Thanks again, dewster! That explains my frustration with trying to play pianissimo. My next piano will be much more dynamic.

The VSL looks pretty awesome. I wonder, can its dynamic range be increased beyond the default 45dB?

Interesting. The groups appear to coincide largely with the pattern of semi/whole tone intervals.

2(x3) = F,F#,G,G#,A,A#

This means the C major is almost free of repeated samples.

Hey dewster,

The last time you reviewed the Roland HP-307 was with v1.4. I don't know if you care to re-review it with your v1.7 DPBSD, but I ran the test as my first exercise in recording audio.

The file is here if you decide it is worthwhile to re-review - hopefully the audio levels are okay:

http://wmsar.info/dewster_dpbsd/

I printed out your latest "readme" file, so now I need to study it.

By the way, as discovered in earlier tests, it doesn't seem that some of the effects on the HP-307 are necessarily active when playing a MIDI file, but I haven't messed around with it very much.

You were waiting for the MIDI spec to be published, and I recently pinged Roland and they said it would be released as soon as it is ready. Hopefully there is a way to program some of the effects within MIDI.

For those interested, I played around with the HP-307 and Test #2 "Key down sympathetic resonance test" and finally have a MIDI file with the necessary conditions/data for the test. I've updated the webpage below:

http://wmsar.info/dewster_dpbsd/

Wow nice discovery!
Would it be possible for you to write a short guide as to how to change regular midi files to contain data required for all features to kick in?

A month ago or so I wrote to polish dealer regarding the issue but they weren't really helpful. They don't even have their own pianos to test things and had to ask some hungarian branch or something. Either polish branch couldn't pass my message or the other branch could not understand what I asked. The reply was some total crap:

"The piano desinger function only works with Grand Piano 1.

If you are not playback with this sound , you will lost your saved settings like damper resonance.

Because the damper resonance (in Piano Desinger) is a Performance Settings, not stored in the song."

After several emails I saw there's no way for them to take this serious and find out what's the issue (I even wrote a step by step guide how to hear what doesn't work...) so I just gave up.
I think I should have taken it to a brittish or american branch directly.

The piano designer works with Grand Piano 1, 2 and 3. If you use the designer you will find that the piano has defaulted to GP1 but any adjustments you make are applied equally to the other pianos...try it and see what you find. A criticism of the HP-307 is that in my opinion there is insufficient tonal variation between GP 1, 2 and 3. Luckily they have done a great job with it and are likely to please most people most of the time. But, something DIFFERENT might have been nice.

Steve

Maybe I wasn't clear enough - the issue was with some piano designer settings (or features) not working at all while playing back a midi file. When you play any normal midi file damper resonance and damper noise isn't part of the generated sound (as if set to 0/10). But when you record something and play before storing as midi file it works as intended. There was a discussion about it some pages ago with conclusion that before midi implementation sheet is published we don't know how to force the piano to playback including damper resonance and noise.

I understand jmmec found a way to control these things directly in midi file so it may be possible now?

BTW dewster do you intend/want to test all previous pianos each time you make a new version of your midi file? Maybe for consistency it would be good but it sure is a lot of work for you?

Most sample libraries use the amount of data as an advertising argument. And most contain a lot of duplicated redundant data ;-)

I think, that is the most reasonable explanation ;-)

Peter

"Duplicated" data is a sinch to "compress". "Duplicated redundant" data can even be disregarded. So what you're saying is that, hypothetically, the sample library could contain a petabyte of uncompressed data occupying only 50GB? I think they've been rumbled.

Regarding my Test #2 analysis for the Roland HP-307:

I'm a newbie to MIDI and there seems to be a mistake in my analysis - I think I should have seen SYSEX events for Test #2 if the Roland HP-307 actually sends proprietary MIDI data for the "string resonance" setting. But the MIDI file didn't capture this; or Cubase doesn't save it if it was sent.

As a test, I turned on the HP-307 this morning and ran the MIDI file I captured last night without making any changes to default settings. The MIDI file played back with damper pedal! I then made the change to "string resonance", and I got the expected results. So something is screwy. (Edit: last night, and this morning, when I set "string resonance", it seems to have configured the piano such that the MIDI file would create the expected result on playback.)

I have to go to work, but I'll try to play around with this more; or else I'll wait for Roland to release the MIDI spec. So ignore the Test #2 MIDI file.

Interesting, I didn't see that pattern. They deviate slightly from it in octave 5.

A #B
1, 2,

C #D #E F #G #A #B
1, 2, 2, 1, 2, 2, 2,
1, 2, 2, 1, 2, 2, 2,
1, 2, 2, 1, 2, 2, 2,
1, 2, 2, 1, 2, 2, 2,
1, 2, 2, 2, 1, 2, 2,
1, 2, 2, 1, 2, 2, 2,
1, 2, 2, 1, 2, 2, 1,

B C
1,1

I think a lack of layer switching at lower velocities is also an issue.

This is something of a work in progress, with changes to the MIDI file as I run across features that can be definitively tested. The main tests for looping, stretching, layer switching, sympathetic resonance, and partial pedaling have been there since the very first version.

