Question for Nord Piano and/or Nord Stage owners regarding the acoustic piano sounds

[zauberklang]

Hi everyone!

I'm trying to find out if Clavia has developed Sample Interpolation Technology for their piano sounds. So my question for Nord owners is: are the velocity layers of the acoustic piano sounds (at least on the newer models) smoothly blended (meaning inaudible)? I'm not planning on buying a Nord Piano anytime soon; it's just the geek in me who wants to know. 🙂

I know that Yamaha, Kawai and Synthogy have developed this very advanced Sample Interpolation Technology for absolutely smooth piano sounds. Roland doesn't use samples for their piano sounds anymore, only physical modeling. I know that at Synthogy it took two highly qualified employees three years to develop this technology! This explains why it is still so rarely used.

My apologies if this topic has been covered before.

[Macsaint777]

Nord piano library samples do use multiple velocity layers, and it is very very difficult to hear any sample switching- so whatever they are doing, it’s done extremely well. 

[cphollis]

^^^^ What he said.

 

I've been using sampled digital pianos for many years, but have never heard that term.  It sounds like something a marketing department would come up with.

 

As above, if anyone could detect sample switching with Nord (or any other DP), it would be unpleasant, you'd hear about it, etc.  If you're still curious, head over to the Nord User Forum and post over there?

[stoken6]

I've played cheaper or older DPs where the switch from one velocity layer was very evident. (It's more detectable on Rhodes samples, because the instrument's timbre changes so dramatically with velocity). But I've never detected anything jarring in Nord's implementation of velocity layers.

 

(Consistent amplitude across every key in the instrument's range, now that's another matter).

 

Cheers, Mike.

[zauberklang]

14 hours ago, cphollis said:

^^^^ What he said.

 

I've been using sampled digital pianos for many years, but have never heard that term.  It sounds like something a marketing department would come up with.

 

As above, if anyone could detect sample switching with Nord (or any other DP), it would be unpleasant, you'd hear about it, etc.  If you're still curious, head over to the Nord User Forum and post over there?

 

This technology is real. Synthogy calls it Sample Interpolation Technology used for ultrasmooth velocity and note transitions https://www.synthogy.com/index.php/products/software-products/ivory-2-studio-grands (see Highlights)

Yamaha first used it on the CP1, I believe. Kawai used it on the MP10 for sure (I owned one). They don't mention it in marketing texts anymore. (I'm not sure they ever did.) People who buy digital pianos from the big three manufacturers (Yamaha, Kawai, Roland) take it for granted these days.

[cphollis]

Thanks for that link.  

 

The geek in me wonders why anyone would create a trade name for it.  The math associated with smoothly interpolating two sample waveforms is well understood AFAIK, and there are lots of different and good ways of doing it.  Perhaps they thought they came up with a somewhat "better" way (more phases or steps in the interpolation process), and realized it wasn't such a big deal.  I can't detect any form of velocity switching on my Nord, nor my Yamaha, etc.

 

 

[zauberklang]

Cphollis, may I ask what Nord and Yamaha models you own? I'm interested in knowning when Clavia (and Yamaha) began do implement the technology.

The math associated with smoothly interpolating two sample waveforms may be well understood, but it still took two specialists at Synthogy 3 years (!) to program it for sampled piano waveforms. (This information is from a reliable source that knows the company.) Kurzweil and Korg do not (yet) have this technology AFAIK. That's one of the reasons they don't sell as many home digital pianos as Yamaha and Kawai. The standard piano sounds on the Kurzweil Forte and the Korg Kronos have 8 clearly audible velocity layers. They do use filters to make the velocity switching less jarring, but that does not make them smooth. Users of synthesizer workstations don't mind the audible velocity switching on the piano sounds as much as users of home digital pianos do. The main piano sound on the Yamaha Montage has 8 layers that are not smoothly blended like on a Yamaha stage or home piano AFAIK.

[Jon E]

3 hours ago, zauberklang said:

 

This technology is real. Synthogy calls it Sample Interpolation Technology used for ultrasmooth velocity and note transitions https://www.synthogy.com/index.php/products/software-products/ivory-2-studio-grands (see Highlights)

Yamaha first used it on the CP1, I believe.

