My volume was set quite low, as I did not think it would affect the WAV file, saved directly to a flashdrive in the piano. The midi file is not very easy listening after all :-).
Did you lose too much dynamic range because of the low levels?

I can try again at normal volume if you like.
I'll need to hook up the laptop again: the piano should be able to play midi and record at the same time, but I could not get that to work.
Martijn

Not to derail this thread, but that sounds awesome. Is there another thread describing this research?

The noise floor is really low so the file should work fine, thanks!

Yes! And where else, but at Theremin World!

http://www.thereminworld.com/Forums/T/28554/lets-design-and-build-a-mostly-digital-theremin

Here's a video of a very early pitch only side prototype on my work bench where I demonstrate setting the null point, basic linearity, and the LED pitch or "tuning" indicator (the square wave audio is rather harsh sounding, I'm currently using a delta-sigma modulated triangle):



Those interested might find my Theremin Excel spreadsheet circuit simulation handy (now with transformers!):

http://www.thereminworld.com/Forums/T/28492/lc-tank-linearizing-coil-excel-simulation

Repository of files for the project for those who want to play along at home:

http://www.mediafire.com/?w36b3brqyg2g3

Theremins are often simple looking circuits that can have deceptively complex behavior. I've developed my own series resonant tank that works very well with digital logic (FPGA) and doesn't require elaborate tuning or alignment of the analog components. I'm a year into the research and I need to start building product to fund my layabout lifestyle!

I've got a big backlog of DPBSD MP3s just sitting here on my hard drive, and I feel pretty bad about not getting to them in a more timely manner. Truth is I want to review them, but the process has become rather tedious - a single review can consume most of a day's free time for me, and that's just too onerous with all of my research, volunteer work (both intentional and unintentional), etc. going on lately.

OK, so here's the deal. I'll review all of the DPs in my backlog, taking the usual snaps of the various analysis views and such, doing the usual pro/con/other text review, and pasting links to the MP3s and pix. These things are easy for me to do in an hour or two per DP. If I encounter anything unusual I may highlight it with picture or two with explanation in the caption, but that's it.

Since posts are editable I will update these "bare-bones" posts to include any user info provided to me either up-front or after the fact, but only after the user has contacted me (PM please) to do so. I'll inform everyone if/when these updates occur.

Sound fair?

Yup - please go ahead! I'm sure many of us are looking forward to your analyses!

Looking forward to it! Wish I had something to contribute after I see all this time and dedication contributed.

The PX-850 especially after all that reading I had on the similar AP-650's "rapid initial note decay" like the PX-350, even though it was stated "- Note decays are nice and long" on PX-350's review.

Sounds great, go for it!

Suggestion for the DPBSD:

I just spent some time at the local music shop, testing string resonance on several acoustic, and several digital, pianos. The digital pianos topped-out around $4K, and there were no Kawai's -- mostly Rolands and Yamahas.

I was solely interested in "pedal-up" resonance between individual undamped strings. So, two tests:

1. Hold down C below middle C (silently), hit C+E+G above middle C (that is, the 3/4/5 overtones) staccato. Listen for the harmonics of the low C (which should be excited by the higher notes).

2. Hold down middle C + G + C above middle C. Hit C below middle C, staccato. Listen for the harmonics of the low C (which should excite the higher C/G/C as 1/2/3 overtones).

All the acoustic pianos showed pretty strong resonance -- best on the little grand, but not bad on the uprights.

All the digital pianos, except for a Roland FP-4F, had either no resonance, or very weak resonance -- audible, but not significant.

The FP-4F sounded much like the little grand -- a nice strong carry for both tests.

So, my proposal:

. . . Make the existing string resonance tests _quantitative_.

That might be worth doing for the pedal-resonance tests as well.

I don't know how much work would be involved. But I suspect that harmonic resonance is one of the things that separates the sheep from the goats, in the digipiano and virtual-piano world.

. Charles

Charles, which DP models did you test?

James
x

Charles, isn't it done, yet? See the picture for the sympathetic (harmonic) resonance, pedal down (left) and up (right).

There's a picture. But no quantitative measure of how much difference there is between:

. . . Hit middle C, pedal up [dampers down]

. . . Hit middle C, pedal down [dampers up]

I've been thinking about how to _measure_ that difference, and (contrary to my first opinion) it's not trivial.

You want to know _how much sound power_ there is, in the "pedal down" case, that _doesn't_ occur in the "pedal-up" case.

Our ears are really good at hearing such differences. Subtracting one sonogram from another, I fear, isn't so easy.

