PAULINE NEWMAN, Circuit Judge.
Tandon Corporation appeals the final decision of the United States International Trade Commission, which held that Tan-don's U.S. Patent No. 4,151,573 (“the ’573 patent”) was not infringed by certain imported double-sided floppy disk drives and therefore that there was no violation of section 337 of the Tariff Act of 1930 as [1019]*1019amended, 19 U.S.C. § 1337. In re Certain Double-Sided, Floppy Disk Drives and Components Thereof, 229 USPQ 968 (USITC 1986). We affirm, on the basis that substantial evidence supports the Commission’s finding of noninfringement.
Standard of Review
In this case that turns on the factual question of infringement, which in turn requires decision of factual questions of technological equivalency as well as the factual underpinnings of claim interpretation, we once more remark on the standard of appellate review that is set by the governing statutes. 19 U.S.C. § 1337(c) provides for appellate review “in accordance with chapter 7 of title 5”. 5 U.S.C. § 706, subparagraph (2)(E), imposes the “substantial evidence" standard of review on Commission findings and conclusions. There is a significant difference between the standards of “substantial evidence” and of “clearly erroneous”, and in close cases this difference can be controlling. See, for example, R.L. Stern, Review of Findings of Administrators, Judges and Juries: A Comparative Analysis, 58 Harv.L.Rev. 70, 80-89 (1944), in which was stated, “Policy, authority and history all thus show that the ‘clearly erroneous’ rule gives the reviewing court broader powers than the ‘substantial evidence’ formula,” at pages 88-89. See also Consolo v. Federal Maritime Commission, 383 U.S. 607, 619-20, 86 S.Ct. 1018, 1026, 16 L.Ed.2d 131 (1966).
It was the intent of Congress that greater weight and finality be accorded to the Commission’s findings as compared with those of a trial court. See S.Rep. No. 466, 96th Cong., 1st Sess. 26 (1979). Indeed, the Senate Report criticized the Court of Customs and Patent Appeals’ use of a more rigorous standard of review on the question of obviousness (i.e., was the decision “clearly contrary to the weight of the evidence”) in Solder Removal Company v. U.S. International Trade Commission, 582 F.2d 628, 199 USPQ 129 (CCPA 1978). Id.
At the same time, the Senate Report accompanying the Trade Act of 1974 made clear that the Commission’s primary responsibility is to administer the trade laws, not the patent laws:
[I]n patent-based cases, the Commission considers, for its own purposes under section 337, the status of imports with respect to the claims of U.S. patents. The Commission’s findings neither purport to be, nor can they be, regarded as binding interpretations of the U.S. patent laws in particular factual contexts. Therefore, it seems clear that any disposition of a Commission action by a Federal Court should not have a res judicata or collateral estoppel effect in cases before such courts.
S.Rep. No. 1298, 93d Cong., 2d Sess. 196, reprinted in 1974 U.S.Code Cong. & Admin.News 7186, 7329. Thus, our appellate treatment of decisions of the Commission does not estop fresh consideration by other tribunals. See Lannom Manufacturing Co., Inc. v. U.S. International Trade Commission, 799 F.2d 1572, 1577-78, 231 USPQ 32, 36 (Fed.Cir.1986).
Background
Tandon alleged unfair practices based on importation by the respondents of certain double-sided floppy disk drives in infringement of certain claims of the ’573 patent; the sale of the accused devices was asserted to have the effect or tendency to destroy or substantially injure an efficiently and economically operated industry in the United States. 50 Fed.Reg. 4276 (1985). The respondents were Mitsubishi Electric Corporation, Mitsubishi Electronics America, Inc., TEAC Corporation, TEAC Corporation of America, Sony Corporation, and Sony Corporation of America. Following a twelve-day hearing the Commission issued a temporary exclusion order. In re Certain Double-Sided Floppy Disk Drives and Components Thereof, 227 USPQ 982, 991 (USITC 1985). Before trial on the question of permanent relief Tandon settled with and granted licenses to all respondents except the two Mitsubishi companies (hereinafter “Mitsubishi”).
The ’573 patent is entitled “Magnetic Recording Device for Double Sided Media”, [1020]*1020and lists inventors Sirjang L. Tandon, Alfred C. Hackney, and Roy A. Applequist. The claimed invention is an apparatus whereby a pair of magnetic heads, also called transducers, receive and transfer information from and to both sides of a floppy disk.1 A drive mechanism moves the transducers along the disk radius to access its concentric data tracks as the disk rotates. The Tandon device was illustrated in the ’573 patent as:
[[Image here]]
Tandon’s first or lower transducer is shown at (14) as a “button head” transducer. As described in the specification, this transducer is fixed to the carriage, and the upper transducer is gimballed.
