NIES, Circuit Judge.
This appeal is from the judgment of the United States District Court for the Western- District of Virginia, Danville Division (Turk, C.J.), in a patent infringement suit brought by Milliken Research Corporation against Dan River, Inc. for infringement of Milliken’s U.S. Patents Nos. 3,254,510 and 3,277,673, to Lesley. The district court held both patents (the Lesley patents) invalid for obviousness under 35 U.S.C. § 103. We affirm.
I.
This litigation has been determinedly pursued and defended, virtually without surcease, since the filing of the complaint in 1970. We are advised that other- cases are suspended pending its resolution. The case was assigned to a special master who attended depositions and held approximately six weeks of hearings during the period from November 1972 through May 1973. In a thorough and detailed report, dealing with numerous defenses, the special master concluded that, while the Lesley inventions were “narrow”, the patents were valid and infringed. Particularly pertinent to our decision here is the special master’s analysis therein of a prior art reference, U.S. Patent No. 3,069,885 issued to Cooper (Cooper ’885), under which both parties to this appeal were licensed. Subsequently, the district court ordered reopening of the trial for consideration of additional evidence including a published article written by Alfred Reisfeld dealing with the state of the knitting art in early 1958. A supplementary report was filed by the special master, again concluding that the patents were valid. A third and final trial before the special master was conducted in October 1980, at which Dan River presented, inter alia, certain fabric samples and construction sheets of a third party, Celanese Corporation, (identified herein as Celanese 7182) as additional evidence of obviousness, if not a complete anticipation of the Lesley inventions. The special master again upheld th'e patents, finding that the Celanese fabrics were not proved to be prior art under 35 U.S.C. 102(b)1 and also holding that the Celanese fabrics appeared to be “of a different structure”, “in a different category” and to have been produced “by methods different from those of Lesley.”
The case again came before Chief Judge Turk on Milliken’s motion under Fed.R. Civ.P. 53(e)(2), for acceptance of each of the three master’s reports, and Dan River’s objections thereto. No additional evidence [589]*589was presented, but extensive argument and briefing was received by the court. It is readily apparent that Chief Judge Turk’s decision was the result of a most careful and thorough, indeed exhaustive, study of the record and of the numerous legal issues advanced by the parties at that time. As far as this appeal is concerned, we need only address the court’s conclusion that the Lesley inventions would have been obvious within the meaning of 35 JJ.S.C. § 103.2
In reaching its conclusion on obviousness, the district court relied on the findings of fact of the special master, denominating as “clearly erroneous” only the special master’s conclusion that the evidence was insufficient to establish Celanese 7182 as prior art.
In accordance with the analysis required under Graham v. John Deere Co., 383 U.S. 1, 17, 86 S.Ct. 684, 693, 15 L.Ed.2d 545, 148 USPQ 459, 467 (1966), the district court delineated the scope and content of the prior art, the differences between the prior art and the claims of the patents in suit and the level of skill in the art. Referring specifically to the teachings of Cooper ’885, and the Reisfeld article, the court concluded that both the Lesley method and fabric claims failed to satisfy the statutory requirements of 35 U.S.C. § 103.
II.
The patents in suit are directed to a method of fabricating knit fabrics (’673) and to warp knit pile fabrics (’510). Warp knit fabrics are produced on warp knitting machines by knitting together two or more sets (warps) of yarn. Lesley requires that the yarn of at least one of the two sets of yarn be “capable of substantial elongation”, that is, be “elastic” and that the elastic yarn be knitted under greater “tension and extension” than the second yarn during the knitting operation. The claims refer to the required tension and extension as “stretching” or “stretched.” Retraction of the- elastic yarn causes very tight stitches in the resultant fabric base and forces the other yarn, of the other warp; away from the foundation to form a “raised pile” or “raised nap” effect. .The Lesley method has the advantage of producing a raised nap or pile effect economically during knitting, without requiring post-knitting processes such as brushing or scouring of the fabric. The pile effect is on one side only and plates or covers the more elastic yarn.
The claims of Lesley ’510 for the fabric itself read:
1. A warp knit fabric having a face side formed from stitches of a first yarn and a back side formed from stitches from a second yarn, said face side stitches being knit such that they contain a length of yarn at least 40% in excess of that required for a balanced fabric construction, said second yarn being capable of substantial elastic elonagtion [sic] and being uniformly closely knit and uniformly raising said face side stitches to form said face side stitches of loosely knit yarn to provide a uniform pile effect on one face of said fabric.
2. The structure of claim 1 wherein said first yarn is capable of elastic elongation.
The method claims 1-4 of the ’673 patent specify:
1. The method of forming a raised pile surface knitting [sic] fabric with a knitting machine having two guide bars and a needle bar, comprising forming two loops from two yarns on each of a plurality of needles on said needle bar, one of said two yams on each said needle being capable of substantial elastic elongation, stretching one loop formed of said one elastic yarn during loop formation on each of a plurality of needles [590]*590while effecting substantially less stretch on the other said loop during loop formation, and permitting retraction of said substantially stretched loop after casting off thereof to thereby form a raised pile surface of said other loops, with the stretched and retracted yarn loops binding other said loops in place, and permitting the stretching during formation of each succeeding stretched loop to pull the preceding retracted said one loop tighter to thereby force the preceding adjacent said other loop further away from the foundation plane of the fabric as formed by said retracted loop.
