United States Court of Appeals, Federal Circuit.
*1291 *1292 *1293 Seth P. Waxman, Wilmer Cutler Pickering Hale and Dorr, LLP, of Washington, DC, argued for plaintiff-appellee. With him on the brief were Edward C. DuMont, Lauren Fletcher, and Daniel S. Volchok. Of counsel on the brief were Morgan Chu, Laura W. Brill, Andrei Iancu, and Alexander C.D. Giza, Irell & Manella, LLP, of Los Angeles, CA; and Herbert F. Schwartz, Ropes & Gray, LLP, New York, NY. Of counsel were Christine W.S. Byrd, Perry M. Goldberg, and Brian E. Jones.
Donald R. Dunner, Finnegan, Henderson, Farabow, Garrett & Dunner, *1294 L.L.P, of Washington, DC, argued for defendants-appellants. With him on the brief were Don O. Burley, Andrew J. Vance, of Washington, DC; and Erik R. Puknys, of Palo Alto, CA. Of counsel on the brief were Harold McElhinny, Rachel Krevans, and Karl J. Kramer, Morrison & Foerster, LLP, of San Francisco, CA.
Before BRYSON, Circuit Judge, PLAGER, Senior Circuit Judge, and KEELEY, Chief District Judge.[*]
BRYSON, Circuit Judge.
TiVo, Inc., owns a patent on technology that enables television users to "time-shift" television signals, that is, to record a television program in digital format and enable the user to replay, pause, fast forward, or reverse while the program is playing on the user's television set. The technology enables time shifting both for previously recorded programs and for programs that are currently being recorded. In 2004, TiVo sued the five appellants, collectively referred to as EchoStar, in the United States District Court for the Eastern District of Texas, alleging that they had infringed various claims of the patent, U.S. Patent No. 6,233,389 ("the '389 patent").
The claims of the '389 patent that were asserted at trial included so-called "hardware" claims (claims 1 and 32) and "software" claims (claims 31 and 61). Those four claims are the only ones at issue in this appeal. The accused devices are two types of EchoStar digital video recorders ("DVRs"), which the parties refer to as the "50X" DVRs and the "Broadcom" DVRs. At the conclusion of the trial, the jury found that the 50X DVRs literally infringed the asserted hardware and software claims. The jury found that the. Broadcom DVRs literally infringed the asserted hardware claims and infringed the asserted software claims under the doctrine of equivalents. The jury awarded TiVo a total of $73,991,964 in damages ($32,663,906 in lost profits and $41,328,058 in reasonable royalties). The district court entered judgment on the verdict and issued a permanent injunction against EchoStar. On EchoStar's motion, this court granted a stay of the injunction pending appeal.
The accused DVRs and the DVRs described in the patent both receive television signals and store data relating to the transmissions on a hard disk in Motion Picture Expert Group ("MPEG") format. An MPEG stream typically has interleaved audio and video components. Both the accused devices and the devices described in the patent can store the MPEG data and subsequently transform it for replay. The storage of the television programs as MPEG data enables users to control the playback of programs, including those that are currently being broadcast, with commands such as pause, fast forward, and reverse.
Television programs are typically transmitted in one of three ways: "over the air" from a ground-based transmitter for reception by a user's television antenna; through a cable; or by satellite. Over-theair and cable transmissions can be used to broadcast either analog or digital television, while satellite services typically broadcast digital television only. The most common format for analog television signals in the United States is the National Television Standards Committee ("NTSC") standard. The corresponding European broadcast standard for analog television *1295 signals is Phase Alternating Line ("PAL"). Digital television standards include the Digital Satellite. System ("DSS"), Digital Broadcast Services ("DBS"), and Advanced Television Standards Committee ("ATSC").
Both analog and digitally encoded television programs that are transmitted over the air are transmitted using high frequency analog carrier waves that are capable of traveling long distances. The Federal Communications Commission has partitioned the analog broadcast signal range into six-megahertz ("MHz") bands for such transmissions. A single six-MHz band can carry a single television channel in analog form or several channels in digital form by interleaving (multiplexing) the signals corresponding to the various channels in a digital format such as MPEG.
The two "hardware claims" at issue in this case relate to the process and apparatus used to effect time shifting according to the invention. The two "software claims" relate to the software process and apparatus used in the invention.
Claim 1 of the '389 patent, the first of the two asserted hardware claims, provides as follows:
A process for the simultaneous storage and play back of multimedia data, comprising the steps of:
[1] accepting television (TV) broadcast signals, wherein said TV signals are based on a multitude of standards, including, but not limited to, National Television Standards Committee (NTSC) broadcast, PAL broadcast, satellite transmission, DSS, DBS, or ATSC;
[2] tuning said TV signals to a specific program;
[3] providing at least one Input Section, wherein said Input Section converts said specific program to an [sic] Moving Pictures Experts Group (MPEG) formatted stream for internal transfer and manipulation;
[4] providing a Media Switch, wherein.said Media Switch parses said MPEG stream, said MPEG stream is separated into its video and audio components;
[5] storing said video and audio components on a storage device;
[6] providing at least one Output Section, wherein said Output Section extracts said video and audio components from said storage device;
[7] wherein said Output Section assembles said video and audio components into an MPEG stream;
[8] wherein said Output Section sends said MPEG stream to a decoder;
[9] wherein said decoder converts said MPEG stream into TV output signals;
[10] wherein said decoder delivers said TV output signals to a TV receiver; and
[11] accepting control commands from a user, wherein said control commands are sent through the system and affect the flow of said MPEG stream.
Claim 32, the other asserted hardware claim, is similar, except that it claims an apparatus rather than a process.
EchoStar's accused DVRs receive analog and digital signals, but they conduct time shifting only for digital satellite signals. EchoStar argues that the district court committed legal error when it failed to construe the two hardware claims to require that the subject devices be capable of time shifting analog as well as digital signals and when the court failed to grant EchoStar's motion for judgment as a matter of law on that ground. EchoStar's DVRs do not infringe, according to EchoStar, because they can process only one type of digital television signal and are incapable of processing analog television *1296 signals at all. In making that argument, EchoStar focuses on the first three limitations of claims 1 and 32, each of which, according to EchoStar, makes clear that the hardware claims do not apply to digital-only DVRs such as its accused devices.