Early floppy disks stored information on only one side, using a single rigidly mounted transducer for data transfer. Opposite the single transducer was a pad that pressed the disk against the transducer and absorbed disk irregularities. The first double-sided floppy disk drive was introduced in 1976 by the International Business Machines Corporation. The IBM disk drive had two symmetric cantilevered transducers, one for each side of the disk, both supported on spring suspensions referred to as gimbal2 springs. The gimbal springs enabled the heads to rotate around the X, and Y axes, and move in the normal direction along the Z axis3. The heads thus followed the irregularities of the disk.
The Tandon disk drive of the ’573 patent was stated to solve certain technological problems that inhered in the IBM drive, particularly misalignment between the heads and disk tracks due to the weak suspension that was designed to compensate for waviness of the disk in the Z direction. Tandon describes its invention as the first successful solution to use of double-sided floppy disk drives, a major factor in the computer revolution, and widely licensed. The Tandon drive uses, as described in the specification, a fixed transducer on one side of the disk and a movable transducer mounted on a gimbal spring on the other side. The movable transducer forces the magnetic disk against the fixed transducer, thus reducing the Z-direction movement of the disk and improving the accuracy of reading/writing to the disk.
The accused Mitsubishi devices are double-sided disk drives which contain two gimballed transducers that, according to the findings of the Administrative Law Judge (“ALJ”), are more stiffly mounted than the IBM transducers, but neither of which is as rigidly fixed as the lower transducer of Tandon. Mitsubishi’s lower transducer has a load point which, according to the findings of the AU, somewhat limits its movement in the Z-direction. Tandon asserts that the stiffness of the gimballed transducers and the limitation on Z-axis [1021]*1021movement brings the Mitsubishi device within the scope of the ’573 claims.
Following the trial on permanent relief, in which Tandon, Mitsubishi, and the Commission’s investigative staff participated, the AU found that the Mitsubishi disk drives did not infringe claims 1, 5, and/or 12 of the ’573 patent, which were all of the independent claims at issue, and that injury had not been proven. The Commission affirmed the ALJ’s determination of non-infringement, and expressly took no position on the question of injury.
Claim Interpretation
The principal issue in the Commission’s infringement analysis was the degree of “fixedness” of the Mitsubishi lower transducer. Tandon argues that the claims are infringed because the Mitsubishi lower head is “fixed” in the Z or normal direction, which Tandon asserts is the only fixedness that the claims require, and that measurements show that the Mitsubishi lower head is as “fixed” as the Tandon lower head.
The Commission thus undertook to interpret the term “fixed” as used in the claims. The Commission concluded that the claims mean that the lower head is fixed not only in the Z direction, but also in the X and Y directions. Tandon ascribes error to this claim interpretation.
Claim interpretation is a question of law, having factual underpinnings. When the meaning of key terms of claims is disputed, as in this case, extrinsic evidence may be adduced including testimony of witnesses, and reference may be had to the specification, the prosecution history, prior art, and other claims. H.H. Robertson, Co. v. United Steel Deck, Inc., 820 F.2d 384, 389, 2 USPQ2d 1926, 1929 (Fed. Cir.1987); SRI International v. Matsushita Electric Corp. of America, 775 F.2d 1107, 1117 n. 11, 1118, 227 USPQ 577, 582 n. 11, 583 (Fed.Cir.1985). Claims may not be construed one way in order to obtain their allowance and in a contrary way against infringers. Autogiro Company of America v. United States, 384 F.2d 391, 398-99, 181 Ct.Cl 55, 155 USPQ 697, 703-04 (1967). When the interpretation of claims requires findings of underlying fact, those factual findings are reviewed in accordance with the appropriate evidentiary standard, i.e., that of substantial evidence. Texas Instruments, Inc. v. U.S. International Trade Commission, 805 F.2d 1558, 1562 n. 2, 231 USPQ 833, 834 n. 2 (Fed.Cir. 1986).
