[00:00:52] Speaker 04: Okay, the next case is number 15-14-18, Rudolph Technologies Incorporated against Cam Tech Limited. [00:01:00] Speaker 04: Mr. McDonald, are you ready? [00:01:08] Speaker 01: Thank you, Your Honor. [00:01:08] Speaker 01: May it please the Court? [00:01:10] Speaker 01: We are here as the patent owner challenging a conclusion of the Board of Obviousness of Certain Claims of the 528 Patent. [00:01:17] Speaker 01: based on its finding that one sentence in the prior element reference is ambiguous. [00:01:23] Speaker 01: The 528 claims briefly compare a substrate such as a dye on a semiconductor substrate to a model trained with images of multiple known good quality substrates. [00:01:36] Speaker 01: That works well to find acceptable variations from what might be perfect and still be able to identify a dye as good and usable. [00:01:45] Speaker 01: So it's a reality-based model [00:01:47] Speaker 01: because it's using actual images, and it's also a diverse model, because it's using multiple real images to see this variation. [00:01:55] Speaker 01: Aliment teaches three other approaches to this inspection process. [00:02:00] Speaker 01: The best summary of that is a sentence in column two at lines 50 to 54 of Aliment, which you'll find at joint appendix 371. [00:02:09] Speaker 01: That sentence talks about a like-to-like or die-to-die comparison, which is not a model and not diverse. [00:02:18] Speaker 01: It also talks about a die, a comparison of a die to a pattern on the same die. [00:02:23] Speaker 01: So it's not even a different die here. [00:02:25] Speaker 01: It's called the repetitive pattern. [00:02:26] Speaker 01: Again, that one's not diverse and it's not a model. [00:02:29] Speaker 01: Then the third one is called a die to database comparison, which is a model. [00:02:35] Speaker 01: But it's not diverse and it's not based on real images. [00:02:38] Speaker 01: As the board noted, that's a synthetic model. [00:02:40] Speaker 01: It's mathematically based, for example, maybe on the specification or the equivalent of a blueprint of what it's supposed to look like. [00:02:46] Speaker 01: But that's perhaps too ideal for the reality of taking the photograph, so it doesn't have that diversity, and it's not reality-based. [00:02:54] Speaker 01: And that description is at the board's decision of Joint Appendix 13, where they call that synthetic. [00:03:00] Speaker 01: Now, the board was correct in finding that Balaman does not necessarily describe any other embodiment than those three. [00:03:07] Speaker 01: And of course, that's where they rejected the anticipation findings by the examiner, and we don't appeal that, of course. [00:03:13] Speaker 01: But their error was to take that same, [00:03:15] Speaker 01: sentence in element that was used for that erroneous anticipation defense and find that it was ambiguous because it might describe yet another approach to this comparison process. [00:03:27] Speaker 01: And for that reason, that's one sentence found at column eight of element in the board's opinion found to render the claims here obvious. [00:03:35] Speaker 01: Now that conclusion of obviousness was infected by two legal errors. [00:03:40] Speaker 01: One, [00:03:40] Speaker 01: The board failed to read Alamit as a whole. [00:03:43] Speaker 01: It took that one sentence out of context. [00:03:46] Speaker 01: And we see cases like Kotzeb or Smith Klein that we cited that say you can't read a sentence or two in the abstract. [00:03:54] Speaker 01: You have to look at the entirety of the reference when evaluating obviousness. [00:03:59] Speaker 01: And in fact, the board agreed that that one sentence could mean that die-to-die comparison that's much different from what we've claimed. [00:04:06] Speaker 01: It's not diverse and it's not a model. [00:04:08] Speaker 01: And it should mean only that. [00:04:09] Speaker 01: It was not ambiguous. [00:04:10] Speaker 01: That's the first error. [00:04:12] Speaker 01: The second error is that ambiguous alternative meaning that the court, the board found referred to what it called a collective or plural to one comparison. [00:04:21] Speaker 01: And even if we assume erroneously that that alternative meaning would apply, it is a leap to go from there to find our claims here obvious because we didn't claim a collective comparison per se. [00:04:32] Speaker 01: We're claiming training or modeling these substrates using multiple good quality [00:04:38] Speaker 01: substrates or wafers. [00:04:40] Speaker 01: I'm paraphrasing because we have a number of independent claims here, but that's the core concept that you see in all of those. [00:04:45] Speaker 01: And to go from this comparison to collective comparison all the way to there, the board makes that leap with no evidence in the record, nothing it has. [00:04:52] Speaker 01: It really just relies solely on hindsight, knowing what our claims look like, to say, oh, could you construe it that way? [00:04:58] Speaker 03: Can I interrupt you just for a minute? [00:05:01] Speaker 03: Sure. [00:05:01] Speaker 03: So you agree, though, that it is a question of substantial evidence. [00:05:05] Speaker 03: Is there substantial evidence to support [00:05:08] Speaker 03: that reliance on that one sentence to conclude that there's these possibilities, that there's two different possibilities of doing it. [00:05:16] Speaker 03: But the question is whether there's substantial evidence, whether that one sentence provides substantial evidence. [00:05:21] Speaker 01: That's one way to look at it, Sean. [00:05:22] Speaker 01: I appreciate that. [00:05:22] Speaker 03: I mean, that's the standard of review. [00:05:23] Speaker 03: I just want to make sure