`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`––––––––––––––––––
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`––––––––––––––––––
`
`HTC CORPORATION, HTC AMERICA, INC., and LG ELECTRONICS, INC.,
`Petitioners,
`
`v.
`
`PARTHENON UNIFIED MEMORY ARCHITECTURE LLC,
`Patent Owner.
`
`––––––––––––––––––
`
`Case No. IPR2015-01502
`U.S. Patent No. 7,542,045
`
`––––––––––––––––––
`
`PETITIONERS’ REPLY
`
`
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Table of Contents
`
`I.
`
`II.
`
`INTRODUCTION ........................................................................................... 1
`
`BOWES AND MPEG DISCLOSE THE CLAIMED VIDEO DECODER .... 1
`
`A.
`
`B.
`
`C.
`
`Bowes Discloses a Video Decoder. ....................................................... 1
`
`The Bowes/MPEG Combination Uses Shared Memory. ...................... 7
`
`Bowes Can Retrieve The Data It Stores. ............................................. 11
`
`III. BOWES AND MPEG DISCLOSE THE CLAIMED ARBITER .................. 12
`
`IV. REASONS TO COMBINE BOWES AND MPEG ....................................... 15
`
`V.
`
`CERTAIN DEPENDENT CLAIMS ............................................................. 21
`
`VI. CONCLUSION .............................................................................................. 21
`
`
`
`
`
`
`
`i
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Exhibit List
`
`1005
`
`1006
`
`Exhibit # Reference Name
`1001
`U.S. Patent No. 7,542,045 (“the ’045 patent”)
`1002
`File History for U.S. Patent No. 7,542,045
`1003
`U. S. Patent No. 5,546,547 (“Bowes”)
`1004
`ISO/IEC 11172-2:1993: Information technology—Coding of moving
`pictures and associated audio for digital storage media at up to about
`1,5 Mbit/s—Part 2: Video,” (1st ed. August 1, 1993) (“MPEG
`Standard”)
`S. Rathnam et al., “An Architectural Overview of the Programmable
`Multimedia Processor, TM-1,” IEEE Proceedings of COMPCON ’96,
`pp. 319-326 (1996) (“Rathnam”)
`R.J. Gove, “The MVP: A Highly-Integrated Video Compression
`Chip,” Proceedings of the IEEE Data Compression Conference (DCC
`‘94), pp. 215-224 (March 29-31, 1994) (“Gove”)
`U.S. Patent No. 5,774,676 (“Stearns”)
`Declaration of Dr. Santhana Chari (“Chari Decl.”)
`International Organization for Standardization, Website of ISO/IEC
`11172-2
`WorldCat Entry for Rathnam
`Patent Owner Claim Construction Brief in Case No. 2:14-cv-690,
`April 7, 2015
`Patent Owner Claim Construction Brief in Case No. 2:14-cv-902,
`June 18, 2015
`District Court’s Preliminary Constructions in Case No. 2:14-cv-690
`Brad Hansen, The Dictionary of Multimedia, 1997
`U.S. Patent No. 8,681,164
`Excerpt of File History for U.S. Patent No. 8,681,164
`RESERVED
`RESERVED
`Shanley, et al., “PCI System Architecture,” Addison-Wesley
`ii
`
`1007
`1008
`1009
`
`1010
`1011
`
`1012
`
`1013
`1014
`1015
`1016
`1017
`1018
`1019
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`1020
`
`1021
`1022
`1023
`1024
`
`1025
`
`Exhibit # Reference Name
`Publishing Company, 1995 (3rd ed.) (“Shanley”)
`Stone, H., “Microcomputer Interfacing,” Addison-Wesley Publishing
`Company, 1982
`RESERVED
`RESERVED
`U.S. Patent No. 5,797,028 (“Gulick 028”)
`“Accelerated Graphics Port Interface Specification,” Intel
`Corporation, July 31, 1996 (Revision 1.0) (“AGP”)
`VESA Unified Memory Architecture Hardware Specifications
`Proposal,” Version 1.0p (“VUMA”)
`U.S. Patent No. 5,712,664 (“Reddy”)
`U.S. Patent No. 5,442,748 (“Chang”)
`U.S. Patent No. 5,432,900 (“Rhodes”)
`Curriculum Vitae of Dr. Harold Stone
`Expert Declaration of Dr. Harold Stone (“Stone ’045 Decl.”)
