Dynamic Processing PDF

Dynamic Processing If a sound engineer uses the term 'dynamics', he usually does not mean the tone strength in musical performance practice, but the so-called technical microdynamics of a signal. Although compression is the best known, it is only one aspect of dynamics. In addition, one should alway...

Dynamic Processing If a sound engineer uses the term 'dynamics', he usually does not mean the tone strength in musical performance practice, but the so-called technical microdynamics of a signal. Although compression is the best known, it is only one aspect of dynamics. In addition, one should always be aware of which tools are available for which tasks. In principle, four control directions can be defined: Output Name Ratio In : Out Control for signals... Comment F A Compression >1 : 1 above threshold B Limiting ∞:1 above threshold Special case of compression A C Downward 1:>1 below threshold B Expansion E D Gate 1:∞ below threshold Special case of Range downward expansion E Upward >1 : 1 below threshold rarely C Compression D Input F Upward 1 : >1 above threshold rarely Expansion The distinction is based on simple criteria: If the signal is processed above (A, B, F) or below (C, D, E) the threshold If the dynamics will be increased (C, F) or reduced (A, B, E) The gate plays a special role, it's just a kind of switch If one change the level of a signal with the fader or the volume automation, one do not change the sound character of the signals. However, if you use Dynamic processors, you have to realize that their 'access time' as well as their 'reset time‘ can be chosen in a way that the control activity has a direct influence on the envelope of the signal, and thus on its sound. The very rapid change of levels thus represents a form of sound design. Dynamic Behaviour (Attack Time) Dynamic processors, including digital ones, are not infinitely fast. Apart from a few special cases, they need a certain amount of time (and they are also often supposed to do this) to bring the input value to the target value. This time is called 'attack time' or 'response time'. It starts immediately when the threshold is exceeded and theoretically never ends. The attack time is defined as the time between exceeding the threshold and reaching 63% of the setpoint. From a practical point of view, one can orient oneself without hesitation to the values stated by the manufacturer and associate these values with the sound result. Notice: Attack is not a 'wait time'. It is a settling time. Typical values are in the range of 0.1 to 50 ms. Release Time If the signal falls below the threshold again, the Processing does not stop immediately. There is a slow return process. This return process is usually much slower than the attack time. We speak of the release time. Practical values are in the range of 10 ms to 3 s. The curve of the release time is a crucial criterion for the sound character of various known devices. There are no limits to creativity when setting attack and release. However, you should keep in mind that extreme settings can have a negative effect on thesound quality. This may be the case, for example, if the attack is selected so that it interferes with the period duration of an oscillation and distorts it: Effects of the attack time on a high and a low-frequency signal Static Behaviour After the attack time, the system is considered as steady. The level changes of the control amplifier can be displayed as static behavior on a characteristic diagram: Output Question: What dynamic processing is represented by the blue and 1:1 pink characteristic curves? 2:1 Threshold 4:1 8:1 Input 1:2 1:4 1:8 Soft Knee A type of compression where the onset of compression is gradual. In normal or hard knee compression when the signal reaches the threshold the unit immediately begins to compress at whatever ratio is set. In some situations the compression becomes very easy to hear (which is often not desirable) as the signal amplitude moves above and below the threshold. This is usually made worse when using high compression ratios Control Type & Sidechain The principle circuit of a control amplifier is very similar for the different applications. An input signal runs via a rectifier into a detector, which compares the measured signal with a setpoint and transmits a corresponding correction value (control voltage) to the control element. There are different ways to gain the signal for the control. The most important are: Feed forward control Feed back control Sidechain If the control signal comes from a sidechain path, this signal can be processed. Some systems offer fixed filters in the sidechain path that can be used, for example, to filter out unwanted signal components from the control signal. Notice: Filters in the sidechain path have no effect on the useful signal! The changes only affect the controlled variable, which may cause the control amplifier to react differently if desired (e.g. stronger or weaker to bass components). Sidechain Input (Key In) Sidechain with EQ Look Ahead In addition, a delay element can be added to the audio path to allow the detector to react faster to signal changes than they actually occur. In this way, a level reduction can be initiated which has already reached its setpoint when the signal peak occurs. This is mainly used in limiters in order to be able to intercept every signal peak reliably. This type of device is called a brick wall limiter. Notice: Be aware that a Brickwall Limiter should be able to intercept intersample peaks as well. Not every Brickwall Limiter has this ability. Feed forward control with Look Ahead delay Gate and Hysteresis Especially expanders and gates are often equipped with a hysteresis circuit. In some cases the characteristic