`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`———————
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`———————
`
`
`
`Cisco Systems, Inc.,
`Petitioner
`
`
`
`———————
`
`
`
`IPR2016-01020
`U.S. Patent No. 9,014,243
`
`IPR2016-01021
`U.S. Patent No. 8,718,158
`
`———————
`
`SECOND DECLARATION OF DR. JOSE TELLADO
`UNDER 37 C.F.R. § 1.68
`
`
`
`
`
`
`
`1
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Jose Tellado
`
`
`TABLE OF CONTENTS
`I. Background ........................................................................................................ 1
`II. Clipping is just one example of a “PAR problem” ............................................ 1
`III. Because Shively’s bit spreading technique employs multiple carriers to
`carry the same data, Shively’s technique increases PAR .................................. 3
`IV. Shively’s bit spreading technique is not limited to 18,000 foot cables ............. 3
`V. Likelihood of phase alignment of random data and data using Shively’s
`bit spreading technique ...................................................................................... 9
`VI. The 12,000 foot cable example shows how Shively’s technique increases
`PAR and the likelihood of signal clipping ....................................................... 17
`VII. A simulation of a transmitter shows that Shively’s technique increases
`PAR and the likelihood of clipping ................................................................. 25
`VIII. A POSITA would have wanted to reduce to cost of a transmitter
`employing Shively’s bit-spreading technique ................................................. 32
`IX. Stopler describes phase scrambling QAM symbols ........................................ 33
`X. Stopler does not require diagonalization ......................................................... 35
`XI. My PhD Thesis ................................................................................................ 36
`XII. Conclusion ....................................................................................................... 36
`
`
`
`2
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`I, Dr. Jose Tellado, do hereby declare as follows:
`
`I.
`
`Background
`1.
`
`I have been retained as an independent expert declarant on behalf of
`
`Cisco Systems, Inc. (“Cisco”) for the above captioned Inter Partes Reviews of U.S.
`
`Patent No. 9,014,243 (“the ’243 Patent”) and U.S. Patent No. 8,718,158 (“the ’158
`
`Patent”). I am being compensated at my usual and customary rate for the time I
`
`spend in connection with these matters. My compensation is not affected by the
`
`outcome of either matter.
`
`2.
`
`I have been asked to provide a supplemental declaration regarding
`
`certain arguments and statements made by the Patent Owner, TQ Delta, and its
`
`expert declarant, Dr. Robert T. Short.
`
`II.
`
`Clipping is just one example of a “PAR problem”
`3.
`
`Dr. Short appears to suggest that a “PAR problem” exists only when a
`
`system encounters an unacceptable level of transmission errors due to signal
`
`clipping. See, e.g., Ex. 2003, ¶30 (“A PAR “problem” exists when the actual
`
`clipping rate exceeds the maximum allowable rate.”). I disagree because Dr.
`
`Short’s conception of a “PAR problem” is too narrow. A person of ordinary skill
`
`in the art (POSITA) would have understood that in addition to causing problems
`
`during a transceiver’s operation (such as clipping), a high PAR is associated with
`
`problems and disadvantages that arise during the transceiver’s design. For
`
`1
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`example, while a transmitter can be designed to handle a high PAR signal without
`
`clipping, the resulting transmitter will generally be more expensive, less efficient,
`
`and larger. From both an engineering and a practical standpoint, the size, cost, and
`
`efficiency of a high-PAR transmitter are disadvantages and potential problems.
`
`For example, the low efficiency of such a transmitter will cause it to consume a
`
`large amount of electrical power, raising its operating expense and generating
`
`significant waste heat. At a telephone company’s central offices, many such
`
`transmitters would be used in close proximity to one another, and their waste heat
`
`would have to be removed through additional cooling equipment—at still further
`
`operational cost—to prevent them from overheating and destroying themselves. A
`
`POSITA would have considered that to be a problem, and therefore a POSITA
`
`would have been motivated to look for ways to reduce the need for high-cost, high-
`
`power, low-efficiency transmitters by reducing the PAR of the signals to be
`
`transmitted.
`
`4.
