Simple questions on synth tech basics...

[JCJR]

As we all know, many developers are trying to recreate the sound of old analog synths.
I have been playing around with the equalizer in Sylenth1 quite a bit, trying to figure out which frequencies to manipulate in order to make it sound more analog.
Is there any consensus on what frequency curve promotes the analog sound?

(Unfortunately my headphones are anything but neutral (they intentionally boost some frequencies to be more analytical), so it is hard to tell whether what I am dialing in on the Sylenth1 equalizer really make it sound more analog.)

Hi fluffy

I've not played with modern softsynths and know nothing about their nature. Guesswork involved.

Though modern softsynths have copious capabilities in anti-aliasing and all kinds of emulated distortion-- I wonder if some of them are "too hi-fi"?

Hiss noise is/was a constant companion to analog, as was hum. Do modern softsynths commonly have controls to emulate hiss and hum?

It wouldn't necessarily need to be loud/noticeable to give a "feel" to the sound. Hiss and hum was always there if you had a quiet listening environment and took the time to specifically listen for it. But usually if the hiss and hum was not stupid-loud, the ear would ignore the noise and listen for the music. However, even if fairly quiet, those noises were still there.

In the 1980's or thereabouts, Bob Carver did interesting work characterizing audio gear especially amplifiers. Linked at end of message.

One detail I recall in a magazine interview or article long ago, Bob claimed that the famous cherished audiophile characteristic of "air" can be duplicated on-demand merely by adding a very small amount of high-frequency white noise. Not a lot of high-frequency white noise. Just a little bit does the trick.

For sake of argument, perhaps call it subliminal noise and subliminal hum. Present if you listen for it, but normally not noticed. Low-level hum in analog synths and sometimes other gear could sometimes subtly interact with audio passing thru the circuitry. So not necessarily just adding hum to a clean signal. More like intermodulating slight hums with the clean signal.

And hum is a 50 or 60 Hz sine wave plus distortion harmonics. Sometimes pure 60 Hz (or 50 Hz, depending on locality). Sometimes "mostly 120 or 100 Hz" the second harmonic of the power line frequency. Sometimes lots of harmonics going up into the spectrum. Once the hum gets fairly broadband it gets called buzz.

Just wondering if trying to make something sound "analog" with an equalizer wouldn't do it-- If the ear is mostly missing certain types of subliminal hiss and/or hum? A clean equalizer couldn't boost hiss or hum if it wasn't in the original signal.

https://en.wikipedia.org/wiki/Bob_Carver
Carver caused a stir in the industry in the mid-1980s when he challenged two high-end audio magazines to give him any audio amplifier at any price, and he’d duplicate its sound in one of his lower cost (and usually much more powerful) designs. Two magazines accepted the challenge.

First, The Audio Critic chose a Mark Levinson ML-2 which Bob acoustically copied (transfer function duplication) and sold as his M1.5t amplifier (the “t” stood for transfer function modified).

In 1985, Stereophile magazine challenged Bob to copy a Conrad-Johnson Premier Five (the make and model was not named then, but revealed later) amplifier at their offices in New Mexico within 48 hours. The Conrad Johnson amplifiers were one of the most highly regarded amplifiers of the day, costing in excess of $6,000 a pair.

Of note that in both cases, the challenging amplifier could only be treated as a “black box” and could not even have its lid removed. Nevertheless, Carver, using null difference testing, (null difference testing consists of driving two different amplifiers with identical signal sources and exact levels, but out of phase by exactly 180 degrees. If the amplifiers were 100% identical, no sound would be heard. If sound was heard, the audio amps had different properties). Bob Carver used "distortion pots" to introduce amplifier characteristics, fine-tuned to null-out any sound differences. His "motel-room" modified amplifier sound was so similar, Stereophile Magazine editors could not tell the difference between his amplifier and one costing more than $6,000. This amplifier was marketed as the M1.0t for about $400.00. Bob Carver may have single-handedly debunked any number of theories about sound quality by using physics, blind and double-blind testing and unbiased measurements, such as "Gold-plated" speaker wires sound better than copper wires, etc.) Carver successfully copied the sound of the target amplifier and won the challenge. The Stereophile employees failed to pass a single blind test with their own equipment in their own listening room.

