Riddle me this: synth filters

[Dr88s]

Here's a question that I have been wondering about for years but never had a good forum to ask about it.

 

When you are reading reviews about an analog or virtual analog synthesizer, the reviewer all often comment about how great a particular synthesizer's filter sounds.

 

I know very little about synthesizer programming, but I do understand the basics, and particularly I understand the ideas behind subtractive synthesis.

 

My question is, if a low pass filter will simply block frequencies above the cutoff, or a high pass filter simply reject frequencies below the cutoff, why would a particular synthesizer's filter sound any different than another's? I presume the answer lies in the use of the word 'simply'.

 

I have the same question about oscillator waveforms in general. Is a sawtooth not a sawtooth not a sawtooth? Does a sine wave on a Voyager sound any different then a sine wave on a Mopho or Nord Lead?

 

This is nothing I have ever lost sleep over. I just figured it was time to ask, what with all the experts around here

[timwat]

I'll take a stab at getting the conversation started.

 

You'll often read about four-pole and two-pole filters, which essentially refer to how "steeply" or abruptly the filter curtails frequencies below the cutoff (for a typical low-pass filter in subtractive synthesis). This is one primary audible difference between filters from different synth models, and from manufacturer to manufacturer.

 

Doubtless there are other reasons specific filter designs sound different from others, but that's the main one I'm familiar with from the olden days when I first learned subtractive on Moogs and ARPs.

 

As far as oscillator waveforms, realize that different basic shapes contain more or less harmonic content (overtones) by definition - a sine wave is essentially the bare fundamental, while the triangle and saw introduce more harmonics, and pulse-width modulation more (I think .... correct me if I'm wrong somebody).

 

One question becomes - if I select "triangle" or "square" from the front panel label on any particular synth, what does an oscilloscope actually reveal in terms of the waveform the machine is outputting? I read reviews and articles where the actual waveform produced by a synth is "more or less" an academically correct triangle, but not 100% - which will account for a unique, "personal" deviation from the conceptually perfect.

 

That's as far as I've gotten so far in my understanding of it. A lot of guys here are much more knowledgeable about this stuff than me. I'm just gonna go practice piano some more.

[AnotherScott]

I think that's basically right. On analog electronics, none of these filters or waveform generators worked perfectly.

 

As I understand it, the basic sound of a Hammond tonewheel was "supposed" to be a sine wave, but luckily, it was hard to get it perfect. Today's Hammonds like the SK1 contain settings that do produce pure sine waves. No real tonweheel model ever sounded like that, but it is ostensibly what they were aiming for. I would guess that most SK owners pick one of the variations that better mirror the character of the real B3s! Similarly, the Moog filter didn't merely have certain filtering characteristics, but was slightly overdriven which introduced a unique distortion. I'm not sure what other differences there may be, between what the circuit did and what the theoretical ideal was.

 

With digital electronics, it's easy to get all these parameters perfect, and I think the key to getting them to sound good is to replicate the equivalent of analog imperfections.

[mate stubb]

No real world filter responds in ideal fashion. The various ways to build a filter cause it to handle feedback, gain changes and overdriven conditions very differently than others.

[Dave Weiser]

This is a topic that is both wide and deep. Countless synth geeks have spent endless hours agonizing over such things.

And yet the variety in oscillators and filters (and the respective character sounds) IS worthy of attention.

 

I'll give you a few basic bits and a few links to check out.

 

With filters there are two basic things to consider:

1. Steepness of the cutoff - this is measured in dB per octave and is described with 1-pole, 2-pole, 4-pole, etc. Each pole gives a 6 dB/octave response. 4-pole was considered the gold standard back in the day of classic analog synths, but plenty of great synths just had 2-pole filters.

 

2. Quality - better filters will provide more of a "brick wall" against the frequencies being blocked. Years ago I had to work with some (what I consider) cheesy rompler synths from the 90s that had supposed low-pass, high-pass synth filters and band-pass filters in their effects sections (for wah effects) where I felt the damned screen was *just lying* to me. I could hear the supposedly blocked frequencies coming through and the "wah" even when set to "100% wet" still gave me some of the original dry signal. With current synths these days (both hardware and software) I don't think this happens as often.

