analysis predominantly relied on the Worstell Patent.
Marvell successfully demonstrated that Dr. Proakis's testimony was substantive, addressing secondary indicia of non-obviousness through expert statements on the novelty of the invention. Disputes arose between Dr. Proakis and Dr. McLaughlin regarding the perceptions of a person skilled in the art and secondary considerations, with the Court finding Dr. Proakis's reasoning not conclusory as a matter of law. Ultimately, the Court determined that the question of obviousness, based on evidence and credibility assessments, warranted jury consideration, leading to the denial of CMU's motion for judgment as a matter of law.
Regarding the defenses of written description, indefiniteness, and enablement, Marvell conceded in its opposition brief that it did not pursue these defenses at trial, resulting in a waiver of such claims. Consequently, the Court deemed CMU's motion for judgment on these defenses moot.
On the issue of anticipation, Marvell sought judgment or a new trial, asserting that the PTO had not considered the Worstell Patent, which it claimed disclosed all limitations of the relevant patent claims. However, the Court noted that Marvell failed to provide authority linking the PTO's omission to the anticipation analysis, emphasizing that the jury maintains the role of weighing evidence, regardless of the patent office's actions. CMU presented Dr. McLaughlin as a rebuttal expert to contest Marvell's assertion of invalidity based on the Worstell Patent.
Dr. McLaughlin testified that CMU's patents are novel due to their method, which utilizes a set of signal-dependent branch metric functions applied to multiple signal samples, distinguishing them from Inkyu Lee and Weining Zeng's focus on a single signal sample addressing only transition noise. He noted that the Worstell Patent only considers one signal-dependent branch metric function, contrasting it with claims from CMU’s '839 and '180 Patents, which require multiple functions. Evidence showed that Worstell's modification of branch metrics for transition noise affects only one branch type, with a constant modification, while the CMU patents implement a variable adjustment across all branches. Dr. McLaughlin concluded that the Worstell Patent does not address transition noise adjustment for multiple signal samples and disagreed with Dr. Proakis on the invalidity of CMU's patents, stating Proakis misrepresented the Worstell method. He ultimately asserted that the claims of the '839 and '180 Patents were not anticipated. The determination of factual disputes regarding invalidity was left to the jury, which concluded that there was sufficient evidence to support the non-anticipation of the patents. Consequently, the defendants' motion for judgment as a matter of law on the basis of invalidity was denied, as was Marvell's motion for a new trial on these grounds. Marvell also sought a judgment on the basis of obviousness, claiming the asserted patents lacked novelty.
CMU presented expert testimony from Dr. McLaughlin to contest claims of patent obviousness. Dr. McLaughlin analyzed various prior art references, including works by Inkyu Lee, Weining Zeng, and the Worstell Patent, concluding that the disputed claims were not obvious. He also evaluated secondary indicia of non-obviousness, such as industry praise and the invention's resolution of a long-standing problem, citing an email from Glen Worstell that highlighted the invention’s significance. The Court found sufficient evidence for a jury to determine that Claim 4 of the '839 Patent and Claim 2 of the '180 Patent were not obvious, leading to the denial of Marvell’s motion for judgment as a matter of law on this issue. The jury found the patents valid concerning the obviousness defense.
Regarding willfulness, CMU's assertion of willful infringement was also addressed. The Court denied Marvell's motions for judgment as a matter of law on willful infringement, allowing the jury to decide on the issue. CMU must demonstrate willfulness by proving, with clear and convincing evidence, that Marvell acted with an objectively high likelihood of infringing a valid patent, and that this risk was known or should have been known to Marvell. The Court will consider the arguments from both parties and the trial record to make a determination on these motions, specifically focusing on the objective reasonableness of Marvell's actions.
CMU provided sufficient evidence to establish that Marvell had a high likelihood of infringing CMU’s patents, supported by Marvell's knowledge of the patents as early as 2002. Key points include:
1. **Knowledge of Patents**: Evidence demonstrated that individuals involved in designing the Accused Technology were aware of the patents-in-suit prior to the lawsuit, with a jury finding that Marvell had actual knowledge of the '180 and '839 Patents before March 6, 2009.
2. **Email Correspondence**: An email from Dr. Aleksandar Kavcic to Marvell on March 8, 1998, indicated his inquiry about Marvell’s detectors and referenced his own publications, which discussed ideas relevant to the patents. This email did not grant Marvell rights to use the ideas.
3. **Engineer Testimony**: Gregory Burd, a Marvell engineer, confirmed he read Kavcic’s papers, including the Viterbi detector model, and informed his supervisors about the CMU patents in January 2002.
4. **Development of Technology**: Burd utilized Kavcic’s model to develop a simulation program at Marvell, initially named KavcicPP and later renamed to MNP.
5. **Awareness Through Patent Application**: Other engineers at Marvell acknowledged awareness of CMU’s patents when applying for a related patent, yet claimed they had not read the patents at that time.
6. **CMU's Communication**: CMU sent two letters to Marvell's executives regarding the patents, which went unanswered, as Marvell expressed no interest in the technology.
Overall, the documentation outlines a clear trajectory of Marvell's knowledge and potential infringement of CMU's patents, supported by multiple testimonies and internal communications.
In November 2004, a customer of Marvell, Fujitsu, inquired about the relationship between CMU’s patents and Marvell's read channel technology, requesting Marvell's opinion by the end of that month. No documents related to this inquiry were found, and Marvell's corporate representative testified that he was unaware of any response. Evidence presented at trial indicated that Marvell made minimal efforts to assess potential infringement of the CMU patents, despite knowledge of their existence. Key personnel, including Dr. Wu, Mr. Burd, and Mr. Doan, admitted to not reviewing the patent claims, even though they were aware of Dr. Kavcic's patented algorithm. Mr. Doan, upon receiving the patents, did not initiate any investigations or consult Marvell's legal team, instead instructing his team to continue development based on the “Kavcic model.”
Marvell's assertions that their product was a suboptimal version of Dr. Kavcic's work contradicted Dr. Wu's acknowledgment of using the original structure proposed by Kavcic. The inaction on the potential infringement risk was inconsistent with Marvell's own intellectual property policy, which mandates forwarding patent information to the legal department. However, Marvell’s Vice President of Marketing, Dr. Armstrong, was uncertain whether the CMU patents were submitted for legal analysis or whether there were any discussions about licensing them following Fujitsu's request. Marvell's lack of inquiry about infringement meant no steps were taken to avoid it, a fact corroborated by Mr. Burd. Despite being a sophisticated company with extensive patents, Marvell produced 2.3 billion chips without investigating the CMU patents, continuing to use the accused technology without plans to discontinue, even seven months after a verdict in July 2013.
Burd presented Marvell with a product called "KaveicPP," which was associated with a detection scheme patented by Dr. Kavcic. Despite being aware of this patent, Marvell did not investigate potential infringement or respond to inquiries from CMU and Fujitsu, indicating a significant risk of infringement. CMU argues that this behavior shows willful infringement, supported by case law stating that failure to investigate patent issues can contribute to a finding of willfulness.