2. The method according to claim 1 including maintaining the input tension of said one elastic yarn forming said one loop substantially above that of the yarn forming other said loop and substantially above the tension required for a balanced fabric construction.
3. , The method according to claim 1 including permitting the feed of the other said yarn to form a loop thereof of a first size upon the completion of the loop-pulling portion of the needle stroke, while permitting the feed of said one elastic yarn to form a stretched loop thereof of size substantially equal to said first loop size but of substantially smaller effective loop size than said first loop size upon permissive retraction thereof after casting off.
4. The method of forming a knitted fabric having a raised nap surface, comprising successively pulling two yarns into loops bn each of a plurality of needles and casting off said loops to form a knitted fabric, one of said yarns being elastically stretchable and being substantially stretched during loop formation while the other of said yarns is at a comparatively light tension and comparatively unstretched during loop formation, whereby the effective retracted length of such said one loop, after casting off, is substantially smaller than the effective length of each said other loop, to thereby yield an elastic foundation knit fabric having a raised nap surface formed of said other loops.
Unlike the claims to the fabric, the above claims do not specify any differential in the lengths of the two yarns, that is, a requirement for “40% overfeed.”
Much of the' argument of the parties below was directed to the meaning of claims. Chief Judge Turk found the claims considered alone ambiguous in toto. However, from reading the claims in conjunction with the specifications, he concluded, as had the master, that Lesley was claiming a pile on the lap side or technical back of the fabric despite the confusing reference to face side in the claims. We understand the back side of the fabric to be the side with arching floats or laps and it is these floats or laps which form the nap or pile.3
The court concluded: “When the patents are read as a whole, they disclose a warp knit fabric in which a lap side pile is formed as a result of knitting two warps in which the yarn fed from one warp through the back guide bars is more heavily ten’sioned or highly stretched than the other yarn.” This was the construction asked for by Milliken, and it was on this interpretation that the court based its conclusion of obviousness. No contention has been made by either party on appeal that the coqrt’s construction of the claims is error.
[591]*591An issue not before us on appeal is whether the two Lesley patents in suit are entitled to the filing date of an earlier filed application (Lesley-1). Nevertheless, the prosecution history is significant in understanding the term “elastic” in the claims, which would otherwise be the source of further confusion. In Lesley-1, claiming fabric and method, Lesley suggested use of nylon, Dacron, or Orion as the yarn capable of substantial elongation. These examples have greater elasticity than some yarns, such as cotton, but have minimal stretch capabilities as compared with others, such as spandex, which became available to Lesley after the Lesley-1 application was filed.
The references to “elastic” elongation came into the claims with the filing of Lesley-2 on May 11, 1962, denominated a continuation-in-part of Lesley-1 with claims to both fabric and method.4 The fabric claims subsequently became the subject of a divisional application (Lesley-3).
In contrast to Lesley’s use of nylon as an elastic yarn, the Cooper ’885 patent, as discussed more fully later, suggests the use of nylon in a stretch fabric as a type of inelastic yarn. Thus, elastic in connection with yarns must be understood as a relative term. Further, Milliken uses the term “elastic” in connection with fabric as a synonym for a stretch fabric, which is yet another concept of elasticity. Despite nylon being identified ás an elastic yarn in Lesley, Milliken does not assert that using nylon as the elastic yarn produces an elastic fabric. Indeed, the Lesley patents do not refer to producing an elastic or stretch fabric. On the other hand, when spandex is used in the Lesley method as the elastic yarn, an elastic or stretch fabric is produced.
The trial court concluded that the original patent application, Lesley-1, supported the patents in suit and that “it would be expected that one skilled in the art would substitute the more elastic yarns for the nylon and other yarns disclosed in.Lesley-1.” The court also commented that the different characteristics of spandex and nylon tended “to support defendant’s position ... that a fabric knitted with spandex is of a different character from one knitted completely of nylon or yarns of comparable elasticity.”
III.
Despite the years of litigation, the extensive trial period and voluminous record, as indicated, the issues have been reduced to a comparatively few before this court.
First, Milliken devotes a major portion of its briefs seeking to establish the proposition that a court of appeals directly reviews the reports of the special master, not the decision of the district court. Under Milliken’s theory, this court would not only uphold the patent, but also the master’s finding of infringement, on which the district court expressly did not rule.
Second, Milliken finds fundamental errors in the district court’s “philosophical approach” to the obviousness issue which “colored” the district court’s judgment.
Next, and more specifically, Milliken attacks (1) the district court findings with respect to what Cooper ’885 discloses, particularly that a Cooper fabric exhibits a “pile”, and (2) the court’s finding that Celanese 7182 is prior art to Lesley.
Finally, Milliken attacks the district court’s legal conclusion of obviousness, asserting that the references do not suggest “the combination of at least 40% overfeeding of the front bar yarns and a high degree of stretching of the back bar yarns” [592]*592in the absence of which “Mr. Lesley’s patented inventions are not achievable.”
Dan River essentially counters each of these arguments. In connection with obviousness, however, its argument is partly based on fabric sample exhibits of Burlington and Lesley which the district court specifically rejected as prior art. No assertion is made that the district court erred in those conclusions, nor is any type of explanation put forth as to why the exhibits should be considered by this court. Accordingly, these exhibits are not discussed herein.
