`
`Jentsung Lin
`In re Patent of:
`7,693,002
`U.S. Patent No.:
`April 6, 2010
`Issue Date:
`Appl. Serial No.: 11/548,132
`Filing Date:
`October 10, 2006
`Title:
`DYNAMIC WORD LINE DRIVERS AND DECODERS FOR
`MEMORY ARRAYS
`
` Attorney Docket No.:
`
`DECLARATION OF Robert W. Horst, Ph.D.
`
`I, Robert W. Horst, Ph.D., of San Jose, CA, declare that:
`
`I.
`
`Introduction.
`
`1.
`
`I summarized my relevant knowledge and experience in my first
`
`Declaration (APPLE-1003). In writing this Second Declaration, I have considered
`
`the following: my own knowledge and experience, including my work experience
`
`in the fields of memory systems and circuit design and my experience in working
`
`with others involved in those fields. In addition, I have analyzed the following
`
`publications and materials, in addition to other materials I cite herein and in my
`
`first declaration:Patent Owner’s Response to the IPR Petition
`
`•
`
`•
`
`•
`
`•
`
`•
`
`Dr. Massoud Pedram’s Declaration (Ex. 2001)
`
`The IEEE Dictionary (Ex. 2002 and APPLE-1014)
`
`Itoh (additional portions) (Ex. 2003)
`
`Horst Deposition Transcript (Ex. 2005)
`
`The Modern Dictionary of Electronics (APPLE-1015)
`
`1
`
`Exhibit 1013
`Apple v. Qualcomm
`IPR2018-01249
`
`
`
`•
`
`•
`
`U.S. Patent No. 4,922,461 (“Hayakawa”) (APPLE-1016)
`
`M. Pedram. “Design technologies for Low Power VLSI,” in Encyclo-
`
`pedia of Computer Science and Technology, Marcel Dekker, Editors:
`
`A Kent, J. G. Williams, and C. M. Hall, vol. 36, 1997, pages 73-95
`
`(APPLE-1017)
`
`•
`
`N. Mohyuddin, K. Patel, and M. Pedram. “Deterministic clock gating
`
`to eliminate wasteful activity in out-of-order superscalar processors
`
`due to wrong-path instructions,” Proc. of Int'l Conf. on Computer
`
`De-sign, Oct. 2009, pages 166-172 (APPLE-1018)
`
`•
`
`Pedram Deposition Transcript (APPLE-1019).
`
`II.
`
`A POSITA would not have limited the interpretation of “clock
`signal” to include only periodic clock signals
`
`A.
`
`A broad plain and ordinary meaning for “clock signal”
`should be used
`
`3.
`
`A POSITA would not have considered a periodic clock signal to be
`
`the broadest reasonable interpretation (BRI) of “clock signal” as used in the claims
`
`of the ’002 patent. There is no indication in the ’002 patent itself that would have
`
`led a POSITA to adopt such a restrictive interpretation because there is no explicit
`
`definition, waveform, or circuit in the ’002 Patent defining or suggesting such a
`
`narrow definition.
`
`2
`
`
`
`4.
`
`Instead of using the BRI for “clock signal,” Dr. Pedram adopts a
`
`restrictive definition inconsistent with the way “clock signal” is used in many
`
`systems, patents and textbooks. He defines “clock signal” as follows:
`
`The term “clock signal,” as recited in each of the
`independent claims of the ’002 patent, should be
`interpreted
`to mean “a periodic signal used for
`synchronization.” This interpretation is consistent with the
`plain and ordinary meaning of the term as understood by
`the POSA.
`
`Pedram Decl. at ¶53, emphasis added.
`
`it is my opinion that a POSA looking at this patent would
`understand the clock to be a periodic signal of fixed
`frequency used for synchronization purposes in the
`context of this memory system
`
`Pedram Depo., p. 78, emphasis added.
`
`5.
`
`Dr. Pedram’s definition appears to be based on his assumption that the
`
`’002 Patent is directed only to synchronous memory systems. I disagree because I
`
`find no such restriction in the specification or drawings of the ’002 Patent. Dr.
