UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE ... · Petition for Inter Partes Review of...

UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD

In Re: U.S. Patent 7,116,710 : Attorney Docket No. 082944.0102

Inventor: Hui Jin, et. al. :

Filed: May 18, 2001 :

Claimed Priority: May 18, 2000 :

Issued: October 3, 2006 : IPR No. Unassigned

Assignee: California Institute of Technology

Title: Serial Concatenation of Interleaved Convolutional Codes Forming Turbo-Like Codes

Mail Stop PATENT BOARD Patent Trial and Appeal Board U.S. Patent and Trademark Office P.O. Box 1450 Alexandria, Virginia 22313-1450

Submitted Electronically via the Patent Review Processing System

PETITION FOR INTER PARTES REVIEW OF CLAIMS 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 OF U.S. PATENT NO. 7,116,710 UNDER 35 U.S.C. §§ 311-319

AND 37 C.F.R. §§ 42.100 ET SEQ. BASED ON DIVSALAR AS LEAD REFERENCE

Petition for Inter Partes Review of U.S. Patent No. 7,116,710

i

TABLE OF CONTENTS

I. MANDATORY NOTICES, STANDING, AND FEES .................................. 1

II. OVERVIEW OF CHALLENGE AND RELIEF REQUESTED .................... 2

A. Publications Relied Upon ........................................................................ 2

B. Grounds For Challenge ............................................................................ 3

III. OVERVIEW OF THE ’710 PATENT ............................................................ 4

A. Summary of the Claimed Subject Matter ................................................ 4

B. Prosecution History of the ’710 Patent .................................................... 4

IV. SUMMARY OF PRIOR ART ......................................................................... 6

A. State of the Art ......................................................................................... 6

B. Summary of References Relied Upon ..................................................... 9

V. CLAIM CONSTRUCTION .......................................................................... 10

A. Level of Ordinary Skill in the Art .......................................................... 11

B. “Repeating” Terms ................................................................................. 12

C. “Irregularly” ........................................................................................... 12

D. “Interleaving” / “Interleaver” / “Scramble” ........................................... 12

E. “Rate close to one” ................................................................................. 13

F. “Stream” ................................................................................................. 13

VI. A REASONABLE LIKELIHOOD EXISTS THAT THE CHALLENGED CLAIMS ARE UNPATENTABLE .............................................................. 14

A. Ground 1: The ‘710 Patent Claims 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 are Obvious Under 35 U.S.C. § 103 over Divsalar in view of the Luby ‘909 Patent .................................................................................... 14


ii

B. Ground 2: The ‘710 Patent Claims 15, 16, 21, and 22 are Obvious Under 35 U.S.C. § 103 Over Divsalar in view of the Luby ‘909 Patent and further in view of Hall ......................................................... 34

C. Ground 3: The ‘710 Claim 20 is Obvious Under 35 U.S.C. § 103 over Divsalar in View of the Luby ‘909 Patent in Further View of Ping ........................................................................................................ 37

D. Ground 4: The ‘710 Claim 20 is Obvious Under 35 U.S.C. § 103 over Divsalar in View of the Luby ‘909 Patent and Ping and in Further View of Hall .............................................................................. 41

VII. CONCLUSION .............................................................................................. 43


iii

LIST OF EXHIBITS

1001 U.S. Patent No. 7,116,710 by Hui Jin, et. al. entitled “Serial Concatenation of Interleaved Convolutional Codes Forming Turbo-Like Codes.” (the “’710 Patent”)

1002 Prosecution History of the ’710 Patent







1009 U.S. Provisional Application Ser. No. 60/205,095 by Hui Jin, et. al. (the “’095 Provisional Application”)

1010 Declaration of Henry D. Pfister, Ph.D.

1011 D. Divsalar, H. Jin, and R. J. McEliece, “Coding Theorems for "Turbo-like" Codes.” Proc. 36th Allerton Conf. on Comm., Control and Computing, Allerton, Illinois, pp. 201-210, Sept. 1998 (“Divsalar”) (published no later than April 30, 1999 at the University of Texas library)

1012 B.J. Frey and D.J.C. MacKay, “Irregular Turbocodes.” from the 37th Allerton Conference (“Frey”) (published no later than October 8, 1999 at the website of D.J.C. MacKay)


iv

1013 E.K. Hall and S.G. Wilson, “Stream-Oriented Turbo Codes.” 48th IEEE Vehicular Technology Conference, pp. 71-76, 1998 (“Hall”) (published no later than June 23, 1998 at the Library of Congress)

1014 L. Ping, W. K. Leung, N. Phamdo, “Low Density Parity Check Codes with Semi-random Parity Check Matrix.” Electron. Letters, Vol. 35, No. 1, pp. 38-39, Jan. 7th, 1999 (“Ping”) (published no later than April 22, 1999 at the Library of Congress)

1015 M. Luby, M. Mitzenmacher, A. Shokrollah, D. Spielman, “Analysis of Low Density Codes and Improved Designs Using Irregular Graphs.” STOC ’98 Proceedings of the Thirtieth Annual ACM symposium on Theory of Computing, pp. 249-258, 1998 (“Luby”) (published no later than July 30, 1998 at the University of Washington)

1016 U.S. Patent No. 6,081,909 by Michael Luby, et. al. entitled “Irregularly Graphed Encoding Technique.” (“the Luby ’909 Patent”) (filed November 6, 1997 and issued June 27, 2000)

1017 F. R. Kschischang and B. J. Frey, “Iterative decoding of compound codes by probability propagation in graphical models.” IEEE Journal on Selected Areas in Communications, 16, 219-230. 1998. (“Kschischang”) (published no later than Febuary 23, 1998 at the Library of Congress)

1018 U.S. Patent No. 7,089,477 by Michael Divsalar, et. al. entitled “Interleaved Serial Concatenation Forming Turbo-Like Codes .” (“the ’477 Patent”)

1019 RA.c code (including RA.c, and supporting files)

1020 J.L. Hennessy and D.A. Patterson, Computer organization and design: the hardware/software interface. 1994. (“Hennessy”) (published no later than November 8, 1994 at the Library of Congress)

1021 Complaint, California Institute of Technology v. Hughes Communications, Inc. et. al., No. 13-CV-07245 (CACD)

1022 Amended Complaint, California Institute of Technology v. Hughes Communications, Inc. et. al., No. 13-CV-07245 (CACD)


v

1023 D. J. C. MacKay, S. T. Wilson, and M. C. Davey, “Comparison of Constructions of Irregular Gallager codes.” IEEE Trans. Commun., Vol. 47, No. 10, pp. 1449-1454, Oct. 1999 (“MacKay”) (published no later than December 3, 1999 at the Library of Congress)

1024 Corrected Claim Construction Order (D.I. 105)

1025 Joint Claim Construction and Prehearing Statement (D.I. 60)

1026 Reporter’s Transcript of Claims Construction and Motion Hearing of July 9, 2014

1027 U.S. Patent No. 4,623,999 by Patricia Patterson, entitled “Look-up Table Encoder for Linear Block Codes .” (“the ’999 Patent”) (issued November 18, 1986)

1028 Luby, Mitzenmacher, Shokrollahi, Spielman, and Stemann, “Practical loss-resilient codes” in STOC '97 Proceedings of the twenty-ninth annual ACM symposium on Theory of Computing, 1997

1029 Richardson, Shokrollahi, and Urbanke, “Design of Provably Good Low-Density Parity Check Codes”

1030 Bond, Hui, and Schmidt, “Constructing Low-Density Parity-Check Codes with Circulant Matrices” ITW 1999, Metsovo, Greece (June 27-July 1 1999).