[fluffy_little_something]

i know some guys here cook bacon
i've been baking some cookies here and was wondering what's the best oven temperature curve that will give them that bacon taste?

what i've learnt is that things aren't so simple
"analog sound" doesn't mean much to me unless you put some more context to it
the best way you can go at this is to pick some specific analog sound (from a recording of an analog synth) and try to recreate it as accurately as you can, analyze the difference, then improve your approximation, then analyze again, improve more, repeat untill you get sick
i've went this route with one of the "simplest" analog synths and it was quite a long way

my advise is:
if you want a specific analog synth sound - try to get the best starting point - a dedicated "emulation" softsynth
if you don't have a specific synth in mind, and you just want to get the "well known (and undescribable) analog sound" - sorry dude, i don't know how to help, but i think you'll be wasting your time chasing that


as for your second post
i'm not familiar with many analog synths, but from those i know - i can't think of even a single one which has oscillator phase-reset on trigger (if that's what you mean)
oscillators naturally run free, so it's simpler and cheaper to not add extra circuitry to reset them at a specified phase

and i am surprised you're using phase-reset to get "punchiness"

I know there are several aspects developers try to emulate when trying to achieve analog sound in software.
And one of them is equalization in my view. For instance it seems to me that hardware has better bass. So I am wondering which frequency range to emphasize with my parametric eq.
I have also heard people say that hardware had a roll-off on high frequencies.

So I was just wondering if there are certain frequency ranges that should be manipulated.


Regarding the phase, I like phase reset, it allows for more consistent percussive sounds. Sounds like the Woodblock preset in Sylenth1 would probably not be possible without phase reset.

[fluffy_little_something]

Hi fluffy

I've not played with modern softsynths and know nothing about their nature. Guesswork involved.

Though modern softsynths have copious capabilities in anti-aliasing and all kinds of emulated distortion-- I wonder if some of them are "too hi-fi"?

Hiss noise is/was a constant companion to analog, as was hum. Do modern softsynths commonly have controls to emulate hiss and hum?

It wouldn't necessarily need to be loud/noticeable to give a "feel" to the sound. Hiss and hum was always there if you had a quiet listening environment and took the time to specifically listen for it. But usually if the hiss and hum was not stupid-loud, the ear would ignore the noise and listen for the music. However, even if fairly quiet, those noises were still there.

In the 1980's or thereabouts, Bob Carver did interesting work characterizing audio gear especially amplifiers. Linked at end of message.

One detail I recall in a magazine interview or article long ago, Bob claimed that the famous cherished audiophile characteristic of "air" can be duplicated on-demand merely by adding a very small amount of high-frequency white noise. Not a lot of high-frequency white noise. Just a little bit does the trick.

For sake of argument, perhaps call it subliminal noise and subliminal hum. Present if you listen for it, but normally not noticed. Low-level hum in analog synths and sometimes other gear could sometimes subtly interact with audio passing thru the circuitry. So not necessarily just adding hum to a clean signal. More like intermodulating slight hums with the clean signal.

And hum is a 50 or 60 Hz sine wave plus distortion harmonics. Sometimes pure 60 Hz (or 50 Hz, depending on locality). Sometimes "mostly 120 or 100 Hz" the second harmonic of the power line frequency. Sometimes lots of harmonics going up into the spectrum. Once the hum gets fairly broadband it gets called buzz.

Just wondering if trying to make something sound "analog" with an equalizer wouldn't do it-- If the ear is mostly missing certain types of subliminal hiss and/or hum? A clean equalizer couldn't boost hiss or hum if it wasn't in the original signal.

https://en.wikipedia.org/wiki/Bob_Carver
Carver caused a stir in the industry in the mid-1980s when he challenged two high-end audio magazines to give him any audio amplifier at any price, and he’d duplicate its sound in one of his lower cost (and usually much more powerful) designs. Two magazines accepted the challenge.

First, The Audio Critic chose a Mark Levinson ML-2 which Bob acoustically copied (transfer function duplication) and sold as his M1.5t amplifier (the “t” stood for transfer function modified).

In 1985, Stereophile magazine challenged Bob to copy a Conrad-Johnson Premier Five (the make and model was not named then, but revealed later) amplifier at their offices in New Mexico within 48 hours. The Conrad Johnson amplifiers were one of the most highly regarded amplifiers of the day, costing in excess of $6,000 a pair.

Of note that in both cases, the challenging amplifier could only be treated as a “black box” and could not even have its lid removed. Nevertheless, Carver, using null difference testing, (null difference testing consists of driving two different amplifiers with identical signal sources and exact levels, but out of phase by exactly 180 degrees. If the amplifiers were 100% identical, no sound would be heard. If sound was heard, the audio amps had different properties). Bob Carver used "distortion pots" to introduce amplifier characteristics, fine-tuned to null-out any sound differences. His "motel-room" modified amplifier sound was so similar, Stereophile Magazine editors could not tell the difference between his amplifier and one costing more than $6,000. This amplifier was marketed as the M1.0t for about $400.00. Bob Carver may have single-handedly debunked any number of theories about sound quality by using physics, blind and double-blind testing and unbiased measurements, such as "Gold-plated" speaker wires sound better than copper wires, etc.) Carver successfully copied the sound of the target amplifier and won the challenge. The Stereophile employees failed to pass a single blind test with their own equipment in their own listening room.

