Akadem Drive AD-1 Manual

What this is

AD-1 produces sound by solving five real circuits numerically, one sample at a time. It does not approximate a distortion curve or look up a wavetable. Turning a knob actually changes a component value inside the circuit, and the solution changes with it.

Because of that, the knobs behave differently on each circuit. The same BIAS control moves cathode resistance on the valve circuits and DC flux offset on the transformer, and whether turning it up adds or removes distortion is reversed between circuits. This is not a lack of polish; it is how those circuits behave. Measured figures are in the BIAS section below.

Install

Double-click install.command in the download. It shows what it will install and asks before doing anything. No administrator password is required.

Everything goes into your own user area:

~/Library/Audio/Plug-Ins/Components/   Audio Unit
~/Library/Audio/Plug-Ins/CLAP/         CLAP

The panel

PRESET

The window above the row of knobs. ‹ and › step through the 18 factory presets. Move any knob and the display changes to USER — not because a preset number is remembered, but because the name is looked up from the knob values every time. It stays correct when you reopen a session, and when host automation moves the controls.

CIRCUIT

Chooses which circuit is actually solved. On a change, the new circuit is rebuilt and warmed up until its operating point settles before it is swapped in, so switching makes no noise. That work happens on a thread separate from audio processing.

DRIVE

The voltage driven into the input stage. Each circuit takes a different amount — several volts for the valves, 0.45 V for germanium — so the knob is an exponential curve with that ceiling as 100%. The actual gain in dB is shown under the knob.

BIAS

It moves a different component on each circuit, and in a different direction. The actual value and unit appear under the knob.

Second harmonic is a proxy for asymmetry (measured at 1 kHz, DRIVE 35%). TRIODE and OP-AMP get more asymmetric as you turn BIAS up; GERMANIUM and IRON go the other way, and PENTODE barely moves at all — on PENTODE what BIAS affects is sag, not asymmetry. We could make them agree, but that would mean bending what the circuits actually do, so we have not.

TONE

The output RC low-pass: 2.2 nF against a series resistance, giving 402 Hz to 20 kHz. The actual cutoff in Hz is shown under the knob.

MIX

The balance of dry and processed. The dry path is delayed to match the processed path's latency and the polarity is matched as well, so nothing cancels at intermediate settings.

OUTPUT

A −24 to +24 dB trim, applied after MIX.

SOLVER

The internal rate at which the circuit is solved. The number is the oversampling factor; below it you get the actual internal rate (session rate × factor) and the integration method.

Aliasing figures are worst-case with 7 kHz in at DRIVE 100%. ×8 is not simply better than ×4. Measured, the only circuit that improves from ×4 to ×8 is OP-AMP (−43.9 to −58.6 dBc); TRIODE, PENTODE and IRON get slightly worse, because the extra stage adds more interpolation and decimation leakage. If in doubt, leave it at the default ×4.

LOCK / OVER

On LOCK, green is normal; it goes out if any sample failed to converge, and when that happens what you get is noise rather than distortion. OVER lights when the output clips.

ENGAGE / BYPASS

This is the host's own bypass. The host's bypass button and the panel toggle are the same control, so they cannot fight each other. Latency does not change while bypassed, so phase against other tracks stays put.

Latency

Fixed at 5 samples, regardless of sample rate or SOLVER setting. It is reported to the host, which compensates automatically.

The reported figure does not move with SOLVER because CLAP has no way to announce a latency change mid-processing, and every change would force the host to rebuild its delay compensation. Any shortfall is padded internally.

Bouncing

When the host reports that it is rendering offline, AD-1 automatically:

• raises SOLVER so that it is never below ×4 (if you chose ×8, it stays there)
• removes the per-sample iteration limit

That limit exists to meet a real-time deadline, and during a bounce there is no deadline. The oversampling factor only affects the amount of aliasing — neither output level nor the amount of distortion moves — so all a bounce changes is that an unwanted by-product is suppressed further. There is nothing for you to do.

Factory presets

There are 18. All of them are level-matched — trimmed by measurement so that a 1 kHz, −6 dBFS input comes out at the same level as bypass, with a worst-case error of 0.05 dB. Switching between them does not jump in volume, so you can compare character alone.

The numbers are knob positions, in percent.

Supported hosts

Pro Tools is not supported either. There is no AAX build.

Specifications

If you run into trouble

See the support page for common symptoms and what to do about them. If that does not solve it, write to support@tsukubaonkyo.com.