`Settlement Agreement
`Reply Declaration of Dr. Harold Stone (“Stone Reply Decl.”)
`
`1026
`1027
`1028
`1029
`1030
`1031
`1032
`[NEW]
`1033
`[NEW]
`1034
`[NEW]
`1035
`[NEW]
`1036
`[NEW]
`
`1037
`[NEW]
`1038
`
`U.S. Patent No. 5,682,484 (“Lambrecht ’484”)
`
`U.S. Patent No. 5,375,068 (“Palmer”)
`
`U.S. Patent No. 5,557,538 (“Retter”)
`
`K. Konstantinides and V. Bhaskaran, “Recent Developements in the
`Design ofImage and Video Processing ICs,” Chapter 2 - VLSI Signal
`Processing Technololgy, Kluwer Academic Press, 1994
`Deposition Transcript of Dr. Mitchell A. Thornton, Ph.D. (June 17,
`2016)
`Information technology – Generic Coding of Moving Pictures and
`
`iii
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Exhibit # Reference Name
`[NEW]
`Associated Audio Information: Systems, ISO/IEC 13818-1:1996,
`ITU-T Rec. H.222.0 (1996) (“MPEG-2 Standard”)
`Srinath V. Ramaswamy and Gerald D. Miller, “Efficient
`Implementation of the Two Dimensional Discrete Cosine Transform
`for Image Coding applications on the DSP96002 Processor,” Proc. of
`the Midwest Conf. on Circuits and Systems, (IEEE 1993)
`U.S. Patent No. 6,081,750 (“Hoffberg”)
`
`1039
`[NEW]
`
`1040
`[NEW]
`1041
`1042
`1043
`[NEW]
`
`RESERVED
`RESERVED
`Curriculum Vitae of Dr. Harold Stone (Revised)
`
`iv
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`I.
`
`INTRODUCTION
`
`The Petition demonstrated it would have been obvious to combine the video
`
`decoding techniques of the MPEG-1 Standard with the system of Bowes and that
`
`such a combined system satisfied the claims, including the requirement that a
`
`decoder receive image data from the main memory. Pet. at 32-56; Ex. 1030 at
`
`¶¶174- 243. As Dr. Stone explained, “[h]aving implemented MPEG, during
`
`decoding in such a system, Bowes·DSP 20 would block-read from the main
`
`memory subsystem 14 data corresponding to at least one previously decoded image
`
`and to a current image to be decoded, in accordance with MPEG Standard.” Ex.
`
`1030 at ¶195. Patent Owner’s formal response presents nothing to overcome this
`
`showing. The claims at issue should therefore be held unpatentable.
`
`II. BOWES AND MPEG DISCLOSE THE CLAIMED VIDEO
`DECODER
`
`A. Bowes Discloses a Video Decoder.
`
`Patent Owner asserts that the DSP 20 disclosed in Bowes is not a “video
`
`decoder,” Resp. at 9-17, that Bowes “does not state that the DSP is suitable for
`
`video compression and decompression,” id. at 10, and that “a POSA would
`
`recognize that a DSP used for image processing is not suitable for video
`
`compression and decompression,” id. at 11.
`
`However, the Board’s interpretation of the claimed “decoder” is “hardware
`
`and/or software that translates data streams into video or audio information.”
`
`1
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Paper 14 at 8-9. “Suitability” for decompression is not part of that interpretation,
`
`and neither Patent Owner nor its expert dispute that the prior art included “off the
`
`shelf” DSPs capable of video compression and decompression pursuant to the
`
`MPEG Standard or that a skilled artisan could have implemented the
`
`Bowes/MPEG combination using such prior art DSPs. Ex. 1032 (Reply Dec. of
`
`Dr. Stone) at ¶¶8-10; see, e.g., Ex. 1006 at Fig. 1; Ex. 2008 at 2; Ex. 1039 at 25.
`
`Indeed, at least one commercially available prior art DSP specifically
`
`designed for decompressing MPEG video data was the Texas Instruments MVP.
`
`See Robert J. Gove, “The MVP: A Highly-Integrated Video Compression Chip,”
`
`DCC ’94, Data Compression Conference, Mar. 29-31, 1994, pp. 215-224 (Ex.
