[sdiy] Odp: Pink?

[Rutger Vlek]

Regarding filters, I have done quite a few experiments with existing and
new filter designs when it comes to their resonance behaviour over the
spectrum. You have some static (wrt cutoff setting) and some dynamic
behaviour in resonance responses. A nice test to perform is to plot the
maximum self-oscillation amplitude at a fixed resonance level and sweep the
cutoff. Some filters show serious roll-off towards high frequencies, which
I find extremely musical. The Yamaha CS filters are a great example.
Tracking inherently suffers as a consequence though. I suspect a small
phase shift occurs in the resonance loop (slew rate? Or capacitance
somewhere? ) that dramatically affects the resulting resonance amplitude.
In the Dendrites multimode VCF I modelled this phase shift in high
frequencies with a RC network for the vintage resonance mode. I've also
heard filters where the opposite effect occurs (e.g. when capacitance of
zener diodes used to soft-limit the resonance path play up), making
resonance at high frequencies ear-bleedingly loud and harsh.

Rutger

Op di 21 aug. 2018 12:06 schreef Mattias Rickardsson <mr at analogue.org>:

> Thanks for bouncing around thoughts about this topic, taking it in
> interesting directions. :-) Bouncing back some comments below:
>
> On Fri, 17 Aug 2018 at 16:45, Quincas Moreira <quincas at gmail.com> wrote:
>
>> The waveforms are what they are, squares and saws simply sound bright due
>> to the nature of the waveforms and their harmonic content. Which is why
>> since the early days filters have been used in synthesis, or waves like
>> triangle and sine with variable waveshaping to achieve the desired amount
>> of higher harmonics. The model 15 for example has a nice non resonant
>> filter bank precisely for these kinds of adjustments. I just use the 3 band
>> eq in my Hexmix mixer to roll off some highs when needed. To do so in the
>> VCO itself would IMO be undesirable, you may need those high harmonics to
>> make a filter sweep really sing, or for FMing other sound aources etc.::
>> My take on this is: Just use eq to taste when needed.
>>
>
> True, it's "just" an EQ adjustment - but my point here is that it's the
> VCO waveform in itself that is too bright, not that the resulting synth
> sound is too bright in the end. What I mean is that the VCO gets too bright
> and also too loud when opening up the filter, compared to when the filter
> is almost closed and everything sounds alright. If this would be
> compensated with an EQ, you'd have to adjust that EQ when tweaking the
> filter. And since the filter is typically tweaked by its envelope every
> time a synth sound is played, the EQ would have to follow the filter
> envelope. All this just to compensate for a problem with the VCO signal.
> Not really efficient! ;-)
>
> I don't fully agree with that history writing regarding the use of filters
> in synthesis, but that's less important here. :-)
>
> BTW, I should also mention that I'm not suggesting that the spectral
> balance of VCOs is necessary problematic in modulation routings, only in
> the audio routing that is routed in the direction of human ears.
>
> On Fri, 17 Aug 2018 at 17:18, <mskala at ansuz.sooke.bc.ca> wrote:
>
>> But any low-pass VCF with at least one pole will roll off the high end at
>> at least 6dB/octave in the limit of high frequencies, which is already a
>> sharper roll-off than traditional 3dB/octave pink noise.  So if the output
>> of the VCF still sounds too bright for you (too much high-frequency
>> content) and you want a sharper roll-off, I'm not sure what to tell you
>> except to try a sharper filter.  And if you don't like the sound of a lot
>> of resonance, try using less resonance.
>>
>
> Full resonance often sounds OK in the low end of the cutoff range, but too
> loud in the highs. Having the same amplitude regardless of the cutoff
> frequency, a whistling resonant filter even has more excess energy in the
> highs than the already too bright VCO waveshapes have. :-)
>
> Using less resonance overall would decrease it also in the lower end.
> Introducing a cutoff-dependent resonance amount (reducing the resonance at
> higher cutoffs) is actually quite useful (and has been used successfully to
> alter the perceived character of a filter), but also has a dilemma in the
> few cases where a whistling oscillator-like sinewave *is* the expected goal.
>
> On Sat, 18 Aug 2018 at 09:24, Roman <modular at go2.pl> wrote:
>
>> You can always limit the slew rate, that will give you more pink, or even
>> red or brown.
>>
>
> This first appeared to me as a less good suggestion, something that often
> happens with the radically best of ideas. But:
>
>
>> For example I enjoy very much a square-ish waveform that is derived from
>> triangle fed to overdriven amplifier. It produces almost a square wave when
>> you look at it at the scope, but it's more like trapezoid. And the slope
>> changes with frequency so it sounds nice, like filtered.
>> Sawtooth may be not so easy - I can imagine a VCO with 2 integrators...
>>
>
> This reminds me of a very minimalistic triangle core VCO that has a mode
> where it is (mis)used as a sawtooth oscillator by letting one of the
> triangle slopes have a much higher rate. The resulting "sawtooth" wave is
> far from perfect, having a much slower reset than normal, but still quick
> enough to give all the audible overtones in almost the right amount. But to