EchoStar begins with the first limitation of the hardware claims, which recites the step of "accepting broadcast signals . . . based on a multitude of standards." EchoStar notes that in the "accepting" limitation the reference to "a multitude of standards" is followed by the words "including but not limited to, National Television Standards Committee (NTSC) broadcast, PAL broadcast, satellite transmissions, DSS, DBS, or ATSC." Because some of those broadcast standards apply to analog signals and some to digital signals, EchoStar contends that the "accepting" limitation requires that the claimed DVR be capable of processing both types of signals. EchoStar does not make clear how many different standards it believes the claimed device must be capable of processing. In its reply brief it suggests at one point that the number must be greater than three. At a minimum, however, EchoStar seems to contend that an accused DVR cannot infringe the hardware claims if it is not capable of processing at least one type of analog signal and at least one type of digital signal.
TiVo responds that EchoStar's argument is based on an unduly restrictive interpretation of the claim language. First, TiVo argues that the hardware claims require only that the accused products "accept" television signals based upon a multitude of standards, not that they "process" all such signals. Because EchoStar's accused devices accept analog television signals and forward them to the user's television set for real-time viewing, TiVo argues that those devices infringe the "accepting" limitation even though they do not process the analog signals by converting them to MPEG streams suitable for internal storage. Second, TiVo argues that EchoStar's devices process signals that are based on numerous standards, such as the MPEG2 Transport standard, the MPEG2 video standard, and in most cases the MPEG1 audio standard. For that reason, TiVo contends, EchoStar's DVRs infringe the "accepting" limitation even if that limitation is understood to require processing of the "accepted" signals.
EchoStar replies to TiVo's characterization of the "accepting" limitation by saying, first, that the claim language must be understood to link the "acceptance" and the "processing" of signals, so that the signals that are accepted are also processed in accordance with the other limitations of the invention. In support of that argument, EchoStar points out that the "tuning" limitation, which sets forth the first step in the processing operation, states that the tuning is performed on "said TV signals," referring generically to the signals first identified in the "accepting" limitation. The patent contains no reference to signals that are simply passed through the DVR without the processing described in the rest of the claim limitations.
EchoStar replies to TiVo's second argument by contending that the reference in the claims to the "multitude of standards" as including, but not limited to, "(NTSC) broadcast, PAL broadcast, satellite transmission, DDS, DBS, or ATSC" is a reference to broadcast standards, and that the "standards" limitation cannot be satisfied by reference to data standards, such as the MPEG2 Transport, MPEG2 video, and MPEG1 audio standards cited by TiVo.
While the language of the "accepting" limitation is not very precise, we agree with the trial court that it does not *1297 require that a DVR be capable of time shifting both digital and analog signals, as EchoStar argues. Even EchoStar does not contend that the claimed DVR must be capable of processing signals transmitted under all of the listed broadcast standards. Although the claim refers to the recited television signals as being "based on a multitude of standards, including, but not limited to" the particular specified standards, the use of the word "or" in the list of standards makes it clear that the claim language does not require that the claimed device be capable of processing signals based on all of the listed standards. The specification likewise demonstrates that the claim language should not be given such a restrictive interpretation. At several points, the specification refers to a DVR that accepts television input streams "in a multitude of forms, for example, analog forms such as National Television Standards Committee (NTSC) or PAL broadcast, and digital forms such as Digital Satellite System (DSS), Digital Broadcast Services (DBS), or Advanced Television Standards Committee (ATSC)."`389 patent, col. 2, 11. 4-10 (emphasis added); see also id., col. 3, 11. 32-37.
Nor is there anything in the specification to suggest that the Claimed DVR must be capable of time shifting at least one type of analog television signal as well as digital signals. At various points, the specification describes how analog signals are processed and then describes how digital signals are processed. The specification also describes, as a "preferred embodiment" of the invention, a DVR that "takes television streams in a multitude of forms," including analog and digital and converts those signals into "MPEG streams." See id., col. 3, 11. 30-44. But nowhere does the specification indicate that in all embodiments of the invention a single DVR must be capable of processing and converting both analog and digital signals into such MPEG streams. In fact, the specification contains one embodiment that seems to contemplate a DVR that processes only digital signals. See id., col. 6, 11. 30-36. That embodiment is based on Figure 7 of the patent, a schematic depiction of the Media Switch that is at the heart of the invention. After describing how Figure 7 operates to process analog signals, the specification adds the following: "If a digital TV signal is being processed instead, the MPEG encoder is replaced with an MPEG2 Transport Demultiplexor, and the MPEG audio encoder and VBI decoder are deleted." Id., col. 6, 11. 30-33. Modified in that fashion, the Media Switch would accept only digital signals. That digital-only embodiment of the Media Switch therefore seems to contradict EchoStar's contention that the claims require that the DVR be capable of processing both analog and digital transmissions.
Even after rejecting EchoStar's main contention that the "accepting" limitation requires that the claimed DVR accept both analog and digital transmissions, the meaning of the "accepting" limitation is still not entirely clear. In particular, it is less than clear what meaning should be assigned to the term "a multitude of standards" in the "accepting" limitation. As to EchoStar's argument that the claimed DVR must be capable of processing signals based on some undetermined number of standards greater than one, it seems unlikely that a claim drafter would use a term of such biblical imprecision as "multitude" if that term were meant to have an important restrictive function in the claim. On the other hand, if the drafter intended to require only that the DVR be capable of processing signals based on "one or more" broadcast standards, including but not limited to those referred to in the limitation, it is difficult to understand why the drafter would not have used that very common (and clear) manner of expressing the idea.
*1298 In the end, we conclude that the interpretation of the "accepting" limitation that is least problematic is the one proposed by TiVothat the reference to the "multitude of standards" includes not only broadcast standards of the type set forth in the claim, but also data standards of the type identified in the written description portion of the specification. See '389 patent, col. 3, ll. 38-43. On several occasions, the written description refers to the broadcast standards as "forms," and refers to the data standards as "standards." See id., col. 2, l. 5; col. 3, ll. 35, 38. While that choice of terms may not be the product of careful and consistent usage, it at least reflects that the patent did not use the term "standard" as a term of art restricted to broadcast standards. In the absence of any more convincing interpretation of the "accepting" limitation, we adopt that interpretation. Because the evidence at the trial showed that, interpreted in that fashion, EchoStar's devices were capable of accepting and processing television signals based on numerous data standards, we hold that the jury's verdict with respect to the "accepting" limitation is supported by substantial evidence.
