Renesas Electronics v. Broadcom Corporation
Renesas Electronics v. Broadcom Corporation
Seth Lloyd argues on behalf of Renesas at the United States Court of Appeals for the Federal Circuit.
Unofficial transcript for users of mofo.com
Judge Lourie (00:01): 2021-1260 and 1362. Mr. Lloyd, please proceed.
Seth Lloyd (00:10): Thank you, and may it please the court. On Renesas’s appeal, the board erred because it read plain disclosures about hardware to refer to software. The disputed limitation is functional. It simply requires checking whether a clock gate is on or off and turning the gate on or off, based on the result. The board focused on whether one reference Van Hook expressly discloses using that functionality. Although Van Hook teaches checking whether a signal processor is off the board thought that that was insufficient because a person of skill could have read that to refer to sending the signal processor a software instruction rather than controlling the processor’s clock. The board based that finding on testimony from Broadcom’s expert, Mr. McNair, who said that another possibility for controlling a processor was to send a software halt instruction. But the problem with that finding and the testimony on which it was based is that the relevant disclosures in Van Hook expressly do not use a software halt instruction.
Seth Lloyd (01:16): Instead, the relevant disclosures at Appendix 970 make clear that Van Hook teaches to use hardware control registers to help the processor, as our expert explained and was unrebutted on this point, “This is a hardware discussion.” That is at Appendix 2774 to 2777. He said, “These are not instructions,” and that skilled artisans would have known that when you’re talking about hardware registers, the hardware registers are “directly controlling something in the circuit.” That’s at the same cite I gave before. And so the problem here is you had testimony that talked about generally. Generally, it is known that you could control a processor using software instructions, but that testimony, that the critical testimony in which the board relied never connected that general knowledge about software to anything in Van Hook, and there would be nothing to connect it to in Van Hook, because in fact, Van Hook uses a hardware mechanism, and our expert went further. He didn’t just explain that hardware would be used, but he explained how a person of skill in the art would understand the specific disclosures of Van Hook. And it was not just that there are control registers, but that in addition to control registers, Van Hook gives a specific architecture for its clock scheme in which what our expert said was a comment.
Judge Stoll (02:44): So counsel, this is Judge Stoll. Looking at this record, it seemed to me that there was testimony from your own witness saying that Van Hook could have disclosed a different way. That there’s different ways that you could implement a halt signal, and that there’s ways for example, by controlling the clock signal, but there’s other ways that the halt signal could operate. Is that true?
Seth Lloyd (03:13): No, I, that’s not true, Judge Stoll. I think the testimony you’re referring to is at Appendix 2775, where our expert was asked whether—can it mean send a control signal to the signal processor rather than simply cutting off the clock. And our expert did say, “I can’t rule that out entirely.” But the question there that it, can it mean, was not referring to something in Van Hook. The discussion had started with our expert saying that in 2773, our expert said that the ordinary meaning, the ordinary usage of the term halt even outside of Van Hook, generally would refer to freezing, or halting would refer to the clock. So, the discussion was about the word halt by itself. And so he did concede that the word halt by itself might have meant something broader. But that was the extent of the concession, because the rest of his discussion, he repeatedly said in, in 2773, 2774, 2775, and 2776, that that’s clearly not what Van Hook means.
Seth Lloyd (04:21): He said 2776, “That it’s unambiguous that Van Hook is not referring to software. This is not a software discussion. This is a hardware discussion.” So, I think you have to read that testimony in context, and in context the only thing he said was that the word halt by itself, but we’re not talking, and I think that’s where the board got off track. Judge Hal—Judge Stoll, simply looked at the word halt and said, “well could the word halt mean something more,” but you have to consider the word halt in the context of the specific disclosure of Van Hook. And in the context of Van Hook, it doesn’t just say halt the processor any way you want. It says halt the processor using these control registers, and on that issue, on what it would mean to halt the processor using control registers, the only testimony was from our expert. it’s like just, I know this—
Judge Stoll (05:17): He, that Professor McNair’s testimony should be ignored
Seth Lloyd (05:22): That, sorry, I missed part of the question.
