Re: acmi accuracy...

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Terry Fritz" <twftesla-at-uswest-dot-net>

Hi All,

Bart wrote:
>I think there is another possibility (probability): When dimensions are 
>entered into acmi, if one of the dimensions are off, the error follows the 
>ratios of measured vs. calc'd. I think the error growth is proportional to 
>dimensional error. 
>
>The more I use acmi, the more I see the importance to enter dimensions that 
>calc to measured inductances in acmi. This is not "making" acmi work, but 
>simply giving acmi accurate information. It is taking out human error when 
>determining dimensions or it can be viewed as entering measured information 
>directly into acmi. When acmi calc's inductances as measured, the mutual 
>inductance and coefficients are very accurate. (except for my coil and 
>therefore I expect it is "my" error somewhere). 

Yes.  I need to retake the physical measurements too.  I was wrong once
about the diameter...  Thanks for pointing out that very basic source of
error.  One must be sure the use the best inputs possible now that it matters!

You also bring up a good point that if one part of the program is wrong, it
can mess up the final result.  Perhaps the coupling calculation is correct
but the primary inductance is wrong...  If the error can be traced to a
subsection by manually inputting known values and all starts to work, then
work can be concentrated to fix that section.

>My primary and secondary are wound in opposite directions. Looking down on 
>the primary, it is wound clockwise. The secondary relative to primary is 
>counter-clockwise. 

Mine are both counter-clock wise (beginning at the center of the primary
and the base of the secondary).  So we do indeed have two different winding
directions.  Perhaps that is causing our two different trend directions.
Thus, one winding direction will have more coupling than the other.  We
could use a new term or definition for this since describing the turns
directions is rather awkward...

>I need to measure inductances as well. I feel much better about acmi's 
>predictions than my own measurements. 

There is a point in computer modeling where one indeed does start trust the
computer more than one's measurements.  That is when you know the model is
working ;-))

>Barton B. Anderson <tesla123-at-pacbell-dot-net> wrote:
>
>> How's that for accuracy of a program?
>
>-Paul replies:
>Looks pretty good from this one set of readings. However I think you
>should review all the available readings - it can be misleading if you
>focus too closely on one set. eg your measurements with this version
>of acmi give 4% to 7% error on k. I have no grounds for preferring one
>set above another, so they all must be considered. Also, you'll notice
>the primary self inductance is a long way out - several percent.

I too must caution against getting too happy with one great set of results.
 E-Tesla had its great moments on just one coil configuration, but would
fall on all others.  I finally came up with a set of eight well known coils
that I tested the program with before saying that it was good.  However,
getting such things to work at all is the hard part :-)  I eventually found
that two configurations seem to always be a bit out of disagrement with the
program.  When I went back and remeasured the coils, I found error in the
orginal data.  That's when I started to trust the program more then me ;-)



Paul Wrote:

>Are you at 60Hz or 1kHz for your measurements? Is one end of the
>secondary grounded when you make the readings?  I'll calculate the
>total low frequency capacitance between your primary and secondary
>and get back to you with the figure.

The current is from my 60Hz AC line.  My LCR meter works at 1kHz.  I can't
remember if the secondary was connected to ground in the old measurements
but I doubt it.  Should it be??

>We'll I'm anxious to avoid the elementary error of reporting more
>digits than are significant. I've added another digit to k - there
>are now 4 after the decimal point, so that k is representable to 0.1%.
>Any more would not be justifiable until reliable agreement to at least
>3 digits is demonstrated.

Thanks!  The last digit really helps when the numbers come out all the same
to detect trends.  Of course, I realize that may be noise but I just feel
better knowing it is there ;-)


John Couture wrote:

>I am pleased to see that coilers on the Tesla List are taking more interest
>in TC testing. However, it should be noted that comparing test measurements
>with computer results can be misleading unless the "error problem" is cleary
>understood.

Yes!  That is why I am trying to search for sources of possible error and
determine the effects they may have.  Without a computer model as sort of a
stick in the ground, no one knows how close or far real coupling
measurements are.  None of use have exactly the same coils so we can't
compare data.  The possible coupling difference between same and opposite
wound primary/secondary configurations is something we would never figure
out without the computer's cold hard numbers (right or wrong) giving use a
common comparison point.  As we try and identify and correct the error, the
program becomes more accurate and our confidence grows.  As Bart mentioned,
soon one trusts the computer's numbers over one's own measurements.  When
you plot your best real data and it scatters in a cloud evenly around the
computers smooth curves, its time to trust the computer ;-))

Cheers,

	Terry