In brief:
During the months of november and december I did a number of tests with different settings of the Drive Pulse Timing.
According to the article of Schumacher et al the intrinsic precession of the elliptical path can be eliminated completely by outwards driving the bob at a very precise moment in the swing. I was wondering what would happen if the timing is to early or to late.
As Schumacher's formula (19) only involves the Q, the length and the amplitude of the pendulum, I wanted to reverse calculate the Q of the pendulum and see if it matches the Q as I measured it more directly in the section Free Run and Kick, Fig 3.
Conclusions:
The precession speed varies from anti-phase with the ellipse amplitude to in-phase while changing the drive timing from early to late.
Somewhere in between the precession speed is almost constant, which is the desired situation.
The full-rotation time ΩF of the swing plane tends to increase with later timing, although from long time observations I found quite some variance in the full FP times. The theoretical value for my location at 51.94 deg North is 30:26 (30.43h).
Reverse calculation of the pendulum's Q from the optimal drive distance using Schumacher's equation (19) yields 6819, 48% more than the 4578 found in the experiment (see fig. 3 there) where I determined the Q from the pendulum's decay time. I do not regard this as a serious problem, seen the inaccuracies in the decay measurement.
The legend for all pictures below is:
White: Amplitude of the minor axis of the ellipse. Plus = CCW, minus = CW, Scale = 45 mm / div.
Blue sawtooth like line: Angle of the major axis of the ellipse
from +180 deg (West) through +90 = North, 0 = East, -90 = South, -180 deg = West.
Blue; sine shaped line: First derivative of the precession angle = precession speed. Zero = midscale.
There can be a small rotational error in the exact orientation of the pendulum's position measuring system w.r.t. the real geographical North.
The drive timing is given in ticks of the 10 kHz timing clock, after the center passage, and in mm from the center, calculated from the timing.
In Fig 8. is a drawing showing the drive positions w.r.t. the dimensions of the drive coil and the magnets in the bob.
Fig 1. Timing = 620. d = 21.7 mm.
180 deg transitions at nov 26, 21:57 and nov 28, 04:48
Full rotation was 30:51
Files: 2024_11_26-27-28.
Fig 2. Timing = 640. d = 22.4 mm.
180 deg transitions at nov 29, 11:37 and nov 30, 18:39
Full rotation took 31:02.
Files: 2024_11_29-30.
Fig 3. Timing = 660. d = 23.0 mm.
180 deg transitions at dec 01, 00:48 and dec 03, 07:07
Full rotation took 30:19.
Files: 2024_12_01-02-03.
Fig 4. Timing = 680. d = 23.7 mm.
180 deg transitions at dec 03, 13:48 and dec 05, 20:55
Full rotation took 31:07.
Files: 2024_12_03-04-05.
Fig 5. Timing = 740. d =25.8 mm.
180 deg transitions at dec 13, 15:41 and dec 14, 23:27.
Full rotation took 31:46.
Files: 2024_12_13-14-15.
Fig 6. Timing = 760. d = 26.5 mm
180 deg transitions at dec 16, 06:25 and dec 17, 14:42.
Full rotation took 32:17.
Files: 2024_12_16-17.
Fig 7. Timing = 780, Shortly after the first full revolution the timing was set to 600.
Here we can clearly see the phase jump in the precession speed.
d = 27.2, d = 21.0
180 deg transitions at dec 19, 01:00, dec 20, 11:17 and dec 21, 18:12.
Full rotation took 34:17 and 30:55.
Files: 2024_12_18-19-20-21.

Fig 8: Arrangement of Driving Coil and magnet in the Bob.
Dimensions are approximately on scale.
We can see that at all driving distances the magnet is very nice within the field of the coil.
Reverse Calculation of the Q of my pendulum.
Schumacher's equation (19) says: 3 / 4 * Q * α2 = sqrt (1 - δ2) * acos (δ) / δ
where α = a / L and δ = d / a.
In my case d = 23.7 mm, a = 230 mm, L = 4370 mm.
so: α = 230 / 4370 = 0.0526 and δ = 23.7 / 230 = 0.103.
Rewrite (19): Q = 4 * sqrt (1 - δ2) * acos (δ) / (3 * δ * α2) .
We find Q = 4 * 3.978 * 1.468 / ( 3 * 0.103 * 0.002767) = 6819.
(exact calculation made in the spreadsheet)
This result is higher than I found when I left the pendulum run freely, 4578 but not far off. See Fig 3 in the experiment