Analog filters, that sound "physical"
-
- KVRian
- Topic Starter
- 513 posts since 3 Sep, 2009 from Poland
more thoughts:
Why such a filters with physical behaviour are good? Well.. get a saxophone and blow stronger
The same with violin, drums, guitars... Of course an amplifier should be used before the filter.
Most of the synths were designed to mimic real instruments though.
Why such a filters with physical behaviour are good? Well.. get a saxophone and blow stronger
The same with violin, drums, guitars... Of course an amplifier should be used before the filter.
Most of the synths were designed to mimic real instruments though.
giq
-
- u-he
- 28061 posts since 8 Aug, 2002 from Berlin
Well I was talking about electronic filters. I have no idea whether or not they are supposed to behave like mass spring systems.
To me it is however clear as to why the filter frequency usually depends on the input gain - because the tanh-ish non-linearities work as a voltage dependent resistor, and the higher the input voltage the higher the resistance, and thus the lower the cutoff. That's why people perceive non-linear filters as "frequency modulated by the input" at high gains.
A simple trick to check this out is to route an LFO signal through a self-oscillating filter. The self oscillation will vary in pitch, controlled by the LFO. Here's a screenshot of ACE doing that (needs quite some gain to start vibrating):
http://www.u-he.com/img/ACE_selfosc_fm.png
To me it is however clear as to why the filter frequency usually depends on the input gain - because the tanh-ish non-linearities work as a voltage dependent resistor, and the higher the input voltage the higher the resistance, and thus the lower the cutoff. That's why people perceive non-linear filters as "frequency modulated by the input" at high gains.
A simple trick to check this out is to route an LFO signal through a self-oscillating filter. The self oscillation will vary in pitch, controlled by the LFO. Here's a screenshot of ACE doing that (needs quite some gain to start vibrating):
http://www.u-he.com/img/ACE_selfosc_fm.png
-
- KVRian
- Topic Starter
- 513 posts since 3 Sep, 2009 from Poland
I am aware of that. Elements work this way and so were designed the circuits many years ago. . And this ignorance is repeated in all modern synths . (despite of they wanted to model synths after real instruments). We all want to emulate these analogs, they wanted to simulate the reality.
Of course there is a sound at the end.. I just have to perform more experiments with saturated filters. My ears suggest me, that tb-303 filter is another exception (at least nonlinear resonating spring sounds quite "squelchy", one may call this a natural ringing).
Another point of view: conservation of energy, that in "the nature" is transformed into increased frequency. Maybe the opposite phenomenon is just an interesting side-nonsense, that must be cloned, cloned, cloned again
. (despite of they wanted to model synths after real instruments). We all want to emulate these analogs, they wanted to simulate the reality.Of course there is a sound at the end.. I just have to perform more experiments with saturated filters. My ears suggest me, that tb-303 filter is another exception (at least nonlinear resonating spring sounds quite "squelchy", one may call this a natural ringing).
Another point of view: conservation of energy, that in "the nature" is transformed into increased frequency. Maybe the opposite phenomenon is just an interesting side-nonsense, that must be cloned, cloned, cloned again
giq
-
- KVRAF
- 3080 posts since 17 Apr, 2005 from S.E. TN
Didn't see drums mentioned, unless I missed it. Hard struck drums, the attack can rise in pitch as well, though perhaps some of that is initial pressure on the head by the stick.
Similar phenomena with hard struck piano strings, tines, guitar strings.
Can't speculate why analog filters wouldn't also behave thataway, or how an analog circuit could be designed to behave as a struck rod. Merely offering more examples.
Well, what happens if the filter is smacked with a strong spike of dc, offsetting the initial impulse response? Stronger nonlinearity one direction versus the other?
Similar phenomena with hard struck piano strings, tines, guitar strings.
Can't speculate why analog filters wouldn't also behave thataway, or how an analog circuit could be designed to behave as a struck rod. Merely offering more examples.
Well, what happens if the filter is smacked with a strong spike of dc, offsetting the initial impulse response? Stronger nonlinearity one direction versus the other?
-
- KVRAF
- 6323 posts since 30 Dec, 2004 from London uk
@ itoa - It seems that your confusing subtractive Synthesis and Physical Modelling. PM is what youre describing :
http://en.wikipedia.org/wiki/Physical_m ... _synthesis
http://en.wikipedia.org/wiki/Physical_m ... _synthesis
-
- KVRAF
- 3080 posts since 17 Apr, 2005 from S.E. TN
I don't know the current course or topic title, but even before the invention of audio synthesisers there was a mid-level undergrad physics module involved with analysis of mechanical systems and drawing the equivalent electronic model, and vice-versa. So I interpreted the question as wondering why the difference in response, mechanical versus electronic. Should have paid more attention in class (and been born with a better memoryUltraJv wrote:@ itoa - It seems that your confusing subtractive Synthesis and Physical Modelling. PM is what youre describing :
http://en.wikipedia.org/wiki/Physical_m ... _synthesis
).-
- KVRian
- Topic Starter
- 513 posts since 3 Sep, 2009 from Poland
UltraJR: pleaaase
JCJR: hehe, I've been just wondering why this important aspect was omitted by designers of analog synhts, assuming they wanted to model real instruments. It turned out that "the correct" behaviour is modelled in synthi filter. I suppose there are more of them (303 dl?). No more theses for now. I gonna spend some time on this, my initial experiments show that filters expressing mechanical behaviour sounds more natural when excited with a single impulse (and this is somehow obvious). Haven't time yet for testing continuous signals.