IV.
. The issue of appellate review of a decision of a district court which rejects findings of a master before whom the case was tried is one of first impression in this court. Milliken would have us review the master’s reports directly; Dan River would have us review the district court only. The parties agree that courts of appeals which have addressed this issue have espoused conflicting views. Milliken relies on decisions of the Second, Fifth, Sixth, Seventh, Eighth, Ninth and Tenth Circuits;5 Dan River on the Fourth, Fifth and Eighth.6 We conclude that the position of neither party correctly defines the role of this court.
Dan River asserts, “the District Judge was in as good position as the Master to extract findings from the wealth of material in the record. In such cases the district judge may evaluate the evidence independently and is free to reach a conclusion contrary to that of the Special Master.” In our view, this premise is erroneous.
The role of the district court vis-a-vis a master is set forth in Fed.R.Civ.P. 53(e)(2), which provides:
In Non-Jury Actions. In an action to be tried without a jury the court shall accept the master’s findings of fact unless clearly erroneous. Within 10 days after being served with notice of the filing of the report any party may serve written objections thereto upon the other parties. Application to the court for action upon the report and upon objections thereto shall be by motion and upon notice as prescribed in Rule 6(d). The court after hearing may adopt the report or may modify it or may reject it in whole or in part or may receive further evidence or may recommit it with instructions. (Emphasis added).
Dan River’s position would eliminate the first sentence of Rule 53(e)(2) from our consideration in determining the propriety of the district court decision. In effect, Dan River argues that the district court conducts a trial de novo on the record. Clearly this is wrong, and it is not the role assumed by the district court in this case. Chief Judge Turk reviewed the master’s findings as an appellate tribunal under the typical restraints imposed by the “clearly erroneous” rule.
Milliken, on the other hand, would have us wholly ignore the role of the district court in the appellate process. We cannot agree with this alternative position. We review the judgment of the district court [593]*593and its underlying findings and conclusions. In reviewing any factual findings adopted or made by the district court, we must comply with Rule 52(a).7 However, the decision of the district court holding a finding of fact by the master clearly erroneous is not itself a “finding of fact”, and our review of that decision is, thus, not controlled by Rule 52(a).
A decision by us that the district court erred in setting aside a finding of the master will necessarily make a contrary finding by the district court ineffectual. But it should be kept in mind that we are evaluating the district court’s decision on an objection to the master’s report, not the district court’s substitute finding. Further, upholding the district court ruling that the master’s finding was clearly erroneous does not ipso facto uphold a substitute finding by the district court. The latter is subject to review in accordance with Rule 52(a). In instances where the district court makes findings on the basis of additional evidence, the two step procedure is apparent, but the same type of analysis applies where the court simply resorts to an existing record.
Thus, we must first review, as a matter of law, the correctness of the district court’s setting aside any factual finding by the master and, if that is upheld, review any substitute or additional findings of the district court under the “clearly erroneous” standard of Rule 52(a).8 These principles will be utilized in assessing the conflicts between the district court and the master as they are dealt with later in the opinion.
V.
With respect to Milliken’s arguments that the district judge erred in his conclusion of obviousness because of errors in his philosophical approach, this can be answered simply. This court reviews judgments not opinions, as we have previously stated on numerous occasions. Stratoflex, Inc. v. Aeroquip Corp., 713 F.2d 1530, 1540, 218 USPQ 871, 880 (Fed. Cir.1983); In re Hyatt, 708 F.2d 712, 715, 218 USPQ 195, 198 (Fed.Cir.1983). Thus, while Milliken is correct that this court has discarded the propositions that it is difficult to find “true inventiveness”, i.e., non-obviousness, in a combination patent, Environmental Designs v. Union Oil Co. of California, 713 F.2d 693, 218 USPQ 865 (Fed.Cir.1984), and that if it was obvious to try combining elements, the combination itself must be obvious, Jones v. Hardy, 727 F.2d 1524, 220 USPQ 1021 (Fed.Cir.1984), Milliken must still convince us the district court reached a wrong result, which it has wholly failed to do. Union Carbide Corp. v. American Can Co., 724 F.2d 1567, 220 USPQ 584 (Fed.Cir.1984).
Milliken also asserts that the district court plainly erred in failing to evaluate each claim of its patents separately. Mil-liken notes that while the district court correctly observed that “Lesley’s inventions relate ... to the stretching of the back bar yarn,” it failed to appreciate that the ’510 fabric patent, as distinguished from the ’673 process patent, is directed to a fabric .having “the combination of not' only a stretched back bar yarn but also at least a ^0% overfed front bar yarn.”9
While we agree with Milliken that claims must be evaluated separately, Med[594]*594tronic, Inc. v. Cardiac Pacemakers, Inc., 721 F.2d 1563, 1582, 220 USPQ 97, 111 (Fed.Cir.1983), rather than as a single unit of claims tied together or “in tandem”, Milliken has failed to show how the district court’s treatment of the claims worked to its prejudice. Given the confusion as to what was claimed, Chief Judge Turk simply gave Milliken the benefit of all possible limitations in all claims, and, nevertheless, held the Lesley patents invalid.
Finally, Milliken asserts that the district court failed to understand that a product may infringe two valid patents simultaneously. In this case, Milliken asserts that the district court failed to appreciate that Dan River’s accused fabric, which exhibits a pile, though licensed and made under the Cooper ’885 patent, could still infringe the Lesley patents.
