Subversion Repositories SmartDukaan

Rev

Rev 37435 | Rev 37437 | Go to most recent revision | Show entire file | Ignore whitespace | Details | Blame | Last modification | View Log | RSS feed

Rev 37435 Rev 37436
Line 401... Line 401...
401
     * These are OPPO's tuning values and are deliberately independent of realme's.
401
     * These are OPPO's tuning values and are deliberately independent of realme's.
402
     * The two classes drive the same dingxiang widget but convert very differently
402
     * The two classes drive the same dingxiang widget but convert very differently
403
     * (oppo ~12%, realme ~60% of completed drags), so a change proven on one is NOT
403
     * (oppo ~12%, realme ~60% of completed drags), so a change proven on one is NOT
404
     * evidence for the other. Tune and measure them separately.
404
     * evidence for the other. Tune and measure them separately.
405
     *
405
     *
406
     * 1.7 was a guess carried over from a single hand measurement. Measured over 3,694
406
     * Back to 1.7 after r37433 set this to 0.67 and it did not help (14% -> 13%).
-
 
407
     *
407
     * real oppo calibrations in production the ratio is 0.67 (realme's is 0.70, on its
408
     * 0.67 was fitted to 3,694 production calibrations -- but every one of those was
-
 
409
     * measured while the slider press was being leaked between attempts (r37435), so the
-
 
410
     * sample was contaminated and the value fitted to noise.
-
 
411
     *
-
 
412
     * With that confound removed the clean evidence is realme: identical widget, same
-
 
413
     * calibration code, ASSUMED_RATIO 1.7, converting 82%. Measured post-fix on similar
-
 
414
     * gaps, 1.7 probes ~31px and closes ~11% of the gap, while 0.67 probes ~52px and
408
     * own 1,683). At 1.7 the probe closes about a fifth of the gap instead of the half
415
     * closes ~37%. The SMALLER probe is the one that lands -- a probe that eats a third
-
 
416
     * of the gap risks crossing the target during calibration, and |c1-c2| cannot tell
409
     * it is meant to, so the ratio is derived from a needlessly small, noisy sample.
417
     * "not there yet" from "gone past", so the derived ratio comes back wrong.
410
     */
418
     */
411
    private static final double ASSUMED_RATIO = 0.67;
419
    private static final double ASSUMED_RATIO = 1.7;
412
    private static final int MIN_NUDGE = 10;
420
    private static final int MIN_NUDGE = 10;
413
    private static final int MAX_NUDGE = 60;
421
    private static final int MAX_NUDGE = 60;
414
    private static final double ALIGNED_PX = 2.0;
422
    private static final double ALIGNED_PX = 2.0;
415
 
423
 
416
    private static int nudgeFor(double gap) {
424
    private static int nudgeFor(double gap) {