1031 Viterbi and Viterbi, “New results on serial concatenated and accumulated-convolutional turbo code performance” in Annales Des Télécommunications 1999

1032 Benedetto, Divsalar, Montorsi, and Pollara, “Serial concatenation of interleaved codes: Performance analysis, design, and iterative decoding” in IEEE Transactions on Information Theory, Vol. 44 (3), 1998

1033 McEliece, MacKay, and Cheng “Turbo Decoding as an Instance of Pearl’s “Belief Propagation” Algorithm”, as published to http://wol.ra.phy.cam.ac.uk/mackay/, under the filename “BPTD.ps.gz” by August 14, 1997


vi

1034 B.J. Frey and D.J.C. MacKay, slide presentation entitled “Irregular Turbocodes” presented at the 1999 Allerton Conference held September 22-24, 1999 (Published Sept 22-24, 1999)

1035 B.J. Frey, slide presentation entitled “Irregular Turbocodes” presented at the 2000 ISIT conference, on June 25, 2000 (Published June 25, 2000)

1036 B.J. Frey, slide presentation entitled “Irregular Turbocodes” presented at the Second International Symposium on Turbocodes and Related Topics in Brest, France in September 2000 (Published June 25, 2000)

1037 D.J.C. MacKay, slide presentation entitled “Gallagher Codes-Recent” presented at the 1999 IMA Summer Program at the University of Minnesota in Minneapolis, Minnesota (Published Aug. 3-5, 1999)

1038 Wayback Machine capture of the May 7, 1999 contents of http://wol.ra.phy.cam.ac.uk/mackay/README.html

1039 D. J. C. MacKay, S. T. Wilson, and M. C. Davey, “Comparison of Constructions of Irregular Gallager Codes” as published to http://wol.ra.phy.cam.ac.uk/mackay/, under the filename “ldpc-irreg.ps.gz” on July 30, 1998 (Published July 30, 1998)

1040 Screen capture of last-modified time information of MacKay website content

1041 D. J. C. MacKay, “Gallager codes — Recent Results” as published to http://wol.ra.phy.cam.ac.uk/mackay/, under the filename “sparsecodes.ps.gz” by July 16, 1999 (Published July 16, 1999)

1042 D.J.C. Mackay, Abstract “Gallager Codes — Recent Results” as published to http://vol.ra.phy.com.ac.wh/mackay/ under file name “sparsecodes0.ps.gz by June 2, 1999

1043 D. J. C. MacKay, “Gallager codes — Recent Results.” Proceedings of the International Symposium on Communication Theory and Applications, Ambleside, 1999, ed. by M. D. B. Honary and P. Farrell. Research Studies Press, 1999 (the “Ambleside Presentation”). (published no later than July 16, 1999 at the website of D.J.C. MacKay)


vii

1044 Portion of electronic log of D.J.C. MacKay dated July 16, 1999

1045 McEliese, et. al., slide presentation entitled “RA” presented at the Institute for Mathematics and its Applications conference on August 3, 1999.

1046 Screen capture of last modified dates of slides from https://www.ima.umn.edu/talks/workshops/aug2-13.99/mackay/mackay.html

1047 B.J. Frey and D.J.C. MacKay, “Irregular Turbocodes.” Proc. 37th Allerton Conf. on Comm., Control and Computing, Monticello, Illinois, Sep. 1999 (published no later than May 11, 2000 at the British Library Boston Spa)

1048 B.J.Frey and D.J.C. MacKay, “Irregular Turbocodes” ISIT 2000 Conference, Sorrento, Italy June 25-30, 2000

1049 D.J.C. MacKay, R.J. McEliece, J-F.Cheng, “Turbo Decoding as an Instance of Pearl’s ‘Belief Propagation’ Algorithm” as appearing on the MacKay websites as of May 7, 1999

1050 D.J.C. MacKay, “Encyclopedia of Sparse Graph Codes” as it appeared on the MacKay websites as of May 7, 1999

1051 D.J.C. MacKay, “Low Density Parity Check Codes over GF(q)” as it appeared on the MacKay websites as of May 7, 1999

1052 D.J.C. MacKay, “Decoding Times of Irregular Gallager Codes” as it appeared on the MacKay websites as of May 7, 1999

1053 D.J.C. MacKay, “Good Error-Correcting Codes Based on Very Sparse Matrices” as it appeared on the MacKay websites as of May 7, 1999

1054 D.J.C. MacKay, “Decoding Times of Repeat-Accumulate Codes” as it appeared on the MacKay websites as of May 7, 1999

1055 B.J. Frey, D.J.C. MacKay, “Trellis-Constrained Codes” as it appeared on the MacKay websites as of May 7, 1999


viii

1056 D.J.C. MacKay, “Turbo Codes are Low Density Parity Check Codes” as it appeared on the MacKay websites as of May 7, 1999

1057 H. D. Pfister and P. H. Siegel, “The serial concatenation of rate-1 codes through uniform random interleavers.” Proc. 37th Allerton Conf. on Comm., Control and Computing, Monticello, Illinois, pp. 260-269, Sep. 1999 (“Pfister”) (published no later than May 11, 2000 at the British Library Boston Spa)

1058 R. J. McEliece, “Repeat-Accumulate Codes [A Class of Turbo-like Codes that we can analyse].” 1999 Summer Program: Codes, Systems, and Graphical Models, University of Minnesota, Institute for Mathematics and its Applications, Aug. 2-13, 1999 (the “IMA Presentation”).

1059 Declaration of Brendan J. Frey

1060 Declaration of David J.C. Mackay

1061 C. Berrou, A. Glavieux, and P. Thitimajshima, “Near Shannon Limit Error Correcting Coding and Decoding.” IEEE International Conference on Communications, ICC '93 Geneva. Technical Program, Conference Record, (1993)

1062 MacKay and Neal, “Near Shannon Limit Performance of Low Density Parity Check Codes.” Electronic Letters(August 29, 1996)

1063 S. Benedetto , G. Montorsi, “Unveiling Turbo Codes: Some Results on Parallel Concatenated Coding Schemes.” IEEE Transactions on Information Theory, vol. 42, no. 2 (March 1996)

1064 Declaration of Robin Fradenburgh Concerning the “Proceedings, 36th Allerton Conference on Communications, Control, and Computing” Reference

1065 Wayback Machine capture of the December 9, 2006 contents of http://wol.ra.phy.cam.ac.uk/mackay/SourceC.html

1066 E-mail from Paul Siegel to Henry F. Pfister


1

I. MANDATORY NOTICES, STANDING, AND FEES

Real Party in Interest: Hughes Network Systems, LLC and Hughes

Communications, Inc. (“Petitioner” or “Hughes”) are the real parties in interest.

Hughes is a provider of broadband satellite services. EchoStar Corporation is the

parent of Hughes Satellite Systems Corporation, which is the parent of Hughes

Communications, Inc.

Related Matters: The ’710 Patent is currently involved in a pending lawsuit

entitled California Institute of Technology v. Hughes Communications, Inc. et. al.,

No. 13-CV-07245 (CACD) (the “Lawsuit”). See Ex. 1015. The Lawsuit

includes the following patents: (i) U.S. Patent No. 7,116,710; (ii) U.S. Patent No.

7,421,032; (iii) U.S. Patent No. 7,916,781; and (iv) U.S. Patent No. 8,284,833.

The complaint was filed on October 1, 2013 and waiver of service was filed on

November 12, 2013. Petitioner is contemporaneously filing petitions for Inter

Partes review for the other patents identified above.

Lead Counsel and Request for Authorization: Pursuant to 37 C.F.R.

§§ 42.8(b)(3) and 42.10(a), Petitioner designates the following: Lead Counsel is

Eliot D. Williams (Reg. No. 50,822) of Baker Botts L.L.P.; Back-up Counsel is G.

Hopkins Guy (Reg. No. 35,886) of Baker Botts L.L.P.

Service Information: Service information is as follows: Baker Botts L.L.P.,

1001 Page Mill Rd., Palo Alto, CA 94304-1007 Tel. 650 739 7500; Fax 650-736-


2

7699. Petitioner consents to service by electronic mail at

[email protected] and [email protected]. A Power of

Attorney is filed concurrently herewith under 37 C.F.R. § 42.10(b).

Certification of Grounds for Standing: Petitioner certifies under 37 C.F.R.

§ 42.104(a) that the ’710 Patent is available for inter partes review and that

Petitioner is not barred or estopped from requesting inter partes review on the

grounds set forth herein.

Fees: The Office is authorized to charge the fee set forth in 37 C.F.R.

§ 42.15(a) to Deposit Account No. 02-0384 as well as any additional fees that

might be due in connection with this Petition.

II. OVERVIEW OF CHALLENGE AND RELIEF REQUESTED

Petitioner challenges claims 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 of U.S.

Patent No. 7,116,710 by Hui Jin, et. al. (“the ’710 Patent”), titled “Serial

Concatenation of Interleaved Convolutional Codes Forming Turbo-Like Codes.”

See Ex. 1001.

A. Publications Relied Upon

Petitioner relies upon the following patents and publications:

Exhibit 1011 - “Coding Theorems for "Turbo-like" Codes” by D. Divsalar,

H. Jin, and R. J. McEliece (“Divsalar”), published at least by April 30, 1999 and

available as prior art under 35 U.S.C. § 102(b); see also Ex. 1064.


3

Exhibit 1013 - “Stream-Oriented Turbo Codes” by E.K. Hall and S.G.