`
`Indeed, PAR reduction was an active area of research in the 1990s. It
`
`was well-known to use a bit-scrambler (or, equivalently, a phase scrambler) to
`
`produce a pseudorandomly phase-aligned multicarrier signal, which (as discussed
`
`above) has an amplitude with a Gaussian distribution. The active research areas
`
`focused on trying to achieve better than Gaussian performance, that is, to achieve
`
`PAR values that are even lower than would occur in a random system. Simply
`
`2
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`achieving Gaussian-level performance—which is all that the simple randomization
`
`techniques of the ’243 and ’158 patents achieve—was trivial and well-known.
`
`III. Because Shively’s bit spreading technique employs multiple carriers to
`carry the same data, Shively’s technique increases PAR
`5.
`
`Shively describes a bit spreading technique that “replicates (‘spreads’)
`
`a k-bit symbol over multiple adjacent bands.” Ex. 1011, 11:17-18. This causes the
`
`bands (which are also called carriers) to carry the same bit or bits, and more
`
`specifically, to be modulated using the same QAM symbol. Because these carriers
`
`carry the same data and are modulated with the same QAM symbol, the phases of
`
`these carriers align. I agree with Dr. Short’s statement that phases of the phase-
`
`aligned carriers will add coherently and create a transmission signal with a spike in
`
`power. Ex. 2003, ¶ 22; Ex. 1027, 97:21-23.
`
`6.
`
`In a system that does not implement Shively’s technique, this spike in
`
`power would not occur because these carriers would be deemed impaired and
`
`would not carry data. Because Shively’s technique causes the impaired carriers to
`
`carry the same data, Shively’s technique increases probability of new spikes in the
`
`amplitude. And, because of these spikes, Shively’s technique increases PAR.
`
`IV.
`
`Shively’s bit spreading technique is not limited to 18,000 foot cables
`7.
`
`Dr. Short considered the application of Shively’s bit-spreading
`
`technique to a line of 18,000 feet with narrow gauge (AWG 26) and therefore very
`
`high attenuation. See Ex. 2003, ¶¶58-68. However, Shively’s technique is not
`
`3
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`limited to lines of only 18000 feet and AWG 26 gauge, nor is it limited to lines
`
`suffering from very high attenuation. Shively broadly describes using its bit-
`
`spreading technique “to compensate for high attenuation and/or high noise in those
`
`parts of the communication channel frequency band that would otherwise not be
`
`unusable due to noise and attenuation effects” and reduce near-end crosstalk noise.
`
`Ex. 1011, 15:50-53, 4:35-37. Thus, Shively’s bit-spreading technique can be
`
`applied to other kinds of line impairments, such as crosstalk noise. Crosstalk noise
`
`is caused by signal coupling between adjacent lines (telephone wire pairs). The
`
`amount of crosstalk noise on a line depends on its proximity to other lines and the
`
`types of signals transmitted over those other lines. There are many sources of
`
`crosstalk (near-end cross talk, far-end cross talk, etc.) and even relatively short
`
`lines can have significant crosstalk noise.
`
`8.
`
`The ANSI T1.413-1995 standard describes both near-end and far-end
`
`crosstalk as possible noise sources in the ADSL system. See, e.g., ANSI T1.413-
`
`1995, pp. 116 & 137-146. A person of ordinary skill in the art would have been
`
`familiar with both near-end and far-end crosstalk noise. The near-end crosstalk
`
`noise is the crosstalk noise on a line that is caused by other transmitters at the same
`
`end of the line as where the noise is being measured. Far-end crosstalk noise refers
`
`to crosstalk noise on a line that is caused by transmitters at the opposite end of the
`
`line from where the noise is being measured.
`
`4
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`9.
`
`The ANSI T1.413-1995 standard provides multiple graphs showing
`
`examples of potential near-end crosstalk (NEXT) noise levels that could impact an
`
`ADSL system. The NEXT noise levels depend on types of signals that are
`
`transmitted on the adjacent communication line (or lines). Figure B.1, below,
`
`shows that having 24 adjacent DSL lines could cause significant noise levels,
`
`particularly in much of the frequency range from 0 kHz to about 530 kHz. As a
`
`reference point, I’ve annotated the diagrams with a horizontal line at -140 dBm/Hz,
`
`which was the “background noise” level shown in the attenuation graph relied
`
`upon by Dr. Short. See Ex. 2003, ¶¶ 59, 61.
