Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards The DPBSD Project!

Generalmusic RP-X Module Review



Yet another interesting DPBSD entry from anotherscott - keep 'em coming! This MP3 is of the default voice "Concert Piano 1" in the Generalmusic RP-X piano module. I've never physically seen a GEM instrument, so it's a real treat to be able to analyze and review this rather rare "bit of kit".

MP3: http://www.mediafire.com/?3bej24sxire5cdx
PIX: http://www.mediafire.com/?z7j3xmnb1z3meg8

Not too surprisingly, the GEM RP-X tests much like the GEM pRP-800 that I reviewed previously (link) so you might want to check that out as well. The main voice here is different than the Steinway or Fazioli however. I'm not sure how one would characterize it (in terms of what brand of physical piano it was sampled from) but I didn't find any technical match in terms of phase signatures and such between this and either of the pRP-800 pianos voices I tested.

GEM talks a lot about the modeled elements in their DPs, but in terms of basic samples they test largely the same as DPs from other manufacturers, with typical note stretching and looped decays. Their attack samples are comparatively longer than most, but their loop samples are unfortunately very short and processed - so processed in fact that I couldn't easily quantify most of the lengths.

Instead of velocity layers, GEM says they use something called "FADE" to alter timbre with velocity. From the description it sounds like a single layer sample with a filter (not sure how complex) and a velocity / time tracking look-up table. It is listed under the heading of physical modeling, but I'm not sure it quite qualifies in my mind as such. For example, Yamaha has a single layer sample with filter too and they don't call it modeling. The effect of FADE on timbre in the RP-X is only obvious over the middle velocities - I would expect quite a bit more timbre change over the higher velocities where everything in a real piano goes more non-linear. The pRP-800 presumably uses the same FADE technology, but has most of its timbre change in the upper 2/3 of the velocity range, which I believe reflects a real piano better.

The strong suit of the GEM approach seems to be with their modeling of pedal and key sympathetic resonance, and also perhaps their algorithm for string damping. So I can see how people are attracted to GEM, and lately to Kawai, as they do many things similarly and well: long attack samples, and nice pedal and key sympathetic resonance effect - unfortunately GEM also has the highly processed loops of Kawai and Yamaha. Looping is really the killer here, as the stretching isn't too obnoxiously audible.

It passes all the pedal / key tests including the quick damping test and the half pedaling test, except for the silent replay test: at pedal up the note is damped and plays at velocity = 1. At 29 dB the dynamic range is fairly un-dynamic, which is likely the reason that softly played notes sound too loud to me. I couldn't hear any "sound effects" such as pedal up/down or key up samples during the test.

One final note, I usually download the manual for the DP I'm testing, and I was struck by the "Technology" section near the end of the RP-X manual. Here they boast about the DRAKE processor they developed, which seems to be some sort of custom 32 bit DSP / RISC special blend running at 50 MIPS. I can only imagine the hardware and software development teams necessary to do this in any sort of competent way. At this point, sadly, uP design is rather hobbyish and a huge resource drain (though it is one hobby I heartily recommend for everyone interested in software or hardware to give a shot as it can be very enlightening - your architectural choices basically force you down very well worn engineering paths). Anyway, I think these creaky old custom processors and all of their legacy overhead are largely to blame for holding back DP development in general, and once they are dispensed with the engineering staff will then be free to concentrate on what they do best: devising better sampling / modeling / hybrid solutions for shoehorning a complex 4 GB piano sample set into 64 MB or less.

Much thanks to anotherscott for this very interesting DPBSD sample!



Figure 1. Spectral phase view of the pedal sympathetic resonance test, stimulus removed, normalized to -1 dB to bring out detail, pedal down (left), pedal up (right). The effect sounds pleasant to my ears.


Figure 2. Spectral frequency view of the key sympathetic resonance test. First "blob" is the C3 & C4 stimulus plus resonance, two seconds later (at cursor) C1 & C2 are lifted, one second later C5 & C6 are lifted, one second later G5 (last "blob") signals end of test. The effect is audibly subtle (as it is on a real piano).


Figure 3. Waveform view of the looping test with vertical zoom applied. Decay times for the low and mid notes are nice and long.