In response to CMU's claim, Marvell contends that its defenses against willfulness were reasonable. Marvell claimed to believe that its technology, covered by U.S. Patent Number 6,931,585, did not infringe on CMU's patents. The court noted that patent infringement is determined by comparing the claims of the patents-in-suit to the accused technology, and that the nature of the infringement only matters for assessing potential enhanced damages. Marvell's assertion that the existence of its own patent negates infringement claims was deemed unreasonable; owning a patent does not exempt the company from infringing on another's patent.
Ultimately, Marvell's reliance on its patents as a defense was found to be misplaced, with the court emphasizing that the focus should be on the accused technology and CMU's patents. The jury concluded that Marvell lacked an objectively reasonable defense, which the court affirmed. Additionally, Dr. Wu's claim of having consulted with Marvell's legal counsel regarding the patent was contested and inadequately developed during the trial, further complicating Marvell’s position.
Marvell’s patent counsel received the '180 and '839 Patents and later obtained patents owned by a Marvell entity. Throughout the litigation, Marvell has clarified that it is not using advice of counsel as a defense against willfulness claims. The Court ruled that Marvell could not argue at trial that receiving a patent suggested a favorable opinion on whether its chips and simulators infringe the asserted claims without introducing actual communications from counsel. Despite this, Marvell’s counsel insinuated during closing arguments that their engineers had consulted with counsel regarding the patent, although the Court expressed doubts about the credibility of this testimony.
Marvell contended that it did not infringe the patents due to their complexity, a claim that does not apply to its acknowledged infringement through the Kavcic-Viterbi Simulator. Marvell argued that even the patent inventors deemed their method too complex for practical implementation, implying that its “simpler, suboptimal” solutions could not constitute willful infringement. However, the Court noted that the distinction between optimal and suboptimal solutions is irrelevant to infringement as defined by the patents, emphasizing that both types would utilize the same method.
Marvell failed to provide legal authority supporting its complexity defense, and the Court found that the reasonableness of Marvell's belief in its non-infringement would be a matter of fact for the jury. The jury was presented with substantial evidence, including expert testimony confirming that Marvell’s MNP method utilized the patented methods, leading them to reject Marvell's defense as unreasonable. The jury's finding on this matter was treated as advisory to the Court's overall determination of willfulness.
Marvell has insufficiently rebutted CMU's claim that the KavcicViterbi Simulator infringes on the patents in question. Mr. Burd asserts that the simulator incorporates the intellectual property detailed in Professor Kavcic's work and patent. Marvell's primary argument against infringement is that the patents do not cover simulators; however, this argument is deemed a “reasonable defense” that requires jury determination based on evidence and credibility. Marvell provided no evidence indicating that its employees believed the patents excluded simulators during the infringement. Instead, Marvell employees regarded the Kavcic algorithm as the “gold standard” for testing, yet failed to justify why their use of patented technology for testing negated the need for a license.
Marvell contends that CMU's delay in filing the lawsuit implies the case for infringement isn't clear enough to support a finding of willful infringement. However, the court argues that the willful infringement inquiry is based on the infringer’s conduct, not the patentee’s litigation strategy. Testimony from Dr. Wu highlights that only Marvell engineers could accurately explain the technology used in their chips. The court notes that Marvell's actions, rather than CMU's timing, are pivotal in determining willful infringement.
Marvell also claims that the court's earlier indication of a "close call" regarding its invalidity defense precludes a finding of willfulness. Nonetheless, the court clarifies that prior rulings do not equate to an objective reasonable defense. Upon reviewing the trial record, the court determines that Marvell's unsuccessful invalidity defense does not protect it from a finding of willful infringement. For Marvell to present a reasonable defense for the period from 2001 to 2009, there must be evidence that the basis for its invalidity claims was known to the infringers or a person of ordinary skill in the art. Marvell's defense largely relied on the Worstell Patent, U.S. Patent No. 6,282,251.
Marvell did not provide any evidence indicating awareness of the Worstell Patent from 2001 until the start of litigation in 2009, undermining its ability to claim a reasonable defense against infringement prior to the alleged infringement. The court noted that Marvell acted without consideration of the potential legal consequences of its actions. Even if Marvell later presented a reasonable defense during litigation, this would not negate the established pattern of conduct reflecting a high risk of infringement. The determination of patent invalidity is a factual matter for the jury, and Marvell's various defenses over the four-year litigation have consistently been deemed for the jury to decide. The court rejected Marvell's argument to deny a finding of willfulness based on earlier defenses that were not submitted to the jury.
The court concluded that the question of Marvell's objective willfulness—whether it acted with an objectively high likelihood of infringing a valid patent—should be assessed as a factual issue by the jury, in line with Federal Circuit precedents. The court ultimately found that CMU had demonstrated by clear and convincing evidence that Marvell disregarded the likelihood of infringement, granting CMU’s motion for a finding of objective willfulness and denying Marvell’s motions for judgment as a matter of law and for a new trial on this issue.
In terms of subjective willfulness, the patentee must show that the risk of infringement was known or should have been obvious to Marvell, a determination reserved for the jury. Marvell argued that no reasonable jury could find the requisite subjective intent for willful infringement; however, the evidence presented at trial did not support this claim.
CMU provided compelling evidence suggesting that Marvell's engineers were aware of Dr. Kavcic’s contributions to the technology, as they worked on projects linked to his name. Marvell’s lack of investigation into potential patent infringement indicated a subjective unreasonableness in believing it was not infringing or that the patents were invalid. Evidence showed Marvell engineers copied technology from Dr. Kavcic and Dr. Moura’s papers, as testified by Dr. McLaughlin, including Mr. Burd’s preliminary write-up of the KavcicPP detector that referenced their work and evolved into the MNP circuit. Despite Mr. Burd's claims of only following the papers and not the patents, Dr. McLaughlin asserted that the papers closely resembled the patents. Furthermore, when Kavcic's name was removed from the project, the computer codes remained unchanged. Dr. Wu acknowledged collaboration with Mr. Burd on a model aligning with Kavcic’s original proposal. The evidence of copying reinforced the Court's conclusion that Marvell acted recklessly regarding the risk of infringing the patents. Marvell's argument that CMU's reputation as an innovator mitigated its knowledge of potential infringement was dismissed by the Court, which found no evidence that Marvell took the 2003 licensing letters seriously or considered the implications of potential infringement warnings. Consequently, the Court denied Marvell's motions for judgment as a matter of law regarding willful infringement, affirming that sufficient evidence existed for a jury to conclude Marvell was aware of or should have been aware of a substantial risk of infringement.
The Court has determined that the finding of subjective willfulness is supported by the evidence, denying Marvell’s request for a new trial and granting CMU's Motion for a Finding of Willful Infringement. Marvell's Motion for Judgment as a Matter of Law, New Trial, and/or Remittitur regarding damages is also under consideration. Marvell argues that the jury’s award of $1.169 billion is legally and factually flawed, while CMU defends the award as compliant with statutory requirements and supported by the evidence.