Indeed, music from DAW's tend to sound perfectly clear, which is not really pleasant, at least not for older people like me that grew up with cassettes and vinyl records
Not sure about hum and hiss. Some synths have saturation controls, which at times introduces hiss and other high-frequency artifacts, but I don't know any deliberate hum feature. Would that even be desirable?

Not sure about the white noise, isn't that something that tape simulation plugins add to entire mixes?

[hugoderwolf]

Phase is meaningless to ear unless it is modulated to create something that can be heard.

This is only true for the phase responses of filters. In this case it's about the oscillator phase when starting a voice, and that's a huge difference. Consider a simple sine oscillator and envelope attack at minimum. If the sine wave is near the zero crossing when the envelope opens, the start of the note is quite smooth. When it's at a maximum, you get a large step at the start of the note, giving the note a much stronger attack transient ("punch").

But usually, the init phase setting in synthesizers is independent of frequency, so the pitch shouldn't change the initial transient much. Could be that VCF cutoff and keytracking play a role here though.

[stratum]

Consider a simple sine oscillator and envelope attack at minimum. If the sine wave is near the zero crossing when the envelope opens, the start of the note is quite smooth. When it's at a maximum, you get a large step at the start of the note, giving the note a much stronger attack transient ("punch").

Well that's probably correct, and it's because the initial attack from a nonzero crossing produces overtones that are audible, not that anybody has heard the phase of that signal. Anybody who generated a stready sine wave at different phases would definitely say phase differences are inaudible, and given that it was audible in the above scenario, my guess is that it should be taken to indicate some level of interaction. One could consider envelope generation to be some kind of amplitude modulation, but I do not think it would be of any help to analyse the situation you have mentioned. That leaves the question, what kind of method would be proper and adequate? STFT result would probably different to begin with, but that would be consistent with the fact that the sound was different in each case anyway and does not give more insight.

[BertKoor]

I know there are several aspects developers try to emulate when trying to achieve analog sound in software. And one of them is equalization in my view.

Is this just a hunch / gut feeling, or do you have any evidence for that? Because in my view that's far-fetched bull dung. In the context of synths at least, could be different for channel strips, tape saturation and compressor simulations and such. The only thing you need to do is let the synth play a full sine sweep or white noise and then look at the spectrum. Ever done that?

For instance it seems to me that hardware has better bass.

What hardware exactly, and what qualifies it as "better" ??

I have also heard people say that hardware had a roll-off on high frequencies.

Quite right, usually above 100 kHz

[hugoderwolf]

@stratum: I was referring to this original question:

Also, why don't synths that have a phase control for oscillators, internally apply keyboard tracking to it?
A phase setting that sounds punchy on C3 sounds wishy-washy on C6 and vice versa.
I can set up a modulation manually, but it seems all but impossible to get the phase to start at the same position of the wave across the keyboard.

There's nothing too complicated about this. The Init Phase parameter has "keyboard tracking" conceptually built in. So if it's set to start at a waveform maximum at C3, it also starts at a waveform maximum at C6.

The reason that it sounds wishy-washy on C6 is probably that due to the high frequency the waveform itself changes quite fast, nearly as fast as the attack transient, and so the contrast between the transient and the static waveform isn't so strong.

By the way: the attack transient doesn't generate overtones, as it's just a one-time occurence. Overtones need periodicity. The attack transient is just a broadband click, similar to a single impulse.

[aciddose]

I have also heard people say that hardware had a roll-off on high frequencies.

Quite right, usually above 100 kHz

The cutoff is typically above 30kHz, with a 20dB/d slope.

decibels(x) = 20 * log10(x)
rc_gain(fc, hz) = fc / sqrt(fc^2 + hz^2)

I use a wrapper function:
rcg(fc, hz) = decibels(rc_gain(fc, hz))

Gain at 20 kHz:
30k = -1.597 dB
50k = -0.644 dB
100k = -0.17 dB
200k = -0.04 dB

rcf(k_ohms, u_farads) = 1 / (k_ohms * 1e-3 * u_farads * 2 * pi)

kilo (1e3) * micro (1e-6) = milli (1e-3)

Let's take the sh-101 as an example. See the bender schematic.

