Voltage controlled filter

Revision as of 11:40, 27 August 2019 by Diydsp (talk | contribs) (Captured many links from a synth-diy mailing list thread about VCF architectures)
Jump to navigation Jump to search

An audio filter is a frequency dependent amplifier circuit, working in the audio frequency range, 0 Hz to beyond 20 kHz. Audio filters can amplify (boost), pass or attenuate (cut) some frequency ranges.

Common filter types used in audio synthesis include:

  • Low-pass (high-cut) filter: removes higher frequencies
  • High-pass (low-cut) filter: removes lower frequencies
  • Band-pass filter: removes frequencies outside a given band
  • Notch filter: removes frequencies within a given band
  • Shelf filters: raise or lower frequencies above or below a cut off point.
  • Peak filters: Raise frequencies within a given band
  • Formant filters: raise multiple peaks, often in such a way that they mimic the human voice.

There are many ways to implement most filters, and each has its own specific audio characteristics or flavour. Some common features across most filter types include:

  • A cut-off frequency - the frequency at which the filter begins to remove frequencies. This is often voltage-controllable, and can changes over time (e.g. controlled by an envelope, or an LFO).
  • resonance - how much the filter boosts the frequency at the cut-off point. This may also be voltage-controllable.
  • frequencies beyond the cut-off often have their phase affected.


Filter architectures (sorry this is crudely laid out, 1st step was capture info, 2nd pass will organize)

This section began from an excellent thread on the synth-diy mailing list in August 2019: https://synth-diy.org/pipermail/synth-diy/2019-August/171529.html


Tillman: Consider a description of a filter as a sort of "taxonomy" with three layers: Top Layer: the filter spec, number of poles, response Second Layer: the topology that implements that filter function Bottom Layer: implementation details, including the control element So a Moog Ladder would be: Top Layer: 4 pole, low-pass, with resonance Second Layer: 4 single-pole low-pass sections in series, with feedback Bottom Layer: the ladder circuit And a State Variable filter would be: Top Layer: 2 pole, multi-mode Second Layer: 2 integrators and an inverter, in a loop Bottom Layer: the circuit, perhaps OTAs


Moog Ladder

The Moog Ladder filter is like that Zen Koan that all students of the synthesizer electronics temple meditate upon. "Oren Leavit"

https://www.allaboutcircuits.com/technical-articles/analyzing-the-moog-filter/

https://www.allaboutcircuits.com/technical-articles/small-signal-open-loop-transfer-function-moog-filter/


ladder filter variations: http://www.till.com/blog/archives/2005/03/ladder_filter_v.html


MS20,

Ian Fritz Threeler,

Mutant Vactrol Filter ...


3-pole, 4-pole, etc - without cascading 2-pole SVFs.

Leapfrog topology as implemented by Matthew Skala. The original ARP2600 filter is a clone of the Moog Ladder. The later ARP2600 filter has the same filter topology, but implemented inelegantly to get around the patent. https://files.northcoastsynthesis.com/msk-007.pdf (page 69 of the PDF) One of the important properties of this topology is that it in some sense minimizes component dependence - which is important for keeping the shape of the curve reasonably consistent when tuning it, given that I'm trying to keep five OTAs tracking each other. If you like implementation details, you might also like the way I'm using different linearizing-diode currents to set the fixed proportion between the different integrator time constants.


Steiner Parker,

The Steiner Parker is a rare exception. It's a classic Sallen-Key filter hacked up with biased diodes as controlled resistors.


Wasp filter.

The Wasp filter is a State Variable with 4069 inverters replacing the inverting opamp in the integrators. Certainly the overdrive characteristics of the 4069 inverter are different than an opamp, but it's in a local feedback loop, and in a global feedback loop, and I think the OTA overdrive will predominate anyway.


EMS diode ladder The EMS diode ladder is the same as the Roland diode ladder, and they're both Moog Ladder knockoffs, knocked off sufficiently to get around the patent.

Arp 2600 filters The original ARP2600 filter is a clone of the Moog Ladder.

The later ARP2600 filter has the same filter topology, but implemented inelegantly to get around the patent.


programmable op amp filters based on the lm4250?

https://modularsynthesis.com/kuzmin/polivoks/polivoks_vcf.htm Okay, that's a State Variable Filter with programmable op amps for the integrators. And a programmable op amp is basically an OTA, with an integration cap, and an output stage.


SVF

The State Variable Filter was introduced by the brilliant Dennis Colin in the ARP 2500 VCF. And his article is one of the classics: Dennis Colin The Electrical Design and Musical Applications of an Unconditionally Stable Combination Voltage Controlled Highpass, Bandpass, Lowpass, Band Reject Filter/Resonator Journal of the Audio Engineering Society, Dec 1971 http://www.guitarfool.com/ARP2500/DennisCollinPaper.pdf He didn't invent the SVF, but he was the first to voltage control it and apply it to music. The State Variable Filter has a long history, back to analog computers. It's also the same mechanism as the simple harmonic motion of a mass, spring, and friction. So it has a wonderful "universal" quality. The Oberheim filter, the Rhodes Chroma filter, and others, are mostly variations on this.


The Colin/ARP SVF puts the integrating cap in the integrator following the OTA. The Rossum/Oberheim SVF puts the integrating cap at the output of the OTA, shunted to ground.


Then there's the ARP 4075 VCF based around the LM3900 Norton amplifier. The first low distortion low noise high fidelity VCF. I just read the patent (US 4,011,466) on the 4075 and it is a fascinating read. That's a unique topology.


One thing the article fails to touch on is the feedback architecture. Feedback has a big impact on the sound, and the feedback design changed between Moog synth models. That is why players back then complained that the new models didn't sound like the Minimoog. My post in SDIY way back in 1997 here: http://search.retrosynth.com/synth-diy/search/lookit.cgi?-v9710.272