`
`1006); Ex. 1032 at ¶10. Thus, the skilled artisan implementing the combination of
`
`Bowes and MPEG could have simply purchased a suitable DSP “off the shelf,” and
`
`would have known to do so. Ex. 1032 at ¶10; see e.g., U.S. Patent No. 6,081,750
`
`(Ex. 1040) at 66:41-49 (filed in June 1995 and noting that a board containing the
`
`MVP chip “is available from General Imaging Corp., Billerica Mass….[and] …
`
`also available from Wintriss Engineering Corp., San Diego, Calif.”).
`
`Bowes, moreover, is directed to a system for carrying out various types of
`
`real-time digital signal processing, including “compressed audio (both high fidelity
`
`audio and speed), high resolution still images, [and] video ….” Ex. 1003 at 1:35-
`
`37. And Bowes states that such technologies “will allow for collaboration at a
`
`2
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`distance such as by video conferencing,” id. at 1:39-41, which a person of skill
`
`would understand to require the translation, including decompression, of data
`
`streams into video and audio, Ex. 1032 at ¶¶5-7. Bowes therefore does disclose a
`
`DSP that is a video decoder capable of compression / decompressions, and satisfies
`
`the Board’s interpretation.
`
`Bowes also explains that “[e]ach of these aspects of real-time information
`
`processing may require dedicated processors designed for their implementation.
`
`However, it is becoming more and more common to use programmable digital
`
`signal processors (DSP) available on the market today.” Ex. 1003 at 1:42-43.
`
`Bowes then explains the DSP 20 used with his invention “may be an off-the-shelf
`
`DSP such as the AT&T® DSP3210.” Id. at 2:21-22 (emphasis added). Thus, the
`
`DSP of Bowes is not limited to an audio or image decoder, but is an “off the shelf”
`
`DSP capable of decoding various types of data, including audio and video data
`
`used for video conferencing, and intended to be used in a technological context
`
`requiring the decompression of video data. Ex. 1032 at ¶¶4-7. Such a device
`
`clearly satisfies the Board’s interpretation of “decoder.”
`
`Patent Owner seeks to avoid these facts by focusing instead on the specific
`
`chip Bowes discloses as used in his “preferred embodiment implementation,” the
`
`AT&T DSP3210. See Ex. 1003 at 6:28-30; Resp. at 11-18. But the obviousness
`
`combination advanced in the Petition and on which the Board instituted trial is not
`
`3
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`limited to the preferred embodiment of Bowes, but rather relied on the more
`
`general description in Bowes of a generic processor referred to as “DSP 20,” see
`
`Pet. at 38-40, a fact the Board recognized in its Institution Decision, Paper 14 at 18
`
`(noting, in response to Patent Owner’s preferred embodiment argument, that
`
`“Petitioner identifies DSP 20 as the decoder …”). Indeed, Bowes explicitly
`
`contemplates using DSPs other than DSP3210. Ex. 1003 at 6:40-44 (“[Bowes’
`
`system design] provides for flexibility as newer technology and faster DSPs are
`
`developed.”).
`
`Moreover, Patent Owner cites no evidence that the DSP3210 could not be
`
`used in the system of Bowes to decode MPEG-1 video and the undisputed evidence
`
`on this record is that it could. For example, Dr. Stone testifies that “the DSP3210
`
`is a high performance computation engine, suitable for numerically intensive
`
`algorithms like those required to decode MPEG,” that the internal circuitry of the
`
`chip “yield[s] 33.4 MFLOP performance” based on “fast conversion between
`
`floating-point and fixed-point representations” and “is therefore perfectly suitable
`
`for decoding MPEG-1 data streams.” Ex. 1032 at ¶¶18-19.
`
`Patent Owner’s expert, Dr. Thornton, never testified otherwise. Rather,
`
`while Dr. Thornton testified that he thought the DSP3210 was not “suitable” or
`
`“appropriate,” see, e.g., Ex. 2009 at ¶45, he pointedly refused to testify that it did
`
`4
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`not satisfy the Board’s interpretation of “decoder” or that it could not be used to
`
`decode MPEG video:
`
`Q. … Your position is that the DSP disclosed in Bowes is an image
`processor, not a video processor, right?
`
`A. I never said that.