Wilson (“Hall”), published at least by June 23, 1998 and available as prior art

under 35 U.S.C. § 102(a).

Exhibit 1014 - “Low Density Parity Check Codes with Semi-random Parity

Check Matrix” by L. Ping, W. K. Leung, N. Phamdo (“Ping”), published at least

by April 22, 1999 and available as prior art under 35 U.S.C. § 102(b).

Exhibit 1016 - U.S. Patent No. 6,081,909 entitled “Irregularly Graphed

Encoding Technique” by M. Luby, et. al. (“the Luby ’909 Patent”), filed on

November 6, 1997 and issued on June 27, 2000. The Luby ‘909 Patent is

available as prior art under 35 U.S.C. § 102(e).

B. Grounds For Challenge

Petitioner requests cancellation of the claims on the following grounds:

1. Claims 1, 3, 4, 5, 6, 15, 16, 21, and 22 are obvious over Divsalar in

view of the Luby ‘909 Patent.

2. Claims 15, 16, 21, and 22 are obvious over Divsalar in view of the

Luby ‘909 Patent and Hall.

3. Claim 20 is Obvious over Divsalar in View of the Luby ‘909

Patent in Further View of Ping

4. Claim 20 is obvious over Divsalar in view of the Luby ‘909 Patent

and Ping in Further View of Hall.


4

III. OVERVIEW OF THE ’710 PATENT

A. Summary of the Claimed Subject Matter

The ’710 Patent relates to irregular repeat accumulate (“RA”) coding for

transmission of communication signals. Claims 1, 15 and 16 describe taking data

or information bits (known in claim 1 as “data elements”) and repeating them

irregularly to determine parity bits (which the claims vaguely refer to as a “second

encoding” or a “further encode”). Claims 5 and 22 describes that these parity bits

are generated using accumulation. Claim 6 describes that the irregular encoding

is performed according to a determined profile. Claim 20 describes that the

information bits are repeated using a low-density generator matrix coder.

B. Prosecution History of the ’710 Patent

The application resulting in the ’710 Patent was filed on May 18, 2001, and

claims priority to U.S. Provisional Application Serial No. 60/205,095, filed on

May 18, 2000. Ex. 1001.

The patent examiner initially rejected various claims over U.S. Patent

6,014,411 to Wang (“Wang”). Ex. 1002 at 58, 61, 63. Applicants thereafter

amended the priority of the pending application to further claim priority from U.S.

Patent Application Ser. No. 09/922,852, filed August 18, 2000, and corrected

informalities. Id. at 71. Applicants further argued that Wang did not disclose or

teach one to “‘repeat’ ‘bits irregularly and scramble the repeated bits.’” Id. at 79.


5

Applicants further argued that “[t]here is no indication in Wang that the rate r is

irregular” and noted that various claims “recite[] that in a first encoding, bits are

repeated ‘irregularly’” or “a different number of times.” Id. at 79.

On March 4, 2005, the patent examiner issued a non-final office action

allowing various claims and objecting to others. Ex. 1002 at 87. The examiner

also rejected various claims over United States Patent 6,396,423 to Laumen

(“Laumen”). Id. at 87, 89. In response, Applicants amended pending claims 15

and 24 to recite “a second coder operative to further encode bits output from the

first coder at a rate close to within 50% of one.” Id. at 99. Furthermore,

Applicants argued that because Laumen’s coders 12 are disclosed with

transmission rates of 1/2, 1/3, 1/4, that they are not “close to one.” Id. at 104.

The patent examiner issued a final rejection on July 21, 2005 maintaining

the rejections. Ex. 1002 at 107. Specifically, the patent examiner noted that

“1/2” was within 50% of one. Id. at 110. In response, Applicants amended

claims 15 and 24 to require that “a second coder operative to further encode bits

output from the first coder at a rate within [[50%]] 10% of one.” Id. at 119.

The examiner thereafter issued a notice of allowance. Id. at 129. On February

24, 2006, Applicants submitted a Request for Continued Examination so that the

patent examiner could consider the article "Efficient encoding of low-density

parity check codes," in IEEE Trans. Inform. Theory, 47: 638-656 (February 2001)


6

by T. Richardson and R. Urbanke, which purportedly post-dated the Applicant’s

provisional filing date, and disclosed the use of irregular LDPC codes. Ex. 1002

at 141. On March 24, 2006, the patent examiner issued a notice of allowance.

Id. at 142. Applicants later requested a change in priority to application

09/922,852 as a continuation-in-part through a certificate of correction. Id. at 165.

IV. SUMMARY OF PRIOR ART

A. State of the Art

The ’710 Patent relates to error detection and correction codes used in

encoding information before transmission as a communication signal over a

communication channel. In particular, the ‘710 patent is directed to irregular

repeat-accumulate (“Irregular RA”) coding techniques. Ex. 1010 at ¶ 34.

During transmission, information contained in communication signals may

be affected by channel noise, leading to potential errors in the information when

received at the receiver. Ex. 1010 at ¶ 15. Accordingly, various coding

techniques were used in the art to generate parity bits, which are then combined

with the information bits into a codeword that is sent in the communication signal.

Id. The recipient of the codeword uses the parity bits to check the integrity of the

information bits and perform subsequent remedial action, such as error correction,

in order to recover the transmitted information. Id. at ¶¶ 15-20, 130.

One prior art technique for generating parity information based on bipartite


7

graphs was known as low density parity check (“LDPC”) codes, which were first

introduced by Robert G. Gallager in 1963 and later refined by David J.C. MacKay.

Ex. 1010 at ¶ 25. Another technique, known as repeat/accumulate (“RA”)

encoding, was published by two of the three inventors of the ‘710 Patent in

September 1998, more than one year before the earliest priority claim of the ‘710

Patent. Ex. 1010 at ¶ 31; Ex. 1011. Turbo codes were also known in the prior art.

Ex. 1010 at ¶ 23. One paper, which Patent Owner has attached to and quoted

from in its parallel district court complaint, published by authors that the Patent

Owner has admitted are “experts” in the field, classified LDPC codes, RA codes,

and turbo codes as members of the field of “random-like codes.” Ex. 1022 at ¶ 24

& at p. 88 (hereinafter, the “Roumy paper”).

It was also known that making a coding technique “irregular,” wherein

different message bits contribute to different numbers of parity bits, would

improve performance of coding techniques. Ex. 1010 at ¶¶ 28-29, 32. For

instance, by 1998 Michael Luby and others investigated whether codes based on

regular graphs would “give rise to codes that are close to optimal” and concluded

that “They do not.” Ex. 1028 at 9. Instead, Luby et al. showed that making

codes irregular yields “much better performance than regular” codes. Ex. 1010.

at ¶ 28; Ex. 1015 at 249; Ex. 1028 at 9. By mid-1999, a paper by Richardson,

Shokrollahi & Urbanke was circulating within the academic community touting


8

new “results indicating the remarkable performance that can be achieved by

properly chosen irregular codes” Ex. 1010 at ¶¶ 28; Ex. 1029 at 621.1 In August

1999, Dr. David MacKay presented a talk at the IMA academic conference on

sparse graph codes, in the speaking slot directly before one of the named inventors

of the ‘710 patent (McEliece). Ex. 1037 at p. 3. In his slides presented at that

talk, Dr. MacKay showed on one page a graph of a Gallager code, a Repeat-

accumulate code, a turbo code, and a recursive convolutional code. Id. at 42. On

the very next slide, the suggestion “make irregular” appears as the second bullet

under the heading “Where to go from Regular Gallagher Codes.” Id. at 43. The

immediate juxtaposition of these sparse graph codes, which includes a repeat-

accumulate code, with a suggestion to “make irregular,” demonstrates that a person

of ordinary skill in the art would recognize that irregularity would improve a

repeat-accumulate code. Ex. 1010 at ¶ 125.

1 A 2001 version of this paper dated after the applicants’ provisional filing date

was disclosed during prosecution. Applicants did not disclose that earlier 1999

versions of this paper were published and well known within the relevant academic

community more than 1 year before the applicant’s priority date. Ex. 1010 at ¶ 28.

By April 5, 1999, the author (Richardson) circulated the paper via Internet link to

colleagues within the academic community by e-mail. Id.


9

Thus, the prior art provided clear motivation to modify encoding schemes

using irregularity to improve performance. Ex. 1010 at ¶ 32. Indeed, the Roumy

paper, which Patent Owner has featured prominently in its district court complaint,

makes clear that this is exactly what happened -- explaining that this prior work

actually motivated the inventors of the ‘710 Patent to modify the prior art regular

RA codes by introducing irregularity: “The introduction of irregular LDPCs

motivated other schemes such as irregular RA . . . and irregular turbo codes.”

