Circuit modeled filter, how to?

[aciddose]

"So you are saying that these differences do make an audible difference?"

audible? only to a person who knows what to listen for. yes, it will make a difference, potentially, to a listener in a sub-conscious way. it might sound better, or worse. it is _NOT_ the specific attributes of a specific circuit which make it better, or worse, however. what you need to do is figure out what sounds good and use that - forget the circuits. if you're only using the circuits as a guide to what sounds good, use your ears and your mind.

again i'm saying "you", when i mean "you, the people". you, specifically, can do whatever you prefer. obviously your intention is to model the circuits, not the fundamental functions they're designed to produce. i just do not think that is a good way to learn. how did i know what to do? i've been doing this stuff for almost a decade. i do not think OP wants to spend a decade learning, so let's just try to start him off with something that works and he can spend time improving his knowledge/skills.

[andy_FX]

If you read any of my posts, you would know that I take circuits as a great starting point for doing modeling. They sound good for a reason, they are real. I use my ears to pick out the good bits, and use my brain to extrapolate the rest and make new ideas.

Wow, 10 years of working this stuff out by trial and error. Ok, so how long was that Electronic Engineering course again? It took me a year of fiddling with trial and error analog modeling like you have said until I found qucs, and now I am learning way faster, and in the last year I've been able to model loads more stuff really well.

I am against spoon feeding people snippets. I'm all for: here are some references, and here is a program to experiment with, you go play and sort it out and come back as ask questions about equations later on etc. This works, and people learn stuff.

Andrew

[aciddose]

ten years of audio in general.

i've never bothered to try modeling anything, i prefer using the "just works" method that i've mentioned and adding the different timbre modifying elements that i feel improve the sound. i've been saying i think that is the best way to learn. get a "just works" function, learn how it works, figure out which elements you want to add and then learn how to do so.

doing a true electronic model isnt going to teach you really how things work. you'll never learn what the fundamental functions behind those circuits are. you'll just start plugging in values and you might become dependent upon such things.

even as a circuit designer - if you stick to circuits themselves and never bother to understand what you're really trying to do you'll never produce anything but shit. the circuits do not exist just to serve themselves - they are an attempt to approximate some ideal function that we want to perform. in your models, if you want to model the korg-35 for example, you'll get noise, clicking, limited slew rate of the cheap opamps will give you problems with upper frequency range. you'll have to implement all the over-complicated kludge solutions like odd scaling (resistor values) and so on just as the designer of the korg-35 did.

do you want to be exact? do you want the clicking, the noise, the muffled timbre, poor control response, inaccurate frequency tracking? if so go on and model your circuit. if you want the timbre and the _positive_ attributes of the circuit you'd better think more in terms of the fundamentals.

[andy_FX]

You take the good bits that you want from a circuit and use them. Drop the rest. I have always said that. When you understand the circuit and what the person was doing that designed it you have a better understanding of the big picture and can make your own designs and extend things. I advocate as much knowledge as possible to make better decisions.

Btw I like self noise of circuits, it's raw and cool sounding to compress, but it should be optional.

Andrew

[andy_FX]

in your models, if you want to model the korg-35 for example, you'll get noise, clicking, limited slew rate of the cheap opamps will give you problems with upper frequency range. you'll have to implement all the over-complicated kludge solutions like odd scaling (resistor values) and so on just as the designer of the korg-35 did.

I think if you badge something as a particular synth (jupiter-8, ms-20) you should deliver. I have been singularly unimpressed with all analog synth models that have guis that imply they are a particular type of analog synth. But if you did do this I would expect a level of detail including slew rate limiting, and odd scaling that you say. I release stuff that is more just "inspired by" particular filters and I have my own take on them. I can post you the Orca technical info if that would help.

Andrew

[aciddose]

i completely understand where you're coming from, i havent really been intending to question what you yourself do. i was only arguing the merits of having somebody like williamk do that sort of stuff.

[andy_FX]

Now, we have had a great rant, and I hope people have been enjoying. I want to help people here, not try to show off or anything. When I put my list up at the start, there are some big scary sounding things, but in the end, that is there name, and I have used it. How about some easy to follow example:

Non-Linear Anaysis

Now really this is a cut down version of full non-linear modified nodal anaysis, but it illustrates the simplicity of the equations for people like me that are just learning this stuff.

It's all easy to follow, it is circuit modeling, and it is the kind of stuff people use to design circuits. It can be simplified as well to something that can run real time.

There is another page which has capacitors on it, but they got one of the equations slightly wrong to do with the equivalent value for the cap. I'll leave it for everyone to spot the mistake:

Transient Analysis

So, there we go, not that hard and all useful in understanding and modeling simple circuits. With these two you can extend to most all other simple solid state circuits since everything is really just a non-linear VI curve, some resistance and capacitance and inductance. When you start getting into non-linear magnetic core saturation of inductors you start getting into another world of fun that I don't really understand as yet, but hope to soon. Getting some cool tube models up and running would be fun too.

