types of filters (based on circuitry)

Anything modular synth related that is not format specific.

Moderators: Joe., lisa, luketeaford, Kent

User avatar
cornutt
Fig juggler
Posts: 1362
Joined: Sun Jan 25, 2009 10:06 pm
Location: Rocket City USA

Post by cornutt » Mon Jul 24, 2017 4:04 pm

The Yamaha GX-1 filter is a Sallen-Key. Some years ago Scott Rider undertook heroic measures to de-pot a GX-1 filter module and dissect it. Before this, everyone assumed they were the same as the CS80's filters, but Rider found that they were of a Sallen-Key topology. He reverse engineered the circuit and made some new modules, and around this work, Paul Schreiber developed the MOTM-485 filter.

Here's a link to Scott's page describing what he did:
http://www.oldcrows.net/~oldcrow/synth/ ... _proj.html
Sequence 15 -- sequence15.blogspot.com

User avatar
addendum
Common Wiggler
Posts: 127
Joined: Mon Mar 17, 2014 1:47 pm

Post by addendum » Mon Jul 24, 2017 4:57 pm

strangegravity wrote: 908 Posts and you don't know what a state variable filter is? These are filters that have any combination of low, band, and high pass outputs.
1118 posts and you don't realize you're quoting a five year old post. For that, you oughta check out the OP's post archive and calculate how many posts they really had at that time. #jk

Btw I think your second comment isn't fully correct. The Wiard Omni Filter I mentioned is different although it does have LP, BP and HP. The fourth output which in SVFs normally is Notch is Allpass in the Omni. Also there are filters with "synthesized" BP, HP etc outputs with variable pole combinations like that Doepfer module that emulates the Xpander filter modes. "State variable" refers to a specific principle.

User avatar
pre55ure
Super Deluxe Wiggler
Posts: 1416
Joined: Tue May 01, 2012 7:15 pm
Location: SoCal
Contact:

Re: types of filters (based on circuitry)

Post by pre55ure » Mon Jul 24, 2017 11:44 pm

strangegravity wrote:
hiawog wrote:hey all. i'm not an electrical engineer, but i want to better understand the different categories of filters as defined by their circuit topologies. obviously there's an infinite number of ways to categorize modules, but i'm hoping to get a better grasp of general approaches to designing a filter and the sonic consequences of those design decisions.

i've heard of a few different types already:

- transistor ladder: thank you bob moog. usually 24db/oct, smooth sound.
- diode: harsher or grittier sounding, i think?
- vactrol-based: vactrol ring and woody tone
- attack-decay: i don't know the right name for this, but i'm talking about the 6db/octave filtering that you can get with a serge-style envelope
- state-variable: don't really understand what this means
- vca-based: i've heard of people designing filters around vca chips, but i don't know how this works.

i'm sure there are plenty of other kinds, and i know that it's never accurate to make sweeping generalizations about a given type of module. i'm just trying to understand the general guidelines and trends, ie. transistor ladder filters tend to have cleaner resonance than diode filters, or whatever. it would also be really cool to know some common or well-known filters of each type.

thanks everyone!
908 Posts and you don't know what a state variable filter is? These are filters that have any combination of low, band, and high pass outputs.
Well he probably wrote that 907 posts ago. No need for giving someone a hard time about not knowing something, especially when they are specifically asking a question to try and understand it better.

User avatar
mskala
Super Deluxe Wiggler
Posts: 2562
Joined: Thu Mar 27, 2014 7:33 am
Location: Toronto
Contact:

Post by mskala » Tue Jul 25, 2017 9:16 am

I'd like to emphasize, though it's been mentioned here before, that state-variable does NOT equal "multiple outputs." Some multi-output filters are not state-variable, and some state-variable filters do not have multiple outputs. Plenty of blind leading blind in this thread.

