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

I've spent enough hours to make at least a week playing the CP1 and V-Piano before deciding on the V-piano for purchase. I believe it is a conbination of the sound (quadraphonic setup) on the V-piano coupled with the keyboard being smoother to play. I own a CLP990 and got used to a heavier feel on the keys, however switching tho the V-piano has improved my playing, at least subconscioulsy..lol.. I was expecting more from the CP1 considering it has come out two years after the V-piano. However I prefer my CLP990 over it if I had to choose. The v-piano keeps calling for me to play it, so I made the choice to get one. For the last six months, I've been using a college owned V-piano, and yesterday I made the purchase for my home. Coupled with my CLP990 and an S90 from yamaha, I will have a new edition to my keyboard family. Thanks for the reviews on the actual sounds. Even though I don't hear a problem with the midrange on the V-piano as others have, it is good to have opinions and there is always room for improvement. One day one of these companies shall achieve perfection. At the moment the V-piano is still the closest thing to that as the CP1 falls short of expectations.

500K VIEWS!



In a mere 9 months the DPBSD thread has doubled its views to reach the half million mark - thanks everyone for the continued interest, and thanks to Pianoworld for hosting this awesome forum! Have some cake (no Kit Kats this time, just pure creamy icing goodness)!

Massive thanks to all of the DPBSD MP3 donors!

The DPBSD Honor Roll
10 pesk
9 anotherscott
7 setchman
3 Glenn NK
3 jve
2 ChrisA
2 CyberGene
2 EssBrace
2 octurn
2 VivatRudolphus
1 Alexank
1 bkmz
1 Derek Andrews
1 Goofball Jones
1 jef_citron
1 jens4711
1 JFP
1 Jumajazu
1 kawaian/mucci
1 kurtie
1 luisdent
1 M.Schreck
1 Martin C. Doege
1 Mawima
1 Melodialworks Music/Lawrence
1 mkhor
1 nan
1 NikkiPiano
1 pkdd
1 Qbert
1 reza
1 R0B
1 sandord
1 sdw91
1 Soren Jorvang
1 tinybox
1 Vincentimes
1 Voltara
1 Volusiano
1 voxpops
1 Yuri Pavlov
1 zaba19

Fantastic achievement! Thanks Dewster!

Great job Dewster

Good Job Dew Meister!

Congratulations! And may the next half million be just as controversial!!!

All hail, dewster!

Congrats indeed...even if the results are meaningless.

(joke)

Cheers,
James
x

Anyone out there with a Roland RD-300NX and/or FP-4F?

I would be very interested DPBSD MP3s of both (the default SuperNATURAL piano with all resonance and sound effects enabled - PM me if you need help doing this). I've read reports lately that the SN piano voice in the RD-300NX sounds inferior to that in the RD-700NX / RD-700GXF / FP-7F / HP307 and I'd like to get to the bottom of this if possible.

[EDIT] I could also use a DPBSD MP3 of the Yamaha CP1 S6 piano voice. And if you're feeling extra charitable, I could use an updated CP1 CF sample as well!

TIA!

And thanks for the congratulations everyone!

dewster: Have you considered testing Synthogy Ivory? (I did a forum search and I couldn't find any mention of Ivory in this thread.)

Yes, I reviewed 4 Ivory pianos quite a while ago at the beginning of this project, nothing new since then. I'd love one or more updates if anyone feels like doing it!

BTW, there is a clickable index on the first DPBSD post which I hope will keep people from needing to use the search function when looking for the review of a specific brand / model within the thread.

Yamaha NP-V60 Review



This is anotherscott's tenth DPBSD MP3 submission - WOW! This one is for the Yamaha Piaggero Series NP-V60. The MP3 is of the default voice "Grand Piano".

PIX: http://www.mediafire.com/?1hip391jdoj0686
MP3: http://www.mediafire.com/?t9rp9h99497h7tf

Anotherscott previously gave us DPBSD MP3s of the NP-V80: the default voice "Live! Grand Piano" and the third voice "Grand Piano". At that time I reviewed the main voice in the NP-V80, but decided not to review the third voice as it tested somewhat inferior to the main patch. The reason anotherscott provided it was due to his speculation at the time that it might be identical to the NP-V60 main piano voice.

