G.Wolny wrote:
Hi Everyone!

I'm not sure if this is the correct group to be asking this kind of
question. If no send me please to a more appropriate.

Could someone explain me how to calculate a very low phase
displacements (c.a. 2-10 minutes ) between two currents?
In detail; I want to calculate the phase displacement between two
currents in a transformer. For now I just perform the calculations in
Matlab on some fictional currents:

i1=I1*sin(w*t+phi1) and i2=I2*sin(w*t+phi2)

whe
phi1 = 0 and phi2 = 5 (multiplied by ‘pi/(180*60)' to get the value in
radians)
‘I' is the current, ‘w' is omega and ‘t' is the vector of time.

My method is this:

A=i1.*i2;
MyA=mean(A);
B=2*MyA/(max(i1)*max(i2));
phi=acos(B) ;

The phase is in radians so I divide it by pi/(180*60) and have the
results in minutes.

The problem is the accuracy of this calculation. To get an phase shift
of few minutes I have to increase the time step in the time vector ‘t'
. Usually I use for the time step a value of 1e-7s (0.0000001s) what
gives 10 million points per one second cycle duration. And that's not
enough because the error between the result and the fixed phase was
between 5-10% and it should be no more than 0,3%.
If I Decrease the duration of the cycle the accuracy drops so only
through increasing the amount of data I can increase the accuracy. I
think that the same problem occurs when I perform the calculations
using FFT.
Increasing the time step takes a very long time to calculate and
generates an out of memory error on my P4 1.7MHz with 128 MB of RAM.
Using FFT it was even much slower.

If you know the exact frequency, then you really only need
one bin of the fft or one spectral line. So, do the dft at
that frequency.

--
local optimization seldom leads to global optimization

my e-mail address is: rb my last name AT ieee DOT org