Questions about PCM tones

[AlenK]

Sure, the XW-P1 is first and foremost a synthesizer. So the PCM tones seem to get little respect. But I sure would like to know more about them. I have questions like: Are any other tones besides some of the pianos and percussion sounds velocity switched? I understand from Mike Martin's past posts that the stereo piano is "triple strike"' meaning I presume that it is based on samples at 3 different velocities. It would be nice to know which tones are similarly velocity switched and how many velocity zones each uses.

Questions like: How are the PCM waves otherwise structured and used by the PCM tones? For example, many of the PCM waves listed in the appendix for the user manual refer to a footnote that says "With these waves one type of wave is allocated to the entire tone range. They cannot be selected by the hex layer tone." Seemingly they can only be selected (by users creating new tones) in the solo synth. But I'm sure they aren't there for that purpose. I'm sure they are used by the PCM tones in some way, but exactly how? For example, there are some 150 piano waves starting at 0799 in the list. How are these used by the PCM tones? I'm going to guess that they include all the multisamples over the piano pitch range and velocities, but exactly how are they organized? It seems curious that all these waves have names but we're not told how the XW-P1 uses them.

Call it insatiable curiosity but I'd like to know.

[AlenK]

Okay, maybe my questions above were too general to be easily answered. Let's start with some simpler ones and build up an understanding of what's going on wth PCM waves from there.

First question: What is the significance of "-L" and "-B" when they are appended to the names of waves? For example, in the synth wave list there is a "PWM" wave, a "PWM-L" wave and a "PWM-B" wave. In fact, most waves in the synth waves list have -L and -B variants (all except for the "VA" waves and a few others). Many waves in the PCM wave list also use the -B and -L suffixes.

[Mike Martin]

L - means low

B - bass

This is done to minimize aliasing artifacts.  So if you're doing a synth bass sound you'd want to use the L or B versions of the waveforms to get the best quality....that is unless you're looking for a particular character. 

[AlenK]

Very helpful. Thanks. If I understand you the harmonic content of the L and B sounds doesn't go as high as the "normal" samples, hence less aliasing. Is that correct? If so I think the L waveforms might be useful for non-bass sounds as well, as long as you don't need to go very high up the pitch scale (which I find I usually don't). The "character" that aliasing gives to a voice isn't the kind of character I'm looking for. I usually want "smooth and silky," which precious few of the presets achieve. ("Screaming" leads aren't my cup of tea.)    

 

Second question (I implied there would be more):  In the list of PCM waves, some names have -1 and -2 variants. Can we assume that like many of the piano waves, these are based on samples at different velocities or intensities? Third, related, question: What does -A and -B signify in the PCM wave list (e.g., Pipe Organ-A, Pipe Organ-B )? 

[Mike Martin]

Yes usually things like the pianos and EPs have multiple variants, these are generally for velocity switches.   Some "Tones" are combinations of waveforms so Pipe Organ A and Pipe Organ B may have been created originally to be used together.  Some of these samples were originally created for products that didn't have realtime filters of any kind so you'll even find variants that have a different tonal character. 

[AlenK]

L - means low

B - bass

This is done to minimize aliasing artifacts. So if you're doing a synth bass sound you'd want to use the L or B versions of the waveforms to get the best quality....that is unless you're looking for a particular character.

After experimenting I must report that I did not find this to be the case, at least not on the waveforms I tried (sawtooth and PWM with a setting of zero for the pulse width, resulting in a square wave). On those waveforms the -L and -B variants had progressively MORE harmonics with progressively more aliasing at any given frequency. I would guess that if the "waves" are multi-sampled to control aliasing (which perhaps was not done all that well, as least for the most harmonically rich ones) then the -L and -B variants shift the "root" waveform down by several octaves. That is, the waveform that would normally be played at, say, C4 is now played at some lower C. What this would do (without getting into a long technical discussion) is cause the waveform to contain "corners" or edges that generate much higher harmonics. These edges, it turns out, are what gave bass notes from the famous PPG Wave synthesizers such an interesting sound. So if you want that "edgy" sound for bass notes (and many people do!) use the -B version. If you want clean(er), use the regular version.

