OPINION
MARGOLIS, Judge.Pursuant to 28 U.S.C. § 1498(a), the plaintiff brought this action to recover reasonable and entire compensation for infringement of its patent by the United States. The defendant argues that it has not infringed the patent and that the patent is invalid. During a four-day trial the parties presented evidence and testimony to a judge of this court, who retired without having made findings of fact or conclusions of law. The parties have waived a new trial and have submitted post-trial briefs. After hearing oral argument on those briefs and considering the entire record, the Court finds that the plaintiff’s patent is valid but not infringed.
BACKGROUND
A. Navy Procurements
On February 22, 1972 Preston R. Weaver was issued the patent in suit, U.S. Patent No. 3,643,513 (the ’513 or Weaver patent). Weaver assigned his patent to the plaintiff, UMC Electronics Company (UMC). The patent discloses an aviation counting accelerometer (ACA) that employs an analog *607transducer, a motion sensing device that emits an analog signal (voltage) proportional to force or stress experienced by a motion sensing element. A complete ACA senses and records the acceleration load levels that aircraft undergo during inflight maneuvers. It should record only acceleration loadings caused by maneuvers and ignore parasitic or transient acceleration signals—that is, vibrations caused by such things as gusts of wind, firing weapons, and the aircraft’s engine. Records of load levels of acceleration can indicate an aircraft’s remaining useful life and show the need for structural overhaul or rotation to different service, e.g., from combat to reconnaissance.
Before the plaintiff entered the ACA field, the U.S. Navy procured its ACA’s under military specification Mil-A-22145 from the Maxson Electronics Corporation and from Giannini Controls Corporation, which has become part of the Conrac Corporation. These ACA’s use electromechanical transducers to generate mechanical signals that indicate levels of acceleration. In such transducers, a seismic mass moves in response to input acceleration, deflecting a spring suspension by a distance proportional to the acceleration. As the mass moves, it touches probes or draws a brush across a set of contacts that are positioned to open and close circuits representing present acceleration levels. The output voltage does not vary with acceleration.
The natural frequency of the mass-spring system limits the ability of the electromechanical transducers to filter out high-frequency acceleration loadings that represent vibration. But an analog transducer, like the one used in the plaintiff’s device, generates a varying electrical signal that can be filtered electronically to remove superimposed vibrations.
Personnel at the Naval Air Development Command were never happy with either the Maxson or Giannini ACA’s because they recorded data that defied common sense, failed to count accelerations, or counted accelerations that never occurred. In November of 1964 the Navy’s Aeronautical Structures Laboratory analyzed the Navy’s ACA systems and found that twenty-five percent of the Giannini units and thirty-three percent of the Maxson units miscounted. As a result of this study, the Navy proposed to remove Maxson ACA’s from its Qualified Products List.
In early 1966, employees of the Navy’s Aircraft Structures Laboratory contacted Weaver, who was then working for UMC. They told him they were having problems with their ACA’s and showed him examples of faulty ACA data. They informed him that the Navy needed additional ACA’s and wanted to buy more accurate ones. Transcript (Tr.) at 24-25, 29.
Later in 1966 the Navy issued requests for proposals to supply ACA’s that met military specification MÜ-A-22145A. On August 25, 1966, UMC submitted a proposal to build such a device (the UMC-A). The UMC-A contains an electromechanical transducer. On December 1, 1966 the Navy awarded UMC a contract to supply approximately 1,600 UMC-A model accelerometers.
In early 1967 UMC discovered that it could not build the UMC-A to meet the static calibration requirements of the specification (that is, when subjected to the same acceleration load on different occasions, the device sometimes counted and sometimes failed to count). In order to prevent UMC from losing the contract, Weaver began work on a different accelerometer between mid-April and mid-May of 1967. This accelerometer (the UMC-B) used an analog transducer, as does the device that Weaver eventually patented. Weaver satisfied himself that the UMC-B could function by building and testing an engineering prototype.
On May 22, 1967 the Navy issued a new ACA specification, MÍ1-A-22145B, which superceded MÍ1-A-22145A. In July it issued a request for proposals to deliver ACA’s built to the new specifications, RFP N00383-68-R-0001 (RFP ’0001). UMC responded to this request on July 27, 1967 with a letter describing an ACA that uses an analog transducer. On August 2, 1967 *608UMC submitted a technical proposal, and on August 9, 1967 UMC demonstrated a prototype of the UMC-B to the Navy. The Navy later cancelled the UMC-A contract for default.
Early in 1968 the Navy cancelled RFP ’0001, and on July 12, 1968 it issued another invitation for bids to supply ACA’s. Thirteen companies submitted technical proposals, ten of which the Navy rejected in the first step of the procurement. On November 29, 1968, as the second step, the Navy issued an invitation to three companies: UMC, Aerodyne Controls Corporation, and Systron-Donner Corporation (SDC). SDC won the contract. UMC has never produced the device disclosed by the Weaver patent.
Since 1970 the Navy has procured ACA’s from only one manufacturer, SDC. The SDC ACA employs an analog transducer. The plaintiff asserts that the SDC ACA infringes the Weaver patent.
B. Weaver’s Invention
In the plaintiff’s device, an analog transducer generates an electric current proportional to the acceleration experienced by the aircraft on which the device is mounted. A filter eliminates transient accelerations from the current, which passes to five voltage dividers (components 45-49 on Figure 4 of the ’513 patent). The dividers determine if the current has exceeded certain levels, the “reset level” and load levels 1-4 (LL1-LL4). Each load level signifies a higher acceleration load. See Figure 3, which illustrates the relation between acceleration and voltage.
When the electrical signal rises to the reset level, the output of amplifier 45b activates timer 53, which begins a timing cycle (point t0 on Figure 3). As the signal reaches LL1-LL4, it successively activates amplifiers 46b-49b, each of which energizes a latch circuit 46c-49c. The latch circuits store the information that the aircraft has experienced a corresponding load level of acceleration. After timer 53 has completed a timing cycle it activates release gate 54, which allows gates 46d-49d to transfer the information stored in the latch circuits to the corresponding counters 46e-49e.