We, of course, agree with Milliken that a product may infringe more than one patent and that one may not be able to practice the invention protected by a patent directed to an improvement of another’s patented article or method except with a license under the latter. However, we do not discern from Chief Judge Turk’s analysis that he misconceived the law in this respect. Rather, he found support for his conclusion that the Lesley inventions would have been obvious from Cooper ’885 (and other prior art) in the facts that the accused fabrics were licensed under Cooper ’885, were made without knowledge of Lesley’s patents, and showed a “pile.”
VI.
A.
The Prior Art
The most pertinent prior art as found by the district court is Cooper ’885, a reference not considered during prosecution. The question whether Cooper ’885, which was a pending application at the time Les'ley-l w„as filed, is prior art is no longer at issue.
Cooper ’885 discloses the production of a warp knit stretch fabric consisting of a spandex foundation “plated”, that is, covered, on both sides by a relatively inelastic yarn, such as nylon. Cooper was directed to utilizing spandex yarn in a manner which concealed certain undesirable characteristics of the bare spandex. Cooper accomplished his purposes by knitting spandex with comparative inelastic yarns with more desirable aesthetic qualities. The patent disclosure does not expressly discuss runner length ratios, overfeed, or pile.
The court made the following findings with respect to Cooper ’885:
Cooper ’885, as noted previously, discloses a warp-knit two-way stretch fabric consisting of an elastic yarn — spandex— plated between a technical back and front consisting of an inelastic yarn, such as cotton, nylon, or others similar. The patent, which resulted from Cooper’s experimentation as documented in his notebook (D-33) represents an attempt— which has been very successful commercially — to utilize spandex yarns in a manner which conceals certain of their undesirable characteristics: they are clammy to the touch, and they tend to yellow with age. Plating accomplishes this design by insulating the spandex yarn from the wearer’s touch, and by hiding the spandex behind yarns with more desirable aesthetic qualities. The yarns are knitted in opposite directions. The patent’s example discloses nylon on the front warp in a 1-0, 1-2 pattern, and spandex on the back bar in a 1-2, 1-0 pattern. ,No runner ratios or tension differentials 25 are specified, although the specification provides that “the amount of compression in the finished fabric may be readily controlled by those skilled in the art.” Col. 3, lines 73-75. The patent also discloses that “[a] balanced lapping movement is used with the two sets of guids [sic] bars making similar movements in opposite directions.” Col. 3, lines 10-13. Filially, the specification provides:
In order to provide an acceptable fabric having a plated construction, i.e., with the yarns of the back guide bar [595]*595being covered by the yarns of the front guide bar on both the face and back of the fabric, the ratios of the original relaxed deniers of the elastic yarns to the inelastic yarns must be below about 4:1. At ratios above this limit, incomplete covering of the bare elastic yarns results. Denier ratios between about 4:1 and 1:6 are preferred. Col. 3, lines 33-40.
Although Cooper does not inherently call for a high degree of stretch of the spandex yarn, it is obvious, from reading the patent, that such a degree of stretch is contemplated; otherwise, the desired plating effect would not be achieved. Further, the patent specifically allows for experimentation by the knitter to achieve the proper “amount of compression.” Col. 3, lines 73-75.
Although plating, and not a pile surface, is called for, the pile is inherent in the plating process.
The master made the following findings, which may add to understanding Cooper ’885:
26. Beginning in 1958, Cooper' engaged in experiments to develop a tricot warp knitted fabric which would have two-way stretch and would also not feel clammy to the wearer (clamminess being a non-desirable characteristic of Lycra [i.e. spandex]); and his solution was to sandwich the Lycra yarn between layers of a non-clammy yarn such as Nylon.
27. This was done by putting Nylon on the front bar and Lycra (or Fiber K as it was known in the experiments of Cooper) on the back bar of a warp knitting machine, or more specifically, a Tricot two-bar machine.
30. Elastomer (or elastomeric) yarns [i.e. spandex] were becoming known in the trade in the early years of the 1960 decade.
31. The same tension on an elastomeric yarn and a Nylon yarn would stretch the former more than the latter.
35. Cooper expected skilled knitters to make adjustments in denier, in knitting construction, and in knitting tensions to produce fabrics which would have the desired aesthetics, characteristics, and compression.
36. Retraction or contraction of the spandex yarn brings the Nylon yarn to the surfaces of the fabric under the Cooper practice. (Emphasis added.)