EchoStar next argues that because its DVRs process only digital signals and not analog signals, they do not satisfy the second limitation of the asserted hardware claims, which recites the step of "tuning said TV signals to a specific program." EchoStar explains that the six-MHz band used to transmit digital television signals carries a multiplexed signal that contains data for multiple television programs. Thus, unlike the case of analog transmissions, in which a single program is carried on a particular segment of the broadcast band, EchoStar argues that tuning to a particular six-MHz band does not constitute tuning "to a specific program." The trial court rejected that argument, holding that "one of ordinary skill in the art would understand `a specified program' to mean `a specified frequency range.'"
The language of the "tuning" limitation is not by its terms limited to locating a place on the broadcast band that carries only a single program. That is, nothing about the phrase "tuning . . . to a specific program" excludes tuning to the six-MHz portion of the band that carries the specific program of interest, even if a further process, such as demultiplexing, is required to extract the data corresponding to that program from the band to which the DVR has been tuned. If a program is contained within a channel that contains multiple, multiplexed programs, then tuning to a program that is contained within the channel is' accomplished by tuning to that channel.
Significantly, the specification of the '389 patent describes an embodiment in which the DVR "tunes . . . to a particular program" by tuning to the particular six-MHz band that contains the multiplexed digital signal that includes the program of interest. At that point tuning is complete. The device then extracts the program from the signal that was carried on the six-MHz band by demultiplexing and generating an MPEG stream. '389 patent, col. 3, ll. 46-49 ("The Input Section 101 tunes the channel to a particular program, extracts a specific MPEG program out of it, and feeds it to the rest of the system."). Thus, the specification clearly uses the term "tunes" to refer to the selection of the six-MHz portion of the band where the multiplexed signal containing the particular program (along with others) is being transmitted. We therefore hold that the district court correctly interpreted the phrase "tuning . . . to a specific program" to include tuning to a specified frequency range that contains the program in question, whether or not that program is the only *1299 one contained within a signal transmitted in that specified frequency range. With the "tuning" limitation construed in that fashion, the evidence is clearly sufficient to support the jury's conclusion that EchoStar's devices infringed that limitation.
EchoStar next contends that the term "converts," as used in the third limitation of the hardware claims, provides a further indication that the claimed DVRs must be capable of processing analog signals, rather than just digital signals as the EchoStar devices do. That limitation requires that the claimed device "convert[] said specific program to an [MPEG] formatted stream for internal transfer and manipulation." The trial court regarded that language as sufficiently clear that it was not necessary to use additional or different language to explain the phrase to the jury. EchoStar argues that the specification uses the term "converts" to refer to the process of encoding a nonMPEG analog program into MPEG format and not to refer to the process of transforming an MPEG signal into an internally storable format. EchoStar supports its argument by pointing out that the specification uses a different term, extraction, to refer to the process of transforming an MPEG signal that is carried on a broadcast transmission to an internally storable format. '389 patent, col. 2, ll. 10-14. EchoStar therefore concludes that the term "converts" does not encompass transforming MPEG data from a satellite signal that already contains MPEG-encoded data. For that reason, EchoStar argues that its 50X and Broadcom DVRs, which process only satellite transmissions that already contain MPEG-encoded data, do not satisfy the "converts" limitation and therefore are not covered by the hardware claims.
TiVo argues that EchoStar's interpretation of the "converts" limitation is at odds with portions of the specification that describe the process of extracting MPEG streams from incoming digital signals containing MPEG data, such as satellite or ATSC transmissions. See '389 patent, col. 3, ll. 46-49. The specification makes clear that if the input signal does not contain MPEG-formatted data, the data is recoded in MPEG format and stored in a digital MPEG format that is suitable for internal transfer and manipulation. If the input signal already contains MPEG-formatted data, as in the case of incoming satellite transmissions, the incoming data is still converted to the internally transferable and manipulable MPEG format. Whether or not the input stream already contains MPEG-encoded data, it is converted to a "uniform, purely digital MPEG format that is suitable for internal transfer and manipulation." The benefit of the uniformity of these streams is that the system does not need to be concerned with "how the signal was obtained." `389 patent, col. 3, ll. 46-52. Because the invention disclosed in the specification and recited in the claims. "converts" the signal to which the device has tuned into an MPEG formatted stream "for internal transfer and manipulation," we hold that the process of converting a digital signal to an MPEG stream is within the scope of the "converts" limitation.
Both the 50X and Broadcom DVRs, even if they do not need to convert signals from an analog format to MPEG, must accept satellite transmissions and convert them into a form suitable for internal transfer and manipulation. With the "converts" limitation construed to cover converting a digital signal to an MPEG stream suitable for internal transfer and manipulation, the evidence supports the jury's verdict that all of the EchoStar DVRs satisfy that limitation. We therefore reject EchoStar's contention that the "converts" limitation, the "accepting" limitation, and the "tuning" limitation all indicate *1300 that the DVRs of the invention must be capable of processing analog as well as digital signals.
EchoStar devotes substantial attention to the fourth limitation of the hardware claims, which requires that the MPEG stream be "separated into its video and audio components." The trial court regarded that claim language as self-explanatory and therefore did not provide a separate interpretation of that language in the instructions to the jury. EchoStar urges this court to construe the term to require that the interleaved MPEG stream be copied into two distinct memory buffers, one for the audio portion of the stream and another for the video portion of the stream. We agree with EchoStar that the written description portion of the specification discloses such an embodiment. See, e.g., '389 patent, col. 4, ll. 23-32. The more difficult question is whether that embodiment represents the limits of the claims' scope.
TiVo argues that physical separation of the audio and video data is not necessary, that logical separation is all that is required, and that indexing of the data achieves logical separation. Under that interpretation, a device would satisfy the "separation" limitation for literal infringement even if the audio and video streams were not copied to separate buffers but were maintained in a single interleaved buffer, as long as the separate audio and video components were indexed with pointers to the components' locations in the interleaved buffer.
The specification states that "[t]he invention parses the . . . MPEG stream and separates it into its video and audio components. It then stores the components into temporary buffers." `389 patent, col. 2, ll. 15-17. In that passage, the specification clearly refers to the separation aspect of the "invention" and not merely one embodiment of a broader invention. As this court recently held, "[w]hen a patent thus describes the features of the `present invention' as a whole, this description limits the scope of the invention." Verizon Servs. Corp. v. Vonage Holdings Corp., 503 F.3d 1295, 1308 (Fed.Cir.2007).