JCJR: hehe, I've been just wondering why this important aspect was omitted by designers of analog synhts, assuming they wanted to model real instruments. It turned out that "the correct" behaviour is modelled in synthi filter. I suppose there are more of them (303 dl?). No more theses for now. I gonna spend some time on this, my initial experiments show that filters expressing mechanical behaviour sounds more natural when excited with a single impulse (and this is somehow obvious). Haven't time yet for testing continuous signals.
giq
-
- KVRAF
- 3080 posts since 17 Apr, 2005 from S.E. TN
A wild guess might be that in the early days it was rocket science just designing proper tracking filters, and the designers were hoping to get the most ideal linear behavior possible, and considered nonlinear behavior a defect to be avoided?itoa wrote:UltraJR: pleaaase
JCJR: hehe, I've been just wondering why this important aspect was omitted by designers of analog synhts, assuming they wanted to model real instruments. It turned out that "the correct" behaviour is modelled in synthi filter. I suppose there are more of them (303 dl?). No more theses for now. I gonna spend some time on this, my initial experiments show that filters expressing mechanical behaviour sounds more natural when excited with a single impulse (and this is somehow obvious). Haven't time yet for testing continuous signals.
All of the "voltage controlled resistor" strategies of which I'm aware, with the possible exception of hf variable width pulse switching (which has its own warts)-- They tend to have very narrow linear regions. So once you pick your poison, the work is to get the circuit to have as wide a clean dynamic range possible. Entering the nonlinear region considered a defect. Exactly as in the early days cranking up a high output gibson geeter into a high gain fender amp, and the resultant distortion, was a horrible defect.
So after a designer has picked his voltage controlled resistor equivalent, the nature of the nonlinearity would be of little concern, except as something to be avoided?
Betcha bob never bragged about the distortion characteristics of his ladder, until after it was discovered by musicians that the distortion had musical applicability and became a bullet point on the feature list.
-
- KVRian
- Topic Starter
- 513 posts since 3 Sep, 2009 from Poland
Indeed, great points!
On the other side: most of us talk about great nonlinear analog filters and recreate (physically model ) them in DSP. Something unintended and somehow counter common sense is recreated again and again. This sounds quite good in certain (existing) setups, but imo this aspect deserves for further experiments and discussion.
On the other side: most of us talk about great nonlinear analog filters and recreate (physically model
) them in DSP. Something unintended and somehow counter common sense is recreated again and again. This sounds quite good in certain (existing) setups, but imo this aspect deserves for further experiments and discussion.giq
-
Zaphod (giancarlo) Zaphod (giancarlo) https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=111268
- KVRAF
- 2596 posts since 23 Jun, 2006
our developers sampled more than 300 analog filters (some of them variations of the same gear, different components and so on). It is true that some of them have a strange frequency response (little peaks here or there) or even phase (old ssl consoles) but what makes an analog filter musical is imho distortion, which balances everything especially in low pass case. When we compared hardware with other software plugins, same model, the plugin one was in most case just a simple cutoff (pure math lowpass maybe with a gentle curve) No harmonic distortion, no phase distortion and so on. Sound is much like when you hear music from an other room, ie in a disco pub from outside. It is a cool and perfect effect, very abused in dance productions, but really far from a minimoog filter, or mutronics (daft punk), they sound very good and balanced even if resonance is zero and they are processing statically your source (without sweeping cutoff or adding other things like pitch shifting or delays). Auto-oscillation is cool... but I repeat, those filters are balanced even when resonance is zero.itoa wrote:Indeed, great points!
On the other side: most of us talk about great nonlinear analog filters and recreate (physically model
my 2ct
-
- KVRian
- Topic Starter
- 513 posts since 3 Sep, 2009 from Poland
300 filters... thats impressive. Have you found more filters that behave this way?
Our trained by the nature ears love nonlinearities, because almost nothing is this world is linear. This is crucial. But as you said most probably circuit developers didn't care about this, other aspects like stability were more important.
(repeated) Most of analog filters have nonlinear characteristic opposite to these found in real vibrating objects. But, having constant drive and input signal this difference is hard to grasp and remains pleasant.
This difference is more obvious with lower frequencies, when we hear separate resonance peaks. The natural and expected effect is a small pitch drop of every "ring"- down-chirps; the vibrating body is losing energy, restoration force gets smaller and more linear. Most saturated analog filters act the opposite and generate up-chrips.
Our trained by the nature ears love nonlinearities, because almost nothing is this world is linear. This is crucial. But as you said most probably circuit developers didn't care about this, other aspects like stability were more important.
(repeated) Most of analog filters have nonlinear characteristic opposite to these found in real vibrating objects. But, having constant drive and input signal this difference is hard to grasp and remains pleasant.
This difference is more obvious with lower frequencies, when we hear separate resonance peaks. The natural and expected effect is a small pitch drop of every "ring"- down-chirps; the vibrating body is losing energy, restoration force gets smaller and more linear. Most saturated analog filters act the opposite and generate up-chrips.
giq
-
- Beware the Quoth
- 33144 posts since 4 Sep, 2001 from R'lyeh Oceanic Amusement Park and Funfair
And if you dont assume that, you have the answer for why it was omitted.itoa wrote: I've been just wondering why this important aspect was omitted by designers of analog synhts, assuming they wanted to model real instruments.
my other modular synth is a bugbrand
