My initial idea was to cover a wide frequency range with one transformer, targeting low frequency under 10 Hz. Because of that i choose to use two stacked T60006L2045V102 toroid cores. 
Initial version of the transformer had 25 turns made of twisted solid core bell wire. With this design the phase error at 2 Hz was less than 2 deg, but in the same time the large capacity and inductance lead to more than 10 deg above 400 khz . 
In the "final" version i have 16 turns, that giving me range of 4.5 Hz to 600 khz with phase error of 5 deg, measured with Venable 3120 FRA. More interestings are the result from the HP 8753C which i belive is a bit more accurate at high frequency giving 5 deg at 1.1 Mhz. 
In both cases the transformer show comparable results as phase shift, and looks much better as magnitude compared to other commercial products. 

Using thin wire may improve the frequency response due to reduced capacity  with the same number of turns, i am planing some more experimenting in a near future.