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

Those certain pianos wouldn't by any chance have a "CP" in the model name, would they?

Hey, thanks for the explanation, Dewster. I understand better where you're coming from on all this now.

Very cool. I should put my YPG-635 to the test.. I wonder how quickly the looping starts on it. Interested in seeing some results for the RD-700GX too

This could be a brilliant resource. If used in conjunction with Purgatory Creek (for subjective analysis), you could make better informed purchase decisions.

If someone would give me an idiots guide (and I really do mean idiot), I would do something with my V-Piano so you could compare the behaviour of a fully modelled instrument. Presumably this is still layered but limited by the technical operating spec of MIDI, ie, 127 layers?

For info I use a Macbook Pro and do have a USB flash drive device if needed so tell me what to do and I'll have a go with the V.

Thanks Dewster for taking the time to create this level of insight for us all.

Steve

dewster - I've been following and immensely enjoying the thread. When time permits (i.e. when Larry turns me loose from the next PB issue) I'll record and UL the AG N2 and the massively sampled Vienna Imperial.

Great job!

Thanks for the encouragement!

OK, I will consider adding a small musical section to the test file.

My problem is I don't exactly know what to put in there, and can't grab random MIDI from the web without possible copyright issues and such.

I'm thinking ideally:
- It shouldn't be longer than maybe 30 seconds (don't want to extend the test too much).
- It should be dynamic (low, mid, and high velocity sections, not necessarily in that order).
- It should cover most of the keyboard note range one way or another.

Anyone have any ideas? I'm willing to listen to any MIDI or MP3 files you might want to direct me to or send my way.

dewster,

I just uploaded the first piano preset "Full Concert Grand" from the Yamaha Motif XS8. It will be interesting to see, as you get more Yamaha DPs, how Yamaha changes their sampled CFIIIS for the various DPs.

http://www.mediafire.com/?sharekey=5da74615b2837725c79b87b207592a1c0742218a332edd73a543906a5faff527

You should feel encouraged! A great response to your project!

A generic MIDI file won't work, in terms of assessing quality, especially if your goal is to compare pianos. The velocities will be inconsistent from piano to piano. That's one of the big problems with purgatory. Each instrument (DP or samples) needs to be played, to create its own unique MIDI file - to be used effectively only to play back the DP or sample in question.

I do agree with you Lawrence. Some MIDI velocity shifting / scaling should have been applied to many of the samples over there.

But there probably isn't a lot of harm in adding some short musical snippet to the end of the DPBSD file, particularly if it is kept fairly short.

Hi setchman,

I reviewed them today and uploaded your MP3 files to the archive. The Steinway in particular is very nice! Not sure what's going on with the partial pedal in both, very odd.

Also, I seem to be running across a fair amount of sound files that I can't do any spectral analysis on at all. It's either a bug in Audition, or extensive phase processing of the sample set (I have suspected the latter even before seeing this issue as it makes sense). Here is a picture:



Makes my job a lot harder. The Vienna Grand was one of these strange phase files, the Steinway was not.


---------------------------------------------
- Galaxy II Vienna Grand (download edition) -
---------------------------------------------
FILE & SETUP:
- dp_bsd_v1.3_galaxy2DE_viennagrand.mp3
PROS:
- Nice long decays, no looping.
- Wide dynamic range (~53.5dB, vel=1:127).
- 7 or 8 velocity layers (from waveform view).
- Fairly good layer matching.
- Some kind of sympathetic resonance going on.
CONS:
- Obviously stretched.
- Stretch distances: 2(x17),4,2(x7),3,2(x10),3,2(x5) = 41 groups.
- No key-up or pedal samples.
- C9 sounds pretty weird.
- No response to partial pedaling.
- Bizarre partial pedal event: pedal 50%=>75% note sounds before being played!
OTHER:
- Linear phase (?) prevents spectral inspection.
- Volume in MP3 file very good.
- Date reviewed: 2010-02-06



-----------------------------------------
- Galaxy II Steinway (download edition) -
-----------------------------------------
FILE & SETUP:
- dp_bsd_v1.3_galaxy2DE_steinway.mp3
PROS:
- Nice long decays, no looping.
- Wide dynamic range (~46dB, vel=1:127).
- Fairly good velocity layer matching.
- 8 velocity layers.
- Velocity switch @ vel=45, 52, 70, 80, 90, 102, 112.
- Very nice sympathetic resonance.
CONS:
- C2 sample just abruptly ends.
- No key-up or pedal samples.
- Obviously stretched (spectral phase view).
- Stretch distances: 2(x44) = 44 groups.
- No response to partial pedaling.
- Bizarre partial pedal event: pedal 50%=>75% note sounds before being played!
- A bit of strangeness near the noise floor.
OTHER:
- Volume in MP3 file very good.
- Date reviewed: 2010-02-06

Incoming:

-------------------------------
- Yamaha YDP223 Grand Piano 1 -
-------------------------------
FILE & SETUP:
- dp_bsd_v1.3_yamaha_ydp223_gp1.mp3
PROS:
- Long note decay.
- Looping isn't too obnoxious.
- Huge dynamic range (~65dB, vel=1:127).
- I believe this is a very smoothly blended 3 layer sample set (spectral pan display).
- Layer switch (?, switch not audible) @ vel=80, 94.
- Responds to partial pedaling.
CONS:
- Obviously looped.
- Lower looping rather realistically wobbly, higher looping unrealistically static.
- Samples are 3 to 1 seconds (lo to hi) with 1 to 0.5 seconds of crossfade.
- Obviously stretched, group transistions farily audible.
- Stretch distances: 2,3(x28),2 = 30 groups.
- No real variation in timbre at higher velocities.
- No obvious sympathetic resonance, though there are subtle differences w/ pedal down
- No key-up or pedal samples.
OTHER:
- Date reviewed: 2010-02-06

Just reviewed it and archived the MP3.

