After a bench trial in this patent infringement action, the District Court upheld the validity of Patent No. 3,454.877 (issued in July 1969 to Loebe Julie), and ruled that defendant-appellant Guildline Instruments, Inc., had infringed claims 1, 5, 8 and 9. The patent relates to a potentiometer, a device for measuring an unknown voltage by producing a known voltage for comparison with the unknown voltage, and covers 'Potentiometer Means for Providing a Standardized Precision Low Voltage.' Plaintiff-appellee Julie Research Laboratories, Inc., owner of the patent by assignment, manufactures and sells precision measurement devices, including a potentiometer embodying the invention set forth in the patent. Guildline Instruments markets potentiometers (and other precision instruments) made in Canada and imported into this country.1 The challenge on appeal is levied against both the holding of validity and the finding of infringement. We reach only the issue of validity.

The technique of potentiometry goes back for well over a century, and has had a number of applications and uses. In the usual classification, potentiometers fall into two modes. In the first a constant current is applied to a variable resistor; changing the resistance permits production of varying known voltages for comparison with the unknown voltage which it is desired to gauge. The second mode reverses the process by calling for application of a variable current to a constant resistor; changing the current permits production of varying known voltages to compare with the unknown voltage. An accepted method of comparison between the known and the unknown voltages is to use a sensitive meter which will read 'zero', or indicate a 'null' condition, when the known voltage equals the unknown. That is because, in that condition, no current flows in the circuit. As the trial judge put it: 'In effect, the known voltage produced by the potentiometer and the voltage of the unknown source cancel each other.'

One consistent effort over many years has been to develop ever greater accuracy in these instruments, and in relatively recent times there has been keen competition to achieve accuracy up to very small fractions of a microvolt. Precise measurements of direct current at this very low level are needed in several activities, including the space program, some other aspects of electronics, and parts of the chemical industry. The Julie patent before us concerns such a microvolt potentiometer for measuring direct current. It exemplifies the first mode described above-- a constant current source applied to a variable resistor.

Very generally, the four claims in issue use a 'feedback' type of constant current source supply presenting an essentially infinite impedance,2 and a transfer means such as a voltage-divider or possibly a series of resistors, permitting the operator, by using fixed input impedance and variable output impedance, to pass varying (but known) amounts of voltage through a constant resistor; the unknown voltage to be measured is connected to the point where the known voltage is introduced to the constant resistor; the transfer device's variable output impedance can be employed to vary the known voltage until it matches the unknown voltage on the 'null' dial and thus reveals the latter's measurement.

It is unnecessary to canvass separately each of the claim components because it is clear, and in effect admitted, that several of them are old, not new. It can also be agreed, at least for purposes of this decision, that the patent device was not wholly anticipated in the prior art. The real struggle is over non-obviousness, and that tussle relates only to two of the elements. Julie Research, representing the patentee, urges non-obviousness on the basis of the combination of (a) use of the feedback source of current, connected with (b) a transfer means using fixed input impedance and variable output impedance.3 Guildline, the alleged infringer, says that both of these elements are obvious in view of the prior art, that their combination with each other and the other components is also obvious, and therefore that the claims are all invalid under 35 U.S.C. 103.

On the first aspect-- the feedback current source-- the record shows conclusively that a feedback-type source of constant current was predicted, known, and actually used in potentiometry well before Julie's patent application in July 1965. Feedback-type constant source generators have been available for at least twenty years, but only in the early 1960's were they developed sufficiently for use in microvolt potentiometers. In 1961-62, Dr. Dauphinee, a recognized potentiometry expert, expressed the need for better constant sources and specifically referred to the probable use of feedback systems 'for stabilizing potentiometer currents.' In 1963 Julie Research and Guildline both offered feedback sources (with essentially infinite impedance) for use in potentiometry.4 The record also suggests that in the same year Minneapolis-Honeywell marketed such a source for one of its potentiometers. In an article prepared in 1964 and published early in March 1965, Dr. Dauphinee wrote that the problem of current stability needed reevaluation 'now that current controllers capable of six-figure regulation for substantial periods and seven-figures over limited periods of time are starting to come on the market.' It seems to us obvious that in 1964 and 1965 it would have been obvious to one skilled in the art of potentiometry to use such feedback-type generators, instead of batteries (with adjusters) or other constant current sources, in an instrument designed as a microvolt potentiometer. The District Court's findings do not expressly touch on the obviousness of the use of a feedback source, but if they are read to find implicitly that such use was non-obvious in 1964 or 1965 they are clearly erroneous on this record.