Judge Stoll (05:25): I’m are you—what is your position with respect to professor McNair’s testimony, which is raised as being contrary to that of your expert? I mean, because it seems like a battle of the experts here. So how do you explain away Professor McNair’s testimony?
Seth Lloyd (05:43): It’s not a battle of the experts, Judge Stoll, because only one expert actually spoke about the relevant issue. So if you look at the testimony that the board relied on, that’s at Appendix 2688 to 89, and this is paragraph 52. The first part of that paragraph does say, Mr. McNair does say, “I disagree with Dr. Colwell’s conclusion about Van Hook,” but then the critical part of the paragraph, the second part simply says, that generally he says, this is 26889. That another possibility would be to send a control signal to the processor, to stop operation. That it is very common for computers to have a halt instruction, and that the halt instruction would generally operate this way, but he never ties that to anything in Van Hook, because there would be nothing to tie it to in Van Hook, just to kind of make this less abstract.
Seth Lloyd (06:37): It would be as if the question were, whether traveling from point A to point B via automobile would involve tires. And you had an expert who came and said, “Well, I disagree with that because another way to travel from A to B would be to take a train. And that would in fact, be a very logical way to do it.” That testimony would not actually speak to the relevant issue, which is whether traveling from A to B via automobile involves tires, and that is the kind of testimony that you have here. You have testimony saying, “Well, another way to halt a processor would be to send a software instruction,” but that testimony does not answer the question about what does it mean to halt a processor using hardware control registers, which is what the art discloses. And this court has consistently held in the cases like Erickson, MobileMedia, and the other cases that we cite, that even when you have an expert that is not the end of the story. The substantial evidence standard of review is a standard of review and requires this court to ask whether a reasonable fact finder could in fact draw the conclusion from that testimony,
Seth Lloyd (07:41): And here the answer is no.
Judge Stoll (07:44): In his testimony, it does rely on something in Van Hook. He relies on Van Hook ’s statement about the display processor 500. Why is that not sufficient?
Seth Lloyd (07:55): Well, I think the display processor 500 was actually for something that the board talked about in its decision. I don’t remember seeing that in Mr. ’McNair’s testimony, Your Honor, but that paragraph, if anything, only supports us and it certainly would not support the contrary conclusion, because when you look at that paragraph, it in fact uses software terms when talking about software. It refers to passing a command to the display processor, which was a different part of Van Hook, but it never once suggests that when you talk about using control registers, that that would involve software. So that's my answer on that question. I would like to reserve the remainder of my time to address the cross‑appeal and rebuttal.
Judge Lourie (08:42): We will save that for you, Mr. Lloyd, and Mr. Johnson.
Brian Johnson (08:47): Yes. Good morning. Again, this is Brian Johnson, may it please the court.
Judge Lourie (08:51): And you want to save two minutes for rebuttal on the cross-appeal?
Brian Johnson (08:57): That’s right, Your Honor.
Judge Lourie (08:58): Right. Proceed.
Brian Johnson (09:02): Your Honor, I’d like to address during this segment, both the appeal and cross-appeal, but I will start with the Renesas appeal in this case, Your Honors. Renesas below relied on Van Hook as a secondary reference. And their reliance was also on a technical misunderstanding of that reference. And the board should be affirmed because it conducted a careful review of all the evidence and came to the right conclusion on the technology. And that’s a conclusion that was fatal below, and fatal to Renesas’s appeal now. Claim 17 relates to a processor that determines the status of at least one gate that controls the flow of a clock signal. And Renesas’s argument was simply that a particular register provided information about a co-processor’s clock. Now, Van Hook in general, didn’t have to do with power management or clock gating. So this this was expert testimony that they believe supported this, and they simply didn’t have that testimony.
