Here is a direct view of the data thanks to blue_energy.
Here is a direct view of more data thanks to blue_energy.
Here is a direct view of even more data thanks to blue_energy!
This is the parsed data from blue_energy. We're both going to try to create plots from this - we'll see who get's there first. Unfortunatly, we both have real work to do!
Second block of data is magnet only where I have added a small neo to the bottom of the arm. It is clamped and glued with caulk, but the caulk isn't quite dry yet.
Third block of data is magnet plus magnetometer which has the last two channels connected to a couple of hall probes. One has 1.3 mV/Gauss and the other has 5 mV/Gauss. You can see the pulses when the neo swings past the probes, and the 5mV/G device looks like it is saturated.
And now for the real thing.. The 1.3mV/Gauss device is placed close to the base magnet. I tried it on top, and it worked, but it starts out saturated. So off to the side it shows how the field drops when the upper magnet goes by. The 5mV/Gauss detector is farther away, and shows when the magnet on the arm goes by. There is a point near the end of each cycle when the upper arm sits over this sensore and sends it into saturation.
Ramuk plotted the data and found a problem. I hope the next batch come out better. This is the same physical set up, but the accelerometer is glued down to the arm for less vibration. The y axis should be clean, and possibly the z axis as well.
Korkscrew found another problem. So here is a re-run of the opposing magnets and here is a re-run of sensors only (no base magnet) with the sensors not moved at all.
Comments on the forum led me to take off the base magnet and replace it with a ferrite coil. The wires have been removed (it was a toroid coil for power in some ancient electronics I long sinced ripped apart) and the ferrite core is placed with the center about where the upper magnet passes over. There should be two bumps when it goes by, we'll see when the data gets plotted!
CWatters suggested adding a weight to the arm. This looks like it helps from the raw numbers alone, but it should be plotted and compared with previous runs. This has the same magnets and sensor positions, but has a nice iron core transformer taped to the arm near the accelerometer. I'm not sure how this affects the magetics (shouldn't be much), but that is not the point of this run. This is to prove mechanical vibration is present in previous runs. The run lasts longer simply because there is more potential energy at the start than in previous runs.
blue_energy has sent me a huge block of data which I have not had time to deal with. It's about 5MB, so unpack with care!
Here are some of sonoboy's pictures. These are from his experiments on Sv and forces betweens coils and permenent magnets.
Sonoboy has sent me some more info to post so here it is:
Hi Dr Mike,
Here are a few shots looking at the current flow in the differential
test rig. The diff.jpg shows the 30 amp referance coil pulse (which is the
same as the current pulse in the measurement coil) and midscreen shows the
output of the differential amplifier. There is a small glitch there with no
magnetic material present. The cermag.jpg shows the current response of the
test rig when a small ceramic magnet is very near the measurement coil. The
cerdemag.jpg shows the current response when a demagnetized ceramic magnet
of the same type is in the same position that the magnetized one was in.
This same current effect is also true with the neos only the waveform is
more complex due to the conductivity of them. As can be plainly seen, a
demagnetized magnet has a much larger effect. BTW there is no noticable
differance when reversing the polarity of the magnetized magnet. What this
has to do with COE is anyones guess...
In their data published around end of October, they have 2 pages in an
Excel spread sheet. I have taken that data and expanded each column
into 12 digit accuracy, then exported it to a tab deliminated text
file. You can see the converted data here
for the first page and here
for the second page.
I have analyzed the data in several ways. The code I used can be seen
here:
for the first page and here
for the second page. Both of these
routines put out stripped down sections of data which remove the
linear interpolation that Steorn put in. This should get us back to
the original or raw data. That raw resultant data is here:
raw_cw0.dat
raw_cw1.dat
raw_cw2.dat
raw_ccw0.dat
raw_ccw1.dat
raw_ccw2.dat
where the 0, 1, and 2 refer to each column block and cw, ccw refer to
clockwise or counter clockwise within each block.
Similarly, the second page load data is stripped down here:
raw_load_cw0.dat
raw_load_cw1.dat
raw_load_cw2.dat
raw_load_ccw0.dat
raw_load_ccw1.dat
raw_load_ccw2.dat
As a next step I looked at the distribution of angular step sizes and
torque step sizes in each of those files. I then fed that into
gnuplot and created the following outputs. I think the 2 degree step
sizes which show up in all the plots come from perfectly linear
torque steps. There are a couple of perfectly linear torque steps of
3 degrees (note the plots that have a mark at 500). The bin sizes
were arbitrarily chosen to be .005, which is 5 steps per actual
angular step, but 5/2 of the torque sensor accuracy. That is why the
plots are smooth on the green curve (torque) and histogram like on
the red curve (angle). The code which created these plots is here:
to generate
statistics. I then plotted the bin data using gnuplot and
got these results:
for file raw_cw0.dat average angular step = 1.011197 angular step deviation = 0.130518 average torque step = -0.000006 torque step deviation = 0.045450
each plot contains the global data and a zoom into the center.
for file raw_cw1.dat average angular step = 1.000000 angular step deviation = 0.079167 average torque step = -0.000448 torque step deviation = 0.045918
note that the last column is no friction and that the steps are less than 1 degree (250 marks exactly one degree in these plots).