Not very impressive, particularly for a newish instrument currently selling for $3.6k. The decay times are fairly short. Otherwise typical sample times, layers, stretching, symp res, etc. for Yamaha (and by that I mean typically mediocre). The P-120 tests better, which is rather shocking.

---------------------------------------
- Yamaha Motif XS8 Full Concert Grand -
---------------------------------------
FILE & SETUP:
- dp_bsd_v1.3_Yamaha_MotifXS8_FullConcertGrand.mp3
PROS:
- Looping isn't too obnoxious @ lows & mids.
- Decent dynamic range (~33dB, vel=1:127).
- 3 layer sample set (waveform & spectral phase displays).
- Layer switch @ vel=70, 106.
- Sympathetic resonance, though it's fairly fake sounding - echoy and reverby.
- Key-up samples of some sort, though it sounds like a strange tone.
- Responds to partial pedaling.
CONS:
- Fairly quick note decay with some buzzing near the noise floor (could be a recording issue).
- Obviously looped.
- Samples are 3 to 1 seconds (lo to hi) with 1.5 to 0.5 seconds of crossfade.
- Lower & mid looping rather realistically wobbly, highs unrealistically loopy.
- Obviously stretched, group transitions fairly audible.
- Stretch distances: 2,3(x10),2,3,4,3,3,4,2,4,2,3,1,2,1,2,3(x5),5 = 31 groups.
- Obvious velocity layer switching, some kind of filter switch as well @ v=50.
- No pedal samples.
OTHER:
- Volume in MP3 file very good, though dynamic range seems like it may be limited.
- Date reviewed: 2010-02-06

If someone out there could run the DPBSD MIDI file thru their Korg SP250 and/or LP350 maybe we could resolve once and for all whether the sample sets are the same or not.

If the file is going to include the use of the hold/damper pedal, again, it will not translate from piano to piano. I know that in the (bad) old days, I would record (MIDI) a piece using one sample, and then try and play it back at a later date with a new and improved sample. Yikes. I remember often having to edit velocities like crazy AND changing the timing of ped ON and OFF messages. Not fun.

In my old age, now, I simply record again. Way better than doing a ton of editing. (I even dislike a bit of editing now). Of course, sometimes the original version was an improvisation, and "spot on" and it's difficult to replicate the magic. (Magic. Part of what made the improv so good in the first place).

Lawrence

I've been in the MIDI controller 64 time nudge heck myself, so I know from whence you cometh, Lawrence.

Sonar has pretty good MIDI filters though, so you just select them all, drive them high or low, and pull them back a bit. The continuous damper controllers are the worst IMO.

Very interesting test (and much work I suspect - my hat's off to Dewster for undertaking this gigantic project - it will be very useful to we DP types).

In regard to Pianoteq and dynamics - Dynamics are adjustable.

If a midi file has a large dynamic range, lowering the setting will reduce the overall dynamic range when the file is rendered to wave. Rendering means Pianoteq converts a previously recorded midi file to a wave file (takes much less time than playing and recording). An advantage of this is that even if the max peak is -30dB in the rendered wave, it can be maxed with no loss in sound quality (the soundcard is not used).

There is also a Limiter which is used to prevent clipping when the output hits the ceiling.

Glenn

Hmmmm... Can you successfully "play" these files in Audition? Maybe the MP3 decoder used by Audition fails? You could try converting the MP3 to WAV externally and then load the WAV file into Audition. You could also try the demo version of iZotope RX. I'm using that myself sometimes for spectrum stuff...

Love the DP BSD project, btw! Can't wait to see Kawai, Roland and Yamaha CP1 results...

Do you want a Kawai MP-5 sample?

If so, I will try to find time to send one, tomorrow (Aussie time)
If you already have one, ignore this post

Rob

I use MIDI to wave rendering a lot in Pianoteq, one of it's best features IMO. Another advantage of this feature is you never get glitches or lost notes on older PCs. Being non-real-time, it's free to take as much or as little real CPU time as necessary to get the job done.

Yeah, they play fine, and the waveform view works fine as well. For these types of files I have to rely heavily on envelopes and my ear to detect stretch and layer groups.

I've only seen this collapsed phase anomaly in PC samplers so far, not in real DPs. Two of the MP3 files I made myself in Audition, one I received from a poster. Here is the list:

- Galaxy II Vienna Grand (download edition)
- VintAudio C7 Close Miked 6 layer
- Windows Media Player

OK, I just went back and re-recorded the Windows Media Player piano *ugh* through my Echo Mia soundcard (SPDIF I/O) and the collapsed phase anomaly exists in the wave file before conversion to MP3. So it seems almost certainly something that exists in the sample set.

Extensive phase processing may help blending, particularly for subtle things like sympathetic resonance of multiple pedal down notes playing together. It could help with velocity layer blending too but, outside of the Yamaha YDP223 so far, I don't see evidence of anyone doing that.