Like others here I'm mostly interested in the latest offerings, as they hold the most promise. But I'll redo any important instruments if people feel like redoing MP3s.

I would not have expected there to be any stretching in the VSL in a million years. To me, the following sentence quoted from the product info strongly implies that every key has been sampled:

"1,200 recorded samples per key represent a magnitude of sampling detail that has been unthinkable up to now."

I guess this actually means that for the keys that they did sample, 1200 samples were taken.

Greg.

Regarding VSL VI, the interface has a dynamic range slider. The analysis mp3 file was recorded with the slider at 50%. At 100%, the quietest note possible is much much quieter.

"Dynamic Range controls the difference in loudness
between velocities of the piano and essentially is
similar (though not identical) to a compressor. The
values which yield results comparable to the sound
of the original piano are located at approximately
55–60%, and are therefore set by default. At 100%,
the samples are played back as they were recorded;
at lower values, keystrokes with lower MIDI velocity are played back louder, so that the piano’s dynamic
range is reduced."

If they really are using a different sample for each velocity, then if they stretched each layer differently you might not easily notice it (wild speculation based on zero evidence).

But all stretching is bad stretching, I don't want any, and it's pretty weird that the VSL has it.

I think I'll start putting the dynamic range measurement in the OTHER section, rather than the PROS or CONS. I shouldn't be giving the impression that more is better.

I'm not sure what the optimal amount would be as it depends on the instrument, how loudly / softly it is played, and how MIDI velocity maps to that.

I would think as long as you can play evenly with whatever keyboard you use, then more is better. Given good technique, once it becomes too difficult to control, then less is better. It'd be interesting to measure the dynamic range of an acoustic piano for comparison.

For me, I think the dynamic range of the timbre is more important than the dynamic range of the amplitude. With a large change of timbre from pp to ff, I often prefer a rather small amplitude dynamic range, because I like to hear the pp timbres quite loudly - it produces a very warm sound, with the expression being provided largely by the change of timbre. Also, small changes of amplitude are easier on the ears.

Greg.

Yes, my thoughts exactly.

It's trivial to change amplitude, and much harder to change timbre, so large dynamic range shouldn't be emphasized as necessarily a good thing.

dewster,

For the betterment of humanity, I did more investigation on the lack of 'sympathetic resonance' for the Roland HP-307 in the DPBSD test. I figured out the MIDI events to enable it, and modified v1.7 of the DPBSD MIDI file.

A new MP3 with sympathetic resonance, and the modified MIDI file to enable it, are at the website below in case you are interested in analyzing it:

http://wmsar.info/dewster_dpbsd/

For Test #2 your "readme" states "Listen for a brief weak G5 note at the beginning of the test, possibly accompanied by the very faint sound of C1 & C3 & ...etc...". I can clearly hear the resonance of the G5 & C-notes when playing the test on HP-307, but it is very difficult to hear in the MP3 [at least for me]. You'll notice a big difference when the loud C2 is played. I don't know how other tests may or may not be affected.

If you do decide to analyze the file, then please let me know if other problems still persist; I think they were pedal up/down & key up?

NOTE: I suspect there is more to sympathetic resonance than just the 'string resonance' setting in "Piano Designer", especially since there are several parameters related to resonance alone. All parameters were at their 'default' setting, with "Reverb" disabled.

I agree that less amplitude dynamic range is better when playing with headphones for the sake of my hearing, but otherwise I want both timbre and amplitude to have wide dynamic ranges. I don't want to sound like a digitally compressed recording. I want to sound live.

I'm able to hear key down sympathetic resonance in there, though it is fairly subtle.


- Pedal and key noises are missing.
- It passes the pedal down silent replay test.
- It passes the quick pedal partial damping test.

The big thing that's missing is pedal down sympathetic resonance - I'm straining and looking but can't detect any.

Also your levels could be better, the peak is -8.65dB, and the noise floor is around -72dB.

Thanks for your efforts to capture this, the SuperNATURAL piano sound is very promising technology.

dewster
I was reviewing mp3 of Yamaha CP1, and I think you missed something important. It has 5 layers, not 4, and the last layer switch is quite audible, not only visible.
With 50 stretch groups, it seems that CP1 has sample set very similar to CLP380.

Here is the picture with layers switches marked:

But isn't that purely conjecture? I'm interested in the seeing the CLP380 mp3, myself. Anyone got one?

Of course, i'm just guessing. I said "similar", not "the same".

Hmm. I looked at it again very closely and the second layer switch you have marked is really subtle - kind of a judgment call. But, as you said, you can hear the last switch. I'll update the review text - thanks!

Funny too how the CP1 has essentially no timbre change from velocity=1 to velocity=54. And the dynamic range is only 38dB. I guess it makes sense to put more transitions on the higher velocity end, where the piano really start going non-linear.

If anyone is interested, here is a highly compressed version of the CP1 layer test with the amplitude variation removed, so all you hear is timbre variation:

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

What a bullshit. Shame to Yamaha for lying to its customers.

This is why the DPBSD project has BS in the title.

It's technobabble like that that pretty much forced me into doing the project in the first place - at some point I just couldn't take it any more.