Almost but not quite! Further analysis shows that the weight of the 2 and 3 degree steps is enough to throw the average off.
Original calculation: for file raw_cw2.dat average angular step = 1.058997 angular step deviation = 0.270945 average torque step = 0.000050 torque step deviation = 0.015977 subsequent calculation: for file raw_cw2.dat in group 0: average angular step = 0.999843 angular step deviation = 0.017705 average torque step = 0.000201 torque step deviation = 0.014786 in group 1: average angular step = 2.007353 angular step deviation = 0.026795 average torque step = -0.007529 torque step deviation = 0.027117 in group 2: average angular step = 2.962500 angular step deviation = 0.012500 average torque step = 0.040500 torque step deviation = 0.004500
So the shift in the picture is caused by the large number of points which are too linear, the actual centroid of the data is a 1 degree step to 4 decimal places.
And the counter clockwise plots look like this:
for file raw_ccw0.dat average angular step = 1.002654 angular step deviation = 0.094728 average torque step = 0.000229 torque step deviation = 0.038951
for file raw_ccw1.dat average angular step = 1.005462 angular step deviation = 0.105958 average torque step = 0.000176 torque step deviation = 0.035408
Again, notice the step size shift under free running conditons:
for file raw_ccw2.dat average angular step = 1.055882 angular step deviation = 0.278062 average torque step = 0.000068 torque step deviation = 0.015750 subsequent calculation: for file raw_ccw2.dat in group 0: average angular step = 0.999689 angular step deviation = 0.016946 average torque step = 0.000112 torque step deviation = 0.014912 in group 1: average angular step = 2.005357 angular step deviation = 0.021503 average torque step = 0.004000 torque step deviation = 0.027370 in group 2: average angular step = 3.008333 angular step deviation = 0.011785 average torque step = -0.023000 torque step deviation = 0.012083
and the second page data all looks similar, i.e. under load:
for file raw_load_cw0.dat average angular step = 0.999861 angular step deviation = 0.076936 average torque step = -0.000097 torque step deviation = 0.048608
each plot contains the global data and a zoom into the center.
for file raw_load_cw1.dat average angular step = 1.005462 angular step deviation = 0.107988 average torque step = -0.000246 torque step deviation = 0.036028
note that the last column is no friction and that the steps are less than 1 degree (250 marks exactly one degree in these plots).
for file raw_load_cw2.dat average angular step = 1.068452 angular step deviation = 0.263867 average torque step = 0.000042 torque step deviation = 0.016314 subsequent calculation: for file raw_load_cw2.dat in group 0: average angular step = 1.000241 angular step deviation = 0.018644 average torque step = 0.000164 torque step deviation = 0.015218 in group 1: average angular step = 1.996875 angular step deviation = 0.020807 average torque step = -0.001542 torque step deviation = 0.026873
And the counter clockwise plots look like this:
for file raw_load_ccw0.dat average angular step = 1.008216 angular step deviation = 0.120804 average torque step = -0.000034 torque step deviation = 0.046292
for file raw_load_ccw1.dat average angular step = 1.011056 angular step deviation = 0.131242 average torque step = 0.000186 torque step deviation = 0.030479
Again, notice the step size shift under free running conditons:
for file raw_load_ccw2.dat average angular step = 1.055882 angular step deviation = 0.276995 average torque step = 0.000068 torque step deviation = 0.015941 subsequent calculation: for file raw_load_ccw2.dat in group 0: average angular step = 0.999922 angular step deviation = 0.016187 average torque step = 0.000351 torque step deviation = 0.014906 in group 1: average angular step = 2.001786 angular step deviation = 0.030567 average torque step = -0.007286 torque step deviation = 0.027283 in group 2: average angular step = 3.000000 angular step deviation = 0.020412 average torque step = 0.004000 torque step deviation = 0.034670
Here is something totally different: An analysis of superparamagnetic fluid particles. Click here for pdf.
As requested, here are pictures of the Steorn flyer handed out at the "demo".