*609
*610
*611If the acceleration signal remains above the reset level at the end of a timing cycle, the appropriate counter will advance one half count when the gates relay the information stored in the latch circuits. For example, if the timing cycle ends at ti on Figure 3, counters 46e and 47e will advance one half count. After the timing cycle, when the current drops below the reset level, amplifier 45 deactivates. This opens reset switch 52, which resets latches 46c-49c. The counters then advance one full count. (If the current rises to a higher load level after the timer has activated the information release gate, the new information can pass immediately from the latches to the counters.) For example, at each of the counters will advance a full count. But if the current sinks below the reset level before the end of a timing cycle, amplifier 45b deactivates. Reset switch 52 then turns off timer 53 and resets the latch switches 46c-49c, erasing the stored information. This timing cycle prevents the device from recording transient signals; the device records only accelerations that persist for a predetermined time or that build upon other accelerations that have persisted for a predetermined time.
VALIDITY
The defendant attacks the Weaver patent on the grounds that the device was anticipated, obvious, on sale more than a year before the application was filed, impracticable, and not invented by Weaver.
A. Anticipation
A person is not entitled to a patent if “the invention was known or used by others in this country ... before the invention thereof by the applicant for patent.” 35 U.S.C. § 102(a). A patented invention is previously “known or used” only if a prior art reference shows each and every element claimed. E.g., General Electric Co. v. United States, 215 Ct.Cl. 636, 678, 572 F.2d 745, 768 (1978). The defendant argues that the Weaver patent is invalid because its claims read on the Maxson, Giannini, and UMC-A devices. All these ACA’s are prior art; none was brought to the attention of the patent examiner who reviewed Weaver’s application.
The Weaver patent makes four claims. They appear below, each claimed means in a separate subparagraph on the left; on the right are the corresponding structures described in the specifications:
Claim Elements
1(a) An accelerometer for counting the number of times each of a plurality of acceleration maneuvering loadings of predetermined magnitude of an aircraft occur (sic), comprising
(b) means adapted to be mounted to an aircraft for sensing acceleration loading thereof,
Structure Shown on Figure 4 of the '513 Patent Corresponding to Claim Elements
(a) A counting accelerometer comprising
(b) A transducer comprising casing (not shown), force coil 17, capacitance bridge 26, oscillator 27, diode bridge 28, and amplifiers 29 and 3 0,
*612(c) means providing a signal proportional to said accelerations, 1/
(d) a plurality of sensing and storing means each responsive to an acceleration signal reaching a different predetermined value for sensing such signal and storing an indication of the value reached,
(e) a plurality of acceleration level recording means,
(f) timing means for timing a predetermined cycle and furnishing a signal indicative of the end of said timing cycle,
(g) means responsive to an acceleration signal reaching a reference level for causing said timing means to initiate a timing cycle,
(h) means responsive to said end of cycle signal arranged to pass
(c) A seismic mass comprising force coil 17 and capacitive displacement sensor (not shown), which supplies an output signal proportional to accelerations (Tr. 325-326; see also Defendant's Exhibit 17 at 6),
(d) Amplifiers 46b-49b and corresponding latch circuits 46c-49c,
(e) Counters 46e-49e,
(f) Timer 53,
(g) Amplifier 45b,
(h) Information release gate 54.
*613a signal from each of said sensing and storing means to an associated one of said recording means.
2(i) The accelerometer of claim 1 further including means for filtering high-frequency components from the signal waveform.
3(e)(i) The accelerometer of claim 1 wherein said means for recording comprise a plurality of counters, each counter arranged to be advanced by one of said sensing and storing means,
(h)(i) said means for passing comprises a plurality of gates, each of said gates being arranged to apply • an advance signal from one of said sensing and storing means to a respective one of said counters.
4(j) The accelerometer of claim 1 further including means for sensing when the signal exceeds said reference value, and disabling said sensing and storing means if the signal falls below said reference value.
(i) Filters 34 and 38-39 (Tr. 331; see also Tr. 275-277).
(e)(i) Counters 46e-49e advanced by amplifiers 46b-49b and latch circuits 46c-49-c,
(h)(i) Although there is no antecedent basis for the "means for passing," the actual structure that applies signals from the latch circuits 46c-49c to the counters 46e-49e is gates • 46d-49d.
(j) Amplifier 45b and reset switch 52.
All the claims of the Weaver patent read literally on the Maxson, Giannini, and UMC-A devices, as the following chart shows. (See the schematic diagrams of these devices printed in the Appendix.)
*614Maxson (App. I)
1(a)
Intended purpose of Maxson ACA
(b)
Transducer
(c)
Spring and mass provide deflection proportional to applied acceleration.
(d)
Several probes ride on surface of mass, closing contacts at different values of input acceleration. Silicon controlled rectifiers CR 4, 6, 8, & 10 store information that each load level has been reached.
(e)
Counters M1-M4
(f)
Rl, R2, Cl, and Q1 time predetermined cycle and furnish to CR12 a signal that indicates end of cycle.
(g)
Reset probe and CR1 apply voltage to timing means.
(h)
Information release gate CR12 responds to end of timing cycle
Giannini (App. II)
1(a)
Intended purpose of Giannini ACA
(b)
Transducer
(c)
Spring and mass provide deflection proportional to applied acceleration.
(d)
Several contacts at predetermined positions sense position of brush moving proportionally to acceleration. Relays Kl-4 store information that load levels LL1-LL4 have been exceeded.
(e)
Counters M1-M4
(f)
Timing means includes Rl, R2, Cl, RT1, and relay K5.
(g)
When reset level is reached, brush leaves reset level contact, beginning timing cycle.
(h)
Relay K5 and associated contacts. Contact pairs 11 & 14,
UMC-A (App. Ill)
1(a)
Intended purpose of UMC-A
(b)
Transducer
(c)
Spring and mass provide deflection proportional to applied acceleration.
(d)
Several probes ride on surface of mass, closing contacts at different values of input acceleration. Relays 1R-4R store information that predetermined values have been reached.