that, because from your brief, I wasn't sure, because it sounded as if you were suggesting it's a legal question, and I don't agree with that. [00:05:31] Speaker 03: So I just want to make sure I'm asking you about it so you can explain to me [00:05:34] Speaker 03: Sure. [00:05:34] Speaker 03: What you think it is. [00:05:35] Speaker 01: Well, I think that, you know, legal, the obviousness finding is a legal conclusion. [00:05:39] Speaker 01: But you're right, there's this issue, what's the scope of the prior art? [00:05:42] Speaker 01: That's a fact sort of finding here. [00:05:44] Speaker 01: But when the board's approach to determining the scope of the art is infected by a legal error, then I think we do cross the line to evaluating as a legal issue. [00:05:52] Speaker 01: There are the cases that say, as a matter of law, you do not evaluate the prior art a one sentence in isolation. [00:05:58] Speaker 01: You have to look at it as a whole. [00:05:59] Speaker 01: So there are legal problems that infected their analysis. [00:06:02] Speaker 01: Moreover, [00:06:03] Speaker 01: Even if we just accept, you know, the element is an undisputed fact here. [00:06:07] Speaker 01: There is no expert testimony or anything else. [00:06:09] Speaker 01: There is a record here within the four corners of that prior art. [00:06:12] Speaker 01: The board didn't rely on anything else. [00:06:13] Speaker 01: And so you have undisputed facts and my second error there I think is taking a leap then to this obviousness conclusion. [00:06:22] Speaker 01: And so I would say it is as a matter of law based on undisputed facts. [00:06:27] Speaker 01: If you actually look at what the board found factually about Alan, that ambiguous alternative meaning, my second point was even if we accept that fact finding as true, merely finding a collective comparison is insufficient legally to reach the conclusion that there's obviousness for our claims that aren't saying collective comparison. [00:06:44] Speaker 01: They're saying something more specific than that about training and modeling with multiple good quality images. [00:06:50] Speaker 01: Collective comparison doesn't get you all the way to there as a matter of law. [00:06:54] Speaker 01: Even if we accept that fact finding by the board, [00:06:56] Speaker 01: in isolation and out of context is true. [00:06:59] Speaker 01: So I think there are some legal problems here as well. [00:07:02] Speaker 01: That's my point. [00:07:04] Speaker 01: So if we go back to that first legal error, the out of context issue, there is absolutely no description and element of collective comparison or anything that would be modeling based on multiple real images. [00:07:18] Speaker 01: That sentence, the key sentence, it's found at Joint Appendix 374. [00:07:21] Speaker 01: It's column 8 of element. [00:07:25] Speaker 01: And it talks about a comparison of an inspected pattern with the light pattern of at least one other die serving as the reference pattern. [00:07:35] Speaker 01: Now that prior, if we go back to that sentence on J371 that describes that, it's a pattern on another like article. [00:07:44] Speaker 01: That's the die to die comparison. [00:07:46] Speaker 01: So it's like to like. [00:07:47] Speaker 01: That's what this means. [00:07:48] Speaker 01: And pattern to pattern. [00:07:49] Speaker 02: Isn't the phrase you just read imply that there's multiple dies? [00:07:53] Speaker 01: The at least one phrase, Your Honor, is I think what you're talking about, right? [00:07:57] Speaker 01: At least one. [00:07:57] Speaker 01: Okay. [00:07:59] Speaker 01: So I would say no to that in context, even in this sentence, because if you look at the entirety of the sentence in column eight, the first part of it begins at line 37. [00:08:08] Speaker 01: As also indicated above, during the phase one examination and also the phase two examination, the pattern, the inspected pattern is compared with the light pattern of at least one other die. [00:08:20] Speaker 01: I paraphrased the second part. [00:08:22] Speaker 01: But the first part talks about what we're doing, we're talking about examination number one and two. [00:08:26] Speaker 01: It's a two-phase processing element. [00:08:28] Speaker 01: They do kind of a rough cut to find the potential problem areas, and then they use a second phase to zero in on those problem areas. [00:08:35] Speaker 01: Well, it's talking about two examinations here. [00:08:38] Speaker 01: So within the four corners of that sentence, when it says at least one other die, it totally works with what we say it means, that this could be two separate steps, not one step using multiple [00:08:48] Speaker 01: collective comparison, but two separate examinations. [00:08:51] Speaker 01: The sentence itself says, phase one examination and phase two examination. [00:08:55] Speaker 01: So I would say even in the four corners of the sentence, you can reconcile that at least one language with multiple one-to-one comparisons. [00:09:03] Speaker 01: And that's, of course, as we were saying, consistent with the whole spec, which never talks about this collective comparison sort of approach to it. [00:09:11] Speaker 01: And there's nothing that Camp Tech has cited here either that would give you any support anywhere in the spec for [00:09:16] Speaker 01: How would you do this collective comparison? [00:09:19] Speaker 01: What is the modeling method? [00:09:20] Speaker 01: Why would you do it that way instead of some other way? [00:09:22] Speaker 01: There's just nothing in there that support that out of context approach to element. [00:09:28] Speaker 01: So that's the first problem is finding that ambiguity is erroneous because it was out