In patent infringement cases, plaintiffs are entitled to damages sufficient to compensate for the infringement, with a minimum of a reasonable royalty as outlined in 35 U.S.C. § 284. CMU, not being a manufacturer or seller of products practicing the claimed methods, cannot claim lost profits and must prove damages through reasonable royalty. This involves a hypothetical negotiation between the patentee and infringer prior to infringement, assuming both parties are willing and the patent is valid and infringed. The Georgia-Pacific case provides factors relevant to determining reasonable royalty, which must be grounded in sound economic and factual analysis.
The jury generally has broad discretion in determining compensatory damages, which requires deference. A remittitur is appropriate if a jury’s verdict is deemed clearly unsupported or excessive, allowing the affected party to accept the reduced amount or opt for a new trial on damages.
In Martik Bros. Inc. v. Huntington Nat’l Bank, Civ. No. 08-83, the case revolves around CMU’s damages claim against Marvell, which has been a recurring topic throughout various stages of litigation, including discovery, summary judgment, and trial. Key arguments from CMU assert that Marvell directly infringed on CMU Patents during its sales cycle and indirectly infringed by encouraging customer infringement in the U.S. Evidence presented included expert testimony from Dr. Bajorek, Marvell’s SEC filings, and a joint stipulation indicating that Marvell engages in an extensive and resource-intensive sales cycle. Successful outcomes, termed "design wins," allow Marvell to become the exclusive supplier for specific hard drive generations. The sales cycle involves simulation programs to design and test chips, which are then manufactured in Asia after customer approval. Notably, most design and testing activities occur in the U.S., utilizing methods covered by the CMU Patents. CMU demonstrated Marvell's direct infringement through the use of the Kavcic-Viterbi simulator for chip testing and performance evaluation.
The document outlines the legal arguments and proceedings concerning alleged patent infringement by Marvell related to the use of patented methods in various chip simulators and engineering samples during research, development, and sales cycles. CMU demonstrated that direct infringement occurred within the U.S. through three scenarios: use of methods in Chip Simulators, Accused Chips as engineering samples, and sales activities. CMU also established claims for contributory and induced infringement based on customers' use of the patented methods.
CMU has not claimed infringement for any activities outside the U.S. nor sought damages for foreign infringement. The challenge of quantifying the use of patented methods during sales cycles has been a major point of contention. Potential valuation methods included calculating reasonable royalties for the simulators, which were deemed impractical due to their lack of market presence, and assessing a fee per use, which proved nearly impossible given the high frequency of use (up to 2.88 trillion times per day per chip).
The Court ultimately permitted CMU to present a theory based on reasonable royalties related to sales generated from the infringing use. CMU's damages expert suggested a $0.50 royalty per chip sold, estimating total damages at $1.169 billion, while Marvell's expert argued for a one-time payment of $250,000. The jury was instructed that Marvell could not be found liable for chips not used in the U.S., but could consider sales resulting from alleged infringement when determining damages, which should adequately compensate CMU without being punitive.
Damages awarded to CMU should restore it to the financial position it would have occupied without the infringement, with a minimum amount equal to a reasonable royalty that Marvell would have paid. CMU bears the burden of proving its damages by a preponderance of evidence, meaning it must show that it is more likely than not that it suffered the claimed damages. CMU seeks a reasonable royalty, defined as the fee that CMU and Marvell would have agreed upon prior to the infringement. The jury found in favor of CMU on all claims, awarding $1.169 billion in damages, which aligns with the expert testimony of Ms. Lawton, who proposed a reasonable royalty of $0.50 per chip sold by Marvell. The jury could have accepted or rejected the testimonies of either expert, and the Court does not reassess factual disputes or the jury's role in determining damages. The awarded amount was within the range proposed by both experts and reflected CMU's request, thus not warranting a new trial or remittitur. Additionally, Marvell challenged the basis for calculating damages, focusing on the inclusion of sales from chips not used in the U.S., but did not raise this issue in a timely manner during pre-trial proceedings.
CMU aims to demonstrate that Marvell's alleged infringement occurs during its U.S.-based sales cycle, involving both engineers and customers within the U.S. CMU seeks damages calculated as a reasonable royalty on all chips produced and sold during this cycle, explicitly excluding any damages for chips not used in the U.S. due to the Court's ruling on extraterritorial sales. Marvell will have the opportunity to contest the reasonableness of this royalty rate at trial.
Marvell argues that the Federal Circuit's decision in *Power Integrations, Inc. v. Fairchild Semiconductor Int'l, Inc.* limits the consideration of all its chips for calculating a reasonable royalty, but the Court finds that *Power Integrations* does not invalidate its prior rulings. In that case, lost profits from foreign sales due to domestic infringement were deemed non-cognizable damages. In contrast, CMU's claims pertain solely to domestic infringement resulting from Marvell's use of patented methods during various stages of chip development and engineering in the U.S. The Court noted that substantial evidence was presented at trial, including testimonies from experts establishing that the infringing activities occurred domestically.
All stages of the sales cycle, excluding the physical production of chips, occur in the United States, as testified by a witness who has visited the design centers of major customers, including Hitachi, Samsung, Toshiba, and Western Digital, all located in San Jose and Lake Forest, California. He provided a demonstrative exhibit listing the infringing sales activities, participants, locations, and citations supporting his conclusions. Another witness, Dr. Sutardja, confirmed that the sales cycle process for major clients like Seagate and Quantum also takes place in the U.S. Western Digital's corporate representative testified that MNP and NLD technology was tested in the U.S.
A stipulation noted that Marvell Semiconductor, Inc. (MSI) has its headquarters in Santa Clara, California, where key management and strategic decisions are made. Most of MSI’s research, design, and development activities for the accused products occur at this location, with a significant portion of its R&D team based there. Key personnel involved in the technology relevant to the litigation are also based in northern California. MSI's design and development documentation for the read channel R&D is primarily housed in Santa Clara. The manufacturing operations group supports R&D rather than commercial production, with many employees dedicated to this effort in Santa Clara. Managerial functions critical to ongoing operations are conducted by key personnel located in northern California, and relevant operational documents are maintained at MSI’s headquarters.
The Court found no evidence that any part of Marvell's sales cycle occurred outside the United States, as Marvell failed to provide substantial evidence to counter CMU's claims regarding infringement, sales, or damages. CMU demonstrated that the infringement for which it seeks damages took place within the U.S., distinguishing it from Power Integrations' foreign exploitation of a patented invention. The Court acknowledged the challenge of quantifying damages for Marvell's use of the disputed method patents, deeming a fee per use unrealistic and the absence of an established fee problematic. Marvell proposed a one-time fee of $250,000 for damages, while CMU suggested a $0.50 royalty per chip for all chips produced during the sales cycle. The Court did not endorse either figure but allowed CMU's theory to be presented to the jury, finding it equally reasonable as Marvell's.
