Analog modeled waveforms. Who's doing it?

[cytospur]

A simple high pass filter will convert a perfect sawtooth in to an analogue-like 'shark fin' sawtooth. Similarly a square will get converted to an alternating positive and negative shark fin sawtooth like wave. All explained by the presence of high pass filters in analogue synths.

[brok landers]

A simple high pass filter will convert a perfect sawtooth in to an analogue-like 'shark fin' sawtooth. Similarly a square will get converted to an alternating positive and negative shark fin sawtooth like wave. All explained by the presence of high pass filters in analogue synths.

indeed. and in essence, that (along other things that happens to the signal while it runs through the whole cirquit) is how it is done in analog synths. the osc itself often does create just a "ping", the easiest way to create a constant rate (i.e. periodic signal), which then is shaped with some kind of filters not only on the osc stage itself, but on the sum of the whole (stringmachies f.e.), to get the desired waveform, whatever that might be. that´s (static summary filtering) also why often the osc shape differs gradually when analysing different pitches and that´s (not only that) also why the osc´s from different synths sound so different.
one needs to get away from the thinking that in digital synths the "perfect saw" or the "perfect square" is something to aim for. instead when developing a software synthesizer, the goal is to use your ear on how it sounds, how the osc interacts with the filter response, how the envs are treating the audible outcome. there is no _strict_ rule, infact some of the most beloved analog synths are quite away from a math perfectness.
though it has to be mentioned, the newer the analog synths got back then, when digital slowly took over, the more accurate they have been produced. it´s the same with mixing consoles - the ones that came towards the end of the analog ära were the ones that are close to digital in terms of accuracy. which of course had its reason. in the analog days you had to put your focus on perfection, as if not done so, the tiny misbehavements of each unit in the chain would sum up to a lower quality in sound. nowadays it´s the opposite - we´re heading backwards to that imperfection, as digital simply is mostly accurate along the pathway of the signal (not counting in bad coding or rounding errors) - so the character that was there in the _whole_ analog chain is missing, therefore we aim for it to be added again in a now controllable degree.

[brok landers]

A simple high pass filter will convert a perfect sawtooth in to an analogue-like 'shark fin' sawtooth. Similarly a square will get converted to an alternating positive and negative shark fin sawtooth like wave. All explained by the presence of high pass filters in analogue synths.

indeed. and in essence, that (along other things that happens to the signal while it runs through the whole cirquit) is how it is often done in analog synths. the osc itself often does create just a "ping", the easiest way to create a constant rate (i.e. periodic signal), which then is shaped with some kind of filters not only on the osc stage itself, but on the sum of the whole (stringmachies f.e.), to get the desired waveform, whatever that might be. that´s (static summary filtering) also why often the osc shape differs gradually when analysing different pitches and that´s (not only that) also why the osc´s from different synths sound so different.
one needs to get away from the thinking that in digital synths the "perfect saw" or the "perfect square" is something to aim for. instead when developing a software synthesizer, the goal is to use your ear on how it sounds, how the osc interacts with the filter response, how the envs are treating the audible outcome. there is no _strict_ rule, infact some of the most beloved analog synths are quite away from a math perfectness.
though it has to be mentioned, the newer the analog synths got back then, when digital slowly took over, the more accurate they have been produced. it´s the same with mixing consoles - the ones that came towards the end of the analog ära were the ones that are close to digital in terms of accuracy. which of course had its reason. in the analog days you had to put your focus on perfection, as if not done so, the tiny misbehavements of each unit in the chain would sum up to a lower quality in sound. nowadays it´s the opposite - we´re heading backwards to that imperfection, as digital simply is mostly accurate along the pathway of the signal (not counting in bad coding or rounding errors) - so the character that was there in the _whole_ analog chain is missing, therefore we aim for it to be added again in a now controllable degree.

[zerocrossing]

A simple high pass filter will convert a perfect sawtooth in to an analogue-like 'shark fin' sawtooth. Similarly a square will get converted to an alternating positive and negative shark fin sawtooth like wave. All explained by the presence of high pass filters in analogue synths.

indeed. and in essence, that (along other things that happens to the signal while it runs through the whole cirquit) is how it is done in analog synths. the osc itself often does create just a "ping", the easiest way to create a constant rate (i.e. periodic signal), which then is shaped with some kind of filters not only on the osc stage itself, but on the sum of the whole (stringmachies f.e.), to get the desired waveform, whatever that might be. that´s (static summary filtering) also why often the osc shape differs gradually when analysing different pitches and that´s (not only that) also why the osc´s from different synths sound so different.
one needs to get away from the thinking that in digital synths the "perfect saw" or the "perfect square" is something to aim for. instead when developing a software synthesizer, the goal is to use your ear on how it sounds, how the osc interacts with the filter response, how the envs are treating the audible outcome. there is no _strict_ rule, infact some of the most beloved analog synths are quite away from a math perfectness.
though it has to be mentioned, the newer the analog synths got back then, when digital slowly took over, the more accurate they have been produced. it´s the same with mixing consoles - the ones that came towards the end of the analog ära were the ones that are close to digital in terms of accuracy. which of course had its reason. in the analog days you had to put your focus on perfection, as if not done so, the tiny misbehavements of each unit in the chain would sum up to a lower quality in sound. nowadays it´s the opposite - we´re heading backwards to that imperfection, as digital simply is mostly accurate along the pathway of the signal (not counting in bad coding or rounding errors) - so the character that was there in the _whole_ analog chain is missing, therefore we aim for it to be added again in a now controllable degree.

interesting. So now I wonder if my ATC's filter is ever open all the way.

