cafit

7/28/2019 cafit

1/84

No. 2012-1677

IN THE

United States Court of AppealsFOR THE FEDERAL CIRCUIT

RAMBUS,INC.

Appellant,

v.

INTERNATIONAL TRADE COMMISSION,

Appellee,

and

GARMIN INTERNATIONAL,INC.,

Intervenor,

and

LSICORPORATION AND SEAGATE TECHNOLOGY LLC,

Intervenors,

and

STMICROELECTRONICSN.V. AND STMICROELECTRONICS INC.,

Intervenors,

and

CISCO SYSTEMS,INC.,Intervenor,

and

HGST,INC.,

Intervenor,

and

HEWLETT-PACKARD COMPANY,

Intervenor.

Appeal from the United States International Trade Commission

in Investigation No. 337-TA-753

NONCONFIDENTIAL BRIEF FOR APPELLANT RAMBUS INC.

Case: 12-1677 Document: 92-1 Page: 1 Filed: 05/23/2013

7/28/2019 cafit

2/84

Doris Johnson Hines

FINNEGAN,HENDERSON,FARABOW,

GARRETT &DUNNER,LLP

901 New York Avenue, N.W.

Washington, D.C. 20001(202) 408-4000

Jason E. Stach

FINNEGAN,HENDERSON,FARABOW,

GARRETT &DUNNER,LLP

3500 Sun Trust Plaza, Suite 3500

303 Peachtree Street, N.E.

Atlanta, GA 30308

(404) 653-6400

John M. Whealan

4613 Merivale Road

Chevy Chase, MD 20815

(202) 994-2195

Jeffrey A. Lamken

Counsel of RecordMichael G. Pattillo, Jr.

MOLOLAMKEN LLP

The Watergate, Suite 660

600 New Hampshire Avenue, N.W.

Washington, D.C. 20037

(202) 556-2010

Counsel for Appellant Rambus Inc.


7/28/2019 cafit

3/84

i

CERTIFICATE OF INTEREST

Counsel for Appellant Rambus Inc. certifies the following:

1. The full name of every party represented by me is: Rambus Inc.

2. The name of the real party in interest represented by me is: Rambus Inc.

3. All parent corporations and any publicly held companies that own 10 percent

or more of the stock of the party represented by me are: None

4. The names of all law firms and the partners or associates that appeared for

the party now represented by me in the trial court or agency or are expected

to appear in this Court are:

John M. Whealan

MOLOLAMKEN LLP:

Jeffrey A. Lamken

Michael G. Pattillo, Jr.

Benoit Quarmby

FINNEGAN, HENDERSON, FARABOW,

GARRETT & DUNNER, LLP:

Doris Johnson Hines

Jason E. Stach

J. Michael Jakes

Christine E. Lehman

Kathleen A. Daley

Naveen Modi

Luke J. McCammon

Michael J. McCabe

Brannon C. McKay

May 23, 2013 /s/ Jeffrey A. Lamken

Jeffrey A. Lamken


7/28/2019 cafit

4/84

ii

TABLE OF CONTENTS

Page

JURISDICTIONAL STATEMENT ............................................................... 1

ISSUES PRESENTED ................................................................................... 1

STATEMENT OF THE CASE ...................................................................... 2

STATEMENT OF FACTS............................................................................. 4

I. The Dally Patents ................................................................................. 4

A. High-Speed Digital Communication Systems AndFrequency-Dependent Signal Attenuation................................. 4

B. The Invention: On-Chip Digital Signal Equalization ............... 8

C. The Technologys Development And Commercialization ...... 13

II. The Decisions Below.......................................................................... 15

A. The ALJs Initial Determination.............................................. 15

1. Invalidity / Claim Construction ..................................... 15

2. Domestic Industry.......................................................... 19

B. The Commissions Final Determination.................................. 21

SUMMARY OF ARGUMENT.................................................................... 24

STANDARD OF REVIEW.......................................................................... 27

ARGUMENT................................................................................................ 28

I. The Dally Patents Claim Of A Pre-Emphasizing Transmitter

Circuit That Can Operate At An Output Frequency Of At Least

1 GHz Is Unmatched In The Prior Art ............................................. 28


7/28/2019 cafit

5/84

iii

A. No Prior Art Of Record Expressly Discloses The Dally

Patents Output Frequency Of At Least 1 GHz.................... 29

B. The Dally Patents Claimed Output Frequency Does Not

Equal the Data Rate Because The Data Rate Is Twice The

Frequency................................................................................. 31

1. The Claims Demonstrate That The Output Data Rate

Equals Twice The Claimed Output Frequency

(D=2F)............................................................................ 31

2. The Specification Supports Rambuss Construction

Of D=2F......................................................................... 34

3. Witness Testimony Confirms That The Output Data

Rate Is Twice The Claimed Output Frequency

(D=2F)............................................................................ 37

4. The ALJs Rationale For Rejecting That Construction

Does Not Withstand Scrutiny ........................................ 39

C. The ALJ Erred In Reading The Patents To Equate Output

Frequency (Cycles-Per-Second) With Data Rate (Bits-Per-

Second)..................................................................................... 43

II. Rambus Established A Domestic Industry For The Dally Patents .... 51

A. The Domestic Industry Requirement ....................................... 51

B. Rambus Established A Domestic Industry For The Dally

Patents ...................................................................................... 52

1. Rambuss Dally-Specific Investments ....................... 53

2. Rambuss Firm-Wide Investments In Licensing ....... 56

C. The ALJ Correctly Found That Rambus Established A

Domestic Industry Under Commission Precedent................... 56


7/28/2019 cafit

6/84

iv

D. The Commissions Reversal Of The ALJs Domestic

Industry Finding Defies The Evidence, Commission

Precedent, And Congresss Intent............................................ 59

1. The Commission Failed To Consider Rambuss

Dally-Specific Evidence............................................. 59

2. The Commissions Allocation Requirement Is An

Unjustified Reversal Of Agency Precedent................... 61

a. The Commission Reached The Opposite

Result, Based On Similar Evidence, In The

Related Rambus 661 Investigation...................... 62

b. Prior Commission Precedent Required Only

A Nexus Between The Patent And Licensing

Expenses, Not Allocation ................................ 65

3. The Commissions Change In Standard Undermines

Congresss Intent ........................................................... 67

CONCLUSION............................................................................................. 70

CONFIDENTIAL MATERIAL OMITTED

Material has been redacted in the Nonconfidential Brief for Appellant

Rambus Inc. This material is confidential business information pursuant to

19 C.F.R. 210.5 and the Protective Order entered by the United States

International Trade Commission on January 4, 2011, and amended on March

17, 2011. Redacted material on pages 2, 14, 20, 22, 23, 26, 55, 56, 57, 59,

60, and 62 contains confidential information regarding licensing terms,

revenues, and/or expenses. Redacted material on pages 2, 13, 22, 53, and 60

contains confidential information regarding a business acquisition.


7/28/2019 cafit

7/84

v

TABLE OF AUTHORITIES

Page(s)

CASES

Allentown Mack Sales & Serv., Inc. v. NLRB, 522 U.S. 359 (1998)............ 61

Alloc, Inc. v. ITC, 342 F.3d 1361 (Fed. Cir. 2003), cert. denied, 541

U.S. 1063 (2004)...................................................................................... 27

Bancorp Servs., LLC v. Hartford Life Ins. Co., 359 F.3d 1367 (Fed.

Cir. 2004) ................................................................................................. 36

Baran v. Med. Device Techs., Inc., 616 F.3d 1309 (Fed. Cir. 2010),

cert. denied, 131 S. Ct. 1607 (2011)........................................................ 36

CCS Fitness, Inc. v. Brunswick Corp., 288 F.3d 1359

(Fed. Cir. 2002)........................................................................................ 35

Changzhou Wujin Fine Chem. Factory Co., Ltd. v. United States,

701 F.3d 1367 (Fed. Cir. 2012) ............................................................... 27

Clifton Power Corp. v. FERC, 88 F.3d 1258 (D.C. Cir. 1996).................... 60

Dominion Res., Inc. v. United States, 681 F.3d 1313 (Fed. Cir. 2012)........ 61

El Rio Santa Cruz Neighborhood Health Ctr., Inc. v. U.S. Dept ofHealth & Human Servs., 396 F.3d 1265 (D.C. Cir. 2005)...................... 60

Honeywell Intl, Inc. v. ITC, 341 F.3d 1332 (Fed. Cir. 2003) ...................... 64

In re Sang-Su Lee, 277 F.3d 1338 (Fed. Cir. 2002) ..................................... 60

InterDigital Commcns, LLC v. ITC, 707 F.3d 1295

(Fed. Cir. 2013)............................................................................ 51, 52, 67

M.M.&P. Mar. Advancement, Training, Educ. & Safety Program v.Dept of Commerce, 729 F.2d 748 (Fed. Cir. 1984) ............................... 67

Microsoft Corp. v. i4i Ltd. Pship, 131 S. Ct. 2238 (2011).................... 30, 37

