This is TRUE given a theoretical environment, but we're dealing with implementation details here. The fact of the matter at hand (with ML speakers) is that you cannot
do without the passive (internal) crossover. Now, the question becomes that do we want to add more processing into the signal path just for the sake of having an external crossover to match with the "theoretical specification" or do we want to keep things streamlined and go with the passive approach. Keep in mind that you're only degrading
your signal by adding components into the path. This is the balancing act we're facing with the ML speaker models today. If your speakers are ML Monoliths which come with external crossover, things will be much easier.
No, the external crossover should be set to the SAME
crossover frequency of the internal one.
You need to match gains between the Bryston and the tube monoblocs. The Bryston overwhelming the monoblocs means that you'll need to attenuate the signal between the preamp and the Bryston, while leaving the signals to the tube monoblocs untouched.
Just to give you an idea, you can take a look at the following link to the "nude attenuators" from EVS:
I went ahead to look at the specs for the ARC VS series and Bryston and here's what I see:
ARC VS55 has 25db gain, VS110 has 28db gain, while the Bryston 14BSST has 29db gain. If you were to go with the VS55, you'll need to 4db of attenuation at the Bryston's input. On the other hand, you'll need 1db of attenuation going the VS110 route. In the practical sense, most people won't notice the 1db difference so you can save some $$ and go with a more standard 3db attenuation value for in the case of the VS55 and forego any attenuation in the case of VS110. The rationale being that a little bit more bass might be preffered by someone in his/her set up. Follow this Parts Express link to see the in-line attenuation unit. A friend of mine is using the 6db attenuation units in his setup with great results.