Re: An Explanation of Dayton Miller’s Anomalous “Ether Drift” Result

Sergey Karavashkin wrote:
I have looked through your paper yesterday. It is written vaguely, no
one to be able to understand, what you have subtracted from what:

<< Therefore if the data from the first turn is subtracted
marker-by-marker from the data of every turn, the result is completely
independent of any orientation dependence, and contains only
systematic(time). >>

I don't know how to state that more clearly, when the nearby text is included.


The matter is, the data of first turn contained errors and masking
effects the same as others. Subtracting, you on one hand as if selected
the systematic error, but on the other you introduced the error of the
first turn into the data for all next, which could sufficiently change
your result.

This is indeed true for the statistical errors related to the data taking. These are quite small (~0.1 fringe), and are accommodated in the fit to the systematic drift. This is not true for the systematic drift itself.

This does NOT "sufficiently change [my] result", it merely contributes to the errorbar of the fit. The statistical errors are inherently included in the fit.


Similarly, suggesting
<< It is convenient to take advantage of the 180 degree symmetry of the
apparatus, and combine the data for markers 180 degrees apart. This
gives 8 orientations, and 8 independent measurements of differences in
systematic(time), shown in Fig. 10 >>,

you proceed from the idea that Miller had to obtain ideal
characteristics. Factually such symmetry is basically impossible,

Not really. The 180 degree symmetry is precise to within a few wavelengths of light -- this is necessary for the fringes to be visible with the central fringe near the image of the pointer. This is far more accurate than necessary for this combination of points 180 degrees apart. For instance, the effect due to the rotation of the earth is VASTLY larger, but is also negligible (footnote 7 on page 6).

The 90 degree angle between arms is not nearly so precise, but is adequate; errors in that angle do not affect the result (as long as it remains fixed, which it does).


Miller was in this sense more correct than you are. He handled the data
of each line separately in supposition that the full-turn effect is
linear (which is seen in your Fig. 10).

You need to READ THE PAPER. Miller's systematic drift is not anywhere close to linear -- for the run I discuss in detail the difference from linearity is at least 0.5 fringe, but the false signal he found is only 0.06 fringe -- the nonlinearity is VERY important, and in the final analysis it is shown to be the source of the false signal he found.

Miller did not "handle the data of each line separately" -- he SUMMED all lines as the first step in his analysis. This is a major problem of his algorithm, and is a major part of the reason his results are flawed.

My fig. 10 does not show any sort of "linearity" -- look at Fig. 3 and the accompanying text, which shows quite clearly that the drift is NOT linear, and deviates from linearity by as much as 1/2 fringe per turn. That is an order of magnitude larger than the false signal Miller found, so his assumption of linearity is just plain wrong at the level of his false signal.


So he avoided to transfer the
systematic errors from line to line. After such handling he already
might average the results without systematic errors which you
introduced to your computation.

You obviously do not understand how systematic errors propagate in Miller's analysis. READ THE PAPER.


<< The 6 km is an UPPER BOUND AT THE 90% CONFIDENCE LEVEL >>

I would say, first, you have undervalued the data. Miller’s speed –
not your 6 but 9,3 km/s – has been derived from same raw data which
you used by way of compensation of the full-turn effect.

The "full turn effect" is irrelevant. Both Miller and I used the 1/2 turn Fourier amplitude for the signal. But his analysis is flawed. READ THE PAPER.

My "6 km/s" is NOT a speed, it is an UPPER LIMIT on any such speed.

Miller obtained his "9,3 km/s" via a flawed analysis algorithm that made his systematic drift look precisely like the signal he expected. You need to READ THE PAPER.


BTW the full turn effect is predicted by certain ether theories, but is predicted to be zero by SR. There is no obvious evidence for it in the data -- see Fig. 7 in which the full-turn amplitude is in bin 1; it is small and reasonably consistent with the values for other amplitudes that clearly cannot be any sort of signal (bins 3-7).


Concerning, whether SRT predicts zero or non-zero effect, could you
relativists first agree among yourselves, and re-read Einstein where he
referred to the negative result of MMX. :)

Everyone who understands SR agrees that SR predicts a null result for the MMX and for Miller's data, neglecting minuscule effects that are significantly smaller then their experimental resolutions.


Tom Roberts
.