Analog computing will come back into fashion


The filter is the element of subtractive synthesis that gives it its name and function. It is used to filter out certain frequencies from an incoming audio signal, but also to add frequency components via resonance.

Parameters of a filter [edit | Edit source]

Filter cutoff frequency [edit | Edit source]

The Cutoff frequency, in German Filter cut-off frequency, is the frequency from which the signal is filtered. It can be regulated and modulated.

If there is a filter envelope, it controls the cutoff frequency; the intensity of this influence can be regulated using the envelope amount.

Resonance [edit | Edit source]

The resonance is a parameter that is also known outside of electronic music as the Q value or the filter quality. It boosts the signal at the cutoff frequency. Some analog filters reduce the rest of the signal, some virtual analog synthesizers can simulate this behavior; if this is not done, the resonance makes the sound louder.

The resonance can also be regulated and, with some synthesizer models, modulated.

Early digital filters were not always capable of resonance. The KORGM1, for example, was criticized as a synthesizer for the fact that its filters could not generate resonance, also because it appeared at a time when classic analog synthesizers were again in fashion.

Self-oscillation [edit | Edit source]

Depending on the filter design, high resonance values ​​can result in the filter generating a sine wave oscillating with the cutoff frequency even without an incoming signal. If the rest of the signal is reduced at the same time, only the sinusoidal signal remains in the case of extreme resonance values. In analog synthesizers, the oscillators themselves cannot generate a sine wave; the self-oscillating filter is used instead.

Edge steepness [edit | Edit source]

A filter does not suddenly cut off frequency ranges, but rather they become weaker the further they move away from the cutoff frequency. This attenuation has a logarithmic curve and is measured in decibels per octave (dB / octave). An audio signal that runs through a low-pass filter at 12 dB / octave, for example, is one octave above the cutoff frequency, i.e. at twice the cutoff frequency, 12 dB weaker than at the input, absolutely only ⅟₁₆ as strong, two octaves above the cutoff frequency, i.e. four times the cutoff frequency, is 24 dB weaker, i.e. only ⅟₂₅₆ as strong, etc.

In the case of analog filters, the edge steepness is predetermined by the design. Virtual-analog synthesizers, on the other hand, have switchable slopes in order to be able to emulate different analog synthesizers sonically. The slope can only really be regulated in very rare cases.

The steeper a filter is, the more likely it is to self-oscillate at high frequencies.

Keyboard Tracking (Keytracking, Keyfollow) [edit | Edit source]

If this parameter is available, it controls how strongly the cutoff frequency depends on the note played. At zero, the filter remains the same over the entire playing range, at 100% the cutoff frequency is scaled in the same way as that of the oscillators. This is what makes the tonal play of a self-oscillating filter possible in the first place.

In analog synthesizers, there is usually no stepless controller for this. If at all, there is a switch with three or four positions that allows various preset tracking to be selected. With four positions these are 0, ⅓, ⅔, ⅟₁. In this way it is relatively easy to ensure that exactly 100% can actually be set (within the framework of analog inaccuracies).

With some analog synthesizers, keyboard tracking is fixed at 100%, for example with the classic models from Yamaha. With these, the cutoff frequency can be set on a harmonic scale instead of an absolute one.

Digital synthesizers allow a reliable selection of even finely scaled values, so many have an almost infinitely variable filter keytracking. Sometimes this can even be regulated to over 100%.

Pole number [edit | Edit source]

The number of filter poles determines the slope of the slope in analog synthesizers: In general, a filter pole has 6 dB / octave, which corresponds to a level reduction to ¼ with each octave. In the case of high and low pass filters, a two-pole filter has 12 dB / octave, and a four-pole filter has 24 dB / octave. A bandpass filter filters out two frequency ranges; a two-pole bandpass filter only has 6 dB / octave because one pole is used for the frequencies above and below the cutoff frequency.

Filter types [edit | Edit source]

Voltage-Controlled Filter (VCF) [edit | Edit source]

Voltage-controlled filter or VCF is the name for analog filters that are completely regulated by control voltages. They work relatively imprecisely, but this contributes slightly to their sound character. Since the analog control is stepless, manual filter runs do not generate any undesired artifacts.

