Preamplifiers are widely used in the music industry to amplify audio signals and improve the signal-to-noise ratio. They also incorporate circuits for non-linear distortion, creating unique tones in iconic amplifiers. As music production goes digital, preamplifiers are being analyzed, modeled, and digitized for emulation. Several studies have confirmed the effectiveness of AI recurrent neural networks in accurately emulating amplifier output. However, traditional network architectures can only fit specific time-series characteristics, requiring re-training and storing different models when preamplifier settings change. This research introduces a conditional input structure, synchronizing knob parameters with input signals into a long short-term memory network, effectively predicting the preamplifier`s output. Experimental design involved implementing five rotary knob parameters and conducting experiments at five different angles for each knob setting. Results showed the proposed model had an average RMSE of less than 0.01, reducing the need to store multiple parameter sets and enhancing AI modeling efficiency for multiple preamplifier characteristics.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=22294
(694KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