of context and in fact that sentence is not ambiguous. [00:09:36] Speaker 01: It's only referring to one-to-one die-to-die comparisons. [00:09:40] Speaker 01: So the second here, if we look at the Joint Appendix and go to what the board found, this goes to your [00:09:45] Speaker 01: point, Your Honor, of what are the fact findings here. [00:09:48] Speaker 01: What the board found here is that that sentence, the comparison is either done individually on a die-to-die, die-by-die basis, i.e. [00:09:59] Speaker 01: multiple one-to-one comparisons, that's what we say is the meaning, or collectively, i.e. [00:10:04] Speaker 01: plural to one comparison, quote. [00:10:07] Speaker 01: That's all they say about what that alternative meaning is. [00:10:10] Speaker 01: Like I just said, I think that's erroneous. [00:10:12] Speaker 01: But if we accept that as factually, we have to live with that, [00:10:15] Speaker 01: That still doesn't get you all the way to finding our claims here obvious. [00:10:19] Speaker 01: None of our claims talk about a collective comparison. [00:10:21] Speaker 01: They talk about, like in claim one, visually inputting a plurality of known good quality images during training, or claim nine, optical viewing of multiple good quality substrates, things like that. [00:10:33] Speaker 01: So you have a leap here. [00:10:34] Speaker 01: And the board cites no evidence or analysis to fill that gap. [00:10:38] Speaker 01: It used a kind of an ipsy dixit hindsight-infected conclusion that CODSAB, [00:10:43] Speaker 01: teaches is insufficient. [00:10:45] Speaker 01: The court needs to show a specific understanding or principle within the knowledge of the person of skill that would make that leap appropriate. [00:10:54] Speaker 01: And in fact what you have here is the opposite because in Alamit it actually teaches at Joint Appendix 371 in its background the opposite of doing a comparison that's the type we talk about that's based on multiple real images where you go pixel by pixel and create these models. [00:11:11] Speaker 01: There at Column 1, Lines 28 to 34 [00:11:14] Speaker 01: It talks about there are some systems out there for inspecting wafers generally based on analyzing high-resolution two-dimensional images of a patterned wafer using a charge-coupled device, CCD, that's the sort of thing disclosed in our patent, on a pixel-by-pixel basis, which is what we talk about. [00:11:30] Speaker 01: However, the extremely large number of pixels involved makes such systems extremely slow. [00:11:36] Speaker 01: They're saying we're not doing it that way. [00:11:37] Speaker 01: That's actually the opposite here in the Four Corners of Aliment. [00:11:41] Speaker 01: So there's actually teaching away from [00:11:43] Speaker 01: going to our system. [00:11:45] Speaker 01: Now in response to that, Campset proffers a number of rationales that the board did not rely on to try to help fill that gap. [00:11:53] Speaker 01: I think really the headline there is that that shows that they can't really support the board's decision within its four corners on its face. [00:12:01] Speaker 01: One thing they do there is they go to this other prior art that was not cited by the board or the examiner or at least relied upon, Chow and Kobayashi. [00:12:09] Speaker 01: Kobayashi was cited by them in the request for re-examination for other claims. [00:12:15] Speaker 01: And those claims did not make it into the re-examination. [00:12:18] Speaker 01: They cited Kobayashi not for this modeling feature. [00:12:21] Speaker 01: They cited it for teaching solder bumps. [00:12:25] Speaker 01: That's at page 897, I believe. [00:12:29] Speaker 03: Could I direct you, I just want to ask you some questions before you run out of time, about the scope of claims 1 and 9. [00:12:37] Speaker 03: you know in its opinion the board noted that there was a difference in the scope there and specifically pointed out that with respect to claim one it was broader and when it talked about how you were developing the model it didn't require that the model be developed from more than one good quality substrate well your honor on claim one it's [00:13:04] Speaker 01: I think it was a little unclear what the board was seeing, because I see on page 9 of their decision, a joint appendix 10, they say, with respect to claim 1, even assuming without deciding that the recited model that the microprocessor develops is trained using plural known good quality substrates, they're reaching their findings. [00:13:22] Speaker 01: So did they decide that it doesn't need to or not? [00:13:26] Speaker 01: I think that sense there actually suggests that they were saying, well, we're not deciding one way or the other. [00:13:30] Speaker 01: We're assuming even if there's multiple ones, [00:13:32] Speaker 01: that these claims are obvious. [00:13:35] Speaker 03: Right. [00:13:36] Speaker 03: But can you explain to me, if I look at the claims and I think that the interpretation here, I look at the claims and I see a difference between claims one and nine. [00:13:44] Speaker 03: And so I wanted to give you an opportunity to explain why claim one isn't broader than claim nine. [00:13:50] Speaker 01: OK. [00:13:50] Speaker 01: So claim one, I would say you have the key there is the model aspect of it and the training. [00:13:56] Speaker 01: Are those two things tied together? [00:13:58] Speaker 01: I think that's the point, right? [00:13:59] Speaker 01: And when you see that section of the 528 pad, it totally intertwines training with the modeling. [00:14:06] Speaker 01: You can't have the one without the other. [00:14:08] Speaker 01: That's in figure three. [00:14:09] Speaker 01: It shows the step-by-step. [00:14:10] Speaker 