The few Harbeths I've heard are pleasant sounding, inoffensive and uninvolving. I feel they are overated and very overpriced.
Alan Shaw states:
The first point is ’cost is no constraint’
we will add components and complexity to achieve, in combination with the shape of the cabinet and the characteristics of the drive units, the smoothest measurable frequency response.
If cost is no object/constraint then what are all those grainy sand cast resistors doing in there??? There is no place in high-end audio for these nasty things including power supplies. He is also parading 'the smoothest measurable frequency response as a badge of honour.
To those gasping in horror, let me explain. Each and every component in an XO is doing some damage to the signal and in the example above, Danny counted about 30 components! Yes the resultant frequency plot is impressively smooth, but at what cost?
I built a little speaker using a small fullrange Fostex driver. Sounded great but had an upper midrange peak. OK no problem, I threw together a notch filter consisting of a cap, coil and resistor all in parallel and wired into one of the speaker wires. This gave the impression of more bass and was a better balanced sound, BUT, it killed the life and dynamics. It was quickly removed.
Understand that this was not an XO as such, just a simple filter. Now try using an XO with about 30 components. Understand also that 30 cheap as chips components will 'measure' exactly the same as 30 better components but will not sound the same. It does however please many who need to see its response graph. More important is a clean CSD and proper phase tracking around XO. The expanded vertical axis in the plots shown in the video appear worse than they are because grads are at 5dB not 10dB, but just look at the waterfalls.
And now a final sobering thought. Place your great measuring speaker in the average room and the acoustics of said room just make a total farce out of the holy grail of 'flat response' Those with measuring kit will know what I mean.