`
`*
`
`
`
`*
`
`
`
`*
`
`You said: It's not appropriate for that because of its architecture.
`
`A. Right.
`
`Q. But that's different from saying: No, it could not be a video MPEG
`decoder. Right? That is a different statement?
`
`A. Yes, that's a different statement.
`
`Ex. 1037 at 66:5-12; 67:8-14; 69:2-8 (emphasis added). Thus, the undisputed
`
`record before the Board is that the DSP3210 could be used to decode MPEG-1
`
`video in the cited combination.
`
`Patent Owner does assert, based on the testimony of its expert, that the
`
`DSP3210 would not be “suitable” for MPEG decoding because it has a floating
`
`point capability and because its fixed point core would supposedly be too slow to
`
`decode MPEG data. Resp. at 12-15; Ex. 2009 at ¶53. As Petitioners’ expert
`
`explains, however, “there are several errors” in Dr. Thornton’s analysis, including
`
`(1) focusing on a supposed processing requirement for MPEG-2 even though the
`
`Board instituted trial on Bowes in view of MPEG-1 (a standard requiring a much
`
`5
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`lower maximum data rate), Ex. 1032 at ¶21(b); (2) relying on a data rate (524
`
`MIPS) inconsistent with the document from which it comes and completely
`
`missing from the citation provided to support it, id. at ¶21(a) & (c); (3)
`
`understating the capabilities of the DSP3210 by at least 100%, id. at ¶21(f); and (4)
`
`ignoring that the processing requirements for converting MPEG data to fixed-point
`
`format are “comparable” to the processing requirements for converting MPEG data
`
`to floating-point format, id. at ¶¶12-17. As Dr. Stone further explains, a skilled
`
`artisan would reasonably have expected the DSP3210 to adequately decode
`
`MPEG1 data based on Prof. Thornton’s own exhibits. Id. at ¶¶22-23.
`
`Finally, Patent Owner contends that its assertions are confirmed because
`
`Apple once produced a product that employed the DSP3210, but used some other
`
`chip for decoding video. Resp. at 15-17. However, Patent Owner cites no
`
`evidence the Apple Quandra was intended to decode MPEG video, and the
`
`document Patent Owner cites demonstrates that Apple did contemplate that the
`
`DSP3210 used in that product would decompress video. Ex. 2005 at 82; Ex. 1032
`
`at ¶¶22-23.
`
`But Patent Owner’s assertions regarding the DSP3210 are beside the point.
`
`As demonstrated above, the obviousness combination at issue is not limited to the
`
`DSP3210, and there is no question that prior art “off the shelf” devices could
`
`decode MPEG-1, such as the prior art Texas Instruments MVP chip, which was
`
`6
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`specifically designed for decoding MPEG video, Ex. 1006, and the HP chip
`
`identified by Prof. Thornton, Ex. 2008. See Ex. 1032 at ¶¶10-11.
`
`B.
`
`The Bowes/MPEG Combination Uses Shared Memory.
`
`Patent Owner next argues that even if one would have combined MPEG
`
`with the system of Bowes, a person of ordinary skill in the art would not have
`
`stored previously decoded images in the main memory of Bowes, but instead
`
`would have stored them in the 8K cache memory of the Bowes’ preferred
`
`embodiment chip, the AT&T DSP3210. Resp. at 17. Patent Owner then points out
`
`that the “8K SRAM cache [] is insufficient to store a previously decoded image
`
`frame, [so] if a POSA were to combine Bowes with the MPEG Standard, he would
`
`have used a larger dedicated memory with sufficient space to store an image frame
`
`and the DSP would retrieve the previously decoded image from this dedicated
`
`memory.” Resp. at 18 (emphasis added).
`
`But the obviousness combination on which the Board instituted trial is not
`
`limited to the preferred embodiment DSP3210 disclosed in Bowes, so Patent
`
`Owner’s arguments, limited as they are to a fictional modification of the DSP3210,
`
`do not address that combination. Patent Owner is not entitled to change the
`
`obviousness combination on which the Board instituted trial. The system of Bowes
`
`was specifically designed so that his DSP used the main memory for real time
`
`digital signal processing, such as for decoding video data. Ex. 1003 at 4:55-64;
`
`7
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`6:58-62. This was the system, combined with the MPEG standard, that was the
`
`basis for the Board’s institution of trial and against which Patent Owner was
`
`required to argue the patentability of its claims. Its arguments directed to some
`
`other system are irrelevant.
