Experiment 2
Description
This experiment considered training the models against the digital phaser data for different LFO rates: fast (\(T_0=0.5\text{s}\)), medium (\(T_0=2\text{s}\)) and slow (\(T_0=8\text{s}\)).
The training window lengths, \(W\), were set relative to the target LFO period:
\[W = T_0 2^b / 100\]The audio examples below demonstrate the shortest, median and longest window lengths used for each rate.
Digital Phaser
Target parameters: feedforward gain: \(g_1 = 1.0\), feedback gain: \(g_2=0.7\), delay-line length \(\phi=1\).
$$T_0$$ | Reference | $$b=0, \quad W/T_0 = 1\%$$ | $$b=2.5, \quad W/T_0 \approx 5\%$$ | $$b=5, \quad W/T_0 = 32\%$$ |
---|---|---|---|---|
0.5s | ||||
2s | ||||
8s |