(e)
Counters 1-4
(f)
R4, R3, Cl, C5, & Q1 time a predetermined cycle; Q1 furnishes signal that indicates end of cycle.
(g)
Opening of contact (not shown)
(h)
Bistable flip-flop and relay 5R (contact 5R1 & coil 5R) respond to end of
*615signal, and diode gates CR15-18 pass signal from probes and CR4, 6, 8, and 10 to counters.
2 (i)
Low natural frequency spring-mass system
3(e) (i)
Counters Ml-4 advanced by probes and rectifiers
(h) (i)
Diode gates CR15-18
4 (j)When mass reaches reset level, reset probe makes contact. When signal falls below reset level, probe is disconnected. This removes power from storing means CR4, 6, 8, 10.
15 & 18, 2 & 6, pass information from storing means Kl-4 to counters.
2 (i)
Low natural frequency spring-mass system
3(e) (i) Counters Ml-4
(h) (i)
The contact pairs listed in 1(h)
4 (j)Brush leaves reset level contact when signal exceeds reference value.
When signal falls below reset value, brush engages reset level contact, disabling sensing and storing means.
cycle; contacts 1R2-4R2 pass signal from relays 1R-4R to counters.
2 (i)
Low natural fre-. quency spring-mass system with appropriate damping
3(e) (i)
Counters 1-4 advanced by probes and relays 1R-4R
(h) (i)
Contacts 1R2-4R2
4 (j)Reset probe senses signal exceeding reference value.
When reset probe switch closes, timer capacitors C5 and Cl discharge through reset probe; 1R reset coil is energized to open contact 1R1 and removes power from latching contacts 2R1-4R1, disabling the storing means.
Element 1(c) of the Weaver patent claims “means providing a signal proporational to ... accelerations.” In each of the prior art references the corresponding element, a moving mass, closes circuits and provides a mechanical signal proportional to acceleration. The specifications of the Weaver patent, however, disclose a device that generates an electrical signal. If the patent is construed to claim means providing an electrical signal, the patent will not read on the prior art. Tr. 596-97.
It is an axiom of patent construction that claims must be read in light of the specifications. See, e.g., Caterpillar Tractor Co. v. Berco, S.p.A., 714 F.2d 1110, 1116 (Fed.Cir.1983) (citing United States v. Adams, 383 U.S. 39, 49, 86 S.Ct. 708, 713, 15 L.Ed.2d 572 (1966)). When all the claims of a patent read literally on a prior art reference, this Court must avoid antici*616pation, if possible, and secure to the pat-entee the just fruits of his actual invention by construing those claims to cover what the drawings and specifications disclose. See, e.g., Dominion Magnesium Ltd. v. United States, 162 Ct.Cl. 240, 248-49, 320 F.2d 388, 394 (1963).2 See also ACS Hospital Systems, Inc. v. Montefiore Hospital, 732 F.2d 1572, 1577 (Fed.Cir.1984) (Court should construe claims so as to sustain their validity, if possible).
The Court concludes that the Weaver patent claims means that generate an electrical signal. Therefore, the prior art does not anticipate the Weaver patent.
B. Obviousness
Even if no single prior art reference anticipates an invention, the inventor may not obtain a patent if “the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains.” 35 U.S.C. § 103. The defendant asserts that Weaver’s device would have been obvious at the time of its invention to a person of ordinary skill in the art.
When considering whether an invention would have been obvious, a court must ascertain the scope and content of the prior art, the level of ordinary skill in the art, and the differences between the prior art and the claims at issue. Graham v. John Deere Co., 383 U.S. 1, 17, 86 S.Ct. 684, 693, 15 L.Ed.2d 545 (1966). But the court must also examine secondary considerations, and may not end its inquiry before giving appropriate weight to all relevant evidence. Stratoflex, Inc. v. Aeroquip Corp., 713 F.2d 1530, 1538-39 (Fed.Cir.1983).
1. Scope and Content of the Prior Art
A counting accelerometer is one of many kinds of motion sensing devices. The prior art includes transducers, indicators (recorders), and associated electronic circuits {e.g., timers, filters, gates, latching circuits, amplifiers), as well as counting accelerometers. The parties agree that prior art comprises the following documents and devices:
Counting Accelerometers by Maxson, Giannini, and UMC
See the chart and schematic diagrams.
References Cited by the Patent Examiner
The patent examiner rejected an earlier version of Weaver’s claim 1 as anticipated by U.S. Patent 2,968,952 (Stalder). Stalder discloses an analog transducer that produces a signal proportional to acceleration, switches that respond when the signal reaches predetermined values, and counters that record when the switches are activated. The examiner found other elements of Weaver’s invention in U.S. Patents Nos. 2,629,030 (Taylor), 2,988,737 (Schroeder), 3,295,634 (Van Dyke), 2,879,-053 (Weaver ’053), 3,448,621 (Magda), and 3,478,605 (Siegel). Taylor shows a timing device and a plurality of acceleration sensors each of which responds to a different level of acceleration. In Schroeder, circuits respond when the acceleration reaches given levels. Van Dyke, like the patent in suit, discloses one signal source that actuates different circuits depending upon the level of acceleration. Both Schroeder and Van Dyke use filters. Weaver ’053 de*617scribes an accelerometer in which four mi-croswitches arranged in counting circuits respond to progressively greater accelerations, and means that prevent the device from recording high frequency components. Magda is a similar acceleration sensing device, and Siegel shows an accelerometer with a physical structure similar to the patent in suit.
The Kistler Manual
In December of 1966 the Kistler Instrument Company published a manual entitled “Operation and Installation Models 305A & 305T Servo Accelerometers.” The manual describes essentially the same transducer as the one shown on Figure 4 of the Weaver patent.
The Kistler Patent
U.S. Patent No. 3,323,372, issued to Walter Kistler, discloses an analog transducer “which may be connected directly ... to recorders____” Id. at col. 4, 7-15. The specifications suggest that a filter network could filter the output of the transducer and produce a signal proportional to acceleration.