Figure 4. Spectral pan view of the note C1, normalized to -1 dB to bring out detail. Attack & loop samples are quite clearly seen here, attack is nice and long, loop is rather short.


Figure 5. Spectral phase view of the note C2, normalized to -1 dB to bring out detail. Attack sample is visible and quite long in duration (as these things go) but the decay loop is so highly processed I can't measure it, though it sounds quite short (< 1 second). Higher notes are similar.


Figure 6. Spectral pan view of the stretch test, normalized to -1 dB to bring out detail, mid notes. 45 samples cover 88 notes. The timbre matching between groups isn't too bad.


Figure 7. Spectral frequency view of the layer test. Timber variation is smooth from approximately 1/3 to 2/3 of the velocity range, rather static above and below that.


------------
- GEM RP-X -
------------
FILE & SETUP:
- dpbsd_v1.8_gem_rp-x.mp3
- Default "Concert Piano 1" patch.
- Setup: Mac with ProTools LE, Mbox interface @ 24-bit, MP3 via Quicktime Pro.
- Recorded by "anotherscott".
PROS:
- Passes the pedal sympathetic resonance test, the effect is pleasant sounding.
- Passes the key sympathetic resonance test, strings both lower and higher subtly respond.
- Passes the quick damping test, though the second damp brings the note decay mostly to the noise floor.
- Passes the half pedaling test, the effect is realistically slightly buzzy.
- Attack samples lengths are fairly long.
- No obvious layer switches, though most timbre change is largely restricted to the range 1/3 to 2/3.
CONS:
- Fails the silent replay test @ pedal up (note plays vel=1).
- Decay times are long for lower notes, a bit short for the highest notes.
- Obviously looped, both visually and audibly.
- Loops are quite short and sound highly processed, could really benefit from longer decay loops.
- Attack sample lengths are (C1:C8): 5.5,6.2,3.6,2.8,2.4,1.5,0.4?,? seconds.
- Loop sample lengths are (C1:C8): 1.8,?,?,?,?,?,?,?,? seconds.
- Stretching is visible over the entire range, audible over the low end.
- Stretch distances: 3,2,1,2(x3),1,2(x3),1,2(x2),1,2(x8),3,2(x3),1,2(x2),3,2(x2),1,2(x9),3(x2),1 = 45 groups.
- No obvious pedal up/down or key up samples.
- Softly played notes sound a bit too loud.
OTHER:
- Notes played @ vel=1 produce no sound.
- Dynamic range 29dB (vel=1:127).
- MP3 levels: peak @ -1.7dB, noise floor @ -76dB.
- Analysis pix taken with L & R swapped, this is fixed in the final MP3.
- Date reviewed: 2010-11-25.

HI, PHI, UPHI? - An Email Exchange With Tom Love

Tom Love is the Senior Manager of the Electronics Division & Online Marketing of Kawai America & Kawai Canada. I recently sent an email to Kawai requesting information regarding the various Kawai sound technologies (Harmonic Imaging, Progressive Harmonic Imaging, and Ultra Progressive Harmonic Imaging - HI, PHI, UPHI) and he kindly responded to it. It is presented here in its entirety (with Tom's and Kawai's official corporate blessing) and I would be interested in comments from anyone here regarding it.

My questions are in italics, Tom's responses are in regular text:


I consider myself to be something of a digital piano enthusiast, and as such I am very interested in the sound technologies employed by the various digital piano manufacturers. Kawai digital pianos occupy an interesting niche here in the US, as many seem to really like the key actions and piano voices, and they are very reasonably priced.

I agree that Kawai has a strong niche market for digital pianos and stage pianos in the US and Canada (which is what I oversee). And yes, when a user likes our instruments, they seem to really like them. As you may judge from any of our printed and online marketing materials, we really focus first on tone and touch, as compared to the real thing. For instance, my assistant in LA, who is our direct liaison to the engineers in Japan, has a 9-foot Kawai EX Grand in his work area to compare to the various prototypes, actions and pianos sounds that are under development at any given time.

But I'm having some difficulty with what Kawai terms "Harmonic Imaging", "Progressive Harmonic Imaging", and "Ultra Progressive Harmonic Imaging" or HI, PHI, and UPHI. The difficulty I'm having is in understanding what these terms actually mean.