Brian Johnson (10:02): And that’s what the board acknowledged. Responding to a couple of arguments that Renesas’s counsel made, they argue that Mr. McNair’s testimony was somehow unrelated or outside of the context of Van Hook, and they refer to his testimony specifically about the fact that stopping a processor by stopping a clock would create errors in transition states, which was pivotal testimony the board relied upon. He highlights paragraph 52, but the full context is 50 to 54, which identifies Van Hook nine times, as I recall. It’s clearly grounded in that context, and that was precisely what he was referring to. And the board also conducted an analysis of Van Hook to further confirm why it supported Broadcom’s expert in that case, specifically finding that the halt command meant allowing a processor to stop its existing tasks, and that’s inconsistent with Renesas’s position because stopping a clock would do so instantly.
Brian Johnson (11:05): So this was a well-reasoned analysis. I’d like to also briefly address the testimony from Mr. McNair or pardon me from Mr. Colwell in deposition. Context does matter. It’s very important. And so is reading that entire section in context, the appendix is 2775 to 76. Dr. Colwell did make a statement about the use of halt generally, and then the questioner specifically asked him to pin down his context to Van Hook, asking what it meant in Van Hook. And that’s where the critical testimony came, where he conceded that it could be done with a control signal and that in the context of Van Hook, he could not rule out that, that’s what it was referring to. So his testimony, and all the expert testimony that the board credited, was within the proper context. If there’s no further arguments on that, Your Honor, or no further questions, I’d like to move on to Broadcom’s cross-appeal on claim 25.
Brian Johnson (12:06): Your Honor, on Broadcom’s cross-appeal, the board’s finding the claims 25 and 26 were obvious over Alben in view of Fallah’s should be reversed. Claim 25 represents an improved clock gating feature, as the ’538 patent explains, historically clock gating has been conducted with hardware control alone. That was the typical technique. And that didn’t work for Broadcom who had customers with specific application needs not anticipated when the hardware was developed. So for that, Broadcom developed a dual control mechanism, both hardware and software control, and the software, which generally resides at the system level, had the ability to directly control the status of clock gates at the circuit level. Now this dual control of this exact same logic gate created conflicts in control, and Broadcom had a solution for that. It developed the overriding feature to resolve those conflicts and all of that’s in claim 25D.
Brian Johnson (13:02): Now the board erred in relying solely on references that disclosed the admitted prior art technique of the ’583 patent, hardware control clock gating. And we believe the board should be reversed for three reasons. First, the board improperly found that software controlled clock gating was obvious, even though it wasn’t disclosed in any asserted reference. And the board also relied on motivation far too generic to render obvious Broadcom’s specific overwriting limitation. And finally, the board did not address reasonable expectation of success at all, and resurrected an abandoned argument by Renesas. On the first issue, the board should be reversed for finding that software clock gating was obvious when it wasn’t disclosed anywhere. Alben, the primary reference, is a hybrid hardware, software, power management scheme. It discloses both hardware and software and software is on the system level. In fact, Alben describes that. It says that it changes the power management mode of the subsystems, and it can do that by changing it between activated and full-power mode, that system level control.
Brian Johnson (14:07): Now, it does describe circuit level control, as well, specifically, clock gating control, and there, hardware control is how it’s conducted. That’s what the board found. In fact, they found that issue undisputed. That Alben uses hardware control unit 12 to control the status of clock gates. And ultimately the board was reasonably forthcoming in what it believes Alben lacked to reach the asserted claims. This is the wording of the board, and the numbering is my own gloss, but the board believed that Alben lacked or Alben must be modified in three different ways. That Alben CPU four must one, directly control the status of a gate independent of control unit 12; two, to override the power management decisions made by control unit 12; and three, directly overwrite a status of off or on that was previously written to by control unit 12. Now, the board relied on two other references to allegedly cure those deficiencies, but they have precisely the same deficiencies.