The Commission's claim interpretation placed great weight on Tandon’s prosecution history in the Patent and Trademark Office. Amendments to claims 1, 5 and 12 were made by Tandon in order to distinguish the claims from a cited reference describing the IBM drive system. The emphasized phrases represent material added by these amendments:
1. A device for maintaining a pair of magnetic transducers in operative relation with both sides of a non-rigid planar magnetic recording media comprising:
a first transducer mounted relative to a first side of the media and having a fixed position in a direction normal to the plane of the media despite movement to different positions along the plane of the media, the first transducer being disposed in data transfer position relative to the media;
a support mechanism adjacent to the second side of the media in a region opposite the first transducer;
a second transducer coupled by gimbal support means to the support mechanism in opposition to the first transducer and movable toward and away from the plane of the media; and
means coupled to said support mechanism for urging said second transducer toward said first transducer and said media to maintain both transducers in operative relation with the intervening media with the first transducer serving as a fixed positional reference despite tendencies of the media to deviate in position from its nominal plane and the second transducer matingly accommodating said tendencies by virtue of its gimbal support.
[1022]*10225. A device for maintaining a pair of magnetic transducers in operative relation with opposite sides of a non-rigid magnetic recording media comprising:
a first non-gimballed transducer mounted on a first side of the media and having a relatively fixed position relative to the plane of the media in which the first transducer is in operative relationship with the first side of the media;
a pivotable support arm having a pivot axis in fixed spatial relationship to the first transducer, and extending along the second side of the media to a region opposite the first transducer;
a second transducer coupled by gimbal means to the support arm in opposition to the first transducer and being movable relative to the plane of the media and to the first transducer; and
means coupled to said support arm for urging said second transducer and the media toward said first transducer with a force such that close operative relationship is maintained between each transducer and the associated side of the intervening media, with tendencies of the media to deviate from its nominal plane being compensated by the second transducer.
12. In a magnetic recording system in which a carriage supporting two magnetic heads is shifted radially relative to a center-driven pliant magnetic disc to provide data transfer with selectable record tracks on either side of the disc, the improvement comprising:
a first transducer fixedly coupled to the carriage for bearing against a first side of the disc in an invariant position in a direction normal to the plane of the disc;
and means including a second gimbal mounted transducer movably coupled to the carriage for bearing against the second side of the disc in opposed relation to the first transducer, the second transducer being movable toward and away from the plane of the disc and the first transducer, said means including resilient means urging said second transducer toward said first transducer with sufficient force to maintain close operative relationship between both said transducers and the disc despite deviation of portions of the disc during movement thereof, with the second gimbal mounted transducer compensating tendencies of the pliant disc to deflect.
Relying on this prosecution history, and on testimony emphasizing the requirement of the claims that the first transducer is in a fixed position, thus serving as a positional reference to the media and the second, gimballed transducer, the Commission interpreted the word “fixed” to mean that the first transducer does not move in any direction. The Commission referred to the patent specification which “repeatedly emphasizes and explains that in the claimed invention the disk is deflected by the fixed transducer and at the same time confined against the fixed transducer” (emphasis in original), and to the testimony of Mitsubishi’s patent expert Jessup and technical expert Lewis that the term “fixed” has the generally accepted meaning in the disk drive industry as “designed to be not able to move.”
Tandon urges that the clauses “fixed position in a direction normal to the plane of the media” in claim 1, and “fixedly coupled ... in an invariant position in a direction normal to the plane of the disc” in claim 12, only limit movement in the Z direction but do not restrict movement about the X and Y axes. Tandon points out that it was excess movement along the Z axis that flawed the prior art IBM heads, and argues that the Tandon claims should not be interpreted as requiring a degree of fixedness beyond that needed to distinguish from this prior art. Indeed, Tandon says, the critical text of the claims stating that the lower transducer is fixed in a direction normal to the plane of the media was required by the examiner as a condition of allowance to distinguish the claims over the IBM reference.
[1023]*1023However, in distinguishing the IBM reference to the patent examiner, Tandon wrote in its remarks accompanying the amendment that the IBM drive
employs two pivotable, gimbaled heads, which ... during operation, ... move with at least two degrees of freedom to attempt to follow the deviations of the floppy disk from its nominal position. In contrast, applicants’ system utilizes one head that has an invariant position bearing against one side of the media without spring loading, and this head is nongimbaled. [emphases added]
Although Tandon states that this argument did not apply to claims 1 and 12, the Commission did not accept that view. The Commission’s position is sufficiently supported in the record, for there is nothing in the prosecution history that limits the quoted comments to any particular claim or claims. Further, Tandon’s specification and prosecution history do not teach any difference in the degree of fixedness of the first transducer among described or claimed embodiments of the invention.