`
`In any event, a skilled artisan would not have modified Bowes as Patent
`
`Owner argues, and Patent Owner cites no evidence to the contrary. For example,
`
`Patent Owner cites the testimony of Petitioner’s expert Dr. Stone (Resp. at 18), but
`
`in the cited passages Dr. Stone opines that a skilled artisan would not use a cache
`
`memory to hold image data as Patent Owner contends, no matter how large,
`
`because such a system would “not guarantee that you could retrieve it” (Ex. 2006
`
`at 159:22-23) and that even if one did the image data “will also be backed up to
`
`main memory,” (id. at 163:22-23). Dr. Stone confirms Patent Owner’s error, and
`
`explains that “[s]uch a modification of Bowes in that situation would be entirely
`
`unreasonable and impractical for several reasons.” Ex. 1032 at ¶¶24-28.
`
`Patent Owner cites testimony from its expert Prof. Thornton to putatively
`
`support its point, but the cited passage merely parrots Patent Owner’s own
`
`response, word-for-word, and cites only the same testimony by Dr. Stone in
`
`support. Compare Ex. 2009 at ¶59 with Resp. at 27. Prof. Thornton’s ipse dixit
`
`opinion, citing only testimony that contradicts that opinion, cannot support the
`
`Board’s fact-finding and should be given little, if any, weight. Cadiocom, LLC v.
`
`8
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Robert Bosch Healthcare Sys., Inc., IPR2013-00451, Paper 65 at 28-29 (Jan. 15,
`
`2015); 37 C.F.R. § 42.65(a).
`
`And there were good reasons why a skilled artisan would not modify Bowes
`
`as Patent Owner contends. For example, the cache memory of the DSP3210 is on
`
`the same silicon as the chip’s other circuitry, Ex. 2006 (Stone Dep. Tr.) at 149:5-6;
`
`Ex. 2001 at 2-2 (Figure 2-1: “On-chip RAM used for kernel storage and
`
`program/data cache”), the cache would have to be made several orders of
`
`magnitude larger to accomplish what Patent Owner suggests, Ex. 1032 at ¶26
`
`(“The sheer complexity and expense of such a redesign, including finding
`
`sufficient space on the DSP die to hold such a massive cache, would be
`
`prohibitive. A person of ordinary skill in the art would not even consider it.”), and
`
`the whole point of the Bowes invention is to use the main memory, Ex. 1003 at
`
`4:55-64; 6:58-62; Ex. 1032 at ¶ 27 (citing Ex. 1003 at 2:52-63, 4:49-60). Patent
`
`Owner’s argument therefore rests on the implausible assertion that a skilled artisan
`
`would have redesigned the entire DSP3210 chip — without any attempt to explain
`
`what would motivate such a massive expense and effort, or how one could even
`
`find space on the chip for such additional cache memory — when the alternative
`
`would be to do exactly what Bowes already says should be done and for which he
`
`specifically designed his system: use the main memory.
`
`9
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Indeed, Patent Owner’s theory is based on the erroneous analysis of its
`
`expert, Prof. Thornton, which concludes that the Bowes DSP 20 includes its own
`
`local RAM (it doesn’t; he was looking at the wrong figure, Ex. 1032 at ¶¶33-35),
`
`and must have an unmentioned larger memory for bus synchronization purposes
`
`and into which it would have stored decoded video frames (it doesn’t, and, even if
`
`it did, it need only store a small multiple of the bus width, Ex. 1032 at ¶¶36-41).
`
`Moreover, as Dr. Stone explained, one would not use a cache memory alone
`
`— of any size — as the only storage for previously decoded MPEG images
`
`because one could not be sure the images would remain in the cache until needed,
`
`since the cache is also used for other types of data. Ex. 2006 (Stone Dep. Tr.) at
`
`158:5-159:23. As Dr. Stone explains:
`
`One could enlarge the cache in the AT&T DSP3210, but could not
`reliably store image data in that cache without also storing it in main
`memory and retrieving it from main memory when necessary. My
`deposition testimony was that if you stored the data in the DSP cache,
`you could not reliably retrieve it from the cache because it could have
`been evicted from the cache when you attempted to retrieve it at a
`later time. Accessing another copy in main memory would be
`necessary.