 

The various producers of classic analog synths - Sequential, Oberheim, Mooog, ARP, etc, all used slightly different components for their filters and oscillators and some of them did have unique characteristics, often caused by "flaws" and imperfections. One example is the clipping in the Minimoog's amp envelope with mixed oscillators that resulted in a "punchy" sound that is still desirable.

 

Another thing to consider is envelope curves - certain synths had linear envelope segments while others had curved responses - either logarithmic or exponential (think convex or concave curves). Certain synths allow you to select between linear, log, and exp response. (Kurzweil's prototype VA1 synth offered these options - we had a blast playing with that section!)

 

Here are a few links to check out - some are heavier than others. The first one is quite heavy, but the first 3 pages are really good.

http://www.analog.com/library/analogdialogue/archives/43-09/edch%208%20filter.pdf

http://www.hollowsun.com/vintage/minimoog/

http://www.lomountsound.org.au/classic_analogue_poly.html

 

Beyond that, you'll want to break out your Google-Fu and have at it with some searches and nerdly research.

 

 

 

 

[ksoper]

And then go play as many as you can.

[Dave Bryce]

No real world filter responds in ideal fashion. The various ways to build a filter cause it to handle feedback, gain changes and overdriven conditions very differently than others.

Same with oscillators. As previously mentioned, they do not behave perfectly. Most of them overshoot the end of their waveform's cycle by at least a little bit, so a square wave does not drop off at perfectly at 90 degrees at the end of it's cycle.

 

Also, the guys doing the programming handle the artifacts of the components in different manners...and different combinations of components produce different results. That's why the same Curtis chip sounds different in different instruments....

 

A synth's sound is the result of it's entire signal path. That's one of the reasons that it cracks me up when I see people posting about listening to "raw waveforms" - you simply can't. First of all, the oscillator itself is inaudible without the VCA....and secondly, even if a filter is all the way open, the signal is still running through it. It's only possible to compare different "raw oscillators" when they're in the same system (AFAIK, only possible with a modular).

 

And how come everyone always talks about the sound of the VCO and VCF....but you rarely (if ever) hear anyone talking about which VCA was used...?

 

dB

[The Real MC]

Care to elaborate on the VA1 that never made production? Would had been an interesting path for Kurzweil.

[Dave Bryce]

Care to elaborate on the VA1 that never made production? Would had been an interesting path for Kurzweil.

You don't wanna go there....trust me.

 

dB

[The Real MC]

The VCA doesn't get as much attention as VCFs and VCOs because vintage VCAs were low fidelity and easy to distort. VCA technology has advanced today in that they are high fidelity/low distortion; there was a point where musicians started noticing something missing in newer synths. Tom Oberheim lamented that the OB-8 was too clean compared to OB-Xa, OB-X, and SEM-based polysynths and you could trace the change in sound to the advances of VCAs with each model.

 

I think a lot of focus on VCOs/VCFs comes from today's modulars. I don't know of any "dirty" VCA modules on the market.

 

But touching on VCFs, when you read a review that comments how great a filter sounds they are often referring to its resonant qualities as you turn up the resonance control. My acid test patches for filters are brass/reeds and choir as these really define the resonant quality of filters. Certain filters are best for choir, certain ones for brass/reeds, and then you find certain filters work well for the subdivision of brass/reeds such as tenor sax and baritone sax. Then you get into wind instruments where another set of filters do them well.

 

Even if you're not into imitating traditional instruments, resonant quality is still relevant for non-traditional sounds.

 

And you can't just assume that resonant quality is consistent between filters of like slopes yet different architectures... the Moog 24dB/oct ladder filter has a vastly different resonant quality than a CEM3320 24dB/oct cascaded OTA. And to add more konphooshun to the mix, the Moog ladder filter in the minimoog sounds different from the ladder filter in the Taurus... the two differences are the feedback architecture of the filter and the coupling of the entire VCO-VCF-VCA chain.