The Anderson Patent
U.S. Patent No. 2,774,535, issued to Lloyd Anderson, discloses a recorder that contains counters, stress sensors set at predetermined levels, gates, and storage elements that are actuated by a continuous electrical signal proportional to stress imposed on an object.
Patents Issued to Harold D. Morris
Harold D. Morris, an employee of SDC, invented and patented what eventually became the accused device, the SDC ACA. Before August 1, 1968, when Weaver applied for a patent on his accelerometer, Morris had already been issued the following patents:
U.S. Patent No. 3,074,279 (Morris ’279) discloses a position detecting transducer or accelerometer that generates a continuous current proportional to acceleration. The patent teaches that the output current can be used directly to operate a recorder.
U.S. Patent No. 3,122,714 (Morris ’714) discloses a fourth-order filter system designed to eliminate all signal components above a predetermined frequency.
U.S. Patent No. 3,168,680 (Morris ’680) discloses a device that actuates a switch contact when acceleration reaches a predetermined value. In the ’680 device an accelerometer provides an analog signal proportional to acceleration that is filtered to eliminate frequencies above 2 cycles. The device also contains circuitry that senses an acceleration signal and stores information that the signal has reached a predetermined value. A gate can pass that signal to any equipment the user connects to the device’s relay. One embodiment of the ’680 patent contains a timing circuit that actuates the contacts of that relay only when the acceleration remains above a reference level for a predetermined period.
Military Specifications
The Navy procured Maxson and Giannini ACA’s under military specification Mil-A-22145A (the A spec.), which required an ACA with two components: (a) a unidirectional acceleration sensing device (a transducer) that completes electrical circuits at discrete load levels; and (b) a counting accelerometer indicator having counter displays that show the number of times an aircraft undergoes preset load levels of acceleration. Around the time Weaver invented his device the Navy published a new specification, MU-A-22145B, which was essentially the same as its predecessor but which changed two requirements: (a) the dynamic response of the accelerometer was to be flat to 2.5 Hertz (instead of the 3 Hertz noted in the A spec.); and (b) the indicator was not to count sinusoidal or square wave inputs of 4 Hertz and above (instead of cutting off counts at 5.5 Hertz, as the A spec, required).
2. Level of Ordinary Skill
As of May 1967, a person of ordinary skill in the counting accelerometer art *618would have had a full academic degree or equivalent experience in either mechanical or electrical engineering, plus technical training and experience in testing or designing electronic circuits, filters, transducers, and recording equipment.
3. Differences between the Weaver Patent and Prior Art
Prior art reveals—in some form—all the elements that compose the Weaver ACA; prior art also teaches certain combinations of those elements. It does not teach a combination of all.
4. Other Considerations
Even to a layman who examines the record of this case eighteen years after Weaver’s invention, it might seem obvious to build the equivalent of the Weaver ACA by combining any analog transducer with, e.g., the filter, modified storing means, and timing circuit of the Morris ’680 and the counting means of the Maxson ACA. The Court, however, must resist the temptation to invent by hindsight. In this case, the best antidote against the charm of hindsight is objective, contemporaneous evidence of what persons skilled in the art considered to be patentable near the time Weaver filed his application. See, e.g., W.L. Gore & Associates, Inc. v. Garlock, Inc., 721 F.2d 1540, 1553 (Fed.Cir.1983), cert. denied, — U.S. —, 105 S.Ct. 172, 83 L.Ed.2d 107 (1984).
The SDC Patent
Harold D. Morris, the defendant’s expert witness, testified at trial that in his opinion the Weaver ACA would have been obvious at the time of its invention in light of the prior art. Morris developed the accused device, the SDC ACA, and was issued U.S. Patent No. 4,016,766 (the SDC patent). The SDC patent essentially discloses the accused device.
Morris has more than ordinary skill in the art. He earned the degree of Bachelor of Science in Electrical Engineering with honors in 1950, studied at the graduate level, is a registered professional engineer in California, and since 1960 has been technical director of the Systron-Donner Corporation. There he oversees all the operations of the Inertial Division, which produces motion sensors—especially accelerometers and related signal processing circuitry.
His SDC patent discloses an ACA that uses an analog transducer, filter, and timing circuit to generate an electrical signal proportional to maneuvering accelerations. Counters indicate how many times acceleration has reached any of four load levels. With few differences, claim 15 of the SDC patent quotes elements l(a)-(e) of Weaver’s claim 1.
At the time Morris invented the SDC ACA, he knew of his own analog transducer (Morris ’279), his filter system (’714), and his switch patent (’680), which discloses storing, gating, and timing means and suggests adding other equipment to an output relay. He also knew of military specification MIL-A-22145, to which the Maxson and Giannini ACA’s had been built. Morris has testified that this specification would have spurred a person of ordinary skill to design circuitry 1) that stores information of an acceleration signal having reached any of four load levels; and 2) that prevents the corresponding counters from advancing until that signal has sunk below a reset level.
Nevertheless, in 1968 SDC offered its ACA to the Navy as a “unique Dynamic Discriminator circuit, which permits counting of true maneuvering loads only.” And on July 14, 1969, Morris applied for a patent on the device because he felt that he “had achieved a rather remarkable new design of counting accelerometer and that it deserved to be protected by a United States patent.” Tr. 729.
Of course, the Weaver and SDC devices are not identical. The Court took the differences into account when weighing Morris’s admission. Still, the SDC and Weaver ACA’s are similar, and an expert who knew all the cited prior art or its equivalent believed in 1968 that one of the devices was *619patentable. This is evidence that the Court cannot ignore.
The Aerodyne Bid
Three companies survived the first round of the Navy’s 1967-68 procurement of ACA’s under specification B. One of those companies, Aerodyne Controls Corporation, offered an ACA with an electromechanical transducer. Yet Morris testified in 1983 that the size limits and strict requirements for transducer response found in specification B could have suggested only one thing to a person of ordinary skill in 1967: an analog transducer with an electrical filter. Tr. 514-16. Aerodyne’s bid tends to show that specification B did not suggest the use of analog transducers in 1967-68.3
Conclusion
What SDC, Morris, and Aerodyne did and said in 1968-69 suggests that hindsight influenced Morris’s testimony in 1983. The defendant has not persuaded the Court that Weaver’s invention would have been obvious to a person of ordinary skill at the time it was made.
