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

Yes it could be that. When reviewing the layer test I hear certain features that become more prominent with increasing velocity. There's the hammer "thunk" sound that gets sharper, and the "clang" sound that I believe is the longitudinal mode becoming more excited. Perhaps they have separated these out somehow for synthesis purposes and are reusing them for a small stretch of notes.

It could also be something as simple as a software error - maybe they have all 88 attack samples in there for this layer, but some spurious entries exist in a stretch table?

My apologies in advance ... I can not get the search to work in this forum ...

Clavinova 340 - check
Laptop - check
USB cable - check
I can download the demo version of Adobe as suggested in the readme - check

Now what?

Is there any screen shot by screen shot type documentation? Do I need to actually play something on the piano? Do I just "play" the MIDI file and it does the rest?

I would love to contribute to this study ... thanks.

You need to play the midi file. You can do this by playing the file with an adequate player on the notebook or I guess, the Clavinova can play a midi file from an USB stick.

To record the sound, you need an additional audio cable to connect the audio out from the Clavinova to the audio in on the laptop. With Audition you can record and save the sound. The record volume is important. This is described in DPBSD readme.

theoak - I sent you a PM...

Generalmusic pRP-800

Fellow Pianoworld DP forum-ite M.Schreck was kind enough to provide us with two DPBSD MP3s of the Generalmusic (GEM) pRP-800 - thanks so much M.Schreck!! The MP3s are of the Steinway D and Fazioli F308 voices. The Fazioli has shorter attack samples so I decided not to do a specific review for it, but the MP3 is at the share point along with the Steinway for the curious.

GEM instruments, at least here in the US, are like the DP equivalent of Sasquatch - no one has really seen one, they are exotic (purportedly a blend of sampling and modeling), and they are shrouded in mystery. So it's great to have an intimate sampling of this rare breed of DP.

At its core, the GEM pRP-800 presents as a sampled DP with attendant garden variety looping and stretching. The attack sample lengths are fairly long but the loop samples are quite short, and I can distinctly hear the looping on the lower notes. Stretching is a bit more than normal for a modern DP, and is audible to me also over the lower notes. Decay time is rather short, particularly for the upper notes.

GEM reports that they use DSP to do additional things like sympathetic resonance, which I have absolutely no reason to doubt. I was expecting to hear more in terms of key down sympathetic resonance - I can clearly hear it in the GEM video so I suspect my test isn't stimulating it very well. This section of the DPBSD could probably use some tweaking to better accommodate the various algorithms found in DPs.

There is a smooth change in timbre with increasing velocity with no detectable velocity steps, particularly in the upper 2/3 range of velocity - below that there isn't much change in timbre.

It fails the pedal down silent replay test @ pedal up (the note plays at velocity = 1). It also fails the quick pedal partial damping test in a rather odd way - a brief damping with the pedal in a real piano removes some energy from the string so there is a quick drop in volume but otherwise normal decay thereafter. In the GEM the decay rate remains high even after the pedal is pressed back down (i.e. after damping is removed).

I'm never sure what to make of dynamic range, which is why it isn't in the PRO/CON section anymore. I used to think bigger was better, but now I'm thinking there is a "just right" range. The 33 dB of the GEM sounds somewhat compressed to me.

Much thanks again to M.Schreck for removing the veil from this interesting DP!


Review below, MP3 and more analysis pics at the share point.



Waveform view of the looping test with vertical zoom applied. Decay times are rather short, looping is also visible here on the two lowest notes C1 & C2.


Spectral pan view of the note C2. Attack & loop samples are quite clearly seen here, loop period is rather short.


Spectral phase view of the stretch test, mid notes. 43 samples cover 88 notes.


Spectral frequency view of the layer test. Timber variation is smooth in the upper 2/3 of the velocity range, rather static below that.


--------------------------
- GEM pRP-800 Steinway D -
--------------------------
FILE & SETUP:
- dpbsd_v1.7_gem_prp800_steinway_d.mp3
- GEM headphone out to Mac line-in, recorded with Reaper, converted with Max.
- Recorded by "M.Schreck".
PROS:
- Passes the pedal down sympathetic resonance test.
- I believe it passes the key down sympathetic resonance test, the response is very subtle.
- Passes the partial pedaling test, the effect is realistically slightly buzzy.
- No obvious layer switches, not much timbre change until 1/3 max velocity, then smooth change with increasing velocity.
CONS:
- Fails the pedal down silent replay test @ pedal up (note plays vel=1).
- Fails the quick pedal partial damping test - oddly, brief damping speeds up decay rate.
- Decay times are somewhat short, particularly in the higher registers (~0.8 to 0.2 Pianoteq, low to high).
- Attack sample lengths are (C1:C8): 5.5,4.2,4.6,4.0,2.6,2.0,?,? seconds.
- Loop sample lengths are (C1:C8): 0.9,0.8,?,?,?,?,?,? seconds.
- Obviously looped, both visually and audibly, could benefit from longer decay loops.
- Obviously stretched, group transitions fairly audible.
- Stretch distances: 5,2,1,2(x4),1,2,1,2,3,3,1(x3),2,3(x3),2,1,2,1,3,1,2(x3),3,2(x3),3,2,2,1,2,3,1,2,2,3 = 43 groups.
- No pedal up/down or key up sounds.
OTHER:
- Dynamic range 33dB (vel=1:127).
- MP3 levels: peak @ -4.9dB, noise floor @ -82dB.
- Date reviewed: 2010-06-29.