After a quick listening comparison when able to subsequently directly compare the two, anotherscott felt the NP-V60 main voice and NP-V80 third voice indeed sounded quite similar, but he heard a difference in how the lower notes were stretched:


Which I confirmed in my testing. Looking at the stretch test for both the V60 (main voice) and V80 (third voice) I'm seeing identical pan/phase signatures, the only exception being the V60 has one extra group located second to last from the bottom:

NP-V60 : 13,5,8,4(x7),3,7,24 = 13 groups (main voice)
NP-V80 : 18,8,4(x7),3,7,24 = 12 groups (third voice)

So both are super duper stretched, particularly on the ends, but the NP-V60 has one extra bass group.

Compared to the main NP-V80 piano voice, The NP-V60 piano voice tests worse in several critical ways: the decays are shorter, the attack and loop sample lengths are generally a lot shorter, and it's stretched more. This poor showing in comparison to an already fairly poor piano puts it firmly near the bottom of the DPs I've tested in terms of sound technology. The fast decays hide the short loops to some degree, but the decay rate itself sounds completely unrealistic. The extreme stretching of the lowest and highest ranges also sounds fake, and the stretch transitions are quite obviously audible. It's hard not to think of this voice as much more than one might find in a decent toy.

Estimated piano voice sample set ROM size:
5.4 sec (attacks) + 4.2 sec (loops) = 9.6 sec / 8 notes = 1.2 sec/note
1.2 sec/note * 13 notes (stretch groups) * 44100 samples/sec * 2 B/sample * 2 (stereo) = 2.75 MB

You would take that number and multiply it by the layers to get the total ROM required, if they are using filtering instead of actual layers the multiplier would then be 1. Any way that you look at this it's probably a really tiny, some would say microscopic, sample size. And while huge sample sets don't necessarily guarantee good sound, I would argue that tiny ones like this (particularly when compressed via stretching and looping) are always associated with poor sound.

Some analysis pix and text review:


Figure 1. Waveform view of the entire looping test, vertical zoom applied to see the noise floor. The decay rate is super fast.


Figure 2. Spectral pan view of the note C3, normalized to -1dB to increase clarity, cursor at attack / loop transition. Attack and loop sample lengths are super short.


Figure 3. Spectral pan view of the stretch test, low end of the notes, normalized to -1dB to increase clarity. Low and high ends are super stretched which makes the notes sound fake. Mids are fairly stretched, with mostly groups of 4. Stretch group transitions are quite audible over the entire range due to timbre and stereo pan mis-matches between stretch groups.


Figure 4. Same view as above but for the NP-V80 third voice (for comparison purposes). Low end is even more stretched. There is a slight timbre change at the cursor that I initially mistook for a stretch group transition.

-----------------
- Yamaha NP-V60 -
-----------------
FILE & SETUP:
- dpbsd_v1.9_yamaha_np-v60.mp3
- This is the first patch: "Grand Piano".
- Setup: Headphone out, Mbox interface @ 24-bit, Mac with ProTools LE, MP3 via Quicktime Pro.
- Recorded by "anotherscott".
PROS:
- Passes the quick partial damping test.
- Passes the half pedaling test.
- Smooth timbre variation with velocity (blended layers and/or filter).
CONS:
- Fails the pedal sympathetic resonance.
- Fails the key sympathetic resonance test.
- Fails the silent replay test (damps @ pedal up).
- Note decay times are very short.
- Obviously looped, both visibly and audibly over the lows and mids.
- Audibly abrupt volume change at the attack/loop transition for note C1.
- Loop sample lengths are fairly short and audible over the lows and mids.
- Attack sample lengths are (C1:C8): 2.?,1.9,0.4,0.3,0.2,0.1,0.3,0.2 seconds.
- Loop sample lengths are (C1:C8): 1.3,0.5,0.3,0.5,0.4,?,?,? seconds.
- Very stretched, visible and audible over the entire range.
- Poor timbre and stereo pan matching between stretch groups.
- Stretch distances: 13,5,8,4(x7),3,7,24 = 13 groups.
- No obvious pedal up/down or key up "clunk" samples.
OTHER:
- Dynamic range 55dB (vel=1:127).
- Notes played @ vel=1 produce no sound.
- MP3 levels: peak @ -2.3dB, noise floor @ -67dB.
- Date reviewed: 2011-09-24

Anyone out there with a Yamaha P-155 feel like providing a v2.0 DPBSD MP3?