C. Prior Use or Sale
An inventor is not entitled to a patent if the device was “in public use or on sale in this country, more than one year prior to the date of the application for patent in the United States.” 35 U.S.C. § 102(b). Weaver filed his application on August 1, 1968. Over a year earlier, on July 27, 1967, UMC had offered to sell the Navy an ACA that used an analog transducer. The defendant argues that this offer invalidates Weaver’s patent. But UMC had not yet produced the device it offered to sell.
The United States Court of Customs and Patent Appeals adopted three prerequisites for § 102(b) attacks against patents on devices offered for sale before they have been produced:
(1) The complete invention claimed must have been embodied in or obvious in view of the thing offered for sale____ Complete readability of the claim on the thing offered is not required because whatever is published (or on sale) more than one year prior to the filing of a patent application becomes part of the prior art over which the claim must be patentable____
(2) The invention must have been tested sufficiently to verify that it is operable and commercially marketable. This is simply another way of expressing the principle that an invention cannot be offered for sale until it is completed, which requires not merely its conception but its reduction to practice____
(3) Finally, the sale must be primarily for profit rather than for experimental purposes____
In re Corcoran, 640 F.2d 1331, 1333-34 (C.C.P.A.1981) (quoting Timely Products Corp. v. Arron, 523 F.2d 288, 302 (2d Cir.1975)) (citations omitted). With one qualification, Corcoran and Timely Products set out the law that binds this Court. See Barmag Barmer Maschinenfabrik AG v. Murata Machinery, Ltd., 731 F.2d 831, 836-37 (Fed.Cir.1984) (citing South Corp. *620v. United States, 690 F.2d 1368, 1370 (Fed.Cir.1982)).
UMC has admitted that it offered its device to the Navy for profit, not for experimentation; but it denies that it reduced Weaver’s device to practice or built a physical embodiment before August 1, 1967.
Reduction to Practice
“[R]eduction to practice requires that an invention be ‘sufficiently tested to demonstrate that it will work for its intended purpose.’ ... On the other hand, ‘[t]here is no requirement for a reduction to practice that the invention, when tested, be in a commercially satisfactory stage of development.’ ” Barmag Barmer, 731 F.2d at 838 (citations omitted).
Weaver built a testing device comprising a centrifuge (to provide a constant acceleration load) and a shaker (to superimpose vibrations upon the static acceleration of the centrifuge). At some time between May and June of 1967 (Weaver could not recall the exact date), he attached an engineering prototype of his ACA to the testing device and determined that he could eliminate superimposed vibrations from the acceleration signal. The prototype contained at least a Kistler analog transducer. The Court finds that these tests reduced Weaver’s invention to practice, because they demonstrated that his invention would serve its intended purpose. Weaver admitted this at trial:
Q. All right, up to May or June 1967, such—
A. Some type of prototype was built that demonstrated that the thing was feasible. That is all I can say, sir.
THE COURT: You were satisfied from that which was built in May of ’67, Mr. Weaver? I just want to make sure I understand your testimony. You were satisfied as a result of the bread-board prototype that was built that the invention, which you conceived, would work?
THE WITNESS: That is right, Your Honor.—more of the invention we worked up, we could satisfy the Navy’s requirements.
THE COURT: I understand.
Tr. 102-03.
Physical Embodiment
One purpose of the “on sale” bar of § 102(b) is to preclude exploitation of the patent monopoly substantially beyond the statutory period. E.g., General Electric, 228 Ct.Cl. at 202, 654 F.2d at 61. To prevent patentees from circumventing this policy, a panel of the Federal Circuit has refused to condition the § 102(b) bar upon the existence of a physical embodiment of the thing offered for sale in every case, because situations might arise in which “commercial benefits outside the allowed time have been great” even without a physical embodiment. Barmag Barmer, 731 F.2d at 837. UMC, however, never produced its ACA; it reaped no commercial benefits before the statutory period. In this case, therefore, the defendant must show either that the claims in the Weaver patent were embodied in the thing offered for sale, or that Weaver’s patented device was obvious in view of the thing offered for sale.
The parties dispute what the UMC prototype contained. Before the offer of sale, July 27, 1967, the only persons who tested the ACA were Weaver and his assistant, Paul Doane. Doane himself tested a prototype that contained only a transducer, a filter, and a voltage sensor that measured one load level. The prototype had no timing circuitry. The claims of the Weaver patent do not read on this device. But Doane joined UMC in June of 1967 and did not know what Weaver might have tested in May.
In answers to interrogatories, UMC described the May 1967 prototype as containing: means providing an analog signal proportional to acceleration loading, a plurality of sensing and storing means, a plurality of recording means, a timing means, means to activate the timer when the acceleration signal reaches a reference level, and means responsive to the end of a timing signal that could pass information from the storing means to the recording means. This *621devicé would embody all the elements Weaver’s claim 1. of
At his deposition Weaver testified that he did not know what the May prototype contained. At trial he testified that the prototype must have contained what UMC’s answers listed. Other evidence, however, tends to show that those answers were wrong.
The answers describe the indicator as “encapsulated.” Both Weaver and Doane testified that UMC’s supplier did not deliver encapsulated circuitry until the fall of 1967, probably after October 20th; and the defendant has admitted that UMC itself lacked the equipment needed to build such circuitry. It follows that in May of 1967 UMC could not have built the device described in its answers.
On this point, the Court mistrusts the ambiguous oral testimony and information submitted for the purpose of litigation. The Court prefers to rely on contemporaneous documents. In its 1967 request, the Navy required bidders to describe their transducers. UMC’s July offer sets out the circuitry to be used with the transducer and adds, “Such a transducer has been constructed and tested in conjunction with voltage sensing and time controlled circuitry.”