These are naming conventions that we use to describe what is a combination of our technology and methodology in creating an acoustic piano sound for our electronic instruments. First came the original HI which had some unique technology that the engineers came up with which allow us to think outside the box in how the sound was recreated within the electronic instrument, The grand piano sound created by this was very well received at the time and it had a very long run in a number of award winning instruments.

A couple of years ago, we introduced the first of our new versions of our concert grand sound. The first was HI with 88-note piano sampling, something that wasn't done in the original HI sound. I'm sure you know that this means that each note in the electronic instrument is playing back its own unique sample from the original instrument. The benefit of this is a more organic piano sound with more character.

PHI is simply a version of this new sound with access to more sonic data to draw from to create the sound. It is not a different piano sound, but has more depth and richness. UPHI is simply an even more robust version of the same sound.

I'm not asking necessarily for sensitive engineering info (though that would certainly be most welcome) - think of me more as an interested consumer looking for more information about a product before deciding on a purchase.

As I like to say to users, artists and dealers alike, "I could tell you, but then I'd have to kill you." Certainly there are aspects to what we do that we consider unique, proprietary and trade secrets. We boast of the unique, speak some of the proprietary and keep the secrets to ourselves. Don't worry, your life's not in danger from this email!

Say I've narrowed it down to two different Kawai DPs, one with older sound and one with newer: how do I judge the dollar value to me of the difference?

With your ears.

I don't mean to sound flippant, but these are musical instruments and as such, their value is subject to the evaluation of the user / listener. There are merits to both the old sound and the new sound. Much depends upon your ears, your playing style, the music you play, and the type of piano sound that you are already familiar and comfortable with.

I firmly believe that there is no one perfect piano sound, and I'm amused at the great lengths that some people will go to try to convince another that their's truly is the ideal piano. This is why we push to get as many different types of piano sounds on our instruments as possible. On most of our instruments, you'll find at least one version of the new sound along with the old.

I get that HI / PHI / UPHI is something of a good / better / best product sound technology tiering, but how will I perceive this if I play the various versions?

It's our collective feeling at Kawai America that much of the difference between them is perceived (you used a good word) through the experience of playing instruments with the different versions more so than listening to them. Personally, I think it is easier to hear the difference between HI and PHI than it is between PHI and UPHI. Yet, to an experienced player (we have a few of these around the office), there is a distinct difference to the experience of playing say a CA63/93 with UPHI and a CN series model with PHI.

Our job in manufacturing is to balance performance with cost, the old value equation. It gets more costly as you move from HI to PHI to UPHI. It's also more costly to move from a plastic key action to a wooden key action. We make a variety of instruments to try to hit certain usages and price points.

PHI is somehow progressive. How or what is progressive? And UPHI, being ultra, is somehow an improved version of PHI. How or what is improved?

Much was explained in an earlier answer. Additionally, as you move up the "product food chain", other aspects, artifacts and characteristics of a grand piano sound are introduced. These include items such as damper resonance, string resonance, key-off effect, key-off noise, etc. We also add capabilities to "tweak" the sound through functions such as Virtual Voicing.

These are subtle enhancements and additions, but they do make a difference to the experience that a knowledgeable player would have.

My theory is that Kawai are using something like a harmonic map for each key. Not separate samples but instead a single data structure (per key) containing the exact *curve* which each overtone is following depending on the velocity. You can easily see that data structure represented as a picture on their USA site:


You can easily see the three-dimensional structure which represents the whole harmonic content per overtone per velocity or something like that.

If I am right, I can't understand why they are not explaining that. GEM's technology is similar if not the same.