01: That's the training section, as you go back to figure two, where it talks about we train and we inspect and so on. [00:14:16] Speaker 01: It has a subflow chart in perspective. [00:14:18] Speaker 01: What is training? [00:14:19] Speaker 01: And it includes developing a model of a die. [00:14:22] Speaker 01: These two things really are inextricably intertwined to each other. [00:14:25] Speaker 01: And even the board itself notes when it's talking about the Means Plus Function Clause, [00:14:29] Speaker 01: that if you have a section in the spec at column 13 where it says the model, I think the word is definitional, a model has at least two die, has to have at least two things to create that statistical model involving a mean in the standard deviation. [00:14:44] Speaker 01: So I think to the extent the board made a mistake there, like I said, I think it's kind of unclear where they were going. [00:14:49] Speaker 01: But if you're just asking me, what does that mean? [00:14:51] Speaker 01: I think it's clear from the district court construction, but also just reading the 528 pattern, whether you have that or not. [00:14:57] Speaker 01: Training and modeling go together. [00:14:58] Speaker 01: And that claim one clearly requires those things to involve multiple dyes or substrates. [00:15:04] Speaker 01: You have that up in the visual inspection device phrase where you have a plurality of known good quality substrates having a user defined level of quality during training. [00:15:16] Speaker 01: So that's claim one. [00:15:18] Speaker 01: What was your other question or was there another one? [00:15:20] Speaker 01: That was it. [00:15:22] Speaker 04: Okay. [00:15:22] Speaker 04: Okay, let's save you rebuttal time and let's hear from the other side. [00:15:25] Speaker 01: That's fine, Your Honor. [00:15:26] Speaker 01: Thank you. [00:15:36] Speaker 04: Mr. Bielsker. [00:15:37] Speaker 00: Thank you, Your Honor. [00:15:41] Speaker 00: And may I please the court? [00:15:42] Speaker 00: There has been, just though you're correct, the standard is a substantial evidence standard. [00:15:49] Speaker 00: And under the case law, that is something that reasonable minds might find acceptable, evidence that reasonable minds might find acceptable. [00:15:57] Speaker 00: And as the applied case of this court stated, just because people may come, there may be opposite inconsistent positions, does not mean the evidence is [00:16:06] Speaker 00: not substantial evidence, and that is applied at 1294. [00:16:11] Speaker 00: And under the Supreme Court's decision in Consolo, it's something less than the weight of the evidence. [00:16:18] Speaker 00: That's at Consolo 620. [00:16:21] Speaker 00: Now, in the Houston case, what this Court says is even when the Board is not very explicit about its reasoning, if there is a reasonable path to discerning what that reasoning is, you can affirm that decision. [00:16:36] Speaker 00: And this situation is very similar to what the Houston court found and is explained at footnote nine of its decision. [00:16:43] Speaker 00: Now, Mr. McDonald has stated that there is no evidence and the board relied solely on one sentence. [00:16:49] Speaker 00: Well, it did focus on one sentence, but there was a significant amount of evidence in the record. [00:16:56] Speaker 00: And what the board said at page nine of its decision, which is at 23-22 of the joint appendix, [00:17:05] Speaker 00: It said, the weight of the evidence on the record supports what the examiner found. [00:17:10] Speaker 00: Now, there actually were two bases that the examiner used to make the rejection. [00:17:15] Speaker 00: One was very explicit and actually has gotten very little attention here, has gotten no attention from Rudolph. [00:17:22] Speaker 00: The second basis was not as explicit and has gotten some attention here in the briefing, and that is those two pieces of prior art, the Chao reference and the Kobayashi reference. [00:17:32] Speaker 00: Let me just go to the first point, which is the [00:17:35] Speaker 00: the basis that the examiner found very explicitly. [00:17:38] Speaker 00: And it relies on something which has been overlooked here. [00:17:41] Speaker 00: And that is what the claims actually say is using multiple substrates, if you interpret all the claims the same. [00:17:50] Speaker 00: The model is made using multiple substrates. [00:17:53] Speaker 00: What is a substrate? [00:17:54] Speaker 00: Well, the claims actually tell you. [00:17:55] Speaker 00: They say a substrate can be a wafer. [00:17:58] Speaker 00: It can be a dye. [00:18:00] Speaker 00: It can be part of a dye. [00:18:02] Speaker 00: It can be a plurality of dyes, or it can be parts of plurality of dyes. [00:18:08] Speaker 00: And actually, what happens here, even in the 528 patent, the way that the examination is done, it is done on a pixel by pixel analysis. [00:18:18] Speaker 00: So when you look at the 528 patent, and it's column 16, line 41 to 45, what you see, the way the model is actually created, [00:18:28] Speaker 00: you take one pixel that you're inspecting, you've got your model is actually created by creating averages of the intensity of each pixel that is going to be compared to the pixel under inspection. [00:18:42] Speaker 00: Now when you look at the Allumont patent, that's actually exactly what's done. [00:18:46] Speaker 00: And that was cited to the examiner, and the examiner actually cited that in his rejections and in fact cut and pasted portions of the Allumont patent that describe that. [00:18:59] Speaker 00: At the joint appendix at page 12, I believe it starts at 1225 to 1233, some of which is actually not in the joint appendix. [00:19:08] Speaker 