I'm also wondering if things like the filter state can effect things upstream, like how a speaker effects a tube amp's power output. Or because it's a transistor based design elements don't have that kind of interaction.

[EvilDragon]

In a lot of analog synths, filters don't open all the way above 20 kHz, which WILL roll off the edges of any raw waveform.

[Urs]

Well, any 4-pole/24-dB lowpass filter will have a 12 dB attenuation at the cutoff frequency (3dB per pole). Therefore, even if the filter opens all the way to 20kHz, some of the audible frequencies are still attenuated.

Usually a little resonance will compensate.

[elassi]

[one needs to get away from the thinking that in digital synths the "perfect saw" or the "perfect square" is something to aim for.

Not too long ago a dev - claiming to provide the best sounding oscillators in the market - proudly presented his "perfect wave" and compared it to other synth's waves.

That was funny.

[Urs]

I don't think any analogues do particularly perfect waveforms?

With few exceptions, an analogue sawtooth is just perfect, and so is a square wave. Triangle is more complicated.

This is so because even very cheap circuits can create perfect waveforms easily.

What's usually not perfect in any analogue oscillator:

- slow frequency drift due to temperature changes or other small effects
- noise injection due to bad shielding

Nevertheless, cycle per cycle the waveforms are close to perfect, but once you sample them into the digital domain, they show the same kind of jaggy edges that any good sounding digital counterpart shows.

Lastly, as was said before, any "round belly shape", cut off corners and bent edges are the effect of follow on circuitry, most notably the DC-filters and waveshapers.

[aciddose]

Most of the synthesizers I've used have no such limit at exactly 20khz, this would be ridiculously difficult to implement a circuit to do.

Instead the cutoff can be placed at any position with no limitation other than the clipping of the CV in some stage before the conversion to current. (This exact position will be fairly arbitrary depending upon the circuits used for CV and current conversion as well as power supply and so on.)

Generally the range of the cutoff control is, given full key tracking, 25hz to 25khz, approximately.

This range is ten octaves. 10 to 10 + 1/2 is just generally a "good idea" for a cutoff control. Say 16hz to 25khz, this would be approximately 10 + 2/3rds.

You might notice MIDI uses 128 note values, 128/12 = 10 + 2/3rds.

amplitude(hz, fc) = fc / sqrt(hz^2 + fc^2)
db(x) = 20 * log10(x)

At 20khz with fc =
20k = -3db
25k = -2.15db
40k = -0.97db
80k = -0.26db

Doesn't make a whole lot of difference really until you get at least two octaves up. It trails off exponentially, so you get diminishing returns.

The most important issue for me is having the ability to specify the 90 degree point (the frequency of oscillation) well beyond the lower and upper limits of hearing. So, quite a bit less than 20hz, quite a bit more than 20khz.

As you say Urs, the 3db can be made up quickly by increasing feedback to produce a butterworth.

[ghettosynth]

interesting. So now I wonder if my ATC's filter is ever open all the way.

I'm also wondering if things like the filter state can effect things upstream, like how a speaker effects a tube amp's power output. Or because it's a transistor based design elements don't have that kind of interaction.

Although, as pointed out, many do not, the "filter" in question doesn't necessarily have to be the synth's filter proper. Something as simple as DC blocking capacitors will act as a high pass filter at some frequency that is dependent on the input impedance of the next stage. There are a lot of details within a circuit that can be approximately modeled as a simple high or lowpass filter that will have some affect on the waveform. In an ideal world, many of these circuits should be transparent, e.g. blocking capacitors, but, in reality, this isn't always the case.

[aciddose]

Btw for anyone interested in the math:

If you want highpass just reverse the order of hz/fc to the function.

amp(hz, fc) = fc / sqrt(hz^2 + fc^2)

low pass = amp(hz, fc)
high pass = amp(fc, hz)
band pass = amp(hz, fc) * amp(fc, hz) (when you involve Q, things are a bit more complicated. Easier to jump directly to transfer functions and complex math at that point.)

To increase the slope, just raise the result to the order you want:

6db = amp(hz, fc)^1
12db = amp(hz, fc)^2
18db = amp(hz, fc)^3
24db = amp(hz, fc)^4

This reflects the fact that indeed the 3db position is shifted when you increase the filter order. Since you can use some basic algebra here, you can get the position you need to set fc in order to get 3db at a specific point if desired.

[Lotuzia]

For those who are interested in figures just take a look at the middle of this page, and you'll see -litterally- that oscillators are not all similar -by far- in analog synthesizers ( And that's *also* part of why they sound different one from each other, and why some emulation sound more accurate than others )

EMS Ramp and Square OSCs

[zerocrossing]

So, then why the hell would M-Audio include a ton of samples of different analog wave forms on the Venom when they could have included cool alternate waves?

[Kriminal]

So, then why the hell would M-Audio include a ton of samples of different analog wave forms on the Venom when they could have included cool alternate waves?

Such as? The venom has more than sine/saw/sq waves.

[hakey]

For those who are interested in figures just take a look at the middle of this page, and you'll see -litterally- that oscillators are not all similar -by far- in analog synthesizers ( And that's *also* part of why they sound different one from each other, and why some emulation sound more accurate than others )

This is too easy,

Comparing the oscillators of a VCS3 and XILS 3 immediately reveals differences. For example, the sawtooth waves generated by the soft synth are brighter than those generated by the vintage synth, and its sine wave is much purer. Likewise, the pulse waves generated by XILS 3 are much more precise than those that you'll obtain from the original [...] Unfortunately, there's a significant error in XILS 3's oscillator 1, whose Shape control is linked to the sawtooth wave, not the sine wave. This is a shocking mistake!