When the Navy received this offer, it requested a technical proposal, which UMC submitted on August 2, 1967. UMC hoped to win the ACA contract. The Court infers that UMC would have included in that proposal all its favorable information about Weaver’s invention. But the test results in the proposal show only that Weaver had attached the filtered Kistler transducer and a Statham accelerometer to the centrifuge. The filtered transducer rejected the superimposed vibrations; the Statham accelerometer did not. These data show the raw output of the filter, a smooth signal rising in proportion to acceleration. No discrete voltage levels were detected.
The Court concludes that the device Weaver tested in May was similar to the one Doane tested in June, except that it contained a timing circuit of some kind. It lacked a plurality of voltage sensors, a plurality of storing and recording means, and means to pass information from the storing to the recording means. The prototype did not embody Weaver’s claim 1. The Court has already held that Weaver’s device was not obvious in view of the Morris ’680, which contains a transducer, filter, voltage sensing circuitry, and timing means. Therefore, Weaver’s ACA was not obvious in view of its prototype, which comprised the same elements as the Morris ’680.
D. Failure to Disclose Best Mode
Patent specifications must describe the invention so clearly “as to enable any person skilled in the art ... to make and use the same, and shall set forth the best mode contemplated by the inventor for carrying out his invention.” 35 U.S.C. § 112. The defendant has shown that when Weaver filed his application he knew of a better filter circuit than the one disclosed.
In Morris’s opinion, capacitor 38 and resistor 39 (shown in patent Figure 4) would not perform as the patent requires; that is, they would fail to compensate for the slightly underdamped characteristics of the LCR filter preceding the operational amplifier 33. He also predicted that the filter would cause the operational amplifier to oscillate and burn out prematurely. This oscillation could even trigger the voltage sensors at wrong points, making the device inoperable.
The plaintiff’s expert, Robert Rea, tested a filter circuit built of modern components adjusted to approximate the system Weaver disclosed. In Rea’s opinion the Weaver device would be stable within the loading range foreseen by the patent.
The Court sees no reason to umpire this battle of the experts. Morris has conceded that in 1967 a person of ordinary skill knew means for filtering acceleration signals, and knew them very well. “A patent is invalid only when those skilled in the art are required to engage in undue experimentation to practice the invention.” *622W.L. Gore, 721 F.2d at 1557 (emphasis in the original) (citation omitted). Even if Weaver's filter malfunctions, a person skilled in the art could easily substitute another.'
And even though Weaver himself knew of a better filter, in a § 112 defense this Court may consider “only evidence of concealment (accidental or intentional)____ That evidence, in order to result in affirmance of a best mode rejection, must tend to show that the quality of an applicant’s best mode disclosure is so poor as to effectively result in concealment.” In re Sherwood, 613 F.2d 809, 816 (C.C.P.A.1980) (emphasis in the original), cert. denied sub nom. Diamond v. Sherwood, 450 U.S. 994, 101 S.Ct. 1694, 68 L.Ed.2d 193 (1981). Weaver did not conceal his invention.
E. Failure to Invent
A person is not entitled to a patent if “he did not himself invent the subject matter sought to be patented.” 35 U.S.C. § 102(f). The defendant contends that Weaver procured each element of his invention from others and designed none of them himself; Weaver therefore is not the inventor.
This argument has no merit. Weaver’s invention combines known elements. A combination is not unpatentable merely because each component is old. E. g., In re Stewart, 222 F.2d 747, 752, 42 C.C.P.A. 937 (1955).
F. Conclusion
A patent is presumed valid, and the party asserting invalidity bears the burden of persuasion. 35 U.S.C. § 282. No matter what the basis for attack, the challenger must overcome the presumption with clear and convincing evidence. E.g., Pennwalt Corp. v. Akzona, Inc., 740 F.2d 1573, 1578-79 (Fed.Cir.1984).
The defendant has failed to persuade this Court that Weaver’s patent is invalid.
INFRINGEMENT
To determine if a patent has been infringed requires two steps. The court must first study all relevant patent documents to ascertain what the claims in issue mean. It must then read those claims on the accused structure. To learn that the claims read literally on the accused structure is to learn little. The structure infringes the patent only if it does the same work in substantially the same way to accomplish substantially the same result. Autogiro Co. of America v. United States, 181 Ct.Cl. 55, 68, 384 F.2d 391, 401 (1967), reh’g denied, 184 Ct.Cl. 801 (1968).
A. Scope of the Claims
Weaver’s claim 1, upon which the other claims depend, originally read as follows:
1. An accelerometer for counting the number of times each of a plurality of acceleration maneuvering loadings of predetermined magnitude of an aircraft occur, comprising means adapted to be mounted to an aircraft for sensing acceleration loading thereof, means providing a signal proportional to said accelerations, a plurality of sensing means each responsive to the signal reaching a predetermined value, and means responsive to each of said sensing means for recording the number of times each of said sensing means senses a predetermined value of the signal.
On April 15, 1970 the patent examiner rejected claim 1 as anticipated by U.S. Patent No. 2,968,952 (Stalder). The examiner then rejected all the claims as obvious in light of prior art. But he added, “Claims setting forth the specific means used for the circuit or counter actuation would be given favorable consideration.”
Weaver did not traverse the examiner’s objections. Instead, Weaver amended his claim 1 and added the following language:
a plurality of acceleration level recording means, timing means for timing a predetermined cycle and furnishing a signal indicative of the end of said timing cycle, means responsive to an acceleration signal reaching a reference level for causing said timing means to initiate a timing cycle, and means arranged to pass a sig*623nal from each of said sensing and storing means to one of said recording means by the signal produced by said timing means at the end of a timing cycle.
Weaver’s attorney argued that the additions made the invention patentable:
Claim 1 as now presented clearly patent-ably distinguishes over the cited references. None of the cited or applied references discloses a system in which time discrimination is made between various loading (sic) and aircraft maneuvering acceleration loading. This feature of applicant’s invention is now particularly pointed out in claim 1 by the recitation of the recording means being effective to receive an acceleration loading signal only if a loading exists through a predetermined time cycle. Claim 1 further brings out that a signal is generated at the end of a timing cycle which enables the passage of the acceleration loading indication from each of the sensing and storing means to warn (sic) of a recording means.