The rub of the case comes with the second alleged novel-unobvious feature of the claims-- a transfer means using fixed input impedance and variable output impedance. To show that this too was obvious in the light of the prior art, Guildline has proffered a reference not considered by the Patent Office, the potentiometer element or ring proposed by Diesselhorst in Germany in 1906. This ring, like the Julie patent, teaches a constant current source, output terminals of a device acting as a voltage divider, a constant or fixed resistor, and input terminals of the voltage divider.

The court below accepted the views of the patentee's witnesses that Diesselhorst discloses a transfer means having a variable input impedance and fixed output impedance, rather than Guildline's witnesses' contrary position. The court considered the issue purely one of 'credibility', citing only factors of demeanor and comparative qualifications; the findings contain no discussion of the substantive or scientific merits of the issue. We do not take that stance. The question should not turn simply on the external indicia of 'credibility'-- as if some ordinary fact like the date of a telephone conversation were involved-- but on analysis of the drawings and documents in the light of the help of the expert testimony.

On that basis, Guildline seems to us clearly right. Everyone agrees as to the physical structure of the Diesselhorst ring-- including an inner ring of resistors, nine divider resistors of the outer ring, and a tenth load resistor of the outer ring. The experts' difference lies solely in their characterization of the various resistors in the process of determining whether the input impedance of the transfer means is fixed or variable. Julie Research's witnesses counted as part of the transfer means the inner ring of resistors, along with the nine divider resistors and the additional load resistor, in coming to the conclusion that the input impedance was variable. Guildline's experts excluded the inner ring resistors and the load resistor, basing their calculations on the nine resistors of the outer ring as constituting the transfer means.

This latter position is correct because it is plain that the inner ring of resistors is not part of the transfer means and has an entirely different function having to do with maintenance of a constant source of current. Since Diesselhorst contemplated batteries as the source and a battery's current production varies with load conditions, it was necessary to provide a compensating device; the inner ring resistors fulfilled this purpose, adjusting the current supplied by the battery to the transfer means of the potentiometer so that the current would be maintained at a constant level.5 The Julie patent, having its own constant feedback source, does not need such compensators; Prof. Abramowitz, the leading expert for Julie Research, admits this at one point in his testimony.

Thus, the inner ring resistors of Diesselhorst have a role analogous to the feedback mechanism in the Julie patent, and do not have a function comparable to the latter's transfer means. Conversely, the nine resistors of the Diesselhorst outer ring are the analogue of Julie's transfer means and perform the same function. And Prof. Abramowitz himself testified that, if the nine resistors of the outer ring defined the Diesselhorst transfer means, the input impedance would be constant. He insisted, however, on including the inner ring in reaching his conclusion that in the Diesselhorst transfer means the input impedance would be variable, without (so far as we can tell) giving any reason and without acknowledging (though he does not deny, and in one exchange admits) the role of the inner ring resistors with respect to maintaining the battery flow as a constant source.

We must conclude, therefore, that the Diesselhorst teaching is a direct predecessor of this aspect of the Julie patent. That teaching remained alive, and was known and used in potentiometry, up to the time of Julie's application. In view of this accepted prior art, it was obvious to a skilled worker, in 1964 and 1965, to use a comparable type of transfer means, though perhaps with a different arrangement of equivalent parts.6 Julie Research says that it would make no sense to start with and modify a Diesselhorst ring so as to make a Julie patent potentiometer. Guildline rightly answers (reply brief, p. 7): 'One skilled in the art need not use an actual Diesselhorst ring to construct a transfer means based on the teachings of Diesselhorst and covered by the claims in issue.' For the nonobviousness criterion of 35 U.S.C. 103, the Diesselhorst teachings can and do extend beyond the specific embodiment disclosed by that inventor. It is enough that, after Diesselhorst and his followers, it would be clear to one with ordinary skill that such a transfer means would work. Similarly, as we have already indicated, it would be obvious to hook up that kind of transfer means with a feedback-type generator. Each of the two assertedly novel features-- the feedback source, and a transfer means with fixed input impedance and variable output impedance-- is obvious by itself, as well as in combination with each other and the other components of the challenged claims.

Plaintiff-appellee stresses the very great precision of its instrument and the need for an exceedingly exact microvolt potentiometer. Neither is a ground for finding nonobviousness here. The notion of microvolt potentiometers was not new; they existed before the Julie conception and the Julie instrument (though not nearly as accurate as the latter). Nonobviousness 'cannot be predicated on superiority alone. Obviousness depends on what those skilled in the art would expect' (Application of Mageli, 470 F.2d 1380, 1384-1385 (Cust. & Pat.App. 1973), and the record suggests that increasing precision was anticipated.