Modine Mfg. Co. v. ITC, 75 F.3d 1545 (Fed. Cir. 1996), abrogated

on other grounds,Festo Corp. v. Shoketsu Kinzoku Kogyo Kabu-

shiki Co., Ltd., 234 F.3d 558 (Fed. Cir. 2000) .................................. 30, 50


7/28/2019 cafit

8/84

vi

Motor Vehicle Mfrs. Assn v. State Farm Mut. Auto. Ins. Co., 463

U.S. 29 (1983).................................................................................... 66, 67

Net MoneyIN, Inc. v. VeriSign, Inc., 545 F.3d 1359 (Fed. Cir. 2008).......... 29

OSRAM Sylvania, Inc. v. Am. Induction Techs., Inc., 701 F.3d 698(Fed. Cir. 2012)........................................................................................ 29

Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc),

cert. denied, 546 U.S. 1170 (2006) ................................................... 31, 34

Retractable Techs., Inc. v. Becton, Dickinson & Co., 653 F.3d 1296

(Fed. Cir. 2011), cert. denied, 133 S. Ct. 833 (2013).............................. 39

Schucker v. FDIC, 401 F.3d 1347 (Fed. Cir. 2005) ......................... 62, 65, 67

SKF USA Inc. v. United States, 630 F.3d 1365 (Fed. Cir. 2011) ................. 62

Wang Labs., Inc. v. Am. Online, Inc., 197 F.3d 1377 (Fed. Cir. 1999)........ 50

STATUTES AND REGULATIONS

19 U.S.C. 1337 ....................................................................................passim

19 U.S.C. 1337(a) (1982) ........................................................................... 51

19 U.S.C. 1337(a)(1)(B)(i)......................................................................... 19

19 U.S.C. 1337(a)(2) ............................................................................ 19, 51

19 U.S.C. 1337(a)(3)(C)......................................................................passim

19 U.S.C. 1337(c)......................................................................................... 1

28 U.S.C. 1295(a)(6) .................................................................................... 1

35 U.S.C. 282 ............................................................................................. 30

19 C.F.R. 210.42(h)(2)............................................................................... 64

ADMINISTRATIVE AGENCY DECISIONS

Certain Integrated Circuits, Chipsets, and Products Containing Same

Including Televisions, Inv. No. 337-TA-786, Commn Op. (Oct.

10, 2012) .................................................................................................. 66


7/28/2019 cafit

9/84

vii

Certain Liquid Crystal Display Devices, Including Monitors, Tele-visions, and Modules, and Components Thereof, Inv. Nos. 337-

TA-749, -741, USITC Pub. 4383 (July 6, 2012)..................................... 66

Certain Multimedia Display and Navigation Devices and Systems,

Components Thereof, and Products Containing Same, Inv. No.337-TA-694, USITC Pub. 4292 (Aug. 8, 2011)...............................passim

ALJ Order No. 21, Certain Semiconductor Chips Having Synchronous

Dynamic Random Access Memory Controllers and ProductsContaining Same, Inv. No. 337-TA-661 ........................................... 63, 64

Certain Stringed Musical Instruments and Components Thereof, Inv.

No. 337-TA-586, USITC Pub. 4120 (May 16, 2008) ............................. 68

OTHERAUTHORITIES

Sinclair, The HarperCollins Dictionary of Electronics (1991) ................ 5, 32

USITC,Facts and Trends Regarding USITC Section 337

Investigations, http://www.usitc.gov/press_room/documents/

featured_news/sec337factsupdate.pdf (Apr. 15, 2013)........................... 69

Websters Third New Intl Dictionary (2002) .............................................. 41


7/28/2019 cafit

10/84

viii

STATEMENT OF RELATED CASES

Appellant Rambus Inc. is unaware of any other appeals taken in connection

with U.S. International Trade Commission Inv. No. 337-TA-753.

Rambus Inc. v. STMicroelectronics NV, et al., No. 3:10-cv-05449 (N.D.

Cal.), involves the same patents at issue in this appealU.S. Patent Nos. 7,602,857

and 7,715,494and may be affected by the Courts decision in this appeal.


7/28/2019 cafit

11/84

JURISDICTIONAL STATEMENT

This appeal arises from an investigation by the United States Inter-

national Trade Commission (the Commission). The Commission had

jurisdiction under 19 U.S.C. 1337. The Commission issued its final deter-

mination on July 31, 2012. Rambus timely filed a notice of appeal on Sep-

tember 21, 2012. This Court has jurisdiction under 28 U.S.C. 1295(a)(6)

and 19 U.S.C. 1337(c).

ISSUES PRESENTED

Although the case below encompassed myriad issues and produced

over 400 pages of rulings, this appeal involves just two issuesone issue of

claim construction, on which the validity issues turn, and another concerning

the domestic industry requirement under 19 U.S.C. 1337(a)(3)(C). The

issues presented are:

1. Does the claim phrase output frequency of at least 1 GHz in

the Dally patents correspond to

(i) 2 gigabits-per-second, where (a) the claim language and

electrical engineering principles indicate that two bits of data are transmitted

per Hertz; and (b) every time the specification discusses both a data signals

frequency in Hertz and its data rate in bits-per-second, it states that the data

rate is twice the frequency; or


7/28/2019 cafit

12/84

2

(ii) 1 gigabit-per-second, as the ALJ and Commission found?

2. Did the Commission err in reversing the ALJs finding of a

domestic industry in the Dally patents where, among other things,

(i) Rambus paid over to acquire certain assets of

an operating company for the purpose of commercializing the Dally patents,

which included a North Carolina research facility, a staff of engineers who

were designing products based on the Dally patents, and the rights to the

Dally patents from MIT;

(ii) Rambus employed licensing personnel who negotiated

numerous licenses covering Dally-based technology, yielding over

; and

(iii) Rambus paid MIT over in royalty payments

for those licenses?

STATEMENT OF THE CASE

Rambus filed a complaint with the Commission under Section 337 of

the Tariff Act of 1930, 19 U.S.C. 1337, alleging that several companies

had imported, sold for importation, and/or sold within the United States after

importation, semiconductor chips that infringe U.S. Patent Nos. 7,602,857

CONFIDENTIAL MATERIAL REDACTED PURSUANT TO PROTECTIVE ORDERCase: 12-1677 Document: 92-1 Page: 12 Filed: 05/23/2013

7/28/2019 cafit

13/84

3

and 7,715,494 (the Dally patents).1

The Commission instituted Investiga-

tion No. 337-TA-753, naming Broadcom Corporation, Freescale Semicon-

ductor, Inc., LSI Corporation, MediaTek Inc., nVidia Corporation, STMicro-

electronics N.V. and STMicroelectronics Inc., and 27 of their customers

(including Cisco Systems, Garmin International, and Seagate Technology)

as respondents.

Administrative Law Judge Essex held an evidentiary hearing and

issued an Initial Determination (ID). A70-453.2 Judge Essex found that a

domestic industry exists for the Dally patents. He found that the accused

products infringed the Dally patents. He ruled, however, that the asserted

Dally claims were anticipated or obvious.

The Commission affirmed in part and reversed in part. A1-69. It

affirmed Judge Essexs relevant infringement findings, claim-construction

rulings, and invalidity determinations. The Commission reversed Judge

Essexs domestic-industry ruling.

Rambus settled with Freescale and with supplier-respondents Broad-

com, MediaTek and nVidia, terminating them from the investigation along

1Rambus also asserted patents in the Barth patent family and another patent

in the Dally patent family. Those patents are not relevant to this appeal.2

A__ refers to the addendum to this brief. JA__ refers to the joint

appendix.


7/28/2019 cafit

14/84

4

with appropriate customers. After filing its notice of appeal, Rambus settled

with LSI. Only supplier-respondent STMicroelectronics (and its customers

Cisco, Garmin, and Seagate) remain as intervenors in this appeal.

STATEMENT OF FACTS

The Commission investigation below was a massive proceeding in-

volving myriad issues of claim construction, infringement, and validity for

61 asserted claims from five patents. This appeal is narrow, presenting only

two questions relating to two patents. The first is a validity issue that turns

on the Commissions construction of claims reciting a transmitter circuit that

pre-emphasizes the output signal at an output frequency of at least 1 GHz.

The second involves the Commissions new test for the domestic industry

requirement under 19 U.S.C. 1337(a)(3)(C), and its application here.

I. The Dally PatentsA. High-Speed Digital Communication Systems And

Frequency-Dependent Signal Attenuation

The invention at issue here dramatically increased the speed of digital

communication systems in consumer electronics, such as personal com-

puters, gaming systems, and mobile phones, JA9753it was a game

changer. JA4835-36. When data is transferred from one integrated circuit

chip to another, the transmitter on the sending chip uses an interface to

send data (in the form of a signal) across a physical channel (e.g., copper


7/28/2019 cafit

15/84

5

wire) to a receiver on the receiving chip. JA9753. Examples include send-

ing data from a PC to a monitor, or from memory to a graphics card.

The performance of digital systems is limited by the speed at which

one chip can reliably send information to another. A639, 1:25-27. For

years, one of the greatest obstacles to increasing the speed of chip-to-chip

signaling was frequency-dependent attenuation. Id., 1:34-37. As ex-

plained below, attenuation is a phenomenon where a signal sent by the

transmitterfor example, a digital 1 or 0degrades as it crosses the

transmission line so that it might not be read as a 1 or 0 when it reaches the

receiver. That degradation is most pronounced at higher frequencies.