This is a super popular download (~150 times!) and all we have is a v1.3, so it would be nice to revisit it. I added it to the begging section of the first post in this thread.

Korg Microstation Review



With this his 11th submission, anotherscott takes top position on the DPBSD Honor Roll! This entry is for the Korg Microstation, the MP3 is of the default voice "L/R Piano".

PIX: http://www.mediafire.com/?uezf2w6txab1ja6
MP3: http://www.mediafire.com/?77ozo8o9ao2k1nm

With its "61-note natural touch mini-keyboard" it may not be everyone's cup of tea, but those 5 octaves fit in a 778 x 210 x 82 mm (30.63" x 8.27" × 3.23") package that weighs only 2.6 Kg (5.73 lbs). At $500 USD this keyboard is both highly portable and affordable.

Some Q&A with anotherscott:


The Korg Microstation web site says: "The oscillator section provides access to 49 MB (calculated as 16-bit linear data) of high-quality PCM sources (sampled at 48 kHz) delivering a rich variety or sounds."

Because the piano sample is mono, it's difficult for me to tell how long most of the attack and loop samples are. But as a very rough estimate let's say it's 1.8 seconds per stretch group.

1.8 sec/group * 16 groups * 48000 samples/sec * 2 B/sample = ~2.8 MB.

Which is 2.8 MB / 49 MB = ~1/17 of the ROM space. Since it shares this space with 359 other sample sets and a bunch of drums, the main piano sample consumes much more than its fair share (as is probably the case in most keyboards) but still it is much too tiny. I wish manufacturers would take the DP aspects of these kinds of products a bit more seriously, thought the synth stuff seems kind of neat if you watch the youtube videos.

Some analysis pix and text review:


Figure 1. Spectral pan view of the entire MP3. The main piano sample set is mono with some kind of mild stereo effect and appropriate L/R pan applied (note the 'S' shape of the sample stretching test on the right).


Figure 2. Waveform view of the looping test, zoomed vertically in order to see the noise floor (so that the decays have a clear visual reference). Decay times of the low and mid notes are too short. Attack samples sound fairly decent, looping isn't too terribly done though the loop sample lengths could be longer.


Figure 3. Waveform view of the entire stretch test (can't use the pan or phase views here due to mono). This sample set is really stretched, with a mere 16 samples covering 88 notes. Most stretch group transitions are audible.


Figure 4. Spectral frequency view of the layer test. The timbre smoothly brightens with increasing velocity, but it sounds kind of bland like maybe a filtered single layer. Dynamic range is rather narrow and compressed sounding.

---------------------
- Korg Microstation -
---------------------
FILE & SETUP:
- dpbsd_v1.9_korg_microstation.mp3
- This is the first patch: "L/R Piano".
- Setup: Headphone out, Mbox interface @ 24-bit, Mac with ProTools LE, MP3 via Quicktime Pro.
- Recorded by "anotherscott".
PROS:
- Passes the quick partial damping test.
- Passes the half pedaling test.
- Smooth timbre variation with velocity (blended layers and/or filter).
CONS:
- Mono sample set.
- Narrow dynamic range makes softly played notes sound too loud.
- Fails the pedal sympathetic resonance.
- Fails the key sympathetic resonance test.
- Fails the silent replay test (damps @ pedal up).
- Note decay times are fairly short.
- Obviously looped, though the attack sample lengths seem adequate.
- Loop sample lengths are somewhat short and somewhat audible.
- Attack sample lengths are (C1:C8): ?,?,?,?,?,0.8,?,? seconds.
- Loop sample lengths are (C1:C8): 1.0,1.2,1.0,1.0,1.0,?,?,? seconds.
- Very stretched, visible and audible over the entire range.
- Stretch distances: 7,5,7,4,5,4,4,5,4,4,5,5,6,5,8,10 = 16 groups.
- No obvious pedal up/down or key up "clunk" samples.
OTHER:
- Dynamic range 27dB (vel=1:127).
- Notes played @ vel=1 produce sound.
- MP3 levels: peak @ -1.9dB, noise floor @ -78dB.
- Date reviewed: 2011-10-08

Windows Media Player Review



Here's a blast from the DPBSD past - creaky old v1.2 as a matter of fact, performed over a year and a half ago at a point where I was still getting my analysis feet wet. It doesn't appear that I posted this previously, so here it is for historical purposes.

