Korg Forums

Hello, I just found this blog post, through the vintage synth radias section
It was interesting so I though I share...according to this guy the radias might not be using real-time synthesis in few osc modes? any thoughts on this?

http://blog.zacharcher.com/2011/03/04/d ... rg-radias/

Well I would rather say that the basic saw, triangle and square waves are mathematically calculated (at least I did so when programmed DSP sounds) and the sine wave is interpolated from a table which is equivalent to a sample. Digital calculated saws will allways sound different to analog saw oscillators because digital ones may be "perfect" linear ramps where analog ones tend to be combed, which is what gives part of the sound character for each analog synth.

The sawtooth is definitely generated in realtime. Its not exactly computationally expensive.
I have a feeling that these sonic differences are somehow related to the anti-aliasing. Perhaps the oscillators are even actually discretely additive? That has always been one way to produce.

There is something about the radias, but I don't know what it is. I often find myself stacking up osc2 up or down an octave to obtain more fatness. And the drive does add a lot of character (not a bad thing) which can help a lot to bring a sound up to the front.

I kinda wanna call BS on that article...

It's true that generating a sawtooth is not computationally expensive, HOWEVER... Consider that Korg may have chosen a sawtooth wave that is not mathematically perfect.

If you sample sawtooth waves from analog synthesizers, they're not mathematically perfect -- the sawtooth slope will be slightly curved, or it will have quirky overtones, or distortion, etc... This is one reason why different synths have their own characters.

On the Radias, you can see that the overtones don't descend in a perfectly straight line, they have little peaks & valleys. This gives the sawtooth some coloration:



Korg may have constructed their own "sawtooth" tone, sampled it, and that's what the Radias uses.

I believe you are misintepreting the data and coming up with false conclusions. In your posting, you write:

"The Radias has 9 oscillator types, including a Formant oscillator. (A formant is similar to a bandpassed sawtooth wave.) Here's the spectrum of a formant produced by the Radias, with the narrowest formant width available. (This spectrum was analyzed in Ableton Live.) This looks acceptable, you can see the prominent peak around 800hz:"
....

"Although the sound is pleasant, this is not a true formant. This behavior indicates that the Radias' "formant" is actually composed of two waveforms. These waveforms have been pre-sampled at various resonant peaks. The Radias produces its "formant" by playing two of these samples simultaneously."

For example, I think you misunderstand the meaning of formants, and therefore the purpose of the formant width and how it's produced. You incorrectly think that it refers to the width of single bandpass filter, apparently. But the data seems to clearly indicate that its a two function formant, which is the minimum necessary for vowels but more are needed for more complex vocal sounds, and the width parameter determines the distance between the center points of the two filter functions, not as you conclude, to two different sampled waves that are mixed together. When you set the width to the minimum value, you see a single peak because both functions are set to the same centerpoint frequency. As you increase the width, the two peaks you see are exactly what you should see, as the two functions become apparent. Therefore, your assertion that it's not a "true formant" is not only incorrect, but proves that the function is more of a formant than you understand. Furthermore, the functions are filters affecting the spectral shape but not the carrier itself, while mixing together two sampled waveforms with different peaks would not be a filter and would change the carrier content.

The use of a 2D spectrogram is also a poor choice for analyzing formants. It would be far better to use a 3D graph to see the changes to the spectrum over time when changing the formant functions, for all the different vowel sounds, for example. The Radias editor shows formant motion data in 3D, for example. Seeing it in 3D might prevent errors due to misinterpreting a single 2D slice of the full formant over time.

zkarcher wrote:It's true that generating a sawtooth is not computationally expensive, HOWEVER... Consider that Korg may have chosen a sawtooth wave that is not mathematically perfect.

If you sample sawtooth waves from analog synthesizers, they're not mathematically perfect -- the sawtooth slope will be slightly curved, or it will have quirky overtones, or distortion, etc... This is one reason why different synths have their own characters.

Indeed, just check out how coloured the Virus classic sawtooth is.



You can clearly see the non-mathematically correct curved 'transient' at one pole of the so-called sawtooth, and also in the FFT scope the frequency response is non-linear, particularly at the higher freqs.

As the Virus is wholly digital the shape and colouration is clearly intentional, or a byproduct of Access initially trying to dodge aliasing.