In general, frequency is the rate at which a waveform action re-

peats, typically measured in HERTZ. Sinclair, The HarperCollins Dic-

tionary of Electronics 123 (1991) (JA17308). The figure below illustrates

this concept:


7/28/2019 cafit

16/84

6

JA17284 (annotated). Frequency, expressed in Hertz, is the rate at which the

signal completes one cycle, transitioning from the midline (A) to one high

level (B), to one low level (C), and back to the midline (D).

In digital systems, transmitted signals typically will have multiple

frequency components. Figure 2A from the Dally patents depicts a hypo-

thetical data signal as sent by the transmitter:

A630; A639, 2:26; JA9756. In the Dally patents, each high swing of the

signal above the dashed line represents a bit of data, a 1, and each swing

below the dashed line represents another bit, a 0. JA9755; A644, 11:27-

29. When the signal remains above or below the dashed line for a sustained

period, it represents a series of consecutive 1s or 0s. JA9755.

When the signal transitions from the midline to one high level, to one

low level, and back to the midline, that represents one cycle, as in the sinu-

soidal wave on p. 5. The rate at which the signal completes those cycles is

its frequency. When the waveform cycles quickly (e.g., 010), it is a high-


7/28/2019 cafit

17/84

7

frequency component of the signal. JA9756. By contrast, when the signal

maintains a single level for a longer period (e.g., 000 or 111), it is a low-

frequency component. Id.

Attenuation is the loss of strength as a signal crosses a physical line

between the transmitter and the receiver. JA9754. Attenuation creates prob-

lems for digital signaling. Receivers require a minimum swing between

the high and low signal levels that represent 1 and 0. JA9756-57. Attenua-

tion reduces the swing between signal levels, making it harder for the

receiver to interpret the signal as a 1 or 0. JA9756-58; A640, 3:58-67.

Figure 2B shows the data signal from Figure 2A as it appears at the receiver

after attenuation:

A630; A639, 2:26; JA9756-57. The dashed line represents the same mid-

point of the sent signal in Figure 2A, but attenuation lessened the signals

swing across the midpoint, reducing the receivers ability to correctly detect

the signal levels and the bit values they represent. A640, 3:58-67.


7/28/2019 cafit

18/84

8

As Figure 2B illustrates, the high-frequency components (the portions

that rapidly transition from high to low or low to high) are more susceptible

to attenuation than low-frequency components (which sustain a single signal

level for longer). JA9754. Because of attenuation, the high-frequency com-

ponents of the signal in Figure 2B may barely cross the midpoint. Atten-

uation thus is frequency-dependent. JA9754. Degradation increases with

signal speed and frequency. Id.

That phenomenon posed severe problems. For years, the processing

speed of semiconductor chips increased steadily, but the speed of signaling

between chips lagged. See A639, 1:25-31. Before the invention here,

frequency-dependent attenuation made most digital systems unsuited for

data with frequencies over 100 MHz. Id., 1:32-39. Chips could process data

at high speed, but attenuation prevented them from reliably transmitting data

to other chips at corresponding speeds.

B. The Invention: On-Chip Digital Signal EqualizationDr. William Dally was a professor at MIT when he solved the prob-

lem of frequency-dependent attenuation in chip-to-chip signaling. JA6043.

Dr. Dally invented an on-chip mechanism in the transmitter that mitigates

frequency-dependent attenuation using a form of signal equalization called

emphasis or pre-emphasis. JA9760.


7/28/2019 cafit

19/84

9

In the invention, the transmitter pre-emphasizes (increases the relative

magnitude of ) portions of the signal to offset the degrading effects of attenu-

ation; as a result, the signal that reaches the receiver better approximates the

signal that was intended. JA9760; A642, 8:48-55. Because high-frequency

components of the signal are more susceptible to attenuation than low-

frequency components, the transmitter pre-emphasizes high-frequency com-

ponents, using a larger amplitude than for low-frequency components. Con-

sequently, despite attenuation, the high-frequency parts of the signal can be

reliably detected by the receiver. JA6041; JA9761.

The invention uses bit history to ensure that high-frequency compo-

nents receive greater emphasis than low-frequency components. See A639,

1:61-67. The transmitter looks to whether the particular bit value is the

same as the preceding output bit signal (e.g.,00), in which case it is

identified as a low-frequency component of the signal, or whether the par-

ticular bit value is different from the preceding output bit signal (e.g.,

01), in which case it is identified as a high-frequency component. A642,

8:56-63. The transmitter uses that information to emphasize high frequen-

cy components of the output signal relative to low frequency components of

the output signal by assigning them different signal levels. A642, 8:52-


7/28/2019 cafit

20/84

10

55. The Dally patents thus claim a transmitter that utilizes bit history to

perform that selective pre-emphasis:

1. A component comprising:

a semiconductor chip;

a processor within the chip; and

a transmitter circuit within the chip, the transmitter circuit

being coupled to the processor to accept a digital input

signal including a plurality of digital values from the

processor, the transmitter circuit being operable to send an

output signal including a series of signal levels representing

the digital values and to emphasize high frequency

components of the output signal relative to low frequencycomponents of the output signalso that:

(i) an output bit signal of the output signal representing a

particular bit value has one signal levelwhen the bit value

is the same as a bit value represented by a predetermined

preceding output bit signal; and

(ii) the output bit signal representing the particular bit

value has another signal level when the bit value is

different from the bit value represented by the

predetermined preceding output bit signal.

A642, 8:45-63 (emphasis added).

By pre-emphasizing the signal at the transmitter based on its fre-

quency, Dr. Dally achieved a largely unattenuated signal at the receivera

1 sent by the transmitter had sufficient magnitude for the receiver to

interpret it as a 1. The following pair of diagrams from the Dally patents

illustrate the benefits of the invention.


7/28/2019 cafit

21/84

11

Absent pre-emphasis, the signal sent by the transmitter (Figure 2A,

A630) would be heavily attenuated by the time it reaches the receiver

(Figure 2B, A630). Indeed, the 1s and 0s transmitted at high frequency may

barely cross the midline:

By pre-emphasizing the high-frequency components of that same data

signal (Figure 7A, A634), the Dally transmitter produces signal swings that

are far enough above and below the midline to be reliably detected by the re-

ceiver despite attenuation (Figure 7B, A634). A641, 5:45-47; JA9761.

Dr. Dally did not invent signal pre-emphasis; techniques existed in

analog telecommunications systems. JA6042. His invention, however,

comprised a novel way of applying on-chip transmitter pre-emphasis to

chip-to-chip signaling in digital systems. Id. As Dr. Dally explained, he


7/28/2019 cafit

22/84

12

invented many particular techniques about how to make a

system that had originally been developed to work at kilohertz

rates [1,000 cycles-per-second] in telecommunications systems

to work at gigahertz rates [1,000,000,000 cycles-per-second] in

digital systems. It was a nontrivial movement of the technology

by . . . six orders of magnitude in frequency that required a dif-

ferent set of techniques in how to build the actual circuits.

Id. Working with Dr. John Poulton, a professor at the University of North

Carolina, Dr. Dally built a chip that embodied his inventions. JA6011.

As Dr. Poulton testified, Dr. Dallys inventions were a game

changerthe foundation of an industry. JA4835-36. They enabled un-

precedented speed for chip-to-chip data transmission, JA4835, multiples

of times faster than what people were using, JA4840. Dr. Poulton explain-

ed that, previously, you would have been lucky to get a . . . 400 megabit per

second signaling system working, whereas Dallys chip was performing

pre-emphasis and running at 4 gigabits per second. JA4839.

Dr. Dallys invention thus enabled chip-to-chip transmission of data

signals with much higher frequencies than previously possible. According-

ly, the Dally patents claim transmitters with the capability to transmit pre-

emphasized signals at very high frequencies:

2. The component as claimed in claim 1 wherein the transmit-

ter circuit is operable to send the output signal with an output

frequency of at least 1 GHzand a bandwidth greater than 100

MHz.


7/28/2019 cafit

23/84

13

857 patent, claims 2, 31, and 49, A642-44, 8:64-67, 10:64-67, 12:53-56;

494 patent, claim 3, A662, 9:5-8 (emphasis added). The validity issue in

this appeal turns on the proper construction of that limitation of an output

frequency of at least 1 GHz.

C. The Technologys Development And CommercializationDr. Dally assigned the patentsthe fruit of his academic researchto

MIT. JA8049-54; A628; A647. In 1999, Dr. Dally founded a company

called Chip2Chip to commercialize the patents. JA6003; JA6030-31.

Chip2Chip obtained an exclusive license from MIT, JA8055-78, and devel-

oped serializer/deserializer (SerDes) circuits for high-speed chip-to-chip

interfaces utilizing Dr. Dallys technology, JA6003.3

Chip2Chip, later

renamed Velio, JA6003, had a research facility in Chapel Hill, North

Carolina, where a team of engineers designed circuits, JA4796; JA6001.