Andrew

[andy_FX]

Btw you never need to model every single last little component, lots of caps and resistors you can leave out since they are to do with making op amps stable and note oscillate at the limits of their slew rate etc. Just play with qucs and check out what is going on, you will learn loads. You can simplify you whole model, and even put in equations directly into qucs. Did I say how much it rocks? It rocks! Part of the simplification process when modeling something is culling all the dead weight and just leaving what you need to get the sound you want.

Andrew

[Aleksey Vaneev]

Whom you'll consult if you want to create a device nonexistent in a form of circuit?

I myself design processes mentally, I design function, not implementation. Of course, various implementations (e.g. I was surprised to see valve cascade DSP code - it's very hard to visualize its function mentally due to its feedbacks) extend complexity of functions I design, but not much since DSP is basically:
1. Summing (which is a magnitude/phase summing)
2. Z-delay (which is a phase warping by means of phasor function)
3. Multiplication by constant (which is a magnitude change followed by frequency-shift by 0 Hz-i.e. no phase shift), multiplication of signal A by signal B (which is a sum of frequency-shifted copies of signal A, shifted by frequencies present in B, all at the same time; negative frequencies are also shifted if real signals are used).

There are no other basis elements available. Schematics simply try to replicate these elements, and do it in a not so easy way since they are forced to be built around system of equations (unless you design your schematic with Tesla's one-conductor/unclosed signal path).

I personally think that today you do not even need to model circuits 'the analog' way: you may start with DSP initially, and bring the model to analog by equating SampleRate to infinity and resolving the limits.

[andy_FX]

Don't downplay some of the brilliant circuit designs out there. People intentionally using components the wrong way around to achieve certain goals. And even the unplanned aspects of circuit leading to a particular characteristic sound and very useful. I've been able to see certain patterns in a few compressors that have led me having a much deeper understanding of how to design my own compressor, and taking these accidental but great sounding features and putting a knob on them.

I consider studying classic circuit designs the same way that a oil painter would study the masters. First you understand the techniques they have used, and then you can use that as a basis to extend and excel. Or even ignore them altogether and do something completely different. But without studying the masters first I feel you leave a hole in your education that leaves you open to others considering you ignorant and not well versed in your field, of which there is much in audio dsp.

Andrew

[Aleksey Vaneev]

But without studying the masters first I feel you leave a hole in your education that leaves you open to others considering you ignorant and not well versed in your field, of which there is much in audio dsp.

Well, in mathematics there are no masters. You make it the way it should be done, or fail (theorem is proven or not). It's engineering, not an art that much as painting. There are no 'patterns' or 'styles' in mathematics/engineering you may follow. Beside that no schematic will give you a pattern of making 'alive' DSP processes: only your own skill and deep understanding of AUDIO will help you tie lifeless DSP code to music (and audio is mostly math, acoustics and psychology - no schematics there).

[hifiboom]

Well, in mathematics there are no masters. You make it the way it should be done, or fail (theorem is proven or not). It's engineering, not an art that much as painting....

really?
I think the other way.
Maths is just the basics. An analog circuit will always be correct mathematically. Its based on the real world physics, which can't be fooled.

Maths doesn't tell you how to design a synth or circuit?

A digital DSP circuit can surpass real world physics, so the maths get more important in digital world, but they don't replace imagination and creativity.

just my opinion. no offense

[Aleksey Vaneev]

A digital DSP circuit can surpass real world physics, so the maths get more important in digital world, but they don't replace imagination and creativity.

I would say the same for circuits. They can't replace imagination and creativity. That's basically what I wanted to say.

[keldon85]

Hmmm, the beauty of [drawn] art is the unique way in which the artist brings out the best elements of the picture. Picasso is a fine example, by removing every single [unnecessary] detail he was able to create wonderful works of art.

Good circuitry is just good circuitry that brings out specific features (often its own artefacts) of the signal path. So like any art the beauty lies in the inexact replication - otherwise you end up with the pure sound back.

Now that's the funny thing, the pure sound; we're simply not used to it so it wounds weird when we are exposed to it! Plus analogue happens in real time without PCM's concept of frames and integration.

[Aleksey Vaneev]

Now that's the funny thing, the pure sound; we're simply not used to it so it wounds weird when we are exposed to it! Plus analogue happens in real time without PCM's concept of frames and integration.

Analogue is probably better represented via PDM (pulse density modulation) since we deal with the flow of equal electrons. So, analog is not analog at all if you perceive it as a flow of electrons spaced apart (it translates to a very high sample rate, though, but it's not infinite).