7. The prior art referred to does not invalidate the putative inventions.
h. The prior art defense is based principally on Lawson, Cooper, the Austrian Publication, two references considered by the Patent Office (Fletcher and Jennings), and also on the Lesley on-sale fabric. Taking these in order, they appear to be distinguishable from the development of the patents in suit:
2) The Cooper patent is for a plated-fabric, developed largely to assist in the commercial -exploitation of duPont’s elastomeric yarn (Fiber K) later given the name of Lycra [spandex]. The great advantage of Lycra is its elasticity, which it imparts to knitted fabrics; its drawbacks are its clammy feel and its tendency (later reduced) to become yellow. The principal object of Cooper and his associate was to cover the Lycra with Nylon (or some other suitable yarn), which would feel good to the wearer and which would not become yellow with age. Even though Cooper himself realized that a pile of the covering yarn might be a greater [596]*596plating protection, since as the fabric was stretched the arched loops would be drawn down to the plane of the fabric, his patent does not call for a pile; and he himself denied that he had created a “true pile” in the fabric a sample of which was attached to p. 41 of his notebook (Vol. VII, p. 94). In addition, the literature distributed by duPont to its licensees under the Cooper patent, while it discusses runner lengths, does not give any instructions for making a pile fabric. Cooper especially emphasizes the relative deniers of the yarns fed to the front and back bars, it being necessary for proper plating that the front bar yarn be of a sufficiently large denier, as compared with that of the back bar Lycra, so that the latter would be completely covered. He does not disclose the production of the lap pile or a lap pile fabric as specified by the patents in suit. [Second emphasis added.]
We discern no disagreement in the above quoted findings of the court and the master on the evidence or in what Cooper ’885 discloses. The district court’s finding that Cooper ’885 inherently produces a pile does not conflict with the master’s report and Milliken does not assert that it does. Rather, Milliken asserts that the district court finding that a pile is inherent in Cooper ’885 is inconsistent with other statements in the district court opinion. Again we are faced with a confusing term — in this instance “pile”.
To illustrate, the record discloses the following testimony with respect to whether Celanese 7182 exhibited a “pile.” Dr. Brown, Milliken’s expert, testified:
[T]he fabric itself upon examination does not show a'pile at all let alone a uniformly closely knit pile.
sjs sfc sj< sfc # sfc
In my opinion if there is any elevation of the laps away from the back fabric it is so minimal that I would certainly not describe it as protruding from that surface. .
Dan River’s witness, Beard, on the other , hand, testified:
Q. How can you tell by looking that it [Celanese 7182] is overfed?
A. Well, if you look at it like this and you turn it like I am turning it here and look at it, you can see the pile sticking up.
Q. And that is what, 7182?
A. Yes.
Q. I wonder if you might show the Judge [special master] just what you have done so that he will know. Would you please describe it to him so that he will know what you are doing?
A. What we are going to do is I am going to show you that there is more yarn in the back; this is not a tight stitch.
Now if you turn this over and you turn it like this, you can see how those little loops stick up; see that loop sticking up there?
THE SPECIAL MASTER: Yes.
THE WITNESS: Now if you stick a pin in there, it can go right through there and right up there easily.
THE SPECIAL MASTER: Yes.
Q. And that is on which side of the fabric?
A. This would be on the technical back.
Q. And is that a pile effect or a novelty fabric effect?
A. That would have been a pile effect.
Milliken provides the following definition for the “raised nap or pile effect” called for by its patent: “Raised, arching laps, protruding, and thus substantially spaced from, the closely knit foundation of a warp knit fabric.”
We understand from the above that the extent to which an arching lap must protrude to be a pile is not a precise amount, particularly where the fabric is of such a close-knit construction that to the casual untrained observer, the fabric appears to be flat.
[597]*597To add to the confusion in use of the term, Dan River asserts that a pile in a stretch fabric (e.g., made of spandex and nylon) is not the permanent pile called for by Lesley, but rather is a transitory pile. “Transitory” to Dan River means the pile exists when a stretch fabric is relaxed, but when the fabric is stretched, the laps forming the pile tend to flatten. Thus, Dan River asserts that the district court correctly said that a Cooper ’885 fabric had a pile in this sense, even though Cooper did not “call for” a pile or did not intend to create a pile, or in Cooper’s words was not a “true” pile. The pile simply is there, according to Dan River, because with the spandex in a related state the excess nylon must push up even without overfeed. Mil-liken, on the other hand argued, “The Lesley pile is precisely what the Dan River pile is. It’s this kind of pile. It’s always there when the fabric retracts again.” Thus, we understand that in Milliken’s view, Dan River’s “transitory” pile is a “permanent” pile and the district court was simply wrong in saying pile was inherent in Cooper ’885, which only calls for plating.
A final matter adding to the confusion: Milliken described Lesley’s inventions as “plating” with a “pile”.
The second reference discussed by Chief Judge Turk as prior art is an exhibit consisting of fabric swatches and instruction cards identified as Celanese 7182. Milliken contests reliance on Celanese 7182 as a prior art reference. Milliken also contends that, if Celanese 7182 is prior art, the disclosure is remote from Milliken’s inventions.. Milliken does not, however, dispute that Celanese 7182 teaches overfeed of the front bar yarn relative to the back bar yarn, albeit, in the amount of 25 percent rather than the 40 percent purportedly required to achieve Lesley’s pile.
In connection with Celanese 7182, we are faced with conflicting determinations by the trial judge and the special master, although not to the extent indicated in Milliken’s brief.
Milliken quotes extensively but selectively from finding 10 of the special master, attributing to Celanese 7182 the specific details of construction described in the finding, Particularly, Milliken asserts the master found that Celanese 7182 does not exhibit a “pile.” As presented in Milliken’s brief, the master purportedly found with respect to Celanese 7182:
10. ... Brown testified generally and specifically that the ... Celanese yarns failed to comply with all the elements called for by the Lesley patents in suit. The fabric swatches ... were inspected ..., and it appeared to the undersigned that none of the fabrics approached the binding in place or the raised surface or uniform pile effect called for by Lesley; on the contrary, the fabrics appeared loose and the laps protruded or rose very little in any of the fabrics, less in some than in others. The elements of Patent 3,254,510 in suit (510) calling for one yarn to be uniformly closely knit and uniformly raising stitches of the other yarns to provide a uniform pile effect on one face of the fabric, read in the light of the specification of that patent, did not appear to be met by the fabrics that were inspected ... by the undersigned. [Emphasis added.]