`
`Pedram’s opinion is based on an absence in the depicted embodiments of the ’002
`
`patent of any timing circuitry he states would be required in an asynchronous
`
`system. However, Dr. Pedram acknowledged during his deposition that the figures
`
`in the ’002 patent omit many necessary components of any memory system, such
`
`as column decoders (44:17-48:1), synchronous memory control logic (54:8-60:7),
`
`and clock drivers (40:10-20). Dr. Pedram’s exemplary synchronous memory
`
`3
`
`
`
`system, as depicted in Itoh, shows many circuit components that are not depicted in
`
`the figures of the ’002 patent. The figures in the ’002 patent depict only a specific
`
`portion of a memory system (e.g., a row address decoder), and the absence of any
`
`one circuit component from the figures does not indicate to a POSITA that the
`
`circuit component must be absent from the entire memory system. As seen below
`
`by the side-by-side comparison of ’002 patent’s Fig. 1 with Itoh’s Fig. 6.15, the
`
`’002 patent does not show any control logic or other circuitry associated with
`
`either synchronous or asynchronous memory systems, but only that necessary to
`
`explain the operation of its row address decoder.
`
`6.
`
`In the embodiment in Fig. 1 of the ’002 patent, the clock input could
`
`be supplied with any type of clock signal—e.g., a periodic clock signal, a non-
`
`4
`
`
`
`periodic gated clock signal, a non-periodic clock signal derived from a periodic
`
`clock (e.g., RS1 as shown in Itoh’s synchronous clock system), or another type of
`
`non-periodic clock signal—without adversely affecting its row decoding
`
`functionality. Similarly, the clock signal supplied to the Sato and Asano prior art
`
`combinations could be periodic or non-periodic while implementing the same row
`
`decoding functionality as the ’002 Patent.
`
`7.
`
`Dr. Pedram states that his definition is the broadest reasonable
`
`interpretation based on the definition of “clock signal” from an IEEE dictionary.
`
`Ex. 2002 and APPLE-1014. The portion of the IEEE dictionary cited in Dr.
`
`Pedram’s declaration defines “clock signal” as “[a] periodic signal used for
`
`synchronizing events,” but this partial quotation omits the part of the definition that
`
`says “Synonyms: clock pulse; timing pulse.” Many systems and devices have clock
`
`pulses and other timing signals that are not periodic.
`
`8.
`
`Also, Dr. Pedram’s declaration omits alternate definitions for “clock”
`
`in the same IEEE dictionary. One of the alternate definitions of “clock” is as
`
`follows:
`
`(2) A signal, the transitions of which (between the low
`and high logic level [or vice versa]) are used to indicate
`when a stored-state device, such as a flip-flop or latch,
`may perform an operation.
`
`APPLE-1014 at 4.
`
`5
`
`
`
`9.
`
`As would be well-known to a POSITA and detailed in references
`
`below, many devices include flip-flops and latches are clocked by non-periodic
`
`clock signals. Thus, this IEEE definition of “clock” is broader than the narrow
`
`definition of “clock signal” given by Dr. Pedram.
`
`10. The Modern Dictionary of Electronics also includes two broad
`
`definitions of “clock”:
`
`1. A pulse generator or signal waveform used to
`achieve synchronization of the timing of switching
`circuits and the memory in a digital computer system”
`…
`
`4. A strobe signal that activates a certain sequence of
`operations.”
`
`APPLE-1015, 3-4.
`
`
`11. These definitions similarly are not restricted to periodic signals and
`
`are broader than Dr. Pedram’s proposed definitions. If anything, a commonality of
`
`these definitions is synchronization, not periodicity.
`
`B.
`
`The ’002 Patent uses “clock” for gated clocks that are not
`periodic
`
`12. The ’002 patent itself uses “clock” and “clock signal” to refer to
`
`signals that are not periodic. In particular, the outputs of the conditional clock
`
`generator are referred to as clocks:
`
`The conditional clock generator may receive a clock input,
`such as the clock input 118 in FIG.1, and may selectively
`apply the clock signal to a selected clock, such as one of
`the clocks 124, 126, 128 and 130 in FIGS.1 and 3. The
`
`6
`
`
`
`conditional clock generator may apply a zero voltage, a
`logic low, or ground voltage signal to the non-selected
`clocks.