 

There is a world of permutations out there.

[mate stubb]

TI think a lot of focus on VCOs/VCFs comes from today's modulars. I don't know of any "dirty" VCA modules on the market.

 

I can't think of many either, but the vactrol VCAs in a Quad Low Pass Gate are pretty funky.

[Dr88s]

I am floored by the depth of responses. Thank you to all, especially Dave Weiser, dB, and Grim.

 

As a child of the digital age, it wouldn't have occurred to me that these filters and VCOs are actually circuits with resistors and capacitors and not mathematical models of a perfect wave. Same with filters as a digital filter can probably brick wall at a set frequency.

 

I have never played anything analog before. I am a piano / EP / organ kind of guy. The only reason these things have become more important to understand is that I am playing an upcoming show where we are trying Bowie's 'Heroes'. The rest of the band thought the usual pad stuff I was playing was too boring so I turned to all of those heretofore ignored buttons with scary words like 'cutoff' and 'resonance' on them to make swishy swooshy moving sounds as the song builds.

 

Who knows - maybe this experience will make we want to be a Voyager...?

[Doerfler]

I am floored by the depth of responses. Thank you to all, especially Dave Weiser, dB, and Grim.

 

As a child of the digital age, it wouldn't have occurred to me that these filters and VCOs are actually circuits with resistors and capacitors and not mathematical models of a perfect wave. Same with filters as a digital filter can probably brick wall at a set frequency.

 

I have never played anything analog before. I am a piano / EP / organ kind of guy. The only reason these things have become more important to understand is that I am playing an upcoming show where we are trying Bowie's 'Heroes'. The rest of the band thought the usual pad stuff I was playing was too boring so I turned to all of those heretofore ignored buttons with scary words like 'cutoff' and 'resonance' on them to make swishy swooshy moving sounds as the song builds.

 

Who knows - maybe this experience will make we want to be a Voyager...?

 

excellent thread. maybe this experience will make you want to buy a Voyager, not just be one. good luck with "Heroes", it's a great tune

[Dave Weiser]

Care to elaborate on the VA1 that never made production? Would had been an interesting path for Kurzweil.

 

dB is right... the "never made" part of the story is long and painful... I would rather have a laser enema than go into the long explanation. It was a tough time that involved a hostile takeover, layoffs, bankruptcy, etc... The up side is that the company miraculously bounced back and is currently doing quite well.

 

Here's the marketing blurb I wrote for it about 10 years ago.

http://www.synthtopia.com/content/2004/02/04/kurzweil-va1-keyboard-announced/

 

Additional good news - Many of the VA1 oscillators and filters made it into the current Kurz architecture (PC3, PC3K, PC3LE, SP4, SP5, Artis).

[Dr88s]

maybe this experience will make you want to buy a Voyager, not just be one. good luck with "Heroes", it's a great tune

 

 

Darn typos.

 

[mate stubb]

When I built my huge modular, I had a great opportunity to explore the different sounds of different circuits. My modular had the following filter designs, all modern but totally analog reproductions - Buchla Low Pass Gate, Moog modular 4 pole ladder, early Prophet SSM 2040, and more.

 

For oscillators, I had sawtooth core, triangle core, digital rom table scanners, and oscillating filters.

 

They all sound distinctly different. It's terribly addicting but therapeutic to stand in front of a modular and patch drones for hours.

 

http://www.hotrodmotm.com/images/case/sod_biggest.jpg

[Dave Weiser]

When I built my huge modular, I had a great opportunity to explore the different sounds of different circuits. My modular had the following filter designs, all modern but totally analog reproductions - Buchla Low Pass Gate, Moog modular 4 pole ladder, early Prophet SSM 2040, and more.

 

For oscillators, I had sawtooth core, triangle core, digital rom table scanners, and oscillating filters.

 

They all sound distinctly different. It's terribly addicting but therapeutic to stand in front of a modular and patch drones for hours.

