Falcon Synth Oscillators – a Close-up Look
UVI’s Falcon hybrid software synth has deep proficiencies as both a sample player and a synthesizer. We examine the capabilities in the latter category herein.
by David Baer, March 2016
SoundBytes Magazine has been giving a lot of coverage lately to UVI’s new software instrument Falcon … and for very good reason. If offers a dazzling array of sound creation capabilities, both as a sample player and as a synthesizer (and, of course, both of those things at the same time if such is your goal). In this article, we are going to engage in a detailed examination of the synth oscillators.
Before we dive right in, it’s appropriate to spend a little time reviewing the architectural structure of Falcon in which the oscillators operate. The oscillators exist in a four-layered configuration: a Program is at the top. Programs contain Layers, Layers contain Key Groups, and Key Groups contain oscillators. Actually there’s an even higher layer, the Multi, which contains multiple programs, but that’s not relevant to this discussion.
Significantly, the Program has a specification for the basic MIDI note mapping. Most of the oscillators do not even have control over this other than a fine tune capability in some cases. The Layer is where Unison operation can be specified: multiple, slightly detuned copies of the underlying sound for rich-sounding timbres. But two of the oscillator types also have a “local” unison feature.
The Oscillator level houses one or more individual oscillators (which can be of different types). It has a shared control for course tuning (semitones) and fine tuning (cents). Then there is the triggering mode, an unusual feature for a synth (as opposed to a sample player) in that oscillators can be invoked in a round-robin fashion.
All the levels have their own gain and pan controls. There you have it.
There are eight oscillator types that can be called into service: basic Analog, Analog Stack, Drum, FM, Noise, Organ, Pluck and Wavetable. We will now proceed to examine each individually.
We start with the easiest of the oscillator types, since there’s really nothing here that most readers will not have seen before … seen before a lot, in fact. There’s a single waveform, the choices being: saw, square, triangle, sine, noise, and PWM (pulse width with modulation of wave shape possible). A silent oscillator is on board that can be used for Sync operation. A second level of Unison operation can be set up here as well. There is no provision for supplying a custom waveform – for that one would use the Wavetable oscillator, about which much more will be found shortly.
The PWM control affects pulse width. It is applicable to all wave types. Below you can see the effect upon a sine wave at 10%, 50% (neutral) and 90% settings.
I think a far more satisfying sound-design experience can be achieved with the Analog Stack oscillator. Here we have most of the capabilities of the basic Analog oscillator, minus the Unison capability. But we have eight oscillators, each with individual tuning, gain and pan. Oscillators 2 through 8 can be synced to oscillator 1 as the master.
Even a four-oscillator sound like that defined in the screen shot above can produce a lovely, rich sound. Generally when stacking multiple waves, sticking with square and triangle waves is rewarding, since they have only odd harmonics. The sound in the screen shot could have been improved upon further by duplicating each oscillator, hard panning the duplicates left and right, and making slight detuning tweaks. And, of course, all the controls are modulation targets, so further animation of the sound is possible. In all, there’s nothing new here, but this is fertile ground for excellent sound creation possibilities.
I must confess to having little interest in synthetic drum and percussion sound creation. So the reader will hopefully forgive me for not seeming all that excited about this one – although it should be said that this is a great source of general FX sound possibilities as the factory presets amply demonstrate.
We have two independent oscillators on board, one a waveform and the other noise. The oscillator can hold a saw, pulse, triangle or sine wave, but frequency is fixed in the oscillator and is not key-tracked. The pitch can be modulated with the pitch modulation controls. The effect of this is hard to describe, so some hands-on experimentation would be fruitful. The envelope is a simple attack/decay shape.
The noise unit is a little more complicated. We have a resonant LP filter and another attack/decay envelope (operating on amplitude, not filter cutoff). Velocity can independently control the levels of the wave oscillator amplitude, the depth of pitch modulation and/or noise amplitude.
Finally, we have several controls that affect the combined output of the two component oscillators. There is an EQ feature with frequency and level. This offers a single peak/notch band with a medium Q and a range of -40 dB to +40 dB. There is an onboard distortion effect, about which the documentation tells us nothing other than its name. It’s doesn’t seem to lay it on too thick and for the noise source, it basically just makes the noise portion of the waveform/noise mix louder. Lastly, there is a bidirectional volume control. Again, the purpose of this is not revealed in the documentation (which is mostly adequate if not quite good most of the time). Since this is a modulation target, that alone may justify this control being present, but then, you would think we’d see this control in all the other oscillator types, and we do not.