`
`’002 Patent at 9:53-58, emphasis added.
`
`
`13. This usage of “clock” and “clock signal” in the ’002 Patent for gated
`
`clock signals is inconsistent with a construction that requires clocks to be periodic.
`
`Non-selected clocks can be low for long, unpredictable, periods of time when the
`
`corresponding memory locations are not being accessed, hence these clocks are not
`
`periodic. Dr. Pedram apparently agrees that the gated clock signals at the outputs
`
`124, 126, 128, and 130 of the conditional clock generator would not be “clock
`
`signals” under his narrow construction of the term as being periodic. Pedram
`
`Depo., 83:9-13, 83:19-84:6-7.
`
`14. Furthermore, the outputs of Sato’s PDCR would have been considered
`
`to be “clock outputs” as used in the claims of the ’002 Patent. A POSITA would
`
`have understood the PDCR to be performing a similar gating operation with Sato’s
`
`timing signal as the ’002 Patent’s clock generator or first logic do with the ’002
`
`Patent’s clock signal. As I explained in my first declaration (¶¶60-63, 65-66, and
`
`70-71), a POSITA would have understood that Sato’s PDCR functions similar to a
`
`demuliplexer by effectively switching the timing signal between each of the four
`
`PDCR outputs ϕx0-ϕx3 according to the values of the two internal address bits ax0
`
`and ax1 and their complements. I even provided an illustration (reproduced below)
`
`7
`
`
`
`of how a POSITA would have understood the PDCR to operate in view of Sato’s
`
`explanation. Dr. Pedram did not dispute the accuracy of this opinion. The clock
`
`gating demuliplexer that I have illustrated switches one input (ϕce) between
`
`multiple outputs ϕx0-ϕx3 according to one or more control signals (e.g., internal
`
`address bits). Thus, functionally, Sato’s PDCR operates the same as the
`
`conditional clock generator and the “first logic” of the ’002 patent. Essentially,
`
`Sato’s PDCR operates similarly to Dr. Pedram’s description of clock gating with
`
`the added functionality of selecting a particular output ϕx0-ϕx3 for the gated
`
`timing signal ϕce based on the memory address inputs (ax0 and ax1) and their
`
`compliments. The functional operation of Sato’s PDCR is undisputed. APPLE-
`
`1019, 31:6-11.
`
`C. Many references use “clock” or “clock signal” to refer to
`non-periodic signals
`
`
`
`8
`
`
`
`15.
`
`In asynchronous dynamic RAMs, the row and column strobe signals
`
`(RAS̅̅̅̅̅ and CAS̅̅̅̅̅) are sometimes described as clocks, even though they are not
`
`periodic or fixed-frequency. An example of this usage is in the Itoh reference:
`
`
`Itoh VLSI, p. 149.
`
`
`16. Static RAMs also may include non-periodic timing signals referred to
`
`as clocks. For instance:
`
`The clock signal generator 14 receives the address
`transition detect signal SATD from the address transition
`detector 13, and produces a clock signal φ S that is
`effective in level for a fixed period. The clock signal φS is
`applied through a clock signal line 23 to the sense
`amplifier 10 and the data output circuit 11, so that those
`circuits are rendered active during the fixed period.
`
`Hayakawa, 3:50-57, emphasis added.
`
`
`17. The clock described in Hayakawa is generated in response to an
`
`address change detected by the address transition detector. This clock depends on
`
`the particular address pattern and is not periodic. Thus, this static RAM clock
`
`requires a definition broader than the definition of “clock signal” offered by Dr.
`
`Pedram.
`
`D. Dr. Pedram used the term “clock” for non-periodic signals
`in his own published papers
`
`9
`
`
`
`18.
`
`In addition to the previously cited references, some of Dr. Pedram’s
`
`own papers use the term “clock” to refer to non-periodic timing signals. Dr.