Weaver amended claim 1 once again “to provide more concise terminology in the last three lines thereof, without changing the substance thereof.” The Patent Office approved claim 1 as amended. The last lines now read: “means responsive to said end of cycle signal arranged to pass a signal from each of said sensing and storing means to an associated one of said recording means.”
The Court must interpret Weaver’s claims in light of their file wrapper or prosecution history in the Patent Office. See, e.g., Farrell Marine Devices, Inc. v. United States, 179 Ct.Cl. 790, 804, 377 F.2d 560, 568-69 (1967) (citing Graham v. John Deere Co., 383 U.S. at 33, 86 S.Ct. at 701). The file wrapper shows that Weaver obtained claim 1 of his patent because he incorporated a “time discrimination feature” into a claim that the examiner otherwise would have rejected as anticipated and obvious. The Court cannot now disregard qualifications upon which Weaver relied to secure allowance of his claim. See id., 179 Ct.Cl. at 805, 377 F.2d at 571.
The time discrimination feature of the Weaver patent comprises means that start the timer when acceleration reaches the reset level, and means responsive to the end of cycle signal that pass information from storing means to recording means. The claim specifies results; it does not describe the structures that produce those results. The Court therefore must construe the means “to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.” 35 U.S.C. § 112.
In this case the Court has two reasons to narrow the range of those equivalents to what the specifications disclose. First, UMC neither produced its ACA nor licensed others to produce it; UMC holds a mere “paper patent.” See, e.g., International Glass Co. v. United States, 187 Ct.Cl. 376, 382, 408 F.2d 395, 398 (1969). Second, the record places Weaver’s ACA in a crowded field. His timing system is similar to those in the Maxson, Giannini, and UMC-A devices. Compare claim elements 1(g & h) on page 8 with the same elements on page 10.
The Court must therefore limit the scope of claim 1 to an accelerometer having the timing means revealed in the specifications. See Farrell Marine Devices, 179 Ct.Cl. at 806-07, 377 F.2d at 572.
B. Reading Weaver’s Claims on the Accused Device
Operation of the SDC ACA
In the accused SDC ACA, a filtered analog transducer generates an electrical signal proportional to acceleration. The current from the filter drives two voltage comparators, one of which determines only if the signal has reached the reset level. Once the signal reaches the reset level, the comparator and associated components raise to +7.8 VDC the voltage on pin 3 (shown on the upper left of the block diagram of the Systron-Donner Indicator). The other comparator works with a binary counter and other components to encode into discrete voltages information that the *624acceleration signal has reached present load levels. For example, when the signal exceeds load level 1 (LL1), the system sends a high level output voltage of +12.5 VDC to pin 3. When the acceleration exceeds LL2, the voltage on pin 4 rises to + 6.3 VDC. The other load levels have associated voltages of +11.5 (LL3) and +18 (LL4).
When the acceleration signal exceeds the reset level, components CR5, Qll, and R23 initiate a timing cycle that is measured by timing means Q12, Q13, and R25-C6. At the end of the timing cyele, Q12 turns off Q13 and Q14. Transistor Q14 then passes a “Master Enable” signal to the four “AND” gates that drive the counters. But the end of the timing cycle alone does not drive the counters.
When the acceleration signal exceeds a load level, the information is stored in the “AND” gates. If the acceleration signal then falls below the reset level after the end of a timing cycle, pulse generator Q4-Q5 produces a 60 millisecond electrical pulse, which travels to the “AND” gates through buffer Q6. By now, the gates have received information that load levels have been exceeded, the timing cycle is complete, and the acceleration level has fallen below reset. Under these conditions, the “AND” gates send signals through voltage comparators that discriminate among the various load level signals. These signals drive the appropriate counters and record the accelerations experienced by the aircraft. The end of the 60 millisecond pulse triggers the 1 millisecond pulse generator Q7-Q8, which resets the device.
No Literal Infringement
If broadly interpreted, Weaver’s claims literally read upon the SDC ACA, which contains means to generate a signal proportional to acceleration, a plurality of sensing, storing, and recording means, means that start a timer when the signal reaches a reference level, timing means that produce a signal at the end of a cycle, a filter, gates associated with counters, and a disabling means.
*625
The defendant argues that nothing in the device corresponds to Weaver’s claim 1(h), “means responsive to said end of cycle signal arranged to pass a signal from each of said sensing and storing means to an associated one of said recording means.” In the accused device, the means responsive to the end of cycle signal passes only a “Master Enable” signal to the gates and does not transfer information from the gates to the counters. Morris has admitted, however, that the counters will not advance—in fact, the device will not operate—without the “Master Enable” signal. That signal is therefore “arranged,” along with other components, to transfer information to the counters.
But a device literally infringes a patent only if the device reflects the claims as properly construed. See, e.g., Atlas Powder Co. v. E.I. DuPont De Nem-*626ours & Co., 750 F.2d 1569, 1579 (Fed.Cir.1984). The Court has already refused to ascribe to the plaintiff’s claim a meaning as broad as the literal purport of its language. If limited by the specifications, the claims of the Weaver patent do not read on the SDC ACA.
No Actual Infringement
Even if Weaver’s claim 1 could bear its literal meaning, the SDC device would not infringe Weaver’s patent, because the two ACA’s do not perform the same work in the same way to accomplish substantially the same result. See Autogiro, 181 Ct.Cl. at 66, 384 F.2d at 399-400.
As soon as the timing cycle of the Weaver device ends, release gate 54 sends information from the latches to the counters, which half index. The counters register a full count later, when the acceleration signal sinks below the reset level. In the SDC device, however, the 60 millisecond pulse transfers information to the counters after the end of a timing cycle, but not until the acceleration signal has sunk below the reset level. This difference in function leads to different results.