 

http://www.hotrodmotm.com/images/case/sod_biggest.jpg

 

Holy sh*t man you built that thing?!

'Tis a thing of beauty.

Angels must have wept when you first powered it on.

If I'm ever in your neck of the woods I'd give one of these to spend some time with that beast.

 

 

[Jim Alfredson]

I never heard about the VA-1. Sounds like Kurzweil missed the boat on that one. VA's were big back then and in many ways continue to be so.

[The Real MC]

I never heard about the VA-1. Sounds like Kurzweil missed the boat on that one. VA's were big back then and in many ways continue to be so.

 

*sigh* If only... played with one at NAMM 2005 and I believe I met Mr Wieser there

 

http://www.sequencer.de/pix/kurzweil/kurzweil_va1.jpg

[J. Dan]

One thing to consider is that many filters actually overdrive, especially at high resonance, at which point you're actually adding harmonics in addition to subtracting. Really, since the resonance is what really brings out the character of the filter, I'd focus more on differences in filter resonance, which is more complex than just the slope of the rolloff.

[The Real MC]

Yup, it was Mr Weiser just found his business card.

[Jim Alfredson]

I never heard about the VA-1. Sounds like Kurzweil missed the boat on that one. VA's were big back then and in many ways continue to be so.

 

*sigh* If only... played with one at NAMM 2005 and I believe I met Mr Wieser there

 

http://www.sequencer.de/pix/kurzweil/kurzweil_va1.jpg

 

Whoa, so it actually existed? Who has that prototype?

[Mike Martin]

Although it evolved dramatically after I left, I was the product manager at Kurzweil during the early stages of the VA-1. Somewhere I have all the sketches and design ideas.

[johnchop]

Who has that prototype?

 

I believe Dave W. does.

 

 

[johnchop]

Also, "laser enema" is a good hardcore synth band name.

[Dave Bryce]

 

Whoa, so it actually existed? Who has that prototype?

I know of a couple of them in the field. Just saw one of them recently, actually.

 

dB

[Theo Verelst]

Oh boy, what a subject. Electronically speaking, probably there are the following main aspects in the historic synth design. The filters are obviously filtering off higher frequencies from a certain frequency onward, and do so with a certain curve as of course said above, but they also have the interesting property that the cutoff frequency is variable, traditionally under voltage control, and with interesting side-effects (and essentially speaking immediate response).

 

Also the filters, depending on kind and drive parameters, can significantly distort the signal in terms of harmonic and/or inter-modulation distortion. The filter itself usually (in case of the widely known ladder and ota variation) filters it's own distortion components to a degree, so depending on setting it can act as a signal exciter, even with subtle and interesting properties, and in other cases (like the Prophet) it can be adjusted and driven in such a way that it arrives at a particularly "clean" waveform, with interesting slight distortion modulations creating a comely spectrum.

 

The modulation of the filter frequency generates transients and transition note/effects that are hard to model mathematically (non-stationary , non-linear multi-pole system with non-trivial modulation sources), as well as sine-waves coming from the resonance which have a humanly interesting phase response behavior, connected with the amount of energy stored in it's various stages, which varies when the self-oscillation frequency changes.

 

Some of the uses of particularly the filter (of course given the waves fed to it and the modulation sources influencing it) are to make certain signal properties relate to the "equal loudness curves" such that the user of the synthesizer can play the patches he or she creates at multiple volumes (you could also read: the more window rattling is produced the funner the tones become...).

 

Finally, the well known synth designers made sure of two things in the whole design, very included also in the filter design, that certain frequency ranges in the most sensitive range for human beings, have properties which aren't creating very dangerous wave forms (the sound smooth, not like a blaring transistor radio can sound for instance), and those waves by "magical" design properties also sound good in reverberant spaces, by design. Often, in the harmonically interesting frequency range, say the middle two octaves of a piano, there is also a designed in "dimension"-connection with the sound, sort of that it sounds without all too wrong side-effects on a warm sounding huge speaker setup.

