============ Nerd alert! Extreme glaze-over danger ahead! ============Iacopo: Many thanks for the raw data. I took the liberty of graphing and analyzing it in Excel. Sorry for the fuzzy screen-grab below.
The plot puts time
t in seconds on the horizontal axis and angular speed
w in rad/sec on a logarithmic vertical axis so as to render purely exponential trends as straight lines. Linear trends appear as curves strongly concave toward the origin. The curve on the right represents the vacuum data.
Each of the tan boxes reports the result of a regression analysis (curve-fitting) on a portion of the data. I ran 3 fits on each data set: An exponential fit at the highest speeds and a linear and exponential fit at the lowest.
Vacuum data: From t = 0 to t = 1,680 sec, decay is purely exponential with a decay constant of
kVHE = -0.00056 /sec and a near-perfect correlation (
RVHE2 = 0.99817). The "VHE" subscript identifies the fit as "vacuum, high-speed, exponential."
From t = 2,540 sec to the end of the series at t = 3,840 sec, the vacuum data best fits a linear decay with a decay constant of
kVLL = -0.0272 /sec and a near-perfect correlation (
RVLL2 = 0.99823). The "VLL" subscript here stands for "vacuum, low-speed, linear."
However, this same tail in the vacuum decay curve also fits an exponential decay with
kVLE = -0.00466 /sec and
RVLE2 = 0.97465.
Air data: From t = 0 to t = 1,800 sec, decay is purely exponential with a decay constant of
kAHE = -0.00115 /sec and a near-perfect correlation (
RAHE2 = 0.99928). The "AHE" subscript identifies the fit as "air, high-speed, exponential."
From t = 2,340 sec to the end of the series at t = 2,580 sec, the air data best fits a purely linear with a decay constant of
kALL = -0.0264 /sec and a near-perfect correlation (
RALL2 = 0.99852). The "ALL" subscript identifies the fit as "air, low-speed, linear."
However, this same tail in the air decay curve also fits an exponential decay with
kALE = -0.00466 /sec and
RALE2 = 0.98142.
Observations: At least 4 observations to note here:
(i) Your top spins down through 3 distinct decay regimes whether or not air is present: (a) Highly exponential decay at the highest speeds, (c) highly linear decay at the lowest speeds, and (b) an apparent combination of the two at intermediate speeds.
(ii) The fact that
kAHE is about twice
kVHE points to a significant boost in high-speed exponential decay due to aerodynamic drag.
(iii) The near-equality of
kVLL and
kALL, and of
kALE and
kVLE, suggests that the same dissipative processes control decay at the lowest speeds in both vacuum and air.
(iv) The low-speed exponential fits to the vacuum and air data aren't quite as tight as the linear fits, but they're still pretty darned good. (
R2 > 0.974 in all cases.) This isn't a big surprise, though. The farther one goes out on the tail of any truly exponential decay curve, the more linear it looks -- even to a regression analysis.
I find observation (i) rather counter-intuitive on the vacuum side.
Preliminary discussion: Overall, the results bear out Alan's statement that spin decay is exponential at the highest speeds and linear at the lowest -- at least in your top.
What really goes on at a top's tip-substrate "contact patch" is far from settled in the physics literature, but the 2 main candidate processes are sliding friction and rolling resistance. Since both generally show little speed dependence, one would expect that the braking torque due to either one or both would be largely independent of speed. The linear decays at the lowest speeds in air and vacuum are consistent with that expectation, but the exponential decay at the highest speeds in vacuum is strongly contradictory.
Observation (i) suggests at least 3 possibilities, not mutually exclusive: (a) For some reason, sliding friction, rolling resistance, or both, are strongly speed-dependent at high speed in your test setup. (b) Other speed-dependent contact patch processes dominate high-speed dissipation in a vacuum. (c) The vacuum was imperfect.
One last point: Now that we've established exponential decay in the high-speed regime, we have to let go of the idea that spinning tops are subject to braking torques that grow with the square of angular speed. Exponential spin decay requires a total braking torque that's linear in angular speed, not quadratic. Linear decay, on the other hand, requires a braking torque independent of speed. I'll back up these statements in a separate math post.