`
`background
`noise floor
`
`
`
`5
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`ANSI T1.413-1995, p.138.
`
`10. Figure B.2 from the ANSI T1.413-1995 standard shows that having
`
`ten adjacent HDSL communication lines could cause significant noise levels in
`
`much of the frequency range below about 380 kHz and from 410 kHz to 650 kHz.
`
`I have similarly annotated this diagram with a background noise threshold at -140
`
`dBm/Hz:
`
`background
`noise floor
`
`
`
`ANSI T1.413-1995, p. 140.
`
`11. Figure B.3 from the ANSI T1.413-1995 standard shows that having
`
`either 4 or 20 adjacent T1 communication lines could cause increasing noise levels
`
`6
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`throughout most of the frequency range used by ADSL (0-1.1 MHz). I have
`
`annotated this diagram with a background noise threshold at -140 dBm/Hz:
`
`background
`noise floor
`
`
`
`ANSI T1.413-1995, p. 142.
`
`12. The ANSI T1.413-1995 standard also includes a graph showing
`
`potential far-end crosstalk in a typical ADSL system. Figure B.4 shows that
`
`having 10 adjacent lines being used for ADSL communications could cause
`
`significant noise levels in the lower half of the frequency spectrum used for ADSL
`
`communications, with the most significant noise levels generally in the frequency
`
`range from 10 kHz to 740 kHz. I have annotated this diagram with a background
`
`noise threshold at -140 dBm/Hz:
`
`7
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`background
`noise floor
`
`ANSI T1.413-1995, p. 142.
`
`13.
`
` These near-end and far-end crosstalk graphs show that crosstalk can
`
`be a significant impairment in a communication system and can significantly
`
`exceed the -140 dBm/Hz background noise level.
`
`14. A POSITA would have recognized that Shively’s bit-spreading
`
`technique could be usefully applied to lines suffering from noise-induced
`
`impairments, such as the examples of crosstalk noise discussed above. As I
`
`mentioned previously, Shively states that bit-spreading is a way to “compensate for
`
`high attenuation and/or high noise.” Ex.1011, 15:50. Since noise can occur on a
`
`8
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`line of any length, a POSITA would not have considered Shively’s bit-spreading
`
`technique to be limited to being used on only long lines.
`
`Likelihood of phase alignment of random data and data using Shively’s
`V.
`bit spreading technique
`15. DMT systems generally sought to randomize the data being
`
`transmitted to reduce the likelihood of phase alignment among the multiple carriers
`
`used to transmit data. To do so, they typically employed a bit-scrambler that
`
`would apply an exclusive-OR operation between each bit being transmitted and a
`
`bit generated by pseudorandom number generator. The resulting stream of bits
`
`encodes the desired data but is mathematically pseudorandom. The ANSI T1.413-
`
`1995 standard describes a typical arrangement in Section 6.3. See ANSI T1.413-
`
`1995, p. 35. A corresponding pseudorandom number generator in the receiver and
`
`another exclusive-OR operation could then be used to recover the original data
`
`bits.
`
`16. When a bit-scrambled data stream is transmitted, the bits are mapped
`
`to QAM symbols that then specify the phase of each carrier. Since the bits are
`
`pseudorandom, the QAM symbols are pseudorandom, and the resulting carrier
`
`phases are pseudorandom.1 This makes the amount of phase-alignment among the
`
`
`1 To be exact, the phase of each carrier is pseudorandomly selected from a discrete
`
`number of possible phase values, which depend on the modulation scheme. QAM-
`
`9
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`carriers also be pseudorandom, so that the sum of the signals (representing the
`
`power required to transmit the summed signal) can be appropriately approximated
`
`using a Gaussian random variable. When the carriers’ phase-alignment is not
`
`random (or pseudorandom), then the Gaussian approximation does not hold and
`
`should not be used. Shively’s bit-spreading technique causes multiple carriers to
`
`be intentionally aligned, and therefore the likelihood of their alignment is no longer
`
`randomized. A system using Shively’s bit-spreading technique cannot be
`
`reasonably approximated by a Gaussian random variable. Thus, Dr. Short was not
`
`correct in applying the Gaussian approximation to a system using Shively’s bit-
`
`spreading technique. See Ex. 2003, ¶66.