(emphasis supplied) Ex. 1022 at page 88 (Exhibit F therein).

B. Summary of References Relied Upon

The Luby ‘909 Patent (Ex. 1016)

The Luby ‘909 Patent (Ex. 1016) is a patent corresponding to work by Luby,

Mitzenmacher, Shokrollahi, Spielman, and Stemann that was academically

published in a paper entitled “Practical loss-resilient codes.” This paper discusses

irregularizing low-density parity check codes. See Ex. 1028 at 4 n.2 (“A good

candidate for the code C is the low-density parity check…”). In the Luby paper,

the authors reported on techniques they had developed to analyze regular codes,

and concluded “that they cannot yield codes that are close to optimal. Hence

irregular graphs are a necessary component of our design.” Id. at 2 (emphasis

added). In view of this, the Luby ‘909 Patent describes irregular codes (i.e. codes

based on irregular graphs) and touts their benefits:


10

[I]rregular graphing of the edges is particularly beneficial for large numbers

of data items. … For example, encoding based upon irregular graphing of

the edges can be used very effectively in high bandwidth video

transmissions.

Ex. 1016 at 11:42-47.

Divsalar (Exhibit 1011)

The Divsalar reference described a rate-1 “accumulate” convolutional

encoder that was shown to produce useful codes that could be easily decoded.

This type of code is known in the field as a “repeat-accumulate” or “RA” code.

Ex. 1010 at ¶ 34.

Hall (Exhibit 1013)

The Hall reference describes a streaming-oriented turbocode, showing how

to use prior art coding and decoding principles which were traditionally “block

coding” approaches in a streaming paradigm “without explicit block boundaries.”

Ex. 1010 at ¶ 96; Ex. 1013 at 71.

Ping (Exhibit 1014)

The Ping reference discloses using a low-density generator matrix (LDGM)

coder to perform low-density parity check coding. Ex. 1010 at ¶ 105.

V. CLAIM CONSTRUCTION

Claims 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 of the ’710 Patent are


11

unpatentable when given their “broadest reasonable construction in light of the

specification.” See 37 C.F.R. § 42.100(b). 2 Consistent with the broadest

reasonable standard, claim terms “are generally given their ordinary and customary

meaning,” as understood by “a person of ordinary skill in the art in question at the

time of the invention.” Phillips v. AWH Corp., 415 F.3d 1303, 1312-13 (Fed. Cir.

2005). The claim terms of the ’710 Patent should be given their plain and

ordinary meaning under the “broadest reasonable construction” with the

considerations discussed infra.

A. Level of Ordinary Skill in the Art

A person of ordinary skill in the art would have a very high skill level, and

would have a Ph.D. in electrical or computer engineering with emphasis in signal

processing, communications, or coding, or a master’s degree in the above area with

at least three years of work experience in this field at the time of the alleged

invention. Ex. 1010 at ¶ 43. The patent owner has accepted this level of skill in

the art in the Lawsuit. Ex. 1026 at 98.

2 Petitioner reserves the right to seek different claim constructions than those

determined by the Board or sought herein in a different forum (e.g., District Court)

that applies different standards of proof and analysis.


12

B. “Repeating” Terms

Claim 1 of the ‘710 patent requires, in part, “repeating the data elements.”

Claim 15 requires, in part, that the claimed coder “repeat said stream of bits.” In

the District Court Action, the court held that “the plain meaning of ‘repeat’

requires the creation of new bits corresponding to or reflecting the value of the

original bits… The Court will refer to this concept as duplication.” Ex 1024

(Corrected Claim Construction Order, D.I. 105) at 10. Furthermore, the plaintiff

in the District Court Action proposed that the ordinary meaning of the repeat terms

was “re-use in forming a code.” Ex. 1025 at 2.

C. “Irregularly”

Claim 15 requires that the coder “repeat said stream of bits irregularly.”

The parties in the District Court Action agreed that the term “irregularly” meant “a

different number of times.” Ex. 1025 at 1. The broadest reasonable interpretation

of these “irregularly” terms would include this definition. Ex. 1010 at ¶ 46.

D. “Interleaving” / “Interleaver” / “Scramble”

Claim 1 requires that the claimed method include “interleaving the repeated

data elements in the first encoded data block.” Claim 15 requires that the claimed

first coder be “operative to repeat said stream of bits irregularly and scramble the

repeated bits.” Claim 19 depends from claim 15 and further requires that

“wherein the first coder comprises a repeater having a variable rate and an


13

interleaver.”

The parties in the District Court Action agreed that the terms “interleaving”

and “scramble” meant “changing the order of data element” and the term

“interleaver” meant “module that changes the order of data elements.” Ex. 1025

at 1. The broadest reasonable interpretation of these terms would at least include

the agreed definitions. Ex. 1010 at ¶¶ 47-48.

E. “Rate close to one”

Claim 1 requires an encoder rate that is “close to one.” That the “rate” is

“close to one” means that the number of parity bits and information bits for a given

codeword are approximately equal. Ex. 1010 at ¶ 49. The specification states:

“[t]he inner coder 210 can have a rate that is close to 1, e.g., within 50%, more

preferably 10% and perhaps even more preferably within 1% of 1.” Ex. 1001 at

2:61-64. The broadest reasonable interpretation of “close to one,” therefore

includes a coder with a rate of 0.50 or more. Ex. 1010 at ¶ 49.

F. “Stream”

Claim 15 recites that the first coder receives a “stream of bits”. The

broadest reasonable reading of the term to a person of ordinary skill in the art

includes “a sequence of bits.” Ex. 1010 at ¶ 50-51. A narrower definition

occasionally used in the art is discussed in the Hall reference, where a “stream” is

distinguished from “blocks” of data, the boundaries of a data block are not explicit


14

and the data are “essentially indefinitely long strings.” Ex. 1013 at 71. The

specification is consistent with the broader (i.e. block-based) definition. Ex. 1001

at 2:35-38 (discussing the formatting of “blocks of data” in the context of the

“stream of data”); Ex. 1010 at ¶ 50-51. The broadest reasonable construction of

“stream” is therefore a sequence of bits.

VI. A REASONABLE LIKELIHOOD EXISTS THAT THE CHALLENGED CLAIMS ARE UNPATENTABLE

Pursuant to 37 C.F.R. § 42.104(b)(4)-(5), all of the challenged claims are

unpatentable for the reasons set forth in detail below.

A. Ground 1: The ‘710 Patent Claims 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 are Obvious Under 35 U.S.C. § 103 over Divsalar in view of the Luby ‘909 Patent

As demonstrated below, each and every element of claims 1, 3, 4, 6, 15, 16,

20, 21 and 22 is disclosed by Divsalar in view of the Luby ‘909 Patent.

1. The cited reference combination discloses all limitations

The following chart shows how all elements of claims 1, 3, 4, 5, 6, 15, 16,

21, and 22 are disclosed by the proposed combination.

‘710 Claim 1 Divsalar in view of the Luby ‘909 Patent

1[p]. A method of encoding a signal, comprising:

1[a] obtaining a block of data in the signal to be encoded;

See, e.g., Ex. 1011 at 5 “An information block of length N is repeated q times, scrambled by an interleaver of size qN, and then encoded by a rate 1 accumulator.”


15


To the extent that the preamble of this claim is limiting, Divsalar’s repeat-

accumulate code meets this limitation, as does the encoder shown in Divsalar’s

Figure 3, reproduced above. Ex. 1011 at 5; Ex. 1010 at ¶ 53. Divsalar states that

“[a]n information block of length N is repeated q times, scrambled by an

interleaver of size qN, and then encoded by a rate 1 accumulator.” Ex. 1011 at 5

(emphasis added). The encoding process is shown in Figure 3, reproduced above,

in which Divsalar obtains a block of length N and applies it to the “rate 1/q

repetition.” Divsalar’s information block of length N is the claimed “block of

data” obtained from a signal to be encoded. Ex. 1010 at ¶ 116.


1[b] partitioning said data block into a plurality of sub-blocks, each sub-block including a plurality of data elements;


See, e.g., Ex. 1016 at 11:32-49 “FIG. 12 graphs the decoding overhead against the average left degree at layer 10 nodes when the edges at layer 10 are irregularly graphed, as shown in FIG. 4, according to the distribution of FIG. 10 for nodes 10 and FIG. 11 for nodes at layer 10'. The FIG. 12


16


1[c] first encoding the data block to from a first encoded data block, said first encoding including repeating the data elements in different sub-blocks a different number of times;

performance with irregularly graphed edges can be compared to the performance with regularly graphed edges indicated in FIG. 9. As shown in FIG. 12, the decoding overhead is significantly reduced as compared with the decoding overhead for regularly graphed edges if the average left degree is 3 or more.”