User avatar
Dr. Sketch-n-Etch
Super Deluxe Wiggler
Posts: 7670
Joined: Thu Jan 21, 2010 1:46 pm
Location: Vancouver, BC, Canada

Post by Dr. Sketch-n-Etch » Tue Jul 25, 2017 11:28 am

A state variable filter is a filter made from a circuit which was originally designed to solve dynamic problems in analog computers. Each output represents one of the "state variables" of the dynamic system (displacement, velocity, acceleration).

If you really want to understand what type of filter is given by a certain circuit, then you really need to analyze the circuit with a current balance and derive the transfer function.
Freedom is wasted on the Free

User avatar
hermbot
Ultra Wiggler
Posts: 764
Joined: Sat Jul 16, 2016 7:42 pm
Location: Wisconsin

Post by hermbot » Tue Jul 25, 2017 12:37 pm

Dr. Sketch-n-Etch wrote:A state variable filter is a filter made from a circuit which was originally designed to solve dynamic problems in analog computers. Each output represents one of the "state variables" of the dynamic system (displacement, velocity, acceleration).
That's a very useful statement and is an interesting corollary to vibration analysis in the mechanical world. We typically measure acceleration because it's easiest and then integrate to get to velocity, double integrate to get to displacement. They each highlight high, mid, and low frequency content respectively.

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1332
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Wed Jul 26, 2017 7:11 am

Also note that state-variable is a generic term that covers a number of filter topologies, not just the common KHN 2-pole design that almost everyone uses.

Neil
Random ramblings :: http://www.njohnson.co.uk

User avatar
mskala
Super Deluxe Wiggler
Posts: 2562
Joined: Thu Mar 27, 2014 7:33 am
Location: Toronto
Contact:

Post by mskala » Wed Jul 26, 2017 7:44 am

neil.johnson wrote:Also note that state-variable is a generic term that covers a number of filter topologies, not just the common KHN 2-pole design that almost everyone uses.
True. It might even be said to be a design method as much as a class of topologies - you can analyse some filters in terms of state variables even if they weren't conceived that way.

User avatar
lilakmonoke
Super Deluxe Wiggler
Posts: 2755
Joined: Thu Oct 21, 2010 6:18 am
Location: berlin

Post by lilakmonoke » Sat Mar 16, 2019 3:18 pm

second thread reanimation, as this is a really important topic. im forever looking for a simple explanation of whats behind the mechanics of a filter. ive done a fair bit of dsp programming and in dsp a filter is usually based on time shifts that translate into the frequency domain.

one of the simplest filter setups is a sine wave and a slightly detuned copy of itself. the detune creates a phaseshift so the waves cancel each other out over time - add more complex waves and you have a comb filter or phaser.

then translate that into a simple filter with a capacitator and a few resistors. the capacitor creates the time shift and there is your filter - so thats how im trying to understand filters because i cant follow the math very well. i might be wrong though ;-)

User avatar
Pelsea
Veteran Wiggler
Posts: 728
Joined: Thu Apr 19, 2018 11:46 am
Location: Santa Cruz CA
Contact:

Post by Pelsea » Mon Mar 18, 2019 9:46 pm

lilakmonoke wrote:second thread reanimation, as this is a really important topic. I'm forever looking for a simple explanation of whats behind the mechanics of a filter.
The basic action of a filter depends on the concept of Impedance, which is a frequency dependent opposition to current flow. There is a simple formula to relate capacitance, inductance, resistance and frequency which I won't go into. The important take away is that capacitive impedance goes down with frequency and inductive impedance goes up with frequency. (Frequency doesn't matter to resistance.) So if we include capacitors or inductors in a circuit, it will behave differently at various frequencies. We can start with that formula and derive mathematical models for how any combination of caps, inductors,* resistors and amplifiers behave.
[*Actually we avoid inductors in audio because they are so big.]

This math generates a lot of buzzwords:
Biquad: The intense math is a biquadratic equation. It can describe practically anything, so it is not much use in this discussion.