In the same context, the specification states that the audio and video components are stored in temporary buffers. Although TiVo does not address that statement in the specification directly, it appears to argue that the statement that the "invention . . . separates [the MPEG stream] into its video and audio components" refers to logical indexing. However, there is no textual basis from which to conclude that the reference to "storing the components in temporary buffers" encompasses logical separation or indexing, and TiVo does not provide any explanation of how the language of the specification can be understood to do so.
The specification's discussion of Figure 3 adds force to EchoStar's argument regarding the "separated" limitation. The specification explains that the video and audio segments of the incoming MPEG stream "must be separated and recombined to create separate video and audio streams or buffers. This is necessary because separate decoders are used to convert MPEG elements back into audio or video analog components." `389 patent, col. 4, ll. 25-29. Because the statement refers to separate streams and buffers, and not logical separation or indexing, the most natural interpretation of that passage is that the audio and video elements are separated into diffrent. *1301 buffers, not simply logically indexed according to the type of data carried by each element.
TiVo's effort to discount the passage of the specification discussing Figure 3 is. unconvincing. TiVo first argues that Figure 3 represents only a preferred embodiment. The detailed description of the invention makes clear that some of the depicted embodiments (such as those in Figure 1 and Figure 2) are merely preferred embodiments. The specification, however, does not characterize Figure 3 as representing only a preferred embodiment. In fact, the discussion of Figure 3 suggests the contrary, as it states that the incoming interleaved video and audio segments "must be separated and recombined to create separate video and audio streams or buffers." The specification further states that the recited separation is "necessary" because of the use of separate decoders to convert the MPEG elements back into audio and video analog components. '389 patent, col. 4, 11. 25-29. Even if the separation of video and audio components into separate streams were regarded as representing only a preferred embodiment, that passage at minimum rebuts TiVo's assertion that "physical cleaving of digital data . . . makes no sense in the context of a computer-based invention."
TiVo's second argument regarding the specification's treatment of Figure 3 is that it relates "to the output stage, when appropriate components must ultimately be placed into distinct streams or buffers for decoding." Again, there is no textual basis for that characterization. The discussion of Figure 3 is not limited to the output stage, but instead refers generally to the processing of data that is interleaved in "the incoming MPEG stream." '389 patent, col. 4, line 23. Moreover, TiVo's strained interpretation of Figure 3 is at odds with the specification's discussion of Figure 4, in which the specification describes the Media Switch at the heart of the invention as operating on separate "video buffers" and "audio buffers." Id., col. 5, 11. 8-9, 26-27.
With respect to Figure 6, which depicts the Media Switch separating data into four buffersa video buffer, an audio buffer, a private data buffer, and an event buffer-TiVo argues in its brief that the accompanying text in the specification shows that "the patent focuses on the logical segmentation and cataloguing of data into an event table to help avoid unnecessary physical data transfers." While TiVo is correct that the specification describes a process of logical cataloguing or indexing of data, Figure 6 and the portion of the specification that describes it make clear that the logical processes described by TiVo occur after the data is separated into distinct audio and video buffers. That is, after the parser in the Media Switch separates the video, audio, and private event data packets and stores them in separate buffers, the parser generates "event" data that identifies the storage location of each of the data packets, That "event data" is stored separately in an "event" buffer. '389 patent, col. 5, 11. 8-27. The program logic reads accumulated events in the event buffer and generates a sequence of "logical segments" that correspond to the parsed MPEG segments. The logical segments contain inforihation including the address and type of the corresponding MPEG segment, the length of the segment, and the time stamp for that segment. Those logical segments are then stored in a separate buffer known as the Packetized Elementary Stream ("PES") buffer. Id., col. 5, 11. 33-58.
The patent explains that using logical segments in the PES buffer means that *1302 "the data associated with the logical segments need not be present in the buffer itself." '389 patent, col. 5, line 66, through col. 6, line 1. The use of the logical segments "has the effect of gathering components of the stream, whether they be in the video, audio or private data circular buffers, into a single linear buffer of stream data on the storage medium." Id., col. 6,11.2-7. Thus, the stream data can be efficiently managed without the need for the computer's central processing unit to copy the underlying data. Id., col. 6,11. 7-15. In sum, TiVo is correct that the specification describes the process of indexing data. Both the specification and the text of claims 1 and 32 make clear, however, that indexing occurs after, and in addition to, the separation of the incoming data into distinct video and audio buffers.
Pointing to dependent claims 8 through 11, TiVo contends that those claims recite "the additional steps specifying the use of distinct video and audio buffers" and thus support its contention that the independent claims do not require separation of the audio and video components of the incoming signal. The dependent claims, however, recite in detail the processing steps that occur after the separation of the video and audio components, such as placing the video and audio components in separate circular buffers, posting events in a circular event buffer indicating the location of the video or audio component in the video or audio buffer, and generating a buffer containing logical segments that point to the buffer location where corresponding video and audio components have been placed. While those details go beyond the recitations of the independent claims, they do not obviate the requirement of the independent claims that the MPEG stream be "separated into its video and audio components." In fact, the dependent claims underscore that the initial separation is separation into distinct buffers, not merely logical indexing.
Finally, TiVo argues that Figure 7 and the discussion of Figure 7 in the specification support its construction of the "is separated" limitation. Figure 7 depicts the Media Switch and its relationship to other components of the invention. Although the specification's discussion of Figure 7 is unclear in at least one respect,[1] the specification unambiguously describes the parser in the Media Switch as detecting "the beginning of all of the important events in a video or audio stream." '389 patent, col. 6,11.39-40. That description supports EchoStar's contention that the video and audio components processed by the Media Switch are separated at the time of processing. For the foregoing reasons, we hold that the term "is separated," as used in the claims, denotes copying the audio and video components into separate buffers, not mere "logical separation" by way of indexing.
EchoStar concedes that its 50X DVRs separate the video and audio streams into distinct buffers. Therefore, the 50X DVRs *1303 satisfy the "is separated" limitation. TiVo's expert testified that the Broadcom DVRs logically separate the video and audio streams by creating a frame index table that indexes or stores pointers to the video data. TiVo does not appear to contest EchoStar's contention that the Broadcom DVRs do not separate the two streams of data into distinct buffers. Because we construe the term "is separated" to require separation into distinct buffers and not to encompass mere logical separation, as performed by the Broadcom DVRs, we agree with EchoStar that the evidence at trial does not show that the Broadcom DVRs satisfy the "is separated" limitation of the hardware claims.
