Codex Corp. v. Milgo Elec. Corp., 534 F. Supp. 418 (D. Mass. 1982)

US District Court for the District of Massachusetts - 534 F. Supp. 418 (D. Mass. 1982)
March 3, 1982

534 F. Supp. 418 (1982)

CODEX CORPORATION and Yellow Freight Systems, Inc., Plaintiffs,
MILGO ELECTRONIC CORPORATION and International Communications Corporation, Defendants.

Civ. A. No. 76-793-S.

United States District Court, D. Massachusetts.

March 3, 1982.

*419 Paul F. Ware, Jr., Goodwin, Procter & Hoar, Boston, Mass., for plaintiffs.

Marcus E. Cohn, and Cornelius J. Moynihan, Jr., Peabody & Brown, Boston, Mass., Harold L. Jackson, Jackson, Jones & Price, Tustin, Cal., for defendants.


SKINNER, District Judge.

The plaintiffs have moved to amend the court's Findings, Rulings and Order of June 12, 1981. Some of these proposed amendments correct predictable technical errors by the court and these have been adopted. Others go to the substance of findings, for some of which I have found support in the evidence, but some of which appear to add findings not expressly dealt with during the trial. The detailed resolution of these proposed amendments are as follows:

1. My memory of the evidence supports the original findings except that transmission at high speeds is not over telephone lines.

2-3. The suggested changes are not supported by the evidence.

4-5. A quibble duly adjusted.

6. A technical correction.

*420 7. A technical correction which was not apparent in the evidence, but to which all parties agree.

8-9. These are technical corrections to which all parties agree.

10. I didn't articulate the difference between symmetrical and asymmetrical spectra, but the suggested amendment is consistent with the evidence and eminently sensible.

11. The precise dating suggested is not significant and I have omitted it.

12. The changed language in paragraph 7 on page 13 reflects the comments of both parties. The phrase "as nearly as possible" was taken from Mr. Whang's testimony.

13. See No. 27, infra.

14. I have combined the comments of the parties in the changes on page 15.

15-16. The reference to the WU 2247 as being the same as the 4400/24 PB was an error, and has been eliminated. The confusion occurred because both the WU 2247 and the 4400/24 PB were the versions of Milgo 4400/24 series.

17. Re-examination of the Milgo spectographs leads me to rewrite the first paragraph of page 17 as it appears in the amended findings. It also is true that the Kansas court's findings with respect to the 3300 and 2200 models referred to patents not at issue in this case.

18. The reference to a doctoral thesis was clearly an error and has been corrected to "master's".

19. The proposed amendment is adequately dealt with by changing "all" to "most".

20. The suggestion to use a more technically accurate description was agreed to by both parties and adopted.

21. The reference to prior art is footnoted.

22. The proposed amendment adds some accuracy.

23-24. I misread the opinion of the Kansas court with respect to the claims in dispute before it, being confused by the discussion of the fixed equalizer. The defendant is correct in its comment that claims of the '023 patent referring to the fixed equalizer were apparently not before that court. Conclusion 2 on pp. 31-32 has accordingly been entirely rewritten.

25. Rather than get into the obsolescence of fixed equalizers, about which the evidence was slight, I have eliminated the reference, which was superfluous in any case.

26. I do not accept the proposed amendment.

27. I was informed by counsel at the outset that I should concern myself with claims 1, 19 and 25 of the '023 patent. There was little, if any, discussion during the trial of the dependent claims. To the extent that the validity of the dependent claims rests on the assertion that the patent teaches narrow skirts, i.e., a composite filter characteristic of less than 50% roll-off, my finding that such a teaching is nowhere contained in the patent would collaterally estop these parties to assert validity on that basis. Accordingly, I do not adopt the proposed amendment.

28. The suggested amendment probably does eliminate the possibility of confusion between "band limiting" as ordinarily used and the special meaning given to the phrase by Mr. Whang. It has been adopted.

In accordance with the foregoing, the Findings, Rulings and Order entered June 12, 1981 are vacated, and Amended Findings, Rulings and Order reflecting the foregoing are filed herewith. The Declaratory Judgment heretofore entered is vacated. No judgment shall enter until the amount of the award of attorneys' fees has been determined and made a part of said judgment.



This is an action for a declaratory judgment establishing the invalidity of three patents held by the defendants (hereinafter collectively "Milgo") as assignees of the inventors:

*421 (1) No. 3,524,023, Sang Y. Whang, inventor, Band Limited Telephone Line Data Communication System ("Whang '023").
(2) No. 3,619,503, Robert G. Ragsdale, Phase and Amplitude Modulated Modem ("Ragsdale '503").
(3) No. 3,783,194, Viesturs V. Vilips, Data Modem Having a Fast Turn-Around Time Over Direct Distance Dialed Networks ("Vilips '194").

By a supplemental complaint, plaintiffs (hereinafter collectively "Codex") sought a similar declaration with respect to Patent No. 3,943,285, Robert G. Ragsdale and Henry H. Parrish, but this patent was withdrawn from the case prior to the trial.