`
`17. The likelihood that a given number of carriers transmitting random
`
`symbols will all be phase-aligned depends on the number of carriers. Basically,
`
`with an increasing number of carriers, it becomes less likely that all of those
`
`carriers will be phase-aligned. The examples below further explain this concept.
`
`18. First, consider the case in which single bits are transmitted by
`
`modulating the phases of individual carriers. This modulation scheme is known as
`
`4, for example, has four phase values. In systems with a sufficient number of
`
`carriers (such as the 256 carriers called for in ANSI T1.413-1995), the difference
`
`between (a) each carrier having one of a set number of possible phases and (b)
`
`each carrier having any phase value, is not significant.
`
`10
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`binary phase shift keying (“BPSK”), because each carrier has one of two phase
`
`values. For example, the bit value of “0” can correspond to a sine wave, while the
`
`bit value of “1” can correspond to a sine wave with a 180º phase shift. If two bits
`
`are modulated onto two distinct carriers, the carriers will have the same phase
`
`when they encode the same bit values. As shown in Table I below, there are four
`
`possible combinations of bit values. In two of those combinations, the bit values
`
`are the same (either both “1” or both “0”). If all four possible combinations are
`
`equally likely (e.g., because the bit values are random), then the likelihood that
`
`both bits have the same value, and hence the likelihood of both carriers having the
`
`same phase, is 2 in 4, or 50%.
`
`Carrier #1
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`Table 1
`
`Carrier #2
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`
`Phases Aligned?
`Aligned
`
`Not aligned
`
`Not aligned
`
`Aligned
`
`
`
`19. When three carriers are phase-modulated with three bits (one bit on
`
`each carrier), there are 23 = 8 possible combinations of bit values on the carriers.
`
`The bits will all be the same only if they are all 0’s or all 1’s. Thus, the likelihood
`
`11
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`of all three carriers carrying the same value, and therefore having the same phase,
`
`is 2 in 8, or 25%. This is illustrated using the Table 2 below:
`
`Table 2
`
`Carrier #1
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`Carrier #2
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`Carrier #3
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`
`
`
`Phases Aligned?
`Aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Aligned
`
`20. Similarly, when four carriers are phase-modulated with one bit per
`
`carrier, there are 24 = 16 possible combinations of bit values. The likelihood of all
`
`four carriers carrying the same bit value, and thus having the same phase, is 2 in
`
`16, or 12.5%. This is illustrated using the Table 3 below:
`
`Table 3
`
`Carrier #1
`Bit = 0
`Phase = 0º
`
`Carrier #2
`Bit = 0
`Phase = 0º
`
`Carrier #3
`Bit = 0
`Phase = 0º
`
`Carrier #4
`Bit = 0
`Phase = 0º
`
`Phases Aligned?
`Aligned
`
`12
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`Carrier #1
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`Carrier #2
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`Carrier #3
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 1
`Phase = 180º
`
`
`Carrier #4
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`Bit = 0
`Phase = 0º
`Bit = 1
`Phase = 180º
`
`Phases Aligned?
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Not aligned
`
`Aligned
`
`21. Likewise, when eight carriers are phase modulated with one bit per
`
`carrier, there are 28 = 256 possible combinations of bit values. The likelihood of all
`
`eight carriers having the same phase is 2 in 256, or 0.78%.
`
`13
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`22.
`
`In general, when n bits are modulated on n carriers, there are 2n
`
`possible combinations of bit values. There are only two possible ways for all of
`
`the bit values to be the same (either all 0’s or all 1’s), so the likelihood of n carriers
`
`all having the same phase is:
`
`
`
`
`
`(cid:2870)(cid:2870)(cid:3289)
`
`Eq. 1
`
`23. Shively’s technique changes the likelihood of multiple carriers having
`
`the same phase significantly. When Shively’s bit-spreading technique is applied to
`
`the carriers, the bits on these carriers are no longer independent of one another.