The phrase “normal to the plane” does not appear in the specification as filed. Tandon argues that the amendment adding this phrase shows the intended absence of restraint of movement in the X and Y directions. However, Tandon represented, as it must, that the amendment was “without the addition of new matter.” The added phrase can not enlarge the scope of the claims beyond that supported in the specification, and can not change the disclosure in a way contrary to its substance as filed.
37 C.F.R. 1.118(a): ... All amendments to the specification, including the claims, and the drawings filed after the filing date of the application must conform to at least one of them as it was at the time of the filing of the application.
The addition of the amendment describing movement in the Z direction did not thereby expand the disclosure and claims with respect to movement in the X and Y directions.
Tandon argues that since claim 5 expressly describes the first transducer as non-gimballed, claims 1 and 12 must be read as encompassing a gimballed first transducer. Tandon argues that the “doctrine of claim differentiation” prevents reading into claims 1 and 12 this limitation of claim 5, citing D.M.I. Inc. v. Deere & Co., 755 F.2d 1570, 1574, 225 USPQ 236, 239 (Fed.Cir.1985).
The Commission held that claim 5 does “not differ in scope with respect to the movement of the lower transducer” from that of claims 1 and 12, and that the term “relatively fixed” in claim 5 “does not have a different meaning from the terms ‘fixed’ and ‘fixedly coupled’ as used in claims 1 and 12.” The Commission held that this claim interpretation did not violate the doctrine of claim differentiation, since it related to only one element of claim 5, and that claim 5 viewed as a whole could still differ in scope from claims 1 and 12.
There is presumed to be a difference in meaning and scope when different words or phrases are used in separate claims. To the extent that the absence of such difference in meaning and scope would make a claim superfluous, the doctrine of claim differentiation states the presumption that the difference between claims is significant. D.M.I., 755 F.2d at 1574, 225 USPQ at 239; Autogiro, 384 F.2d at 404, 155 USPQ at 708. At the same time, practice has long recognized that “claims may be multiplied ... to define the metes and bounds of the invention in a variety of different ways.” Bourns, Inc. v. United States, 537 F.2d 486, 492, 210 Ct.Cl. 642, 187 USPQ 174, 178 aff'd. per curiam, 199 USPQ 256 (Ct.Cl.1976). Thus two claims which read differently can cover the same subject matter. Further, as this court stated in D.M.I., 755 F.2d at 1574 n. 2, 225 USPQ at 238 n. 2, “[c]laims are always interpretable in light of the specification that led to the patent.” See also Autogiro, 384 F.2d at 397, 155 USPQ at 702 (“No matter how clear a claim appears to be, lurking in the background are documents that may completely disrupt initial views on its meaning”). Whereas under the facts of D.M.I. there was “simply no basis in either the specification or prosecution history” for limiting the claim beyond [1024]*1024its literal terms, 755 F.2d at 1574 n. 2, 225 USPQ at 238 n. 2, in Tandon’s case the Commission held that the specification and the prosecution history require that the claims be limited to a non-gimballed, fixed, lower tranducer. Whether or not claims differ from each other, one can not interpret a claim to be broader than what is contained in the specification and claims as filed. See Autogiro, 384 F.2d at 397, 155 USPQ at 702-03 (“[W]ords must be used in the same way in both the claims and the specification.”); 37 C.F.R. 1.75(d)(1):
[T]he terms and phrases used in the claims must find clear support or antecedent basis in the description so that the meaning of the terms in the claims may be ascertainable by reference to the description.
Thus, as a matter of claim interpretation, the Commission correctly held that the inclusion of the term “non-gimballed” in claim 5 did not require that claims 1 and 12 be read to encompass a gimballed first transducer.
We have considered all of the arguments offered by Tandon, and conclude that on this record the Commission correctly interpreted claims 1, 5 and 12 to mean that the first transducer is fixed in all directions.
Infringement
Applying the claims as it interpreted them, the Commission found that the Mitsubishi drives do not infringe claims 1, 5 and 12 of the ’573 patent, either literally or under the doctrine of equivalents. We review these findings to determine whether they are supported by substantial evidence.
A.
