Directional coupler transformer wiring

I am building a directional coupler for HF and I need it to be as flat as possible on 1-30MHz because it will drive an ALC circuit of an oscillator.
It terms of isolation and flatness, is it better to use the single binocular core version http://michaelgellis.tripod.com/direct.html or the two toroids version http://www.hoaglun.com/blog/2010/1/1...rsion-two.html ?

I think option two is better, having flat coupling vs frequency, and also have better directivity.

How is that achieved? Using shielding between the two transformers?
Otherwise what is the problem with the binucular core version?

Also, what if I need less sampling output (eg 30db), should I increase the 10T?

Yes, EM separation between couplers is the main reason of flatness and better directivity.

In this type of coupler changing the turns ration would not change too much the coupling. For example doubling the 10T to 20T on both transformers, the coupling will change only 3dB.
If you need less sampling output, just use a pad attenuator which will do the job, and keep the flatness. Actually a pad attenuator will improve the flatness even more, because the coupler will see a better resistive 50 ohms impedance at the port.

Where shall I connect the attenuator, at the input of the coupler, between the transmitter and the coupler?
I guess this attenuator pad should be 50R I/O impedance right?

Also you mention adding more turns will change the coupling by 3db, wou mean more turns mean less sampling right?