Wolfgang
Guest
|
 Posted:
Mon Dec 20, 2004 12:56 pm Post subject:
Re: implementation of PLL |
|
|
Dear jkm,
Thanks for the title but I don't have it and at the moment have no ambition
to get it.
I don't know if your overlaying control will be done in dq-cooridnates.
If so you'll see that you have to do some calculations to get them from
uvw-values.
Your pll is - compared to that - easy.
You know dq-coordinates ?
Search on google about it.
If you don't want to use them your voltages are sinussoidal.
You can detect the phase of a voltage if you make a fourier calculation
(look at Bronstein for Fourier-series calculation an calculate coefficiens
only for one frequency)
You get two coefficients:
ak=factor * integral over (U(t)*sin(w*t))
bk=factor * integral over (U(t)*cos(w*t))
Dependening on the angle you can calculate the phase relation of U(t) to
sin(w*t) (you give t you give w it's your sin(w*t) only U(t) is
measured)
(e.g. phi = arctan(ak/bk) for ak>0 and bk>0)
You can finde the factor in Bronstein but it is not of real interest for you
....
As far as i know people who apply such simple control calculate only ak
if ak>0 U(t) is leading sin(w*t) and for the next period they introduce a
step
if ak<0 U(t) is not leading sin(w*t) and for the next period they leave out
a step to "be faster".
So they don't do a real PLL which changes frequency they have a fixed
frequency (50/60Hz)
and if they would normally have e.g. 180 steps per period they make 179 or
181 steps to
compensate for the small drift.
To calculate ak you need an integral which is some sort of summation of the
discrete values you have.
MAC is a special DSP command which you MUST know if you want to do DSP
assembler.
Its called "Multiply and ACummulate". You multiply your sin(w*t) from ram
table with your U(t) value
and ACummulate. The result is ak (disregarding the factor).
But you should first find out what you want.
If you need dq coorinates you must do a transformation an you get "stable"
values for the line voltage.
the d component should be U and you can directly use q component for phase
compensation.
q>0 ... q<0 ... (Normally to avoid jittering q is filtered over a period and
the correction is only
done once per period).
But may this all does not apply to you as you have something complete
different in mind than I have...
Good luck, Wolfgang |
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in