The old but never satisfyingly answered question

[Spogg]

Interesting Martin.

In my mind a morph is a change from a starting waveform to a destination waveform. A linear cross-fade would be the simplest morph.
But of course you could travel through any number of wave shapes on the way. One example is wave folding, where the intermediate waveforms are not a mix of the start and finish shapes. But is that strictly morphing?

Maybe this is all about semantics; what we mean by the term.

Cheers

Spogg

[martinvicanek]

Yes, you could view crossfading as a special form of morphing. Although in film morphing usually involves some deformation so the caracteristic features of the source move toward the target characteristic features.

[trogluddite]

Maybe this is all about semantics

Sensory perception plays a big part too, I think. In my earlier example, I showed how two detuned oscillators can sound the same as a single oscillator being amplitude modulated, and vice versa. But if the separation in frequency is big enough (or, equivalently, the modulation frequency is high enough), we do perceive two separate tones. I don't know of any experiments into where the point lies at which our "description" would change, nor how consistent it is between different listeners; but my hunch is that it's where the partial frequencies are separated far enough to fall into different "bins" in an internal, biological/neural "FFT analyser" (the spiral of the cochlea in the inner ear acts as a kind of physical frequency separator, so it may not be entirely down to the brain).

Two sine waves is only the very simplest case, of course - more commonly we're working with waveforms where there are a multitude of partials interacting, so it's maybe not surprising that there are effects which are difficult to describe unambiguously, and even some we might call "auditory illusions" (e.g. rising/falling shepard tones). This is a big part of my fascination for DSP; even processes which are well known and mathematically simple can produce interesting perceptual effects if you just happen to throw the right combination together.

In the case of tula's thought experiment: if the modulation speed were high enough, and contained suitable harmonic content (and was aliasing-free!), it might well be perceived by most people as a kind of morphing rather than AM/cross-fading - in the same way that FM synths don't usually sound like a bunch of discrete oscillators with lots of vibrato applied. Whether "morphing" should be considered an accurate description in a technical sense is an open question, I think - personally, I lean more towards Martin's definition (PWM being possibly the simplest form), but I don't think it's written in stone.

(PS: Thanks for the lovely welcomes back - though I'm not sure that living in a "Yellow submarine" would be too pleasant in the dingy, peat-stained, industrially-polluted waters of Bradford Beck! I may well have been in some other place where "the sun doesn't shine" [metaphorically or otherwise], however! )

[trogluddite]

Just re-reading the thread, and this got me thinking...

But if you do the same blending on a per sample basis...

Consider the case where both oscillators produce exactly the same wave-form...

If we conceptualise it as "summing per sample", it's easy to see that the amount terms would cancel out, leaving only (1.0 * oscillator), no matter how madly we modulate the "amount".

But what if we conceptualise it as "two waveforms being mixed"? We now have two amplitude modulated oscillators which will produce additional frequency components, yet we still don't hear any. The answer is that they are still there; which listening to either one in isolation would easily prove. However, the amount components of the modulators are inverted between the two terms, so the added partials are also inverted and thus cancel each other out (even if they're aliased). The original signal is still heard because there's a modulation component at zero Hz (the DC offset of the unipolar modulators), where (fo + fa) and (fo - fa) both equal fo. So the result is exactly the same even though, at first glance, this might seem to be a different kind of processing.

It's a nice little example of how the time-domain and frequency-domain ways of looking at audio are totally equivalent perspectives on the same phenomenon.

So which of these equivalent "models" is most intuitive for the end-user may only be a matter of the subjective qualities of the output and whatever hints we might give in the user-interface design - they might even be more inclined to hear it as "morphing" just because we added a control labelled "morph"!

[k brown]

I certainly don't follow the maths, etc. being discussed, but there certainly is a difference between crossfading and morphing waveforms and it's VERY audible. Take the example of sine into saw. With a crossfade the top harmonics of the saw become audible immediately - just at low level. With a morph (such as phase modulating a sine into a saw; along with the feedback required), the mid-point of the morph is much softer-sounding (and is it's own unique waveform; not just a mix of two) than the mod-point of a crossfade - just look at the two cases on a scope; they look nothing alike and they sound nothing alike. I consider a morph where the shape changes in a continuous way (as Martin described re- visual morphs), rather than simply combining two waveforms at varying levels, which is quite visible on a scope.

Another example is the Tri Saw wavetable morph someone came up with way back in SM days (and I think Spogg did a simpler/leaner version of). The mid point looks and sounds very different from a crossfade from a tri to a saw.

Likewise, Martin's WaveShape and Spread Partials oscillators I've been playing with - true seamless morphing from sine to complex shapes.

[adamszabo]

Here is the difference between Morph and Crossfade. I think its pretty self explanatory
Both change from Saw to Triangle but in different ways.

[k brown]

Bingo !

And that TriSaw morph at about 1/3 of the way toward saw is one of my favorite sounds; just can't be had by crossfade or filtering a saw.

[tulamide]

So which of these equivalent "models" is most intuitive for the end-user may only be a matter of the subjective qualities of the output and whatever hints we might give in the user-interface design - they might even be more inclined to hear it as "morphing" just because we added a control labelled "morph"!

I'm afraid, this is the ultimate answer to the term confusion!

Here is the difference between Morph and Crossfade. I think its pretty self explanatory
Both change from Saw to Triangle but in different ways.

YES! It is self-explanatory. And the reason why I keep being confused. My main orientation, through my job and private interests, has been visuals for almost 20 years, before I started VST programming. And in visuals, the morph example you gave is also called morrph, while the second is a blend. So that's the concept I'm used to. For VSTs I tried to understand, why the second example is called morph instead, and just couldn't get it.

At least I now see, that I'm not alone with my concept of morphing as I knew it from visuals. Martin and Adam have the same concept, and trog just gave the ultimate answer: a big knob labelled "morph" introduced my confusion.

It may be that this time it's actually satisfyingly answered.

[RJHollins]

Wait until someone brings up 'crossfade'.

[Spogg]

A very illuminating discussion guys!

Dare we conclude that cross-fading is type of morphing, or should I duck for cover?

Cheers

Spogg