Q: Does the Neve 33609 plug-in include a model for saturation in the circuit, or is the signal distorted only due to the action of the compressor?
A: There has been much debate among users on this topic. The Neve 33609 plug-in does model distortion in the audio path separately from distortion caused by gain modulation. For the 33609SE, the audio path distortion is left out, leaving the rest of the behavior intact.
The 33609 hardware uses a diode bridge to accomplish compression. The diodes in the bridge are forward-biased, and the incremental resistance of the diodes determines the amount of gain reduction. The resistance of the diodes is in turn dependent on the amount of bias current going through the diodes. Bias current is determined by the output of the 33609 sidechain.
At large amplitudes, the audio signal itself can cause appreciable changes in the amount of current flowing through the diodes in the bridge. When the audio signal moves the diodes away from their bias points, the incremental gain of the circuit changes, causing distortion. Because the circuit’s gain depends both on the bias currents (determined by the sidechain) and the audio signal itself, the amount of distortion depends on both the signal level and the instantaneous amount of gain reduction being applied.
When the unit is not in gain reduction, the incremental resistance of the diodes is relatively large. In this case, bias changes due to the audio signal do not cause appreciable changes in incremental gain, so the amount of distortion is very low, even for large signals. Similarly, when there is a large amount of gain reduction being applied, the diodes have a relatively low resistance, and effects of the audio signal again are not dramatic. However, when gain reduction is between 3 and 7 dB, diode resistances are comparable to surrounding circuit elements, and large audio signals can cause THD levels approaching 10%. It’s important to keep in mind that while distortion is often considered a dirty word in audio, and 10% levels of distortion could be considered alarming, the unique non-linear characteristics that give the 33609 its beloved sound are technically distortion. In other words, not all distortion is created equally.
The following table shows the amount of distortion present for the 33609 plug-in, with an input at -0.5 dBFS, and the headroom control set to 22, for different levels of gain reduction. Note that the only way to achieve low levels of gain reduction for a signal of this amplitude is to modify the plug-in so that the threshold of compression can be set very high. For the hardware or the plug-in, it would be impossible to observe distortion levels this high in real-world use, because signals of this amplitude would immediately cause the unit to go into heavier gain reduction. However, this experiment gives insight into how transients can be beautifully softened by the action of the 33609. This mechanism limits peak levels of sudden transients until the attack circuit can cause gain to be reduced.
|Gain Reduction dB||THD %|
This experiment gives insight into how transients can be beautifully softened by the action of the 33609.
It is important to make a distinction between distortion caused by gain modulation and distortion caused by other mechanisms. For any compressor, there will be signal distortion caused by the fact that the gain changes as a function of time. For most types of level detection, including most peak detection and RMS detection schemes, level estimates for sinusoidal inputs will have ripple components at harmonics of twice the input frequency. These ripple components can cause gain modulation, which produces harmonic distortion. Usually, ripple amplitude will increase with shorter attack and release times for a given input frequency.
In order to hear the effects of distortion in the audio path, it is critical to suppress the gain modulation usually caused by compression. There is a simple way to do this with the 33609 plug-in. If the plug-in is put into heavy compression with a large-amplitude sine wave, and then compression is disabled using the Compress In switch, the compressor’s gain will return to unity, following the compressor’s release curve. When the Compress In switch is placed in the “out” position, the audio signal is no longer routed to the sidechain. After this point, the compressor can no longer attack, and there will be no ripple present in the sidechain output. If the compressor is set to a very slow release time, varying amounts of harmonic distortion can be heard as the compressor releases monotonically through its gain reduction curve.
This test was performed with the 33609 and 33609SE plug-ins. As can be heard in the output files, harmonic energy is present during the release at much higher levels for the 33609 than the 33609SE plug-in. If the distortion is neglected, both compressors share the same release dynamics.
The chart above shows a portion of the outputs for the 33609 and 33609SE plug-ins. As can be seen, the 33609SE has an output which is nearly sinusoidal, while the 33609 shows measurable distortion.
This test can be duplicated as follows:
Input at 436 Hz, -0.5 dBFS
“Limit in” set to “out”
Compression threshold set to +2
Compression recovery set to 1500
Compressor gain set to 0
Ratio set to 6:1
Output gain set to 0
Headroom set to 22
Toggling the Compress In switch will produce the test result. This test can also be performed using the 33609 hardware.
— Dr. Dave Berners