Yamaha used SCM (Spectral Component Modeling) in the CP1. To me this seems to be a technology to manipulate the samples to sculpt the characteristics of pianos and epianos. A description of this can be found here: http://sandsoftwaresound.net/spectral-component-modeling/

 

 

[drohm]

"The math associated with smoothly interpolating two sample waveforms is well understood..."

That statement is a major oversimplification of the actual problem or application, imo.  From experience, I know that doing this properly highly depends on the application (Audio, RF, Imagery, Lidar, etc...real-time, non real-time.). In the case of piano sampling and sounds for real-time performance, I can imagine this being very challenging to generalize. Although I have spent my career in digital signal processing, I'm also now more curious to the methods being used by different manufacturers. Had not thought about it prior to this thread. 

[AnotherScott]

Casio calls it "Linear Morphing," saying it "creates smooth transitions between the softer and louder sounds." They also use the term "Multi-dimensional Morphing" which they say "smoothly expresses changes in volume and tone characteristics with variations in touch and passage of time" -- which sounds like linear morphing for the touch part (how the sound changes with velocity), plus the "added dimension" of also doing something similar for how the sound changes over time.

 

32 minutes ago, cphollis said:

The geek in me wonders why anyone would create a trade name for it. 

Marketing. ;-) Really, since not every board does it, if you want to boast about how your board does, well, you have to call it something.

 

On 5/5/2022 at 6:44 AM, zauberklang said:

Roland doesn't use samples for their piano sounds anymore, only physical modeling.

They still use both. Originally, V-Piano was their modeled piano, while SuperNATURAL added modeling to samples, in part, to do exactly what you're talking about, eliminating audible velocity shifts. This clear distinction then got blurred, with something called  "SuperNATURAL Piano Modeling" (FP-90/90X and LX-17) which is completely modeled (and can also be distinguished from the "regular" SuperNATURAL pianos in that they have unlimited polyphony).  Then there are the LX700 console models where they now have something called "Pure Acoustic Piano Modeling" which I believe is also completely modeled. But Roland is still using both approaches. Pretty recently, they released as new SuperNATURAL Piano (see https://www.roland.com/us/products/rc_supernatural_acoustic_piano_2/ ) which appears to be another implementation of the "traditional" SuperNATURAL (aka SuperNATURAL Acoustic).

 

Getting back to the original question, AFAIK, Nord does not do any sample interpolation. I don't think Yamaha does either, except possibly in the SCM pianos which I don't believe are available in any current product. And I don't think Korg or Kurzweil do it either. I think Kawai does... they don't have a name for it, but they use the phrase "Harmonic Imaging" as a kind of catch-all for their various underlying technologies, without telling you what they are. ;-)

 

Though even if a company is not doing sample interpolation, it is still possible to have transition points that are very hard to discern.

 

It's also possible that by employing velocity based filters on the samples, one could conceivably achieve smoother transitions without actually doing any sample interpolation.

[zauberklang]

1 hour ago, AnotherScott said:

I don't think Yamaha does either, except possibly in the SCM pianos which I don't believe are available in any current product.

 

Yamaha uses sample interpolation on all or most of their digital pianos. They just don't mention it anymore because it has become a standard feature. The Yamaha AvantGrand N3 from 2009 had no audible velocity layer switching. http://forum.pianoworld.com/ubbthreads.php/topics/1391594/Re: The DP BSD Project!.html#Post1391594

[AnotherScott]

28 minutes ago, zauberklang said:

 

Yamaha uses sample interpolation on all or most of their digital pianos. They just don't mention it anymore because it has become a standard feature. The Yamaha AvantGrand N3 from 2009 had no audible velocity layer switching. http://forum.pianoworld.com/ubbthreads.php/topics/1391594/Re: The DP BSD Project!.html#Post1391594

That link doesn't say they use sample interpolation. It does say, "Velocity appears to be a smoothly blended multi-layer sample set. - No audible velocity layer switching." But there are different ways to achieve that effect.

