[LAD] What KvR didn´t understand.

[Ove Karlsen]

On 1/7/2013 1:57 PM, Ove Karlsen wrote:
> What KvR didn´t understand 10 yrs ago, and still don´t understand.
>
> Why does digital synths often sound so bad? Either stale, or harsh etc.
>
> Let me tell you in complete truth and honesty, is has got nothing to 
> do with digital. It has something to do with the engineers making the 
> algorithms.
>
> When I was a newbie DSP engineer, the first thing I tried was making a 
> TB-303 filter. Which is what a lot of people do first. I talked to the 
> people on #musicdsp, and they had little clue, some had tried and said 
> it was difficult or impossible, some say they had succeded but their 
> filters didn´t sound too good.
>
> On a few days, not having touched code, since I was 12 years old, I 
> did a resonance filter, that screamed and shreaked. Some engineers in 
> the KvR forum, said it was a bad thing to do, because their job now 
> got so much more difficult.
>
> When in reality, it was not difficult at all. And this is typical for 
> those kinds of engineers. They don´t get into the algorithm. They 
> don´t understand what is going on. Instead very unecesary high-level 
> theorems, they try to fit into what is simple analog feedback paths.
>
> One of the guys even worked with supposedly professor for many years, 
> and they did not come up with anything good.
>
> They argue it is something to do with frequency-response, for 
> instance, why the analog filters sound the way they do, and it cannot 
> completely be done in digial.
>
> All this is just crazy trash.
>
> Later I actually looked at the schematics for the 303, and realized 
> there was just four feedback-paths with one negative feedback-path 
> around. It is as simple as that. That is all "analog vintage" 
> synth-filters. There is absolutely no obscurity going on, it is as 
> simple as it can be.
>
> Knowing that analog has a certain headroom, and that components are a 
> bit inaccurate, and there is often some highpassing going on, due to 
> the frequency-response of the components, you can model that, VERY 
> SIMPLY, and without much cpu use. Some of the stuff released on KvR 
> uses extreme cpu, and even sounds bad.
>
> Try this ok, in your synth, and you will realizing that digital can 
> sound just as good as analog, and without the inaccuracies. And analog 
> often has characteristics you DON´T want. So it is even better.
>
> Released under The Beneficient Open-source licence. Please google it. 
> Since this licence allows for functions alone, to be released as 
> opensource you can make it a function, and use it alongside whatever 
> else you use.
>
>
> //licenced under The Beneficient Open-source Licence.
>         // Osc lo-emph.
>         b_lo = b_lo + ((-b_lo + b_v) * b_lfr); // for emulating the 
> analog-charateristic of more saturation in the low-freq. (due to 
> saturated buffers)
>         b_v = b_v - b_lo;
>         b_v = b_v + (b_lo * b_lgn);
>
// there was some earlier code here that was not intended in the paste.
>         if (i_ftype == 1) { // 24dB lowpass ("ladder")
>             double b_rez = b_aflt5 - b_v; // sub = no attenuation with 
> rez.
>             b_v = b_v - (b_rez*b_fres); // negative feedback for 
> resonance.
>             b_v = b_v * b_off2; // gain offset
>             b_v = b_v + ((fvar90-0.5)*2); // bias
>             if (b_v > 1) {b_v = 1;} else if (b_v < -1) {b_v = -1;} // 
> clip
>
>             //sat/soften clip.
>             double b_vr = b_v; if (b_vr < 0) {b_vr = -b_vr;}
>             b_vr = 1-b_vr;
>             b_vr = pow(b_vr,fvar91*10); // something I tested at the 
> time, this is a filter from my synth "Abdullah", and work in progress.
>             b_vr = 1-b_vr;
>             if (b_v < 0) {b_vr = -b_vr;}
>
>             b_v = b_vr;
>             b_v = b_v - ((fvar90-0.5)*2); // bias
>             b_v = b_v / b_off2;
>
> // you can also do clipping at 0.0001 for instance, and mix, and get a 
> little resonance buildup, before resonance hits the audible range. A 
> bit similar to how some zero-cross distortion works.
>
>             b_aflt1 = b_aflt1 + ((-b_aflt1 + b_v) * b_fenva);
>             b_aflt2 = b_aflt2 + ((-b_aflt2 + b_aflt1) * b_fenva);
>             b_aflt3 = b_aflt3 + ((-b_aflt3 + b_aflt2) * b_fenva);
>             b_aflt4 = b_aflt4 + ((-b_aflt4 + b_aflt3) * b_fenva);
>             b_v = b_aflt4;
>
>             b_hp = b_hp + ((-b_hp + b_v) * b_fhp); // highpass to 
> emulate analog, and get nice resonance, and also remove DC.
>             b_v = b_v - b_hp;
>             b_aflt5 = b_v;
>         }
>
>
> That is the ultimate "analog" filter, completely digital, and without 
> inaccuracies, and ofcourse with perfect keytracking etc.
>
> Forget all the obfuscating arrogant atheist KvR-nerds. This is the 
> real deal.
>
> And all my DSP is just as perfect, and they never did anything of that 
> either.
>
> And Unix-philosophy is really close to my philosophy of "least 
> obscurity". So it would be natural for this to develop and etablish 
> itself on Linux. I was a "hacker" in my teens, and I guess many who 
> have been into hacking, and brilliant programming, really celebrates 
> God, and ofcourse comes to the same idea of least obcurity, which is 
> also very much like (non-idolaterous) religion.
>
> Instead ofcourse KvR bans the brilliant, who even talks about a 
> peacebringing religion, and peaceful meditation, according to Gods 
> praises, and the highest of intelligence, infinite human unfolding and 
> rights, if you wish. And  that is the incoherent idolater/faithless.
>
> Peace Be With You.
>