The jury was instructed that Marvell could not infringe with chips not used in the U.S., but could consider sales resulting from alleged infringing use during the sales cycle. Marvell sought to limit CMU's royalty to chips imported into the U.S., but this argument lacked evidentiary support. No evidence was presented that any company, including Marvell, has licensed patented technology based on the number of chips imported into the U.S. The only license submitted indicated a fee structure not tied to the chips' location. Furthermore, limiting royalties based on U.S. sales posed a problem since tracking the number of chips returning to the U.S. is not feasible. Estimations presented at trial suggested between approximately 329 million and 557 million chips imported into the U.S. from Marvell's major customers were relevant to the jury's consideration.
Marvell contended that the jury should value its use of the patented method based solely on the number of chips used in the U.S., while CMU argued for consideration of all chips produced during the sales cycle, regardless of their return to the U.S. The jury had substantial evidence to support CMU's theory. The Court noted that a hypothetical negotiation would likely find a per-chip royalty agreement illogical, as it would require extensive and impractical accounting of import/export data, which is often unavailable. According to Federal Circuit precedent, if precise apportionment between infringing and non-infringing items is impossible, the infringer bears the risk; thus, Marvell must face the consequences of its record-keeping failures. CMU's damages claim includes those for Marvell's indirect infringement, which could have increased the award significantly.
Regarding the location of sales, CMU seeks damages related to sales arising from Marvell's infringement during the sales cycle. Marvell, in its response to a Rule 50(b) Motion, claimed CMU did not provide sufficient evidence that sales occurred in the U.S., referencing the Power Integrations case. However, Marvell's liability is based on "use" under 35 U.S.C. § 271(a), meaning that U.S. sales are not strictly necessary for CMU's claims. Marvell failed to present evidence that its sales occurred outside the U.S., and the Power Integrations case addressed lost profits rather than reasonable royalties. Finally, Marvell argued that there was a lack of evidence regarding the location of delivery, buyers, legal title transfer, and sales contracts, suggesting a potential waiver of its claims.
The jury was not instructed on the legal standards for determining the location of sales, an argument raised by Marvell only in its reply brief, which CMU contests due to Marvell's failure to present this issue in its two Rule 50(a) motions during trial or the initial post-trial Rule 50(b) motion. Under relevant case law, a defendant waives the right to raise an issue in a Rule 50(b) motion if it was not sufficiently specified in a Rule 50(a) motion, preventing the opposing party from adequately preparing. Throughout the trial, Marvell did not contest the sufficiency of evidence regarding the location of sales, which the Third Circuit mandates must be raised early to allow for potential rectification. The court noted that Marvell’s claims regarding jury instructions on the legal definition of sales were also waived, as these were not included in proposed jury instructions or preserved during the charge conference. The court's instructions required CMU to demonstrate that Marvell "sold or offered to sell" infringing products in the U.S., and no request for a specific definition of sales was made by Marvell. A party’s failure to preserve objections to jury instructions bars later challenges unless the error is fundamental and results in a miscarriage of justice. Even if Marvell had not waived this argument, the court found ample evidence for a jury to conclude that sales occurred in the U.S., particularly given that Marvell's headquarters is in California, as agreed upon by the parties.
Management and strategic decision-making for MSI primarily occur at its headquarters in Santa Clara, California, where the majority of design and development documents related to MSI's read channel products are also located. Key personnel, including Alan Armstrong and Bill Brennan, who have essential roles in sales and marketing for these products, are based in this location. Nearly all sales and marketing documents regarding the accused products are found in Santa Clara.
Dr. Armstrong testified that Mr. Brennan approved sales from California, while Dr. Bajorek noted that Marvell’s sales predominantly occur within the United States, identifying the domestic offices of major customers. A previous court decision outlined factors for determining sales in relation to 35 U.S.C. 271, which include the locations of sales negotiation, performance, and delivery, among others.
Evidence presented suggests that sales tied to Marvell’s cycles were made in the U.S., with no contradictory evidence indicating sales occurred elsewhere. Marvell did not present any evidence of international sales during the trial and did not challenge the assertion that sales occurred exclusively in the U.S. The court's decision in prior cases, such as Lake Cherokee v. Marvell Semiconductor, which excluded evidence of foreign sales, is referenced but distinguished based on the context of the legal analysis of what constitutes a sale under 35 U.S.C. 271(a).
Judge Gilstrap concluded that Marvell's foreign activities were undisputed, noting that purchase orders were submitted to MAPL in Singapore, with accused products manufactured and delivered abroad without entering the U.S. The absence of evidence from Marvell, such as purchase orders, delivery receipts, or testimony regarding overseas activities, hindered Marvell's ability to control the narrative presented to the jury. Consequently, the jury found Marvell liable for contributory damages, necessitating a finding that Marvell sold or offered to sell the accused products in the U.S. The court determined there was sufficient evidence to support this finding, rejecting Marvell’s requests for judgment as a matter of law (JMOL), a new trial, or remittitur.
Marvell argued that Carnegie Mellon University (CMU) failed to meet the "but for" standard for damages, claiming CMU could not prove that sales of the accused chips were solely due to Marvell's infringement. Marvell provided several points, including that no reasonable jury could determine the claimed methods were essential for its customers, that non-infringing components contributed to its success, and that the sales cycle's location was irrelevant. CMU countered that sufficient evidence existed for a jury to find Marvell's infringement was the "but for" cause of sales. The court agreed, suggesting Marvell's arguments exaggerated the requirements for causation. It reaffirmed that CMU’s theory was reasonable, viewing the evidence favorably for CMU, and clarified that sales could indicate the value of Marvell's use of the patents, provided a causal link to the infringement was established. The court concluded that CMU met its burden of proof regarding the "but for" causation.
The parties agreed, as noted in the trial record from December 11, 2012, that Marvell's sales cycle, lasting eighteen to thirty months, is crucial to its operations and primarily conducted within the United States by staff based in Santa Clara. Most of Marvell's research, development, and customer interactions occur there. A "design win" is achieved when a customer selects Marvell for chip production, leading to orders for integrated circuits for entire product generations. Expert testimony from Dr. McLaughlin indicated that Marvell's simulators and NLD and MNP-type chips infringe on CMU's patents, as they utilize claimed methods during the sales cycle, supported by evidence showing the use of infringing chips throughout this period. Additionally, expert Dr. Bajorek opined that all sales of NLD and MNP-type chips stem from this U.S. sales cycle, with the technology being essential for those sales. Sales data from Marvell, which was not disputed, indicated over 2.3 billion chips sold between March 2003 and July 2012, suggesting that Marvell's use of the claimed methods was a necessary factor for its sales. Marvell's arguments regarding customer demand for the MNP technology and the impact of other product lines on sales raised factual issues for the jury to consider. Despite Marvell's claims of customer disinterest in the MNP, testimony from Western Digital's Iftiqar Baqai countered this, as he indicated demand for the MNP during his tenure, though his involvement was limited to read channel activities until 2005.