`
`Ex. 1032 at ¶32. Thus, the claims would still be satisfied even if the preferred
`
`embodiment DSP of Bowes were modified as Patent Owner suggests, since the
`
`10
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`ability to access image data in main memory would necessarily have been part of
`
`such a modified system.
`
`C. Bowes Can Retrieve The Data It Stores.
`
`Patent Owner also argues that the Bowes DSP can not retrieve from main
`
`memory the same data that it stores to main memory. Resp. at 22-24. This
`
`extraordinary interpretation of the Bowes DSP — that it is not capable of retrieving
`
`data that it had previously written into memory — is based on a single passage in
`
`Bowes which describes how a “block write mode” can be used to store data into
`
`main memory “so that some other parts of the computer system can utilize it.”
`
`Resp. at 23. But the cited sentence actually says that “[i]n many cases it will be
`
`necessary to push that data back out to the DRAM so that some other parts of the
`
`computer system can utilize it,” and that the block write mode could be used for
`
`that purpose. Ex. 1003 at 7:6-12 (emphasis added). Thus, Bowes does not state
`
`that his DSP cannot retrieve data it previously wrote to main memory, but instead
`
`merely states that the bock write mode can be used “in many cases” where data
`
`needs to be written to main memory so other parts of the system can use it. Id.; see
`
`Ex. 1032 at ¶¶ 43-44.
`
`Patent Owner also argues that Petitioner has not identified any portion of the
`
`MPEG Standard that discloses reading image data from memory. Resp. at 24. But
`
`the MPEG Standard discloses that previously decoded images must be used to
`
`11
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`decode at least some video images consistent with the MPEG standard, Ex. 1004 at
`
`8, 42-48, 66-67; Fig. 4, and the Petition demonstrated that Bowes discloses a DSP
`
`20 that will access main memory as needed to carry out real time signal processing,
`
`such as the processing required for video conferencing, Pet. at34-36. To a person
`
`of ordinary skill in the art, the combination therefore discloses a video decoder that
`
`writes and reads image data to and from main memory in order to decompress
`
`video images pursuant to MPEG. Pet. at 36; Ex. 1030 at ¶¶179-195; Ex. 1032 at ¶¶
`
`43-44. That is sufficient to show obviousness, as the Board recognized in its
`
`Institution Decision. Paper 14 at 23-24; see also KSR Int’l Co. v. Teleflex Inc., 550
`
`U.S. 398, 418 (2007), (“As our precedents make clear, however, the analysis need
`
`not seek out precise teachings directed to the specific subject matter of the
`
`challenged claim …”).
`
`III. BOWES AND MPEG DISCLOSE THE CLAIMED ARBITER
`
`Patent Owner next argues that “[i]ndependent claims 1 and 5 recite an
`
`arbiter that controls access to the main/system memory” and that the “memory
`
`controller and arbiter (MCA) 200” of Bowes does not control access to memory.
`
`Resp. at 24-29. According to Patent Owner, the MCA of Bowes instead controls
`
`access to the memory bus. Id.
`
`But the Petition demonstrated that the “memory controller and arbiter
`
`(MCA) 200” of Bowes controls access to the memory by controlling access to the
`
`12
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`bus, Pet. at 39-40; Ex. 1030 at ¶¶197-199; Ex. 1003 at 5:28-29, 6:45-54, 9:64-65,
`
`10:2-8, and claims 1 and 5 do not recite how access to the memory is controlled, so
`
`any technique of controlling access to the memory would satisfy that claim
`
`language. Indeed, the Board rejected Patent Owner’s argument in the Institution
`
`Decision, noting that “Patent Owner does not explain, however, why controlling
`
`access to the memory bus does not, in turn, control access to the main/system
`
`memory,” Paper 14 at 24-25, and Patent Owner provides no further explanation on
`
`this point in its formal response.