 

Digitally speaking it's a hell of a job to get everything right, mainly because of the limitations imposed on the Digital Signal Processing by the sampling and signal reconstruction theory (theoretical EE university subjects), which many programmers aren't aware of, and because the whole signal path of the filter can in most practical cases only be simulated properly by solving matrices instead of "simple" software iteration schemes, so that the interactions in the filter components are properly modeled.

 

I know for a fact, because I worked on it, measured some properties and have some results already, some digital simulations are fairly complicated, in some cases, like certain software packages, because they use certain not-so-simple mathematical approximation techniques, and in other cases, like the existing Kurzweil VA synthesis, because the main architect(s ?) have put in some very convoluted and difficult to make come out right EE building blocks, which cover subjects between sampling issues and signal averaging properties, and allow all kinds of wonderful filter simulations to take place, but as the OS221 is, it is such a mess and mixture and so far away from picking a oscillator and a filter kind, nobody should find it very funny to listen to the blurps and blips and muted screams it produces, with only a few exceptions.

 

T

[JeffLearman]

As far as oscillator waveforms, realize that different basic shapes contain more or less harmonic content (overtones) by definition - a sine wave is essentially the bare fundamental, while the triangle and saw introduce more harmonics, and pulse-width modulation more (I think .... correct me if I'm wrong somebody).

A square wave is only even harmonics (or only odd, depending on how you count them). A triangle has only odd harmonics (or even, again, depending). Amusingly enough, the Wikipedia page for square wave uses a different way of counting than the Wikipedia page for triangle wave. Sigh. The "depends" part depends on whether you count the fundamental as zero or one.

 

I don't know about the details of PWM waves, but varying the pulse width varies the harmonic content. I couldn't say whether it introduces new ones or deletes old ones or just changes the relative amplitudes of various ones.

 

 

As a child of the digital age, it wouldn't have occurred to me that these filters and VCOs are actually circuits with resistors and capacitors and not mathematical models of a perfect wave. Same with filters as a digital filter can probably brick wall at a set frequency.

Actually, no filter can be a perfect brick wall, as far as I know. A digital filter can come *way* closer, thanks to not being restricted to "IIR" (infinite impulse response) filters. [One of the biggest single improvements to digital audio was the introduction of so-called "one-bit A/D's" -- that is, delta-sigma modulation. This happens in the early stages of an A/D converter, and works at a very high sample rate (megahertz). It allows the signal to be digitally brick-wall filtered before conversion to the usual format. The digital filter works WAY better than the analog filters that we had to use previously.]

You can do even better if you're allowed to introduce latency.

 

The first virtual analog synths (and a lot of ROMplers and other sample players) do use software filters that are pretty close to the ideals. Those worked fine for sample players, but sounded boring for virtual analogs, so they worked on making VA digital processing sound more like the classic analog circuits.

 

It's just a guess -- I'd like to hear what Mate and others say -- but I suspect that the biggest differences are with resonance and when circuits are saturated. I'd suspect that, without resonance and without saturation, one manufacturer's 2-pole filter would sound a lot like another's.

[mate stubb]

Square waves AND triangle waves are generally agreed to have only odd harmonics, and sawtooth or reverse sawtooth waves have only even harmonics.

 

The triangle wave's harmonics are much lower in amplitude than the square, so that it sounds more rounded and less buzzy.

[Dave Bryce]

I don't know about the details of PWM waves, but varying the pulse width varies the harmonic content. I couldn't say whether it introduces new ones or deletes old ones or just changes the relative amplitudes of various ones.

The latter.

 

A pulse wave contains only odd-numbered harmonics. Square waves are pulse waves with 50% duty cycles...or, the ON time is equal to the OFF time. Changing the duty cycle ratio changes the balance of the harmonics, so every pulse wave has it's own harmonic structure/sound.

 

If you get a chance to work with one of the analog synths that have continuously variable waveforms (as opposed to discrete switching, where only fixed values are available) you can really hear what the effect of changing the pulse width does.

 

dB