Western Digital purchased one billion chips that utilized the infringing technology. Mr. O’Dell, a witness for Marvell, noted customer interest in removing the MNP feature but lacked direct experience during the critical 2004-2010 timeframe. CMU provided expert testimony from Dr. Bajorek highlighting the necessity of the technology, internal Marvell communications indicating the MNP as a critical requirement, and testimony from Western Digital’s corporate representative, Teik Ee Yeo, confirming the technology's beneficial use. Additionally, data showed that the majority of Marvell's chips sold contained the infringing technology. Thus, Marvell's claims of undisputed evidence were found unconvincing, leading to factual disputes resolved in favor of CMU.
Regarding damages, Marvell argued that CMU failed to provide evidence for a running royalty linked to product sales. CMU countered that sufficient evidence existed for a jury to determine that a running royalty was appropriate based on the hypothetical negotiation between the parties. The court noted that there was no established royalty for the patents involved, and both parties would need to negotiate the financial terms, including whether to adopt a running royalty or lump sum payment. CMU presented evidence that Marvell had previously engaged in running royalty agreements, suggesting that a reasonable jury could conclude similarly for this case.
ARM provided testimony regarding CMU's past licensing agreements, including both running royalty and lump sum licenses, which Marvell did not contest. Evidence included a Showa Denko license where CMU agreed to a running royalty, and a "Highly Speculative Forecast" document that listed CMU's involvement in running royalty agreements. Dr. Wooldridge corroborated this by stating CMU was a party to such agreements. The Court concluded that sufficient evidence existed for a jury to reasonably determine that the hypothetical negotiation between the parties might have resulted in a running royalty agreement.
Marvell's argument that CMU improperly included sales to Seagate Technologies in the royalty base was rejected, as the Court found no evidence that Seagate exercised its "have made" rights under the license, which sustained the potential for damages against Marvell.
Regarding the royalty rate, Marvell challenged Ms. Lawton's $0.50 per chip estimate as speculative and unsupported. The Court held two Daubert hearings to evaluate Ms. Lawton's qualifications and methodologies, ultimately determining that she was qualified and applied a reliable methodology for her damages opinion. Consequently, the Court denied Marvell's motions for judgment as a matter of law, a new trial, and remittitur concerning the reliability of Ms. Lawton’s testimony.
Ten minutes prior to Ms. Lawton's testimony, the Court excluded her prepared slides, viewing them as overly summarizing for a trial context. Ms. Lawton's calculations indicated a price-per-chip of $4.42 and an operating profit-per-chip of $2.16, derived from Marvell’s internal sales data. She referenced comments from Marvell’s CEO and a marketing executive suggesting a target gross margin of 50%, while her analysis of gross margins from 2000 to 2013 revealed an approximate profit of 59.6% on MNP and NLD chips. Based on her findings, she determined that Marvell earned $0.42 in “excess profits” per unit sold above the target margin.
Ms. Lawton also performed an operating profit premium analysis, comparing sales data from 2003 for Marvell’s chips sold to Maxtor and Toshiba, which yielded an operating profit delta of $0.06 to $0.72 between chips with and without the MNP/NLD. This range was considered indicative of potential negotiation outcomes. Ultimately, she calculated a reasonable royalty of $0.50 per unit, leading to total damages of $1.169 billion based on 2,338,280,543 accused chips sold since March 2003.
In contrast, Marvell’s damages expert, Mr. Hoffman, did not analyze sales data and instead suggested a one-time payment of $250,000 based on prior agreements, lacking the depth of Ms. Lawton’s analysis. His testimony was perceived as unfocused and merely repetitive of existing agreements, leading the jury to favor Ms. Lawton's comprehensive analysis. The jury even requested access to both experts' full reports during deliberations, highlighting their serious approach to evaluating the expert opinions presented.
The Court ruled that the determination of the weight and sufficiency of Ms. Lawton's testimony is exclusively within the jury's purview. It found no impropriety in the data points she utilized, thus denying the motion for judgment as a matter of law (JMOL) and rejecting the need for a new trial or remittitur. Marvell contended that CMU's claim of $1.169 billion was disproportionate to its prelitigation value assessments and argued that JMOL was justified due to Ms. Lawton's alleged failure to consider various evidentiary points, including DSSC membership licenses, letters indicating lack of interest in licensing, a 2004 Intel Subscription Agreement, and projections deemed speculative. However, the Court noted that Ms. Lawton adequately addressed these concerns in her expert report and testimony, explaining how the agreements influenced her opinion. The jury had access to all evidence that Marvell claimed was overlooked, and the weight of this evidence was ultimately for the jury to determine.
Determining the admissibility and weight of Ms. Lawton's testimony is solely the jury's responsibility. Marvell's motions for judgment as a matter of law regarding disproportionate damages are denied. Marvell asserts that CMU lacks evidence linking the claimed algorithm to customer demand, failed to apportion the algorithm's value to the accused MNP chips, that a calculated price premium for a minor customer does not apply universally, and that CMU's excess profits analysis is disconnected from the algorithm's value. The Court finds these arguments pertain to the weight of Ms. Lawton's testimony rather than its admissibility, as she addressed each point and was cross-examined by Marvell. The Court has previously evaluated her testimony and the validity of her price premium and excess profits analysis, noting she based her reasonable royalty opinion on Dr. Bajorek’s assessment that the technology was industry standard and crucial for Marvell’s survival. It is permissible for an expert to rely on another expert's opinions to form their own under Rule 703, provided they do not merely replicate those opinions. Ms. Lawton also relied on Dr. McLaughlin’s assessments regarding alleged algorithm infringement to establish liability. It is standard for damages experts to assume liability when testifying about damages. Challenges to the factual bases of expert opinions affect the weight of the testimony, not admissibility. The case exemplifies competing expert testimonies where the jury's belief in the underlying facts determines the outcome. The Court concludes that the jury’s damages award was supported by substantial evidence, resulting from a fair trial.
All evidence and expert testimony presented by Marvell were submitted to the jury, who adequately evaluated the credibility and weight of this evidence in determining damages, ultimately ruling in favor of CMU. Ms. Lawton’s expert opinion remains reliable despite its unfavorable conclusion for Marvell. Marvell's motion to challenge the jury's decision is denied. The court highlights that Marvell knowingly engaged in strategic risks, including using technology potentially covered by CMU's patents and failing to maintain sales records. Despite knowing the stakes, Marvell chose to proceed to trial without mediating, disregarding the significant potential damages communicated as early as January 2012.
Marvell's litigation strategy, characterized as "all or nothing," included a claim that the parties would have settled for a one-time payment of $250,000, which the jury rejected. By not presenting evidence of a lower running royalty or details about its business practices, Marvell limited its own defense and allowed CMU’s expert, Ms. Lawton, to dominate the case. Her analysis on Marvell's sales of the accused technology was pivotal, while Marvell's expert, Mr. Hoffman, failed to provide robust counterarguments or evidence. Consequently, the jury's decision was influenced by Marvell’s lack of substantial evidence regarding its business operations and products.