Brian Johnson (15:05): Both of them describe hardware, software, power management schemes that are similar to Alben. Alben is already a fine tuned hybrid system. And it understands the efficiencies of both hardware and software, and nothing within those references’ would’ve motivated Alben or one of skill to modify Alben in any way. Fallah is the secondary reference and Fallah is a textbook, and it describes precisely the same types of procedures. It refers to what are called DPM algorithms and explains what those are. They refer to selectively slowing down or shutting off system components that are idle. That’s system components, and changing them between active and standby mode. That’s precisely the same thing Alben does when it changes the power management mode of the subsystem there, it uses full power and automatic mode, but it’s the same system level control. Now Fallah does describe circuit control, or at least describe clock gating, but that comes several sections later, 13.3, that discusses circuit level power management, and there clock gating is exclusively in hardware. As we describe in our briefing, there are two forms of hard or clock gating described, both of them being in hardware. So this reference is cumulative at best and with respect—
Judge Stoll (16:24): Well, this is Judge Stoll. I was reading Fallah as saying that the dynamic power management, which includes the clock gating, could be implemented in software or hardware. That’s at J8777. Do you disagree with that?
Brian Johnson (16:41): I disagree with that characterization. What it refers to is DPM algorithms and it’s on Appendix 774, that it describes what those are, where they selectively slow down and shut off system components. And it’s the next section 13.2, that describes how the DPM algorithms work. Taking a step back, Fallah is a textbook about power management and the title itself distinguishes these two levels I’m referring to. The title is referred to as Circuit and System Level Power Management. So 13.2, which describes the system level power management, does describe DPM algorithms. The clock gating comes in the next section, that’s 13.3, and that is the circuit level, and there is no discussion of software whatsoever. No mention of DPM algorithms whatsoever. So what the introduction section that, Your Honor, is referring to is in Appendix 777, is referring to system level control and precisely the same thing that Alben does and Benini is even worse.
Brian Johnson (17:43): And in fact, teaches a way in that it describes clock gating in a section that the board never addressed. And there, it actually explains why hardware control clock gating is best. And it describes that as being ideal because it’s faster. And that’s again, precisely the ’583 patent admitted prior art. And so these references are cumulative at best and teach a way at worst. And either way, they wouldn’t motivate one of skill to modify Alben at all at the fine-tuned system that understood the efficiencies of hardware and software and already utilized it to maximize those efficiencies. But even if we accept what the board has said about motivation, and we accept the specific motivation that we found, one of skill would not have actually recreated the overriding limitation. So the motivation that they found was additional software in power management control, and power management is a general concept, particularly when in the context that that motivation is found, it’s hard to understand why one of skill would’ve applied that to clock gating at all,
Brian Johnson (18:48):
and certainly would’ve done that with respect to the overriding limitation. Recall, the overriding limitation had to do with dual-control and resolving a conflict in two types of control of the same gate. So simply adding software wouldn’t have gotten to that claim element. And the facts here are similar to that [inaudible]. And that case had to do with thermal conductivity at a certain level, 42. All the prior art references in that case disclosed something less, but the board found that limitation obvious because one of skill would’ve been motivated to increase thermal conductivity. Now this court vacated. Even accepting that motivation is this court explained one of skill wouldn’t necessarily have reached 42, and the board has the same problems here. Even if we accept that one of skill would want to use software, how would they have actually—one of skill wouldn’t have applied that to clock gating or the overwrite technique.
Brian Johnson (19:45): And in fact, the facts here justify reversal even more so [inaudible], because unlike Thermo conductivity, hardware and software is not a sliding scale where if you use more and more and more software, you somehow arrive at the claimed invention. This is a specific implementation, not disclosed anywhere. And it’s improper for the board to try to disclose motivations or to supply motivations through, or pardon me, undisclosed elements through obviousness. And that’s the hair wear case that explains if you’re going to supply missing elements, not disclosed anywhere, it can only be done on very peripheral issues. And this is not a peripheral issue at all. This is precisely the issue that Broadcom uses to distinguish the prior art that I referred to that’s in column two and column five of the ’583 patent. This is also the claim element that the examiner found distinguished the prior art there. It’s the reason Broadcom has a patent.