`
`Moreover, Patent Owner has never requested a narrowing interpretation of
`
`this claim language; nor could it consistent with the specification. As Dr. Stone
`
`explains, the arbiter disclosed in the specification controls access to memory by
`
`controlling access to the bus. Ex. 1032 at ¶¶46-51, 54-61. This is confirmed by,
`
`for example, claim 7 of the 368 Patent, which requires “an arbiter circuit … for
`
`controlling access to the bus,” IPR2015-01500, Ex. 1001 at 17:9-11 (emphasis
`
`added), and which the Board must assume is supported by the only arbiter
`
`disclosed in the patent. Ex. 1032 at ¶¶63-69. Thus, just as in Bowes, the way the
`
`arbiter disclosed in the specification controls access to the memory is by
`
`controlling access to the memory bus. Even under Phillips, the claims must be
`
`construed broadly enough to cover that functionality. Phillips v. AWH
`
`Corporation, 415 F.3d 1303, 1313 (Fed. Cir. 2005) (“Importantly, the person of
`
`13
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`ordinary skill in the art is deemed to read the claim term not only in the context of
`
`the particular claim in which the disputed term appears, but in the context of the
`
`entire patent, including the specification.”).
`
`Patent Owner also argues that in Bowes “a device can be allowed to access
`
`the peripheral bus without being granted access to the main memory.” Resp. at 26.
`
`But controlling access to the memory bus is still one way to control access to the
`
`memory, even if the bus may be used for non-memory accessing operations as
`
`well. Indeed, as Dr. Stone explains, the specification describes that the bus is also
`
`used for non-memory accesses, Ex. 1032 at ¶48 (citing Ex. 1001 at 12:13-14), so
`
`the claims cannot be fairly read to distinguish such a system. Vitronics Corp. v.
`
`Conceptronic, Inc., 90 F.3d 1576, 1583 (Fed. Cir. 1996) (a claim interpretation that
`
`excludes a preferred embodiment “is rarely, if ever, correct”).
`
`Patent Owner also asserts that the MCA of Bowes grants access to bus
`
`cycles, not to memory requests, Resp. at 26-27, but the claims recite neither. All
`
`that is required is an arbiter for controlling access to the memory, and whether that
`
`is accomplished by granting access to bus cycles or memory requests is beside the
`
`point.
`
`Finally, Patent Owner argues that granting access to memory is supposedly
`
`more efficient than granting access to a bus, Resp. at 27-29, but that also is
`
`irrelevant, even if one were to incorrectly accept it as a distinction from Bowes.
`
`14
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`The claims recite an arbiter for controlling access to the memory—they do not
`
`require that such controlling be carried out in any particular manner.
`
`IV. REASONS TO COMBINE BOWES AND MPEG
`
`The Petition demonstrated it would have been obvious to combine Bowes
`
`and MPEG because to do would have been the use of known techniques to achieve
`
`predictable results, and also because “an ordinary artisan would have been
`
`motivated to combine the MPEG Standard’s highly efficient compression to
`
`address Bowes’ bandwidth requirement.” Pet. at 33-34. Patent Owner does not
`
`dispute these two reasons to combine included in the Petition. See Resp. at 29-38.
`
`Rather, Patent Owner asserts that “the DSP of Bowes is not suitable for
`
`video decoding,” citing its earlier argument. Resp. at 30. However, as
`
`demonstrated above, Patent Owner’s suitability arguments are focused solely on
`
`the preferred embodiment DSP3210 of Bowes, not on the more general “DSP 20”
`
`on which trial was instituted, the prior art included other DSPs capable of decoding
`
`MPEG video, and the DSP3210 was also capable of performing that function. See
`
`supra § II.A; Ex. 1032 at ¶¶8-23.
`
`Patent Owner also argues that “[a] POSA would not deem using shared
`
`memory between a decoder and another device as being advantageous,” Resp. at
`
`31, citing its expert’s unsupported assertion relating to “power consumption and
`
`space saving benefits,” Ex. 2009 at ¶73. Patent Owner is mistaken. Whether
`
`15
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`“using shared memory between a decoder and another device [is] advantageous” is
`
`irrelevant, since Bowes already discloses that functionality. See, e.g., Ex. 1003 at
`
`2:50-64; 6:22-26. The issue here is whether there was a reason to combine Bowes
`
`and MPEG, which the Petition demonstrated and Patent Owner does not dispute,
`
`not whether there was a reason to combine Bowes, MPEG and shared memory.
`
`In any event, Bowes discloses that the advantage obtained from using shared
`
`memory is avoiding the cost of adding “an expensive block of static random access
`
`memory (SRAM)” which “greatly reduces the cost of computer systems.” Ex.