 

Of course, as long as it sounds good, that's what matters. :-)

[Tusker]

My experience has been that while sample interpolation is desirable, bad sample interpolation can be fake sounding. As shocking as Princess Leia’s manipulated face in the otherwise excellent movie: Rogue One.
 

 

 

Let your ears be the judge. 

[AnotherScott]

Getting back to Yamaha... the CFX piano in the Montage/MODX doesn't have sample interpolation... you can tell because you can actually get into all the parameters via the edit menus, right down to the wavesamples. I'd be curious to know if people find the CFX to have smoother velocity transitions when implemented on the non-user editable pianos, like CP88, P515, various consoles...?

[cphollis]

Fun topic.  I now want to go do a bit of homework on the psychoacoustic side, e.g. how do our ears perceive different approaches to morphing audio waveforms?   

 

A version of this arises when debating digital sampling rates vs. real analog, or whether 127 values are enough for velocity midi information vs playing a real piano.

 

If the ear can't detect the quantization, does it matter?

[zauberklang]

1 minute ago, AnotherScott said:

Getting back to Yamaha... the CFX piano in the Montage/MODX doesn't have sample interpolation... you can tell because you can actually get into all the parameters via the edit menus, right down to the wavesamples. I'd be curious to know if people find the CFX to have smoother velocity transitions when implemented on the non-user editable pianos, like CP88, P515, various consoles...?

 

This is exactly what I'm dying to find out. I'd love to hear from Yamaha CP88/P515 and Nord Piano/Stage users if there are any audible velocity transitions. (I should have included Yamaha in the title of my original post.) Why did Yamaha decide not to include this sample interpolation or related advanced technology in their top of the line synthesizer/workstation Montage. Are there technical restrictions or was it a product policy decision.

[zauberklang]

9 minutes ago, cphollis said:

Fun topic.  I now want to go do a bit of homework on the psychoacoustic side, e.g. how do our ears perceive different approaches to morphing audio waveforms?   

 

A version of this arises when debating digital sampling rates vs. real analog, or whether 127 values are enough for velocity midi information vs playing a real piano.

 

If the ear can't detect the quantization, does it matter?

 

I believe Vienna Symphonic Library sampled 127 or more layers for their latest pianos https://www.vsl.co.at/en/Synchron_Package/Synchron_Pianos_Bundle

I haven't played them, but I'm pretty sure the transitions between the layers are very smooth.

[Docbop]

3 hours ago, cphollis said:

Thanks for that link.  

 

The geek in me wonders why anyone would create a trade name for it.  

 

 

It was probably created by some Marketing weenie trying to come up with another bullet point for ad copy. 

[funkyhammond]

31 minutes ago, AnotherScott said:

Getting back to Yamaha... the CFX piano in the Montage/MODX doesn't have sample interpolation... you can tell because you can actually get into all the parameters via the edit menus, right down to the wavesamples. I'd be curious to know if people find the CFX to have smoother velocity transitions when implemented on the non-user editable pianos, like CP88, P515, various consoles...?

 

Can you clarify? Why does the ability to edit deep parameters imply no sample interpolation? Is it because you can make the wavesamples sound so different as to make the velocity switching very audible? Wouldn't it depend on how the interpolation algorithm is implemented? I don't own a Montage or MODX so I'm just curious.

 

I'll be heading into the store today to test out the X Piano again so I'll also check out if I can hear sample switching in the CFX in the CP88. It's usually more in EPs rather than APs that I can hear that. 

[El Lobo]

wow. geeking out on being in the weeds on this. you can hear the Nord piano library here: 

https://www.nordkeyboards.com/sound-libraries/nord-piano-library

[zauberklang]

15 minutes ago, El Lobo said:

wow. geeking out on being in the weeds on this. you can hear the Nord piano library here: 

https://www.nordkeyboards.com/sound-libraries/nord-piano-library

 

You have to play the sounds on an actual Nord keyboard to hear it.

[AnotherScott]

1 hour ago, funkyhammond said:

Can you clarify? Why does the ability to edit deep parameters imply no sample interpolation?

