Funny for Krell to recommend capacitive coupling! Did they say whether it is input or output coupling you need? I assume input since you mention your tube preamp (and yes, this is a good idea) but I have read some speaker designers say that lack of output amplifier coupling is the reason for glare - and not input coupling. I am not sure which is right but I would place my bet on output coupling, especially with B&W impedance curves. Output coupling can also be done with a capacitor but it is best to use a transformer (like tube amps and McIntosh amps). The transformer also allows for impedance decoupling which is a much more elegant solution that adding a simple cap (although it costs a lot more).
In many direct-coupled amps, the DC often gets shunted in the voltage gain stage anyway. The input coupling cap just makes sure that the input differential current mirrors are have proper midpoint bias. Adding a cap at the input can really distort the waveforms if the circuit isn't designed for it at the outset due to nonlinear frequency gain. This is also why I think output coupling might be what they were referring to. Hard to say which is best without a schematic of the amp however.
Also, you mention having 500 caps in the signal path. This is incorrect. The caps you see and think of are actually in parallel to the path and/or in the bias loop and not signal loop. The cap Krell is talking about is actually in SERIES with the signal path, whether input or output. Huge difference.
Arthur
In many direct-coupled amps, the DC often gets shunted in the voltage gain stage anyway. The input coupling cap just makes sure that the input differential current mirrors are have proper midpoint bias. Adding a cap at the input can really distort the waveforms if the circuit isn't designed for it at the outset due to nonlinear frequency gain. This is also why I think output coupling might be what they were referring to. Hard to say which is best without a schematic of the amp however.
Also, you mention having 500 caps in the signal path. This is incorrect. The caps you see and think of are actually in parallel to the path and/or in the bias loop and not signal loop. The cap Krell is talking about is actually in SERIES with the signal path, whether input or output. Huge difference.
Arthur