The FM oscillator brings us a four-operator FM capability, with eleven topologies (or algorithms in DX7-speak). The topologies on tap are shown below.
Although we have only four oscillators with which to work, we could easily create many of the DX7 six-operator patches by employing two Falcon FM oscillators and dividing the work between them. But this won’t work in all cases. Consider the representative DX7 algorithms in the following graphic.
Algorithm numbers 3, 7 and 25 could be easily duplicated in Falcon. For number 4, the feedback routing is not in the Falcon FM repertoire. The same is true for numbers 8 and 26. Nevertheless, the Falcon FM capability is plenty powerful and should satisfy most FM sound designer’s needs, so I don’t mean to imply that this is a serious shortcoming.
Each operator has a toggle labeled HZ. This changes the frequency from a relative ratio to an explicit frequency. If in ratio mode, then Snap locks the ratio to exact harmonic intervals. Level is just what it says. And remember, all of these knobs can be modulation targets.
Boy have we got noise in Falcon! Just check out the choices in the screen image below.
There are fifteen options, eight of which have an additional control. For example, the Band option is white noise run through a band filter, and the associated control is Bandwidth. The center point frequency of that filter is key-tracked at 100%. For a sufficiently narrow bandwidth, one can hear a ghostly melody when playing a sequence of notes on a MIDI controller. Of course, many of the other noise types could produce this same effect with an external, key-tracked band filter appropriately applied.
The other controls are as follows: Sample and Hold provides Rate, Crackle provides Color, Logistics provides Chaos, and Static I, Static II, Dust and Velvet each provide Density.
So what are all these noises like? Many are familiar variations on filtered white noise – that is the “colored” noises. The image below shows the frequency spectrums of these. Amusingly, Brown noise is not named for a color but a 19th century Scottish botanist, but Brown noise is also sometimes known as red noise. Got that? 😀
Now, it could be pointed out that many of the noise types on tap could be produced with white noise and suitable EQ and/or filtering, but where’s the fun in that?
Lorenz and Rossler noise may not be familiar to you. No surprise, since an explanation of either would require some rather intimidating mathematical heavy-lifting. Crackle has a controllable Color property. The Lorenz, Rossler and two variations of Crackle spectrums are seen in the following image.
The Organ oscillator provides a straightforward Hammond-Organ-like set of “drawbar” controls that work just like the real thing. The 8’ drawbar produces a sine wave at concert pitch. 16’ and 5⅓’ provide sine waves that are one-half and three-halves of concert pitch respectively. If the note played on the keyboard is middle C, the remaining drawbars correspond to pitches that are respectively (and approximately): the octave above middle C, the G above that, the next C above that, the E above that, the G above that and the C above that … just like the real thing.
A Hammond-like percussion option completes the story. That’s really all there is to it. Nice organ sounds from a not-very-complicated interface.
Now we get to by far the most complex oscillator to describe, the Pluck oscillator. As you might guess, it excels at pluck-type sounds, but the factory presets surprisingly contain some bowed-string-type sounds as well. The Falcon documentation leaves a bit to be desired for this oscillator type, so we’ll try to make up some lost ground here.
Before we go any further, though, we need to take a detour back to 1983, at which time a paper was published in the Computer Music Journal that described a very clever way to produce reasonably realistic synthetic plucked-string sounds. The authors were Kevin Karplus from Cornell and Alex Strong from Stanford. The Falcon documentation does briefly mention Karplus-Strong, and it’s helpful to understand this background.
The approach that the research team discovered is actually pretty straightforward. Start with a wavetable initialized with random noise (here “wavetable” takes the historical meaning of an array of data holding sample values describing a single waveform cycle). Repeatedly “play” this waveform, replacing the current sample value with the sample value of one cycle previous averaged with its preceding neighbor. We start with noise, but due to the averaging, we rapidly converge on a consonant timbre. The repeated averaging has the added benefit of applying a natural decay to the sound in which higher partials decay faster than lower ones.
This work was done at the dawn of the personal computer era, and the processor speeds were appallingly slow by today’s standards. Because only two numbers were involved in the averaging operation, a division could be avoided, being replaced by a much, much faster right-shift operation. This was actually a pretty big deal given how inefficient division operations where on the computer chips that were state-of-the-art in the early 1980s. The paper speaks of being able to produce maximum playback speeds of 20KHz on the Intel 8080 PC-flagship chip at that time. How very, very far we have come in the intervening third of a century!