`
`Instead, Shively’s bit spreading technique intentionally modulates multiple carriers
`
`with the same bit, which results in these carriers having the same phase.
`
`24. Shively suggests using groups of 4 carriers. Ex. 1011, 13:49-52. This
`
`means that the 4 carriers in each group will carry the same data and be modulated
`
`using the same symbol. Shively contemplates that the 4 carriers will collectively
`
`transmit a single bit of data, and thus they will have a 1-bit symbol. The phase-
`
`modulation of a single bit of data is known as BPSK. If 4 carriers use Shively’s
`
`technique, the likelihood that these 4 carriers will phase-align is 100%. It is a stark
`
`contrast to 12.5% chance of phase alignment when the 4 carriers carry random bits.
`
`If 8 carriers use Shively’s technique (two groups of 4), then the likelihood of all 8
`
`14
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`carriers being phase-aligned is 50%. Again, it is a stark contrast to the 0.78%
`
`chance if these same carriers carried random bits.
`
`25.
`
`In general, when n bits are modulated onto n carriers that are divided
`
`into groups of 4, and each group carries the same bit values, there are 2n/4 possible
`
`combinations of bit values. So, the likelihood of n carriers all having the same
`
`phase if they are employing Shively’s technique is:
`
`(cid:2870)(cid:2870)(cid:4672)(cid:3289)(cid:3120)(cid:4673)
`
`
`
`Eq. 2
`
`26. Graphing equations #1 and #2 illustrates that probability of n-carriers
`
`having perfect phase alignment when carriers carry random bits and bits using
`
`Shively’s bit spreading technique is never close, and only grows further apart as
`
`the number of carriers increase:
`
`Graph 1
`
`likelihood of perfect phase alignment
`
`number of carriers
`
`15
`
`
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`
`27. As illustrated above, the probability of carriers phase-aligning is
`
`radically different between the random-data case (equation #1) and the Shively’s
`
`bit spreading technique case (equation #2).
`
`28. The enormity of the differences also becomes apparent when
`
`considering how often the phases of carriers align in a multicarrier system, such as
`
`ADSL as defined by the ANSI T1.413-1995 standard. ADSL transmits 4000 DMT
`
`symbols per second. ANSI T1.413-1995, p. 24. Table 4 below summarizes the
`
`frequency with which n carriers will have a perfect phase alignment when a given
`
`number of carriers carry random BPSK data (equation #1) and when they carry
`
`BPSK data using Shively’s bit spreading technique (equation #2). To determine the
`
`frequency with which a set of carriers have perfect phase alignment, the
`
`likelihoods of equation #1 and equation #2 are multiplied by the DMT symbol rate
`
`(4000 symbols/second).
`
`Number of phase-
`aligned carriers
`4
`8
`16
`24
`32
`48
`52
`
`
`
`Table 4
`
`Random data frequency Shively frequency
`
`500 times per second
`31 times per second
`Once every 8 seconds
`Once every 35 minutes
`Once every 6 days
`Once every 1100 years
`One every 17,850 years
`
`4000 times per second
`2000 times per second
`500 times per second
`125 times per second
`31 times per second
`2 times per second
`Once per second
`
`16
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`29.
`
`In his analysis of an 18,000 foot cable, Dr. Short determined that a
`
`maximum of 16 carriers (4 groups of 4) carried bits using Shively’s bit spreading
`
`technique. Ex. 2003, ¶66. Dr. Short’s analysis is flawed because, in assuming a
`
`Gaussian approximation, he treats Shively’s carriers as if they carried random data
`
`and thus grossly underestimates the likelihood that those carriers will have their
`
`phases align. Dr. Short erroneously assumed that Shively’s bit spreading technique
`
`will, on average, cause all 16 carriers to have the same phase about once every 8
`
`seconds. But Table 4 demonstrates that Shively’s bit spreading technique will
`
`cause all 16 carriers to align approximately 500 times per second. Thus, Dr. Short’s
`
`numerical analysis is based entirely on an erroneous assumption and is unreliable.
