The Tech Talk column takes you “under the hood” of our plugins.
Michael Massberg is responsible for analog modeling at Brainworx and its partner companies. He writes the software algorithms that recreate legendary analog units in the digital domain. His work involves lots of circuit analysis, math, measuring, soldering and – finally – coding. In this column, Michael will talk about what's so special about the hardware devices we model and what steps are necessary to faithfully lift them into to the digital plugin world.
Under the Hood of the SPL De-Esser
By Michael Massberg
Traditionally De-Essers are constructed using simple sidechain compression: the incoming signal is filtered so that only the higher-frequency part of the spectrum, containing the unwanted sibilance, remains. The filtered signal is then used as the sidechain input to a compressor which will attenuate the unfiltered signal depending on the level of sibilance at its sidechain input. However, since the compressor will attenuate the signal over a wide range of frequencies and not just reduce the sibilant noises themselves, these simple designs can cause substantial artifacts such as lisp or nasality.
Fig. 1: The classic Model 9629 SPL De-Esser hardware
The SPL De-Esser – invented by Wolfgang Neumann – takes a different approach: instead of using simple compression, it works more like a dynamic filter.
Fig. 2: Simplified circuit block diagram
Fig. 2 shows a very basic block diagram for the De-Esser's circuit. As a feed-back design, the input to the sidechain is taken from the already-processed output.
The signal is filtered using a combination of band-pass and high-pass types where the frequencies of operation (6.4 kHz and 7.6 kHz) are selected using the “Male/Female” switch on the front plate. Only the band-pass part is rectified and used to generate a DC control voltage, where the input level to the rectifier is determined by the “S-Reduction” pot.
Attack and release time constants are being applied by a complicated envelope generation circuit. The release envelope is actually a two-stage process. While the initial trajectory of the release curve is quite flat, it significantly increases speed once the control voltage has dropped below a certain level.
The resulting envelope then controls the gain of an OTA (Operational Transconductance Amplifier). The input to the OTA is again the filtered signal and the OTA’s output draws current from the De-Esser’s output stage. Put simply, the filtered signal is reversed in phase and dynamically mixed in with the original signal, causing the unwanted sibilance to phase-cancel out.
Fig. 3: Magnitude response for various control voltages
Fig. 3 shows the magnitude response of the SPL De-Esser for various constant control voltages. As can be seen, the De-Esser only attenuates the problematic parts of the spectrum and leaves the lower frequencies completely untouched.
Another specialty of the SPL De-Esser is its Auto Threshold circuit. Using rectification and filtering, the Auto Threshold stage generates an level estimate of the input signal. The level estimate is then used to control the gain of the main rectification circuit. Essentially Auto Threshold increases the processing depth once the input signal weakens in level and decreases it once the input gets stronger. As a result, the amount of sibilance reduction stays fairly constant even with fluctuating levels of input (which, for example, could be caused by a singer moving closer to or away from the microphone).
Classic & Dual-Band De-Essers
SPL offers their De-Esser in two different form factors: there’s the classic Model 9629 SPL De-Esser and the 1070 Dual-Band De-Esser RackPack module. While both units generally share most of their circuitry, there are some differences that made us want to model both of these units.
The biggest difference is of course the second band of the Dual-Band De-Esser. It’s basically a second De-Esser circuit operating at a fixed frequency of 11.2 kHz and wired in series with the base band. The second band allows to further reduce higher-frequency noises.
Another difference is that the Auto Threshold function is switchable in the classic design, while it is always activated in the dual-band unit.
The Dual-Band De-Esser also adds a unique post filter borrowed from SPL’s Transient Designer. The filter adds some high-end shine in order to make up for a possible loss of perceived high-frequency content due to the dynamics processing.