But back to the present. Below is the screen image of the Pluck oscillator. The Falcon solution uses Karplus-Strong as a starting point, but takes it considerably further from there.
For one thing, there are three sources of sound excitation: Sample, Synth and Noise. They can be used in solo fashion or mixed together however desired. Let’s start with noise and work our way left. Noise is just that: random sample values. This is where the Karplus-Strong approach is employed and it’s surprising at how good this technique sounds. Look at the three wave displays below. We see images at just after note-on, 200 ms later, and finally one second later. Even in the first few dozens of milliseconds, the averaging operations can be seen to profoundly tame the noise into a more dulcet tone.
There’s more in this case. There are a number of string modeling processes that can be optionally applied like specifying a pick position – plucking near the middle of any string is going to produce a mellower sound that plucking near a termination point.
I will not go into every parameter here. There are quite a few and these are covered in reasonable detail in the documentation. I will point out something that puzzled me for a bit. I could not get the Decay Release control to have any effect. It turned out that there’s an amp envelope also present that trumps the pluck decay parameter. Silly me! 😀
The plot potentially thickens in that a second string can be brought into the picture that is excited by the first (shades of AAS Chromophone!). Learning to program sounds for this oscillator will take some dedicated experimentation. Some tricks are far from obvious, but there are some good factory presets to show you the way.
But back to the excitation sources – let’s look next at the Synth option. Here we start with a waveform that is governed by the Brightness setting (which has no effect on Noise or Sample sources). At minimal brightness, we have what appears to be basically sinusoidal. The image to the right shows the waveforms at respective brightness levels of 0%, 10%, 50% and 100%). This waveform shape appears to stay constant through the decay phase and only the amplitude diminishes in my testing. By itself, Synth is pretty tame, but nicely combines with Noise to mellow out the final product.
Lastly there is the sample excitation source. This can be either a single-cycle waveform or an extended sample. The factory content makes use of both types. You can supply your own wave file via drag and drop. Just don’t expect it to sound like what you’d expect. I tried using a piano sample, for one thing, and it ended up sounding nothing like the original once it had traversed the pluck signal processing path.
All in all, this is a very capable oscillator that does especially well on its namesake sounds. Pluck is an ambitious addition to the feature set. It does not make me contemplate abandoning my excellent collection of AAS physical modeling instruments by any means. But as an additional element in the sound palette, the pluck oscillator is a welcome and somewhat unexpected bonus.
Lastly we arrive at the ever-so-capable Wavetable oscillator type. By “wavetable” here, we apply the more current meaning of the term implying a capability to morph between multiple single-cycle waveforms during playback. But as you’ll see, the morphing is only one of the two dimensions of waveform alteration available.
The factory wave content includes both single and multi-wave offerings. What good is a wavetable with just one waveform, you ask? The Phase Dist control answers that question and we’ll get to that shortly. You can see the factory options in the Single/Sawtooth category below. Eight other single categories include Square, Triangle, Parabolic, etc.
As to multi-wave material, there’s a fair amount of factory content included. Some of the factory-supplied offerings in the Multi/Analog category appear below.
User-supplied waves can be dropped on to the wave window, either single or multi-wave files, wherein the “slices” are just concatenated together to form a multi-cycle sample file. For a multi-cycle file, the file name must end with an underscore followed by the number of samples in each waveform – this is contrary to the instructions in the current documentation, which is in error on this point.
Images can be dropped onto the file window for conversion into audio data – there appears to be an undocumented maximum size for this to work, however.
When working with multi-waves, the Index control, which begs to be modulated with an LFO or envelope, can be used to traverse the individual waveforms. A smoothing option would normally be engaged to make a seamless morph happen. Otherwise an abrupt transition occurs, which might occasionally be preferred for aggressive sound designs.
The other dimension of sound alteration is Phase Distortion. This does things to the linear (time-dimension) spread of the waveform. The image below shows the effect various distortion types have on a basic triangle wave. Here a picture truly is more useful than a thousand words. The Bend+/- option is not shown because it just combings Bend+ and Bend- into one composite operation.
Finally, the right-hand area of the interface provides a unison capability. What’s particularly interesting in this case is the ability to use multiple waveforms in a multi-wave sample set to develop an interesting stereo image.
What more is there to say? Falcon has most impressive synth sound generation capabilities that are well worth taking the time to explore and master. And remember, we’ve only been looking at the sources of synthetic sound. It goes without saying that the modulation possibilities and arsenal of FX modules just makes the whole story all that more impressive.