`
`VI. The 12,000 foot cable example shows how Shively’s technique increases
`PAR and the likelihood of signal clipping
`30. Dr. Short stated that his analysis of an 18,000 foot narrow gauge and
`
`high loss (AWG26) cable without crosstalk is based in part on Figure 6 in Exhibit
`
`2009. Ex. 1027, 32:2-8. Figure 6, replicated below, shows line attenuation for five
`
`different cable lengths.
`
`17
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`
`
`Ex. 2009, p. 31, Fig. 6.
`
`31. Figure 6 shows line attenuation in the frequency band from 0 kHz to
`
`1100 kHz, which generally corresponds to the frequency range used for
`
`communication in the ANSI T1.413-1995 standard. The ANSI standard employs
`
`256 carriers, with each carrier spaced 4.3125 kHz apart. ANSI T1.413-1995, p. 46.
`
`Because low frequency carriers, such as carriers #1 through #6, are often reserved
`
`for analog voice and a guardband, ADSL services typically start with carrier #7.
`
`Ex. 1016, p. 187. Thus, out of 256 carriers there may be 250 downstream carriers
`
`18
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`that can carry ADSL data.2 It is appropriate to focus on the use of Shively’s
`
`technique in the downstream direction because it generally uses more carriers than
`
`the upstream direction, so the engineering considerations regarding signal power
`
`and PAR are generally greater for the downstream direction.
`
`32.
`
`In his analysis of an 18,000 foot cable, Dr. Short categorizes the
`
`carriers into three groups:
`
`
`
`Unimpaired carriers are those carriers above a threshold that can be
`
`used for ordinary ADSL communication (i.e., above the 2-BIT
`
`THRESHOLD (DMT) shown in Figure 6). Unimpaired carriers can
`
`2 The availability of 250 downstream carriers is for systems employing echo
`
`cancellation, allowing the upstream subchannels to also be used for downstream
`
`data. If echo cancellation is not used, then there will be fewer carriers used for
`
`downstream data. See ANSI T1.413-1995, p. 46:
`
`The channel analysis signal defined in 12.6.6 allows for a maximum
`of 255 carriers (at frequencies nΔf , n = 1 to 255) to be used. If echo
`cancelling (EC) is used to separate downstream and upstream signals,
`then the lower limit on n is determined by the ADSL/POTS splitting
`filters; if frequency division multiplexing (FDM) is used the lower
`limit is set by the down – up splitting filters. The cut-off frequencies
`of these filters are completely at the discretion of the manufacturer
`because, in either case, the range of usable n is determined during the
`channel estimation.
`
`19
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`carry data with high probability of success. Ex. 2003, ¶¶ 27-28; Ex.
`
`1027, 27:9-16.
`
`
`
`Impaired carriers are those carriers that cannot be used for ordinary
`
`ADSL communication (i.e., below the 2-BIT THRESHOLD) but are
`
`above the -140 dBm/Hz background noise floor. Impaired carriers
`
`can carry data using Shively’s bit spreading technique. Ex. 2003, ¶¶
`
`49-50; Ex. 1027, p. 22:24-23:7 and 90:12-14.
`
`
`
`Unusable carriers are those carriers below the noise floor (-140
`
`dBm/Hz). Unusable carriers are not used to carry data. Ex. 2003, ¶51;
`
`Ex. 1027, 28:17-20.
`
`33. Below I have applied Dr. Short’s categories to the 12,000-foot
`
`example shown in Dr. Short’s source for attenuation information. To assist with
`
`applying Dr. Short’s categories, I have annotated Fig. 6 with a horizontal line
`
`at -135 dBm/Hz, which approximates the minimum received signal strength
`
`required to reliably use a carrier to transmit one bit, when that carrier is part of a
`
`group of four carriers employing Shively’s bit-spreading technique. I arrived at the
`
`-135 dBm/Hz value as follows: First, the 1-bit threshold would normally lie about
`
`11.3 dBm/Hz above the -140 dBm/Hz noise floor, or at about -128.7 dBm/Hz.
`
`That value is approximately where Dr. Short drew his 1-bit threshold on Fig. 6.