The above quotation gives an inaccurate impression with respect to the finding by the master. The second and third sentences of the above quotation relate specifically to a Burlington fabric and other fabrics inspected under a microscope. That part of the sentences following the word “inspected” has been deleted.' Moreover, in finding 15, the master specifically stated that finding 10 applied generally to the Celanese fabric except that he “had no recollection of viewing Celanese under a microscope.” Our review of the record •confirms that, at least during the trial, the master did not make a microscopic examination of Celanese 7182. Thus, with respect to the Celanese fabrics, the district court aptly adopted only the master’s more general finding 20, which is as follows:
20. Considering the testimony as a whole and the appearance of the fabrics [598]*598as observed without magnification and the appearance of some under magnification, the Burlington and Celanese fabrics appear to be of a different structure and in a different category from the fabrics called for by Lesley Patent ’510, and to have been produced by methods different, and in a different category, from those called for by the Lesley Patent ’673, taking into consideration the elements of the claims of both patents read with (but not enlarged by) the specifications.
Thus, as far as the disclosure of Celanese 7182 is concerned, the master and the trial court had no disagreement. The district court accepted that the Celanese fabric, which was asserted as a complete anticipation, was different. Contrary to Milliken’s assertion, the master did not make any findings with respect to the disclosure of Celanese 7182 not adopted by the district court. Milliken’s elliptical presentation of the record was a serious disservice to this court.
The conflict between the district judge and the special master arises over whether Celanese 7182 should be considered at all. At the third trial, the master was presented with Celanese 7182 along with certain Burlington fabric exhibits and made the following pertinent findings of fact:
1. The Defendant introduced, as prior art and to show obviousness, experimental fabric samples (with construction sheets relating thereto) produced during the period from 1952 to 1958 by Burlington Industries, Inc., identified as numbers DX-285-291, inclusive; and two experimental fabric samples (with construction sheets) produced prior to 1954 by Celanese Corporation of America, identified as cards numbers 7165 and 7182, each contained in a book of Celanese fabrics with accompanying construction sheets, a copy of which book was furnished to a subsidiary of the Plaintiff and is identified as DX-298, and a copy of which was furnished to Burlington identified as DX-299.
* * * # * *
5. ... On cross-examination, Beard [an employee of Milliken subsidiary] testified that he had stated that it was impossible for him to say when he had received the Celanese cards (T., pp. 153-54)....
sjs # Jfc # * *
11. The Celanese cards numbers 7165 ands 7182 contained fabric samples and construction sheets relating respectively thereto, which would enable a knitter to produce the fabrics.
% sis # sjs sfc *
13. The Celanese fabrics were disclosed to Burlington and Milliken primarily to foster the sale of Celanese yarns. Although Celanese for a few years manufactured and sold fabrics, it was not established that the fabrics shown on cards 7165 and 7182 were on sale more than one year prior to April 6, 1959.
14. Both fabrics had overfeeding of the front bar, but neither as much as 40%, number 7182 having an overfeed of 25% and number 7165 of 7%.
16. Beard testified that a pile could be seen on number 7182, and extracted from the fabric what he termed a loop. Taylor described number 7182 as slightly overfed on the front bar, said that there was a slight unbalance in the fabric with the back bar yarn fairly taut and the front bar slightly looser, with a very slight protrusion of the front bar yarn. He candidly described number 7165 as almost a balanced fabric, and for all practical purposes a balanced fabric. Brown, noting that neither fabric had as much as a 40% overfeed, testified that the overfeed of 25% in 7182 went into both the overlap and the underlap portions of the front bar stitches, that the fabric showed no pile, or so minimal a protrusion of the excess yarn that it could not be described as protruding, and otherwise placed both fabrics beyond the scope of the patents in suit.
17. Burlington and Celanese engaged in experimental work in the production of [599]*599various kinds of fabrics. and the use of various types of yarn.
As a conclusion of law, the master held:
2. The evidence shows that there was some likelihood that some of the experimental fabrics were shown to customers or potential customers of Burlington and Celanese, at about the time that they were produced in the period from 1952 to 1958, but as a whole is insufficient to establish that the fabrics were on sale more than a year prior to April 6, 1959, when the Lesley application was filed. [Emphasis added.]
The district court adopted the master’s findings and conclusion with respect to the Burlington fabrics. With respect to ■ Celanese 7182, the court found error in the master’s conclusion that the evidence was insufficient to demonstrate that Celanese 7182 had not been “on sale or in public use.” (Emphasis added.) The district judge then went on to hold that Celanese 7182 had been “in public use.” Directly compared, it can be seen that the district court did not, in fact, overturn the master’s findings and conclusion with respect to the asserted “on sale” bar. The district court relied on a different statutory ground, namely, “public use.” However, since the court treated the master’s findings and conclusion as effectively dealing with the “public use” defense as well, we will do the same.