In 2003, Rambus acquired many of Velios assets, paying over

for Velios research facility, the engineers in its circuit design team,

and its license from MIT for the Dally patents. JA8825; JA8869; JA8897;

JA8931-39; JA12628-29; JA4795; JA6001. Until at least 2009, Rambus

employed 15-20 engineers at that facility and continued developing SerDes

3A serializer/deserializer accepts parallel data inputs, serializes the data into

a single bit stream, and transmits it at high speed to another chip, where it

deserializes the bit stream again into parallel data.


7/28/2019 cafit

24/84

14

technology. JA4796-97. Rambus incorporated that work into cell designs

schematics for semiconductor circuitrythat practice the Dally patents.

JA3593-94; JA3607; JA12661-62. Rambus developed data sheets and

technical guides to implement those designs for semiconductor chip fabri-

cators. JA3607-08; JA8462-512; JA8513-59; JA8560-615; JA8616-73.

Rambus commercialized the Dally technology through licensing, typi-

cally as part of a larger SerDes portfolio. Rambus employed roughly 30 in-

dividuals who made presentations to and negotiated licenses with potential

customers. See JA12625-27; JA9346-49; JA12656-57; JA8324-94; JA8395-

461; JA8768; JA8764; JA8765; JA8766. Rambus has entered into six

SerDes technology license agreements covering products Rambus designed

that incorporate Dally technology. JA12641-46; JA8079-116; JA8150-57;

JA8158-208; JA8209-31; JA8232-58; JA8259-76. Those licenses have

yielded over in royalties to Rambus. JA12646-47. Rambus also

negotiated a separate patent license with Samsung covering the Dally pat-

ents. JA8277-302.

Under Rambuss license from MIT, Rambus must pay MIT

on received by Rambus. JA8124. Rambus

has paid MIT over under that agreement. JA12646; JA12630-

31.


7/28/2019 cafit

25/84

15

II. The Decisions BelowA. The ALJs Initial DeterminationAfter an evidentiary hearing, the ALJ (Judge Essex) issued his Initial

Determination (ID), finding no violation of Section 337. A70-453.

1. Invalidity / Claim ConstructionThe Dally output-frequency claims recite a transmitter circuit that not

only pre-emphasizes high-frequency components of the output signal rela-

tive to low frequency components, but also is operable to send the output

signal with an output frequency of at least 1 GHz. A642, 8:64-67. The

ALJ found that respondents infringe those claims. A153. But he found the

claims invalid as obvious and anticipated based on his construction of the

term output frequencyspecifically, based on his equating output fre-

quency (in Hertz) with output data rate (in bits-per-second). A226-29;

A268-69. That was pivotal: The Dally claims recite the signals output

frequency (F)the rate at which the signal can transition from the midline to

one high level, to one low level, and back to the midline. A642, 8:66;see

pp. 5-7, supra. The cited prior art, however, did not reference output

frequency. Instead, it quantified an output signal as a data rate (D) ex-

pressed in bits-per-second (b/s). See A227-28; A268.


7/28/2019 cafit

26/84

16

The ALJs validity analysis thus turned on the data rate in the prior art

that corresponded to the claimed output frequency of at least 1 GHz. The

ALJ found that the output data rate achievable by the relevant prior art

was 1.0625 Gb/s. A228; A268. Consequently, the ALJ ruled that the Dally

output-frequency claims would read on the prior art if they corresponded to

an output data rate encompassing 1.065 Gb/s.

The ALJ rejected Rambuss argument that the claimed output fre-

quency of at least 1 GHz corresponds to an output data rate of twice the

frequency, such that a frequency of 1 GHz corresponds to an output data rate

of at least 2Gb/s.4

In electrical engineering, one cycle corresponds to two

signal levelsi.e., the signal moving from the midpoint, to the high point, to

the low point, and then back to the midpoint. A99-100. Moreover, the

Dally patents claim a signaling type in which each high signal level

represents one bit of data (a 1), and each low signal level separately

represents another (a 0). See, e.g., A643-44, 10:39-41, 11:27-29; A100.

Because two signal levels, and thus two bits of data, are represented in a

single cycle or Hertz, Rambus urged, the output data rate equals twice the

4Before the Commission, Rambus urged that the output frequency equaled

one-half the data rate. Rambus takes the same position here, but believes it

simpler to express the relationship as data rate = 2 x output frequency, or

D=2F.


7/28/2019 cafit

27/84

17

output frequencyi.e., D=2F. A100. Concomitantly, an output frequency

of 1 GHz corresponds to a data rate of 2 Gb/s.

That was consistent with the fact that, when the specification address-

es both signal frequency and data rate, the data rate was twice the frequency

(D=2F). For example, the specification discusses an operating frequency of

2 GHz corresponding to a bit rate of 4 Gb/s. A640, 4:18-19 (emphasis

added). Similarly, the specification discloses that for a data signal of 4

Gb/s, the highest frequency of interest is 2 GHz. Id., 3:60-62. In both

instances, a signal with a data rate of 4 Gb/s is equated withafrequency of 2

GHz, i.e., D=2F.

Rejecting Rambuss reliance on electrical engineering principles, the

ALJ stated that Rambuss position requires a number of assumptions that

may or may not be true and certainly are not mandated without review of the

other intrinsic evidence. A101. Rambuss arguments might be valid in

light of the other intrinsic evidence. A102. The ALJ, however, never

analyzed Rambuss arguments in light of the other intrinsic evidence.

With respect to the specification, the ALJ did not dispute that it

repeatedly expresses a 2-to-1 ratio between data rate and frequency. But he

dismissed those portions as not particularly persuasive because they are

discussing the relationship between data rate and operating frequency, not


7/28/2019 cafit

28/84

18

output frequency, which do not appear to be the same thing, or at the very

least, Rambus has not presented evidence that they are the same thing.

A108-09. The ALJ cited no intrinsic evidence to the contrary. And he did

not explain what the specifications reference to operating frequency

means if it differs from output frequency.

Instead, the ALJ concluded that the claimed output frequency is

equal to the output data rate. A109. He based his decision primarily on Fig-

ures 1 and 12 of the Dally patents, both of which reference 400 MHz with

regard to parallel data inputs (equivalent to a 4 GHz serial input), and show a

4 Gb/s serial data output. A106-07. The ALJ concluded that those figures

both appear to use a 1:1 ratio between bits-per-second and Hertz. A107.

The ALJ disputed that the references to 400 MHz address not the data

signals frequency, but the clock operating on the signal(which operates at

twice the highest frequency of the data signal). See A108. The specifica-

tions discussion of Figure 1 distinguishes between the data signal and the

400 MHz clocks operating on that signal: The receiver accepts the signal

and its own 400 MHz clock. A640, 3:24-25 (emphasis added). The

receiver also generates 4 GHz [i.e., the 1-bit serial equivalent of 10-bit

parallel at 400 MHz] timing signals aligned to the received data, samples the

noisy signal, decodes the signal, and produces synchronous 8-bit data out.


7/28/2019 cafit

29/84

19

Id., 3:26-28 (emphasis added). The ALJ nevertheless stated, without

citation, that Rambus is simply incorrect that these examples are merely

referring to clock rate. A108.

Having equated the claimed output frequency in Hertz with the

output data rate in bits-per-second, the ALJ concluded that the Dally patents

recited output frequency of at least 1 GHz corresponds to a data rate of at

least 1 Gb/s. Because prior art allegedly could achieve an output data rate of

1.0625 Gb/s, the ALJ held that the claims were anticipated or obvious.

A228-29; A268-69.

2. Domestic IndustrySection 337 makes it unlawful to import, or to sell after importation,

articles that infringe a valid and enforceable United States patent. 19

U.S.C. 1337(a)(1)(B)(i). That prohibition is subject to the domestic in-

dustry requirement, i.e., proof that an industry in the United States, re-

lating to the articles protected by the patent, . . . exists or is in the process of

being established. 19 U.S.C. 1337(a)(2).

Applying Commission precedent, Judge Essex concluded that the

evidence shows that a licensing-based domestic industry . . . exists as to the

asserted Dally Patents. A434. Judge Essex found that Rambus had made

substantial investments specifically related to the Dally patents. Rambus


7/28/2019 cafit

30/84

20

had paid over in royalties to MIT for licensees use of those

patents. A439; A435. Judge Essex also found that was

attributable to royalties from technology license agreements which license

technology related to the Dally patents; under Commission precedent, that

was further evidence of significant investments. A436.

Judge Essex also found that Rambus had made substantial firm-wide

investments in licensing. It was undisputed that Rambus ha[d] over 30 em-

ployees involved in its U.S. based licensing practice at an expense of over

from 2006 through the first half of 2010. A439. The Com-

mission, in a prior Section 337 investigation relating to the Barth patents,

had found a domestic industry based on similar firm-wide evidence, and had

stood by that finding before this Court. A423 n.40;seepp. 62-65, infra.