[AlenK]

After further listening I have to amend what I wrote above. I had guessed (assumed) that the synth waves were multisampled like a good number of the PCM waves. In other words, different waveforms or samples across different ranges of pitches. That's one common way to bandlimit a waveform to ensure no aliasing and has been used since the 80's in digital and hybrid (digital/analog) synths, as well as in some virtual-analog synths today. Given a synth design that plays back waveforms from memory (which is what is happening here after all) this is easy to do.

 

But I guessed wrong. As far as I can tell most of the synth waves aren't multi-sampled. They just play the same waveform for every key you press. The "normal" waveforms (i.e., not the -B or -L waveforms). contain just enough harmonics in most cases to not alias too obviously at the highest non-transposed pitch on the keyboard (C7). The -L version of a given waveform contains more harmonics and plays an octave lower. If you transpose it up an octave it aliases much more than the "normal" waveform. The -B waveform contains even more harmonics than the -L version. It only plays up to a middle frequency (C5) and can't be transposed up (only down).

So now I finally understand what Mike meant by the -B and -L waveforms being there to minimize aliasing. The -B and -L waveforms don't actually do anything themselves to minimize aliasing. Rather, NOT using them for notes higher than C5 will minimize aliasing. If you use them for a bass tone you get a brighter, edgier sound. If you don't use them the sound will be comparatively dull.

So how can you tell if you've got aliasing? One easy way is to pitch bend a note quickly back and forth. If you hear anything other than the tone rising and falling in frequency with the bend, such as a warbling or whistling sound or you hear frequencies that move in a direction opposite to your bend, you have aliasing.

In my opinion aliasing is never a good thing, even if you want your lead sound to "scream" (and according to Casio's ads, that's what lead sounds do ). Certainly, if you want an authentic analog sound you have to avoid it. Analog synths don't alias. Some of the XW's synth waves alias less than others. For instance, many of the "normal" CZ waves seem to be okay whereas the "normal" PWM wave and the "normal" sawtooth wave alias very badly. Some of the "normal" CZ waves are among the few that appear to be multisampled, the waveforms for higher notes actually having weaker high harmonics. I suppose Casio was motivated to make sure that at least these waveforms sounded the best they could.

[Jokeyman123]

Hello again AlenK-based on your info, apparently Casio needed to do some "stretching"-using one sample across a longer range of notes which allows for less internal memory needed in the sound rom, but sacrifices some depth in the acoustic instruments. Having done some sample creation, I can say that Casio seems pretty clever in that most simple sines, sawtooths etc. can be stretched without sacrificing to much quality (aliasing excepted) and use minimal rom memory and it shows, the synth sounds in the XW-P1 are real nice IMO. Not so with PCM samples which even with the best programming "ears" and skills still need multi-samples to sound right across even a limited span of 2-3 octaves. I have played with soundfonts in software and with keyboard samplers and have gotten pretty good results with relatively little memory. (Korg DSS-1, GEM SK-76, Ensoniq TS-12). That is where the XW is a little weak-alot of the acoustic instruments are OK flat-out but in a 500 dollar instrument I didn't expect a state-of the-art sample library! The fun is-think about mutating a clarinet, ep, trumpet or even an organ wave into a beastly synth sound-or putting one into a hex layer with even 1-2 hairy synth sounds, and on the fly-how cool is that!

[Display Name]

 The fun is-think about mutating a clarinet, ep, trumpet or even an organ wave into a beastly synth sound-or putting one into a hex layer with even 1-2 hairy synth sounds, and on the fly-how cool is that!

 

 

That's kind of the approach I take.. figure out the most creative way for to use an instrument... Aliasing,  noise, distortion, etc. can all be used musically in the right application.