The district court concluded that the master erred since it was “undisputed that the [Celanese] fabrics were furnished, in D-298 and D-299 [i.e., the Celanese notebooks], to third persons, and that no restrictions were put on their use.” The court rejected the argument that such distribution could be for “experimental purposes” since the invention was out of the hands of the inventor and any testing was not “under his direction”, citing City of Elizabeth v. American Nicholson Pavement Co., 97 U.S. 126, 134, 24 L.Ed. 1000 (1877).
With respect to the date they were received, the district court said there was “no significant question about whether they were received by Celanese customers prior to 1959.”
Milliken argues here that the master’s finding was directed to the lack of evidence to establish the date of receipt by third parties. However, like the district court, we read the finding of the master more broadly. The master stated that there was only “some likelihood” that the “experimental” Celanese 7182 fabric was shown to Celanese customers. That the evidence was insufficient to prove third parties had been shown Celanese 7182 appears to us, as it did to the district court, to be the basis for the master’s treatment of Celanese 7182 as “experimental” fabrics, and we agree with the district court that this finding was clearly erroneous. To view the record as a whole as establishing no more than “somé likelihood” that these customers were shown Celanese 7182 ignores the fact that the exhibits were obtained from the customer’s files. Thus, the district court’s rejection of the theory that the Celanese fabrics were “experimental” in a legal sense cannot be faulted.
Turning to the district court’s alternative findings, on which it based its conclusion that Celanese 7182 was in public use more than one year prior to the filing, Milliken bears the burden on appeal of establishing clear error. Lear Siegler, Inc. v. Aeroquip Corp., 733 F.2d 881, 885, 221 USPQ 1025, 1030 (Fed.Cir.1984). We have no quarrel with the court’s pragmatic evaluation of the evidence.
As an initial matter, from a reading of the testimony, the impression is gained that a pre-1959 date for Celanese 7182 was not seriously in dispute, as stated in the district court’s opinion. Voluminous testimony was presented directed to whether the Celanese 7182 instructions were sufficiently detailed to be able to produce the fabric and to whether Celanese 7182 fabric showed a lap side pile. Indicative that the date was not a critical issue, for example, is the following exchange between counsel for Dan River and Milliken’s witness Dr. Brown:
[600]*600THE WITNESS: ... [T]he only difference I can see at all between these two sample cards [of Celanese fabrics] are the dates on them. One is dated May 25, 1954 and. the other is dated June 22, 1954.
t)c % :{s s}c sjt !{:
Q. They are complete in all details [for making the fabric] as far as you are concerned?
A. I can’t bring anything more to mind now that would be needed.
Q. As far as the date, I think we have already covered that.
A. As I said, the only difference is the date and that is the transmittal date of some kind.
While neither recipient of Celanese 7182 could recall the precise date on which it was received, this is not a fatal defect of proof. At the time of receipt, Celanese 7182 was no more than another sheet to be added to the Celanese notebook of the kind Celanese regularly sent out. There is no suggestion that the dates on the exhibits were altered or added later, or that the dates bear no relationship to reality or the date of manufacture. Thus, we do not find that the factual inference that the exhibits were “received prior to 1959,” made by the district court from the evidence with respect to the Celanese notebooks, to be clearly erroneous. Celanese 7182, accordingly, will be considered on the issue of obviousness.
The principal remaining prior art is the Reisfeld article. Reisfeld wrote a series of articles, containing an extensive summary of the knitting art, which was published on February 24, 1958, more than a year before the first Lesley patent application. Reisfeld describes warp knit fabrics and discloses the production of long floats on the lap side by overfeeding the front bar. One example given by Reisfeld demonstrates a 12 percent overfeed; but Reisfeld does not confine himself to that percentage as a maximum. In more general terms, Reisfeld describes overfeeding one yarn of a two-yarn fabric and states:
Thus, if on a two-bar fabric the runner of one bar is made very long, the laps of that bar will assume a form of loose floats or even loops and protrude from the surface of the fabric. This is sometimes made use of for obtaining novel texture fabrics or for production of towel cloths. The loose floats may even be napped to convert them into a deep pile. [Emphasis added.]
Reisfeld then describes creating the rough texture suitable for towelling by using cotton on both the front and back bar.
In another section of his article, Reisfeld discusses warp tension as a means for creating pattern effects although, admittedly, he made no attempt to discuss it in detail because of its complexity. He did state, however, the following:
One beam threaded 6 in, 3 out is kept under normal tension, while the other, threaded 6 out, 3 in has its warp sheet maintained under a strong tension by heavy weights. As a result of tension differential the wales produced by the taut ends stand up from the surface of the fabric lending it a three dimensional effect____
Its principle involves subjecting selected groups of warp threads to fluctuating tension in a manner where one group of threads is made taut while the other slack. [Emphasis added.]
B.
The Differences between Lesley and Prior Art
Based on the foregoing references, and others which were deemed cumulative, the master and the district court were in agreement that the differences between Lesley and the prior art were twofold: (1) feeding of at least 40 percent more yarn than conventional through the front guide bar than through the back bar,10 and (2) substantial[601]*601ly greater stretching of the elastic back bar yarn as compared with the stretching of the front bar yarn.
Apart from stretch and overfeed, the district court held that the “remaining disclosures [limitations] in the claims either simply describe warp knitting or describe effects which flow from the stretching differential.” This is not disputed and, in any event, we agree.