Judge Essex rejected respondents contention that Rambuss firm-

wide investments were not specific to the Dally patents. The Commis-

sion does not require . . . strict mathematical proof of the investment in li-

censing certain patents among others in a portfolio. A433 (quotation marks

omitted). Rather, the complainant need only show a nexus between the

licensing activity and the asserted patent. A421. When the asserted patent

is part of a patent portfolio, and the licensing activities relate to the portfolio

as a whole, the Commission requires that the facts be examined to determine


7/28/2019 cafit

31/84

21

the strength of the nexus between the asserted patent and the licensing

activities. Id.

Examining that nexus underCertain Multimedia Display and Naviga-

tion Devices and Systems, Components Thereof, and Products Containing

Same, Inv. No. 337-TA-694, USITC Pub. 4292 (Aug. 8, 2011) (final) (Nav-

igation Devices), Judge Essex considered the size of the licensed portfolio;

the relative value contributed by the Dally patents; the prominence of the

Dally patents in licensing negotiations; and the scope of the technology

covered by the portfolio compared to the scope of the Dally patents. A421-

22; A435-38. Judge Essex found that those factors demonstrated a strong

nexus between the Dally patents and Rambuss licensing activities; he there-

fore concluded that Rambuss firm-wide expenses could be considered when

evaluating Rambuss investment in licensing the Dally patents. A439. The

evidence, he stated, shows that Rambuss investments in licensing the

asserted Dally Patents were and continue to be substantial. A439. Judge

Essex ruled that the domestic industry requirements have been satisfied.

A418.

B. The Commissions Final DeterminationOn review, the Commission affirmed the construction of output fre-

quency as equaling output data rate, adopting the ALJs analysis. A11-12.


7/28/2019 cafit

32/84

22

The Commission, however, reversed Judge Essexs finding that Rambus had

established a domestic industry for the Dally patents. A44-51.

The Commission stated that what is wanting . . . is evidence specific-

ally demonstrating investment made in the licenses upon which Rambus

relies. A47 n.19. The Commission, however, never mentioned Rambuss

specific investments in the Dally patents. It did not address:

the in royalties Rambus paid to MIT for licenseesuse of those patents, which the ALJ had cited, A435; A439;

the Rambus paid to acquire a research facility, therights to the Dally patents from MIT, and a staff of engineers at the

facility who were designing products based on the patents;

the engineers Rambus employed to develop Dally-based SerDestechnology; or

Rambuss efforts to negotiate licenses covering the Dally patents.The Commission acknowledged that Rambus had received over

in royalties for licenses encompass[ing] the Dally patents, A46

and that, under Commission precedent, royalties are circumstantial evi-

dence that an investment was made, A48but it gave that evidence no

weight. It also ignored Rambuss license to Samsung, and that Rambus paid

MIT in royalties for that license alone. A436.

The Commission instead focused solely on Rambuss firm-wide ex-

penditures. The Commission acknowledged that, with respect to that firm-


7/28/2019 cafit

33/84

23

wide evidence, Rambus was notrequired under Commission precedent to

provide a precise allocation of its licensing investments on a patent-by-

patent basis. A49 (emphasis added). But the Commission ruled that Ram-

buss proof failed because the Commission [did] not know what portion of

the in total licensing expenditures incurred by Rambus, or what

portion of time or expenses incurred by the 30 Rambus employees, might be

allocated to the Dally . . . patents. Id. Without proof to support such an

allocation, the Commission claimed, it was unable to find grounds upon

which these unknown investments could be considered substantial. A50.

The Commission did not dispute that, under its decision inNavigation

Devices,Rambus ha[d] demonstrated a nexus between the . . . Dally patents

and the portfolios that include them, A46 n.19, or that, as the ALJ found, a

nexus exists between the Dally patents and Rambuss licensing activities,

A439. The Commission did not dispute that, three years before, it had found

that Rambus established a domestic industry for the Barth patents based on

the same type of firm-wide licensing expenses. A50. The Commission

nevertheless concluded that Rambus did not provide sufficient evidence of

its investments in licenses specific to the Dally patents to establish a do-

mestic industry. A51.


7/28/2019 cafit

34/84

24

SUMMARY OF ARGUMENT

I. The Dally claims at issue recite a chip with a transmitter that

pre-emphasizes output signals and is operable to send the output signal with

an output frequency of at least 1 GHz. A642, 8:64-67 (emphasis added). It

is undisputed that no prior art in this case expressly discloses a transmitter

that sends a pre-emphasized output signal with an output frequency of at

least 1 GHz. Indeed, all of the cited prior art discloses signals in terms of a

data rate expressed in bits-per-second, not frequency expressed in Hertz. In

finding the Dally patents limitation of an output frequency of at least 1

GHz anticipated and obvious, the Commission relied upon prior art that

discloses an output signal with a data rate of approximately 1 Gb/s. For that

finding to stand, it must be shown that the data rate equals the claimed

frequency (D=F).

The Dally patents, however, disclose something different. Frequency

expressed in Hertz means cycles per second. In electrical engineering, a

single cycle includes two different signal levelsone high and one low:

The signal begins at the midpoint, moves to the high point, moves to the low

point, and then finishes at the midpoint where it started. As set forth in the

patent, those two signal levels correspond to twobits of data, not one. Thus,

a frequency of 1 GHz (1 billion cycles-per-second) corresponds to 2 Gb/s (2


7/28/2019 cafit

35/84

25

billion bits-per-second), not 1 Gb/s. The specification makes that clearer

still. Every time the specification discusses the data signals frequency in

terms of a data rate, it states that the data rate is twice the frequency (D=2F).

Thus, the claimed output frequency of 1 GHz equals an output data rate of 2

Gb/s. That data rate was not achieved by the prior art.

The ALJ relied on figures in the patents that, in his view, appear to

use a 1:1 ratio between output frequency and output data rate. But the ALJ

mistook the clock rate for the output frequency. Only Rambuss construc-

tion reconciles the figures in the Dally patents with the language of the

specification.

II. After an 8-day hearing, the ALJ determined that Rambus had

established a domestic industry in the Dally patents. The domestic indus-

try requirement is satisfied if the complainant has made a substantial

investment in the exploitation of the patents, including investments in

engineering, research and development, or licensing. 19 U.S.C.

1337(a)(3)(C). Rambus presented investments specific to the Dally pat-

ents. Rambus invested millions of dollars to acquire and maintain research

facilities and a staff of engineers for designing chips that utilized Dally

technology, and to acquire the exclusive rights to the Dally patents from

MIT. Rambus also invested in negotiating licenses for Dally-based tech-


7/28/2019 cafit

36/84

26

nology. And Rambus paid MIT over in royalties when Rambus

successfully licensed its Dally-based technologies.

Rambus also showed substantial firm-wide investments in licensing.

It employed over 30 individuals in its licensing department and spent over

on licensing operations. As required under Commission prece-

dent regarding portfolio licensing, Rambus also demonstrated a nexus be-

tween those firm-wide licensing activities and the Dally patents. The ALJ

thus properly found that Rambus established a domestic industry.

In reversing the ALJs finding, the Commission never addressed

Rambuss Dally-specific investments. That alone is grounds for reversal.

The Commission also refused to accept Rambuss evidence of firm-wide

licensing expenses as proof of substantial investment in the Dally patents.

The Commission did so even though, in a related Section 337 investigation

three years earlier, it had found that Rambus established a domestic industry

based on the same type of firm-wide evidence. The Commission simply

changed its rules and, with no notice or explanation of its reasoning,

declared that its former nexus standard was no longer sufficient; instead,

the complainant now must present evidence showing how its firm-wide

expenditures can be allocated to the asserted patents. That is arbitrary and

capricious agency action. It also contravenes Congresss intent. Congress


7/28/2019 cafit

37/84

27

amended Section 337s domestic industry requirement to make clear that

universities and companies that license, rather than manufacture, their

patented technology may enforce their rights before the Commission. Yet

the Commissions new allocation requirement would bar the Com-

missions doors to anyone who licenses its patents as part of a portfolio.

Entities that license patentswhether universities, small research firms, or

large technology companies that hold tens of thousands of patentssimply

do not track employee time and expenses in terms of particular patents and

cannot provide the proof of allocation the Commission now requires.

STANDARD OF REVIEW

This Court reviews the Commissions legal determinations, including

claim construction, de novo. Alloc, Inc. v. ITC, 342 F.3d 1361, 1367 (Fed.

Cir. 2003), cert. denied, 541 U.S. 1063 (2004). It reviews the Commissions

factual determinations for substantial evidence, but the Commissions

reasoning . . . is reviewed under the arbitrary and capricious (or contrary to

law) standard. Changzhou Wujin Fine Chem. Factory Co. v. United States,

701 F.3d 1367, 1377 (Fed. Cir. 2012).


7/28/2019 cafit

38/84

28

ARGUMENT

I. The Dally Patents Claim Of A Pre-Emphasizing TransmitterCircuit That Can Operate At An Output Frequency Of At Least

1 GHz Is Unmatched In The Prior Art

Before Dr. Dallys invention, the speed of chip-to-chip signaling was

greatly restricted by frequency-dependent attenuationthe degradation of

high-frequency components of a data signal as they cross a physical line

between the transmitter and receiver, which makes those signals difficult to

interpret. See pp. 7-8, supra. Dally solved that problem by inventing a

digital transmitter that enabled unprecedented new speeds, JA4835; a

game chang[ing] advance, id. Claim 1 of the Dally 857 patent recites a

transmitter that uses bit history to identify high-frequency components of the

signal and emphasize[s] those high frequency components of the output

signal relative to low frequency components of the output signal, offsetting

the effects of frequency-dependent attenuation. A642, 8:48-55.