Digitally Controlled Filter (DCF) [edit | Edit source]

The Digitally Controlled Filter or DCF still works analog, but with digital control, which makes it more precise and compact. Except that it can be tuned more cleanly than a VCF and the raster of the digital control can produce step effects when filtering, if the control is not smoothed, there are hardly any sound differences due to the principle; instead, the DCF can be better integrated into a digitally controlled and storable synthesizer, as was common in the 80s.

The DCF has gone out of fashion with new analog and hybrid synthesizers. Even if these have DCOs, VCFs are mostly used again.

Digital filter [edit | Edit source]

A Digital filter is part of a digital signal path. It simply calculates frequency components from the audio signal and does this in a basically sound-neutral manner. Digital filters could also abruptly remove frequency bands; Gradually falling filter edges are basically historical.

While digital oscillators were already to be found in synthesizers in the 1970s, digital filters did not appear until the second half of the 1980s together with samplers and ultimately rompers. Before that, synthesizers with digital tone generators and samplers either had analog filters or no filters at all. Some early digital filters, such as those at KORG, are not yet able to generate resonance.

The increasing interest in analog synthesizers in the 1990s led to the development of virtual-analog synthesizers. These made digital filters necessary, which behaved like analog filters as much as possible and sounded like analog filters. At first these were still relatively generic, but from the beginning attention was paid to a variable filter characteristic including different slope steepnesses as well as the greatest possible elimination of digital artifacts. Later even synthesizers appeared that imitated specific classic analog filters, for example the four-pole cascade filter from Moog or the Oberheim two-pole filter.

Time-Variable Filter (TVF) [edit | Edit source]

Time variable filter or EDF is Roland's name for a digital filter with an assigned multistage envelope.

Low pass filter [edit | Edit source]

A Low pass filter (Low-pass filter, LP) lets the frequency components from the cutoff frequency downward pass unchanged and filters out those above. In terms of sound there is a change to dull.

High pass filter [edit | Edit source]

A High pass filter (High-pass filter, HP) lets the frequency components from the cutoff frequency upward pass unchanged and filters out those below. In terms of sound, the bass foundation decreases, the sound becomes thinner.

Bandpass filter [edit | Edit source]

A Band pass filter (Band-pass filter, BP) allows only the frequency components at the cutoff frequency to pass unchanged and filters out both those above and those below. In practice it consists of a series-connected low-pass and high-pass filter. Usually both have the same cutoff frequency, but with some models the spread can also be regulated. Analog bandpass filters thus have poles for the low-pass and high-pass components and therefore require twice as many poles as low-pass or high-pass filters for the same slope.

Notch filter [edit | Edit source]

A Notch filter (Band stop filter, Notch filter, Band rejection filter, BR) filters out the frequency components at the cutoff frequency and lets those above and below pass unchanged. Similar to the band-pass filter, the notch filter consists of a low-pass and a high-pass; However, these are connected in parallel here and far enough apart in the cutoff frequency to be able to downregulate a frequency range far enough. With regard to the number of poles and the edge steepness, the same applies as for the bandpass filter.

Filterbank [edit | Edit source]

A Filter bank consists of several parallel-connected filters with different cut-off frequencies, which, similar to additive synthesis, allow the targeted manipulation of certain frequency ranges. Depending on the intended use, the lowest filter can be a low-pass filter and the highest a high-pass filter, but in general all filters are band-pass filters. In synthesizers, filter banks are usually only used as part of formant filters, equalizers or vocoders.

Formant filter [edit | Edit source]

The Formant filter already existed in the Trautonium. It is a special type of filter bank that consists of filters capable of resonance and is primarily used to generate resonances, i.e. to amplify the input signal at several frequencies. Formant filters are often used as Vowel filter used to synthetically generate vowels of human speech.

Multimode Filter, State Variable Filter [edit | Edit source]

These are filters with variable characteristics.