C.
Level of Skill
With respect to level of skill in the art, there is also no dispute that, as found by the master:
63. The knowledge and skill of a skilled knitter are of a high caliber. He is able and expected to adapt and supplement general instructions in the construction sheet for a particular fabric in its manufacture.
VII.
From the teachings of the prior art references, we agree with .the district court that the Lesley inventions would have been obvious to one of ordinary skill in the art.
Stretching and retraction of one loop is the import of Cooper ’885 which expressly teaches that the back bar yarn should be capable of substantial elastic elongation. Although Cooper does not specify a high degree of stretch of the spandex yarn, nevertheless, as the district court found, “a degree of stretch is contemplated; otherwise, the desired plating effect would not be achieved.” [See also special master’s finding of fact 36: “Retraction or contraction of the spandex yarn brings the Nylon yarn to the surfaces of the fabric.”] Whether Cooper ’885 fabric does or does not exhibit a “pile” in its relaxed state is not the issue, and this court has not attempted to re-resolve a factual matter which Milliken tells us would require microscopic comparison of the exhibits and then be understood only with expert guidance. The method of knitting determines the effect achieved. Accordingly, we will look to the words of the references and to the arguments.
The necessity for stretching the back bar yarn, Milliken tells us, is that upon retraction, the elastic yarn loop will then create tight stitches which forces the arching lap portion in the stitch being made in the other yarn away from the foundation to form the pile effect. Cooper ‘885, thus, must also have tight stitches, as the district court found. Otherwise, the plate would hot be brought to the surface, or, to paraphrase the above description, “away from the foundation to form the plate effect”.
There is no question that several prior art references teach overfeed, not found in Cooper. Reisfeld, in particular, was relied on by Chief Judge Turk, and “other prior art references,” presumably meaning Celanese 7182 which calls for overfeed of 25 percent.
Milliken argues before us, as quoted at the beginning of this opinion, that 40 percent overfeed is a significant minimum in [602]*602the Lesley invention. However, the district judge found the 40 percent cutoff point chosen by Lesley appears to be arbitrary, relying upon a fabric [D39] as support for Dan River’s contention that nothing particularly unusual occurs when the overfeed is 40 percent. Moreover, the court noted, “According to Dr. Brown, Lesley must have ‘thought that at 40 percent these fabrics were commercially acceptable fabrics and that at less than' 30 percent they were not.’ ” Also, the court noted that Dr. Brown did not know what, if anything, unusual would occur at the 40 percent overfeed level. The court specifically rejected Dr. Brown’s opinion that “the Reisfeld article and other prior art disclosing a lesser degree of overfeed teach" away from Lesley by suggesting that higher degrees of overfeed are not possible.” 11
Milliken argues that the prior art does not suggest a combination of overfeed and stretch as in Lesley. As Milliken would have it, the combination would not have been obvious in the absence of an express suggestion in a prior art reference. This is clearly not the law. “The proper test is whether the references, taken as a whole, would suggest the invention to one of ordinary skill in the art.” Medtronic, Inc. v. Cardiac Pacemakers, Inc., 721 F.2d at 1581-82, 220 USPQ at 110.
In any event, Milliken’s argument must be premised on exclusion of consideration of Celanese 7182 which is a single reference which teaches one of skill in the art to combine overfeed of one yarn and greater tension on the other. In delineating the teachings of Celanese 7182, out of the voluminous record, the district court specifically noted the testimony of Milliken’s own expert that the instructions of Celanese 7182 meant the back bar yarn would be knitted under greater tension than the front bar yarn. Thus, overfeed and different tensions are found in a single reference.
■Milliken asks that we find legal error in the district court’s conclusion on obviousness because the court said it was somewhat inconsistent for Milliken to suggest that the Cooper patent would not have made Lesley obvious and that Dan River’s accused fabrics, though plated, would infringe Lesley because they have pile. We find in the trial judge’s analysis only an appreciation that the breadth of the Lesley claims as asserted against the infringers makes Cooper ’885 a much more pertinent reference than any considered by the Patent Office. Milliken rejects that the Lesley inventions are limited to the specific embodiments described in its patent applications, using nylon for the elastic yarn. Had that view been urged, the validity of Milliken’s patents could well have been affirmed by the court below. However, with that reading of the patents, Milliken co.uld not now reach the infringers, and this is what we discern from the trial judge’s analysis.
The district judge accepted the clear responsibility on him to resolve the question of obviousness. With respect to that issue, he was in no way bound by the conclusion of the special master that Lesley had obtained valid, though narrow, patents.
The district court did not choose to discuss all references which would lend support to its conclusion on obviousness, nor to address all issues, recognizing the possibility that in a case of this magnitude, an issue of controlling importance, in the view of the parties, may have been omitted. The highly evident thoroughness of the court’s attention to the case, however, makes the latter possibility unlikely. The parties here mainly dispute the court’s legal conclusion on the obviousness of the inventions. Based on the same art considered by the district court, we are led to the same conclusion that the claims of the Lesley patents would have been obvious within the meaning of 35 U.S.C. § 103.
[603]*603VIII.
For the foregoing reasons, the decision of the district court holding U.S. Patents 3,254,510 and 3,277,673 invalid is affirmed. The parties shall bear their own costs.
AFFIRMED.