EchoStar next contends that its 50X DVRs do not satisfy the seventh limitation of the hardware claims, which provides: "wherein said Output Section assembles said video and audio components into an MPEG stream." EchoStar argues that the "assembles" limitation covers only the assembly of audio and video components into a single, interleaved MPEG stream. TiVo argues that the "assembles" limitation also covers the assembly of each component, audio and video, into its own separate stream. Because EchoStar does not dispute that the Broadcom DVRs satisfy the "assembles" limitation, even under its construction, the dispute over the "assembles" limitation applies only to EchoStar's 50X DVRs.
TiVo agrees that the 50X DVRs do not assemble audio and video components into a single interleaved MPEG stream, but instead assemble the audio and video streams into two separate MPEG streams. Nonetheless, TiVo argues that the 50X DVRs satisfy the "assembles" limitation because the reference to "an MPEG stream" must be interpreted to mean "one or more MPEG streams."
As a general rule, the words "a" or "an" in a patent claim carry the meaning of "one or more." Baldwin Graphic Sys., Inc. v. Siebert, Inc., 512 F.3d 1338, 1342-43 (Fed.Cir.2008). That is particularly true when those words are used in combination with the open-ended antecedent "comprising." See, e.g., Abtox, Inc. v. Exitron Corp., 122 F.3d 1019, 1023 (Fed.Cir.1997). However, the question whether "a" or "an" is treated as singular or plural depends heavily on the context of its use. See Norian Corp. v. Stryker Corp., 432 F.3d 1356, 1359 (Fed.Cir.2005). The general rule does not apply when the context clearly evidences that the usage is limited to the singular. Baldwin Graphic, 512 F.3d at 1344.
Unlike the case in Baldwin Graphic, where the claims and the written description could be read to encompass either a singular or plural interpretation of "a" or "an," the claims and written description in this case make clear that the singular meaning applies. The pertinent claim language refers to "assembl[ing] said video and audio components into an MPEG stream," which in context clearly indicates that two separate components are assembled into a single stream, not that the video components are assembled into one stream and the audio components into a second stream. Earlier limitations in the claims refer to converting specific programs into an MPEG stream, parsing that MPEG stream and separating it "into its video and audio components." Later limitations refer to sending the reassembled MPEG stream to a decoder and converting it into TV output signals. Those steps clearly describe the separation of a single *1304 stream into two components and then reassembly of the components into a single stream for conversion into television signals. Moreover, although the open-ended term "comprising" is used to refer generally to the limitations of the hardware claims, the "assembles" limitation itself does not contain that term. Rather, the claim language simply refers to the assembly of two components into "an MPEG stream."
The corresponding explanation in the specification refers to separating the MPEG stream into video and audio components, which are then "reassembled into an MPEG stream." '389 patent, col. 2, 11. 15-29. That language describes the process of separating one stream into two sets of components and then reassembling the two sets of components into a single stream. In particular, when discussing the term "assembles," the specification states that "when the program is requested for display, the video and audio components are extracted from the storage device and reassembled into an MPEG stream which is sent to a decoder. The decoder converts the MPEG stream into TV output signals and delivers the TV output signals to a TV receiver." '389 patent, Abstract; id., col. 2, 11. 27-32. The specification also provides that the "MPEG stream has interleaved video . . . and audio . . . segments." Id., col." 4, 11. 23-26. That language, like the claim language, describes the MPEG stream that is "reassembled" as a single MPEG stream, not two different streams. The concept of reassembly indicates that the MPEG stream is restored to an earlier state, which in this case would be the state in which it was initially received by the system, as a single stream of data. Accordingly, we agree with EchoStar that the claim would be interpreted by one having ordinary skill in the art to require the reassembly of the audio and video components into a single interleaved stream.
TiVo argues that the "assembles" issue presents a question as to the sufficiency of the evidence and that TiVo's evidence of "assembly" has not been shown to be insufficient. But the evidence cited by TiVo did not establish infringement under the narrower test for satisfying the "assembles" limitation that we have adopted. Rather, TiVo simply argues that the evidence showed that EchoStar's devices assemble video and audio components into separate streams or buffers.
For example, TiVo contends that EchoStar's 50X DVRs infringe the "assembles" limitation when performing a function referred to as "trickplay," which involves fast-forwarding or rewinding a television program. The evidence at trial, however, showed that trickplay involves only video data and does not result in the assembling of video and audio components into a single stream of data. TiVo's evidence therefore does not show that EchoStar's 50X DVRs satisfy the "assembles" limitation, as properly construed.
Because we hold that the term "assembles," as used in the hardware claims, requires interleaved assembly of the audio and video components, and because the evidence at trial did not show that the 50X DVRs perform interleaved assembly, we necessarily conclude that the 50X DVRs do not literally infringe the "assembles" limitation of the hardware claims. Inasmuch as the Broadcom DVRs do not satisfy the "is separated" limitation and the 50X DVRs do not satisfy the "assembles" limitation, we must reverse the portion of the judgment upholding the jury's verdict that *1305 EchoStar's DVRs literally infringe the hardware claims.
At several points, TiVo argues. that even if this court were to overturn the jury's verdict of literal infringement, there would still be ample evidence of infringement under the doctrine of equivalents. There are two problems with upholding the judgment on the hardware claims on that basis. First, the jury was told that if it found literal infringement it should not make a determination as to whether there was infringement under the doctrine of equivalents, so there was no verdict on the issue of equivalents with regard to the hardware claims. Second, we have construed two of the claim limitations more restrictively than the trial court's instructions permitted. For that reason, even if the jury had reached a verdict with respect to the doctrine of equivalents we could not sustain that verdict merely upon finding that substantial evidence supported it. At this juncture, we could uphold the judgment on the basis of the doctrine of equivalents only if we were to conclude that no reasonable jury, given proper instructions, could reach any verdict other than to find infringement by equivalents. The parties, however, have, not briefed that issue in any detail, and we therefore do not address it. More generally, we do not decide what further proceedings, if any, are appropriate in the district court regarding the equivalents issue. Instead, we leave that issue for the district court to resolve in the event that, on remand, TiVo decides to continue to pursue the hardware claims in light of this decision.
With respect to claims 31 and 61, the so-called "software claims," EchoStar argues that its DVRs do not satisfy several of the claim limitations and that the jury's verdict of infringement must be reversed.