`
`Pedram lists both of these papers in his CV. See Ex. 2001, CV citations [BC18],
`
`[C217]. The following quote is from a chapter entitled “Design Technologies for
`
`Low Power VLSI” which Dr. Pedram contributed for the book Encyclopedia of
`
`Computer Science and Technology. Encyclopedia of Computer Science and
`
`Technology. In this chapter, which was published before the ’002 Patent priority
`
`date, Dr. Pedram writes:
`
`Moreover in many cases the switching of the clock causes
`a lot of additional unnecessary gate activity. For that
`reason, circuits are being developed with controllable
`clocks. This means that from the master clock other clocks
`are derived that can be slowed down or stopped
`completely with respect to the master clock, based on
`certain conditions. The circuit itself is partitioned in
`different blocks and each block is clocked with its own
`(derived) clock. The power savings that can be achieved
`this way are very application dependent, but can be
`significant.
`
`APPLE-1017, page 90, emphasis added.
`
`
`19. The bolded text describes “clocks” derived from a master clock, but
`
`these clocks are slowed down or stopped and hence would not meet Dr. Pedram’s
`
`narrow definition for clock signal because these clocks, when stopped, are not
`
`periodic.
`
`10
`
`
`
`20. A conference paper that Dr. Pedram’s co-authored was published
`
`shortly after the ’002 Patent priority date in Proc. of Int'l Conf. on Computer
`
`Design. In the paper, titled “Deterministic clock gating to eliminate wasteful
`
`activity in out-of-order superscalar processors due to wrong-path instructions,” Dr.
`
`Pedram writes:
`
`Clock gating is a well-known technique used to reduce
`power dissipation in clock associated circuitry. The idea
`of clock gating is to shut down the clock of any
`component whenever it is not being used (accessed). It
`involves inserting combinational logic along the clock
`path to prevent the unnecessary switching of sequential
`elements. The conditions under which the transition of a
`register may be safely blocked should automatically be
`detected. This problem is the target of our paper.
`
`APPLE-1018, p. 1, emphasis added.
`
`
`21. The clock of a component that is shut down is not periodic. Hence this
`
`paper is also describing non-periodic signals as clocks. Again, this usage of
`
`“clock” fails to fall under Dr. Pedram’s definition for clock signal.
`
`E. A BRI of “Clock Signal”
`
`22. The previous sections give several reasons that the plain meaning of
`
`clock should not include a requirement that the signal is periodic. If a construction
`
`of the term “clock signal” is required, it is my opinion that a person of ordinary
`
`skill in the art would have understood “clock signal” as used in the ’002 patent to
`
`mean: “a signal used for synchronization.” This proposed construction is identical
`
`11
`
`
`
`to Dr. Pedram’s construction, except that it removes his unwarranted requirement
`
`that the clock signal be “periodic.” This proposed construction encompasses the
`
`broad usage of the term in ’002 patent to refer to signals that are not periodic, the
`
`broader scope of the dictionary definitions that include non-periodic clock signals,
`
`and the usage of “clock” and “clock signal” in prior art references. In particular,
`
`this construction correctly allows for gated clocks including the conditional clock
`
`outputs in the ’002 patent.
`
`
`
`III. There is no race condition related to Sato’s clock signal
`
`23. The Pedram declaration has several paragraphs addressing a potential
`
`race condition that could occur if Sato’s clock signal, ϕce, was disconnected from
`
`decoders NAG0-NAGk. Pedram Decl. at ¶¶110-113. In my opinion, this is purely
`
`a strawman argument because Sato’s Fig. 3 and text clearly show that timing signal
`
`ϕce is connected to those decoders, and none of the ’002 Patent claims would
`
`preclude this connection from a clock to the decoder. The 4:16 decoder (second
`
`logic) of ’002 Patent claims do not specify whether this decoder is dynamic or
`
`static, and the Sato decoder as shown (including the connections to ϕce) satisfies
`
`this claim element. The operation of the decoder without the clock is irrelevant. In
`
`the actual operation of Sato as described below, there are no race conditions.
`
`12
`
`
`
`IV. Sato discloses a timing signal that synchronizes operations of its
`row decoder in the same way the ’002 clock signal synchronizes
`operation of its row decoder
`
`24. Pedram is incorrect when he states:
`
`The above-described function performed by the selection
`control signal ϕce of Sato—ensuring that the memory
`system does not enter into an unstable race condition—is
`entirely different than any function performed by the
`clock signal of the ’002 patent claims. The clock signal of
`the ’002 patent claims does not prevent an unstable race
`condition from occurring.