The representative claim of the Dally patents at issue here, which de-

pends on Claim 1, adds a frequency-based limitation in view of the inven-

tions ability to pre-emphasize at an extraordinary rate. For example, Claim

2 of the 857 patent recites a speed limitation such that the transmitter

circuit is operable to send the output signal with an output frequency of at


7/28/2019 cafit

39/84

29

least 1 GHz and a bandwidth greater than 100 MHz. A642, 8:64-67

(emphasis added).

The validity issue in this appeal turns on the proper construction of

that output-frequency limitation. As explained below, the output frequency

of at least 1 GHz claimed in the Dally patents corresponds to a data rate of

at least 2 Gb/s, as each cycle transmits two bits. The Commission erred in

holding that the limitation was met by a data rate of half that speed.

A. No Prior Art Of Record Expressly Discloses The DallyPatents Output Frequency Of At Least 1 GHz

Anticipation exists only where a reference discloses within the four

corners of the document . . . all of the limitations of the challenged claim.

Net MoneyIN, Inc. v. VeriSign, Inc., 545 F.3d 1359, 1371 (Fed. Cir. 2008).

Obviousness similarly requires an analysis of the scope and content of the

prior art to determine whether a skilled artisan would have been motivated

to combine the teaching of the prior art references to achieve the claimed

invention, and whether he would have had a reasonable expectation of

success in doing so. OSRAM Sylvania, Inc. v. Am. Induction Techs., Inc.,

701 F.3d 698, 706 (Fed. Cir. 2012).

In this case, respondents asserted that the Dally claims are anticipated

and obvious because prior-art references disclosed digital transmitters that

performed the signal pre-emphasis claimed in the Dally patents and were


7/28/2019 cafit

40/84

30

operable to send the output signal with an output frequency of at least 1

GHz, as recited in the Dally output-frequency claims. A642, 8:64-67. But

the respondents did not identify a single prior-art reference purporting to

disclose such a transmitter with an output frequency of at least 1 GHz.

Instead, respondents relied on prior-art references that described their output

data signal in terms of a data rate expressed in bits-per-second.

Patents, however, are presumed valid, 35 U.S.C. 282, and anyone

asserting invalidity must prove it by clear and convincing evidence, Micro-

soft Corp. v. i4i Ltd. Pship, 131 S. Ct. 2238, 2242 (2011). Respondents fell

short of meeting that burden. To the contrary, their effort to equate the limi-

tation of 1 GHz (1 billion cycles per second) with prior art operating at 1

Gb/s (1 billion bits per second) defies the text of the claim and the specifica-

tion, which repeatedly expresses the data rate as twice the frequency. It con-

travenes basic electrical engineering principles as embodied in the operation

of the patents, under which (a) two signal levels may be transmitted per

cycle (or Hertz); (b) each signal level corresponds to one bit; and

consequently (c) one cycle corresponds to two bits. And it contravenes the

rule that claims should when reasonably possible be interpreted so as to

preserve their validity. Modine Mfg. Co. v. ITC, 75 F.3d 1545, 1557 (Fed.


7/28/2019 cafit

41/84

31

Cir. 1996), abrogated on other grounds, Festo Corp. v. Shoketsu Kinzoku

Kogyo Kabushiki Co., Ltd., 234 F.3d 558 (Fed. Cir. 2000).

B. The Dally Patents Claimed Output Frequency Does NotEqual the Data Rate Because The Data Rate Is Twice The

Frequency

The ALJ ruled that the highest output data rate in the prior art is

1.0625 Gb/s. See A228 (discussing Widmer and Ewen); A268 (discussing

the SL500 art). The invalidity challenges to the output-frequency claims

thus hinged entirely on whether the claimed output frequency of at least 1

GHz exceeds an output data rate of 1.0625 Gb/s. Because it doesthe

claimed output frequency of 1 GHz corresponds to a data rate of 2 Gb/sthe

prior art does not satisfy the claim limitation. The Commissions invalidity

findings must therefore be reversed.

1. The Claims Demonstrate That The Output Data RateEquals Twice The Claimed Output Frequency (D=2F)

[T]he words of a claim are generally given their ordinary and cus-

tomary meaning, as a person of ordinary skill in the art in question at the

time of the invention would have understood them. Phillips v. AWH Corp.,

415 F.3d 1303, 1312-13 (Fed. Cir. 2005) (en banc) (quotation marks omit-

ted), cert. denied, 546 U.S. 1170 (2006). Here, the text of the Dally patents

claims, and ordinary electrical engineering principles, make clear that the


7/28/2019 cafit

42/84

32

output data rate, in gigabits-per-second, is twice the recited output

frequency, in gigahertz. They demonstrate that:

(a) Two signal levels are transmitted per Hertz (the unit by which

the Dally patents express frequency);

(b) Each signal level represents one bit of data in the Dally patents;

and consequently

(c) Two bits are transferred per cycle or Hertz.

Thus, at least 2 Gb/s of data corresponds to an output frequency of at least

1 GHzi.e., D=2F.

The Dally patents recite an output frequency of at least 1 GHz.

A642, 8:66. As a dictionary definition cited by the Commission Staff makes

clear, frequency, measured in HERTZ, is the rate at which a waveform

action repeatsi.e., the rate at which the signal transitions from the midline

to one high level, down to one low level, and back up to the midline. Sin-

clair, The HarperCollins Dictionary of Electronics 123 (1991) (JA17308).

Thus, one Hertz corresponds to one full high-low transition, or one cycle,

per second. JA3786-87.

The text of Claim 1 (on which Claim 2 depends) demonstrates that

each of the two signal levels constituting a cycle corresponds to one bit of

dataand that, as a result, one full cycle corresponds to two bits. In

describing how pre-emphasis operates, Claim 1 specifies that an output


7/28/2019 cafit

43/84

33

signal representing a particular bit value [i.e., a 1 or 0] has one levelwhen

its value is the same as a preceding bit value, and it is given another signal

level when it differs from the preceding value. A642, 8:56-63 (emphasis

added). That the signal for a particular bit value has one level forecloses

the notion that it takes a full cycle to represent a single bit. A full cycle has

two signal levels (one high and one low). It simply makes no sense to refer

to a particular bit value having one signal level if it in fact is

communicated by the two different signal levels that together make up a

cycle. If the inventor had meant to communicate otherwise, he would have

referred to the signal representing the bit value as having a set of signal

levels, or he would have referred to the bit value as having signal levels of a

particular magnitude. That he did not speaks volumes.

Other claims likewise demonstrate that, in the Dally patents, two

signal levels, comprising a single cycle, represent two bits of data. For

example, Claims 24 and 35 of the 857 patent recite that each digital value

in the digital input signal [1 or 0] is represented by one of the signal

levels in the output signal. A643-44, 10:39-41, 11:27-29 (emphasis added).

Thus, because each signal level represents one bit in the Dally patentsthe

high signal level representing a 1 and the low signal level representing a

0, JA9755-56one high-low cycle corresponds to two bits of data. In


7/28/2019 cafit

44/84

34

terms of frequency (cycles-per-second) and data rate (bits-per-second), that

means that one cycle per second corresponds to two bits per second. Thus, 1

GHz corresponds to 2 Gb/s.

2. The Specification Supports Rambuss Construction OfD=2F

The specification makes that clearer still. Usually, [the specification]

is dispositive; it is the single best guide to the meaning of a disputed term.

Phillips, 415 F.3d at 1315 (quotation marks omitted). Here, every time the

specification expressly correlates the frequency of the output signal with a

data rate, it confirms that the output data rate equals twice the output fre-

quency (D=2F)conversely, that one cycle corresponds to two bits.

First, the specification discusses Figures 2A and 2B, which illustrate

the effects of frequency-dependent attenuation on a hypothetical data signal.

A640, 3:58-62. The specification describes the data signal in terms of a par-

ticulardata rate4 Gb/s. Id., 3:60. It then describes the highest fre-

quency of interest in that signal as 2 GHz. Id., 3:62 (emphasis added).

That discussion expressly confirms that the rate of data being output in that

signal (4 Gb/s) is equal to twice the frequency of the data signal (2 GHz);

i.e., D=2F.

Second, the discussion of Figures 3A and 3B makes the same correla-

tion. Those figures demonstrate the phenomenon of frequency-dependent


7/28/2019 cafit

45/84

35

attenuation by charting resistance-per-meter of wire and attenuation-per-

meter of wire as a function of the frequency of the data signal. A640, 4:11-

19. The specification explains that the example in those figures is at our

operating frequency of 2 GHz corresponding to a bit rate of 4 Gb/s. Id.,

4:18-19 (emphasis added). That confirms that, in the Dally patents, the

output data rate (in that example, 4 Gb/s) is twice the output frequency (2

GHz); again, D=2F.