The defendants have by answer and counterclaim asserted the validity of their patents, alleged infringement of their patents and seek injunctive relief and damages. Both sides claim they are entitled to attorneys' fees.

During the trial I ordered summary judgment for the plaintiff on the Ragsdale '503 on the basis of the admission by Mr. Ragsdale that the only novelty in his patent was in a claim which was not at issue in this case.





The patents in issue deal with devices called modems, the function of which is to convert the discrete digital signals of computers into analogue signals suitable for transmission over telephone lines, and at the receiving end, to convert the analogue signals back to digital signals which can be "read" by the receiving computer. "Modem" is short for "modulator-demodulator."

Computer signals are binary, that is, they consist of but two electrical impulses, one positive, one negative, usually represented as "1" and "0". Each 1 or 0 is known as a "bit." Information, or "data," is coded by assigning words, letters or numerical values to various combinations of bits. Common coding practice employs the "tribit," a "word" consisting of three bits. Combinations of tribits are then transmitted according to a pre-set code to convey data. There are eight available tribits: 000, 001, 011, 111, 101, 110, 100, and 010.

A modem which can transmit eight identifiable signals can transmit tribits; a modem which can transmit only four identifiable signals can transmit only dibits, 01, 11, 10, 00; a modem which can transmit only two identifiable signals can transmit only the single bit "words", 1 and 0. Use of tribits obviously increases the range and speed of data transmission.

The aim of modem designers is to increase the speed of transmittal so that the capacity of large computers, which are capable of processing data at the rate of 150,000 bits per second ("bps"), may be fully utilized by remote terminals. Such speeds may be approached over special private or leased lines which are tuned to provide minimal signal distortion over a wide range of frequencies. Such lines are expensive. Accordingly, the focus of design efforts over the last twenty years has been to create a modem capable of transmitting large quantities of data per second over the ordinary dial telephone lines.


Problems of the Telephone Network.

The commercial telephone system is designed to transmit signals in the frequency range from 300 to 3000 Hertz (Hz) or cycles per second, known as the voice band. At the upper and lower ends of this frequency range the telephone systems exhibit characteristics of amplitude distortion and delay distortion. Since the signals transmitted by modems are distinguishable by minute differences of amplitude and timing, these distortion characteristics are fatal to accurate transmission of data. These distortions may be corrected by equalizers, which retard and suppress the fast and strong signals to the level of the signals in the frequencies in which there is the most distortion. Unfortunately, every pair of telephone lines has different distortion characteristics, and because of the automatic routing system in the national distance dial telephone network, it is impossible to predict *422 what distortion characteristics will appear in a given transmission. There are three types of equalizers which have been used to deal with this problem: (1) the adjustable equalizers, which must be manually tuned for each transmission, (2) the fixed compromise equalizer employed by Whang in his '023 patent, and (3) the automatic adaptive equalizer used by the plaintiff, Codex, and also by the defendant in its present line of modems.

The central range of the voice band is relatively free from distortion, however, and less stringent equalization is required to transmit accurately in this range, referred to at trial as the "sweet spot." In the early 1960's, when the modem described in the '023 patent was devised, the "sweet spot" was thought to be a band about 1000Hz in range between the frequencies of 1200Hz and 2200Hz. Since that time the telephone network has improved and the usable "sweet spot" for modems may be somewhat wider.


Characteristics of the Signal

The carrier of the signal is an alternating current which is ordinarily depicted as a sine wave, the undulations of which reflect the positive-negative alternations of energy:


In fact, however, the wave form should be visualized as a helix, like the thread of a cylindrical bolt. This is important in understanding the measurement of phase modulation, and the structure of combined phase and amplitude modulation, quadrature modulation and the operation of the automatic adaptive equalizer. In the literature, accordingly, the carrier signal is depicted both as sine wave and as a circle.

Transmission of data is accomplished by modulating the carrier signal in accordance with a pre-established code. Modulations may be in the amplitude (strength) of the signal (AM), in its frequency (FM) or in its phase (PM), or by a combination of modes of modulation. This case deals primarily with phase modulation, or combinations of phase and amplitude. Amplitude modulation simply makes the oscillations of the carrier signal more or less pronounced. Phase modulation is more difficult to visualize, but is well-pictured in Whang '023, Figure 8.


In practice, the phase modulation is accomplished by (for instance) eight different carrier transmitters each in a different phase 45° apart. At each modulation a different carrier transmitter enters the circuit according to a pre-set code.

Phase modulations are expressed as ± x°. The importance of visualizing the wave as a helix is now apparent. Returning to the simile of a cylindrical bolt in a threaded sleeve, rotation of the bolt moves the thread forward or back depending on the direction of the rotation. Modulation is expressed as changing the phase in terms of completion of the 360° cycle.

For purposes of data transmission each value of change of phase is assigned a word. In the case of an eight-phase modem, the words are tribits, as illustrated in Whang '023, col. 7, 1. 13-21. For instance, according to Whang's example, the phase shifts in Fig. 8 of -180°, -45° and -90° represent the tribits 111, 000 and 100 respectively.