00: I think we only included 1231 to 133. [00:19:12] Speaker 00: But you can actually see at the examiner's rejection, which starts at page 1294, he says, Camtech's arguments have been considered, and I am rejecting the claims because what happens in Alumat is pixel information [00:19:29] Speaker 00: in the reference is compared to... Sorry, did you say that's at 1225? [00:19:33] Speaker 00: 1225 is Camtech's... is Camtech's response to the patent owner saying the claims are not obvious and don't teach the model. [00:19:46] Speaker 00: That starts at 1225, goes to 1233. [00:19:49] Speaker 00: The portion that I'm quoting right now is not actually in the Joint Appendix, but it's part of that. [00:19:58] Speaker 00: But you can look at what the examiner said to actually glean what CAMTEC argued to the examiner. [00:20:03] Speaker 00: So if you see the joint appendix at 1294, what the examiner actually says is that when you look at the pixel information in Alumat, there's two pieces of information. [00:20:15] Speaker 00: There's the intensity. [00:20:16] Speaker 00: So I take the inspected, I take the reference, I look at the intensities. [00:20:20] Speaker 00: Do I stop there? [00:20:21] Speaker 00: No. [00:20:22] Speaker 00: Because to make a determination as to whether that intensity information is really valuable, [00:20:28] Speaker 00: I have to figure out what that pixel's actually representing. [00:20:32] Speaker 00: Is it representing the center of the die? [00:20:35] Speaker 00: Is it representing an edge of the die? [00:20:37] Speaker 00: And to make that determination, the pixel type information, what Alumar teaches is you look at all these pixels around it. [00:20:44] Speaker 00: It's actually a three by three neighborhood. [00:20:47] Speaker 00: And you look at that three by three neighborhood, you take averages, you use histograms, and based on that, you make a determination of what the pixel type is. [00:20:57] Speaker 00: And what the examiner said is, hey, that's actually a collective comparison. [00:21:02] Speaker 00: Because what the claims say is a substrate can be a portion of a die. [00:21:06] Speaker 00: In other words, a pixel, a picture element. [00:21:10] Speaker 00: So the examiner cited that to the patent owner. [00:21:12] Speaker 00: And the patent owner comes back and says, ah, that can't be a model because that information is not stored. [00:21:18] Speaker 00: And that is at the joint appendix 1313. [00:21:23] Speaker 00: Examiner comes back and the examiner says, no way, you're wrong. [00:21:26] Speaker 00: That information is stored. [00:21:28] Speaker 00: It's stored in memory 75. [00:21:32] Speaker 00: And that's at 1365 of the joint appendix. [00:21:35] Speaker 00: And the examiner, in his actual rejection on the model, so at 1375, he makes that argument. [00:21:43] Speaker 00: And then he cuts and pastes the actual, you know, he does it like a cut and paste and puts the actual text from the patent in there, an image. [00:21:51] Speaker 00: of the Alumont patent in there, it cites to that, and then in the actual rejection, he cites to column 12, lines 28 to 30 of Alumont, which is, column 12, 28 to 30 of Alumont is actually the joint appendix at 376. [00:22:07] Speaker 00: So there was this very specific reference to that, and CAMTEC, in its briefing before the PTAB, actually referred to that. [00:22:17] Speaker 00: It got referred to at the hearing, [00:22:20] Speaker 00: And it's actually referred to, again, before this court, the blue brief at page six. [00:22:25] Speaker 00: So there actually is information on the collective comparison, because what you're actually comparing is pixels. [00:22:33] Speaker 00: It's one pixel to another pixel, and the reference pixel on Alumont is not just based on that one pixel that corresponds. [00:22:43] Speaker 00: That one pixel, you figure out the type, and it's based on this whole neighborhood. [00:22:47] Speaker 00: If you were to argue that, [00:22:49] Speaker 00: In fact, it can't just be a pixel, contrary to what the claims say that it has to be a die. [00:22:53] Speaker 00: Allumont also addresses that, because it says when you look at the image that you take, actually is not necessarily just one die. [00:23:01] Speaker 00: It can overlap multiple dies. [00:23:03] Speaker 00: And you can see that in figures 9 to 11 of the Allumont patent. [00:23:07] Speaker 00: You actually can see the figure. [00:23:08] Speaker 00: There's a circle. [00:23:10] Speaker 00: And you can see that circle overlapping multiple dies. [00:23:13] Speaker 00: So in fact, when it does the 3 by 3 analysis, [00:23:17] Speaker 00: Some of those pixels may be on one die, some of the pixels may be on another die. [00:23:21] Speaker 00: It's fully consistent. [00:23:23] Speaker 00: So when the board said that the weight of the evidence supports what the examiner found, he certainly knew that the exam, or the board knew that the examiner was relying on this pixel data that was created from multiple pixels, not just one. [00:23:40] Speaker 00: And just to show you that that actually is exactly like what the 528 patent does, [00:23:46] Speaker 00: Column 16, 41 to 45 of the 528 patented issue says, and I think this is the direct quote, but I might have a few words wrong, each pixel on unknown quality waivers is viewed whereby the model is used to determine if the pixel and or group of pixels are deemed good quality or questionable. [00:24:09] Speaker 00: And you'll see there's an example in that 528 patent where it says, well, [00:24:13] Speaker 00: On one pixel, the intensity value is 98. [00:24:16] Speaker 00: On one of them, it's 100. [00:24:18] Speaker 00: On the other one, it's 105. [00:24:19] Speaker 