In any event, the more significant point is that the claims in issue do not call for the highest degree of microvolt accuracy, merely a potentiometer 'useable for precision microvolt range measurements and calibrations' (claims 5, 8, and 9) or 'useable for precision microvolt range measurement and providing standardized-precision low voltages effectively free of spurious random noise and thermal emf's' (claim 1). The invention itself is simply described in the patent as a 'Potentiometer Means for Providing a Standardized Precision Low Voltage.' The patentee cannot support the patent by the fact that its actual device, as made and sold, may be superior in ways not related to, or not shown on this record to follow from, the patent claims.

For the same reason, the fact that the actual Julie potentiometer, as fabricated, may have filled a need for an extremely accurate instrument is not in itself persuasive of nonobviousness. The Julie mechanism clearly embodies more than the Julie patent, and it has not been shown that the factors which make the Julie device popular flow from the patent claims rather than from these other aspects. Moreover, 'once it has been established by the prior art references that the difference between the patent in suit and this prior art is as unsubstantial as it is in this suit, such secondary evidence ('that the particular 'invention' before the court filled a long felt need in the particular industry and had become a commercial success') cannot be held to demonstrate lack of obviousness.' Formal Fashions, Inc. v. Braiman Bows, Inc., 369 F.2d 536, 539 (2d Cir. 1966).

In holding that the claims are void for obviousness, we necessarily disagree with the District Court's contrary result. Insofar as that conclusion stemmed from crucial factual findings-- particularly as to the Diesselhorst ring-- we have exercised our authority, as in other patent-validity cases involving obviousness, to review the record for ourselves and to determine that the findings below were clearly erroneous.7 Cf. Lemelson v. Topper Corp., 450 F.2d 845 (2d Cir. 1971) cert. denied, 405 U.S. 989, 92 S. Ct. 1253, 31 L. Ed. 2d 456 (1972); Shaw v. E.B. & A. C. Whiting Co., 417 F.2d 1097 (2d Cir. 1969), cert. denied, 397 U.S. 1076, 90 S. Ct. 1518, 25 L. Ed. 2d 811 (1970); Watsco, Inc. v. Henry Valve Co., 404 F.2d 1104 (2d Cir. 1968), cert. denied, 396 U.S. 821, 90 S. Ct. 61, 24 L. Ed. 2d 72 (1969); Taylor Wine Co. v. Celmer, 397 F.2d 784 (2d Cir. 1968), cert. denied, 393 U.S. 1016, 89 S. Ct. 620, 21 L. Ed. 2d 561 (1969); Preuss v. General Electric Co.,392 F.2d 29 (2d Cir.), cert. denied, 393 U.S. 834, 89 S. Ct. 105, 21 L. Ed. 2d 104 (1968). The ultimate decision of invalidity/validity for obviousness/nonbviousness is one of law (Graham v. John Deere Co., 383 U.S. 1, 17, 86 S. Ct. 684, 15 L. Ed. 2d 545 (1966); Shaw v. E.B. & A. C. Whiting Co., supra, 417 F.2d at 1102), and the District Court's final conclusion on that question is, of course, not binding. Lemelson v. Topper Corp., supra, 450 F.2d at 848.

Our confidence in our conclusion is strengthened by some additional factors. This is a crowded art, with several people working in it and considerable activity in the '50's and '60's. The currents and cross-currents of thought seem to have been generally available, and there was a great deal of common knowledge. It was the kind of milieu from which developments like the Julie patent and the Guildline instrument were not unlikely to emerge. Cf. Indiana General Corp. v. Krystinel Corp., 421 F.2d 1023, 1030-1031 (2d Cir.), cert. denied, 398 U.S. 928, 90 S. Ct. 1820, 26 L. Ed. 2d 91 (1970). Then, too, the patent claims, as we see them, are basically directed to a combination patent and must therefore be scrutinized with special care (General Radio v. Kepco, Inc., 435 F.2d 135, 137 (2d Cir. 1970), cert. denied, 404 U.S. 874, 92 S. Ct. 28, 30 L. Ed. 2d 121 (1971)), and the usual presumption of validity is weakened because the significant Diesselhorst ring was not considered by the Patent Office (Reeves Bros. v. U.S. Laminating Corp., 417 F.2d 869, 872 (2d Cir. 1969)).

For these reasons we reverse the judgment below with directions to enter judgment for appellant Guildline Instruments, Inc. on the ground that claims 1, 5, 8 and 9 of the Julie patent No. 3,454,877 are invalid for obviousness.8

Reversed.