The time interval for each modulation is known as a baud. If each modulation is coded to represent a tribit, the modem is said to be able to transmit three bits to the baud. The baud is the unit of time used by each modulation, but the actual time interval of a baud is expressed in microseconds, i.e., millionths of a second. The term T, however, represents the baud interval expressed in seconds. If the baud is 1250 microseconds, T has a value of .001250.


Problems in the Signal.[1]

The first problem arises from the consequences of modulating the carrier signal. Modulation of an unfiltered carrier signal produces a tremendous dispersal of energy at frequencies above and below the carrier frequency. The distribution of this energy on the frequency scale is mathematically predictable. Because this energy appears over a wide range of frequencies it is affected by the relatively large delay distortion characteristic of the upper and lower ends of the voice band of the telephone lines. The delayed reception of this energy by the receiving modem will distort the apparent phase and amplitude of the received signal and thus result in inaccurate translation of the coded data.

This problem is solved by filtering the signal by a series of filters (of which the telephone system is itself one) which attenuates, or filters out, energy at all but the desired frequencies. This is called the composite filter of the modem. Some components *424 of the composite filter may be in the transmitter and some in the receiver.

The solution of the first problem creates the second. While the composite filter attenuates the energy which appears outside the desired frequency band width, it spreads the signal out over time. The energy representing the signal in one baud is smeared in time so that some of it is still "ringing" or echoing in the bauds that succeed. The cumulative ringing from prior bauds creates such confusion during the succeeding bauds that the receiving modem cannot distinguish the modulation of the succeeding baud. This is called intersymbol interference.

One solution of this second problem is to use a composite filter which restricts the energy which is passed at a certain level within a frequency range (the passband) known as the Nyquist band. Nyquist, a Bell Laboratories researcher, discovered as early as 1928 that if a filter were used which permitted a passband of a range of Hz, the intersymbol interference would all be at a null point at the center of the succeeding baud. If the receiver samples the signal at the center of the baud, it will be able to read the actual signal of that baud. This technique is referred to as "center sampling."

A third related problem is to time the receiver so that it samples at the center of the baud. The superimposition of the data signal on the carrier signal produces a pattern of overall energy observable over time which is known as the signal envelope. The graphed shape of the signal envelope will vary according to the nature of the composite filter used in the modems. These shapes are illustrated by Figures 3 through 7 of the Whang '023 patent. Figure 3 is significant because, according to the description at col. 2, 1. 68-9, it "illustrates a typical desired response of signal in Fig. 2 when subjected to proper band limiting filter;" i.e., a filter producing a band width limited to Hz (Nyquist band). A timing impulse may be derived from the envelope illustrated in Figure 3 which will so govern the receiver as to cause it to sample at the center of each baud. This technique is referred to as "recovery of clock from the envelope."


Characteristics of Filters

For purposes of this case, the composite filter to be considered is one which limits the band width to the Nyquist band ( Hz). Such filters are characterized by the spectral configuration of the signal which the filters pass, showing the energy measured in decibels at different frequencies.[2]

The ideal Nyquist band filter would produce a rectangular spectrum.


If such an ideal filter could be built, no energy would be transmitted except in the frequency range of the passband. Such a filter is not a practical possibility. The Nyquist goal of achieving nulls of intersymbol interference in succeeding bauds may be achieved however with a spectrum such as the following, where the area of a equals the area of b, and the area of c equals the area of d, and the passband is Hz:


In the common case a and c would also normally be approximately equal, but the curve may be slightly asymmetrical in this respect, as evidenced by spectrographs produced in court.

There are a number of filters which will meet this criterion, but which have slopes of different sharpness. They are described as having different percentages of "roll-off" or wide or narrow "skirts". The skirts refer *426 to the frequency ranges outside the passband at which some significant energy is transmitted, and I have labeled them S' and S" in the above diagram.

A symmetrical filter is said to have 100% roll-off when S' plus S" equals the passband; 50% roll-off when S' plus S" equals 50% of the passband, and so on. For an asymmetrical filter, the roll-off of each skirt is calculated by comparing S' and S" to one-half the passband. Roll-off in excess of 100% would not satisfy the geometry of the Nyquist criterion described above. The narrower the skirts, i.e., the lower the percentage roll-off, the more efficient the filter is at attenuating the energy outside the passband.

Sharp roll-off also increases the delay distortion within the passband, as well as increasing the difficulty in recovering "clock" from the signal envelope, which depends upon energy outside the passband.


History of the Whang '023 Patent

The '023 patent is the result of research undertaken by Whang in 1966. Whang was at that time a filter engineer employed by the defendant Milgo; he had little specific experience in modem design. There was no commercially successful high-speed modems then available which could operate reliably on the commercial dial network. The Bell Systems Laboratory had developed a four-phase 2400 bps modem which was in use but not entirely reliable. Codex was concentrating on a 9600 bps modem which was not commercially successful, being expensive and unreliable. It had several experimental models in process, but did not bring out its first commercially successful product until after the issuance of the Whang '023.

Whang was assigned by Milgo to devise a modem to satisfy the requirements of Western Union, which was seeking a 2400 bps modem for use on lines with characteristics similar to the commercial dial telephone system. Western Union specified that the modem should operate within a 1000Hz band width, at the "sweet spot."