Claim 31 of the '389 patent is the first of the two software claims. It provides as follows:
A process for the simultaneous storage and play back of multimedia data, comprising the steps of:
[1] providing a physical data source, wherein said physical data source accepts broadcast data from an input device, parses video and audio data from said broadcast data, and temporarily stores said video and audio data;
[2] providing a source object, wherein said source object extracts video and audio data from said physical data source;
[3] providing a transform object, wherein said transform object stores and retrieves data streams onto a storage device;
[4] wherein said source object obtains a buffer from said transform object, said source object converts video data into data streams and fills said buffer with said streams;
[5] wherein said source object is automatically flow controlled by said transform object;
[6] providing a sink object, wherein said sink object obtains data stream buffers from said transform object and outputs said streams to a video and audio decoder;
[7] wherein said decoder converts said streams into display signals and sends said signals to a display;
[8] wherein said sink object is automatically flow controlled by said transform object;
*1306 [9] providing a control object, wherein said control object receives commands from a user, said commands control the flow of the broadcast data through the system; and
[10] wherein said control object sends flow command events to said source, transform, and sink objects.
Claim 61, the second of the two asserted software claims, is similar, except that it recites an apparatus rather than a process.
As a matter of claim construction, EchoStar challenges the district court's definition of the term "object." That word is used in the claim terms "source object," "transform object," "sink object," and "control object."
EchoStar's expert stated in his declaration that the term "object," as used in the computer arts, means "an item written in an object-oriented computer programming language (for example, C + +) that is an instance of a class from which it inherits properties, and that includes both data and all procedures that operate on the data." Based on that declaration, EchoStar urged the court to define the term "object" as meaning an item written with an objectoriented computer programming method such as C + + "that encapsulates data and the procedures necessary to operate on that data and can inherit properties from a class or another object."
TiVo's expert offered a broader definition for the term "object," referring to it as "a software term that describes a collection of data or operations." He explained that "objects" can contain both data and operations, but that they may contain only data or only operations. He added that the term "object" does not imply the use of an object-oriented computer language, and that objects and object-oriented techniques can be used in many programming languages. As support for its proposed definition, TiVo cited a technical dictionary, which gave "a collection of data and operations" as one of the definitions of the term "object." IEEE 100: The Authoritative Dictionary of IEEE Standard Terms 752 (7th ed.2000).
After assessing the competing expert declarations and the evidence to which the parties directed the court's attention, the district court accepted as the proper definition of "object" the definition offered by TiVo, i.e., "a collection of data and operations." The court concluded that TiVo's definition represented the plain meaning of the term "object" to one of ordinary skill in the art. The court also concluded that persons of ordinary skill in the art would "readily understand the meaning of `source object' upon a reading of the claim language and its context in the specification." Based on both intrinsic and extrinsic evidence, the court concluded that "in accordance with its ordinary meaning" the term "source object" means "a collection of data and operations that (1) extracts video and audio data from a physical data source, (2) obtains a buffer [memory where data can be temporarily stored for transfer] from a transform object, (3) converts video data into data streams, and (4) fills the buffer [memory where data can be temporarily stored for transfer] with the streams." Additionally, the court defined "transform object" to mean "a collection of data and operations that transforms the form of the data upon which it operates"; it defined "sink object" to mean "a collection of data and operations that (1) obtains data stream buffers [memory where data can be temporarily stored for transfer] from a transform object and (2) outputs the streams to *1307 a video and audio decoder"; and it defined "control object" as "a collection of data and operations that receives commands from a user that control the flow of broadcast data."
On appeal, EchoStar argues that the district court erred in failing to construe the term "object" to require the use of object-oriented software. EchoStar points out that the patent describes how three conceptual components that are featured in the software claimssource object, transform object, and sink object work in the context of software written in the C + + programming language, which employs an "object-based approach" that collects together logical operations and software elements that perform those operations.
We discern no error in the district court's claim construction. While the patent specification includes an embodiment showing the use of "a C + + class hierarchy derivation of the program logic," '389 patent, col. 8, 11. 9-10, and uses terms characteristic of object-oriented programming in connection with that example, neither the written description nor the claims anywhere state or imply that the invention must use object-oriented programming in general, or C + + in particular. Without more, the use of an example that employs object-oriented programming is not sufficient to require that the claims be limited to embodiments using C + + or a similar programming language. Moreover, while EchoStar criticizes the court for using an alternative definition in a technical dictionary as the basis for its definition of the term "object," TiVo offered evidence other than the dictionary that supported that definitionin particular, its expert's declarationand the district court concluded that persons of ordinary skill in the art would understand that term according to its ordinary meaning, which accorded with TiVo's definition.
Importantly, the term "object" was not used by itself in the claims, but rather as part of the terms "source object," "transform object," "sink object," and "control object." The district court defined each of those terms by reference to the functions performed by the collection of data and operations, and aside from its contention that the claims should be read to require object-oriented programming, EchoStar does not object to those definitions. Because the intrinsic evidence did not limit the scope of the software claims in the manner that EchoStar urges, and because the district court's construction of each of the claim terms was soundly based on the extrinsic evidence proffered by TiVo, we find no error in the court's decision not to limit the software claims to embodiments employing object-oriented programming such as C + +.[2]
We next address EchoStar's argument that the jury's finding of infringement *1308 is not supported by substantial evidence even under the district court's construction of the critical language of the software claims. EchoStar argues that in order to be a "collection" of' data and operations, all data and operations that constitute the "collection" must be grouped together within the software code or perhaps within a single file. The word "collection," however, brings with it no such requirement of colocation. TiVo's expert explained that the EchoStar software at issue constitutes "a collection of data and operations" because the relevant subroutines are part of the same program and are "able to interact and get access to the data they need to." He testified that because the EchoStar software allows the subroutines to "come together to achieve a particular function in each case, and . . . share data," they constitute "a collection of data and operation[s]." EchoStar offers nothing other than its own expert's testimony to rebut TiVo's expert on this issue. Particularly in light of the absence of any objective support for the testimony of EchoStar's expert that a "collection" of data and operations must appear in the same subroutine or subdirectory within the software code, we cannot agree with EchoStar that the jury's verdict was unsupported by substantial evidence.