As Multimode filter they are called in digital synthesizers. The most basic form can be switched between low-pass, band-pass and high-pass. Some rompers also have a notch filter. However, these filters do not have a defined character and a uniform slope of 12 dB / octave.

In virtual-analog synthesizers, they perform the same task as in analog synthesizers. So higher demands are made on them. A combination often used in synthesizers from the 90s are low-pass, band-pass and high-pass with 12 dB / octave as in Oberheim and low-pass with 12 dB / octave as in Moog. For the sake of simplicity, however, some synthesizers have separate options for low-pass, band-pass and high-pass and for the slope with two to four positions.

On the one hand, there are more special forms with the AccessVirus, which has two filters per voice; There is no band-pass setting, instead the band-pass is generated from two filters with a jointly adjustable cutoff frequency and resonance and, since one filter can be switched to 24 dB / octave, it can be asymmetrical. On the other hand, there are synthesizers such as the AlesisIon, Alesis Micron and AkaiMiniak or the KORG KingKORG, which, among other things, imitate specific filter classics and their characteristics, which therefore sound different despite the same slope. These synthesizers come with more than a dozen types of filters.

State variable filter is the term used for analog synthesizers. Here, too, it can turn out differently: In modular synthesizers, for example, there are separate output sockets for low-pass, band-pass and high-pass that can be used simultaneously with the same input signal. The implementation in the Oberheim SEM is unusual: Here, the three types of filter can be cross-faded continuously - from low-pass to band-pass and on to high-pass.

Filters and polyphony [edit | Edit source]

With polyphonic synthesizers, each filter in the signal path must be present once per voice. Ideally, each voice also has its own filter envelope. The settings are the same for each voice, apart from a few exceptions.

The reduction to paraphony can simplify this structure, which can reduce the structural effort, especially in the case of analog synthesizers, in which the filters must be physically present as components in the synthesizer. However, this also cuts the tonal possibilities. If the synthesizer only has one filter for all voices, there are hardly any more possibilities to modulate it depending on the key played. Keyboard tracking becomes pointless if the filter can only follow one of several keys played. The filter envelope is also only started by the first key pressed and stopped again by the last key released.

In purely digital synthesizers, filters are little more than mathematical models. The computational effort for generating a filter is therefore included from the outset in the computing power required for a voice. With rompers, each wave generator is usually assigned its own filter anyway, and this unit forms a voice with all associated envelopes.

Common filter modulations [edit | Edit source]

Filters usually only have two parameters, the change of which has a noticeable effect on the sound: cutoff and resonance. However, a lot can be done by modulating the cutoff frequency.

Envelope [edit | Edit source]

When a subtractive synthesizer has multiple envelopes assigned to fixed destinations, one of them is usually the filter envelope.

In the case of simpler synthesizers, there is often only one envelope that controls the amplifier in a fixed or switchable manner. This can then also be used to modulate the filter.

Freely usable envelopes, for example in modular synthesizers, can also be used for filter modulation. In this case in particular, modulation of the resonance is often possible.

Wah-wah [edit | Edit source]

Wah-wah occurs when the cutoff frequency is repeatedly increased and then reduced again in a resonating low-pass filter, so that the frequency peak resulting from the resonance moves up and down in the frequency spectrum. The modulation can be done by hand (or foot), but also by LFO. In synthesizers themselves, this effect is not often used, and even less often is it called it.

Filter FM [edit | Edit source]

Filter FM is a kind of increase in wah-wah or a combination of this with an essential element of FM synthesis. The modulating LFO oscillates here in the audio range, i.e. over 20 Hz; If the synthesizer supports it, a normal oscillator can also be used as a modulator. Due to the high speed, the opening and closing of the filter is no longer perceived as such, but as a completely different, diffuse sound component. Resonance is not even necessary here, but it creates new sound changes.

As long as no atonal noises are sought, Filter-FM like FM-Synthesis only works if the keyboard tracking for the frequency of the modulator is set to exactly 100%. In order to obtain a largely consistent sound, the keyboard tracking should also be set to 100% for the filter.