Whang developed a 2400 bps modem with the following salient characteristics:

1. An eight-phase signal.
2. A baud interval of .001250 seconds (T).
3. A composite filter having a passband width of 800Hz, i.e., Hz and a roll-off characteristic of 100% or less, i.e., a composite filter which complied with the Nyquist criterion.
4. A carrier frequency of approximately 1700Hz so that the passband was located within the 1000Hz "sweet spot" of the BEX system and commercial telephone dial systems.
5. A fixed compromise equalizer.
6. A device for center sampling of the equalized signal.
7. A timing device which derived a pulse from the envelope of the data signal, translated this pulse to a very high frequency signal to time the receiver to achieve sampling as nearly as possible at the center of beach baud.

While the experimental modem first exhibited to Western Union may have had a composite filter roll-off of less than 50%, the 175 modems actually sold to Western Union had composite filter roll-off substantially in excess of 50%. They worked very well.

Whang's solution of Western Union's problem was a radical departure from the approach of other contemporary modem designers, and successfully utilized a method of data transmission generally regarded as unfeasible. The accepted philosophy among modem engineers skilled in the art called for development of suitable automatic equalizers so that transmission could be accomplished without distortion over a wide spectrum of frequencies. Speed could be achieved over the wider band widths without encountering the narrow tolerances of eight-phase transmission.

In July, 1966, Whang filed an application for what became the '023 patent, listing the defendant Milgo as his assignee. After many amendments, including the addition of center sampling to the claims, and after considerable discussion with the examiner, the patent was issued August 11, 1970.

*427 I find the following to be significant prior art which was not considered by the patent examiner:

Bennett & Davey, Data Transmission McGraw-Hill, N.Y., N.Y. (1962) Chapters 5, 8, 10, 11
Widl, W., "An Experimental data Transmission System", Ericsson Review, Vol. 39, No. 3, 1962, pp. 62-71
Evans, G.L., Enriquez, E. and Wilson, Q. C. "A High Speed, Serial, Four-Phase Data Modem for Regular Telephone Circuits", Convention Record Global Communications, May, 1961, pp. 100-104

I further find that the applicant misrepresented the work of Irland as using full modulation interval decoding rather than center sampling. The assumed novelty of combining center sampling with the other elements of present claim 25 was apparently a significant factor in securing the allowance of present claims 1, 19 and 25 of the '023 patent. It is upon these independent claims that defendants principally rely in their counterclaim for infringement.

After the issuance of the patent, Milgo affixed labels asserting the protection of the Whang '023 patent on its modems with the following model numbers:

WU 2247 (the original Western Union Model) 4400/24 PB 2200/20 2200/24 3300/36 20 LSI 24 LSI 201 LSI 201C LSI 96 MM 26 LSI

The label was removed from the 2200/20 and 2200/24 on January 28, 1976. On April 11, 1977, Milgo removed the labels from all models except the WU 2247 (which was by that time obsolete), the 4400/24 PB, (which was related to the WU 2247, but which had somewhat narrower skirts) and the 96 MM. Both the labeling and the unlabeling were done at the direction of Attorney Stanley Jones, Milgo's patent attorney, who prosecuted the '023 application in the patent office and who has represented Milgo in all subsequent litigation concerning that patent, including the present case.

Of the Milgo models listed above, only model 96 MM has "narrow skirts", i.e., a combined filter roll-off of 50% or less.

Mr. Whang presently asserts that his patent covers Milgo 4400/24 PB, 4400/48, 4600/48, 48 MM and 96 MM, and also asserts infringement by various models of other manufacturers (Chart FF). Mr. Whang testified that all of these were designed with narrow skirted filters. I find that all of these modems do have a combined filter characteristic of less than 50% cosine roll-off except the 4400/24 PB. I find that most of the 4400/24 PB produced and sold had a cosine roll-off of greater than 50%, in fact on some modems close to 100%. A few individual modems in the 4400/24 PB series may have had a cosine roll-off close to 50%, but this appears to have been a haphazard consequence of the vagaries of filters.

Prior Litigation: Milgo Electronic Corporation v. United Business Communications, Inc., 189 USPQ 160 (D. Kansas 1976); aff'd per curiam, 623 F.2d 645 (10th Cir. 1980).

On July 19, 1971, Milgo brought suit against United Business Communications, Inc. ("UBC") for infringement of the Whang '023 patent. The court found that UBC had purchased a Milgo 4400/48 and had copied it prior to the issuance of the '023 patent. Thereafter, the defendant produced a "chinese copy" of the 4400/48 and continued to market it after the effective date of the patent. Asserted invalidity of the patent was the principal defense.

Whang testified and the court found that the patent covered Milgo Models 4600, 4400/24,[2a] 4400/48, 24 LSI and 20 LSI. He also testified that the composite filter of all of those modems had cosine roll-off of less *428 than 50%. This was untrue, as Whang now admits, except for the 4600 series and the 4400/48. The expert witness for UBC never examined or tested any of the modems, so that Whang's assertion went unchallenged.