Transfer function: Fancy term for the shape of the filter response curve. This refers to the family of formulas used to design or analyze the filter. There are many of these, including some that exist only in the digital world. All transfer functions have order-- first order, second order etc. These are also loosely called "poles", and correspond to the number of frequency sensitive components. So a filter built around 3 capacitors would be a third order filter. (The easiest way to visualize this is as a filter that filters a filter that filters a filter. Or a cascade of filters if you like.) The curves we commonly see in synthesizers include:

Bessel-- Filters always change the phase of the signal. Bessel designs have the least phase change and very gentle slopes. This makes them popular with audiophiles, but you need a high order to do the kind of filtering we like.
Butterworth-- For electronic music, we are looking for steep slopes. The ultimate slope is determined by the order of the filter (in steps of 6 dB/Octave) but the slope in the first octave or two can be steeper than this. The transition from the flat (passband) region to the initial slope is called the knee. A Butterworth design has a relatively broad knee, but this can be tightened up in higher order circuits and by adding feedback (called resonance or occasionally Q) to the citcuit. This is probably the most common shape in the modular world.
Chebyshev-- The Chebyshev function has a steeper initial slope and a sharp knee, but the passband is not perfectly flat, with ripples at subharmonics of the cutoff.
Elliptical-- also known as a Cauer filter. This is similar to Chebyshev but has a slightly steeper initial slope and less ripple in the passband. These are complex circuits that require many closely matched parts.

Beyond the math, we also distinguish filters by the basic design of the circuit. We will run into:

Passive:-- This is just a capacitor and resistor (or inductor and resistor, just exchange the words low pass and high pass)-- if in series, it's a high pass filter, if the capacitor has one end connected to ground it is low pass. There are lots of these in our modules-- not in the signal path but along the power busses. We also use the high pass configuration to block DC and convert gates to triggers.
Active: The problem with passive filters the signal we want is reduced in amplitude. We can fix this by adding an opamp or discrete transistor amplifier. This makes up a basic block, which is combined with other blocks for more advanced transfer functions.
Sallen-key: This circuit features an easy way to get two poles from one opamp or transistor (one pole is in the positive feedback path). The low parts count is attractive. You can get a 24 db/Oct slope with two opamps. The Sallen-Key circuit can be configured to generate any function, but only one at a time. It can easily be made to oscillate by adding feedback, which may or may not be a good thing-- just before oscillation there is a pronounced bump at the cutoff frequency which can produce a quack or wood block sound.
State variable: As someone mentioned, this is an electrical model of a pendulum. It needs at least three opamps arranged in what amounts to a circle. The main advantage over Sallen-key is that lowpass, highpass and bandpass are available from different points in the circuit. (Notch is available by mixing high and low pass.) Most designs allow very high resonance without oscillating.

We also identify filters by the method used for voltage control of frequency. The most common approach today is to replace frequency determining resistors in Sallen-Key or State variable designs with operational trans-conductance amplifiers or OTAs. These are the ubiquitous 3080 designs or various filter IC chips.

A classic approach that is still venerated is the ladder design. This design features capacitors enfolded in transistor pairs in such a way that a changeable current gradually increases the cap's effect on the signal. These elements are usually drawn "stacked" on the schematic which is the origin of the term ladder. This was Bob Moog's first patented invention. You can do the same thing less elegantly with stacked diodes, a design that turned up in competing instruments, presumably to avoid paying royalties.

The most recent approach uses something called switched capacitors. Since a capacitor works by absorbing current over time (then giving it back) if you disconnect one from the circuit repeatedly, you are effectively changing its capacitance. Further, you can make a tiny cap behave like a large one. We can now buy integrated circuits which the frequency is controlled by the duty cycle of a very high frequency clock. This is not hard to voltage control.