00: The average of those three is 102 plus or minus something. [00:24:24] Speaker 00: And that's actually exactly what the Allumont patent does. [00:24:28] Speaker 00: In fact, it mentions averages of the pixel data at column 14, line 10. [00:24:35] Speaker 00: And that's in the joint appendix at 377. [00:24:38] Speaker 00: Now, let me get to the second point of evidence. [00:24:41] Speaker 00: which supports the conclusion that the board and the examiner came to, which is when you read that sentence about at least one die, is there any support in the world for a person of ordinary skill interpreting that to mean more than one die is used in creating the reference? [00:24:58] Speaker 00: And the answer is absolutely. [00:25:00] Speaker 00: As this court knows, when making an obviousness determination, what you must do is look at the scope and content of the prior art. [00:25:08] Speaker 00: And when you look at the scope and the content of the prior art, [00:25:11] Speaker 00: It was very well known that you want to use multiple inputs to create your model. [00:25:18] Speaker 00: And where was that known from? [00:25:20] Speaker 00: Well, CAMTEC told the examiner that it was known from the Chao and Kobayashi references. [00:25:27] Speaker 03: Where is that in the record? [00:25:29] Speaker 00: That is in the record at Joint Appendix 1233. [00:25:32] Speaker 00: 1231 and 1233. [00:25:39] Speaker 00: And the reason that CAMTEC made that statement was to rebut an argument that the patent owner made, saying, hey, creating this model by using multiple inputs is not known. [00:25:52] Speaker 00: CAMTEC responded as it is allowed to do under 37 CFR 1.948 A sub 2. [00:26:01] Speaker 00: And as the case which Rudolph actually cited, the Belkin case at 1383 says it can do. [00:26:09] Speaker 00: So Camtech responded to an argument that the patent owner made saying, hey, it's not obvious to make a model based on multiple inputs. [00:26:17] Speaker 00: Camtech responded, of course it is. [00:26:19] Speaker 03: Can you point to me exactly where that is at 12, 1331, and 1333? [00:26:23] Speaker 03: Exactly what language you're relying on? [00:26:25] Speaker 00: So that would be, I believe there's two paragraphs. [00:26:28] Speaker 00: And one paragraph refers to the Chao patent, and one paragraph refers to the Kobayashi patent. [00:26:35] Speaker 00: At least that's what my note said. [00:26:40] Speaker 03: Oh, I was looking at the wrong page. [00:26:45] Speaker 03: I was looking at 1331 instead of 1231. [00:26:54] Speaker 00: So there are two paragraphs. [00:26:56] Speaker 00: One refers to Chao. [00:26:57] Speaker 00: One refers to Kobayashi. [00:26:58] Speaker 00: And it's explained in there. [00:27:00] Speaker 00: The same argument was actually made in the PTAB brief, which is Joint Appendix 1878. [00:27:07] Speaker 00: And it was also discussed briefly at the hearing [00:27:10] Speaker 00: Rudolph has said, you can't consider that evidence. [00:27:16] Speaker 00: And they actually cited the Belkin case and they say, hey, this is a totally new ground. [00:27:20] Speaker 00: We're not talking about a combination. [00:27:22] Speaker 00: What we're talking about is what this court and what the PTAB was required to do, which was understand what the scoping content of the prior ARC is so that when they read that sentence at least one die, they are informed of, well, how would a person of ordinary skill in the ARC understand that? [00:27:39] Speaker 00: Given the fact that [00:27:40] Speaker 00: It was well known in the prior art that you want to use multiple inputs to create your model. [00:27:46] Speaker 00: A person of ordinary skill in the art reading that at least one die would say, of course, yeah, use multiple inputs, multiple die to make the one to model comparison where your model is based on multiple things. [00:27:59] Speaker 00: And if you look at the CHEL patent and you, it's CHEL column four, 23 to 27, I don't have a joint appendix site because [00:28:10] Speaker 00: Rudolph objected to that being included in the joint appendix, but the clerk of this court said we file the motion for the court to take judicial notice of it. [00:28:19] Speaker 00: It was granted. [00:28:21] Speaker 00: But if you look at the child patent at column four, line 23 to 27, it says the reason that you want to take multiple inputs to make your model is to make sure that you're not going to come up with all these false defects. [00:28:35] Speaker 00: It really helps in accounting for that. [00:28:39] Speaker 00: That language, the reason why it's done, the way that that problem is solved, is identical to what the 528 patent does. [00:28:49] Speaker 00: I would ask the court to, I don't have enough time to actually go through the language verbatim, but column 4, 23 to 27 of CHAL, compared to column 13, lines 10 to 15 of the 528 patent, you're going to find the exact same reasons for making the model based on multiple things. [00:29:08] Speaker 00: Now, given the state of the prior art, the arguments that were made to the PTAB, which was in the record, it was completely reasonable for the Patent Office to come to the conclusion that based on that one sentence, the model was created from collective information. [00:29:29] Speaker 00: Let me just get briefly to one thing that was not addressed, and that's the standing or the ripeness issue of claims [00:29:37] Speaker 00: 30 to 34 and 36. [00:29:40] Speaker 00: We did cite to the ACME case the letter of the notice of appeal that the PTAB issued cited to the provisions in the CFR into the statute that says this is a new rejection. [00:29:53] Speaker 00: When you have a new rejection, you have two options. [00:29:57] Speaker 00: A new