The district court found that the '023 patent was valid as a novel and non-obvious combination of old elements of the art. Findings 32 and 33 are as follows:

32. Prior to the Whang '023 invention, the modem industry believed that the modem that performs the best with greatest noise tolerance margin over a good line or without a line (back-to-back) would also perform the best over a poor line. For that reason, most prior art modems employed wide band energy spectrum and two or four level modulation. [Citations omitted].
33. The Whang '023 invention made a breakthrough because Whang approached the problem with a completely opposite philosophy. His approach was based on the theory that a modem that would perform the best and adequately over a lousy line would perform adequately over any line (Whang '023, Col. 2, lines 38-43). Whang's invention combined both eight level modulation and extreme band limiting into one modem. The Whang approach represented a new and novel conceptual theory that went against the state of the then existing modem art. [Citations omitted].
189 USPQ at 173.

The court also concluded that a composite filter characteristic of less than 50% cosine roll-off was included in the claims. The district judge ruled that an inventor may be his own lexicographer, and that Mr. Whang had written his own specifications without benefit of some of the existing textbooks. Finding 61 states in part:

61. ... The Whang invention teaches and claims his composite filter means which, when expressed mathematically in terms of roll-off of DX J-1, requires a roll-off of from about 50% roll-off down to the ideal or zero percent roll-off of the 1/T Nyquist limit. [Citation omitted]. The features of a narrowband composite "filter means" of the Whang '023 patent represent a significant and non-obvious improvement over the widebandwidth filters recommended for phase modulated systems in DX J-1.

The Court of Appeals for the Tenth Circuit affirmed the judgment of the district court. In an extended per curiam opinion, the court picked up on the divergent philosophy of contemporary modem designers compared to that embodied in the '023 patent, the failure of Whang's predecessors to find a practical solution to high-speed transmission of data over ordinary long-distance switch voice-grade lines and the commercial success of Milgo modems. (The evidence of commercial success in the district court, however, included evidence of the sale of modems now said not to be covered by the patent because of their wide-skirted filters.)

The Court of Appeals made no mention of roll-off characteristics of filters in its opinion, but presumably in affirming the judgment it affirmed the district court's findings relative to filters.


Defendants' Position at Trial

The defendants' principal witness on the '023 patent was Sang Whang, the inventor. On cross-examination Mr. Whang conceded several times that differential phase modulation, limiting the passband to Hz and center sampling were old, not only separately but in combination.[3] He says further his patent is a modem patent, not a filter patent, and does not teach anything about designing filters. He says the novelty of his invention is the combination of these elements with severe band limiting. By severe band limiting he means a passband of Hz and confinement of substantially all the passed energy *429 within the "sweet spot". The energy level at the "sweet spot" edges should be no more than "a couple of percent". According to Mr. Whang, a person skilled in the art would recognize that a combined filter roll-off of about 50% or less would be required. In drawing the specification in the '023 patent he claims to have used the phrase " Hz passband" in a special sense to include the confinement of substantially all of the energy within the "sweet spot".

In support of his present reading of the patent he points out that the preferred embodiment calls for a signal time of .00125 second, a passband width of 800Hz, i.e., Hz, and a carrier frequency of 1700Hz, located in the middle of the "sweet spot" between 1200Hz and 2200Hz. According to Whang, specifications made it clear that substantially all of the energy outside of the "sweet spot" should be attenuated.

If it were true that the patent claims called for a composite filter with the above characteristics, and energy at the "sweet spot" limits was down around 2%, I find that a person skilled in the art would infer the necessity of constructing a composite filter with a cosine roll-off of about 50% or less.


Facts Which Impeach Whang's Testimony

1. In the art of filter design, at least as applied to data transmission, "passband" has a generally accepted meaning. It is used to denote the range of frequencies which a filter passes at amplitude levels down to about -3db. Whang was a filter specialist before he undertook the invention of modems. He is, moreover, a highly educated man, with considerable facility in technical language (despite his present disclaimers). He wrote a master's thesis on filters. I find it highly unlikely that he used the term "passband" in his patent in anything other than the ordinary sense. Furthermore, the use of "passband" in the generally accepted sense is consistent with the rest of the language in the patent, and Mr. Whang's new special meaning is not. I further find that a person skilled in the art would read the word "passband" as used in the '023 patent as meaning the range of frequencies passed by the composite filter of the modem at amplitude levels down to about -3db.

2. Alternate embodiments of Whang's invention as described in col. 14, 1. 68-72 of the '023 patent would permit the edges of the passband to coincide with the edges of the "sweet spot". Since a perfect zero roll-off filter is impossible, the necessary implication is that there will be substantial amounts of energy transmitted outside the "sweet spot". Accordingly, it is not even a permissible inference from the language of the patent that all of the energy except "a couple percent" would be within the "sweet spot". Whang's conclusion that a teaching of "narrow skirts" is to be implied from the centering of the passband in one of his suggested embodiments is completely undermined by the lines cited.

