L. HAND, Circuit Judge.

This is a suit in equity:in the:usual form, for the infringement of a patent ion a machine to fashion wooden heels for woft men’s shoes. The only issue is as to the priority of invention between Sawyer, the patentee of the patent in suit, and Pope and Mann, the joint patentees of the patent under which the defendant manufactures; No. 1,669,144 issued May 8, 1928. In the District Court the defendant argued in addition that it did not infringe, and that the art had anticipated Sawyer’s invention;, but these defences are not pressed here and we shall disregard them. Thus the controversy comes down to which of two inventors was the earlier, and whether he was, duly diligent in reducing his work to practice. The dates are as follows: Sawyer filed his application on January 12, 1927;’ the date of his invention was April, 1923; it was reduced to practice in June, 1924. Pope and Mann’s application was filed on September 29, 1925; it may be assumed to have' been reduced to practice in March, 1925, because by that time a machine had been-constructed which was capable of turning out heels and did so. As there can be no just complaint of delay after discovery, the whole case turns upon the actual date of ■ Pope and Mann’s invention, which the defendant claims was fully conceived and put into drawings and a machine made, by ■ March 18 or March 28, 1923. There is at the outset a dispute as to the degree of proof: necessary upon this issue. The defendant, •> relying upon Willard v. Union Tool Co., 253 F. 48 (C. C. A. 9), maintains that the bur *264 den rests upon the plaintiff to show beyond doubt that Sawyer as junior applicant antedated Pope and Mann; but that the defendant in its turn may carry Pope and Mann’s date back of any date shown for Sawyer by a mere balance of probability. But the law is not so, whatever may be the practice in the Patent Office. When an inventor’s date is to be carried back beyond his application, courts regard the effort with great jealousy, and must be persuaded with a certainty which is seldom demanded elsewhere; quite as absolute as in a criminal case, in practice perhaps even more so. Brooks v. Sacks, 81 F. 403 (C. C. A. 1); Dey Time Register-Co. v. W. H. Bundy Recording Co., 178 F. 812 (C. C. A. 2); Bearings Co. v. Harris Hardware & Mfg. Co., 299 F. 782 (C. C. A. 2). It makes no difference how the question arises; whether the patentee is carrying back his own invention, or a supposed infringer is carrying back his; the burden is the same as the proof necessary to establish a prior use. We agree with the judge in his statement of law, but it seems to us that the documentary and physical exhibits establish the earlier date of Pope and Mann’s invention beyond reasonable doubt.

Sawyer’s machine was much the more elaborate; it makes the whole “breast” of the heel from a rude block in one operation. Not so Pope and Mann’s, which receives the heel when the upright portion of the “breast” has been already formed and only finishes it. The heels in question are so high that the foot is at a very substantial angle to the floor. As the “breast” should be perpendicular, the “lip” will be at an angle to it, for the “lip” is a thin continuation of the top of the “breast,” where it is fastened to the outside of the “shank” or sole. The outside of the “shank” is convex in transverse section; and as the heel must fit upon it, its top surface must be concave, but Pope and Mann’s machine has nothing to do with making it so. The “lip” extending forward along the “shank” makes the transition between “shank” and “breast”; to do this agreeably, and receive a strip of leather which covers the “shank” and runs down the “breast,” its outer surface must be convex transversely like the “shank,” and the transition curve between “lip” and “breast” must also be agreeable in longitudinal section. Pope and Mann’s machine made this convex surface of the “lip,” and the longitudinal transition between “lip” and “breast,” both of which had formerly been done by hand. The workman held the nearly completed block and rotated the unfinished “lip” against the edge of a rotating rasp which planed its outer surface into the proper transverse convex arc to blend with that of the “shank,” and into a longitudinal arc to blend “lip” with “breast.” Mann conceived the notion of a holder upon which the heel should be mounted and which should similarly rotate it. There is no doubt that he did construct a machine of some sort which would do this, and the issue is narrowed to the precise kind of machine he made, particularly whether the angle of the axis of the holder to that of the rasp could be varied. This last feature is important in order to accommodate the machine to heels of different heights. The cutting surface of the rasp is not circular in cross-section and does not touch the heel with the whole of that surface. It therefore follows that the curve by which the “lip” blends into the “breast” will vary in form according to what section of the rasp does the cutting. The higher the heel the smaller the angle between the “lip” and the “breast”; and the smaller the angle, the shorter should be the radius of the arc of transition. This accommodation depends upon the power to vary the angle between the axes.

Mann swore that he had conceived the notion of finishing the “lip” by a machine as early as 1917, but this we disregard. He had a rasp, made in 1921, which Pope produced and Mann identified; Mann also produced a cone which he swore' that he had mounted on a holder, and which he used to revolve the heels against the surface of the rasp. The mounting he did not produce; at the trial the cone had been fitted into a confessedly new holder. We have no adequate proof of how old the cone was. The defendant called two witnesses, Chase and Stone, mechanics indeed, but not acquainted with this kind of machine. They declared that in the autumn of 1922 they had visited Mann and seen him operate some kind of machine for finishing heels, but their identification of the cone machine was not certain enough to serve. Pope visited Mann late in 1922, and Mann undoubtedly interested him in some machine which he had then constructed. Both say that the cone had by then given place to another kind of holder; it had in any case been an extremely clumsy makeshift; it is not apparent how the heel could be fastened to it at various places. According to Mann it was soon discarded and Pope never saw it mounted. *265 In its place the heel was set on a holder which rotated in an arc of variable radius; but whether it had power to change the arc of transition between “lip” and “breast” depends, as we have shown, upon whether the axes were in variable angular relation to each other.