So with four common circuits used to generate four common transfer functions, each configurable for low, high, or band pass we already have 48 possible filters to talk about, and I've left out many alternative designs. Naturally no two of these filters sound the same. In fact, a golden ear might be able to hear differences between two apparently identical circuits because the most critical components are difficult to manufacture to spec. 1% precision capacitors exist, but not necessarily in the values a careful design might require. Most caps in our modules will have a 5% tolerance.
Books and tutorials on modular synthesis at http://peterelsea.com
Patch responsibly-
pqe

User avatar
lilakmonoke
Super Deluxe Wiggler
Posts: 2755
Joined: Thu Oct 21, 2010 6:18 am
Location: berlin

Post by lilakmonoke » Wed Mar 20, 2019 3:31 am

that is a brilliant writeup, thanks! i can follow all that except for the impedance part but like i said i come from dsp. where do OTA based filters fit into this? i really like the sound of these maybe because they are connected to "the roland sound"?

im currently developing a wavetable synth with a lot of harmonics and am looking for an analog filter that has low distortion so the harmonics dont get blurred. the switched capacitor filter sounds like an interesting option. somehow i feel the simpler the circuit the better the sound.

i would like to experiment with a breadboard and simple filter topologies where do i start? you develop modules yourself because you mention "our modules"?

here is the diy switched capacitor filter module by big ed, the man with an arduino in every module :-)

[video][/video]

User avatar
mskala
Super Deluxe Wiggler
Posts: 2562
Joined: Thu Mar 27, 2014 7:33 am
Location: Toronto
Contact:

Post by mskala » Wed Mar 20, 2019 4:40 am

lilakmonoke wrote:im currently developing a wavetable synth with a lot of harmonics and am looking for an analog filter that has low distortion so the harmonics dont get blurred. the switched capacitor filter sounds like an interesting option. somehow i feel the simpler the circuit the better the sound.
Although harmonics being "blurred" isn't a precise term with a specific meaning, I wouldn't reach for a switched-capacitor filter first if "blurred harmonics" were what I was trying to avoid. Because they operate on the basis of sampling, they're capable of aliasing effects (put in one frequency, get out a different frequency) that don't happen with other methods. And if you're taking the steps to properly mitigate that then the circuit is not going to end up really simple.

User avatar
lilakmonoke
Super Deluxe Wiggler
Posts: 2755
Joined: Thu Oct 21, 2010 6:18 am
Location: berlin

Post by lilakmonoke » Wed Mar 20, 2019 5:51 am

Because they operate on the basis of sampling, they're capable of aliasing effects
you might be right, on the other hand the sound of my synth is based on aliasing, foldback frequencies, amplitude modulation and integer relationships so the clock frequency of the filter might just fit in perfectly.

here are sounds out of the pure data prototype, all the frequencies are conharmonic, meaning in whole number relationships. it includes a complex sequencer based on integers and a delay that is run through an analog eq. if you listen closely the eq in the delay progressively isolates certain frequencies.

prototype soundbites:
https://soundcloud.com/lilakmonoke/in-3 ... e-mountain

so thats one of the functions im looking for in a filter, clean frequency isolation. the other one is some analog sound shaping, just the right amount of noise etc. to counter the deadly clean digital signal - but little distortion because there is so much detail in the frequency spectrum.

User avatar
Pelsea
Veteran Wiggler
Posts: 728
Joined: Thu Apr 19, 2018 11:46 am
Location: Santa Cruz CA
Contact:

Post by Pelsea » Wed Mar 20, 2019 11:55 am

Impedance is simply a description of how a circuit responds to an AC signal. It is a combination of:
Resistance, opposition to the flow of current
and:
Reactance, opposition to changes in the flow of current. (Exhibited by devices that store charge such as capacitors and inductors.)
Impedance is measured in Ohms. There are tutorials on impedance all over the web.

Conductance is the inverse of resistance (i.e. the capacity for current flow). g = i/v where as you probably remember, r = v/i. I miss the days when transconductance was measured mhos, as that was always good for a laugh in otherwise kind of dull lectures. (Siemens has its own humor, but for a different sort of mind.)

An operational transconductance amplifier is similar to an opamp, but has an extra pin that is biased to control the conductance. Thus you can use one anywhere a variable resistor would be appropriate. The advantage is easy voltage control, the disadvantages are limited signal amplitude and a tendency to distort. There are plenty of OTA tutorials out there too.