Mitsubishi’s accused devices included several models, of which one embodiment was illustrated as:
In the Mitsubishi devices the lower transducer (3) is supported by a gimbal-spring (4) having a load finger (2). A pad (5) is affixed between the gimbal-spring and the load finger. As the AU found, the gimballing of the lower transducer enables roll and pitch movement, while the load finger restricts its movement in the Z-axis direction. The gimbal movement of the upper transducer is not at issue.
Neither Mitsubishi nor the Commission disputes that the patent claims read on the Mitsubishi devices except as to the lower transducer.
Tandon argues that the Mitsubishi devices literally infringe claim 1 since the lower heads of the devices, according to Tandon’s measurements, do not move significantly in the Z direction, as discussed below. However, the claim interpretation that the Commission adopted and that we have upheld requires the lower head to be fixed in all directions; further, the AU and the Commission found that the lower head of the Mitsubishi device moves significantly in the Z-direction. Substantial evidence supports the Commission’s finding that literal infringement has not been proven.
Although the Mitsubishi lower head is supported on gimbal springs and a load point, Tandon argues that the Mitsubishi [1025]*1025heads4 in actual use move no more in the X, Y, and Z directions than do the Tandon heads made in accordance with the claimed invention, and thus that claims 1 and 12 are literally infringed. Tandon submitted measurements showing that the Mitsubishi lower heads moved comparably to the Tandon heads in the Z or normal direction (in microinches):
Mitsubishi Tandon (5 1/4") Z (normal) 29 17 Tandon (8") IBM 59 3,139
and that the lower heads also moved comparable amounts in the X (pitch) and Y (roll) directions (in arc minutes):
Mitsubishi Tandon (5 1/4") X (pitch) 3.8 0.30 Y (roll) 0.60 0.30 Tandon (8") IBM 0.45 22.62 0.45 18.17
Tandon argues that its data show that the Tandon and Mitsubishi drives all display relatively small movements in the three directions, and that the gimbal spring of the Mitsubishi lower head is purely “cosmetic”. Tandon argues that its calculations are based on Mitsubishi’s own data, that Z-axis movement in the Mitsubishi device is negligible due to the presence of the “load finger” making the lower head effectively “fixed”, and that X and Y-axis movement of the Mitsubishi device is comparable to that of the Tandon device. Tandon provided the IBM data to show the comparatively large movements in all three directions for drives that are “truly gimballed”.
Mitsubishi submitted conflicting data (in micrometers and degrees):
Drive Z (normal) X (pitch) Y (roll)
Tandon 0.26 0.0017 0.00067
Mitsubishi
M4841-00M 16.6, 24.1 0.19, 0.65 0.13, 0.42
-112M 19.8, 19.3 0.083, 0.145 0.16, 0.27
-001M 42.5, 36.9 0.19, 0.12 0.058, 0.024.
Tandon’s and Mitsubishi’s separate measurements of the movement of the different heads produced data that appear to be in bold conflict. For instance, Mitsubishi’s measurement of 16.6 micrometers for Z-axis movement on one accused device converts, according to Finding of Fact 216, to 664 [sic: 654] microinches, compared to Tandon’s data showing 29 microinches for the same model. Although “facts” in judicial proceedings can be found many ways without offending the natural law, it is of small aid when two sets of scientific “facts”, ostensibly measuring the same physical phenomena, show dramatically different results.
The Commission, remarking on the conflicting data, stated that it “afforded greater weight” to the Mitsubishi tests because “those tests considered the loading forces” exerted by the upper arm through the upper transducer against the disk and the lower transducer. The ALJ found that Tandon’s data did not take into account these loading forces, and that failure to take these forces into account would result in flawed data. There was expert testimony to the effect that the Tandon device [1026]*1026holds the disk flat against the lower head in the absence of the loading force, but that the Mitsubishi does not. We conclude that there was substantial evidence supporting the Commission’s reliance on the Mitsubishi tests over the Tandon tests.
Based on the Mitsubishi data, and on videotapes and expert testimony, the Commission found that “the lower or first transducer of the Mitsubishi drives moves significantly during operation of the drives.” The Commission further found that “the first (lower) transducer of the Mitsubishi drive does not act as a fixed positional reference for the disk as expressly required by claim 1. Rather, the first transducer in the Mitsubishi drives follows the movements of the disk, i.e., the disk is not forced to conform to the first transducer.”
We uphold, as supported by substantial evidence, the factual findings of the AU and the Commission, who had details of the measurements and how they were obtained, and the examination and cross-examination of witnesses as to the data’s genesis and significance. Based on these findings, the Commission’s finding that there was no literal infringement is supported by substantial evidence, and is affirmed.