Here is the quote from the site:


What I understand is this: You record say 100 velocity levels per key. You can graphically put that on a 3D map like this: on X-axis is the corresponding velocity, on Y-axis is the time and on Z-axis is the amplitude of the wave-form. In other words, imagine many waveforms from ppp to fff stacked to each other which would create a 3D image. If you apply Fourier transform to such a map (i.e. extract overtones in each moment) you will transform that into a 4D structure which is a little bit hard to imagine, so let's simplify. Take for example the fundamental. It's a sine wave which gets quieter with the time. So, we will not draw the fundamental sine wave, but instead will draw a line to denote only the current amplitude of it. This requires substantially less data than the wave file, since the amplitude is smooth and not cyclic. This way we can represent each overtone for a specific velocity by a smooth curve representing the change in time of its amplitude. By doing so we can have really precise map of the whole harmonic content by using less data. But that's just a theory

For all I know they're doing strenuous Fourier gymnastics in there, but the upshot is it tests the same and sounds the same (to me anyway) as garden variety looping. The test of the pudding is in the eating.

It seems to me that the various technologies given names by manufacturers are first and foremost sample set compression methods. As such their primary value-add is in reducing the size and cost of sample memory, and not so much in adding realism to the sound. One can get fantastic realism with conventional sampling methods provided the sample set is made sufficiently large (>500MB, ~$1 retail).

Well, that's the heart of the matter, I think. I mean, one side of the argument says "tell us what this technology does" and the other side says "it doesn't matter, what matters is whether it sounds better" (aka "the test of the pudding is in the eating"). If you can't hear the difference, then explanation of the technology doesn't really matter. That said, if there is a particular sonic difference that we knew specifically to listen for, that might make it more likely that we would actually hear it, once we were focussed in the right direction.

That's exactly the kind of information that I was fishing for from Kawai. What should I be listening for that differentiates the various technologies?

That's some elusive answers you got there, I'm a tad bit more confused now. It's almost like I would like to send them an Excel spreadsheet with the features of other manufacturer techs checked and just ask him to check the features of theirs.

But since that not going to happen we have no choice but to follow their advice and just try it for ourselves. Sadly so few have the possibility to do so :-/

Compare it to the Roland SN video, where they painstakingly point out exactly what you should listen for, with phase scope shots and everything:

[video:youtube]X0jY-NR_WG4[/video]

HI vs. PHI vs. UPHI: The Listening

I thought it might be interesting to listen to a couple of mashups of three different Kawai DPs representing their three differently monikered sound technologies. These three DPs are:
1. MP5 which uses Harmonic Imaging (HI)
2. CN33 which employs Progressive Harmonic Imaging (PHI)
3. CA63 which utilizes Ultra Progressive Harmonic Imaging (UPHI)

First off, the CN33 and CA63 sound like they may have been derived from the same sample set or same physical piano, while the MP5 sounds quite different, particularly the lower notes. Second, the incidental sounds such as key and pedal noises of the CA63 weren't captured via MIDI playback for some reason.

I. Attacks & Loops


Figure 1. Waveform view of the first mashup.

MP3: http://www.mediafire.com/?qv1i56yom6wnc76

The first mashup consists of the notes C1, C2, ..., C7, C8 as follows:
C1: MP5, CN33, CA63 (9 seconds each)
C2: MP5, CN33, CA63 (8 seconds each)
C3: MP5, CN33, CA63 (7 seconds each)
C4: MP5, CN33, CA63 (6 seconds each)
C5: MP5, CN33, CA63 (5 seconds each)
C6: MP5, CN33, CA63 (4 seconds each)
C7: MP5, CN33, CA63 (3 seconds each)
C8: MP5, CN33, CA63 (2 seconds each)

I limited the sample lengths so that the loop sections are adequately represented but not so long as to induce ennui in the listener. All notes are individually peak normalized to -1dB except for the C8 samples of the CN33 and CA63, which I had to manually attenuate some to sound about the same loudness as the MP5.

What I hear:
C1, C2, C3:
- MP5 - Loop is kind of obvious and muffled sounding, loop decay is maybe not fast enough.
- CN33 - Loop is more obvious than MP5, though brighter sounding.
- CA63 - Loop is less obvious, but loop decay is maybe too fast.
C4: All loops have a rather dull timbre.
C5, C6, C7, C8: All sound fairly identical and OK to me.

I must say that I don't get the dull decay timbre thing. Our Young Chang grand piano gets slightly duller as the notes decay, but it's not nearly as pronounced as any of these three DPs. Though I've noticed that lots of DPs do this, not just Kawai.


II. Dynamics


Figure 2. Waveform view of the second mashup.