`
`See Ex. 2003, ¶59. The annotated Figure 6 below shows the 1-bit threshold as a
`
`20
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`dashed red line. Using Shively’s bit-spreading technique across four carriers
`
`provides approximately 6 dB of coding gain, meaning that the 1-bit threshold when
`
`using Shively’s technique is 6 dBm/Hz lower, at about -135 dBm/Hz.3
`
`1-bit (non-repeating)
`threshold
`
`1-bit, 4-carrier-
`repeating threshold
`
`
`
`Unimpaired
`Carriers
`
`Impaired
`Carriers
`
`Unusable
`Carriers
`
`
`
`Ex. 2009, Fig. 6 (annotated).
`
`3 This analysis does not take into account a ± 3dB ripple permitted in the ANSI
`
`T1.413 power spectral density mask. See ANSI T1.413-1995, p. 48, Fig. 17. The
`
`power spectral density mask allows slightly more power (up to 3dB) to be put into
`
`carriers whose attenuated signal strength would be close to the -135 dBm/Hz
`
`threshold. By exploiting the ±3dB ripple, a transmitter could potentially extend the
`
`application of Shively’s bit-spreading technique to even more carriers.
`
`21
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`
`34.
`
`In effect, consideration of the minimum signal strength required to
`
`reliably transmit one bit when using Shively’s technique (i.e., the -135 dBm/Hz
`
`threshold) differs from Dr. Short’s description of the delineation between
`
`“impaired” and “unusable” carriers. Dr. Short’s description would categorize
`
`carriers between -135 dBm/Hz and -140 dBm/Hz as “impaired,” whereas I am
`
`categorizing them as “unusable” when Shively’s technique is applied to groups of
`
`four carriers. However, using Shively’s bit-spreading on these “unusable” carriers
`
`(for example, by spreading over more than 4 carriers to achieve greater coding
`
`gain) would only cause PAR to climb even higher.
`
`35. When I apply Dr. Short’s carrier categories to the 12,000 foot
`
`AWG26 attenuation graph, the carriers with frequencies from 0 to approximately
`
`810 kHz are “unimpaired” carriers that can carry random data using ordinary
`
`ADSL modulation (e.g., QAM-4). The carriers with frequencies from
`
`approximately 810 kHz to approximately 1040 kHz are “impaired” carriers that
`
`could carry data using Shively’s bit spreading technique. The carriers with
`
`frequencies approximately 1040 kHz to 1104 kHz are “unusable” carriers when
`
`Shively repeats one bit on four carriers.
`
`36. To determine the number of unimpaired carriers, I divide the carriers
`
`in the unimpaired frequency range (0 kHz to 810 kHz) by the frequency width of
`
`22
`
`
`Cisco v. TQ Delta
`CSCO-1026
`
`
`
`Second Declaration of Dr. Jose Tellado
`
`each carrier (4.3125 kHz). This analysis demonstrates that the 12,000 foot cable
`
`has approximately 188 unimpaired carriers (810 kHz / 4.3125 kHz = 188).
`
`37. To determine the number of impaired carriers, I divide the carriers in
`
`the impaired frequency range (810 kHz to 1040 kHz, or 230 kHz) by the frequency
`
`width of each carrier (4.3125 kHz). This analysis demonstrates that the 12,000 foot
`
`cable has approximately 53 impaired carriers (230 kHz / 4.3125 kHz = 53.3).
`
`Using a similar approach, I calculate that the unusable frequencies between 1040
`
`kHz and 1104 kHz correspond to approximately 15 unusable carriers: (1104 kHz –
`
`1040 kHz) / 4.3125 = 14.8 carriers.
`
`38. As I mentioned previously, the six lowest-frequency carriers were
`
`commonly left unused to avoid potential interference with ordinary analog voice
`
`communications. Thus, the 12,000 foot cable would use approximately 182
`
`unimpaired carriers (188 – 6 = 182). Shively suggests an example that uses the bit-
`
`spreading technique with carriers arranged in groups of four, which would allow
`
`for 52 of 53 impaired carriers to carry repeated data (arranged in 13 groups of 4
`
`carriers each). The leftover, 53rd carrier would be unused.
`
`39. When 52 carriers (13 groups of 4) carry repeated data, varying
`
`numbers of carriers will align at different DMT symbols, thereby creating a spike
`
`in the transmission signal amplitude. For example, Table 4