`
`Pedram Decl. at ¶113, emphasis added.
`
`
`25.
`
`In fact, as shown below, ϕce performs the same function required by
`
`the clock recited in ’002 Patent claims and as described in the ’002 Patent
`
`specification – determining when a selected word line is driven low or high.
`
`26. Dr. Pedram may be arguing that the “same function” must mean that
`
`the clock signal is not allowed to be used for other functions as well, but no such
`
`restriction is part of the ’002 Patent claims.
`
`27. Both the ’002 Patent and Sato drive the inputs to the final inverting
`
`wordline driver in a similar manner. A particular wordline driver is held in a static
`
`precharge state with a high input when a different group is selected or a different
`
`clock is selected. The clock selection is done by the conditional clock generator
`
`(called the predecoder PCDR in Sato). Clock ϕce of Sato performs the same
`
`function as CLOCK in the ’002 Patent claims. It transitions the wordline driver to
`
`13
`
`
`
`the active evaluation state by driving the word line driver input low when the clock
`
`is high and the corresponding address is selected.
`
`28. Table 1 compares four operational states for wordline driver XWL0 of
`
`the ’002 Patent and wordline driver WD0 of Sato in reference to the annotated
`
`figure below (described in the petition at pages 30-32 and in my first declaration at
`
`¶83). The four operational states represented in Table 1 are based on 1) whether
`
`the wordline driver group of driver XWL0 or WD0 is selected or not and 2)
`
`whether the clock generator output/PDCR output (e.g., CLK<1> or ϕx0) associated
`
`with the wordline driver (XWL0 or WD0) is high or low. In all four cases, the
`
`’002 Patent’s row address decoder (as shown in the embodiment of Figs. 1 and 3)
`
`performs the same operation as Sato’s row address decoder (as shown in Fig. 3).
`
`In each case, transistors force the input to the word line driver high or low. The
`
`input is driven low (“activating” the wordline by driving it high) only when the
`
`group is selected (by the 4:16 decoder), the clock is selected (by the 2:4
`
`predecoder), and the clock is high.
`
`
`
`14
`
`
`
`#
`
`Selection Status
`
`’002 Word Line Driver Input Sato Word Line Driver Input
`
`(ddh0)
`(input to WD0)
`
`Nonselected group
`(Address Line
`304/SO̅̅̅̅ = HIGH)
`
`Pulled HIGH by LOW CLK<0>
`turning Mp0 ON and by wordline
`via 320
`
`(Mn0 is OFF blocking Address Line
`304 signal)
`
`Wordline WL<0> is LOW
`
`0
`
`1
`
`2
`
`3
`
`Nonselected or
`LOW clock
`(CLK<0>/ϕx0 =
`LOW)
`
`Nonselected group
`(Address Line
`304/SO̅̅̅̅ = HIGH)
`
`Selected HIGH
`clock
`(CLK<0>/ϕx0)
`
`Selected group
`(Address Line
`304/SO̅̅̅̅ = LOW)
`
`Nonselected or
`LOW clock
`(CLK<0>/ϕx0 =
`LOW)
`
`Selected group
`(Address Line
`304/SO̅̅̅̅ = LOW)
`
`Selected HIGH
`clock
`(CLK<0>/ϕx0)
`
`Pulled HIGH by ϕx0 turning
`Q29 ON
`
`(Q19 is OFF blocking NAND
`decoder output SO̅̅̅̅ signal)
`
`Wordline W0 is LOW
`
`Unselected HIGH address line 304
`maintains ddh0 at a HIGH level
`through Mn0.
`
`Unselected HIGH NAND de-
`coder output SO̅̅̅̅ maintains the
`wordline driver input at a HIHG
`level through Q19.
