Synth Filters and VCF Filter Families
Exploring differences in Synthesizer Filter Types and Implementations
Subtractive Synthesis and the Role of the Filter
The filter circuitry used in subtractive synthesizers is one of the defining factors for the character and final sound of a synth. Subtractive synthesis means you start with tone generation - ie: Oscillators or Noise playing at full volume, and you subtract from that sound to get your final output. The Filter section represents the most significant subtraction from that sound. The filter is responsible for shaping / cutting out frequencies to mold the tone. In addition, the amp circuit is used to control the overall volume. Both of these circuits are controlled by Envelopes that shape their change over time (Attack, Decay, Sustain, Release - ADSR)
The Low Pass filter mode is the most often used in synthesizers, though some synths offer a High Pass mode, or combination like Band Pass or Notch. For now, we'll focus on the Low Pass mode - Many synths offer switchable 2-pole and 4-pole low pass filtering per voice, giving you 12db per octave fall off, or 24db per octave fall off respectively. Some, like Oberheim OBX / Seq OB6 just have a 2-pole filter, resulting in more upper harmonics / treble being present in the sound. Others like Prophet 5 only have 4-pole filter modes, resulting in a darker, more bass heavy sound.
More Info on Some Legendary VCF Filter Chip Models
CEM 3320 - Oberheim, Sequential, Roger Linn, PPG, Fairlight, Behringer, Elka, etc
SSM 2040 - Sequential, E-mu, Groove Synthesis, Buchla, etc
SSM 2044 - Korg, PPG, E-mu, Fairlight, Waldorf, Sequential, Behringer, etc
The Main Filter Families, and Specific Implementations
Within the realm of low-pass filters, you've got several different filter families / topologies:
Transistor Ladder, (Moog style)
Diode Ladder, (Roland acid style, EMS)
Operational Transconductance Amplifier OTA, (Sequential, Oberheim, Roland and others)
Sallen Key, (Korg 35 style)
Steiner Parker,
etc...
Each of those families has several different implementations that have specific characteristics when it comes to resonance and exact curves of the filter. Some people have very strong preferences for some specific implementations - the debate about the Prophet 5 OTA v1/2 filter, versus OTA v3 filter is a great example of this.
You can usually get in the ballpark of a given sound even with different families of filter, by just selecting the correct slope and adjusting parameters to get close. If you've got a filter in the same family (ie: Transistor Ladder, OTA), but with slightly different implementation, you can usually get extremely-close to a given sound, even though the specific implementation of the filter might be different.
Comparing Implementations of Prophet OTA Fitlers
Taking a look at the modern Sequential Prophet 6 vs Prophet Rev2, you're comparing two synths with OTA filters. You've got the same basic family of filter, but the specific implementation is different. The P6 has a 4-Pole SSM-2040 (Solid State Micro) type of implementation, whereas the Rev2 has a switchable 2-pole / 4-pole CEM-3397 (Curtis) type of implementation. This is similar to the difference between the Prophet 5 v1/2 and Prophet 5 v3, where the first two had SSM OTA filters and the ladder had a Curtis OTA filter. For the majority of people, the difference between these two implementations of an OTA filter are minimal to negligent, however, to experienced sound designers and synthesists, there are some slight, but noticable differences.
These differences are most evident in two scenarios: In Bass/Lead/Mono style patches where a single filter being clearly articulated at a time, or in Polyphonic patches where filter key-tracking and filter velocity is off (since you get each voice articulating the filter at the same points temporally, emphasizing/multiplying the unique character of the filter) When playing polyphonic chords with either key tracking on, or velocity sensitivity on, the specific character of a given filter implementation is somewhat blurred and almost impossible to differentiate. It's somewhat rare to design patches that have 0s for both key tracking and velocity sensitivity... though definitely sometimes used.
Voice Modeling for Filters and Envelopes
One additional that should be considered when comparing classic synths to more modern implementations is Voice Variance and Voice Modeling. There is a seperate article here that discussed Voice Modeling in detail: http://www.VoiceComponentModeling.com
In addition to per-voice variance in oscillator tuning, classic synths tend to have more per-voice variance in the fitler and envelope circuits. Classic synths tend to have electronic circuits with looser electrical tolerances, leading to small variances per voice in exact cutoff point of the filter and exact timing of envelope stages on a per voice basis. Also, circuitry degrades slightly over times, so a synth built with electronics components 30-40 years ago (the 1980s) will have far more voice variance than a modern designed synth using the same electrical topology for oscillators (VCOs), filters (VCFs) and envelope circuits.
Modeling Filter and Envelope Variance with the Prophet Rev2 or Deepmind
This is another characteristic you can model with the Rev2 or Deepmind. You can either design in modulation that directly alters the precise cutoff point on a per-voice basis, or, even better is to alter the Attack/Decay by a very small amount per voice... this essentially kills two birds with one stone, as you're designing in a slight voice variance to the attack/decay stages of the envelope, and when applied to the filter cutoff, it is also varying the cutoff per voice, to emulate classic synths with imperfectly matched envelope and filter circuitry per voice. Really, you only need to alter the Attack per voice and the downstream effect is that the decay and release will have variance as well as the filter cutoff point.
More info on Voice Modeling with the Prophet Rev2:
http://www.VoiceComponentModeling.com
More info on Voice Modeling with the Behringer Deepmind:
https://www.presetpatch.com/articles/Voice-Component-Modeling-with-Behringer-Deepmind