Analog Obsession: Compression Technology and Topology
By Dave Berners
Compression has become commonplace in virtually every step of recording: tracking, mixing and mastering all usually involve use of compression at some point. The diversity of applications for compression and the wealth of sonic variety that can be created by use of compression have caused an explosion of design. Compressors have been built using variable-mu tubes, optical elements, FETs, diode bridges, VCAs, and probably scores of other devices. Feedback, feedforward, and mixed topologies have been experimented with, and every different design seems to have a unique flavor. Compression, perhaps more than any other kind of signal processing, can lend character to a recording.
In some cases, the technology used to build a compressor will dictate much of its sonic character. In other cases, details of the compressor's attack and release will be explicitly designed by auxiliary circuitry. In all cases, the technology used will create certain constraints for design that may limit the applications of a given compressor. Let's examine some of the more popular technologies in terms of their generic properties.
Compression, perhaps more than any other kind of signal processing, can lend character to a recording.
Variable-μ compressors use vacuum tubes designed to have a gain that can be controlled by a sidechain signal. These compressors feature smooth compression curves and can be made to have significant warmth or saturation. Attack, release and ratio are designed explicitly by external components. Disadvantages are that poorly matched tubes or poor maintenance can result in significant distortion of audio. Two famous examples of variable-μ compressors are the Universal Audio 175 and the Fairchild 670.
Optical compressors were first introduced to prevent the gain reduction signal from leaking into the audio path. Because of this, they are often billed as zero-distortion compressors. A big advantage of optical compression is that, depending upon the physics of the optical device used, a program-dependent release can be had "for free," with no additional components needed in the circuit. Disadvantages are that the attack and release times are coupled, meaning that unless the release time is artificially increased (which would negate the program dependence), the attack and release times cannot be set separately. Attack times are generally not exceptionally fast. Also, for stereo applications, it can be difficult to get precise tracking of compression between channels.
Famous example: Teletronix LA-2A.
FET compressors are somewhat similar to variable-mu compressors, but with properties of solid-state saturation rather than tube warmth. The nonlinearities associated with the FET can be largely eliminated by a clever trick of circuitry. FET compression knees tend to be somewhat more compact than vari-mu tubes. Attack, release and ratios are configurable as with vari-mu designs. Famous example: UREI 1176.
Diode bridge compressors can be fully configurable in terms of compression parameters, in that the compression curves and the attack and release can be designed independent of the compression element. Diode bridges also provide good immunity between the control signal and the audio path. Compressors built on this technology are tonally distinctive, because the nonlinearities associated with the diodes can reach up to ten percent harmonic distortion under certain conditions. These designs are flexible and can add a pleasant character. Famous examples: Neve 2264, Neve 33609.
VCA compressors are the most modern designs, using large-scale circuit design to produce a variable-gain amplifier with high linearity and immunity to crosstalk from control signals. Character for these designs is obtained exclusively by design of the attack and release behavior and the static compression characteristics. VCAs are probably the most flexible compression technology, and they can be extremely transparent. Care must be taken in design, as some of the "accidental" good properties of other technologies will not appear for free here. Famous examples: dbx 160S, SSL G-Master Buss Compressor.
Using any of the technologies above, you can design a compressor that uses either the input or the output signal as the control for the amount of gain reduction. If the input is used, the compressor is referred to as feedforward. If the output signal is used, the compressor is feedback. Technically, it is not possible to get ratios of infinity to one or higher using a feedback topologies, although there are feedback limiters, such as the Neve 33609, that have extremely high compression ratios. Feedback topologies are popular because the signal being sensed by the sidechain has a smaller dynamic range when feedback is used. For noise gates and expanders, feedforward topologies are popular for the same reason. For compressors with no program dependence and constant ratios of compression, it is possible to make feedforward and feedback designs that have identical attack and release behavior.
With certain caveats, any of the above technologies or topologies can be used for most any application. Nuances of tone, however, can depend heavily on the technology used. Next month, we will take an in-depth look at one of the more distinctive ways of compressing audio: the diode bridge attenuator.
Questions or comments on this article?