EchoStar also argues that its DVRs do not satisfy the portion of the "source object" limitation that requires the source object to "extract[ ] video and audio data from said physical data source." EchoStar argues in part that TiVo and its expert, Dr. Jerry Gibson, addressed this limitation with respect to only one model of EchoStar's DVRs. While it is true that Dr. Gibson testified in detail with respect to only one type of device, there is nothing improper about an expert testifying in detail about a particular device and then stating that the same analysis applies to other allegedly infringing devices that operate similarly, without discussing each type of device in detail. Dr. Gibson testified that the other models of EchoStar's DVRs operate similarly with respect to the "source object" limitation, and EchoStar's own expert, Dr. V. Thomas Rhyne, also discussed the various models of EchoStar DVRs collectively and made no relevant distinction among them with respect to this limitation. Furthermore, Dr. Rhyne specifically stated that the 50X DVRs use the same hardware push to a temporary memory buffer as is used in the device that was analyzed by Dr. Gibson, and in its brief EchoStar points to no distinction among the DVRs that would require separate analysis. We therefore analyze all of the DVRs together with respect to the "source object" limitation.
EchoStar argues that its DVRs do not contain software that extracts data. Instead, it contends that its devices use a "hardware push" to move data from the physical data source to a temporary data buffer. EchoStar also argues that its Broadcom devices push the data to a temporary buffer and not to the Broadcom chip. Finally, EchoStar argues that, for its Broadcom DVRs, the "Ioctl" command identified by TiVo's expert cannot "extract[] video and audio data from said physical data source" because it does not *1309 extract data from any circuitry on the Broadcom chip, but instead extracts data from the temporary data buffer that is connected to the Broadcom chip.
As an initial matter, software alone cannot extract data from a physical device; it can only control hardware that extracts data. Therefore, when a device "extracts video and audio data from [a] physical data source," it is necessarily the case that certain hardware operations are performed. For that reason, the hardware/software distinction made by EchoStar is unhelpful. What matters is whether the operations performed by the interaction of software and hardware in the accused DVRs, taken as a whole, are covered by the claim term. In order to assess EchoStar's argument, we must determine whether the evidence was sufficient to show that the particular process employed in EchoStar's accused devices is covered by the "extracting" limitation. That is, we must determine whether the process of pushing data from the physical data source to a temporary data storage buffer and then extracting the data from the temporary data storage buffer, all under the general control of the DVR software, constitutes a process that is fairly described as "extract[ing] video and audio data from said physical data source," as recited in the software claims.
After consideration of the parties' arguments and the evidence to which the parties have drawn our attention, we conclude that there was sufficient evidence from which the jury could find that the "extracting" limitation was satisfied. In particular, it was reasonable for the jury to find that the two-part process of moving data from the physical data source to the source object, as practiced in the. EchoStar devices, constitutes extraction by the source object of video and audio data from the physical data source, as those terms are used in the "extracts" limitation of the software claims. In particular, the jury could have rejected the argument made by EchoStar, based on the testimony of its experts, that data is not extracted from the physical data source because it is extracted from the temporary data storage buffer that is next to, but separate from, the chip that the experts characterized as constituting the physical data source.
EchoStar points out that TiVo's expert, Dr. Gibson, did not refer to the temporary data storage buffer in the course of his general characterization of the components of the physical data source in the EchoStar DVRs. Instead, he stated that "the physical data source is on the Broadcom chip," and he identified a number of circuits on that chip on which the physical data source "relies." In that context, he did not specifically state that the temporary data storage buffer to which the data was pushed was part of the physical data source or part of the Broadcom chip. EchoStar's experts, Drs. Rhyne and Johnson, testified that the data is "pushed" out of the elements on the Broadcom chip to the temporary data storage buffer adjoining, but separate from, the Broadcom chip. Even though EchoStar does not contest that the data is extracted from the temporary data storage buffer, it contends that the temporary storage buffer is not part of the physical data source and therefore that the data is not extracted from the physical data source. Dr. Gibson also testified, however, that the Ioctl command extracts the video and audio data from the "physical data source." Under EchoStar's reasoning, his two statements, made one after the other, are in conflict. Even assuming a conflict in his testimony, however, the jury was entitled to weigh the evidence and find that Dr. Gibson's failure to specifically *1310 identify the temporary data storage buffer as part of the physical data source is outweighed by his much clearer statement that the Ioctl command extracts data from the physical data source, which necessarily includes the adjoining temporary data storage buffer.
Based on evidence that the Broadcom chip and the temporary data storage buffer operate together in the process of moving data from the physical data source, it was reasonable for the jury to find that the temporary data buffer was simply an extension of the physical data source where data was stored pending its extraction for further processing. From that evidence, the jury could permissibly find that the "Ioctl" command, the relevant portion of the source object in the EchoStar DVRs, extracts video and audio, data from the physical data source. As there is substantial evidence that the required data and operations are performed by the accused DVRs, and because we agree that the pertinent data and operations do not need to be housed within a particular file or grouping of lines of code, the EchoStar DVRs satisfy the "extracting" limitation. We therefore uphold the jury's verdict that the EchoStar DVRs infringe the software claims of the '389 patent.
EchoStar argues that TiVo made "grossly misleading, if not false, [statements]" in its argument to the jury and that those, statements were so prejudicial that they require that EchoStar be granted a new trial. EchoStar contends that TiVo's counsel told the jury that EchoStar never sought a written opinion, and that the reason it did not do so was that any independent lawyer would have said that the EchoStar devices infringed. TiVo responds that its lawyer's remarks about the failure to seek a written opinion applied only to a particular law firm, the Bozicevic firm, and that there was nothing false or misleading about the comments when viewed in that context.
In discussing EchoStar's interactions with the lawyers from the Bozicevic firm, TiVo's counsel argued to the jury that the evidence showed that for two years EchoStar had failed to give the lawyers the technical information they needed to provide a written opinion on infringement and that "if [EchoStar] gave the lawyers, the independent lawyers, the technical information, . . . they would say that there is infringement." EchoStar argues that it did not pursue a written infringement opinion from the Bozicevic firm because its own internal investigation satisfied it that there was no infringement and because it was engaged in negotiating a business deal with TiVo that made the issue of possible infringement less important. Moreover, EchoStar points out that after TiVo filed suit, EchoStar sought and obtained two formal opinions of noninfringement from another outside law firm. Nonetheless, EchoStar argues, TiVo took unfair advantage of the district court's refusal to admit the outside law firm's opinions by suggesting to the jury that EchoStar had avoided obtaining an infringement opinion, even though TiVo's counsel knew that EchoStar had sought and obtained written advice regarding infringement from another firm.
