cannot assume a pi/4 shift and divide

No I am trying to avoid division for you. The pi/4 is due to the fact the to get a tan I need sin/cos. This is implicit division. But you have two sine waves. So to get a cos wave you have to shift by pi/4. You would also have to shift by the actual phase diff in the case of the sine waves concerned.

An alternative math process is to consider them as exponentials. This reduces division to subtraction of exponents.

So you need your box to do the equivalent of taking the tan or calculating exp() and subtracting. This could be achieved with an old analogue computer, but I don't know how your box works.

 

Steve, if your inductor Q is high (XL>>Rseries), then the phase is not unpredictable. The current should lag the voltage by 90 degrees, regardless of (presumably small) changes in inductance.

 

i think the problem lies not in the division procedure. a table or a common shift-and-add loop is ok. a uP can do table lookup just fine too.
the major problem is in getting the peak values in less than 10 cycles of the waves, right?
how about using the dsp unit to get it done?