rejection is not something final that you can appeal. [00:29:59] Speaker 00: You have two options. [00:30:01] Speaker 00: One option, request rehearing. [00:30:03] Speaker 00: Option number two, request further prosecution. [00:30:07] Speaker 00: Neither one of those was done. [00:30:09] Speaker 00: Those claims are not properly before this court. [00:30:13] Speaker 00: And last, since I have 27 seconds left, Your Honor, you asked about the difference in scope between claims one, claims nine, and I would also add claims 14. [00:30:24] Speaker 00: You are correct. [00:30:25] Speaker 00: Claim one and claim 14 do not require multiple inputs to create the model. [00:30:31] Speaker 00: Claim nine actually does. [00:30:34] Speaker 00: Thank you so much. [00:30:35] Speaker 00: If you have any questions, I will end it there. [00:30:38] Speaker 04: Okay. [00:30:39] Speaker 04: Good. [00:30:39] Speaker 04: Thank you, Mr. Bielski. [00:30:47] Speaker 04: Mr. MacDonald. [00:30:48] Speaker 01: Thank you. [00:30:48] Speaker 01: Do I have 43 seconds for him? [00:30:51] Speaker 01: Okay, four minutes. [00:30:53] Speaker 01: Good. [00:30:53] Speaker 01: Thank you. [00:30:55] Speaker 01: The points they make here I think are not justification for affirming this decision because the board, there's just no analysis in the board's decision that's consistent with what he was just talking about. [00:31:05] Speaker 01: There's no reliance on Chao or Kobayashi in there. [00:31:09] Speaker 01: Now Kobayashi is a printed circuit board inspection system. [00:31:12] Speaker 01: That's like what your old AM radios would have solder bumps in them from a board that has little transistors that you could actually see back in the 60s or 70s. [00:31:21] Speaker 01: It's not looking at semiconductor substrates. [00:31:24] Speaker 01: was Camtech tried to use Chao to invalidate the sister patent. [00:31:28] Speaker 01: It's also in litigation we mentioned in here that's in that decision reported at 655F3rd. [00:31:34] Speaker 01: And it didn't find these claims that also involved modeling of multiple substrates or wafers, sorts of things. [00:31:41] Speaker 01: The patent was found valid over Chao. [00:31:43] Speaker 01: There's differences. [00:31:44] Speaker 01: Chao does not do a pixel to pixel comparison. [00:31:48] Speaker 01: It looks for dramatic changes on a substrate, an image of it, [00:31:54] Speaker 01: where there's a dramatic change, but that would indicate some sort of an impurity on it. [00:31:57] Speaker 01: That's what our expert explained when you go into the context. [00:32:00] Speaker 01: I can't imagine how this court's going to sort out what Chow means or what Kobayashi means here. [00:32:05] Speaker 01: The point is that the board did not rely on that. [00:32:08] Speaker 01: Even though it had that as part of its record, it deemed it insufficient to support its findings. [00:32:12] Speaker 02: Irregardless of the prosecution history, just looking at Alamod and back to that term, [00:32:22] Speaker 02: at least one other die. [00:32:23] Speaker 02: Why isn't that sufficient to inform one skill in the art that the use of multiple dies is an option? [00:32:32] Speaker 01: Well, it informs them that use of multiple dies in multiple comparisons is an option, because the first part of the sentence says we have this in both the phase one and the phase two examinations. [00:32:42] Speaker 02: But to do something... You would admit that at least the use of the word at least one other die implies or one [00:32:51] Speaker 02: skilled in the art would look at that and say, well, multiple dyes, I can rely or use multiple dyes. [00:32:58] Speaker 01: It would include one dye, or it could include more than that. [00:33:02] Speaker 01: Or more than one. [00:33:03] Speaker 01: That's multiple one to one, either single one to one. [00:33:06] Speaker 02: Why wouldn't that be substantial evidence? [00:33:11] Speaker 01: Well, because it's inconsistent with the rest of the specification and that sentence. [00:33:16] Speaker 01: Because there's a lot more to comparing with multiple dye, if you're not going to talk about [00:33:20] Speaker 01: multiple one-to-one. [00:33:22] Speaker 01: If I were going to compare one to a lot, how do you do that? [00:33:24] Speaker 02: I mean, our bad times about that. [00:33:26] Speaker 02: The board did go through the prosecution history and looked at all the other references and concluded that, at least on the basis of this one sentence, that multiple dies was an option to one skilled in the art. [00:33:44] Speaker 01: It found that, as it put it, a collective comparison. [00:33:48] Speaker 01: That's the words that the board found, right? [00:33:50] Speaker 01: And so my second point there is that I think that's inconsistent with the spec, which nowhere talks about such a thing. [00:33:56] Speaker 01: But that still doesn't get you all the way to modeling and training with images of multiple known good quality wafers. [00:34:04] Speaker 01: That's a lot that the board had packed into just that phrase, collective comparison, without any citation to the record or anyplace else that would support that legal conclusion of obviousness. [00:34:16] Speaker 01: And it is very parallel, I think, to that COTSAB case in that sense. [00:34:19] Speaker 03: The other things I just wanted to mention... Can I ask you about another argument that CAMTAC's counsel made? [00:34:27] Speaker 03: He pointed out that the claim, when it talks about a substrate, defines that as being just even part of a wafer. [00:34:34] Speaker 03: And so, therefore, I guess the idea is that you don't even have to compare multiple wafers. [00:34:39] Speaker 03: You can compare multiple