DPBSD Version 1.8 Released

Yet another fine tuning of the key down sympathetic resonance test, and a slight tweak to the silent replay test.

Please use this new MIDI file (located in the root directory of the DPBSD share point) instead of all previous versions.

====================
= Revision History =
====================
v1.8 - 2010-07-04:
- Changed key down sympathetic resonance test, stimulus is now C3 & C4, then C1 & C2 lift together, then C5 & C6 together.
- Shortened pedal down silent replay test so events happen every second, rather than every two.
- Modified peak recording level (test 0) to accomodate above changes.

>100K VIEWS

The DPBSD thread has had over 100,000 views - thanks everyone for the interest! And big thanks to Pianoworld for hosting this forum and rashly giving me free-run of a thread!

This project would have died long ago if it had to rely solely on my own piddly handful of sample files, so huge thanks to those individuals who went way out of their way in order to provide DPBSD MP3 files of the various DPs reviewed! (In order of the number of samples reviewed, then in alphabetical order within groups, please let me know if I've missed anyone):

= The DPBSD Honor Roll =
10 pesk
7 setchman
3 jve
2 ChrisA
2 CyberGene
1 bkmz
1 EssBrace
1 Glenn NK
1 Goofball Jones
1 kawaian/mucci
1 M.Schreck
1 Mawima
1 Melodialworks Music/Lawrence
1 nan
1 NikkiPiano
1 pkdd
1 R0B
1 sandord
1 sdw91
1 Voltara
1 Volusiano
1 Yuri Pavlov
1 zaba19

Woo-hoo! Cake and champagne for everyone!


Get a snootful and tickle those Kit-Kats! (They're triple sensor with escapement and ivory feel milk chocolate feel!)

Dewster, what you've done with this project is a tremendous work! Thank you and keep on with the analysis! Even if they will not admit it, I am sure you have made at least some impact on digital piano manufacturers. (Roland SN-piano seems to be suspiciously close to following your criterion for example).

Dewster, It's a great achievement!

Best wishes,

Steve

Thanks CyberGene! People have been aware of and complaining about old-school sound problems for quite a while now - to their credit, Roland stepped up to the plate and addressed them head-on. Time will tell if there are any long-term issues with their particular implementation.

Thanks Steve! And thank you both for the MP3s!

Thanks for your work dewster!
I came to this forum roughly a couple of days after purchasing my hp307. Shortly after that I saw your thread where you attempted to technicaly evaluate the piano sound generating systems of digital pianos. At first I was reluctant to do it ("Be warned however, we'll may sit around taking endless pot shots at your expensive shiny new DP!" ) but then I thought "oh what the heck, let's hear it! this is a really good piano afterall!") Little did I know what effects would the test bring

That's the spirit!

The truth (as best we can ascertain it without help from DP manufacturers) wherever it leads (including inconvenient truths and catcalls from the peanut gallery).

congrats also from my side!

Thanks mucci, your Kawai CA63 was a very nice addition to the DPBSD project.

I, and I'm sure others, appreciate the effort you put into it!

I just watched the Roland SuperNATURAL piano video again - it's like a breath of fresh air to my engineering/musical self.

Those guys really kick ass and take names when it comes to stretching, layer switching, and looping. They let it all hang out for literally anyone to see and hear. I couldn't have done a better instructional video myself, quite amazing.

The only thing I would add would be to talk about partial pedaling and sympathetic resonance, both of which the SN piano does quite nicely.

Roland, please make a stage piano with PHAIII (I would prefer straight plastic, non-ivory feel), several SN pianos (Yamaha, Steinway, Bosendorfer), good strings, good harpsichords, and your positive pipe organ (preferably thrown in, but I would gladly buy it as an expensive expansion board). I sincerely beg you to make such a product.

Speaking of Roland SuperNATURAL piano, I noticed some posts over at Roland Clan Forums for the RD-700GX that discuss problems rendering MIDI. Link 1 Link 2.

This has been an issue for the DPBSD project and the HP-307 because key up, pedal up/down, and sympathetic resonance sounds don't work for MIDI playback unless you enable them via SYSEX, as jmmec discovered. For the RD-700GXF the key up and pedal down sounds are missing with MIDI playback.