MP3: http://www.mediafire.com/?dgv3adwdd967m2k

The second mashup consists of the note C4 (all durations 0.5 sec) as follows:
MP5 @ velocity 1, 3, 5, 7; CN33 @ velocity 1, 3, 5, 7; CA63 @ velocity 1, 3, 5, 7
MP5 @ velocity 41, 43, 45, 47; CN33 @ velocity 41, 43, 45, 47; CA63 @ velocity 41, 43, 45, 47
MP5 @ velocity 81, 83, 85, 87; CN33 @ velocity 81, 83, 85, 87; CA63 @ velocity 81, 83, 85, 87
MP5 @ velocity 121, 123, 125, 127; CN33 @ velocity 121, 123, 125, 127; CA63 @ velocity 121, 123, 125, 127

Each DP sample set (velocity 1:127) is bulk normalized so that velocity 127 is -1dB peak.

What I hear:
MP5: The highest velocities are quite "clangy" and metallic sounding.
CA63: I can hear fine gradations of timbre within the contiguous velocity groups.


What do ya'll hear?

"What do ya'll hear?"

I hear the same thing, but the difference PHI and UPHI is difficult for me to detect. The lower notes are much improved between the MP-5 and MP-6.

Thanks for this side-by side. One of the difficulties of comparing instruments is that ears adjust or "forget" exact tones between sitting at one piano and sitting at another; it doesn't take long for the ears to forget a tonal quality. Your mashup alleviates that problem.

Could so this same mashup for the RD-700NX substituted for the MP-5? This would put Kawai's two technologies -- UPHI and PHI -- next to a Roland supernatural piano.

Kawai MP10 Review



We recently obtained several DPBSD MP3 files for the new Kawai MP10 from Piano World forum member "tinybox" - thanks so much! There are three files, one each for the "Concert 1", "Jazz 1", and "Pop 1" voices. I place them all at the share point, along with some pix of the analysis.

MP3's:
- http://www.mediafire.com/?z1pkk24z3pkxp3m (Concert 1)
- http://www.mediafire.com/?5hom36uo63zx3x2 (Jazz 1)
- http://www.mediafire.com/?964gtaeq6l8onc7 (Pop 1)
PIX:
- http://www.mediafire.com/?i432edh8hnpjvqy

As usual, since it is the first preset, the "Concert 1" voice will be reviewed here. I plan to do a separate post highlighting the three piano voices, as they sound and look (in terms of phase signatures and such) like fundamentally different sample sets.

The DPBSD MIDI file was played (sequenced) directly on the MP10 (sitting on a thumb drive I believe) with the resulting MP3 files recorded directly by the MP10 (also to thumb drive) - the MP10 can similarly directly render MIDI => WAV. Hats off to Kawai for providing this very convenient rendering method, one I hope soon becomes an industry standard. I'd like to see a bit more detail in the resulting WAV file as Adobe Audition reports it as 16 bit resolution, but the dynamic range is really pretty good. I prefer to micro manage audio compression so I generally record to WAV and then perform MP3 or OGG compression on my PC. Windows reports the MP3 files generated on the MP10 as 192 kbps, and the manual says these are constant bit rate (CBR) which, coincidentally, is what I usually recommend for the DPBSD test.

The MP10 passes all of the piano / key tests, except for the end of the silent replay test: at pedal up the note incorrectly damps. I hear a "clunk" sound effect when keys are released to the up position, and a "loom of strings" sound when the pedal is depressed - the sounds are not obnoxious and add to the realism. Note that these sounds are now reproduced when playing a MIDI file, which is something Kawai apparently fixed.

As usual for a Kawai DP, the pedal and key sympathetic resonance is well done and realistically pleasant sounding, which I believe adds quite a bit to their sonic appeal. To their credit, Kawai seems to have sworn off note stretching some time ago, which is quite laudable. But they (as well as most other DP manufacturers, I'm not singling out Kawai here) still utilize looping for the note decays and, though it is fairly well done, I can hear evidence of it on notes C3 and C4, and somewhat hear it on C1, C2, and C5. The pedal sympathetic resonance helps mask simultaneously playing loops to some degree, but I would like to see/hear longer, less highly processed loops or preferably no looping at all.