> my great surprise there was a different tone to it that people actually
> liked more than ordinary oscillators.
>
> Designing such a slewed sawtooth VCO in a more thorough fashion might be
> somewhat of a heavy exercise. I'd want the slewrate to be a controllable
> parameter; I'd want it to follow exponential and linear control - something
> that would need to work regardless of the slewrate used; I'd want the
> timing of other generated waveforms to follow the zero crossings and not
> the peaks of the sawtooth; and finally, for FM duties and freedom of thumps
> I'd want the saw to be DC-free, meaning that the start and end of the
> slewed reset slope would need to be equally far from zero - which wouldn't
> be the case of an ordinary sawtooth with a slew function hung onto its
> output.
>
> The overdriven triangle wave is actually a great parallel example to this,
> thanks for mentioning it!
>
> On Sat, 18 Aug 2018 at 11:12, Rutger Vlek <rutgervlek at gmail.com> wrote:
>
>> I share your opinion, when it comes to solo sounds from analog
>> oscillators. I've often noticed how these sounds are EQ'd and otherwise
>> processed (e.g. Lots of reverb) at mixing stage such that the resulting
>> master actually follows a rather pink (3 dB/oct) frequency spectrum.
>>
>
> Yes, and it would be nice to get a more "production ready" sound out of
> synths to begin with.
>
> There's a really accurate static pink filter implementation on the Esp
>> (Elliot) website, but I prefer more control so I designed a spectral
>> tilting module for Eurorack. It allows you to tilt the spectrum (even CV
>> controlled!) of a sound up to a positive and negative 3dB/oct slope. I've
>> achieved great results with it on my VCO's, being able to make them
>> gradually more pink. Combining a pink-ish VCO voicing with overdrive also
>> sounds wonderful, and avoids harshness. Putting tilting under CV control is
>> also cool, as it gives a very gentle and natural modulation of timbre (e.g.
>> modulation with lfo or brightening with higher velocity). I've even
>> achieved a very passable analog leslie simulator with it in a stereo
>> configuration, partially because my implementation of tilting also involves
>> small phase shifts that are modulated along with timbre.
>>
>
> I've studied that tilting module of yours before, and was struck by how
> much more useful it was compared to my initial reactions. Good ideas!
>
> On Sat, 18 Aug 2018 at 19:16, KD KD <pic24hj at gmail.com> wrote:
>
>> Some of the Roland polys have a overall "designed" gain and spectral curve
>> JP6/MKS80 i vaguely recall. Its the only poly synths i have seen doing
>> this but
>> i can imagine stoneage stuff like PolyMogg, ArpQuddra, Bork Triddent might
>> have some of this implemented due to crap PA's back in those days, but
>> none
>> of above mentioned for the sake of pinkyfy "certain" function modules as
>> such.
>>
>
> Are you referring to the slight bass boost that is introduced if the HPF
> slider is set to zero? I think this is quite an ingenious trick, not only
> to make the sounds more impressive but also to actually introduce some kind
> of EQ that balances up the sound in a similar fashion to our discussions
> here.
>
> On the other hand the brightness you experience might be a physical
>> condition you have deteriorated into? you no longer a youth, i for
>> instance
>> entered glass-eye condition some years ago and 0603 is like flys fart to
>> solder! :-)
>>
>
> "Glass eye" is a... let's say... INTERESTING way of translating that word
> to English. People will be quite impressed that you're still into building
> stuff. ;-)
>
> On Mon, 20 Aug 2018 at 12:57, Roman Sowa <modular at go2.pl> wrote:
>
>> After breakfast I usually come up to conclusion that morning's idea was
>> over complicated and can be done much simpler.
>> It wasn't different this time either :)
>>
>
> Most brilliant ideas are made mainly out of breakfast. Breakfast is one of
> the most important basic elements, after carbon, oxygen and sillycon. :-)
>
> Actually, considering my wish to correct the spectral balance of both VCOs
> and VCFs, a Roland-inspired (but extended into also attenuating the highs)
> post-EQ might be the best (or at least simplest) choice after all - for
> pure synth sounds. Its only big drawback would be that the sound of the
> filter is EQed, giving it an even sharper slope. Wouldn't be much of a
> problem I guess.
>
> If a VCF is used to process samples or other "correct" sounds that
> shouldn't be EQed further, the story is different.
>
> /mr
>
>
>
> > On Aug 17, 2018, at 09:09, Mattias Rickardsson <mr at analogue.org> wrote:
> >
> > I often experience VCOs "too bright", having too much high end in the
> sawtooth & pulse waves.
> > I often experience VCF resonance peaks "too bright", being too strong at
> higher cutoffs.
> > I often experience white noise "too bright", having too much high end...
> >
> > ...while pink noise feels more balanced. Equal power per octave, not per
> Hz.
> > Also, good sounding loudspeakers/studios tend to have a bit pink-ish
> spectrum roll-off from the sound source to the listener position,
> suggesting that some sort of pinkification of clean sounds could be
> desirable.
> >
> > So, why aren't VCOs and VCFs more spectrally pink?
> > Have there been any attempts historically to alter their characteristics
> a bit in synths? :-)
> >
> > /mr
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20180821/cd98e1da/attachment.htm>