See, e.g., Ex. 1016 at Figure 17

Divsalar teaches a method that includes repeating each input bit the same

number of number of times. Ex. 1011 at 5; Ex. 1010 at ¶ 118. The Luby ‘909

Patent teaches that sparse graph codes can be improved by using “irregular

graphing.” Ex. 1016 at 11:23-49 (discussing the advantages of irregular graphing

of edges). The “irregular graphing” encoder used by the Luby ‘909 Patent

constitutes first “partitioning said data block into a plurality of sub-blocks, each

sub-block including a plurality of elements” and then “repeating the data elements

in different sub-blocks a different number of times”, and finally computing parity-

check sums of the repeated bits. Ex. 1010 at ¶ 118.


17

For example, this process is represented graphically in Figure 17 of the Luby

‘909 Patent, reproduced above. In this figure, the circles on the left represent

information bits to be encoded and the circles on the right represent parity bits

computed for these information bits. Ex. 1010 at ¶ 119. In Figure 17 of the Luby

‘909 Patent, each parity bit on the right is computed by summing together (modulo

2) all of the information bits connected to that parity bits by an edge in the graph.

Id. Let d_i be the degree of the i-th information bit. The degree of an

information bit is the number of edges connected to the information bit. Id. In

Fig. 17 of the Luby ‘909 Patent, some information bits are connected to two parity

bits (i.e., have a degree of two) and other information bits are connected to three

parity bits (i.e., have a degree of three). Id. A person of ordinary skill in the art

would understand that assigning a first group of two or more input bits a first

degree (e.g., two) and a second group of input bits a second degree (e.g., three) is

within the broadest reasonable interpretation of “partitioning said data block into a

plurality of sub-blocks, each sub-block including a plurality of elements.” Id.

In Figure 17 of the Luby ‘909 Patent therefore, the input bits with a degree of two

are one sub-block and the input bits with a degree of three are a second sub-block.

Id. Each of the sub-blocks includes at least two input bits. Id.

In Figure 17 of the Luby ‘909 Patent, the parity bits are computed by first

repeating the i-th information bit d_i times and then interleaving the repeated bits


18

based on the edge connection in the graph. Ex. 1010 at ¶ 146. Next, each parity

bit with degree d is computed as the modulo-2 sum of d repeated bits. Id.

Because Fig. 17 of the Luby ‘909 Patent discloses some information bits with

degree two and other information bits with degree 3, the disclosed method of the

Luby ‘909 Patent meets the broadest reasonable interpretations of “partitioning

said data block into a plurality of sub-blocks, each sub-block including a plurality

of elements” and then “repeating the data elements in different sub-blocks a

different number of times.” Id.


1[d] interleaving the repeated data elements in the first encoded data block;


The fourth limitation of the claim is “interleaving the repeated data elements

in the first encoded data block.” The first paragraph of Section 5 in Divsalar

describes how “An information block of length is repeated times, scrambled

by an interleaver of size , and then encoded by a rate 1 accumulator.” Ex. 1011

at 5 (emphasis added). This operation is also described graphically in Figure 3 in

Divsalar (shown above). It would be clear to a person having ordinary skill in the

art that the method described in Divsalar teaches a method that includes


19

“interleaving the repeated data elements in the first encoded data block.” Ex. 1010

at ¶ 126.


1[e] and second encoding said first encoded data block using an encoder that has a rate close to one.

See, e.g., Ex. 1011 at 5

Divsalar’s Figure 3 shows a block that is labeled “rate 1 (linebreak)

1/(1+D).” Ex. 1011 at 5. This block refers to the rate-1 accumulate code of

Divsalar. The ‘710 Patent uses the identical term (i.e., “1/(1+D)”) to refer to a

rate 1 encoder. Exhibit 1001 at 2:67-3:2, Ex. 1010 at ¶ 127.


The method of claim 1, wherein said first encoding is carried out by a first coder with a variable rate less than one, and said second encoding is carried out by a second coder with a rate substantially close to one.

See, e.g., Ex. 1011 at 5


4. The method of See, e.g., Ex. 1011 at 5 “An information block of length


20


claim 3, wherein the second coder comprises an accumulator.

N is repeated q times, scrambled by an interleaver of size qN, and then encoded by a rate 1 accumulator.”

Divsalar’s repeater, as modified by the Luby ‘909 Patent, is a first coder

with a variable rate less than one. Ex. 1010 at ¶ 132. As described above,

Divsalar’s repeater, as modified by the Luby ‘909 Patent, repeats input bits a

different number of times. Id. Because certain input bits are repeated two or

more times, the rate of the first coder is less than one. Id.

Divsalar’s accumulator is a “second coder with a rate substantially close to

one” as discussed above with regards to “rate 1 accumulator.” Ex. 1010 at ¶ 133.

The combination of Divsalar and the Luby ‘909 Patent therefore discloses all

limitations of claim 3.

Section 5 in Divsalar describes how “[a]n information block of length is

repeated times, scrambled by an interleaver of size , and then encoded by a

rate 1 accumulator.” Ex. 1011 at 5. Divsalar’s accumulator is a disclosure of the

claimed “second encoder compris[ing] and accumulator.” Ex. 1010 at ¶ 135. The

combination of Divsalar and The Luby ‘909 Patent therefore discloses all

limitations of claim 4. Id.


5. The method of claim See, e.g., Ex. 1011 at 4 “There is a similar conjecture


21


4, wherein the data elements comprises bits.

for the bit error probability which we do not discuss in this paper.”


6. The method of claim 5, wherein the first coder comprises a repeater operable to repeat different sub-blocks a different number of times in response to a selected degree profile.

See, e.g., Ex. 1016 at 7:53-8:30 “FIG. 4 depicts a graph of a partial set of edges irregularly graphed from the layer 10 to the first redundant layer 10'. The lines connecting the respective nodes in layer 10 with nodes in layer 10' dictate how the redundant data items at nodes 10' are computed based upon the information in the data items associated with layer 10. In the example shown, if the respective data items at layer 10 nodes have values a-p as indicated and are graphed to the nodes at layer 10' as shown in FIG. 4, then the values of the redundant data items associated with the top four nodes in layer 10' will be computed as shown using an exclusive-or operator. For example, the top or first node will have an associated value representing a exclusive-or b exclusive-or f.

As will be recognized by those skilled in the art, computing systems typically include an exclusive-or function as a standard operation which performs a sum of bits mod2 on a bit by bit basis. It will further be recognized that conventional computing systems are typically capable of performing an exclusive-or operation on each bit within a word in parallel, i.e., in one operation. Accordingly, the exclusive-or operation adds relatively little overhead to the encoding processing. Hence, if the information within each of the data items at the message nodes 10 consists of several words, as is customary, then an exclusive-or would be taken of each word in data item a with the corresponding words of data item b and an exclusive-or of the corresponding words of data item f. This will result in a value at the associated node 10' of the same length as the information in data items a or b or f and the value will


22


equal the value of a exclusive-or b exclusive-or f.

Each of the data items and redundant data items at nodes 10, 10', 10" and 10"' includes an associated index which is received with the item. The index corresponds to the information within the item. Accordingly, a recipient is able to determine that a received redundant data item associated with the top node of layer 10' includes a x-or b x-or f. More particularly, the index identifies the node with which the item is associated

See, e.g., Ex. 1016 at Figure 4:

Both Divsalar and the Luby ‘909 Patent receive input in the form of data

elements comprising bits. Ex. 1010 at ¶ 136. In Divsalar, the repeater repeats

different sub-blocks the same number of times. The Luby ‘909 Patent, however,

discloses the repetition of different sub-blocks a different number of times in


23

response to a selected degree profile. Id. at ¶ 137. This is shown graphically,

for example, in Figure 4 of the Luby ‘909 Patent, reproduced above. Id. The Luby

‘909 Patent describes Figure 4 as reproduced above in the claim chart.

A person of ordinary skill in the art would understand that Divsalar’s

repeater, as modified by the Luby ‘909 Patent’s selective repetition of input bits a-

p is a “repeater operable to repeat different sub-blocks a different number of times

in response to a selected degree profile,” within the broadest reasonable

construction of that term. Ex. 1010 at ¶ 139. The “selected degree profile is

disclosed by the number of times each input bit contributes to an output bit. Id.