For the RD-700GXF, it seems all you need to do is change the "Part Mode" from "16PART+PERF" to "16PART" (see page 111 of the user's guide) and the MIDI should play back using the piano designer settings.

If anyone can try this and report back I'd be very interested. I'd also be very interested in an updated DPBSD MP3 of the main piano in the RD-700GXF with the key up and pedal down sounds enabled in this manner, as that was the only CON in its DPBSD review - and it would be pretty amazing to have a DP with a clean bill of health in the technical sound department!

Yamaha P-85 Review

Martin C. Doege was kind enough to provide us with a DPBSD MP3 of the Yamaha P-85. Big thanks to Martin! This is another DP I've been anxious to review since the beginning of the DPBSD project, mainly due to my curiosity regarding the performance of the supposed single layer sample, and also due to the constant praise heaped on this DP by a certain member of the forum.

Martin re-ran the test and was able to increase the SNR by ~6dB, so if you grabbed the initially posted MP3 file for listening purposes you might want to go to the share point and get the latest one as it is a substantial improvement.

But after the long wait it seems there aren't too many surprises. The timbre variation with velocity is nice for a purportedly single layer instrument, though I found the timbre variation limited more to the middle velocities, with less variation over the lowest and highest velocity ranges. It passes the silent replay and brief damping tests like a champ, and it has a serviceable partial damper effect.

As for the negatives, I couldn't detect any kind of sympathetic resonance associated with the damper pedal or keys held down. With 30 samples covering 88 notes there is a fair amount of stretching going on for the main voice in a modern DP. And I could hear the rather short loop samples pretty clearly. During the analysis I didn't encounter any mechanical noise samples for the keys or pedal.

Decay times are somewhat short, and Martin raises the good point that my calibration of the DPBSD to Pianoteq is a rather hazy specification - and I agree, the vagueness of the timing of this test is something of a weak area for the DPBSD. In my defense though, the lowest note is given 30 seconds of decay time, which doesn't seem overly onerous and jibs fairly well with our acoustic grand. I've tweaked these times in the past and may do so again at some point, as realistic decay time to me is very important.


Review below, MP3 and more analysis pics at the share point.



Waveform view of the looping test with vertical zoom applied. Decay times are rather short.


Spectral pan view of the looping test, note C2. Attack and loop samples are clearly seen, cursor is located at the transition point, looping is fairly audible.


Spectral phase view of the stretch test, mid notes. 30 samples cover 88 notes. Low and high note ranges display the same groups-of-3 stretching.


Spectral frequency view of layer test. Timber variation is smooth with velocity over the middle range, with less variation at the lowest and highest velocities.


---------------
- Yamaha P-85 -
---------------
FILE & SETUP:
- dpbsd_v1.8_yamaha_p-85.mp3
- Sequenced with MU.LAB, MacBook line-in, recorded with Audacity.
- Recorded by "Martin C. Doege".
PROS:
- Passes the pedal down silent replay test.
- Passes the brief pedal partial damping test.
- Responds to partial pedaling.
- Timbre variation is fairly smooth with increasing velocity, though most of the effect is restricted to the mid velocities.
CONS:
- No visible or audible pedal down sympathetic resonance.
- Fails the key down sympathetic resonance test.
- Note decay times are somewhat short (on the order of 2/3 to 3/4 Pianoteq).
- Stretching is visible over the entire range, audible over the lows and mids.
- Stretch distances: 2,3(x28),2 = 30 groups.
- No obvious pedal up/down or key up samples.
- Obviously looped, both visually and audibly.
- Attack sample lengths are rather short.
- Attack sample lengths are (C1:C8): 2.0,2.0,1.8,1.7,1.3,1.1,?,? seconds.
- Loop sample lengths are short and fairly audible over the lows and mids.
- Loop sample lengths are (C1:C8): 0.6,0.6,0.6,0.4,0.3,0.3,?,? seconds.
OTHER:
- Purportedly a single velocity layer sample set.
- Dynamic range 47dB (vel=1:127).
- Notes played @ vel=1 produce no sound.
- MP3 levels: peak @ -1.3dB, noise floor @ -51dB.
- Date reviewed: 2010-07-20

Just wondering what the procedure is for getting a particular thread a sticky? With the info available in this DPBSD thread I think it might help if it were on the front page for any newcomers to see initially, and of couse easy access for those of us that like to reference it.

Thanks to Dewster and Martin for this latest entry.

It's actually a good sign for a thread when you never see a message from a moderator--- it usually means tempers have gotten away from members and the thread has gone into a death spiral, with the posts edging toward being actionable.

But, to make a thread sticky (and always at the top of the list), a mod has to do it. I guess the 'Notify' button at the bottom of the thread window would do it. They might do it just out of surprise at getting a "Notify" that wasn't a complaint about deportment.

thanks clef,

I sent a "notify" note requesting the sticky and we'll see from here.