In context, we think it was clear that the statements by TiVo's counsel were directed only to the Bozicevic firm and that the district court was correct to conclude that counsel's comments did not deny EchoStar a fair trial. Counsel's comments about the failure to obtain an opinion on the issue of infringement were made with respect to *1311 the lawyers referred to immediately before that comment during counsel's argument, that is, the lawyers from the Bozicevic firm. The trial court agreed that the statements were directed to the Bozicevic law firm. Moreover, the court cautioned the jury several times that lawyers' statements are not evidence. The trial judge, who was present during the argument and could assess the accuracy and impact of the statements, was in a far better position than we are to determine how those statements would have been understood by the jury and whether they were unfairly prejudicial to EchoStar's defense. In any event, as TiVo points out, the disputed argument bore only on the issue of willfulness. Because the trial court declined to enhance damages or award attorney fees to TiVo, the dispute over willfulness is moot.
EchoStar insists that the trial court's treatment of the issue of willfulness, including the court's failure to strike counsel's argument about the Bozicevic firm, or at least to admit evidence regarding EchoStar's other steps to obtain legal opinions regarding infringement, unfairly prejudiced EchoStar's defense. So far as is revealed by a review of the cold record, however, we conclude that the transcript does not support EchoStar's claim of prejudice. The statements made by counsel did not imply that EchoStar never sought. a written opinion or that any reasonable lawyer would say that there is infringement, as EchoStar urges. In addition, and importantly, the statements related to the issue of willfulness and did not directly affect the issues of infringement and invalidity that ended up affecting the trial court's judgment. In the end, despite EchoStar's vigorous urging on this point, we are not persuaded that counsel's conduct was both improper and so prejudicial that we must override the trial judge's exercise of discretion with respect to the supervision of the trial proceedings and direct that a new trial be held.
EchoStar next argues that the district court improperly limited the testimony of its expert, Dr. Nathaniel Polish, on the issue of the asserted invalidity of the '389 patent. EchoStar wanted, Dr. Polish to be allowed to state that the infringement analysis provided by TiVo's expert Dr. Gibson, if accepted, would compel A finding of invalidity. The trial court allowed Dr. Polish to give his opinion as to validity based on the court's construction of the claim terms, but it refused to allow him to testify about how Dr. Gibson's infringement analysis would affect the issue of invalidity. The district court explained its ruling by stating that "Dr. Polish's testimony should be limited to the Court's claim construction and the prior art, not the criticism of Dr. Gibson's report or previous testimony."
We agree with TiVo that the district judge did not abuse his discretion in the way he ruled on the proper scope of Dr. Polish's testimony. An expert must "[compare] the construed claims to the prior art." Helifix, Ltd. v. Blok-Lok, Ltd., 208 F.3d 1339, 1346 (Fed.Cir.2000). Dr. Polish was allowed to testify regarding how the prior art related to the claims as construed. He was prohibited from testifying about how the prior art related to "Dr. Gibson's view of the claims." Dr. Polish was free to testify about whether the prior art practiced the claimed invention. In light of the court's construction of the asserted claims, EchoStar was not denied the benefit of Dr. Polish's testimony on the issue of invalidity. The court's *1312 ruling therefore did not deprive EchoStar of any evidence it was entitled to introduce; at the same time, the court's ruling avoided possible jury Confusion by ensuring that the invalidity inquiry focused on the relationship between the prior art and the claims, as construed by the court. We hold that the trial court's handling of Dr. Polish's testimony did not prejudice EchoStar's right to a fair trial.
In sum, because of a failure of proof of literal infringement, we reverse the judgment of infringement of the hardware claims with respect to all of the accused devices. We remand for any further proceedings that may be necessary with respect to those claims. We affirm the judgment of infringement of the software claims with respect to all of the accused devices. Because the damages calculation at trial was not predicated on the infringement of particular claims, and because we have upheld the jury's verdict that all of the accused devices infringe the software claims, we affirm the damages award entered by the district court.
The district court's injunction was stayed during the course of these proceedings. The stay that was issued pending appeal will dissolve when this appeal becomes final. At that time, the district court can make a determination as to the additional damages, if any, that TiVo has sustained while the stay of the permanent injunction has been in effect.
Each party shall bear its, own costs for this appeal.
AFFIRMED IN PART, REVERSED IN PART, and REMANDED.
[*] Honorable Irene M. Keeley, Chief Judge, United States District Court for the Northern District of West Virginia, sitting by designaion.
[1] The specification at one point refers to the Media Switch outputting "streams to the MPEG video decoder 715 and a separate audio decoder 717." '389 patent, col. 6, 11. 63-65, which suggests separate video and audio streams. Shortly thereafter, however, the specification refers to the "decoded audio output from the MPEG decoder" being "digitally mixed 718 with the separate audio signal." Id., col. 7, 11. 1-3. Neither party has offered an explanation for why there is "audio output" from what was previously described as a "video decoder" and was distinguished from a "separate audio decoder." In the absence of any explanation by the parties for these seemingly inconsistent references, we draw no inferences from that passage of the specification.
[2] As noted, the district court based its construction of the software claims on its conclusion as to what the critical claim terms would mean to a person of skill in the art. That 'conclusion in turn was largely based on the court's assessment of extrinsic evidence. Although we have characterized claim construction as a question of law even when it involves competing presentations of extrinsic evidence, Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1456 (Fed.Cir.1998) (en banc), we recognize that there is substantial force to the proposition that such a conclusion is indistinguishable in any significant respect from a conventional finding of fact, to which we typically accord deference. See. Amgen Inc, v. Hoechst Marion Roussel, Inc., 469 F.3d 1039, 1041 (Fed.Cir.2006) (Michel, C.J., dissenting from denial of rehearing en banc); id. at 1043 (Newman, J., dissenting from denial of rehearing en banc); id. at 1044 (Rader, J., dissenting from denial of rehearing en banc); id. at 1045 (Gajarsa, J., concurring in denial of rehearing en banc); id. at 1046 (Moore, J., dissenting from denial of rehearing en banc). Applying our governing non-deferential standard of review, we uphold the district court's conclusion in this case. If we were to treat that ruling as a finding of fact, we would uphold the district court's ruling a fortiori in light of the more deferential "clear error" standard applicable to factual findings.