The jury faced a dilemma in determining damages, limited to either $250,000 or $1.169 billion. Marvell's strategy should have centered on persuading the jury of a lower royalty if liability was established, rather than solely attacking CMU’s witnesses. CMU presented three credible experts who effectively supported their claims, while Marvell’s employee testimony lacked persuasiveness against evidence indicating infringement and patent validity. Notably, Marvell’s technology sales data revealed that a significant portion of their chips relied on the Accused Technology over a decade, generating $10.34 billion in revenue and $5.05 billion in profit, leading to Marvell dominating 60% of the market. The Court, acting as an impartial reviewer, found sufficient evidence to uphold the jury's substantial damages award, emphasizing that the jury was appropriately tasked with determining the nature and extent of damages. Consequently, the Court denied Marvell's motions for judgment as a matter of law, a new trial, and remittitur, allowing the case to proceed to jury deliberations.
The Court has denied Marvell's motions for judgment as a matter of law and for a new trial concerning both non-damages and damages issues. It partially granted CMU's motion for a finding of willful infringement but reserved its ruling on enhanced damages for a later opinion. The Court has also postponed its decisions on several of Marvell’s motions, including those related to laches, CMU's requests for a permanent injunction, post-judgment royalties, supplemental damages, and interest, pending further written opinions. CMU’s request for attorney fees was denied without prejudice. The Court emphasized that in considering motions for judgment as a matter of law, it must view evidence favorably towards the nonmoving party. The document references expert testimony and technical details relevant to the case, particularly relating to the magnetic recording disk drive technology, including the transition from longitudinal to perpendicular recording and its implications for data storage. Additionally, it notes changes in CMU's leadership, with Dr. Cohon stepping down as president.
Dr. Kryder, a professor of electrical and computer engineering at Carnegie Mellon University (CMU), holds a B.S. and a Ph.D. in electrical engineering and physics from Caltech. His prior experience includes teaching at the University of Regensburg and working at IBM's TJ Watson Research Center, with expertise in magnetic recording heads and media, though he lacks high-level signal processing expertise. The legal parties dispute the relevant timeframe for the inventions and the associated rights, as well as which DSSC companies utilized the patents under their DSSC licenses.
Dr. Moura, originally from Mozambique, earned his electrical engineering degree from the Technical University of Lisbon and a Ph.D. from MIT, later becoming a CMU professor in 1986. Dr. Kavcic, from Yugoslavia, received his B.S. in electrical engineering from a German university and his Ph.D. from CMU. Robert Wooldridge has served as the Director of the Center for Technology Transfer and Enterprise Creation at CMU since 2001. Mr. Mahler was not called as a witness.
CMU’s partnership with Intel extended beyond a subscription agreement, involving significant investments in research and development totaling approximately $150 million, including sponsorships and the establishment of Intel Science and Technology Centers on campus. An email from Dr. Moura to the Intel Program Manager requested inclusion of patent 6,201,839 in a deal, highlighting that both he and Dr. Kavcic are inventors.
Dr. Sutardja, from Indonesia, holds a B.S. from Iowa State University and a Master's and Ph.D. from UC Berkeley, has worked at several tech companies, and is named on approximately 300 patents. Mr. Burd, originally from Moscow, has a strong academic background in mathematics and joined Marvell in 1999. Dr. Nazari did not participate in the deposition or trial but is currently employed at Vital Connect with colleagues Mr. Doan and Mr. Brennan.
An email from an individual named Alex to Nersi discusses the existence of a detector developed by Marvell, which allegedly incorporates ideas from a previous talk given by Alex and a GLOBE-COM 98 paper he shared. Alex inquires if there is a write-up available about the detector and mentions he is seeking employment as he is graduating soon. The email includes a broken URL for his resume and publications. The GLOBE-COM paper relates to ideas in certain patents but is distinct from a later IEEE paper referenced by Burd. The court suggests that Dr. Nazari may have shared Dr. Kavcic's work with Marvell.
In evaluating evidence for a motion for judgment as a matter of law (JMOL), the court must favor the non-moving party's reasonable inferences. The response to Alex's email was not submitted at trial, limiting the court's ability to consider it in the current motion. However, during post-trial hearings, Marvell provided Dr. Nazari's response, indicating they did not have a product relevant to Alex's work at that time but were hiring. It is noted that Marvell could not implement iterative coding on chips until 2007-2008.
The document references various components of System on Chip (SoC) architecture, including the central processing unit and hard drive controller, and defines a "foundry" as a semiconductor fabrication business. Mr. Brennan, Marvell's Vice President of Sales, approved all deals before leaving the company. The acquisition of Maxtor by Seagate in 2006 is mentioned, along with testimony from Teik Ee Yeo and Iftiqar Baqai on behalf of Western Digital.
The parties agree on several facts presented at trial, though the implications of these facts were hotly contested. There are approximately 206 accused chip models. Dr. McLaughlin, an expert in electrical engineering, testified without challenge from Marvell and has professional connections within the small community focused on this technology. He has extensive experience in research, academic instruction, and commercial chip implementation, and provided an expert report and deposition on infringement opinions in compliance with Rule 26(a).
The Court accepted Dr. McLaughlin as an expert in signal processing, acknowledging the challenges in clarity from witness testimony regarding circuit diagrams during the trial. To enhance comprehension, the Court mandated the submission of all demonstrative materials used, including Dr. McLaughlin’s presentation labeled P-Demo 7. Dr. Blahut, a professor at the University of Illinois with significant experience in signal processing and read channels, was also accepted as an expert after passing a Daubert challenge. The Court cited the legal standards for patent invalidation due to anticipation and obviousness, referencing relevant case law. Dr. Proakis, an adjunct professor with extensive experience in digital signal processing and read channel technologies, was accepted as an expert without any Daubert challenges against him. The hypothetical negotiation date between the parties was established as March 6, 2003. Dr. Bajorek, an expert in the hard disk drive industry with a Ph.D. from Cal Tech, was accepted as a witness despite a partial Daubert challenge from Marvell. The Court defined an "industry standard" technology as one adopted by the majority of drive makers across two or more generations. Ms. Lawton, a damages consultant with 27 years of experience, was accepted as an expert in IP damages following a thorough examination of her credentials, amid a partial Daubert challenge from Marvell. Mr. Hoffman, a CPA specializing in intellectual property negotiations and damages consulting, was also mentioned as part of the expert testimony landscape.
CMU's Daubert challenge against Mr. Hoffman was partially granted and partially denied, with the Court accepting him as a damages expert for intellectual property. Following this, on July 7, 2009, Marvell's motion to transfer the case to the Northern District of California was denied on September 21, 2009. Marvell submitted an amended answer on April 29, 2010, and CMU responded on May 28, 2010. A seven-hour Technology Tutorial took place with experts Dr. McLaughlin and Dr. Proakis, leading to a two-day Markman hearing, resulting in the Court issuing an order on the disputed claims' meanings.