[Edited to correct sloppy use of terminology.]
Last edited by Pelsea on Thu Mar 21, 2019 7:47 pm, edited 1 time in total.
Books and tutorials on modular synthesis at http://peterelsea.com
Patch responsibly-
pqe

User avatar
acidbob
Super Deluxe Wiggler
Posts: 2681
Joined: Mon Sep 19, 2011 2:48 am
Location: Denmark

Post by acidbob » Thu Mar 21, 2019 8:02 am


User avatar
lilakmonoke
Super Deluxe Wiggler
Posts: 2755
Joined: Thu Oct 21, 2010 6:18 am
Location: berlin

Post by lilakmonoke » Thu Mar 21, 2019 2:43 pm

Impedance is simply a description of how a circuit responds to an AC signal.
thanks again, i think im close to understanding it. that reminds me of my highschool years when i bought lots of books on electronic circuits and looked at the pictures dumbfolded. that was before the internet ;-)
great link! tim has a switched capacitor filter design that sounds great and seems really easy to build. you can filter down to 40 hz before the clock noise becomes audible at 16 khz, in my case way lower ;-)

http://www.timstinchcombe.co.uk/index.php?pge=lmf100

listen: http://www.timstinchcombe.co.uk/synth/m ... 00_otr.mp3

Image

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1332
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Mar 21, 2019 3:44 pm

Pelsea wrote:Impedance is simply a description of how a circuit responds to an AC signal. It is a combination of:
Resistance, opposition to the flow of current
and:
Reactance, opposition to changes in the flow of current. (Exhibited by devices that store charge such as capacitors and inductors.)
Impedance is measured in Ohms. There are tutorials on impedance all over the web.

Transconductance is the inverse of resistance (i.e. the capacity for current flow). g = i/v where as you probably remember, r = v/i. I miss the days when transconductance was measured mhos, as that was always good for a laugh in otherwise kind of dull lectures. (Siemens has its own humor, but for a different sort of mind.)

An operational transconductance amplifier is similar to an opamp, but has an extra pin that is biased to control the conductance. Thus you can use one anywhere a variable resistor would be appropriate. The advantage is easy voltage control, the disadvantages are limited signal amplitude and a tendency to distort. There are plenty of OTA tutorials out there too.
Kind-of-right. If I may offer some clarifications:

Capacitors store electrical energy in the form of electrical charge between two plates (E = half-C-V-squared), inductors store electrical energy in the form of magnetic flux generated by current flowing in the windings/turns (E = half-L-I-squared).

The inverse of resistance is conductance, not transconductance. And for completeness, the inverse of impedance is admittance, and susceptance is the inverse of reactance. And if you think "mho" is funny, wait until you find out the unit of elastance....

Transconductance is the ratio of a circuit block's output current to its input voltage (typically an amplifier). FETs are characterised by their transconductance (gm) to indicate the small-signal ratio between changes in the current flowing through the Drain-Source and corresponding changes in the voltage at the Gate (relative to the Source).

There are also transresistance amplifiers - input current drives output voltage - often used as current-to-voltage converters.

A typical OTA produces an output current proportional to the difference in input voltages, scaled by the third input signal (in the case of the LM13700 and CA3080 this control is a current fed into the IABC pin, where ABC stands for Amplifier Bias Current). There is typically no feedback path (unlike most op-amp circuits), unless you're making something like a filter.

Neil
Random ramblings :: http://www.njohnson.co.uk

jorg
Veteran Wiggler
Posts: 617
Joined: Fri Apr 03, 2015 9:38 am
Location: East Coast USA

Post by jorg » Fri Mar 22, 2019 3:06 pm

This will answer a ton of questions:
https://www.analog.com/media/en/trainin ... apter8.pdf

User avatar
Grumble
Super Deluxe Wiggler
Posts: 1619
Joined: Thu Dec 17, 2015 5:24 am
Location: U aint much if U aint Dutch
Contact:

Post by Grumble » Fri Mar 22, 2019 3:53 pm

http://sim.okawa-denshi.jp/en/

and here you can calculate your filter and more

Post Reply

Return to “Modular Synth General Discussion”