ROSS VU

Every plugin VU promises 300 ms. This one was measured.

The movement is a real second-order system with average rectification, exactly like the electromechanical instrument: zeta 0.826 e omega zero 14 rad/s. It is not an RMS follower with smoothing on top.

The tests do not use a copy of the code. They extract the Movement e TruePeak structs from the very source that compiles the plugin, and measure. If anyone touches the constants, the number in the table changes.

$ ./rodar.sh
structs extracted from the plugin
48k CAL-18 meio    -> rest -0.012 VU | 99% in 302 ms | overshoot 0.86%
44.1k CAL-18 meio  -> repouso -0.000 VU | 99% em 299 ms | overshoot 1.00%
96k CAL-18 meio    -> rest -0.003 VU | 99% in 300 ms | overshoot 0.96%

The peak that blows the limiter lives between the samples.

The LED and the reading are dBTP, by 4x polyphase oversampling, 12 coefficients per phase, in the spirit of ITU-R BS.1770-4. The classic case is a full-scale sine wave at fs/4 with 45 degrees of phase: every sample lands on 0.707, so a sample-peak meter swears there are 3 dB of headroom left. There is none.

The lamp follows that same number, and its threshold is yours: -12, -9, -6, -3, -1 or -0.1 dBTP. A lamp that only lights up after the clip warns you of nothing, it just confirms. This one blinks for 430 ms, short enough to read as a warning and not as a stuck lamp.

The scale was not drawn. It was calculated.

In an iron VU the needle deflects linearly with the current, and it is the printed scale that comes out crooked. That is why the -20 to -10 stretch is tight and the -3 to 0 one is wide. Here every mark is positioned by p = 10^(VU/20) / 10^(3/20), the real deflection of the movement. It is not a curve picked by eye to look vintage.

Because it is average-rectified and not RMS, it reads high crest factor material below the RMS, exactly like an iron VU. That is the correct behaviour, not a defect.

Five controls. No decoration.

• Cal-20 / -18 / -16 / -14 / -12 dBFS. Defines what 0 VU is. One click cycles.
• Response600 ms to 100 ms. In the middle it gives the normative 300 ms, and the number shows under the knob.
• PeakLamp threshold: -12 / -9 / -6 / -3 / -1 / -0.1 dBTP. One click cycles. Blinks for 430 ms.
• Input-24 to +24 dB. It is a real gain stage: it touches the audio, not just the meter.
• Output-24 to +24 dB on the output. Double click on any knob resets it.

Five reading modes: L+R, LEFT, RIGHT, SIDE e L | R, the last one with a second needle for the right channel. Six factory presets:

• Mix Bus -18
• Streaming -14
• Broadcast -20
• Tracking Fast
• Program Slow
• Stereo L | R

One codebase, four formats, zero exotic dependencies.

• PluginVST3, CLAP and LV2, all from the same source, via DPF and NanoVG.
• StandaloneStandalone application: JACK/PipeWire on Linux, WASAPI on Windows. Runs with no DAW at all.
• Linuxx86_64. Tested on REAPER 7 over Fedora 44 with PipeWire.
• Windowsx64, VST3, CLAP and standalone. Built with mingw-w64, with no external DLL dependency. See the caveat below.
• InterfaceResizable keeping the aspect ratio, drawn in vector. No stretched bitmaps.
• FontsEmbedded in the binary. The plugin does not depend on any font installed on the system.
• LicenseGPL-3.0-or-later. Source included.