Good to be home
The musicality of Hugo is a topic definitely worth discussing. I have no analogue experience, I'm a child of CD. Hugo seems to share a smoothness with analogue but not at the expense of almost hyper detail. Was that your intention Rob?
Hmm. The over reaching intention is to try to get Hi-Fi sound real - shut your eyes, and you are there, center stage row 10. When I listen to live un-amplified music, I am always struck by how far audio needs to come. The power of live instruments, with the huge palette of timbre qualities, the exact placement of instruments in space, the fact that you can easily hear individual instruments - these are the virtues that I cherish. Now live music does generally sound smooth - but then listen close to a live trumpet, or a crash of cymbals, and smooth is not a term one would use!
On musicality, this is of course the primary goal, we listen to music to enjoy it. My view is that if it's easy for the brain to understand what is going on in an audio scene, then you will improve musicality. But musicality is not something one can evaluate easily, particularly if you have a lot of listening tests to get through. Now, my goal is not to artificially sweeten the sound, but to make it completely transparent, with no character of its own. Problem is, nobody knows what a perfectly transparent product is. So how does one do listening tests to refine transparency? How do you know you are not adding something artificial and just enjoying distortion? The key to making something transparent is testing for variation. By that I mean, if a device you are listening to exhibits more variation, then it is better in a transparency point of view. So if instruments have more variation in timbre qualities, then its better. If there is more variation in sound stage, its better. If you can hear more instruments at once, its better, if you can resolve changes in timing and rhythm, its better.
There is also another important element - if I hear a difference, I need to fully understand why, from an aberration perspective, it is making a difference. Then if you understand why something makes a difference, you can make further improvements by engineering a reduction in those aberrations. The tap length on the interpolation filter is a case in point; maths prove that if you use an infinite tap length filter you will perfectly reconstruct the bandwidth limited signal - it will be as if you had not sampled the signal. I knew that the conventional assumption of a few hundred taps for the reconstruction filter would be fine was wrong; but the only way of proving that is too listen carefully. I will keep on increasing tap length until I can't hear an improvement, then I will add some for good measure. I know that Hugo's 26,368 taps is not at the limit yet. That's what makes my job so satisfying and exciting - I know much more improvements are coming!
Rob