`
`1003 at 4:55-64; see also Ex. 1032 at ¶¶72-74. Neither Patent Owner nor its expert
`
`dispute this advantage to the use of shared memory.
`
`Patent Owner also asserts, block citing seven full pages of testimony, that
`
`“Dr. Stone testified, at the time of the filing of the ’045 Patent, a POSA would not
`
`have concluded that using a shared memory for video decoding is advantageous as
`
`compared to using a dedicated memory.” Resp. at 31. The cited Stone testimony
`
`says nothing of the sort. Rather, Dr. Stone explained that a skilled artisan would
`
`consider many factors when deciding to use a shared memory and that counsel’s
`
`incomplete hypotheticals did not provide enough information to give definitive
`
`answers to those questions. See Ex. 2006, 134:23-141:22; see also Ex. 1032 at
`
`¶¶31-32. The testimony is irrelevant, in any event, since the question is whether to
`
`16
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`combine Bowes and MPEG, not whether to combine Bowes and a shared memory,
`
`which Bowes indisputably already includes.
`
`Patent Owner next asserts that the Bowes arbitration scheme is incompatible
`
`with MPEG “because it only provides for an inflexible statically fixed priority
`
`scheme” and MPEG decoding “requires variable amounts of required decoding
`
`computations and memory usage.” Resp. at 32-35. Neither Patent Owner nor its
`
`expert, however, state that the system of Bowes was incapable of decoding MPEG
`
`bit streams, see, e.g., Resp. at 35 (arguing only that the Bowes arbitration scheme
`
`would “likely” prevent the DSP from achieving sufficient throughput), and never
`
`explain what they mean by “incompatible.”
`
`Moreover, as Dr. Stone explains, the analysis of Prof. Thornton on which
`
`this argument is based is saturated with errors. For example, while Prof. Thornton
`
`opines that the Bowes arbitration scheme “only provides for an inflexible statically
`
`fixed priority scheme,” Ex. 2009 at ¶76, that is contrary to the disclosure of Bowes.
`
`Ex. 1032 at ¶¶75-79. Dr. Stone explained, after analyzing the various states of
`
`Bowes arbitration scheme as depicted in Figure 3, that Prof. Thornton had mis-read
`
`the state diagram as requiring each state to be entered (i.e., the supposedly
`
`“inflexible statically fixed priority scheme”), when in fact the diagram shows that
`
`states may be skipped depending on whether other system components are
`
`requesting the bus (i.e., a flexible and variable priority scheme). Ex. 1032 at ¶¶77-
`
`17
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`78. As Dr. Stone concludes, “Prof. Thornton’s misinterpretation of the arbitration
`
`state diagram to contain an ‘inflexible arbitration cycle’ is the basis for his
`
`testimony …. Consequently, none of the conclusions reached in those paragraphs
`
`is based on a correct interpretation of Bowes.” Ex. 1032 at ¶79.
`
`Patent Owner also points to the “watchdog timer” of Bowes, asserting that
`
`“[a]bsent the availability of a dedicated local memory in Bowes” the watchdog
`
`timer “could render” the combined Bowes/MPEG system “nonviable.” Resp. at
`
`36-37 (emphasis added).
`
`However, the Bowes preferred embodiment DSP does have a “dedicated
`
`local memory” (i.e., the 8k cache, Ex. 1003 at 2:24), and neither the Petition nor
`
`the Board’s Institution Decision indicated that the combination on which trial was
`
`instituted precluded use of such a memory along with a shared main memory from
`
`which image data would be read. The argument that the system might not work in
`
`the absence of such a local memory is therefore irrelevant.
`
`Moreover, the Bowes’ arbitration scheme can decode images in real time,
`
`even with the presence of a watchdog timer. Ex. 1032 at ¶¶80-82. As Dr. Stone
`
`explains, Patent Owner and its expert have again misread Bowes as requiring that
`
`an entire image must be read from main memory before the watchdog timer runs
`
`out, while actually the system could make multiple reads of data from the memory,
`
`each within the time period permitted by the timer:
`
`18
`
`
`
`IPR2015-01502
`
`
`
`Petitioners’ Reply
`
`Prof. Thornton also opines that an entire previously decoded image
`has to be retrieved from memory within one wa