Those statements in the specification should be dispositive. See CCS

Fitness, Inc. v. Brunswick Corp., 288 F.3d 1359, 1366 (Fed. Cir. 2002) (pat-

entee may act[ ] as his own lexicographer). But the ALJ rejected them

without meaningful analysis. The ALJ dismissed the statements as not par-

ticularly persuasive because they are discussing the relationship between

data rate and operating frequency, not output frequency, which do not appear

to be the same thing, or at the very least, Rambus has not presented evidence

that they are the same thing. A108-09.

With respect to Figures 2A and 2B, the ALJ simply erred. The

accompanying discussion refers to the highest frequency of interestnot

operating frequency, as the ALJ supposed. A640, 3:62. And the fre-

quency of interest discussed is the frequency of a component[ ] of the

[data] signal outputby a transmitter to a receiver across a transmission


7/28/2019 cafit

46/84

36

line. Id., 3:50-62, 4:1-9. It thus refers to the same output frequency of the

output signal sent by a transmitter circuit recited in the asserted claims.

A642, 8:64-66.

The specification did refer to an operating frequency in connection

with Figures 3A and 3B. But even there, the specification used that term

interchangeably with the claim term output frequency. Another patent in

the Dally family incorporated by reference in the 857 patentU.S. Patent

No. 7,099,404makes that clear. The 404 patent shares a specification

with the asserted Dally patents and likewise references an operating fre-

quency of 2 GHz corresponding to a bit rate of 4 Gb/s. JA17372, 4:24-25.

Claim 1 of the 404 patent, however, recites an output operating frequency

of at least 1 GHz, JA17374, 8:57-61 (emphasis added), showing that the

Dally patents regularly use the phrases operating frequency, output fre-

quency, and output operating frequency to refer to the same concept. Cf.

Baran v. Med. Device Techs., Inc., 616 F.3d 1309, 1316 (Fed. Cir. 2010)

(patentee may use different terms interchangeably to refer to the same

concept), cert. denied, 131 S. Ct. 1607 (2011);Bancorp Servs., LLC v. Hart-

ford Life Ins. Co., 359 F.3d 1367, 1373 (Fed. Cir. 2004) (similar). The

ALJs conjecture that the references to frequency in those parts of the

specification do not mean output frequency is unfounded.


7/28/2019 cafit

47/84

37

The ALJ in any event inverted the burden. It was respondents burden

to prove invalidity by clear and convincing evidence, i4i, 131 S. Ct. at 2242,

not Rambuss burden to present[ ] evidence demonstrating validity, A108-

09. Any disputes should have been resolved by the specifications plain

termsand any residual doubts resolved in favor of a construction that

preserves the claims validity.

3. Witness Testimony Confirms That The Output DataRate Is Twice The Claimed Output Frequency (D=2F)

Respondents counsel and witnesses confirmed Rambuss construc-

tion. Respondents counsel illustrated the nature of frequency, and that one

cycle encompasses two signal levels, with the following exhibit:

JA17284. Respondents then asked Rambuss expert, Dr. Singer, to confirm

that exhibits accuracy:


7/28/2019 cafit

48/84

38

Q: [I]n this example . . . the frequency is one cycle per

second or one Hertz. Did I illustrate that right?

A: Yes, you did.

. . .

Q: So when you are saying the frequency, how many of

those high/lows are there within a given period of time,

right?

A: Right, within a given period of time, one Hertz corres-

ponds to one high [signal level] and one low [signallevel].

JA3786-87 (emphasis added).

Respondents own witnesses agreed that, consistent with electrical

engineering principles, 2 bits of data are transferred for every Hertz. One of

respondents engineers testified:

Q: What do you see as the distinction between a link run-

ning at 800 megabits per second versus 800 megahertz, ifthere is one?

A: Megabit is the data sent per second. Megahertz is the

cycles per second.

Q: Do those happen to be the same in the links you were

working on?

A: No.

Q: And why not?

A: It requires . . .two bits of data to make a full cycle.


7/28/2019 cafit

49/84

39

JA6320-21 (emphasis added). Similarly, Dr. PoultonDr. Dallys colla-

borator and now an nVidia scientistconfirmed that [t]he maximum fre-

quency in a four gigabit data stream is two gigahertz. JA6287 (emphasis

added). As each of those witnesses understood, a cycle includes two signal

levels; each signal level corresponds to one bit; so each cycle per second

corresponds with two bits per second.

4. The ALJs Rationale For Rejecting That ConstructionDoes Not Withstand Scrutiny

The ALJ rejected all of that proof. With respect to the nature of a

cycle and the fact that it comprises two different signal levels, the ALJ

dismissed that as assumptions that may or may not be true and certainly are

not mandated without review of the other intrinsic evidence. A101. But

they are fundamental electrical engineering principlesconfirmed by

respondents exhibits and witnessesnot assumptions. And the ALJ never

performed a review of the other intrinsic evidence to confirm or refute

them. That alone renders the ALJs reasoning arbitrary and unsupportable.

Retractable Techs., Inc. v. Becton, Dickinson & Co., 653 F.3d 1296, 1305

(Fed. Cir. 2011), cert. denied, 133 S. Ct. 833 (2013).

The ALJ faulted Rambus for assum[ing] that the signal is a simple

sinusoidal wave that transitions from high to low and back again. A101.

The ALJrelying on the following (incorrectly) annotated version of Figure


7/28/2019 cafit

50/84

40

2Anoted that data signals will not always simply alternate between high

and low signal levels (e.g., 101010), but will sometimes maintain a high

signal level (e.g., 111111) or low signal level (e.g., 000000) for a period of

time:

A102.

That misses the point. Data patterns are infinitely variable and will

contain both high-frequency components that rapidly transition between 1

and 0, and low-frequency components that transmit multiple consecutive 1s

or 0s. See pp. 5-8,supra. Because of that variability, it makes no sense to

analyze the output-frequency claims in terms of one particular hypothetical

data signal. Rather, it is the highest frequency achievablethat is, how fast

the signal is capable of transitioning from 1 to 0 or from 0 to 1that

matters. That is why the claims recite transmitters that are operable to send


7/28/2019 cafit

51/84

41

the output signal with an output frequency of at least 1 GHz. A642, 8:65-

66 (emphasis added). The word operable indicates that the transmitter is

capable of transmitting a pre-emphasized signal at a 1 GHz frequency

transitioning from high to low and back 109

times per secondwhile recog-

nizing that the device will generally operate at a lower frequency with fewer

transitions per second when it transmits consecutive 1s or 0s. See Websters

Third New Intl Dictionary 1580 (2002) (defining operable as fit or

possible to use).

Another figure invoked by the ALJ illustrates the point. The ALJ

cited the following diagramwhich is notfrom the Dally patents, but was

produced by the Commission Stafffor the proposition that a one-Hertz

signal could also be generated using 4 bits:

A101. That particular signal represents a 1100 data pattern with a single

high level and a single low level, corresponding to one Hertz, that transmits

4 bits of data. But the device is capable oftransmitting a 1010 data pattern


7/28/2019 cafit

52/84

42

in the same time span, in which case 4 bits are transmitted in 2 cycles per

second or 2 Hertz:

Because the actual data pattern (1100 or 1010) can vary, what matters

is the maximum frequency in the example, 2 Hertz for 4 bits. Indeed, if one

accepted the ALJs argumentthat a one-Hertz signal could also be gener-

ated using 4 bits, A101the figure he cited would represent a data rate that

was four times the output frequency. The claimed output frequency of at

least 1 GHz thus would yield a 4 Gb/s data rate in the example (D=4F)

placing the Dally patents even further beyond the prior art.

Finally, the ALJ questioned whether, in the Dally patents, each signal

level represents one bit of data. Referencing the figure belowagain, not

found in the Dally patentsthe ALJ stated that the evidence at the hearing

established that in other coding schemes that exist within the art, such as

Manchester or bi-phase coding, a single bit is represented by two signal

levels, a low followed by a high. A102.


7/28/2019 cafit

53/84

43

A103. While other coding schemes exist within the art, the ALJ cited

nothing to suggest the Dally patents utilize such schemes. To the contrary,

in the Dally claims, each digital value in the digital input signala 1 or

0is represented by one of the signal levels in the output signal, not two

signal levels as in the ALJs diagram or Manchester encoding. A643-44,

10:39-41, 11:27-29 (emphasis added). That is clear from the claims at issue

here. Seepp. 31-34,supra. The ALJ offered no answer.

C. The ALJ Erred In Reading The Patents To Equate OutputFrequency (Cycles-Per-Second) With Data Rate (Bits-Per-

Second)

Invoking two figures and one passage in the Dally specification, the

ALJ concluded that the term output frequency should be construed to

equal the output data rate. A99; A11-12 (affirming without discussion). But

those examples do not support the ALJs conclusion. The ALJ never ex-

plained why his view of those examples should receive greater weight than


7/28/2019 cafit

54/84

44

the portions of the specification directly addressing the correlation between

the data signals frequency and data rate, which support a 1-to-2 ratio. See

pp. 34-37,supra.