In terms of velocity layers, I find the MP10 to have a fairly smooth timber change with velocity over the entire range, with two slightly audible timbre steps, and one fairly jarring timber step near max velocity. It's very difficult to tell audibly or visually how many velocity layers are employed, but anywhere between 4 and 8 or so wouldn't surprise me. During my testing I usually compress the layer test 20:1 so I can listen for timbre variation without having my eardrums blown out. If anyone else wants to hear this for themselves I trimmed it out of the DBPSD MP3 and put it here:

http://www.mediafire.com/?iq84nbkodr8drar

Much thanks to tinybox for submitting test samples of this very new and interesting DP!


Some analysis pix and text review:


Figure 1. Spectral pan view of the pedal sympathetic resonance, stimulus removed and peak normalized to -1 dB to bring out detail. The effect sounds subtle yet pleasing to me, and helps to hide the note decay loop sound.


Figure 2. Spectral frequency view of the key sympathetic resonance test, vertical zoom applied. First "blob" is the C3 & C4 stimulus plus resonance, two seconds later (at cursor) C1 & C2 are lifted, one second later C5 & C6 are lifted, one second later G5 (last "blob") signals end of test. The effect is audibly subtle (as it is on a real piano) with strings both lower and higher responding. You can also see evidence of the incidental key-up sound effects here.


Figure 3. Waveform view of the looping test, vertical zoom applied to see the noise floor. You can see easily see the C1, C2, and C4 loops here, the applied decays to them, and the incidental key-up sound effects. Decay times are nice and long, looping is fairly well done, and doesn't sound as artificial as it looks here.


Figure 4. Spectral pan view of the note C2, test peak normalized to -1 dB to bring out detail. Attack and loop samples are visible, with cursor positioned at the end of the crossfade. Loop sample lengths are rather short and processed so they don't "wobble" much. Other notes are similar.


Figure 5. Spectral phase view of the stretch test, mid notes, test peak normalized to -1 dB to bring out detail. No obvious evidence of stretching, low and high groups are similar.


Figure 6. Spectral frequency view of the layer test. Timbre changes fairly smoothly with velocity, and over most of the velocity range, but there is an abrupt sounding timbre step around velocity 124 (at cursor). Floating "blob" in the black and white view at the upper left signals another two timbre steps I can hear, but they are much less audible.


--------------
- Kawai MP10 -
--------------
FILE & SETUP:
- dpbsd_v1.9_kawai_mp10_concert1_v2.mp3
- This is the "Concert 1" piano voice.
- MIDI => MP3 directly via MP10.
- Recorded by "tinybox".
PROS:
- Passes the pedal sympathetic resonance test, the effect is pleasant sounding but fairly subtle.
- Passes the key sympathetic resonance test, strings both lower and higher subtly respond.
- Passes the quick damping test.
- Responds to half pedaling.
- Note decay times are nice and long.
- No obvious stretching visible or audible.
- Timbre variation is fairly smooth with increasing velocity, layers fairly well blended.
- Key up "knock" and pedal down "loom of strings" samples (also rendered via MIDI).
CONS:
- Fails the silent replay test, note damps @ pedal up.
- Obviously looped, both visibly and audibly, though fairly well done.
- Attack sample lengths are (C1:C8): 2.5,3.0,2.2,1.8,1.7,1.5,1.1,? seconds.
- Loop sample lengths are fairly short and somewhat audible over the lows and mids.
- Loop sample lengths are (C1:C8): 1.4,1.3,0.66,0.83,0.65,0.75,0.65,? seconds.
- Layer timbre steps are visible and audible @ velocity = 22,36,124, with vel=124 quite abrupt sounding.
OTHER:
- Visual evidence of possible layer switching @ velocity = 30,46,64,74,102.
- Notes played @ vel=1 produce a sound.
- MP3 levels: peak @ -4.3dB, noise floor @ -84dB (very approx.).
- Dynamic range 52dB (vel=1:127).
- Date reviewed: 2010-12-12; corrected 2010-12-14.