On this the critical issue, the evidence so far stated will hardly serve. The original holder has quite disappeared and no satisfactory explanation is made of its disappearance ; we know nothing of it except what we get from the testimony of interested, or not well advised, witnesses and even this is not clear. One of them, Hart, for example, identified the cone as the holder which he saw in Pope’s shop in the spring of 1924, clearly an error. But on physical exhibits we may more safely rely. Emerson & N. Co. v. Simpson Corp., 202 F. 747 (C. C. A. 1); Drum v. Turner, 219 F. 188 (C. C. A. 8). Mann produced three photographs of what he said was the machine which he gave to Pope; and these were taken just before Pope carried it off to his shop to be perfected. The date of these remains in doubt; Mann says that they were taken early in 1923, in March so far as he could recall; Pope says that he took the machine to his shop in January of that year. We do not find the date established with that certainty which is necessary, but it does appear to us that the photographs, whenever taken, were of a machine in which the holder was at an angle to the rasp. This is apparent especially in Exhibit F. 1, which was taken along the axis of the holder which faced the camera; the rasp certainly is not perpendicular. The pictures are not themselves identified by any date; but they were taken in the air outside some clapboarded building, painted white. -Mann swore that it was the outside of his dwelling in a small hamlet; and certainly the building has not the appearance of an “old machine shop” in Haverhill, Mass., where Pope’s shop was. Unless they were fabricated for the purpose, we should believe that they were taken before Pope took the machine to Haverhill. That does not, however, fix the date, and without more we should still be at fault.

To fix it we rely upon certain drawings of Pope which are dated in March, 1923. These were working drawings based on the model, and being established as of March 28, 1923, at the latest (one was on the eighteenth and nineteenth), they fix the time before which Mann’s machine came to Pope’s shop in Haverhill. They also tell us of the construction of the machine at that time, or at least of the machine which it was proposed to build. Exhibit P. is plainly a working drawing for a heel holder; and although it cannot be positively identified as the same, it looks very much like the corresponding part of the machine shown in Exhibit C. It throws no light upon the angle between the axes, but it does show that the details were already worked out, and that there was to be a holder which rotated the heel against a rasp. Exhibit Q. reveals much more; it contains two drawings of which one is a plan of the base on which both rasp and holder are to be set, and the other an elevation of the rasp holder which tells us nothing. The plaintiff cannot reasonably dispute any of this, but it does assert that the angle of the holder to the rasp nowhere appears, and that the crux of the invention therefore stands as unsupported as before. We cannot agree. It seems to us that from the date of these drawings, it necessarily follows that the machine of the photographs was in Pope’s shop before April, 1923, and as we have said, the axes of that machine were set at an angle. But we are not confined to this, for the drawing of the base shows to some extent how the holder was to be used. In the center is an opening in the form of a cross with equilateral arms, certainly meant to allow for the adjustment of the holder in two directions normal to each other. This cross does not point along the axis of the rasp; although its center is on that axis, the arms are at an angle to it. No one has suggested any reason for such an angle, if the axes were to be perpendicular. Furthermore, if they were to be, it is not apparent why the holder should have had any lateral adjustment to the axis of the rasp. While the axes are perpendicular the rasp must always cut with the same arc of its surface, and the “back-gauge,” which appears on Exhibit P., would seem to give room for such lateral adjustments as might be required for heels of varying depth. On the other hand, the cross, if set angularly to the axis of the rasp, would have a function if the angle of the axes was to vary, which it could not have if it were not so set. In order to decrease the radius of the transition cut, the angle of the axis of the holder to that of the rasp must become more acute; the cutting surface of the rasp is so made. But as the axis of the holder is swung to an angle acute to that of the rasp, the heel will move in a circle, that is, in two directions relative to *266 the rasp ; it will move towards the rasp in the line of its axis; and towards its center in a line perpendicular to the axis. Both 'these1 motions cannot be compensated for by a movement of translation of the holder along either arm of a cross set perpendicularly to the axis of the rasp; but both may be compensated for by a diagonal 'movement which, like the circular movement of the heel, is a component of two 'movements, one parallel with the axis, of the ’rasp’,' the other perpendicular to that axis. 'In short, a cross set at an angle will give 'four possible compensatory movements of translation to the holder, each of which can 'neutralize the change in position of the heel 'düé fo its circular swing when the axes change’their angle. Thus not only was the 'angular position of the cross meaningless if the axes were to be perpendicular, but it 'was necessary if the angle between them 'was-to vary.

Óür conclusion is consonant with the probabilities. If Mann ever used a cone, it had to be at an angle to the rasp, for it 'woúld not operate otherwise at all, except ■‘perhaps when the heel was set at the very ■base. Moreover, it is scarcely possible to 'suppose that any one should long have used 'such a machine as Mann’s second one without"'observing that unless the relation of 'the axes could'be changed the transition between "lip” and “breast” must be uniform no •matter what the angle was between the two. ■Of course it is possible that Mann or Pope, "or both, may have been unable to correct '•this defect until after April, 1923, though 'it did not call for much ingenuity. But the ■physical exhibits seem to us to lay any •doubts, and, in spite of the severe measure of proof required, to establish the defendant’s priority. It follows that the decree .must be reversed and the bill dismissed.

Decree reversed; bill dismissed.