3. Figure 3 of the '023 patent is described as depicting "a typical desired response of signal in FIG. 2 when subjected to proper band limiting filter." (Col. 2, 1. 69-70). In a subsequent patent authored by Mr. Whang (Scott-Whang 3,988,540), the '023 patent is referred to as prior art. Figure 3 is reproduced as Figure 1A and described as "an illustration of waveforms from the aforementioned prior art patent" ('540, col. 2, 1. 67-8). At the trial, under repeated questioning by the court, Mr. Whang unequivocally stated that he intended this figure to represent the signal actually transmitted by the modem described in the '023 patent. This figure depicts the signal transmitted through a composite filter having a roll-off of 100%, i.e., skirts of the maximum width consistent with the Nyquist criteria described above.

4. Milgo's prototype modem, the 4400/24 (or WU 2247), the development of which led to Whang's purported invention, had wide skirts. Subsequent Milgo modems with wide skirts were labeled with the '023 patent number on them. Mr. Whang was Milgo's principal modem designer, and these modems constituted his claim to fame in modem design. He was a specialist in filter *430 design as well. I find his present testimony that he was unaware of the filter characteristics of the WU 2247 and other Milgo modems carrying the '023 label not to be credible. I find that the concept of narrow skirts as the novel teaching of the '023 patent to have been devised by Mr. Jones and Mr. Whang after the fact, for purposes of establishing the validity of the '023 patent in the Kansas litigation.


Ultimate Finding of Fact Concerning Filter Roll-Off Characteristics of the Whang '023

I conclude that the Whang '023 patent, neither expressly or by implication, on neither the claims nor the specifications, describes a composite filter roll-off characteristic of less than 50% as an element of the invention. On the contrary, I find that, as Whang himself testified at one point, Whang left it to the design engineer to determine the roll-off characteristics of the composite filter anywhere within the limits of the Nyquist criterion, i.e., from zero to 100 percent, depending upon the requirements of the particular application.


Characteristics of the Allegedly Infringing Codex Modems

Most of the present line of Codex modems have composite filter roll-off of less than 50%. They have in fact very "narrow skirts" of the order of 25% roll-off. They employ the Nyquist band width of Hz. Defendants urge that they also employ center sampling, an element of all of the independent claims of the Whang '023 patent upon which the defendants rely in this action, namely claims 1, 19 and 25. Plaintiffs claim that their use of a digital automatic adaptive equalizer permits a different system.

Center sampling takes advantage of the Nyquist bandwidth resulting from the composite filter of the modem. Use of this bandwidth ( Hz) produces a null in the intersymbol interference at the center of succeeding bauds. The pattern of these various signals which shows on an oscilloscope vaguely resembles a human eye, viz.,


and the process is referred to as sampling at the "eye". The delay distortion of the commercial switched dial telephone lines, however, will obliterate the "eye", because the null points of various elements of intersymbol interference will be out of synchronization. In the '023 patent, Whang specifies the use of a fixed compromise equalizer to reduce the effect of this distortion and to permit reasonably accurate sampling at the center of each baud. The utilization of the fixed equalizer is also described in some of the dependent claims of the patent (e.g., claims 6-10). The essence of the center sampling process is sampling every time at the center of each baud.

The sampling system employed by Codex reads the signals before they are equalized. The sample is taken at a point perceived by the decoding device as the point that maximizes band edge signal energy. In a perfect line this would indeed be in the center of the baud, but in a real line it may be in the center or substantially off-center on either side. For purposes of the Codex system it does not matter. The distortion is *431 later resolved in the automatic adaptive equalizer.[4]

There is a resemblance to center sampling, in that the sample is taken at intervals of T, but I would suppose that to be a likely characteristic of any sampling technique.

It is also true that the Codex decoding system, i.e., sampling at the point that maximizes band edge signal followed by adaptive equalization and integration, achieves the same result as center sampling. Indeed, the Codex and Milgo modems are compatible.

I find, however, that the means employed by Codex are basically and conceptually different from center sampling in the sense employed in the '023 patent, and indeed are antithetical to the entire concept and purpose expressed by Whang in the '023 patent.

I conclude that the plaintiff Codex does not employ center sampling in its modems, and accordingly does not infringe the independent claims 1, 19 and 25 of the Whang '023 patent.



One of the characteristics of long-distance commercial telephone lines which affects data transmission is the presence on the line of echo suppressors. These are devices which permit signals to be transmitted in only one direction at a time, the preference being given to the stronger signal. Without them an ordinary telephone user will receive echoes of his own voice. The problem with them is that they require about 100 milli-seconds to reverse direction. Data is transmitted in small blocks. The receiving unit acknowledges receipt and signals any errors. If there is an error in transmission, the block of data is repeated; if not, a new block is transmitted. This cycle is repeated at a very rapid rate. The additional turnaround time of the modems resulting from the time required to reverse the echo suppressors cumulates over a succession of transmission cycles until it represents a substantial loss of valuable time in the transmission of data.

The purpose of the device described in the Vilips '194 patent is to deactivate the echo suppressors. A special tone is transmitted over the line which disables the echo suppressors. Echo suppressors remain disabled if there is energy constantly on the line, even though the special tone has ceased. There may be gaps in transmission which would permit the echo suppressors to start up again, however, requiring a repetition of the initial disabling procedure. This is prevented by a second tone generated on the line as soon as the disabling tone stops, which remains on during the entire transmission. This tone has a frequency within the band width of the telephone system which does not overlap the frequencies of the data transmitting signal. This tone has no purpose other than to disable the echo suppressors in order to facilitate fast turn-around of the data transmitting signals.

