P-Q series filter - better understanding

Hello everyone.

This is my first post and it has to be a request for help. Here is my problem, if you have a bit of patience to read:
I'm learning about power conditioning and applications for a while now, mainly active power filters (I'm not a student, that was a while ago). Electronics being what it is, practice is needed, and what better way to understand a circuit than using an electronics simulator. I chose one freely available -- LTspice -- and it's the basis of my visualization.
Now, having learned about the "p-q theory", I wanted to better understand its ways, so I proceeded into simulating. I will not go into details and just say that whatever topology I tried, shunt, series or hybrid, without considering the power stage, the p-q theory never worked for me and(!) the moment I tried replacing the "middle" part with d-q rotating frame calculations, everything just worked!
My English not being perfect, at all, I'll try to put it like this: I'm trying to simulate (for example) a series active filter based on the p-q theory. Considering only the abc-αβ0 matrix, then the calculation of power, then selection of powers and then the transposed matrix, the output value, say va* ('a' voltage, output after inverse tranformation), added to va ('a' voltage at the input) should yield a nice sine-wave, or, at least, the desired effect. It doesn't. Removing the power calculations and selection and inserting αβ-dq transformation, then selection of powers, then dq-αβ (inverse Clarke is still there), adding va* to va results in a pure sine.
I changed signs, multiplication blocks, ...suffice is to say I tried *many* combinations, I'm at a point now where Mrs. Frustration threatens to take over. I know I am doing something wrong since for everyone else the results are good.
So, if there's anyone willing to help me, I've attached the schematic, it's in LTspice, but (unfortunately) you'll need some custom models, made to help me reduce the schematic. Every block is made with ideal sources & co for the easiness of simulation (after all, I want to see a result first, then try the "real thing").
A small explanation of the schematic: the result -- the resulting sine -- should be v(x) (to the right of the page) and the 3 resistors in the middle-top were supposed to be in parallel with 3 G-sources; i took them out for a quicker simulation since with or without them, v(x) should be the same. The rest, from left to right, lower side: abc-αβ0 for V and I, power calculations, selections of powersand inverse matrix. The 3-phase generators are current source based, that's why I can parallel them. In the zip file -- the attachement -- there are also the formulas after which I made this.
About the custom models, they are in the LTspice Yahoo group (I'm afraid you'll need to register for that) in " Files > Examples > Educational > Math_blocks- Filter-Power " (a moderator moved them there), and the needed libraries are Pwr, Filt and Math. I know, it's too much but, as I said, this is only for those that are willing to help. Still, without them there would have been mostly behavioral sources filled with big lines of expressions which would have made debugging a pain. (all the blocks work quite well, I use them everyday; if you have doubts, theres an example file "abc-ab-abc.asc".)

So, without further ado, you have my problem(s) and my request(s). Kind souls, please stop by...

[edit]
For some reason, the file didn't upload, I only noticed now. Fixed.
[/edit]

Re: P-Q series filter

...anyone?

Are the formulas good? Is the schematic, at a first glance, good? Did anyone try something based on the p-q theory?

Re: P-Q series filter

...I managed to figure it out.
It seems that the p-q theory is very unappropiate when distorted/unbalanced voltages/currents are involved so, in addition, I have to add a positive/negative sequence voltage detector so as to generate two αβ signals that are pure sine, unity amplitude, as reference, to avoid using the distorted, transformed signals. They are generated by either the PLL, or the voltage detection block. PLL also has the ability to generate ωt signal.
BUT(!), if that's the case, the schematic, now, with the two signals given by the positive voltage detection block, would turn into a regular d-q case. After all, the ωt signal used in d-q transformation is used to generate the "sin" and "cos" signals (the Park matrix) which are then multiplied with the αβ signals that are to be "worked" on (the sensed voltages). The same happens here, only there is no (?) ωt signal but there are the two unity amplitude αβ signals which, further on, are multiplied with the sensed voltages.
Now, from a practical point of view, that would be fine by me as I get the desired effect, but how does this change the theory? After all, even the Park matrix now resembles the calculation of powers in the p-q theory. I know they are very similar, but is this still p-q or it's just the d-q "in disguise"? Does this prove that p-q isn't to be used for distorted/unbalanced signals and one should consider switching from static α-β frame to rotating d-q? Is this the wolf instead of grandma' ?
Confusion...