value can be influenced, but often it is fixed and'invisible‘ to the user in the system. In principle, it represents a second threshold whose switching state depends on the first threshold. In most cases, it cannot be set as an absolute value, but as a deviation from the 'main threshold'. Let's look at a gate and see what happens. First, we set the threshold to -6 dBFS. Signals below this value are removed, values above it remain unprocessed. Without hysteresis, this limit is at the same position when opening and closing. However, this hardly ever makes sense, because it cuts off the end, for example. If the hysteresis control is now set to -6, the signal must first fall below -12 dBFS for it to be processed. With rising levels, the threshold value remains unchanged. Gate without Hysteresis Gate with Hysteresis Peak vs. RMS The detector of a dynamic processor can react to different signal components. As already described, it is sometimes possible to use an external, independent source for the control signal. But even if the control signal is obtained from the input signal, the detector can either react quickly to all peaks or the control voltage can follow the energy content (RMS value) of the signal more slowly. Crest Factor The crest factor describes the ratio between the RMS value and the peak value of a signal. A low crest factor approaches the control signal to its RMS value, a high crest factor approaches its peak value. Assuming the signal is 6 dB above the threshold and the detector reacts to the peak levels, the control element lowers the signal at the output to 3 dB (at ratio 2:1). However, if the detector analyses the integrated RMS value, it is only 3 dB above the threshold for the detector and is reduced by 1.5 dB. The output signal is therefore at 4.5 dB. The crest factor therefore has an effect on the real ratio. The RMS level control is also more constant than the peak control. This is due to the integration, which 'offsets' individual peaks and thus balances the overall control behavior. Since the hearing also reacts less to peak levels and more to the rms value, the control changes to a more 'aurally-accurate' control. The Figure shows a music signal, first as uncompressed original [blue], then compressed with Peak-Crest [yellow] and RMS-Crest [green]. All other parameters remained unchanged. It becomes clear that peak crest peak feedback control is much stronger than RMS crest. The crest factor is only adjustable with very few compressors. Parallel Compression (New Yorck Compression) In parallel compression, processed signal components are mixed into the unprocessed original signal. This means that the source signal must first be split (or the original track is copied in the DAW). One of the two signals is compressed, the other remains unprocessed. The mixing ratio is determined by the two channel faders. It is recommended to consider the original signal as the main signal and add the compressed signal carefully. The most important condition for this processing must be that the latency of the compressor plug-in is compensated with sample accuracy. If the DAW does not offer this option automatically, compensation must be performed manually. If there is an offset between the unprocessed and processed signal, comb filter effects and clearly audible sound disturbances can occur. It is recommended to sum the two tracks in a separate audio group so that their this signal can be regulated in the mix. This also makes it possible to process the entire signal with equalizers, for example. The parallel compression is an approximation to the upwards compression. Here, too, quiet signal components are compressed and added to the original dynamics from below. Notice: The parallel processing described here is also interesting for other effects! Multiband Compression A broadband device has the limitation that its control processes react equally to all frequencies contained in the signal. This can be mitigated, for example, by filtering the sidechain. However, the control itself always has an effect on the overall signal. If one does not want this, one must divide the input signal into different spectral ranges and conduct these into independent control stages. In the simplest case, the necessary filtering is nothing else than a crossover as it can be found in almost every loudspeaker. Alternatively, the division can also be carried out via so-called FIR filters (finite impulse response filter), which buys the advantage of phase neutrality with the disadvantage of the long latency. The outputs of the processing stages are again mixed together. The many parameters of a multiband device result from the fact that the control elements must be multiplied by the number of bands and the functions of the crossover are added. The splitting of a signal into several spectral ranges always represents a danger for the phase position at the intersections of the bands. It is important that this also applies to devices that work with phase-linear filters (FIR). There is always an overlap area between the bands. If the signal only passes through the filter stage and is then summed, the result corresponds to the input signal, at least in the amplitude frequency response. However, as soon as different processings are applied in the bands, level and sometimes phase differences inevitably result, which influence the overall signal during the following summation. Bildnachweis: Abbildungen Friedemann Kootz, 2011, 2012 Außer: Seite 2, Handbuch der Tonstudiotechnik, Band 1 Dickreiter et.al., Saur 2008 https://www.sweetwater.com/insync/soft-knee-compression/ Studio Magazin 04/11 https://www.soundonsound.com/techniques/multi-band-compression-tips Abbildungen Friedemann Kootz (Studio Magazin), 2011-2012