Both ACA’s use electromechanical counters, which react to electrical stimulation only after a delay. That delay varies with temperature. A drive pulse lasting 40 milliseconds would drive the counters at any likely temperature. The SDC device’s 60 millisecond pulse guarantees that the counters will never fail to count. But in the Weaver ACA, the drive pulse from release gate 54 begins at the end of a timing cycle and stops when acceleration sinks below reset. If the timing cycle should end and the acceleration signal should sink below reset at the same moment, the counters would not advance. And if, for example, the signal should sink below reset only 5 milliseconds after the end of the timing cycle, the counters probably would not advance. Pulses of other short durations might activate one counter but fail to activate another.
On the other hand, after the end of a timing cycle the Weaver device continues to produce a counter drive pulse until acceleration sinks below reset, no matter how long the period. For example, if the aircraft should bank in a circle and exceed an acceleration level for 20-30 seconds, the pulse would continually energize the counters, subjecting the ACA to heat and stress.
The Court concludes that the two ACA’s transfer and record information in substantially different ways and can produce substantially different counts.4
The SDC ACA File Wrapper
During the prosecution of his ACA patent, Morris tried several times to provoke an interference with the Weaver patent by copying Weaver’s claim 1. To one of his amendments Morris added a Declaration under Rule 131, swearing that he conceived his own invention before Weaver’s filing date. To another amendment he added the following argument:
The counter actuation in Weaver may occur at an earlier point in time than in MORRIS, but the accumulated counts are the same in number. Whether the latching voltage comparator signal is joined by the necessary signal to produce a count earlier or later is of no consequence once the predetermined digital delay period has elapsed. Thus the limitation in Claim 1 of Weaver stating that a signal arises “indicative of” the end of the timing cycle is not a material limitation. As shown in MORRIS it may appear at or subsequent to the end of the *627timing cycle and produce identical results.
According to the plaintiff, Morris has admitted that his ACA and Weaver’s work in substantially the same way to accomplish substantially the same result.
The Court disagrees. Morris copied Weaver’s claim to provoke an interference. The Court cannot interpret this standard applicant’s maneuver as an admission that the two devices have similar structure, operation, and results. See, e.g., Foster Cathead Co. v. Hasha, 382 F.2d 761, 766 n. 5 (5th Cir.1967), cert. denied, 390 U.S. 906, 88 S.Ct. 819, 19 L.Ed.2d 872 (1968). In fact, Morris’s argument that Weaver’s claim 1 reads on the SDC device failed to convince the patent examiner (the same patent examiner who reviewed Weaver’s application). The examiner found Morris’s copy of Weaver’s claim “misleading in that it reads as if timing means 23 activates gates 21 when time runs out whereas gates [are] activated when acc. [acceleration] signal drops.” Like this Court, the patent examiner found that the two devices do not perform in the same way.
CONCLUSION
The defendant has not persuaded the Court that the Weaver patent is invalid, but the plaintiff has failed to show that the patent is infringed. The Clerk will dismiss the complaint.
*628
*629
*630APPENDIX III
. The plaintiff argues that the filter 34, 38-39— not the motion sensing element—provides a signal proportional to acceleration, because the phrase "acceleration maneuvering loadings” in the preamble modifies the word "acceleration” in element 1(c) (though it does not modify the same word in 1(b)). The plaintiff also argues that the unfiltered signal would reflect transient accelerations and would not be proportional to maneuvering loadings; therefore, element 1(c) claims the filter and does not read on the cited prior art.
This strained linguistic argument does not persuade the Court. See, e.g., Lemelson v. United States, 752 F.2d 1538, 1551 (Fed.Cir.1985). Furthermore, the filter appears in claim 2.
. The United States Court of Claims was usually willing to sustain a patent by reading the claims in light of the specifications. (A rare exception is General Electric Co. v. United States, 228 Ct.Cl. 192, 197, 654 F.2d 55, 58 (1981) (en banc)). Other courts refused to read into a claim restrictions that it did not contain. The Court of Claims seems to have adopted the less common approach to claim construction. See generally SSIH Equipment S.A. v. U.S. International Trade Commission, 718 F.2d 365, 384-86 (Fed.Cir. 1983) (dissent from en banc decision) (discussing "majority" and "minority” positions). The Federal Circuit has resolved the conflict by holding that the Patent Office should interpret claims without limitations found only in the specifications, because the applicant can narrow his claims during the examination; courts, however, should construe claims to sustain their validity. In re Yamamoto, 740 F.2d 1569, 1571-72 (Fed.Cir.1984).
. The plaintiff offered no evidence that Aero-dyne's device was designed by a person of ordinary skill. The defendant suggests that the designer may have been an inexperienced high school graduate, and argues that Aerodyne’s bid is therefore irrelevant.
The finder of fact, however, may reason from any inference that reasonable minds can draw from the evidence, even if the inference relies in part on conjecture. E.g., Daniels v. Twin Oaks Nursing Home, 692 F.2d 1321, 1326 (11th Cir. 1982), reh'g denied, 698 F.2d 1238 (1983). The finder of fact may even draw inferences from inferences provided that each is reasonable. Id. at 1324.
Here the record shows that the Navy sent the 1968 IFB to 33 companies and received thirteen technical proposals, which Navy technical personnel reviewed. They rejected ten as unacceptable; they admitted Aerodyne to the second stage of the competition.
The Court infers that Aerodyne submitted a bid hoping to win the contract, and that Aero-dyne acted as a prudent business organization. The Court concludes that such an organization would not endanger its chances of winning a contract by entrusting its technical proposal to a person of less than ordinary skill. Aerodyne’s success in the first step of the procurement supports this conclusion.
. The plaintiff argues that SDC has merely improved Weaver's device by adding features. The Court disagrees, but would find no infringement even if it accepted the plaintiffs contention. Weaver himself obtained a patent on a narrow improvement. "Accordingly, the patentee cannot prevent others from making improvements on the prior art unless they use substantially the very novelty which is the basis of his patent.” Maytag Co. v. Murray Corp. of America, 318 F.2d 79, 84 (6th Cir.1963); accord Thomas & Betts Corp. v. Litton Systems, Inc., 720 F.2d 1572, 1580 (Fed.Cir.1983).