The Neve 33609 is a solid-state limiter/compressor that uses a diode bridge to accomplish gain reduction. The bridge is used in a shunt configuration, much like the FET in an 1176. Audio is applied across what would normally be the two AC inputs to the bridge. The incremental resistance of the bridge is altered by passing a DC current from the top of the bridge to the bottom, keeping all four diodes forward-biased most of the time. Conceptually, the gain control is similar to a variable-mu tube compression scheme, where bias points are altered according to the desired amount of gain reduction.

With variable-mu circuits, single-ended topologies can achieve compression, but have an undesirable artifact called thump, which is due to control signals coupling into the audio. The thump problem can be solved for variable-mu compressors by using a push-pull topology, so that a differential output can be taken from the circuit, while the control signal leakage is entirely common mode (for perfectly matched tubes). In the case of the 33609, the thump problem is solved in a similar way, by using four diodes, with the audio being a differential signal, and the control signal being common mode. To keep the circuits differential, multiple transformers are required, which increases the cost of production for units such as the Fairchild 670, Universal Audio 175, and Neve 33609; however, fidelity is much improved with this method.

The diode bridge gives the 33609 a unique sonic character. In terms of linearity, the diode is probably on the warmer end of the spectrum. At the clean end are optical gain reduction elements and VCAs. FETs can also be made to have low distortion by feeding a portion of the audio signal into the control port, as on the 1176LN. Variable-mu tubes and diodes are less clean, and can be anywhere from barely noticeable to very warm, depending upon circuit topology and biasing. Because diodes can be used for clipping, many people associate diodes with large amounts of distortion. However, as used in the 33609, the diodes are forward biased virtually all of the time, and the distortion due to the diodes is minimal, leading to sonic warmth rather than excessive distortion. The result is a very distinctive but pleasant sound that is achieved by no other circuit.

The other unique feature of the 33609 is the Auto Release function. In most cases, auto release is accomplished by making the release circuit a second-order filter, leading to a variable release time depending upon input signal statistics. The second-order release is used by compressors such as the 1176 and the Fairchild 670 (in modes five and six). For the 33609, auto release is also accomplished by a second-order filter, but a nonlinearity placed within the filter changes the behavior of the circuit. Because of this nonlinearity, the 33609 Auto Release behaves differently from other compressors.

When creating the DSP-based emulation of the 33609 for Universal Audio, we were given tremendous support from Neve, which provided access to engineering documents, as well as hardware units that we could disassemble for testing. Because of our ability to test subcircuits independently, we were able to produce highly detailed and accurate models of the 33609 circuits, resulting in a very good match for both the audio path and the sidechain of this compressor. The 33609 has more significant, distributed nonlinearities than any other unit we have modeled, and we collected much more data than ever before to model those nonlinearities. The result is a DSP model that, while costly, should accurately reflect the behavior of the hardware in any situation.

As with all compressors, the attack and release behaviors, as well as the static compression curves, critically affect the sonics of the device. The diode bridge of the 33609 will add nonlinear processing through an entirely different mechanism than the gain reduction, and this processing is also important to the sound of the unit. Not only will transients be affected by the behavior of the diode bridge, but the nonlinearities will create additional harmonic energy in the steady-state signals.

Modeling the diode-bridge behavior requires upsampling the audio path, which adds considerable expense to the algorithm. Because of this, we chose to implement two versions of the 33609. The full version includes complete models of the compression characteristics, as well as all relevant nonlinearities in the audio path. A second version, the 33609SE, will include the full dynamics-processing model, but will not include all of the signal-path nonlinearities, greatly reducing DSP cost. The SE version will sound very similar to the full version for many sources and, in all cases, will provide a useful and unique addition to the range of compressors available on the UAD.