Montage is fully user editable, there are no "hidden" pieces to how the sounds are put together, you can examine what happens with every key strike, right down to the wave level. You can swap waves in and out, specify the exact velocity ranges for each, audition/solo/mute them independently, see exactly how they behave. You can build any sound you like out of any of the wave data. But there's no function for adding interpolation between velocity-adjacent wave elements. A sample will sound the same regardless of its adjacent sample. 

 

1 hour ago, zauberklang said:

Why did Yamaha decide not to include this sample interpolation or related advanced technology in their top of the line synthesizer/workstation Montage. Are there technical restrictions or was it a product policy decision.

IIRC, Yamaha guru Bad_Mister kind of addressed this in a discussion about why Montage doesn't have some of the advanced Genos sounds (things that are not specifically tied just to its arranger functionality), i.e. the SA2 sounds. Extrapolating from what he was saying, I think those sounds are built with deep engineering that does not lend itself to a consumer-level user interface. Maybe parameters that are too numerous, complicated and/or inter-related in their behaviors, different for different sounds, difficult to explain or present to an end-user, etc.. Since Montage is designed around letting users manipulate the sounds to the Nth degree (and again, dig in and see exactly how every sound is created), SA2 sounds don't work in that paradigm. Sample interpolation could be something similar. 

 

Roland, as it happens, has a similar situation, but they took a different approach. All the PCM sounds on their rompler-synths/workstations can be deeply edited in a straight-forward manner, just as sounds are on a Montage, but Roland also has their equivalent of SA2 sounds with various behavior modeling and such... the SuperNATURAL Acoustic tones. And Roland's solution, when putting both kinds of sound into the same board (as in the FA and Fantoms) was to make the PCM sounds fully editable as usual, but have almost no editing options available for the SuperNATURAL Acoustic sounds. I think that's a great way to go, and it would be cool if Montage had a set of sounds you could call up where you sacrificed deep editability in order to get those more "exotic" sonic behaviors. But whereas Roland is fine with the idea of "special casing" sounds (you can't edit this kind of sound, you can only load one instance of that kind of sound, you can't use seamless transitions with that one, this one can only work in Part 2, etc.), Yamaha doesn't special-case. There are no such rules, practically anything that works somewhere works everywhere. I understand the appeal of that philosophy, but personally, I'd like to have the option of including SA2 sounds (or SCM, or whatever other special sound-specific tech they may have in other boards) in a Montage/MODX Performance, even with limitations of not being able to do quite everything with them that you can do with all the other sounds. But Yamaha's approach here is not indefensible, especially from a company who has often been taken to task for making products that are too complicated!

 

 

[pjd]

Ah, here's a blast from the past. From the Roland XP-60/XP-80 Owner's Manual, page 50, tone velocity range. In addition to lower and upper velocity switch levels, one can specify crossfade at a velocity switch point:

 

"Crossfade specifies how the volume of the Tone changes when the velocity of a note is outside the specific velocity range. Higher settings will result in a more gradual attenuation. If you don't want the Tone to sound at all for velocities outside the specified range, set this parameter to 0."

 

1998. Quoted the XP-60 manual because I still have it on the shelf. 🙂I haven't done Roland in a while, but this likely dates back even earlier.

 

As to Super Articulation 2 (S.Art2), I agree with AnotherScott. If you think AMW2 is complicated, S.Art2 is insane. It also requires a level of software scripting to recognize gestures and it must respond in a timely way without noticeable latency. Definitely not for n00bs. 🙂

 

All the best -- pj
 

Update: Just found the JV-90 manual on the same shelf. No crossfade in the -90.

 

[mate stubb]

Roland SA instruments had completely seamless velocity response for the 128 midi velocity values way back in the 80s. Not the same technology probably as what the OP is talking about, but it achieved the same practical results.

[stoken6]

So can I ask - what is "sample interpolation"? If you just interpolate between the waveforms at each sample point ("MIDI velocity 60 is represented by interpolating 50% of our velocity-40 sample and 50% of our velocity-60 sample"), you're just superimposing waves - you'll just hear both samples together, possibly with some nasty phasing. 