From VCA OUT (#5).
Output = 2.2 uF (C2) into 100k (R30)
Forms a high-pass filter with fc = 0.72 Hz. (Switch coefficients to compute high-pass gain with the rc_gain function.)

hp_freq = rgf(100, 2.2)
hp_gain = rcg(20, rcf(100, 2.2))
Gain at 20 Hz = -0.005 dB

2.2k (R29) into 1 nF (C12).
Forms a low-pass filter with fc = 72343 Hz.

lp_freq = rcf(2.2, 0.001)
lp_gain = rcg(rcf(2.2, 0.001), 20000)
Gain at 20 kHz = -0.32 dB.

[fluffy_little_something]

I am not sure we are talking about the same roll-off. I think there is one at about 16 kHz as well, i.e. well inside the audible range.

[aciddose]

No.

[JCJR]

Indeed, music from DAW's tend to sound perfectly clear, which is not really pleasant, at least not for older people like me that grew up with cassettes and vinyl records
Not sure about hum and hiss. Some synths have saturation controls, which at times introduces hiss and other high-frequency artifacts, but I don't know any deliberate hum feature. Would that even be desirable?

Not sure about the white noise, isn't that something that tape simulation plugins add to entire mixes?

Hi Fluffy

I don't know. Was just wondering if the perception of hiss and hum are sometimes user-desirable analog artifacts, missed in digital synthesis that is "too clean"? Even "very clean" old analog gear had some hiss and hum not difficult to perceive if you listen for it. Sometimes might have to turn up a little to clearly hear it, and often not necessary to turn up to hear it if you listen for it.

Distortion is often user-desirable. Lots of folks like distorted better than clean. In some cases folks will prefer the signal with "barely measurable" additional distortion, which "just sounds better" for whatever reason. And some folks seem to prefer quite a lot of added distortion. I've known a few fellas who would always turn up the stereo "just enough" so that it was annoyingly distorted to my ear, regardless whether it is a low-power radio, a medium-power boom box, or a loud stereo or sound system.

Long ago I worked with an audio tech in his repair shop for awhile and he would drive me nuts thataway. He was a great fella, a music lover with good taste in music. But whatever he was listening to, small or large system, he would always turn up "just enough" that it was annoyingly nasty, then he liked it. It must have been a preference for 5 or 10 percent distortion, because otherwise at least occasionally he would NOT have turned it up loud enough to distort. But he ALWAYS turned it up JUST LOUD ENOUGH to have easily perceived distortion.

So I was wondering, given that distortion is often user-desirable, then maybe a certain amount of hiss and hum can also be user-desirable?

I generally like playback "as clean as I can get it". Except for distorted guitar and occasional "slight tastefully distorted" bass. Or I tell myself that, though perhaps I like subtle distortion more than my self-image would allow. Dunno. I spent lots of effort over the years minimizing as best possible distortion, hiss and hum. But maybe a certain amount of hiss and hum sounds "better" for some folks than absolutely noise-free clean?

[fluffy_little_something]

Ai, distortion is not for me As soon as I hear a distorted guitar, adiós

I do use slight filter drive occasionally (like, 1 or 2 on a scale from 1 to 10), but as soon as there is distortion as such, it is too much for me.
With many synths distortion sounds ugly with chords, anyway. On some it is even periodic (as if it were controlled by an LFO), which also sound ugly and artificial.

Not sure why I like slight filter drive. Maybe because it adds some edge, sizzle, and liveliness. On some synths slight resonance does pretty much the same thing.

[stratum]

With many synths distortion sounds ugly with chords, anyway.

I guess they weren't properly EQ'ed. Putting a tanh(x) after a signal that contains a bass-heavy chord is not a very good idea, it's necessary to boost treble and reduce bass before applying distortion, as done in guitar amps.

[fluffy_little_something]

Not sure EQ is the problem.
I get the feeling it has more to do with the location of the distortion in the audio chain. On Waves' Element for instance one can switch between pre and post-filter mode, which seems to make a difference.


Anyway, a totally different question:
Good synths tend to use a lot of CPU power. Specifically, what is more important, higher integer or floating-point performance of the CPU?

[stratum]

Not sure EQ is the problem.
I get the feeling it has more to do with the location of the distortion in the audio chain. On Waves' Element for instance one can switch between pre and post-filter mode, which seems to make a difference.

Yes location is important, the signal path that tends to sound good with distortion is the one in guitar amps: Generate signal => reduce bass, boost treble => overdrive => tonestack (V curve, boost bass and treble, the notch is around 500hz-1khz) => lowpass filter around 5khz (cabinet). A guitar pickup behaves like a lowpass filter so the first EQ block (that 'reduce bass, boost treble' also compansates for that).

Specifically, what is more important, higher integer or floating-point performance of the CPU?

In the past floating point math was very slow compared to integer math but that was a long long time ago.