Roland HP-307 Update - MIDI Playback:

Okay everyone, the website is updated with the latest and greatest understanding of the HP-307 MIDI playback behavior. No promises that this is 100% correct, but I promise that I'm not making it all up:

http://wmsar.info/dewster_dpbsd/

The bottom line is that I'm able to enable all the effects that I played with (the main effects).

Sometime this weekend, I plan to run a new DPBSD test.

Regards

Really awesome work jmmec
As for duplex scale - you can hear it when you strike the low bass notes. There's a clear distinction between on and off. If I recall correctly you can play the demo (I mean the demo songs of piano designer under [Twin Piano]+[Functions], p.19 in the manual) that will demonstrate it for you.

Ah! I had forgotten about the 'demo' which is very useful.

I just played around with 'duplex scale' and I hear it very clearly now. doh....

I also need to modify my impression of "key off resonance" since it extends across the keyboard, and not just from Middle-C and lower.

Thanks

Listening to so many DPs rendering the DPBSD test gets me to thinking how might be the best way to make a realistic piano-like sound with a combination of samples and modeling. Here's what I would try:

1. Record a real piano with only one string per note allowed to vibrate (either damp the unison strings or remove them).
2. Interpolate in the frequency domain between 4 or more velocity layer attack samples per note (recorded in step 1).
3. For each note make multiple waveguides: one each for the transverse vibration of each unison string, and at least one more for the longitudinal vibrations.
4. To play a note, excite its waveguides with the corresponding recorded & interpolated attack sample.
5. Couple the waveguides together to create sympathetic resonance.

Voila! Small sample set, realistic inter-beating of unison strings, unlooped natural sounding decay, no velocity switching, all the sympathetic resonance you could ever want.

You forgot the soundboard in your model :), but recording the attack/response and playing it back sounds like commuted waveguide synthesis. There's a similar idea here, if you haven't already seen it:

https://ccrma.stanford.edu/~jos/pasp/Piano.html
https://ccrma.stanford.edu/~jos/wav/pno-cs.wav

Yes, thanks for the pointer, saw that back in my grad school days. Julius O. Smith has some awesome web pages on taking the Karplus-Strong algorithm to new heights. His initial paper in CMJ on that is a must-read.

Since the soundboard response is part of the note sampling, it would naturally be included - or in the parlance, commuted.

A long time ago I was working on someone's electronic stage piano. It was analog and had one oscillator and one envelope generator per note, along with some filtering. The problem was some keys weren't responding, and it turned out that some connections to the tiny individual PCBs were slightly oxidized. Pulling them out and cleaning the contacts with an eraser fixed it. But while I was in there I played with it a bit, and discovered a pot on the oscillator board that adjusted the frequency (ugh). I detuned one note 1/2 step and played it along with the note next to it and the sound became much more piano-like. Amazing how little it takes to improve a simulation.

Roland HP-307 & MIDI Update:

Hey dewster,

There are 3 new MP3's at the website below, where two use a newly modified DPBSD v1.7b. At last, the recording levels should be okay, but please do confirm.

There was a new critical discovery about the HP-307 and MIDI playback. Pretty frustrating and a waste of 6+ hours.

You'll see that the 'pedal down sympathetic resonance' test should pass now, though it is much more obvious with "Damper Resonance = 10" instead of the default setting (5). But even at the default setting, the "sqiggly lines" clearly show a difference.

http://wmsar.info/dewster_dpbsd

The webpage above will be updated in the next day or two with new information and corrections, and also include the modified DPBSD for the HP-307.

regards

The noise floor is excellent, but the peak level could use another 5 dB or so.


Didn't you get the memo? Graphs of squggly lines are meaningless!

I'm hearing both pedal down and key down sympathetic resonance. No pedal down sound effects though.

heh... I also discovered the "soul" of the HP-307 and it is MIDI Channel 4. My fear is that some will argue that the person-hood of the HP-307 is being violated by revealing this truth (or is it a... bug?). One can never be too careful in avoiding the sensitivities of others, especially in these days and times.

There was clipping on the modified v1.7b Test #1, so I turned things down to stop that section from clipping. So just turn it back up and live with Test #1 clipping on future recordings?

I can improve the levels for the MP3 for the baseline DPBSD v1.7 since that didn't have clipping since 'sym resonance' is disabled (but this was recorded last, so it used the same, lower levels, as the other 2 MP3's).

This may be outside the scope of your intentions, but any other thoughts about the 'damper down' and 'key down' sympathetic resonance tests? Do you consider the sound good, decent, bad, indifferent, unexpected, revealing, mysterious, profound, awful, sickening? Do you need a better recording?

When it comes to applied science, any appeals to pure subjectivity can and should be disregarded.


Turn it up to the point where clipping is just about to happen, with maybe 1 dB headroom left over. When I open the full DPBSD MP3 with pedal down sympathetic resonance set to 5 (dpbsd_v1_7b_Roland_HP307_DR5_jmmec.mp3) Audition tells me the peak level is -5.53 dB. So you could turn that test up 5 dB more without clipping. The noise floor looks very good now, with only +/- 1 bit of noise, so the peak level is less important than it was in your previous recordings. There is some kind of mosquito noise going on down there though that is kind of odd, could be MP3 artifacts I suppose.


IMO, key-down symp res is so subtle that it's way down on my list of things I need in a piano sound. Though I do appreciate it when it is implemented in some form. To me it's much more exciting that looping, stretching, and layer switching are absent on the HP307, and that it passes the pedal down silent replay test, the quick pedal partial damping test, and the partial pedaling test with flying colors.