MP3: http://www.mediafire.com/?xtxgomzm3x1

------------------------
- Windows Media Player -
------------------------
FILE & SETUP:
- dp_bsd_v1.2_wmp.mp3
- Adobe Audition.
- Recorded by "dewster".
PROS:
- Free.
CONS:
- Mono sample.
- Unnaturally quick decay - notes dive straight into the dirt.
- Truly wretched looping.
- Ghastly stretching of entire octaves.
- No layer switching (here a bad thing - obviously a one-layer sample w/o filtering).
- Sickening sympathetic resonance.
- No response to partial pedaling.
- Highest notes sound really flat.
OTHER:
- Dynamic range 66dB (vel=1:127).
- Start here if your ears need training (or abuse!).
- This is quite possibly the worst DP voice in the world.
- Date reviewed: 2010-02-03

DPBSD Routine Maintenance

- The DP model name links (within manufacturers) on the Main Post page have been alphabetized (take that, EssBrace!).

- Ditto within the text reviews file located at the Share Point.

Looping vs. SuperNATURAL: The Decaying

I tend to go on and on about the evils of looping. I have a personal loathing of it which I express perhaps too often in PW forum threads. PW member Brent_H recently described looping as "that uncanny effect where it sounds very like a piano for the first three quarters of a second and then exactly like someone turning down a volume knob as it decays." This is a very apt analogy because, with few exceptions, DP synthesis consists of exactly that.

Here is what most DPs do when you play a note:

1. They play back a few seconds of a recorded sample of the first part of the note - the attack sample. The bass notes are almost always allotted more sample time than the treble notes because it takes longer for bass notes to "settle down" into their decay phase.

2. This is followed by a short crossfade, which is a smooth transition that goes fully from one sound to another via simple mixing, over to:

3. a short piece of the recorded sample played over and over - the loop sample - with an envelope applied in order to form the dropping volume over time decay sound. The looped decay may be accompanied by a gradual introduction of a low pass filter, which dulls the highs over time.

Given a little thought, the problems with this approach should be immediately obvious. Piano note harmonic decay is much more complex than a simple reduction in volume and a dulling of the overall brightness. Often the loop sample period is audible because it isn't long enough to capture the "wobbling" sound of a string with multiple modes and one or more unison strings interbeating with it. So if you can hear the loop period it almost always sounds fake, but even if you can't it's still a real tone killer.

One of the technically best loopers I've tested in this thread is the Yamaha AvantGrand N3, which is also the most expensive DP I've tested. Let's look at some numbers from that review:

- Attack lengths are (C1:C8): 4.4,4.3,3.9,3.5,3.0,2.1,?,? seconds.
- Loop lengths are (C1:C8): 0.7,0.67,0.67,0.67,0.56,0.43,?,? seconds.

The AG has nice long attack (the crossfade time is included in the attack numbers) sample times. But it is hobbled by very short loops. In order to do this, Yamaha has to highly process the loops in order to remove any audible periodicity introduced by the looping itself, and in so doing they must remove long-term phase/pan changes due to the interbeating of unison strings, longitudinal modes, etc.

Many will tell you that looping doesn't matter much because it gets largely hidden during normal play. Indeed, this is the reason almost all demo songs in DPs consist of a flurry of quickly played short duration notes. They're intentionally emphasizing the positive by biasing the playback in favor of the most realistic sounding portion of each note, which is the attack sample. Bass notes that are held for any length of time are usually are accompanied by a bunch of higher notes playing, and it's hard not to interpret this as a conscious attempt to mask the sonic deficiencies of the looping bass notes.