`
`Wordline WL<0> is LOW
`
`Wordline W0 is LOW
`
`Pulled HIGH by LOW CLK<0>
`turning Mp0 ON and by wordline
`via 320
`
`(Mn0 is OFF blocking LOW Ad-
`dress Line 304 group select signal)
`
`Pulled HIGH by ϕx0 turning
`Q29 ON
`
`(Q19 is OFF blocking LOW
`NAND decoder output SO̅̅̅̅
`group select signal)
`
`Wordline WL<0> is LOW
`
`Wordline W0 is LOW
`
`Pulled LOW by LOW (selected) ad-
`dress line 304 signal through Mn0
`being turned on by HIGH CLK<0>
`
`Wordline WL<0> is HIGH
`
`Pulled LOW by LOW (se-
`lected) NAND decoder output
`SO̅̅̅̅ signal through Q19 being
`turned on by HIGH ϕx0
`
`Wordline W0 is HIGH
`
`TABLE 1
`
`15
`
`
`
`
`In summary, Sato synchronizes the operations (moving between rows
`
`29.
`
`
`
`in the table) and the transitions of different row line drivers based on timing signal
`
`ϕce. The clock signal of the ’002 patent also synchronizes operations (moving
`
`between rows in the table) and the transitions of different row line drivers based on
`
`the Clock signal. Hence, both Sato’s ϕce signal and the ’002 Patent clock signal
`
`fall under the earlier suggested definition of clock signal as “a signal used for
`
`synchronization.”
`
`16
`
`
`
`V. The row address circuit of Sato could be applied in synchronous
`memory systems
`
`30. Dr. Pedram states that:
`
`Sato’s selection control signal ϕce does not meet the
`“clock signal” limitation because it is not a periodic signal
`used for synchronization but rather one of multiple non-
`periodic timing signals (ϕce, ϕwe, ϕsa, ϕoe) used in an
`asynchronous system.
`
`Pedram Decl. at ¶114.
`
`
`31. As my previous analysis shows, the “clock signal” limitation should
`
`not be restricted to periodic signals. However, even if such a limited construction
`
`were adopted, the timing signal ϕce of Sato would still satisfy the limitation for
`
`“clock signal” because the Sato row decoder circuit would still function with
`
`periodic external signals. In other words, there is nothing in Sato suggesting to a
`
`POSITA that the periodicity of the timing signal would adversely affect the
`
`operation of Sato’s row address decoder.
`
`32. Sato shows the ϕce signal generated from the TC block in Fig.4 in
`
`response to an unspecified function of external signals CE̅̅̅̅, WE̅̅̅̅̅, and OE̅̅̅̅. No logic
`
`or timing diagrams are included in Sato to give the details of how is ϕce is
`
`generated. Dr. Pedram seems to assume that ϕce has the same or similar waveform
`
`to CE̅̅̅̅, but then incorrectly assumes that the external signal must not be periodic.
`
`33. Dr. Pedram and I agree that Sato’s timing signal must be an
`
`oscillatory signal for Sato’s row address decoder to function properly. Compare
`
`17
`
`
`
`my first declaration at ¶¶68-69 with Dr. Pedram’s declaration at ¶¶120-122. In his
`
`declaration, Dr. Pedram confirms that, while Sato does include an obscure
`
`statement that the timing signal is “kept at a high level” when the RAM is in a
`
`selection state (Sato, 3:59-65), the timing signal “ϕce cannot stay in a selected state
`
`while the row address is changing.” Pedram Decl. at ¶120. Dr. Pedram further
`
`confirms the analysis I presented in my first declaration:
`
`The petition also argues that “[a]lthough Sato includes an obscure
`
`statement suggesting that the timing signal could be kept at a high level,
`
`in contrast, a POSITA would not have read this statement as one
`
`embodiment, distinct of other aspects described within the Sato
`
`disclosure,” and that “POSITA would have understood that the Sato
`
`system would perform poorly or malfunction if the timing signal in Sato
`
`were kept at a high level throughout operation.” Petition at 16-17, citing
`
`Horst declaration ¶¶68-69. Dr. Horst’s example of how Sato’s address
`
`decoder XDCR would fail if it was supplied with two consecutive
`
`addresses 000000 and 111100 (the rightmost two bits correspond to ax0
`
`and ax1, respectively) without the timing signal ϕce changing state
`
`from a selected state to a non-selected state to a selected state (high to
`
`low to high) simply confirms the fact that ϕce cannot stay in a
`
`selected state while the row address is changing. In fact, there is a
`
`return to non-selected state requirement after a row address is correctly
`
`decoded and before a new row address comes in, and the dynamic
`
`NAND gates NAG0- NAGk would fail if one was to evaluate them
`
`multiple times without precharging them in between evaluations.