Brian Johnson (20:39): So this is an important issue. And ultimately, those two claim elements are not disclosed anywhere and it’s improper for the board to create them. That’s hindsight. And finally, one other issue the board should be reversed for not conducting any analysis on reasonable expectation of success, and that’s the board’s burden to do. And that’s particularly problematic in this case, because as we’ve explained in our briefing, they created a combination of Alben and Fallah that no party advocated for. Broadcom identifies a particular conflict in the combination that Renesas tried to create, and Renesas affirmatively walked away from that combination. The board resurrected that and resolved the conflict on their own, creating a new combination, making Broadcom unable to discuss reasonable expectation of success at all. And, Your Honor, with that, I’ll reserve the rest of my time for rebuttal.
Judge Lourie (21:34): Thank you, Mr. Johnson, we will hold it for you. Mr. Lloyd has considerable time remaining.
Seth Lloyd (21:43): Thank you, Judge Lourie. I’d like to begin where my colleague left off on Broadcom’s appeal and then touch a couple of rebuttal points on Renesas’s main appeal. On Broadcom’s appeal, Broadcom’s case depends on asking this court to draw different inferences from the references and to read the references in a cramped way that the board rejected on the facts. Start with the point that Broadcom says that the prior art, only discloses hardware control of clock gates. The board in fact found the opposite. The board at Appendix 31 to 32 expressly found that Alben teaches a hybrid approach. It similarly found for Appendix 20, at Appendix 26, Appendix 28, and Appendix 33, that Fallah also discloses a hybrid approach. Fallah teaches, in the board’s words, that the algorithms for controlling power management can be implemented in software and/or hardware. Broadcom asks the court to ignore the “and” part of that and read it as only disclosing one or the other.
Seth Lloyd (22:42): But in fact, Fallah teaches you can use both software and hardware, the same kind of hybrid approach that Broadcom now claims to have invented. Broadcom also argues that, well, even if those have hardware approaches, they’re at the system level and not the component level, but Broadcom’s arguments are contradicted by the references themselves, and at least the board reasonably rejected that based on the references. If you look at Alben, the Alben reference, and this is at Appendix 854 at column seven, it talks about how—well previously talked about how software can intervene to override what the hardware is doing, the overriding that Broadcom wants you to focus on. And it explains that the intervening happens not just for the entire system at column seven, but at the subsystem level. And when you look at Alben, at Appendix 850, those subsystems are down at the device level.
Seth Lloyd (23:41): It’s the individual component. So Alben is not talking as Broadcom suggests about simply some kind of broad software control. And Broadcom says, well, Broadcom claims that the board found against that, but that was not the board’s finding. The only difference that the board found between Alben and the prior or, and the claims, is in how the elements were arranged. The board believed that Alben taught only software overriding that was indirectly done, and the board read the claims to require directly overriding. But even if there is that difference between Alben, that’s the kind of difference, that obviousness, that generally looks at. That’s KSR, where you have the elements that are known in the prior art, and the only question is whether a person of skill in the art would’ve thought to arrange them in the manner claimed. ’That’s ordinary obviousness. This is not a here/where or a DSS type case where you have some kind of missing limitation.
Seth Lloyd (24:36): The board found all limitations known in the prior art. In the following reference, there was already some discussion with the court about this, but I just wanted to respond to this idea that Fallah is only talking about software and hardware hybrids for system level and not component level. At Appendix 776 to Appendix 777, Fallah in fact talks about how the DPM, that's the acronym they use for dynamic power management—the DPM algorithms operate for systems and their constituent components. It’s talking about using power management at the system and the component level. And then it goes on to talk about how those algorithms can be implemented in both software and/or hardware. And it gives the same exact benefits that Broadcom claims were what drove its invention. It identifies that hardware is more efficient, but software can be more flexible.
Seth Lloyd (25:34): And then, and this of course is all in the introductory section when there are only two sections in this entire chapter. One is system level, which Broadcom says is the only thing you should focus on from Fallah. But one of them is clock gating, the exact thing that Broadcom claims to have invented here. And at Appendix 786 to 787, when it talks about clock gating, it refers to that as dynamic power minimization. The same type of dynamic power control that Fallah already tells you can be implemented in both hardware and software, not just one or the other—a hybrid approach. Broadcom also complains that somehow the board required these very detailed modifications with the, I think Broadcom’s counsel’s had one, two, and three modifications, but the board’s finding was in fact that there was only one change required that said Appendix 27, where the only change was to have the CPU in Alben directly control clock gates, instead of indirectly doing that.
