This is a set of EQ-neutral models captured from a Vox AC30S1 using an SM57 microphone (placed about 3 inches off-center, 2 inches from the grill cloth, and slightly off-axis).
The S1 variant of the AC30 has a single Top-Boost-like channel with Volume, Reverb, Treble, Bass, and Gain controls. These models were captured with the factory speaker installed (a VX12). The factory tubes were replaced with a matched quartet of JJ EL84s and 2 JJ 12AX7s.
Volume was kept between 1 and 2. Reverb was set to zero for all captures. To keep the models sounding EQ-neutral, Bass and Treble were gradually lowered together in-sync as the Gain was increased. On the AC30S1, Treble and Bass essentially work like Presence and Resonance controls, only in reverse. So just like you'd want to raise Presence/Resonance to compensate for losses when the gain is increased on a Marshall, you want to lower Treble/Bass here to tame fizz and harmonic buildup.
No reactive load devices were used, and no manual processing was done to the sweep signal input file (T3K-sweep-v3.wav) nor the output audio files used to train these models. They're listed below in-order of increasing gain level from 1 to dimed. Hope you guys enjoy.
01. Pedal Clean - Gain 1, Treble 6, Bass 6 (tight and sparkling, max headroom)
02. Historic Chime - Gain 2, Treble 5.5, Bass 5.5 (dynamic, warm and articulate)
03. Edge of Breakup - Gain 3, Treble 5, Bass 5 (light grit with harmonic shimmer)
04. Hair Clip - Gain 4, Treble 4.5, Bass 4.5 (smooth crunch, touch responsive)
05. Denney Crunch - Gain 5, Treble 4, Bass 4 (mid-focused, open crunch)
06. Dartford Drive - Gain 6, Treble 4, Bass 4 (tight drive, controlled overtones)
07. Bubble Bath - Gain 7, Treble 3.5, Bass 3.5 (slight compression, still articulate)
08. Jennings' Organs - Gain 8, Treble 3, Bass 3 (saturated crunch, aggressive lead)
09. Korg Collar - Gain 9, Treble 2.5, Bass 2.5 (edge-of-dirt, harmonically complex)
10. Cytherea Soak - Gain 10, Treble 2, Bass 2 (full saturation, thick mids)
Capture Notes:
These models were trained between 60 and 140 epochs, producing ESR values ranging from 0.002-0.005. I use the Python NAM trainer gui app on my local machine, so you won't see those values listed here on T3K. In general, I find that the models feel more alive if I use less epochs and more input gain on my audio interface with a good mic placement. I used a Focusrite 2i2 (4th gen) and a passive Radial ProRMP box.