A Pretrial Order was entered on October 10, 2011, scheduling jury selection and trial preparations. Marvell's motions for partial summary judgment of invalidity regarding U.S. Patent Nos. 6,201,839 and 6,438,180 were denied, including a motion for reconsideration. In spring 2012, both parties filed summary judgment and Daubert motions, with hearings held on July 10-11, 2012. The Court granted and denied parts of Marvell's motions regarding non-infringement and damages. Specific claims from U.S. Patent No. 6,201,839 and U.S. Patent No. 6,438,180 were found not to infringe. Various motions to exclude expert testimony were granted or denied in part. CMU and Marvell filed their Pretrial Statements in September 2012.
Witness Lists and Offers of Proof were submitted by CMU and Marvell on October 8 and October 15, 2012, respectively. In accordance with the Pretrial Order, Motions in Limine were filed on September 24, 2012, with responses submitted on October 9, 2012. Prior submissions, including exhibits and deposition designations, were stricken by the Court's Order on October 24, 2012, due to extensive objections by both parties. The parties initially presented over 2,700 exhibits but narrowed their disputes to 335 by November 9, 2012, reserving deposition issues for later resolution.
The proposed final jury instructions and verdict slips were also struck for failing to comply with the Court’s requirement for joint submissions, as the parties submitted only twelve agreed instructions out of thirty-seven total. The parties were instructed to resolve these disputes, leading to the submission of joint instructions and verdict slips by email on December 19, 2012, with the Court ruling on disputes during the Charge Conference.
Several motions were addressed, including the denial of Marvell's request to submit the issue of laches to the jury and the denial of various motions regarding the testimony of expert witnesses. The Court granted Marvell’s request for an extension of trial time despite previously limiting each side to twenty hours for presenting their case.
The Court granted a motion allowing each party twenty-five hours for trial proceedings, which includes opening statements, examinations, and closing arguments. A "gentlemen's agreement" was established between the parties to streamline witness testimony, permitting certain witnesses to be called only once, with the provision that cross-examination could extend beyond direct examination. The Court was informed that this agreement specifically involved testimonies from Marvell engineers Mr. Doan, Dr. Wu, and Mr. Burd. The case, although related to patent law, adheres to the Third Circuit's legal standards for motions for judgment as a matter of law and for a new trial, referencing precedent cases such as Leader Technologies, Inc. v. Facebook, Inc. and others that establish consistent standards for Rule 50(a) and Rule 50(b) motions.
During a deposition, Mr. Burd, as Marvell's Rule 30(b)(6) witness, acknowledged the variability of parameters across different branches of the Viterbi trellis. The parties did not present information about Michael Madden to the jury but included his deposition in court filings, noting his history at Marvell from 2000 and his reporting line to Mr. Burd. Mr. Han, another potential witness, was neither deposed nor called to testify, although he was identified as a Marvell employee at the time relevant to the case. Mr. Burd confirmed that the KavcicViterbi simulator used by Marvell during testing was based on Dr. Kavcic's work, as detailed in Marvell's documentation. Additionally, two representatives from Western Digital testified, with Mr. Baqai sharing his experience in read channel chip development from 1997 to 2005, although his testimony did not address chip enablement in specific modes.
Dr. Moura acknowledged during his deposition that the branch metric equation from the '180 Patent represents a set of branch metric functions, noting its variance depends on signal-dependent artificial intelligence (AIs). Dr. McLaughlin confirmed that a modified metric described in the patent accounts for transition noise, which varies based on the type of transition. The Court denied Marvell’s request to supplement expert reports post-summary judgment, emphasizing that expert discovery was to conclude by April 6, 2012, and that Marvell should have anticipated the Court's ruling in favor of CMU based on previously known positions.
Throughout the trial, jurors were provided with notebooks containing patent details and a glossary. The Court previously addressed CMU's challenge regarding Dr. Proakis’s new opinion and deemed it unnecessary to revisit. Factors influencing the jury's evaluation of expert testimony included the length and clarity of presentations, with Dr. McLaughlin's thorough and methodical approach favorably contrasting with the more rushed testimonies of Dr. Blahut and Dr. Proakis. Dr. McLaughlin presented clear examples supporting non-obviousness, which Dr. Proakis did not contest. The Court indicated that the jury could reasonably infer that Dr. Nazari might have shared Dr. Kavcic’s work with Marvell, according to jury instructions allowing for inferences based on common experience and reasoning. When questioned about sharing information with Marvell, Dr. Kavcic denied any understanding that Marvell could freely utilize the paper’s content.
A patent was already filed, and unauthorized use without a license is prohibited once a patent is filed and ultimately granted. Kavcic communicated with Dr. Nersi Nazari to discuss his work as part of his job search. Mr. Doan, who left Marvell in October 2009, testified via deposition, and his perceived lack of bias compared to current employees was noted. Mr. Burd referenced Kavcic’s approach as a benchmark in his deposition, likening it to Ronald Reagan's association with a historical event. The '585 patent is owned by Marvell International Ltd., which is not a party to the case, leading CMU to argue that the patents are irrelevant. CMU asserted that the introduction of these patents warranted discussion of Marvell’s corporate structure and tax strategies, despite not receiving discovery on Marvell’s royalty payments to itself. The Court precluded CMU from arguing that Marvell’s tax strategy was illegal during the trial. It was emphasized that owning patents does not serve as a defense against infringement claims, and the jury was instructed accordingly. Additionally, it was noted that a patent may later be declared invalid, as Marvell has attempted to invalidate CMU's patents. The judge can allow the jury to determine underlying facts when the defense involves factual questions. This practice has been observed in similar patent cases.
The Court determines that there are likely mixed questions of law and fact regarding Realauction's defenses, necessitating jury evaluation of underlying facts before the Court can address the objective prong of willful infringement. CMU argues that Dr. Wu's deposition clarified that communications with a lawyer about the patents were solely for his prior art search related to his 2002 patent. The Court noted Dr. Wu's demeanor during testimony—his physical discomfort and repeated glances at Dr. Sutardja—could lead the jury to question his credibility. The jury was tasked with assessing witness testimony and assigning appropriate weight or potentially discrediting it. In response to CMU's motion to exclude certain testimony and arguments about Marvell’s pre-suit communications with counsel, the Court mandated Marvell to submit relevant documents for in camera review, which it considered in its analysis of willfulness but did not base its decision on these documents. The Court also acknowledged that no inference could be drawn from the documents. Furthermore, Marvell’s designee stated that a specific class written by Marvell engineers related to Professor Kavcic’s patents. Dr. Proakis testified that the concept of suboptimality is not mentioned in patent claims, indicating that whether a suboptimal version infringes depends on circumstances. Multiple witnesses discussed simulators but did not address how their opinions influenced Marvell’s decisions regarding the patents. Dr. Wu recounted that he did not inform Dr. Kavcic about naming files after him, likening it to his casual interactions with other prominent figures like Dr. Viterbi.