The conclusions about the sawtooth wave also seem to be in error. That 2D chart is using dB for the vertical axis. dB is a logarithmic function. That makes the higher harmonics seem more important than they really are. Every 3dB drop is actually a 50% drop in the signal. Once you get out above the 20th harmonic, there's so little frequency content that any difference is insignificant - it can't be heard above the base frequency and the earlier harmonics. On a sawtooth wave you can use the harmonic number to determine the relative frequency content in the result. The first harmonic would have 1/2 the level, second 1/3rd, third 1/4, and so on. So, for a low note like 100 Hz sawtooth, the harmonic at 1,000hz would have 10% of the level of the primary. At 2,000 it would be 5%, 4,000 - 2.5%, 8,000 - 1.25%, 16,000 0.75%.

A more significant problem with the author is the following statement:

"The most important feature of the Radias -- which you will either love or hate -- is the character of its sound. My opinion is that anyone who describes the Radias as "deep" or "warm" is wrong, just plain wrong. The Radias is the opposite of warm, fuzzy analog. It is deeply cold, in an air-sanitizing sort of way. If iciness can be sexy, this is it.

The sound has a certain sonic distance, as if the phase of one speaker is inverted (but of course, it's not). The basic waveforms seem a bit sterile, even after summoning some analog trickery (detuning the oscillators, etc.) The sound is so smooth that it's practically transparent. This synth sits well in the background. With its bright character, I've been able to drop it into many mixes, and it magically sits well with minimal equalization. But my attempts to feature the Radias as the frontman have never felt satisfactory."

I think his testing is undertaken with the goal to support his preconceived (and wrong-headed, might I add) notions about it. I think he should learn it better.

Timo wrote: Indeed, just check out how coloured the Virus classic sawtooth is.

Nice! I think this is standard practice now, you "season" your waveforms because the mathematical models sound a bit unexciting?...

The sine wave on my V-Synth XT has upper harmonics, you can lowpass it and audibly change the timbre Heheheh

zkarcher wrote:....
....
On the Radias, you can see that the overtones don't descend in a perfectly straight line, they have little peaks & valleys. This gives the sawtooth some coloration:



Korg may have constructed their own "sawtooth" tone, sampled it, and that's what the Radias uses.

The first law of engineering measurement is: never trust a measure you haven't done by yourself!

The second law: Don't trust any measurement system until you have crossreferenced it with others and you are sure you have no side effects.

Since I supose that your Radias has no digital output (like all others, includin mine), you are ignoring the audio stages between the digital wave and the "measurement unit". Here you have at least 2 analog amplifier stages (Radias Line Out, DAW Line in) that also are not linear in frequency and have peak and valleys too. Unless you are able to digitize the direct digital values from the oscillators, all affirmations on this topic are no more than speculations.

By the other side if for example samples would be used for the basic waves, the programmers would have a really hard work to achieve PWM which mathemathically is very easy to implemt.

[EDIT]
From the Blog:
-- "Other features are likable, despite a few flaws: The 39 (!) knobs on the front panel don't rotate continuously." --

Do I understand this right or is the sentence a little confusing? Do you think it is a flaw that the knobs don't rotate endlessly like the value wheel (or Dial)?

If this is what you meant, what would be the use of 39 endless rotating knobs if they don't show you an estimated amount of each parameter value at a glance when programming? In that case 1 (!) endless (dial) knob and a bunch of select buttons would be sufficient for the same task which is what the majority of synths offer and where you get lost if your sight gets untied from the edit screen.

axxim wrote:Since I supose that your Radias has no digital output (like all others, includin mine), you are ignoring the audio stages between the digital wave and the "measurement unit". Here you have at least 2 analog amplifier stages (Radias Line Out, DAW Line in) that also are not linear in frequency and have peak and valleys too. Unless you are able to digitize the direct digital values from the oscillators, all affirmations on this topic are no more than speculations.

Sad but true. I cannot extract the digital output of the Radias. But, I'm still curious as to why the treble drops off when the waveforms are played with low notes.

My Radias is connected to analog inputs of an RME Fireface 800 with 3' cables, as are most of my other synths. (The Fireface converts the signal to digital audio, which is sent to my computer.) I haven't noticed any missing treble from the other synthesizers.

I agree that pulse width modulation is computationally cheap. But, if Korg wanted to "season" the output, it might be easier to use samples. I just recorded the square wave with pulse width set to 126 (almost the maximum), here's the spectrum of C3:



This looks (and sounds) good! Nice and bright. But at C0, the treble rolls off:



Here's one octave lower, and now there's some (very quiet) peaks (artifacts?) in the treble region:



Is it just me? Is it possible that my studio setup is scrubbing the treble out? I would rather have the high frequencies present, because I feel they are an important part of the sound. Sawtooth waves sound different when they have 32, or 64, or "all" (mathematically accurate sawtooth shape) of their harmonics...