Seth Lloyd (26:38): And the board rejected at Appendix 31 to 32 Broadcom’s argument that there would be some kind of complicated change. And then again at Appendix 35, that this would somehow require completely changing everything about Alben. So Broadcom’s arguments on motivation to combine rely on this court finding that there were gaps in the prior art that the board rejected and then somehow filling those gaps. And because Broadcom is wrong about there being any gaps to begin with, its motivation to combine arguments fails. On reasonable expectations of success, Broadcom’s arguments fail for at least one of two reasons. First, the petition and our expert testimony, in fact, did address reasonable expectation of success, and the board itself, at Appendix 28, quoted the portions of our petition that addressed this, where we explained that hybrid approaches to clocking were known in the art and can be used for clock gating.
Seth Lloyd (27:36): And we supported that with expert testimony, which the board also cited, so Broadcom’s wrong on the facts. But then of course, this issue Broadcom clearly waived across four filings over more than 100 pages, Broadcom never once said anything about reasonable expectation of success. So that issue is waived. On Broadcom’s appeal. I think we had the discussion already about the experts, and that the law is clear that just because an expert gives a conclusion, is not enough to be substantial evidence when that conclusion is not based on anything in the record, and in fact, contradicts the plain record evidence. And for that reason, we’d ask that you affirm on Broadcom’s appeal and on Renesas’s appeal, you reverse as to those claims. I’m happy to answer any other questions the court has.
Judge Lourie (28:24): Not hearing any. I thank you very much, Mr. Lloyd. Mr. Johnson has a couple of minutes for rebuttal and the cross-appeal.
Brian Johnson (28:35): Yes. Thank you, Your Honor. Council cited KSR, pointing out that as long as there are known prior art elements, that case so supports combining, but that’s not our case because both software controlled clock gating and the overriding implementation are not disclosed anywhere. And I understand that counsel for Renesas has an argument that software controlled clock gating is disclosed in Fallah, which we disagree with and explain that in our brief, but where he seems to have no argument is the overriding limitation. The board specifically looked at whether Alben disclosed overwriting and found that it did not. And that’s the specific implementation that Broadcom claimed, and I’ve seen no evidence and no argument and no finding that the overriding limitation was disclosed anywhere in Fallah. It’s’’ not there. Renesas’s counsel also points out that Alben is a hybrid approach. So is Fallah. The problem is that it cuts against their argument. Because it’s already a hybrid approach, one of skill would not have been motivated to modify Alben. Alben already recognized the importance of software and power management, and that’s why they conduct it in a way to maximize those efficiencies. It is system level, and that’s precisely what the board found. The control with respect to the clock gating, the board found correctly, was undisputedly hardware control, and Alben is already a fine-tuned system. And as we explain in our brief, it touts the advantages of software in the way it implements them. So this cuts against their motivation argument. And then counsel for Renesas suggests that we say that we invented clock gating. Not so, it’s a particular control within clock gating and that’s important. And he identifies Fallah in pages 786 to 787 as discussing clock gating, and that’s important because that entire discussion does not mention software.
Brian Johnson (30:25): The word software does not exist, and there’s no discussion of DPM algorithms that is the form of control that was discussed in the introduction. And Renesas would like this court to infer that the introduction was referring to this section 13.3, telling that there’s no discussion of software whatsoever. And so this too cuts against obviousness. And finally, with the reasonable expectation of success, the testimony that counsel refers to is just general testimony about use of certain things in the prior art is not testimony about reasonable expectation of success and Broadcom did not waive it because the kinetic case which we cite in our brief was rejected. Precisely this argument is an affirmative requirement that the board identify and support reasonable expectation of success. They did not because they had no expert testimony to rely on. Thank you, Your Honors.
Judge Lourie (31:19): Thank you to both counsels. We will take the case under submission.