Mr. Burd and Dr. Wu indicated that their mention of Dr. Kavcic was in relation to the "media noise problem" he discussed in his paper. The Court referenced Rule 11 of the Federal Rules of Civil Procedure, which mandates that attorneys ensure their written submissions to the court have factual support, based on reasonable inquiry. Specifically, Rule 11(b)(3) requires certification that factual claims have evidentiary backing or will likely have such support after further investigation. Counsel is expected to exercise caution, akin to the advice to "stop, look, and listen" at a railroad crossing. Under Rule 56, summary judgment is only appropriate if there is no genuine dispute regarding material facts. A defense may still be unreasonable even if a genuine dispute was previously identified. For motions for judgment as a matter of law or new trials under Rule 50, the Court evaluates trial evidence favorably towards the non-movant. The document argues that the discovery of prior art, years after infringing activity began, does not negate claims of willful infringement, as such a defense contradicts the law's intent. Willfulness generally hinges on prelitigation conduct. The Court would err by concluding that no reasonable defense existed before the jury assessed the facts of infringement and invalidity. Allowing defenses to be presented to the jury does not imply a finding of no willful infringement. The evidentiary basis for a jury's willfulness finding remains intact even when multiple defenses are raised at trial. In the dissenting opinion of Highmark, Inc. v. Allcare Health Management Systems, Inc., Judge Moore criticized the notion that the objective prong of willfulness should always be ruled on by the judge.
The court is urged to reconsider the Bard decision en banc due to concerns raised by judges Reyna and Mayer regarding its implications on fact-finding roles and its inconsistency with precedent. The court initially aimed to present factual interrogatories to the jury regarding willfulness but found it impractical due to the complexity of the claims and the number of disputed facts. The jury evaluated Marvell’s awareness of the patents and whether Marvell had a reasonable defense against infringement claims based on its understanding of the patents' complexity and its own activities. The substantive law governing patent damages is analyzed under Federal Circuit precedent, while remittitur decisions fall under the law of the regional circuit. The court highlights a significant change in Marvell's legal representation before trial, which introduced new issues post-trial. Although Power Integrations' request for rehearing was denied, they plan to seek Supreme Court review regarding the damages issue. CMU presented strong evidence establishing that a sales cycle occurred in the U.S., and expert testimony indicated Marvell's operational necessity to remain in California for business success, despite potential infringement risks.
Moving research, design, testing, sales, and marketing operations overseas was not considered viable. The Georgia-Pacific factor regarding rates paid for comparable patents indicates that while the DSPG technology may not be directly comparable, it reflects standard contract terms relevant to Marvell and the broader industry. Evidence presented included Dr. Armstrong's testimony and CMU's global licensing strategy, which did not differentiate based on geographical use. Mr. Hoffman highlighted the uncertainty of tracking products through the supply chain, emphasizing the complexity of commerce. Under existing precedent, any adjustments based on product location would refer to Ms. Lawton's chip estimate. CMU submitted a supply agreement between Western Digital and Marvell to counter Marvell’s motion regarding extraterritorial conduct. The court noted that for judgment as a matter of law (JMOL), only properly admitted evidence is considered. Dr. Bajorek provided the sole evidence of Marvell's foreign activity, identifying Taiwanese foundries for chip production. The court allowed CMU to counter any evidence Marvell might present at trial related to tax strategies. Marvell's parent company and its Asian subsidiary were not involved in the Texas lawsuit. CMU’s liability theory is based on Marvell’s use of the patented method, not sales or offers to sell. The issue regarding Mr. Baqai’s presence in court centered on the validity of the subpoena, with acknowledgment that he may have felt compelled to appear despite the technical limitations of the subpoena's jurisdiction. Additionally, Seagate, as a member of the DSSC, holds rights to the patents-in-suit under the Associates Agreement, including the right to have inventions made.
The Court noted that Marvell did not provide adequate evidence of a valid Seagate Associates Agreement for the patents at issue, as only unsigned copies were submitted, which contained conflicting terms for the years 1992-1997. The agreement relevant for 1995-1997 would determine Seagate's "have made" rights concerning the patents. Consequently, Marvell failed to demonstrate that Seagate had a license for the patents, nor did it provide any evidence indicating Seagate exercised its DSSC rights in dealings with Marvell. The Court reasoned that if Seagate had used its rights prior to Marvell's sales, Marvell would not be liable for infringement; however, if Marvell lacked such rights, it would have infringed the patents while making non-infringing sales.
Testimony from Ms. Lawton was highlighted as exceptionally thorough, demonstrating significant mastery of the case details. Her expert report spanned over 500 pages and included 3,000 footnotes. Marvell contested her comparative analysis of chip prices during cross-examination, which was based on advice from Dr. Armstrong regarding the need for comparability among the chips. Ms. Lawton outlined specific criteria for comparability, such as identical data rates and sales timing. She assessed price differences for Fujitsu, Hitachi, and Seagate but found them lacking in comparability.
The Court also directed the jury that neither the slide presentation nor expert reports were admissible for their deliberations. Ms. Lawton acknowledged consideration of the DSSC agreements, which included a specified annual fee, during her analysis.
An expert in the case concluded that a specific measure of royalty was inappropriate due to the existence of special agreements. The Intel subscription agreement, dated September 2004, was noted as occurring approximately three-and-a-half years after the hypothetical negotiation date of March 13, 2001. This agreement was included in the expert's analysis. A highly speculative memo from January 2006 was also referenced, occurring nearly five years after the hypothetical negotiation date. Marvell's counsel argued that non-infringing storage products had higher margins than infringing ones, but Ms. Lawton pointed out that some of these products, like a Seagate chip, sold at a premium and that she excluded certain chips from her analysis due to lack of confirmation. Mediation attempts by Magistrate Judge Infante and settlement conferences were held, yet the case persists. The presence of numerous attorneys suggested to the jury the significant value of the technology involved. Marvell attempted to seal trial slides, asserting their importance, and during jury deliberations on December 21, 2012, a request for a calculator was made, which led to a delay in the proceedings until December 26, 2012. The Court later agreed to provide a calculator upon request.
The jury did not request a calculator during the trial, which the Court believes could have favored Marvell by allowing consideration of lower per chip royalties. The Court noted Marvell's counsel appeared to treat the trial as a precursor to a potential new trial, as evidenced by two of Marvell’s experts reportedly sleeping during proceedings. Key testimony from former Marvell employees, who opted for deposition designations, was beneficial to CMU, revealing a lack of engagement with the patents from former employee Mr. Doan. The Court restricted discussions of total revenue and profit figures, allowing only average values, but jurors could calculate totals based on provided data. If a new trial occurs, Marvell will be limited to an expert damages opinion of $250,000. Since the litigation began in March 2009, Marvell's failure to remove the infringing technology has resulted in significant damages, estimated at over $716 million, including nearly $20 million in potential attorney fees, which could have been avoided through reasonable settlement discussions. The role of judges is emphasized as being to apply the law impartially, akin to umpires in a game.