The issue is whether this device is sufficiently novel and non-obvious as to constitute an invention.

I find that it was well known in the prior art that echo suppressors, once disabled, would remain disabled as long as there was energy on the line, and that this could be accomplished by a signal on a second frequency channel. In the prior art as it related to modems, this was accomplished by a reverse channel, which transmitted in the direction opposite to that of the main data signal and at a different frequency. This reverse channel was also used for data transmission for different purposes in different applications, but it was recognized that it could be used solely for the purpose of disabling the echo suppressors in order to reduce the turnaround time.[5]

*432 The device described in the Vilips '194 patent eliminates the data carrying aspect of the secondary channel and causes the continuing tone to be activated by the modem rather than the business machine to which the modem is attached. In the prior art, the reverse channel was activated automatically, even though there was no data to transmit, but the device for doing so was located in the business machine.

The Vilips device is smaller and cheaper because the data transmitting and receiving components are eliminated.

I find that the device described in the Vilips '194 patent represents (1) the exploitation of one well-recognized aspect of a device utilized in the prior art to the exclusion of other aspects, and (2) a business judgment that the market would forego additional data transmittal capacity in the secondary channel in return for a more compact and less expensive modem. In my view, changing the activating device from the business machine to the modem was an improvement, but an improvement of the order of a mechanic's expedient rather than an invention.



1. The claims of a patent, in order to be valid, must be novel and non-obvious over the prior art. 35 U.S.C. §§ 102, 103.

2. The protection of a patent does not extend beyond its claims. United States v. Adams, 383 U.S. 39, 86 S. Ct. 708, 15 L. Ed. 2d 572 (1966).

3. Terms used in a patent shall be given their ordinary meaning among persons skilled in the art, unless it is plain from the context that a special meaning has been assigned to them by the inventor. Eastern Electric, Inc. v. Seeburg Corp., 310 F. Supp. 1126 (S.D.N.Y.1969), aff'd 427 F.2d 23 (1970).

4. The exploitation of one generally known aspect of a device utilized in the prior art and the elimination of other aspects of the device does not constitute an invention, and the purported patent of such a device is invalid. Shu-Conditioner, Inc., v. Bixby Box Toe Company, 294 F.2d 819 (1st Cir. 1961).

5. A patent is presumed to be valid, but that presumption is negated to the extent that relevant prior art was not presented to or discovered by the patent examiner.

6. The presumption of validity of a patent is also negated to the extent that the prior art was misrepresented to the patent examiner and this misrepresentation affected the examiner's decision.

7. Comity should be extended to the prior decision by another court concerning the validity of a patent, even though the prior litigation did not involve a party to the present case. While the decision of the other court is not binding on parties who did not participate in the prior case, it is to be considered strongly persuasive in the absence of convincing new evidence or clear conviction that the prior decision is incorrect as a matter of law. Spray-Bilt, Inc. v. Ingersoll-Rand World Trade, Ltd., 350 F.2d 99 (5th Cir. 1965); New York Scaffolding Co. v. Liebel-Binney Construction Co., 243 F. 577 (3d Cir. 1917), aff'd 254 U.S. 24, 41 S. Ct. 18, 65 L. Ed. 112 (1920).

8. The decision of another court concerning the validity of a patent may be considered less persuasive if it was based on material misrepresentations of fact by the prevailing party.

9. A device is equivalent to a patented device, and thus infringes, if it produces a similar result by similar means.



Applying the foregoing rulings of law to the findings of fact, I come to the following conclusions:

*433 1. The presumption of validity of the Whang '023 patent has been rebutted by evidence of relevant prior art not before the examiner and by evidence of a misrepresentation of part of the cited prior art by the applicants.

2. The persuasive effect of the findings of the District Court of Kansas is adversely affected both by a misapplication of law and a misrepresentation of facts. The court's conclusion that the '023 patent claims included a narrow-skirted filter appears to be based on mistaken application of the rule that an inventor can be his own lexicographer. The rule only applies with respect to terms with an accepted meaning if the inventor clearly signals that he intends to apply an idiosyncratic definition.[6] Furthermore, the court based its conclusions concerning narrow skirts on a misrepresentation of fact, namely, that the first commercial embodiment of the invention in fact had a narrow-skirted filter.

The Kansas court was also impressed with the fact that Whang had achieved a practical breakthrough in the 4400/24 and WU 2247 modems. In my view, Whang did indeed make a practical breakthrough, but it was not the result, as the Kansas court thought, of the narrow-skirted filter. As previously stated, the practical break-through was accomplished by the WU 2247 which concededly did not have a composite filter with less than 50% roll-off. In my view, the breakthrough was the result of Whang's discovery that within a narrow band width all pairs of telephone lines would have such similar distortion characteristics that they could be reliably equalized with a fixed equalizer. This in fact is stated in the introduction to the '023 as the principal advantage of his invention. Col. 2, 11, 27-49. Neither in Kansas nor in the present case, however, has Milgo sued on any of the claims of the patent which incorporate a fixed equalizer, and in this case, at least, has specifically and forcefully argued that the use of the fixed equalizer is not to be read into claims 1, 19 and 25 from the embodiment in the specifications.