I suspect that if I don't shut my trap, a Korg lynch mob will be pounding on my front door

[EDIT]:
Yes, "endlessly" was the word I was looking for! Me is not good writer. *fixed* Call me crazy, but I would prefer knobs that rotated endlessly, so that when you switched programs, the knobs would magically be positioned correctly.

To quote the article:
"The insert effects are impressive, but some of the best effects (the amp simulators) are restricted to monoaural output."

All guitars are mono.

I believe the analog waveforms on the Radias are generated in real-time because of the random phase location on a note-on command. If it were a sample the phase of the oscillator should technically play back at the same starting point on midi note-on every time. Sure the sample could be free-running and looped but that would mean there would need to be two gates. One for the amp envelope, and another to turn off the "free running sample" if you switch waveforms. Its simply impractical. Not to mention that free-running is not how the [truly] sampled pcm waveforms inside the Radias function. Why would the Korg R&D team waste time to make one free-running set of samples and another non free set? Its possible but just unlikely.

Axxim's point about the pwm or waveform modulation also seems enlightening. It would simply be easier to calculate the different oscillator modulations in real time rather than apply the effects of these modulations to a sample.

zkarcher wrote:Is it just me? Is it possible that my studio setup is scrubbing the treble out? I would rather have the high frequencies present, because I feel they are an important part of the sound. Sawtooth waves sound different when they have 32, or 64, or "all" (mathematically accurate sawtooth shape) of their harmonics...

It is a mystery as to why the treble seems to be tapered. I REALLY don't think its the Fireface... When you ran your test were you sure to put the filter on "Thru"?

If you need those last few harmonics you can boost the high end with the two band EQ or try increasing the resonance on a serial lowpass filter.

CharlesFerraro wrote:To quote the article:
"The insert effects are impressive, but some of the best effects (the amp simulators) are restricted to monoaural output."

All guitars are mono.

I believe the analog waveforms on the Radias are generated in real-time because of the random phase location on a note-on command. If it were a sample the phase of the oscillator should technically play back at the same starting point on midi note-on every time. Sure the sample could be free-running and looped but that would mean there would need to be two gates. One for the amp envelope, and another to turn off the "free running sample" if you switch waveforms. Its simply impractical. Not to mention that free-running is not how the [truly] sampled pcm waveforms inside the Radias function. Why would the Korg R&D team waste time to make one free-running set of samples and another non free set? Its possible but just unlikely.

Axxim's point about the pwm or waveform modulation also seems enlightening. It would simply be easier to calculate the different oscillator modulations in real time rather than apply the effects of these modulations to a sample.

zkarcher wrote:Is it just me? Is it possible that my studio setup is scrubbing the treble out? I would rather have the high frequencies present, because I feel they are an important part of the sound. Sawtooth waves sound different when they have 32, or 64, or "all" (mathematically accurate sawtooth shape) of their harmonics...

It is a mystery as to why the treble seems to be tapered. I REALLY don't think its the Fireface... When you ran your test were you sure to put the filter on "Thru"?

If you need those last few harmonics you can boost the high end with the two band EQ or try increasing the resonance on a serial lowpass filter.

The Radias is not a guitar The unison spread is stereo...

Switching waves could yield some interesting data, hmm... They could switch at the beginning of the waveform cycle. (The PCM waves are pretty short, it wouldn't have to wait long, I think?) And/or you could crossfade the wave samples when the change occurs.

[EDIT]:
Also, the filter was indeed on THRU when I captured all these spectrums.

Hey, Timo, could you put a steep 24dB low pass filter on that Virus sawtooth, then sweep it from 20Khz until you start hearing a significant effect on the sound you hear? A snapshot at that point will help to show when the upper harmonics don't matter because they can't be heard.

Sometimes pretty pictures are meaningless compared to what you actually can hear. The primary note of low bass sawtooth will tend to dominate the upper harmonics. There's just a lot more sonic energy at the lower frequencies than the higher ones.

pheeew...! and i started to believe that korg tricked me heh. Well thank you guys, some really interesting knowledge exchange happening here.

a bit of a side topic but zkarcher, I agree with you on the endless dials...I find that the normal potentiometers are a bit of a limitation when it comes to live performance, especially when you need to switch back and forth between sounds, etc. . . But then the radias is a tricky little beast, with a lot of "hidden" qualities.. and it makes up for many limits or the initial "cold clear" sound