Figures 1 and 12 of the Dally patents both reference (a) parallel data

inputs accepted at 400 MHz, which the ALJ found to be equivalent to a 4

GHz serial input, and (b) a 4 Gb/s serial data stream. Based on that, the ALJ

found that the figures appear to use a 1:1 ratio between bits-per-second and

Hertz. A107. The ALJ, however, failed to heed the specifications explan-

ation of those figures.

Figure 1 shows:

A630. The specification explains that the transmitter module 22 in Figure 1

accepts 8-bit parallel data at 400 MHz, and code[s it] into 10 bits before

outputting it serially. A640, 3:16-19. Without further referencing the speci-

fications text, the ALJ assumed that the 400 MHz in the figure (repro-


7/28/2019 cafit

55/84

45

duced in red) refers to the frequency of the input data signals at the trans-

mitter. The ALJ then extrapolated that 8-bit parallel data signals at a fre-

quency of 400 MHz would be the equivalent of a 1-bit serial data signal at a

frequency of 3.2 GHz (8 x 400 MHz = 1 x 3.2 GHz). A106. The ALJ next

assumed that transmitter module 22 is processing and outputting the equiva-

lent of 10-bit parallel data at a signal frequency of 400 MHz, or the equiva-

lent of a 4 GHz 1-bit serial signal. A106. Because transmitter module 22s

serial output is labeled as 4 Gb/s in the figure (reproduced in green), the ALJ

assumed that 4 GHz must equal 4 Gb/s, and thus that the claimed output

frequency equals the output data rate. A107.

The ALJ erred at the outset. The specification shows that Figure 1s

references to 400 MHz are not to the frequency of the data signals.

Rather, 400 MHz is the frequency of the clocks used to process the input

data signals. The specification does not say a 400 MHz data signalis sent to

the transmitter module; it says that the module acceptsi.e., processes

the data at 400 MHz. A640, 3:16-17 (emphasis added). The discussion of

the receiver is clearer still: The receiver accepts the signal andits own

400 MHz clock. Id., 3:24-25 (emphasis added). The receiver also gener-

ates 4 GHz [i.e., the 1-bit serial equivalent of 10-bit parallel at 400 MHz]

timing signals aligned to the received data, samples the noisy signal, decodes


7/28/2019 cafit

56/84

46

the signal, and produces synchronous 8-bit data out. Id., 3:26-28 (emphasis

added).

The specifications explanation of Figure 1 thus distinguishes between

the received data signal and the clocks or timing signals that operate

on the data signal. And it indicates that the references to 400 MHz are to

those clocks or timing signals, not the data signal. That explains why

Figure 1 uses the different units of MHz and Gbpsif all references

were to the data rate, as the ALJ assumed, the Figure would have uniformly

referred to bits-per-second everywhere. The ALJ misread Figure 1 because

he divorced his analysis from the specifications explanation.

The ALJ made the same mistake regarding Figure 12, which appears

as follows:


7/28/2019 cafit

57/84

47

A637. The ALJ mistook the frequency of the clock that accepts and

operates on the data signal, A642, 7:5-6, for the frequency of the data input

signal itself, A107. As a result, the ALJ again mistakenly interpreted the

400 MHz reference in Figure 12 (which, with a 10-bit parallel input,

would be the equivalent of a 4 GHz 1-bit serial input) as proof that the

frequency of the data signal has a one-to-one correspondence with the 4

Gbps output data rate labeled in Figure 12. A107.

The ALJ stated that Rambus is simply incorrect that these examples

are merely referring to clock rate. A108. But he offered no explanation

why that was so. Nor did he attempt to reconcile his interpretation with the

specifications text. The ALJs confusion of the clock rate with the output

frequency of the data signal on which it operates was fatal to his

interpretation.

Properly considered, both Figures 1 and 12 support Rambuss con-

struction. As explained above (and as the specification states), a 4 Gb/s data

signal corresponds to a 2 GHz output frequency (D=2F). See pp. 34-37,

supra. Applying the ALJs math correctly, the 10-bit parallel data in Figures

1 and 12 is processed using 400 MHz clocks, or the equivalent of a 4 GHz

clock. A clock signal has a rising edge and a falling edge. When only one

clock edge is used to sampleas was common when the Dally applications


7/28/2019 cafit

58/84

48

were filedthe clocks frequency is twice the maximum frequency of the

data signal. Thus, the equivalent of a 4 GHz clock in the figures corres-

ponds to a maximum signal frequency of 2 GHz.5

For that reason, Figures 1

and 12which could have a maximum data signal frequency of 2 GHz

given the clock speed and show a data rate of 4 Gb/ssupport Rambuss

position.

Finally, the ALJ relied on the specifications reference to a 4 Gbs

serial channel in which a single serial channel (two pins) replaces 40 100

MHz pins. A108 (quoting A642, 8:13-16). The ALJ concluded that

[f]orty parallel channels at 100 MHz equates to one serial channel at 4

GHz, and then equated a 4 GHz output frequency with a 4 Gb/s data rate.

Id. But the ALJ assumed that the 40 100 MHz pins configuration refers

5Literature contemporaneous with the Dally patents notes that a digital sig-

nal is first sampled at a rate fs Hz (or samples/sec), which must be greater

than twice the highest frequency component in the signal. JA14990. As

the figure below illustrates, a 4 GHz clock is required to sample a 2 GHz

data signal on only the rising edge (shown in red):


7/28/2019 cafit

59/84

49

to single-ended signaling, which requir[es] one pin for a given channel.

A108. That assumption is unfounded. Single-ended signaling is nowhere

mentioned in the Dally patents. Rather, as the ALJ noted, the Dally patents

examples refer to differential signaling, which requir[es] two pins for a

given channel, not one. A108. The ALJ acknowledged that the 4 Gbps

single channel described in this embodiment he invoked is itself a two

pin implementation that uses differential signaling. Id.

For those reasons, one of ordinary skill in the art would have under-

stood the cited 40 100 MHz pins to be configured in pairs for differential

signaling. That would yield 20 two-pin differential interfaces, not 40 one-

pin interfaces for single-ended signaling as the ALJ assumed. Applying the

ALJs math, 20 differential interfaces x 100 MHz = 2 GHz. The serial

channel thus supports D=2F: A 2 GHz signal frequency again correlates

with a 4 Gb/s data rate.

The ALJ hedged that, even if Rambus is correct, the fact that one

embodiment is inconsistent is not fatal to his effort to equate an output

frequency of 1 GHz with a data rate of 1 Gb/s. A108. But the ALJ no-

where explained why that embodiment would exhibit a different and unique

relationship between output frequency and data rate present nowhere else in


7/28/2019 cafit

60/84

50

the patent. All of the evidence is to the contrary. See pp. 31-39, supra.6

And, even if ambiguity remained, that ambiguity should be resolved in favor

of the construction that renders the claims validnot the ALJs invalidating

construction. See Modine, 75 F.3d at 1557; Wang Labs., Inc. v. Am. Online,

Inc., 197 F.3d 1377, 1383 (Fed. Cir. 1999).

* * * * *

For the reasons above, the Dally patents claimed output frequency

of at least 1 GHz equals an output data rate of at least 2 Gb/s. According to

the ALJ, the highest output data rate in the relevant prior art is 1.0625 Gb/s,

about half the claimed rate for the invention here. See A228 (discussing

Widmer and Ewen), A268 (discussing the SL500 art). Because no prior art

reference discloses the claimed output frequency or corresponding data rate,

the Commissions invalidity findings must be reversed.

6Respondents also argued that Rambuss construction of output frequency

would render another limitation in the claimsa bandwidth greater than

100 MHz, A642, 8:66-67superfluous. A103. That argument was based

on the fact that nonmodulated signals have a bandwidth equal to twice the

highest frequency. JA3825. Consequently, if the output frequency is at

least 1 GHz, as claimed, the bandwidth of the nonmodulated signal will

always be greater than 100 MHz, leaving no need for the bandwidthlimitation. The superfluity argument fails on the merits, for reasons Rambus

explained below. See A104. In any event, the ALJ acknowledged the issue

is not dispositive on the claim-construction question. Id. If respondents

view were correct, it would render the 100 MHz bandwidth limitation

superfluous undereitherconstructionwhether the data rate equals the out-

put frequency, or is twice the output frequency.


7/28/2019 cafit

61/84

51

II. Rambus Established A Domestic Industry For The Dally PatentsA. The Domestic Industry RequirementAlthough the Commission has authority to exclude from the United

States any articles that infringe a U.S. patent, its focus is not intellectual

property but trade: The Commissions purpose is to adjudicate trade dis-

putes between U.S. industries and those who seek to import goods from

abroad. InterDigital Commcns, LLC v. ITC, 707 F.3d 1295, 1302 (Fed.

Cir. 2013) (quotation marks omitted). Accordingly, the Commission can

grant relief only if an industry in the United States, relating to the articles

protected by the patent, . . . exists or is in the process of being established.

19 U.S.C. 1337(a)(2). That is the domestic industry requirement. Inter-

Digital, 707 F.3d at 1297.

Documents

cafit