. Charles

I cannot think of the strong similarities between the 2 pictures other than this is a very primitive sustain implementation with mere prolongation effect only.

SR with pedal up is much more a phenomenon to measure quantitatively, than with pedal down - where the quality of the SR implementation seems to be a more qualitative one.

Yes, I agree.


Good suggestion, but I'm not exactly sure how to do that. Here is the current DPBSD key sympathetic resonance test from the readme file:

2. [C1,C2,C5,C6v1] & [G5v15] @ t0; [C3,C4v100t12d0.25]; [C1,C2v0t14]; [C5,C6v0t15]; [G5v15t16].

- C1, C2, C5, & C6 are simultaneously played as silently as possible and held, and are given 12 seconds to decay if they do play. A brief light G5 is also played at the same time to indicate the beginning of the test.
- C3 and C4 are simultaneously played loudly.
- Two seconds later C1 and C2 are simultaneously lifted.
- One second later C5 and C6 are simultaneously lifted.
- One second later a brief light G5 is played to indicate the end of the test.

This is what it looks like on our NX:

Figure 1. Here the last three steps of the test are zoomed up in the spectral frequency view, so the horizontal axis is time and the vertical axis is frequency. You can see the C3 and C4 excitement on the left, the lifting of C1 and C2 at the cursor, the lifting of C5 and C6 one second later, and the G5 end of test signal on the right.

This is a very subtle effect that is often difficult to hear, hence my bludgeoning it with lots of opportunity to resonate and lots of stimulus. Other than reporting what it does in the various cases (in this case the lower and higher notes both respond) I'm at a loss as to how to quantify it further in a way that might be consistent across DPs. Finer tests that might produce interesting results on one DP might produce inconclusive results on another.

Key sympathetic resonance is interesting and adds to the rich sound of the piano, but the gorilla in the room is pedal sympathetic resonance, which absolutely must be implemented in some fashion in order to have a realistic DP experience IMO.

Sorry for the delay. I went back to the store (Long & McQuade, Richmond BC) today, with a notepad.

The Roland FP-4F was (sadly) gone -- it gave good results when I tried it last week -- maybe in the "Wow" category.

Test 1: Hold C below middle C silently, strike CEG above middle C. Listen for resonant sustain.

Test 2: Hold middle C / G / C silently, strike C below middle C. Listen for resonant sustain.

Results:

Young Chang 7' "Performance Grand"
. . Test 1: WOW! Test 2: WOW!

Yamaha U1H upright, refurbished:
. . test 1: good
. . test 2: good

Yamaha Arius YDPV 240 - "Natural Grand" voice
. . test 1: dead
. . test 2: dead

Yamaha Arius YDPC 71
. . test 1: dead
. . test 2: dead

Roland CK Home K RG-1-SB (using the first piano voice)
. . test 1: OK
. . test 2: good

Roland RP301-SB (voice GP1)
. . test 1: mild effect
. . test 2: OK

Roland CK Home K- KR107-MH ("Natural Grand")
. . Test 1: WOW!
. . test 2: good

Roland CK Home K HP-207 - MH
. . Test 1: ? very faint, if present
. . Test 2: Effect present, but weak

Korg SP-250 -- "Piano 1"
. . Test 1: dead
. . test 2: dead

I have access to some higher-end DP's at another shop, and will continue to experiment. Hearing that Young Chang was a real eye-opener!

. Charles

You can't shut a real piano up!

Like Nigel Tufnel's Les Paul, "you can go out for a bite and you'll still be hearing that one."

I also thought it was a "subtle effect", until I tried it on the acoustic grand! Maybe it's the big soundboard, back-exciting the open strings. Pianoteq may be hated, but it _does_ capture that sound nicely.

"Pedal resonance" is appearing on low/mid-range DP's (e.g. my PX-350). You're right -- without it, a DP's sound will be much cleaner (and "deader") than an acoustic's.

I suppose I should try to play the DPBSD test on the Young Chang, and send you a recording for analysis . . .

. Charles

I agree with Charles, that's one appropriate way to test sympathetic resonance. Testing 2 notes onee octave apart will not reeveal if all the possible harmonics are correctly generated.

If you download the Kawai ES7 manual you will see a chapter in there that explains how to fully test sympathetic resonance (including other keys beyond C2 and CEG3 as Charles is doing, which is anyway a very good reference)

I also agree that this is not secondary if we look for a real piano emulation, but of course is not that relevant if you play piano in a pop/rock band..

@dewster: welcome back at DPBSB work and a big big thanks again for all you do about it

Hey thanks for valuable URL to share my music with the world!!!

