Real vs Fake Sympathetic Resonance

dore_m

As some of you know, I've been trying to model sympathetic resonance as a new feature in my sample libraries. It's been an interesting process and I've learned alot by just trying. I've finally got the keyboard and pedal behavior working somewhat reliably, so I can now begin to optimize the harmonics to mimic real recordings of the piano that I sampled. I'm basically doing what Pianoteq does, but doing it only for the SR.

Before I take a deeper dive into optimizing the SR for the Steinway B, can you tell the difference between my rough model and the real thing?

One recording is me playing the real Steinway (holding an open C and striking an F, and then G)

The other is a rendering of the sample instrument (the attacks are samples played back) and the sustained sound are triggered sine waves (modeled).

Which do you think is which?

sr-2.wav
sr-1.wav

QuasiUnaFantasia

sr-2.wav seems to have much longer lasting harmonics, very nicely behaved. Maybe a little too nicely behaved.

navindra

sr-1 has so much more mechanical noise and to me the strings vibrate forever. It also doesn't sound perfect as QuasiUnaFantasia said. I'm accustomed to this experience from an acoustic piano -- extraneous noise, strings vibrate forever, doesn't always sound perfect.

sr-2 sounds idealized. It felt like the resonance died down a bit too abruptly at the end.

vagfilm

On sr-2 there is a larger spectrum resonance (almost a shimmering echo...) that starts when the main resonance is already decaying. This happens both times, after F or G. I don't hear the same effect on sr-1. But since I have limited experience with grands SR, I have no clue if the presence of that late "shimmer" is or isn't a real feature. My gut feeling is that it is not, and that sr-1 is the actual recording.

dore_m

I can't fool you guys. I'm currently tuning the sine wave loudnesses to get them to match better. The "shimmering echo" is too much because I'm exciting all of the resonant frequencies by a relatively uniform amount, and I can see on the real piano's spectrogram that some frequencies are louder than others. Hopefully it will feel more real as I figure this out.

dore_m

Real:

Fake:

You can see clearly my sine waves on the lower picture - represented as multiple bright horizontal lines. Brighter means louder at that frequency. I need to tamp these down.

CyberGene

I guess you're holding the middle C, right? I did the same test with the Yamaha N1X, binaural CFX with the default VRM at 5, then at the max value of 10.

n1x-vrm5.mp3
n1x-vrm10.mp3

P.S. Oops, I messed it up, I'm playing F4 and G4 whereas you're playing F2 and G2. I'll redo later.

CyberGene

I noticed that if I hold silently an octave C2 - C3, then playing F2 and G2 produces very nice resonance beating, probably because the matching partials are not overlapping but have slight difference due to the inharmonicity (octave stretching). So I recorded it (first default VRM 5, then increased to 10):
n1x-vrm-5-10.mp3

dore_m

@CyberGene Here's how our spectrograms compare. N1x first. First 2 partials look similar.

I'm using an LFO to fake the beating - maybe I should turn it up as my first partials don't show any beating at all.