B.
Tandon argues that even if the claims are not literally infringed, they are infringed under the doctrine of equivalents. Tan-don asserts that the accused devices, as compared with Tandon’s claims, perform “substantially the same function in substantially the same way to obtain the same result.” Graver Tank & Mfg. Co. v. Linde Air Prod. Co., 339 U.S. 605, 608, 70 S.Ct. 854, 856, 94 L.Ed. 1097, 85 USPQ 328, 330 (1950) (quoting Sanitary Refrigerator Co. v. Winters, 280 U.S. 30, 42, 50 S.Ct. 9, 13, 74 L.Ed. 147 (1929)).
Although Mitsubishi argues that Tandon relinquished the possibility of asserting equivalency that would encompass the Mitsubishi devices because of the amendments to the claims and representations made to the PTO during prosecution of the Tandon application, Tandon is not estopped from claiming any equivalents at all. See Mannesmann Demag Corp. v. Engineered Metal Products Co., Inc., 793 F.2d 1279, 1284, 230 USPQ 45, 48 (Fed.Cir. 1986) (“Amendment of claims during patent prosecution does not necessarily bar all benefit of the doctrine of equivalents”). However, claims may not be enlarged by equivalents to encompass the teachings of the prior art. Perkin-Elmer Corp. v. Computervision Corp., 732 F.2d 888, 900, 221 USPQ 669, 678 (Fed.Cir.1984), cert. denied, 469 U.S. 857, 105 S.Ct. 187, 83 L.Ed.2d 120 (1984). Based on Tandon’s statements to distinguish the IBM prior art before the PTO, the Commission held that Tandon could not broaden the scope of its claims to include devices utilizing two non-fixed heads.
The Commission found that while the Mitsubishi devices performed the Tan-don function of data transfer, they performed the function in a different way. The claimed invention, as described in the specification and in the prosecution history, transfers data utilizing one fixed head which serves as a positional reference on one side of the disk, and one gimballed head on the other side. The accused devices transfer data utilizing two gimballed transducers. The AU found that although the two Mitsubishi heads do not move to the same extent as in the IBM device, they still “follow the deviations of the floppy disk”.
Tandon argues that the Commission found that both the Mitsubishi and Tandon devices suppress the large-scale waviness of the disks, and that this suppression of large-scale waviness by both devices is alone sufficient for finding infringement by equivalents. Tandon’s proffered data, which compares the Z-axis movement of both devices with the much larger movement of the IBM device, supports its argument that the IBM head follows the large-scale waviness of the disk, which the other two devices suppress. Tandon thus asserts that the Mitsubishi device is much closer to the Tandon device than to the IBM device, which was the most pertinent prior art, and is within a reasonable interpretation of equivalency.
[1027]*1027The AU found that the Tandon device suppresses both the large-scale and small-scale fluctuations, while the Mitsubishi device suppresses the large-scale fluctuations only, while following the small-scale fluctuations of the disk. Thus the AU found that the Mitsubishi device transfers data in a different way than either the IBM or the claimed devices. These findings are supported by substantial evidence. They are important factors in the determination of non-equivalence.
The AU cited test data to the effect “that the degree of movement in the lower head of the Tandon drive does not vary for [various brands of] media. In contrast, in the Mitsubishi drive, the degree of movement of the lower head varied by over a factor of 3:2 depending upon the type of media used. In the Tandon drive, it is the upper head alone which responds to the thickness variations and fluctuations of the media. In the Mitsubishi drive, both heads cooperate to accommodate the thickness variations and fluctuations of the media.”
Both the Mitsubishi and the Tandon devices were found to remain within the “critical spacing” from the media of 0.32 microns (12.6 microinches). This spacing is the maximum distance between the media and the head that will allow accurate data transfer. It was uncontroverted that both devices achieve accurate and reliable data transfer. Even so, the AU found that a Mitsubishi Mark I &h inch drive head moved 16.6 micrometers (664 microinches) in the normal direction, “about 50 times that of the critical spacing”, and “several hundred times more than the lower head movement in the Tandon 5V4 inch drive”.
This evidence supports the Commission’s findings that the Mitsubishi device works in a different way from that described and claimed in the ’573 patent, as the Tandon device smoothes out both the large-scale and the small-scale waviness in the movement of the disk, while the Mitsubishi device follows the small-scale waviness after smoothing out the large-scale waviness.