Another visit -- to Tom Lee / Richmond BC -- and more "string resonance" testing.

As before, test 1:
. . . Hold down C below middle C, strike (staccato) CEG an octave above middle C, and listen for sustain of the low-C harmonics.

Test 2:
. . . Hold down middle C / G / C above middle C.
. . . Strike (staccato) C below middle C, listen for harmonics.

Roland:
. . . RD700NX -- (1) adjustable (and strong), (2) adjustable (and strong)

. . . RD300NX -- (1) adjustable (and strong), (2) adjustable (and strong)

. . . DP90SB -- (1) yes, (2) yes

. . . KR107 PE -- (1) yes, (2) yes

. . . F120 - - (1) maybe, very faint (2) yes

Yamaha:

. . . CP-5 -- (1) no (2) no

. . . CP-1 - - (1) no (2) no

. . . P155 -- (1) no (2) no

. . . CP300 -- (1) faint (2) faint

. . . CLP440 -- (1) yes (2) yes

. . . Modus -- (1) no (2) no

. . . YDP141 -- (1) no (2) no

. . . YDP-S51 -- (1) no (2) no

. . . N3 ($$$$ !) - - (1) no (2) no

I'll write some comments later . . .

. Charles

Comments:

a) I _thought_ I heard substantial differences, model-to-model, among the Roland "supernatural" primary "grand piano" sounds. But I was listening through the pianos' loudspeakers (or outboard amp/speakers) -- it could have been my imagination. And I may have accidentally had EQ or reverb enabled.

b) The adjustability of several Roland models is neat. With the "string resonance" effect at maximum (127), they had (roughly) as much resonance as the baby grand in the showroom. With the default setting (64 out of 127), the effect was weaker.

c) I was surprised that the Yamaha N3 showed no string resonance. I'd be interested in having results on the N1 and N2.

Thanks for your forbearance --

. Charles

More (and the last chapter of) "string resonance" tests, after a visit to a Kawai / Roland showroom:

[As before: (1) hold C below middle C, strike CEG above middle C; (2) Hold middle C/G/C, strike C below middle C]

Kawai:

. . CL26: (1) no (2) no

. . CS3: (1) no (2) no

. . CN24 (1) no (2) no

. . CN34 (1) bad implementation (2) bad implementation

. . CP119 (1) yes (2) yes, strong

. . CE220 (1) bad impelmentation (2) yes, strong

. . KDP80 (1) no (2) no

. . F120 (1) yes, weak (2) yes

. . MP6 (1) bad implementation (2) bad implementation

. . ES6 (1) yes, strong (2) yes, strong

Roland:

. . RP-301SB (1) yes (2) yes

. . RP301R (1) yes (2) yes

Let me describe Kawai's "bad implementation". When I held down low C, and struck high C/E/G, the only pitch that _really sustained_ was the high C. Holding down C, and striking high E/G, only the high E sustains. Holding down high G, _it_ sustains.

Holding down middle C/G/C, and striking low C, the only harmonic that sustains is middle C.

. . . That's not how an acoustic piano behaves.

That glitch was disappointing -- these were _not_ inexpensive DP's.

Next project:

. . . Build a testing machine to quantify "action feel".<G>

. charles

Spectral panning is often discussed in these reviews. May I ask for a clarification? Is it correct to say that this term can be used in two ways:

1. The static differences in the frequencies on each stereo channel. (The direction in which the freqs remain panned throughout the length of the note.)
2. The dynamic changes in the frequencies present on each channel. (The way that some freqs may swing between channels or fade in and out of channels.)

In either case, is there research on this subject, pertaining to pianos, that someone could point me to? By which I mean studies that have analyzed notes on a real piano, looking at the degree to which the amplitude of freqs in each ear or channel seems to change as the note decays?

(Should I start a separate thread on this subject?)

From the Adobe Audition help file:

"Spectral Pan Display shows the pan (left‑right stereo) position of every frequency in an audio file, so you can visualize the sound locations in the stereo plane. The x‑axis (horizontal ruler) measures time and the y‑axis (vertical ruler) measures pan position, with the top of the display (‑100%) representing full left and the bottom of the display (100%) representing full right. Brighter colors represent louder audio signals. You can modify display settings for Spectral Pan Display using the Spectral Controls panel."

I haven't done a search, but there are likely few if any references out there with respect to spectral pan and pianos.

In terms of L/R pan, there are obviously multiple factors in play. The main one is how far left or right the physical strings being played are from the microphones (or your ears). The next would be phasing effects from unison strings playing nearly the same frequency slowly interbeating.