15[p]. A coder comprising:

15[a] a first coder having an input configured to receive a stream of bits, said first coder operative to repeat said stream of bits irregularly and scramble the repeated bits

See, e.g., Ex. 1016 at 1:12-15 “In downloading data from storage or communicating data, for example over a communications network such as the INTERNET, data is transmitted in streams of message packets.”

See, e.g., Ex. 1011 at 5

See, e.g., Ex. 1016 at 11:32-49 “FIG. 12 graphs the decoding overhead against the average left degree at layer 10 nodes when the edges at layer 10 are irregularly graphed, as shown in FIG. 4, according to the distribution of FIG. 10 for nodes 10 and FIG. 11 for nodes at layer 10'. The FIG. 12 performance with irregularly graphed edges


24


can be compared to the performance with regularly graphed edges indicated in FIG. 9. As shown in FIG. 12, the decoding overhead is significantly reduced as compared with the decoding overhead for regularly graphed edges if the average left degree is 3 or more.”


To the extent that the preamble of this claim is limiting, Divsalar’s repeat-

accumulate code meets this limitation. Ex. 1011 at 5; Ex. 1010 at ¶ 141. As

discussed above in section V.F, while the ’710 Patent does not define the term

“stream,” it would be clear to a person having ordinary skill in the art that encoders,

such as Divsalar’s, have “input configured to receive a stream of bits” step as part

of encoding or decoding. Ex. 1010 at ¶ 142. Furthermore, Divsalar includes a

first coder in Figure 3 that receives a block of length N and performs a rate 1/q

repetition. Ex. 1011 at 5. Divsalar, however, does not “repeat said stream of bits

irregularly,” as required by claim 15. As discussed above with respect to claim 1

(limitations 1[b] and 1[c]), the Luby ‘909 Patent teaches that sparse graph codes


25

can be improved by using “irregular graphing.” Ex. 1016 at Figure 17 and 11:23-

49. It would have been obvious to a person having ordinary skill in the art to

combine “irregular graphing” with Divsalar’s “repeat and accumulate codes.” Ex.

1010 at ¶ 147.


15[b] a second coder operative to further encode bits output from the first coder at a rate within 10% of one.


Divsalar’s second coder (i.e. the accumulator) has a rate of one, as discussed

above with respect to claim 1 (limitation 1[e]). Ex. 1011 at 5; Ex. 1010 at ¶ 148-

150.


16. The coder of claim 15, wherein the stream of bits includes a data block, and wherein the first coder is operative to apportion said data block into a plurality of sub-blocks and to repeat bits in each sub-block a

See, e.g., Ex. 1016 at 11:32-49 “FIG. 12 graphs the decoding overhead against the average left degree at layer 10 nodes when the edges at layer 10 are irregularly graphed, as shown in FIG. 4, according to the distribution of FIG. 10 for nodes 10 and FIG. 11 for nodes at layer 10'. The FIG. 12 performance with irregularly graphed edges can be compared to the performance with regularly graphed edges indicated in FIG. 9. As shown in FIG. 12, the decoding


26


number of times, wherein bits in different sub-blocks are repeated a different number of times.

overhead is significantly reduced as compared with the decoding overhead for regularly graphed edges if the average left degree is 3 or more.”


This limitation is disclosed by Divsalar’s repeater, as modified by the Luby

‘909 Patent. Ex. 1010 at ¶ 158. Divsalar teaches a method that includes repeating

each input bit the same number of number of times. Id. The Luby ‘909 Patent

teaches that sparse graph codes can be improved by using “irregular graphing.” Ex.

1016 at 11:23-49 and Figure 17, as discussed above with respect to claim 1

(limitations 1[b] and 1[c]). Id. at ¶¶ 158-160. The combination of Divsalar with

the Luby ‘909 Patent therefore meets the broadest reasonable interpretation of a

first coder “wherein the first coder is operative to apportion said data block into a

plurality of sub-blocks and to repeat bits in each sub-block a number of times,

wherein bits in different sub-blocks are repeated a different number of times.” Id.


27

at ¶ 161. Claim 16 is therefore obvious over the combination of Divsalar and the

Luby ‘909 Patent.


21. The coder of claim 15, wherein the second coder comprises a rate 1 linear encoder.

See, e.g., Ex. 1011 at 5


22. The coder of claim 21, wherein the second coder comprises an accumulator


Divsalar’s RA encoder is shown graphically in Figure 3. Ex. 1011 at 5.

Divsalar’s Figure 3 and the associated text explicitly state that the second encoder,

which is an accumulator, is “rate 1” Ex. 1010 at ¶¶ 169-171; Ex. 1011 at 5.

Thus, the combination of the Luby ‘909 Patent with the rate-one seconding

encoding of Divsalar discloses all elements of claim 22.

2. One of skill in the art would be motivated to combine the references

A person of ordinary skill in the art would have been motivated to combine

Divsalar and the Luby ‘909 Patent. As discussed above, Divsalar discloses a

repeat and accumulate code while the Luby ‘909 Patent discloses irregularity.


28

The combination of Divsalar with the Luby ‘909 Patent would produce a

predictable result. Ex. 1010 at ¶ 121. Any performance improvement to the

“repeat and accumulate code” of Divsalar due to “irregular graphing” in the Luby

‘909 Patent was predictable based on the knowledge of a person of ordinary skill in

the art. Id. Indeed, documents cited in Patent Owner’s Complaint in the

Litigation confirms that Luby’s work motivated these inventors to add irregularity

to the RA code in Divsalar. Supra at section IV.A; Ex. 1022 at page 88 (Exhibit

F therein).

Furthermore, these references are in the same field of inquiry and are

directed towards solving the same problem. Ex. 1010 at ¶ 121. In particular, the

Luby ‘909 Patent itself suggests the use of “irregular graphing” to improve on

known results with “regular graphing” and describes how “[w]ith irregular

graphing of the edges, the decoding overhead is approximately 2.3%...With regular

graphing, over 14% overhead is required.” Ex. 1016 at 11:58-61. Thus, it

would have been obvious to a person having ordinary skill in the art to combine

“irregular graphing” with Divsalar’s “repeat and accumulate codes.”

The Luby ‘909 Patent shows that coding performance is significantly

improved by allowing the code bits of an LDPC or LDGM code to have irregular

degrees. Ex. 1010 at ¶ 128. By the time of the invention, it was well known in the

art that one could improve code performance by adding irregularity to iteratively


29

decodable codes. Id. Thus, the combination of Divsalar with the Luby ’909

Patent would produce a predictable result. Any improvement to the “repeat and

accumulate” code in Divsalar due to irregularity was therefore predictable based

on the general knowledge in the art. Id.

For example, Divsalar teaches a method that includes repeating each

message bit the same number of times and that this results in a regular Tanner

graph. Ex. 1010 at ¶ 129. A Tanner graph is a bipartite graph representation of

parity checking operation of a given coding scheme. Ex. 1023 at 2. The Luby

‘909 Patent teaches that codes defined by Tanner graphs can be improved by using

“irregular graphing.” Ex. 1016 at 11:23-49. It would have been obvious to a

person having ordinary skill in the art that making the information bit degrees

irregular in the Tanner graph of a “repeat and accumulate” code is identical to

using an irregular repeat pattern. Ex. 1010 at ¶ 129. Thus, the combination of

Divsalar and the 909 Patent produces an irregular “repeat and accumulate” code

whose performance improvement was predictable from the cited prior art. Id.

A person of ordinary skill in the art would further be motivated to modify

Divsalar by the teachings of the Luby ‘909 Patent based on a Tanner graph

representation of the RA code of Divsalar. Ex. 1010 at ¶ 129. The Tanner graph

of a binary parity-check code defines and represents the additive relationships

between all of the bits involved in the code. Ex. 1010 at ¶ 129. In Kschischang,


30

a Tanner graph is described as follows (Fig. 1(b) is shown below the text):

A Tanner graph is a bipartite graph representation for a check structure, similar to the one described above. In such a graph, there are two types of vertices corresponding to the variables and the “checks”, respectively, with no edges connecting vertices of the same type. For example, a Tanner graph corresponding to the Hamming code described above is shown in Fig. 1(b). Each check vertex q in the set of check vertices Q is shown as a filled circle. In this case, a check vertex ensures that its set of neighbors satisfies even parity in a valid configuration.

Ex. 1023 at 2.