These findings are also consistent with the amended claim language. Claim 1 was amended to read, in part, “with the first transducer serving as a fixed positional reference despite tendencies of the media to deviate in position from its nominal plane and the second transducer matingly accommodating said tendencies by virtue of its gimbal support”. The Tandon lower transducer is claimed as fixed “despite tendencies of the media to deviate”, while the gimballed upper transducer follows the deviations.
Tandon’s inventors, when confronted with the IBM prior art, distinguished their invention as one which suppresses both large-scale and small-scale waviness. Having successfully argued this position before the PTO, it is simply too late for Tandon to reclaim the ground which it yielded during prosecution. Since the Mitsubishi head moves along with the small-scale fluctuations of the media while remaining within the critical spacing, it falls at worst into the range that Tandon is estopped to claim. The Commission’s factual findings supporting this conclusion are supported by substantial evidence.
Tandon’s claimed invention suppresses small scale waviness by penetration of the fixed head into the plane of the media. The ’573 specification states that “the fixed transducer is positioned to penetrate slightly into the nominal plane of the record element”. The AU found, based on the testimony of Tandon’s patent expert, that “[a]n important aspect of the ’573 patent ... is that the lower head must be ‘fixed’ so that the media tends to conform to the head rather than the head following any pitching and rolling of the media____” Mitsubishi witnesses Lewis and Hayashi testified to the same effect.
Tandon asserts that the Mitsubishi lower transducers also provide a positional reference for the media in the same way as the claimed device, and therefore infringe as equivalents by performing the same function in substantially the same way. There was evidence, referred to by the AU, that “the penetration of the Tandon drive is about five times that of the Mitsubishi drives”, and that “the penetration of the lower head in the Mitsubishi devices is not sufficient to establish a physical reference for the disk”. The Commission, applying prosecution history estoppel and referring to testimony, concluded that “the Mitsubi[1028]*1028shi drives, although performing the same function, do not perform it in substantially the same way.”
Tandon points out that only claims 2 and 13, not here at issue, expressly require penetration of the transducer into the plane of the disk. Invoking the doctrine of claim differentiation, Tandon argues that claims 1, 5, and 12 are not limited to devices that penetrate the plane of the disk. However, before the Patent Office the applicants stated that penetration of the plane is what provides a “reference position” for the disk, and in response to the examiner’s rejection of all claims the applicants emphasized a “significant feature” of the “fixed head which serves as a positional reference.” As we observed supra, the doctrine of claim differentiation does not allow unrestrained expansion of claims beyond the description of the invention in the specification, and explanations and representations made to the PTO in order to obtain allowance of the claims. See Autogiro, 384 F.2d at 399, 155 USPQ at 703-04.
Tandon also argues that the Commission’s reliance on penetration by the transducer was in error in that it was based on an impermissible comparison of Tandon’s commercial embodiments with the alleged infringing devices, rather than the correct procedure of applying the claims to the accused devices. Amstar Corp. v. Envirotech Corp., 730 F.2d 1476, 1481-82, 221 USPQ 649, 653 (Fed.Cir.1984), cert. denied, 469 U.S. 924, 105 S.Ct. 306, 83 L.Ed.2d 240 (1984). The Commission, criticizing the AU for having made this error, explicitly disclaimed having based its final determination on a comparison of devices.
The Commission’s finding is, on this record, adequately supported by substantial evidence that the way the Mitsubishi devices work is not substantially the same as that defined in the claims. Thus we affirm the Commission’s determination that the Mitsubishi drives have not been proven to infringe the claims of the ’573 patent literally or under the doctrine of equivalents.
Procedural Error
Tandon asserts that the Commission erred in striking Tandon’s designation of and reliance on certain deposition testimony in its reply brief. Tandon had designated before the AU certain portions of the inventors’ testimony, which designated portions were included in the record before the Commission. Tandon’s argument is that after Mitsubishi filed its brief before the Commission it became necessary to enlarge the designation. This the Commission rejected as contrary to its rules, which provide that the record before the Commission shall not be enlarged beyond that before the AU.
The Commission did refer to the inventors’ testimony in connection with its decision as to the meaning of “fixed” in the ’573 specification and claims. However, Tandon has not demonstrated a sufficient basis for the extraordinary relief requested. We discern no arbitrary ruling or abuse of the Commission’s discretion, and decline to require that the Commission reopen proceedings on this basis.
AFFIRMED.