`
`These facts were well known to a POSA.
`
`18
`
`
`
`34. Our primary disagreement is merely over the timing of the oscillations
`
`or pulses in Sato’s timing signal ϕce. Dr. Pedram suggests that Sato’s memory
`
`would only be accessed at random intervals giving rise to only non-periodic timing
`
`signals such as that shown in Dr. Pedram’s timing diagram reproduced below.
`
`
`
`
`
`However, as I noted above, this is not my experience. In my experience, it is
`
`common to use asynchronous memories in systems that have a synchronous global
`
`clock, and in such systems, one would expect memory selection signals to be
`
`synchronous to the global clock. Depending on the number of memory banks,
`
`such selection may not always be periodic, but in systems with a single memory
`
`bank, one would expect the selection signal to be periodic. In such a system,
`
`Sato’s timing signal would be periodic (as shown by the modifications to Dr.
`
`Pedram’s diagram below), at least during memory access operations. This
`
`application of the Sato circuit would meet the claim limitations under Dr. Pedram’s
`
`restrictive definition of “clock signal.”
`
`19
`
`
`
`
`
`35. Even in synchronous systems, such as those illustrated by Itoh,
`
`memory control signals (e.g., RS1 and CS1) are not periodic indefinitely. See
`
`Itoh’s Figure 6.15. These signals are often gated (e.g., turned off) when the
`
`memory is not being accessed (e.g., the chip select signal is in a deselected state).
`
`In fact, Sato’s row address decoder would function fundamentally the same
`
`whether the timing signal was periodic or not. The only difference would be
`
`regularity at which Sato’s predecoder performed its operations, which are the same
`
`as the operations recited in the claims of the ’002 patent.
`
`36. Furthermore, Sato states that his address decoder could be modified
`
`(if necessary) for use in clocked synchronous systems. Modifications, such as
`
`adding a global clock input to the TC block would, in my opinion, be well within
`
`the capabilities of a POSITA. Indeed, I agree with Dr. Pedram’s deposition
`
`testimony that a POSITA would have been able to build a clock generator from
`
`well-known components in order to accomplish Sato’s clear suggestion that his
`
`20
`
`
`
`address decoder could be used in clocked synchronous systems. Pedram Depo.,
`
`48:2-21.
`
`VI. Combination of Asano and Itoh
`
`37.
`
`I agree with Dr. Pedram’s explanation of how one would have built
`
`separate address decoders for partitioned groups of address bits. Pedram Depo.,
`
`18:7-25:12. I also agree with Dr. Pedram’s explanation that the use of a
`
`programmable logic array (or “PLA”) to implement separate decoders was well-
`
`known prior to the priority date of the ’002 Patent. Pedram Depo., 23:16-25:12.
`
`Even in a single PLA, however, separate and distinct logic paths within the PLA
`
`would generate each decoder output.
`
`ADDITIONAL REMARKS
`
`
`38.
`
`I currently hold the opinions set expressed in this declaration. But my
`
`analysis may continue, and I may acquire additional information and/or attain
`
`supplemental insights that may result in added observations.
`
`39.
`
`I hereby declare that all statements made of my own knowledge are
`
`true and that all statements made on information and belief are believed to be true.
`
`I further declare that these statements were made with the knowledge that willful
`
`false statements and the like so made are punishable by fine or imprisonment, or
`
`both, under Section 1001 of the Title 18 of the United States Code and that such
`
`21
`
`
`
`willful false statements may jeopardize the validity ofthe application or any
`
`patents issued thereon.
`
`Dated: Tuy 13, 2019
`
`By:
`
` Poet
`
`Robert W.Horst/ Ph.D.
`
`De,
`
`