3. Claims 1, 19 and 25 of the Whang '023 patent are invalid. Mr. Whang has admitted that differential phase modulation, the use of the Nyquist band width, center sampling and the derivation of clock from signal envelope were all old, both singly and in combination. The evidence corroborates his admission. The asserted novelty, a composite filter roll-off of less than 50%, is nowhere stated in any of the claims or specifications of the patent, either expressly or by implication. Since there is no novelty, there is no invention.

4. Even if the '023 patent were valid, the Codex modems do not infringe because they do not employ one of the critical elements of claims 1, 19 and 25, namely center sampling.

5. The Vilips '194 patent is invalid for lack of invention; i.e., it is merely a mechanic's adaptation of prior art that is both old and obvious. In accordance with the stipulation of the parties, I conclude that if the '194 patent were valid, the Codex modems would infringe it.



Plaintiffs seek their attorneys' fees on the ground that this is a special case because defendants have deliberately fabricated the narrow skirt theory and have attempted to assert the Ragsdale '503 patent which the inventor himself admitted had no novelty in the claims asserted in this trial. I regretfully come to the conclusion that the plaintiffs are correct. The evidence is very strong that Whang knew that the WU 2247 modem had wide skirts, and quite persuasive that he also knew that a number of Milgo modems labeled as covered by the '023 patent did also. Mr. Whang's present posture of ignorance of the meaning of terms in his own field of specialty is not at all persuasive. Both Mr. Whang and Attorney Jones are highly trained and sophisticated people. I cannot escape the conclusion that both of these men have deliberately misrepresented the narrow skirt issue *434 to both the District Court of Kansas and to this court. One of the unavoidable hazards of patent litigation is the fact that district judges are likely to have no background in the technology involved. This places a heavy burden on patent lawyers and their expert witness to do their best to mitigate the situation rather than exploit it. Defendants' post-trial brief strikes me as doing just the reverse; in fact, it verges perilously close to double-talk.

Accordingly, I rule that this is a special case and the plaintiffs are entitled to reasonable attorneys' fees.



A judgment shall enter declaring that claims 1, 19 and 25 of the Whang '023 patent are invalid in accordance with the foregoing; declaring that plaintiffs' modems do not infringe said claims; declaring that claims 1 and 5 of the Ragsdale '503 are invalid for lack of novelty in accordance with the motion for summary judgment allowed during trial; declaring the Vilips '194 patent invalid for lack of invention; and declaring that if the Vilips '194 patent were valid, the Codex modems would infringe it. The defendants' counterclaim shall be dismissed on the merits, and plaintiffs awarded their attorneys' fees after further hearing.


[1] In the following discussion, some of the comments and diagrams will have reference to a continuum of frequencies ("the frequency domain") and some will have reference to a continuum of time ("the time domain"). In diagrams, the horizontal axis may be either a scale of frequency or a scale of time. The reader should take care to identify which is which. The vertical axis always represents amplitude, or the strength of the signal, usually expressed in decibels (db), a convenient logarithmic measure. A diagram in the frequency domain says nothing about time, although the event depicted may occur over some continuum of time, and similarly, a diagram in the time domain says nothing about the frequencies at which the depicted energy levels are fluctuating through time (except by calculation through the Fournier analysis).

[2] These graphs are in the frequency domain. The amplitude is measured in descending decibels from the assumed maximum amplitude of the signal. Decibel, as the term is used in measuring electrical energy, is a comparative measure defined as follows:

A unit for expressing the ratio of two amounts of electric or acoustic power equal to ten times the common logarithm of the power ratio. Webster's Third New International Dictionary

There is an apparent inconsistency in the testimony concerning the relationship of the logarithmic scale to actual levels of energy at Tr. 145-146. -12 decibels should represent less energy than -10 decibels.

[2a] This series included the Western Union WU 2247 and the 4400/24 PB.

[3] Whang also conceded that derivation of clock from the signal envelope existed in the prior art.

[4] The automatic adaptive equalizer is rather complex. The signals are stored and integrated, that is, averaged out, and released in an integrated form which makes sense to a decoding device.

[5] "Disablement of Echo Suppressors", CC ITT Supplement No. 85, Extract from AT&T Contribution Com. Sp. A/No. 75, July, 1963, pp. 674-688; Bell System Technical Reference, "Data Sets 402C and 402D Interface Specification", November, 1964. This prior art was not disclosed to the Patent Office.

[6] Ellipse Corporation v. Ford Motor Company, 452 F.2d 163, 167 (7th Cir., 1971); Chemical Construction Corp. v. Jones & Laughlin Steel Corp., 311 F.2d 367, 371 (3rd Cir., 1962); Rubbermaid Incorporated v. Contico International, Inc., 381 F. Supp. 666, 671 (E.D.Mo., 1974).