parts of a wafer. [00:34:41] Speaker 03: How do you respond to that? [00:34:43] Speaker 01: Well, we had that first case we had at 655 f-thirds. [00:34:48] Speaker 01: I talked about different parts of things. [00:34:50] Speaker 01: And that was where it was remanded in part because the court said, well, when you have that, it has to be discrete parts. [00:34:56] Speaker 01: And so if you have just parts of the same dye that's physically connected to each other, that's not discreetly separated. [00:35:02] Speaker 01: So that was the whole issue in the first case where we had to redo it. [00:35:06] Speaker 01: And that's working its way up again here after we got another finding of infringement, even with the new [00:35:11] Speaker 01: claim construction there, but that's where it's being taken out of context here and just not consistent. [00:35:16] Speaker 01: I mean, the board didn't rely on that at all. [00:35:18] Speaker 01: What the board found at its decision at Joint Appendix 13 and 14 is that Element does not necessarily describe that modeling based on one-to-one, you know, or modeling with two or more images, a comparison of one to a model based on one. [00:35:32] Speaker 01: It said it did not necessarily find that. [00:35:35] Speaker 01: What counsel is trying to tell you here is that the board really did find that Element has that when they were talking about the pixels and things. [00:35:41] Speaker 01: But the board's decision says they didn't find that element necessarily describes that. [00:35:45] Speaker 01: The pixels there are very different. [00:35:47] Speaker 01: If you look, for example, at figure seven of element, you'll see there's an array of light collectors, about eight of them, in that embodiment that are looking at reflected light coming off. [00:35:56] Speaker 01: What the patent talks about with these pixels, for example, at column nine of element, at 45 to 50 at page 375 of the appendix, is that that's from sampling these reflected patterns. [00:36:09] Speaker 01: Element, as we pointed out in the reply brief, specifically says we're trying to minimize actually collecting data of the image. [00:36:14] Speaker 01: We just want to see the reflection. [00:36:16] Speaker 01: We don't want to see the actual image, that pattern on the dye itself, or that would be actually a distraction from looking for some impurities on the surface. [00:36:23] Speaker 01: It's a very different thing. [00:36:24] Speaker 01: And again, you know, this is an issue that the board recognized that this did not support. [00:36:29] Speaker 01: That's the dye to dye comparison, that the board said that's not enough. [00:36:33] Speaker 01: to render these claims obvious. [00:36:35] Speaker 01: There has to be something else. [00:36:36] Speaker 02: Could you comment on reject the claims 30 through 34 and 36? [00:36:40] Speaker 01: Yes. [00:36:40] Speaker 01: You know, that I think had a unique circumstance here because the board rejected that subset of the claims that I was anticipating. [00:36:48] Speaker 01: The board reversed that. [00:36:49] Speaker 01: And then they substituted in this 103 ground, which in a sense is then new. [00:36:53] Speaker 01: But in another sense, it's the exact same ground they relied on to reject all the other claims, 1 and 9 and 14 and so on, for obviousness. [00:37:00] Speaker 01: They in fact say, in their opinion as they're analyzing those, [00:37:03] Speaker 01: As we discussed above, element suggests this collective comparison and therefore renders this multi, the model based on multiple images obvious. [00:37:13] Speaker 01: As we discussed above, it's the same one we talked about for 1, 9, and 14. [00:37:17] Speaker 01: And so it's kind of in this unique circumstance here. [00:37:20] Speaker 01: And when you look at the context of what was the option here, what were we going to say to the board or to the examiner? [00:37:26] Speaker 01: You know, they heard our argument. [00:37:27] Speaker 01: We think that that sentence and element is insufficient. [00:37:31] Speaker 01: They dealt with that. [00:37:32] Speaker 01: And they said that they were wrong about that. [00:37:34] Speaker 01: But there was nothing to it. [00:37:35] Speaker 01: There's no amendments to the claims. [00:37:36] Speaker 03: There was no new evidence. [00:37:37] Speaker 03: Isn't the decision quite clear about what you're supposed to do? [00:37:41] Speaker 03: Our decision includes a new ground of rejection. [00:37:44] Speaker 03: And then they tell you specifically what you must exercise one of two options. [00:37:49] Speaker 01: You know, certainly looking back at that, Your Honor, I understand you took it at one sense. [00:37:53] Speaker 01: I think when I say it's a unique circumstance, reading the decision as a whole, when there were the exact same grounds relied upon, I mean, the reality is we looked at that opinion and didn't appreciate [00:38:02] Speaker 02: Well, it says a new ground of rejection. [00:38:05] Speaker 02: I mean, it seems to me that that leaves room at the basis. [00:38:08] Speaker 02: The rationale for the rejection could have been the same, but now it's a new ground. [00:38:14] Speaker 01: Right. [00:38:14] Speaker 01: But when you look at what's the point of doing these other steps before taking an appeal is to present new evidence or to modify their claims. [00:38:20] Speaker 01: And there was this nothing else that we were going to do. [00:38:22] Speaker 01: The board understood our argument. [00:38:24] Speaker 01: They rejected it. [00:38:26] Speaker 01: But just to repeat the same thing again, [00:38:28] Speaker 01: didn't make sense to us. [00:38:31] Speaker 01: That's what happened. [00:38:34] Speaker 04: Thank you.