 

A more intelligent (but complex, and computationally intensive) approach is derive the samples' harmonic content through Fourier transform, and interpolate between the harmonics. 

 

EDIT - or avoid samples altogether, and model the sound, taking into account the impact of velocity (e.g. Roland's SA Synthesis, as Moe just posted).

 

Does anyone know how manufacturers implement what the OP calls sample interpolation?

 

Cheers, Mike

[funkyhammond]

2 hours ago, AnotherScott said:

You can swap waves in and out, specify the exact velocity ranges for each, audition/solo/mute them independently, see exactly how they behave. You can build any sound you like out of any of the wave data. But there's no function for adding interpolation between velocity-adjacent wave elements. 

 

Yup, that would do it. 

 

2 hours ago, AnotherScott said:

Extrapolating from what he was saying...

 

Can I infer from what you are extrapolating regarding the lack of interpolating that this is getting a tad complex? 😉

 

I like Roland's hybrid approach that you described. Makes sense. Have some sounds that are fully editable and others that aren't so that engineers can make assumptions and provide features like sample interpolation, etc.

 

Anyway, I was back in the store today and briefly tested out the CFX and a couple of Rhodes in the CP88. I can hear the velocity sample-switching in the APs if I really focus and just repeatedly hit a key going gradually from very soft to very hard and back down. But it's subtle and I have to really focus. The EPs are a little easier to hear the velocity switching (especially on some of them like Rhodes patch 02) but it's still quite good. I doubt I would notice it under normal playing conditions. And I really doubt I would ever notice it with the APs.

 

 

[zauberklang]

7 hours ago, mate stubb said:

Roland SA instruments had completely seamless velocity response for the 128 midi velocity values way back in the 80s. Not the same technology probably as what the OP is talking about, but it achieved the same practical results.

 

Roland's SA technology is actually a precursor of their SuperNatural synthesis. This is from an Electronic Musician article from 2008:

 

"After the press conference, I talked with Toshio Yamabata, the director of
Roland R&D; Shun Takai, an engineer who has been working on the technology
directly; and Mike Kent, manager of technical relations for Roland
R&D. They confirmed that SuperNatural, which the company has been quietly
developing over the past two decades, is indeed an outgrowth of SA.
As was the case with SA, the first step in creating a SuperNatural program
is separating the various elements of the target sound—for example, the
string and bow of a violin as well as the frequency, time, and phase components.
These elements are then utilized to reconstruct the sound using an
additive technique with complex waveforms from the original sound, along
with modeled components and wave-table synthesis."

[Iconoclast]

Been playing the Stage 3 since it came out and I've never once thought I could hear the difference between the samples.  As far as acoustic pianos go, it's one of the strong points of the board. It's got a great connection between player and sound, unlike a lot of vst pianos that may sound great but feel weird.

As to how they do it...I don't have a single clue.

[Docbop]

28 minutes ago, Iconoclast said:

As to how they do it...I don't have a single clue.

My guess is it sampling vs modeling.   Comparing boards before buying my Nord Stage 3 that only thing I came up with why to me the Nords were fuller sounding and more natural.   Sure modeling you can tweak every little detail, but samples have all the natural overtones to fill in the sound. 

[AnotherScott]

On 5/6/2022 at 3:55 PM, stoken6 said:

So can I ask - what is "sample interpolation"? If you just interpolate between the waveforms at each sample point ("MIDI velocity 60 is represented by interpolating 50% of our velocity-40 sample and 50% of our velocity-60 sample"), you're just superimposing waves - you'll just hear both samples together, possibly with some nasty phasing. 

 

A more intelligent (but complex, and computationally intensive) approach is derive the samples' harmonic content through Fourier transform, and interpolate between the harmonics. 

The first approach you describe does not sound like interpolation, it sounds like cross-fading (as pjd mentioned in his post). Interpolating is not combining two waveforms, but created a new interpolated waveform that is "part way between" the two waveforms. Like the way you can morph one picture into another. You never see (or overlay) the two pictures at the same time, one just gradually turns into the other as points move from point A to point B.