[EDITS]
2010-12-15:
- New MP3 links to new MP3 files, all were the "Concert 1" voice.
- Corrected my comments regarding MP3 vs WAV - original WAV => MP3 was done via Audacity.
- Corrected my comments regarding MP3 dynamic range - it is in fact very similar to WAV.
- Concluded that silent replay test failure is note damp @ pedal up.
- Corrected the text review to reflect the above.

Once I get the NX I might try that, thanks for the suggestion!

It might be interesting to pit top of the line Roland (NX), Kawai (MP10), and Yamaha (AG?) against each other in some way.

I would suggest that the CP5 would be a closer model for comparison.

Cheers,
James
x

For stage piano price points, I agree.

I guess I was thinking of comparing the technically best current AP voice from each manufacturer, regardless of price or the specific target application. Unlike the AG, the CP5 is stretched, which makes it somewhat inferior IMO (though the loops on the AG are fairly short - I need to give the test MP3's another close listen).

The entire industry seems to be holding back for god knows what.

I think Roland and Kawai are offering some pretty outstanding digital pianos for those wanting as an authentic as possible piano experience. The RD700GXF/RD700NX and Kawai MP10 look to be class leaders in action and sound so far as I can tell. I've yet to play the MP10, but I'd be very happy with either the GXF, the NX, or the MP10 for a great stage piano. The MP10 may be a bit heavy actually... I will say that I really did like the CP5 a lot. More than the CP1 oddly enough, but I prefer the action of my GXF and the NX, and I'm sure the MP10. If Yamaha's action was graded the way the GXF/NX/MP10, I think it'd be right there with the others.

In terms of keys and timing and convenience, perhaps, but sound-wise PC samplers from a decade or more ago leave modern DPs in the dust. We're talking pennies here in terms of sample memory cost, maybe a couple of bucks when it's all said and done. I understand product tiering, and I can see not wanting to cause competition among one's various models but, come on, the existence of looping / stretching in a >$2k pro class instrument is outrageous.

I'm not specifically harping on Kawai here (or you either PianoZac) - Yamaha does it too, as does Korg, Kurzweil, Casio, etc. and I'm sure Roland SN isn't the final word in terms of sample memory space compression. Let us stop this sample set butchery, there really isn't a compelling financial / engineering argument for it anymore.

Listening to the three MP3s (and viewing their waveforms in an audio editor), it is clear that the recordings are indeed all identical.

I believe this is because the SMF (which does not contain any programme change messages) must be reloaded whenever the panel voice is changed, otherwise the playback sound will remain the same.

I'll obviously confirm this point with the MP10 engineers tomorrow.

Kind regards,
James
x

Maybe my point of view is not relevant here in this more technical context (it's all about memory, looping, stretching etc.), but I want to share it anyway:

You might know that I recently shared some experience to you regarding Pianoteq play and some layering with my KAWAI CA63.
I have to say that I changed my mind: Over time I no longer prefer to use my Netbook for layering CA63 and Pianoteq play. I completely switched back to just use the plain CA63 sound for my playing experience. Why? Well, I really don't know. Maybe the authenticity of the playing experience now is more important to me than getting some additional sound effects and realism from Pianoteq play. I could never stand the sole sound of Pianoteq anyway because it's too artificially sounding to me. Playing it layered with a sample-based piano the sound experience was way improved to me. But then the overall sound somehow seems to be too "thick", I don't know how to express this.

Maybe it's "back to simplicity": Currently I like the simple original "concert grand" sound very much, with reverb "hall 1". I also often use TADutchman's "Blended Concert Grandeur", "Shigeru Kawai" and "Steinway Grand Devotion" (in this order), depending on my mood and the piece I'm playing.

I really like my KAWAI CA63, despite of the objective technical sound flaws, e.g. looping, short decay! And I especially like the keyboard touch better and better from day to day!

No wonder they tested very similarly! I should have spent more time listening for differences.

There is an "under construction" notice at the top of the post until we clear this up, sorry everyone for the confusion.

I second that. I am mostly using the default Concert Grand sound. I used to play with heavy touch, however I have recently discovered that a better way to get rid of the occasional jumps in velocity is not the heavy touch but instead a normal touch with piano timbre set in the designer to "mellow 2". So, that's the only change I apply to the default sound. (And I add a "hall 2" reverb if I play with headphones.)