Pedal down symp res is much more important to me as it is not subtle at all on a real piano, and indeed for me is the most beautiful part of a real piano sound. Since it is often implemented as a delay or dispersion effect, it can often interact poorly with looping, which seems to be the case in many Yamaha DPs.

I don't hear that kind of interaction in your HP307 MP3s, which is good, but it does sound rather like a reverb effect. How does it strike you?

And how does it strike others who have listened to the MP3 files, particularly dpbsd_v1_7b_Roland_HP307_DR10_PARTIAL_jmmec.mp3 where it is turned up to 10?

Even on real pianos resonance sounds like a reverb effect.

Hmm, not to me.

What I mean is, when I press the pedal down on a real piano and play some notes it doesn't seem as though the piano were suddenly placed in a reverberant room. It instead sounds like it is in the same room its always been in but with the string all vibrating and interacting with each other.

Well I remember when I played on some upright piano, pressing pedal down gives an effect very similar to reverb, and - yes, to a "reverberant hall".

I'm pretty sure about it, because event when I played on acoustic guitar, the resonance between guitar strings sounded kinda like of reverb too.

Oh, here is a good example with acoustic piano:



Hear this? It's clearly a reverb-like.

And here is a quote from HP307 manual, "Troubleshooting" section

Well, here is what I'm talking about:

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

It's an MP3 mashup of the pedal down sympathetic resonance test for the VSL Vienna Imperial (default setting) and the HP307 (with the effect set to max).

The first sound is the Imperial with pedal up, the second with pedal down, the third is the HP307 with pedal up, the fourth with pedal down.

The HP307 pedal down sounds fairly reverby to me, while the Imperial does not.

Here is another mashup where the HP307 damper resonance is instead set to 5 (default):

http://www.mediafire.com/?yizn3zhdzjz

Not nearly as reverby sounding as when it is maxed out (IMO).

Ok, lets repeat it.

Even on real pianos resonance sounds like a reverb effect.

I listened to your files. HP307 is more authentic.

Correct me if I'm wrong, but the Imperial pedal down sympathetic resonance is a real recording of a piano with the pedal down. By very definition that is more authentic than a dispersion effect.

Excuse me, but I think that mix of a separately recorded notes won't give a real recording.

What can you say about the video above with real acoustic piano? Are you denying that this effect is very similar to reverb?

So far the sympathetic resonance of the HP-307 sounds fine, though I'd say the Imperial may be a bit better. There are various settings on the HP-307 that influence the sound.

The very nice thing about the HP-307 is that there isn't any looping/buzzing, and it sounds pretty 'natural' to me (not "superNATURAL" and not "superFAKE").

I took dewster's mashup of the Imperial and added a new sample of the HP-307 since maxing out parameters isn't something you'd probably do in real life.

http://wmsar.info/dewster_dpbsd/comp/imperial_hp307MOD_hp307MAX.mp3

The MP3 above plays 3 samples:

• 1. The Imperial (from dewster),
• 2. The HP-307 with 'moderate' settings (see below),
• 3. The HP-307 with MAX settings (see below).

dewster, I'm curious about the 'mosquito noise' - how are you determining that, and would you post a picture? I installed Adobe Audition and was looking at the 'frequency analysis', and was wondering if you were referring to the noise down below 1000Hz, or the entire range (seems more noisy below 1kHz). I did a quick web search on mosquito noise, so maybe I'm misinterpreting or looking at the wrong thing?

Regards

I agree - I think the HP307 pedal down sympathetic resonance sounds pretty good when it isn't turned up too much i.e. when it is set to a reasonable, realistic level. Beyond that it sounds more like an effect, but even then it doesn't sound bad.


Maybe I'm misusing the term. It's non-white noise - a bubbling slurry of tones - rather like the random chatter from a computer in an old science fiction film. I hear it in your files mainly when the recorded sound is getting near the noise floor.


For instance, here is a waveform view of the looping test in the file dpbsd_v1_7b_Roland_HP307_DR5_jmmec.mp3. The vertical dimension has been greatly zoomed up so we can see the noise floor, and a portion of the C1 note decay is highlighted. In it I can hear something like random tones. Here is an MP3 of the highlighted section, highly compressed so you don't have to turn the volume up:

http://www.mediafire.com/?hgagxodtw2j

It could be external noise getting into your signal, or noise from the analog to digital converter on your soundcard or in your digital recorder, or produced somehow by the MP3 conversion process, or it could be coming from the digital to analog converters in your HP307, or from the SN process itself.

Point taken. But sympathetic resonance in a piano is apparently linear enough that a simple sum of pedal down samples works in a realistic manner. If it wasn't, PC based DPs wouldn't sound very good.


Interesting video! Sympathetic resonance is very similar to reverb, and that's why dispersion effects can be effectively used to simulate it, at least to a first order.

But it isn't reverb. For one thing it's much more frequency selective.

I fear this discussion is entering xkcd territory:

http://www.mediafire.com/?fmzt2gaaw2y

MP3 for Yamaha CLP-S308

Recorded via aux out. I suspect it may have recorded in mono? Haven't had time to dig into that yet.

Thanks!

It is in mono actually.

I sent you a PM.

Hmm, this explains a lot

There is no line-in. I'd expect more from a mid-range business laptop (Lenovo T61p).


Audio
High Definition (HD) Audio
Built-in stereo speakers
Software control volume
Monaural microphone jack
Stereo headphone jack

How about recording left and right channels separately and then join them as a stereo track?

I suspect I'd have a hard time syncing them up precisely. Will probably just grab a long headphone extension cable and plug into my desktop...

MIDI timing isn't generally tight enough to correctly align audio phase.

How about importing both left and right mono tracks into Audacity, zooming in closely, then aligning the two until they are perfectly synced?

Cheers,
James
x