At the time of the alleged invention, a practitioner of ordinary skill in the art

could map between a description of a binary code defined by parity-check

equations or by a function description of the encoder (as provided in Divsalar) and

a description of the same code as a Tanner graph. Ex. 1023 at 2-3 (discussing

Tanner graphs representing codes); Ex. 1010 at ¶ 123. In a Tanner graph, each bit

involved is mapped to a variable node (i.e., numbered circle) and each equation is


31

mapped to a check node (i.e., filled circle). Ex. 1010 at ¶ 123. The implied

constraint is that all variable nodes attached to check node must sum to zero

modulo 2. Ex. 1010 at ¶ 123. A Tanner graph of a decoder for Divsalar’s RA

code would be equivalent to the Tanner graph provided in Figure 10 of U.S. Patent

No. 7,089,477 (“Divsalar ‘477”), which is shown below.

Ex. 1018 at Figure 10; 5:47-48 (“The Tanner Graph realization for an RA code is

explained with reference to FIG. 10.”); Ex. 1010 at ¶ 123.

In the figure above of the non-irregular RA code of Divsalar, each

information node is connected to three check nodes. As discussed above, the

Luby ‘909 Patent demonstrates the advantage of “irregular graphing,” where

information nodes are connected to different numbers of check nodes. Ex. 1010 at


32

¶ 124. For example, this process is represented graphically in Figure 17 of the

Luby ‘909 Patent.

Ex. 1016 at Figure 17. In this figure, the circles on the left represent information

bits to be encoded and the circles on the right represent parity bits computed for

these information bits. Ex. 1010 at ¶ 125. For example, in Figure 17 of the Luby

‘909 Patent, each parity bit on the right is computed by summing together (modulo

2) all of the information bits connected to that parity bits by an edge in the graph.

Id. In Fig. 17 of the Luby ‘909 Patent, some information bits are connected to

two parity bits (i.e., have a degree of two) and other information bits are connected

to three parity bits (i.e., have a degree of three). Id. A person of ordinary skill in

the art would therefore be further motivated to modify the encoder of Divsalar

based on its Tanner graph representation to have an “irregular graphing” between

information and check nodes. Ex. 1010 at ¶ Id. As discussed above, the Luby


33

‘909 Patent demonstrated the advantages of such an irregularly graphed code. Id.

Also, both of these references are in the same field of inquiry and are

directed towards solving the same problem. Specifically, both of these references

focus on the design of binary error-correcting codes and both seek to provide

reliable communication with minimal overhead. Ex. 1010 at ¶ 130. For example,

Divsalar states on page 5 that “[i]n this section we will introduce a class of turbo-

like codes which are simple enough so that we can prove the IGE conjecture. We

call these codes repeat and accumulate (RA) codes.” Ex. 1011 at 5. The technical

field defined by the Luby ‘909 Patent is “[t]he present invention relates to loss

resilient and error correcting codes and, more particularly, to a technique for

creating loss resilient and error correcting codes having irregular graphing between

the message data and the redundant data.” Ex. 1016 at 1:5-10. Because the Luby

‘909 Patent suggests the use of irregularity to improve performance, a practitioner

having ordinary skill in the art would therefore have been aware that one could

improve performance by adding irregularity to iteratively decodable codes. Ex.

1010 at ¶ 130. A person of ordinary skill in the art would have been motivated to

modify an existing code to introduce irregularity to improve the performance of the

code. Id.


34

B. Ground 2: The ‘710 Patent Claims 15, 16, 21, and 22 are Obvious Under 35 U.S.C. § 103 Over Divsalar in view of the Luby ‘909 Patent and further in view of Hall

As demonstrated below and in the claim charts in Section VI.A.1, supra,

each and every element of claims 15, 16, 21, and 22 is disclosed by Divsalar in

view of the Luby ‘909 Patent and in further view of Hall. This ground is

presented in the unlikely event the Board construes “stream” in the narrower

manner that the term is used in the Hall reference, rather under what petitioner

contends is the broadest reasonable construction.


The first limitation of claim 15 is “a first coder having an input configured to

receive a stream of bits, said first encoder operative to repeat said stream of bits

irregularly and scramble the repeated bits.” Divsalar in view of the Luby ‘909

Patent discloses this limitation as described above in section VI.A.1.

If, however, the Board adopts a narrower construction of “stream,” then Hall

discloses “streaming-oriented turbo codes” by replacing the block permuter of

Divsalar by a convolutional interleaver. Ex. 1013 at 71; Ex. 1010 at ¶ 96. Section

I of Hall states that turbo codes “have been traditionally viewed as a block coding

technique. With this viewpoint, research has focused on the block interleaver

design and block decoding architectures. In this paper, we explore a stream

encoding paradigm, without explicit block boundaries.” Id. The benefits of


35

stream-oriented turbo codes are described in Hall. Section 6 of Hall states that

“[c]onvolutional interleavers were shown to be attractive for low decoding latency

applications, and with higher delay, provide a structured interleaving option with

near-capacity performance.” Id. at 76; Ex. 1010 at ¶ 96.

The limitation further requires that the “first coder [is] operative to repeat

said stream of bits irregularly and scramble the repeated bits.” Divsalar in view

of the Luby ‘909 Patent discloses this limitation as described above in section

VI.A.1.

Claim 15 further requires “a second coder operative to further encode bits

output from the first coder at a rate within 10% of one.” Divsalar in view of the

Luby ‘909 Patent discloses this limitation as described above in section VI.A.1.

Claim 16 depends from claim 15 and adds the limitation that “wherein the

stream of bits includes a data block, and wherein the first coder is operative to

apportion said data block into a plurality of sub-blocks and to repeat bits in each

sub-block a number of times, wherein bits in different sub-blocks are repeated a

different number of times.” Divsalar in view of the Luby ‘909 Patent discloses this

limitation as described above in section VI.A.1.

Claim 21 depends from claim 15 and further requires, “wherein the second

coder comprises a rate 1 linear encoder.” Divsalar’s accumulator is a rate 1 linear

encoder. Divsalar in view of the Luby ‘909 Patent discloses this limitation as


36

described above in section VI.A.1.

Claim 22 depends from claim 21 and further requires, “wherein the second

coder comprises an accumulator.” Divsalar in view of the Luby ‘909 Patent

discloses this limitation as described above in section VI.A.1.



Hall with Divsalar and the Luby ‘909 Patent. It would have been obvious to

combine Divsalar with the Luby ‘909 Patent for at least the reasons discussed

above in section VI.A.2.

Furthermore, it would be obvious to a person having ordinary skill in art that

the “repeat and accumulate” codes in Divsalar could be implemented in a

streaming fashion (such as that disclosed in Hall) by replacing the permutation

block in Figure 3 of Divsalar by a convolutional interleaver. Ex. 1010 at ¶ 97.

Any improvement to the latency or performance of the “repeat and accumulate”

codes due to stream encoding is therefore predictable based on the general

knowledge in the art. Id. These references are in the same field of inquiry and

are directed towards solving the same problem. Id. The references themselves

suggest the use of streaming to improve performance and a practitioner having

ordinary skill in the art would have been aware of the benefits and drawbacks of

streaming operation. Id.


37

The encoder of Divsalar, as modified by Hall does not “repeat said stream

of bits irregularly.” It would have been obvious, however, to modify the encoder

of Divsalar, as modified by Hall to “repeat said stream of bits irregularly,” at least

based on the Luby ‘909 Patent for the same reasons as discussed above in section

VI.A.2.

C. Ground 3: The ‘710 Claim 20 is Obvious Under 35 U.S.C. § 103 over Divsalar in View of the Luby ‘909 Patent in Further View of Ping

Each and every element of claim 20 is disclosed by Divsalar in view of the

Luby ‘909 Patent and in further view of Ping.


Claim 15, from which claim 20 depends, is disclosed by the combination of

Divsalar in view of the Luby ‘909 Patent as shown above in section VI.A.1. The

following chart shows how all elements of claim 20 are disclosed by the proposed

combination by further reference to Ping

‘710 Claim 20 Divsalar in View the Luby ‘909 Patent and in Further View of Ping

20. The coder of claim 15, wherein the first coder comprises a low-density generator matrix coder.

See. e.g., Ping Ex. 1014 at 38 “An LDPC code is defined from a randomly generated parity check matrix H...In this Letter, we report a modified approach to LDPC code design. We adopt a semi-random technique, i.e. only part of H is generated randomly, and the remaining part is deterministic. The new method can achieve essentially the same performance as the standard LDPC encoding method with significantly reduced complexity.”

“Based on eqns. 1 and 2, p = {pi}can easily be


38

‘710 Claim 20 Divsalar in View the Luby ‘909 Patent and in Further View of Ping

calculated from a given d = {di} as

and .(mod2)(4)”

The ’710 Patent includes an embodiment where the repeater and interleaver

are replaced with a low-density generator matrix coder, in Figure. 4.