The DPBSD test doesn't provide much opportunity to the listener who wants to hear what that particular DP sounds like in an actual performance. The pedal sympathetic resonance test plays several notes at once, but that's about it. So I decided to take the AG DPBSD MP3 file, which has all the C notes in it for the looping test, and perform a multi-channel mix. I arranged C1 (the lowest C on the piano) to start at time zero, then arranged C2 through C6 to have their attack-to-loop transitions aligned with that of C1. I then normalized it to -1dB peak, and applied a small amount of hiss reduction because the SNR of the source MP3 isn't the best. I played around with muting various combinations of the notes, but the worst loop sound I could come up with was that of the C1 all by itself, which sounds kind of like a motorboat.

For comparison, I took the Roland RD-700NX "Concert Grand" DPBSD MP3 file and did the same exact thing, and even though it wasn't necessary I applied the hiss reduction to keep the playing field as level as possible. In order to highlight the differences in the decays rather than the attacks, I took the period from 5 seconds to 10 seconds in each of these and put them at the end of the file For a more direct comparison.

MP3: http://www.mediafire.com/?dmdfj21y1k11hv3

And here is a view of it in Audition:


Figure 1. Spectral pan view of a mix of the notes C1 through C6, all MIDI velocities 63, stacked so that all attack/loop transitions are aligned. 1st segment (from the left) is the Yamaha AvantGrand N3; 2nd segment is the Roland RD-700NX (no looping but identical timing of the notes); 3rd segment is the central 5 seconds of the 1st segment; 4th segment is the central 5 seconds of the 2nd segment. Both note mixes were independently normalized to -1dB, then a light hiss reduction was applied to push down the audible noise floor. Full transition to looping of all notes for the AG begins at 4.4 seconds into the file. All samples taken from the associated DPBSD MP3 test files.

- What I hear -
What you see in the spectral pan view above is pretty much what you get. The NX decays have that slow random phasey thing going on like a real piano, the ensemble decay of many notes sounds rich and thick. The AG loops sound completely dead in comparison, and putting a bunch of them together doesn't seem to improve the sound significantly - granted this is stereo, but I can't imagine 4 channels of that improving things much either. To my ears the differences in the decays between looped and SN are like night and day.

- Disclaimer -
We've owned the NX for over a year now but readers shouldn't interpret this post as a plug for it in any way. I'm only using it here to highlight the sonic deficiencies of one of the best and most expensive loopers on the market. The NX has a lot going for it, but once a fully sampled stage/slab DP hits the market with good keys, good sound, and good user interface, we'll most likely sell the NX and get that.

- DP wish list -
Uploadable fully sampled Yamaha, Bosendorfer, Steinway, Chickering, etc. that pass all the DPBSD tests, EPs, harpsichords, clavichords, pianofortes, strings, mixable pipe organ stops; 76 light hammer action ungraded keys starting at A0; a "box" shape with no unnecessary length / width / height / weight / weird angles; 3 sheet music rest; small internal speakers; simple rational UI; MIDI recording / playing; WAV recording; MIDI => WAV rendering; gorgeous reverbs; <$3k

Tut! We're all doomed.

What is that high-pitched whine in the 3rd segment?

James
x

Something I didn't realize before an explanation of dewster, is that this pan view really mesure the "motion" of the frequencies in the stereo pan. So the decay is twice flawed :
1) the decay is looped (in a sens of periodic / in time), so wrong "melody" of sound (artifical)
2) the decay is static "physically" in stereo pan, so the harmonics are not rolling / moving around our ears, but are completely inert. so wrong "presence/motion" of sound (artificial).

Correct me if I mistake.

Probably just an artifact from normalizing the amplitudes, which often messes with the noise floor. Or perhaps somehow injected during recording. I seriously doubt it exists in the AG source.


Yes, the various colored lines in the spectral pan view represent harmonics, and the vertical axis is left / right pan. So when the lines move vertically the harmonics are moving around in the stereo field. Lots of that going on with real pianos and with SN, none really with the AG.

1. With looping there is no complex evolution of the individual harmonic amplitudes themselves.

2. In something with loops as short as the AG, harmonic pan movement really isn't possible without audibly revealing the looping cycle. So they squash it by over processing the loops, but end up throwing the baby out with the bathwater.