Ex. 1001 at Figure 4. In this figure, no interleaver is shown for the embodiment

with an LDGM code. Ex. 1010 at ¶ 163. The ‘710 patent at 3:57-59 states that

“[t]he interleaver 204 in FIGURE 2 may be excluded due to the randomness

already present in the structure of the LDGM code.”

Ping discloses a low-density generator matrix coder. The low-density

generator matrix of Ping is not irregular. Ex. 1010 at ¶ 164. That is, the low-

density generator matrix of Ping does not include “first coder operative to repeat

said stream of bits irregularly,” as required by claim 15. Id. The Luby ‘909 Patent

teaches that sparse graph codes, such as an LDGM, can be improved by using

“irregular graphing.” See section VI.A.1; Ex. 1016 at 11:23-49 (discussing the

advantages of irregular graphing of edges); Ex. 1010 at ¶ 164. A person of

ordinary skill in the art would modify the low-density generator matrix of Ping by


39

making the matrix irregular. Ex. 1010 at ¶ 164.

To show this, a mathematical definition of LDGM codes may be considered.

Ex. 1010 at ¶ 165. This is given by basic coding theory under the assumption that

the generator matrix is sparse (or low density). Id. Thus, the modulo-2

sum defines a mapping from an information vector

to a codeword . Id. This is called a regular LDGM

code if all rows of have the same number of ones and all columns of have

the same number of ones. Id. Likewise, the output of an

accumulator with input is defined by for

where . Id.

The encoder of claim 20 is shown in Figure 4 of the ‘710 patent (shown

above). Using the above notation and this Figure, it can be observed that the

input would be mapped to the pair of outputs

and . Ex. 1010 at ¶ 166. If a systematic code is desired, then both

sequences are used; otherwise, only the sequence is used. Id.

To see that Ping discloses all elements except for irregularity, it should be

shown that the equations in Ping are mathematically identical to the encoding

equations above. Ex. 1010 at ¶ 167. Ping states that “the codeword as

where and contain parity and information bits, respectively.”

Ex. 1014 at 38. Using for the information bits and


40

for the parity bits, Equation 4 in Ping defines the encoder

with the two modulo-2 sums: and . Id.; Ex.

1010 at ¶ 167. In Ping, the matrix is random matrix with ones per column

and ones per row. Ex. 1014 at 38; Ex. 1010 at ¶ 167. Based on the

equation , is the generator matrix of a regular LDGM code. Ex.

1010 at ¶ 167. If , then is output of the LDGM

encoder and is the output of an accumulator with input

Id. By relabeling , , and and comparing

with the previous two paragraphs, Ping’s Equation 4 is mathematically identical to

the combination of a LDGM encoder and accumulator. Id.

The combination of Ping with the Luby ‘909 Patent and Divsalar therefore

discloses all elements of claim 20 under their broadest reasonable constructions.



the Luby ‘909 Patent, Divsalar, and Ping. Motivations to combine the Luby ‘909

Patent, and Divsalar, are discussed supra in Ground 1. One of ordinary skill in

the art would have been motivated to combine these with Ping by modifying

Ping’s encoding system to use an irregular LDGM code based on the teaching of

Frey. Ex. 1010 at ¶ 168. For example, the abstract of Ping states that “[a] semi-


41

random approach to low density parity check code design is shown to achieve

essentially the same performance as an existing method, but with considerably

reduced complexity.” Ex. 1014 at 38.

Furthermore, the combination of Ping with the Luby ‘909 Patent or Divsalar

produces a predictable result. Ex. 1010 at ¶ 168. Specifically, any improvement

to the Ping code due to irregularity in the Luby ‘909 Patent produces a predictable

result. Id. Any improvement to the Ping code due to irregularity was therefore

predictable based on the general knowledge in the art. Id. A person of ordinary

skill in the art would have been motivated to combine these references because

those in the field of coding theory had recognized the general improvement that

irregularity adds to an existing non-irregular code. Id. One in the field would

understand the advantages of making a code irregular. Id.

D. Ground 4: The ‘710 Claim 20 is Obvious Under 35 U.S.C. § 103 over Divsalar in View of the Luby ‘909 Patent and Ping and in Further View of Hall

This ground is presented in the unlikely event the Board construes “stream”

in the narrower manner that the term is used in the Hall reference, rather under

what petitioner contends is the broadest reasonable construction.


As explained above in Ground 3, claim 20 is obvious over Divsalar in view

of the Luby ‘909 Patent and Ping. If, however, the Board adopts a narrower


42

construction of “stream,” then (as discussed above in connection with Ground 2)

Hall discloses “streaming-oriented turbo codes” by replacing the block permuter of

Divsalar by a convolutional interleaver. Ex. 1013 at 71; Ex. 1010 at ¶ 96. Section

I of Hall states that turbo codes “have been traditionally viewed as a block coding

technique. With this viewpoint, research has focused on the block interleaver

design and block decoding architectures. In this paper, we explore a stream

encoding paradigm, without explicit block boundaries.” Id. The benefits of

stream-oriented turbo codes are described in Hall. Section 6 of Hall states that

“[c]onvolutional interleavers were shown to be attractive for low decoding latency

applications, and with higher delay, provide a structured interleaving option with

near-capacity performance.” Id. at 76; Ex. 1010 at ¶ 96.



Hall with Divsalar, the Luby ‘909 Patent and Ping. It would have been obvious

to combine Divsalar with the Luby ‘909 Patent and Ping for at least the reasons

discussed above in connection with Ground 3.

Furthermore, as discussed above in connection with Ground 2, it would be

obvious to a person having ordinary skill in art that the “repeat and accumulate”

codes in Divsalar could be implemented in a streaming fashion (such as that

disclosed in Hall) by replacing the permutation block in Figure 3 of Divsalar by a


43

convolutional interleaver. Ex. 1010 at ¶ 97. Any improvement to the latency or

performance of the “repeat and accumulate” codes due to stream encoding is

therefore predictable based on the general knowledge in the art. Id. These

references are in the same field of inquiry and are directed towards solving the

same problem. Id. The references themselves suggest the use of streaming to

improve performance and a practitioner having ordinary skill in the art would have

been aware of the benefits and drawbacks of streaming operation. Id.

VII. CONCLUSION

Petitioner respectfully requests that inter partes review of the ’710 Patent be

instituted and that claims 1, 3, 4, 5, 6, 15, 16, 20, 21, and 22 be cancelled as

unpatentable under 35 U.S.C. § 318(b).


44

Respectfully submitted,

BAKER BOTTS L.L.P.

October 14. 2014 _/Eliot D. Williams/___________________

Date Eliot D. Williams (Reg. No. 50,822) G. Hopkins Guy III (Reg. No. 35,866)

1001 Page Mill Road, Bld. 1, Suite 200 Palo Alto, California 94304-1007 650.739.7510 Attorneys for Petitioner, Hughes Network Systems, L.L.C. and Hughes Communications, Inc.

CERTIFICATE OF SERVICE

In accordance with 37 C.F.R. §§ 42.6(e) and 42.105, the undersigned

certifies that on the 14th day of October 2014, a complete and entire copy of the

PETITION FOR INTER PARTES REVIEW OF CLAIMS 1, 3, 4, 5, 6, 15, 16,

20, 21, and 22 OF U.S. PATENT NO. 7,116,710 UNDER 35 U.S.C. §§ 311-319

AND 37 C.F.R. §§ 42.100 ET SEQ. BASED ON DIVSALAR AS LEAD

REFERENCE (“petition”) including exhibits and testimony relied upon were

served on the patent owner at the correspondence address of record for the subject

patent,

California Institute of Technology 1200 E.California Blvd.

M/C 201-85 Pasadena, CA 91125

via Express Mail and to counsel for patent owner in the Lawsuit,

Quinn Emanuel Urquhart & Sullivan, LLP

James R. Aspberger 865 S. Figueroa St., 10th Floor Los Angeles, California 90017

via Express Mail.

October 14. 2014 _/Eliot D. Williams/___________________ Date Eliot D. Williams (Reg. No. 50,822)

G. Hopkins Guy III (Reg. No. 35,866) 1001 Page Mill Road, Bld. 1, Suite 200 Palo Alto, California 94304-1007 650.739.7510 Attorneys for Petitioner, Hughes Network Systems, L.L.C. and Hughes Communications, Inc.

UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE ... · Petition for Inter Partes Review of...

Documents

Transcript of UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE ... · Petition for Inter Partes Review of...