Problem with PTQ sound

Thank you very much for your reply and for the opportunity to clarify both points.

I am using the internal speakers of the Yamaha NU1 via its line-in input or the Philips SHP9500 headphones (directly from the computers, no piano involved). I tried also with cheap computer speakers or making a recoding (digitally in PTQ itself) and playing it on my stereo amplifier where I listen to CDs (in another room, so cannot made a direct connection short of moving the hefty NU1). The problem is exactly identical.

By "dull" I mean something that does not sound at all like a piano, not even like a felted piano. More like the sound of a flute! Let me say this: if it weren't for the abrupt attack, which obviously is not flute-like at all, and to lesser degree for the slow decay, which is uncommon for a flute (but not impossible), the sound could be mistaken for a double contrabass flute rather than a piano. Increasing hammer hardness makes those notes sound like a piano to me, so that's something you can test easily on your own.

You can hear what I mean "sort of" (I'll explain this in a moment) in my recording here. This sound has been digitally recreated from PTQ from its own MIDI recording of the very same performance -- no editing of either MIDI, or sound or video has been performed. Only thing done has been mixing and aligning of sound to video (only at the beginning extra noise from the outside in addition to the digital rendering of PTQ, which then fades away during the performance):

https://www.youtube.com/watch?v=G-_3XvwvqAA

at second 20", 36" and 1'37" there is a low D which is a little bit "dull". I said "sort of" because this is by no means horrible. As I wrote the worst is in the octave below, but even this is still a bit too dull. The problem is not too severe in this case also because it's the last note of the phrase, so it's supposed to be soft and "concluding", hence some "dullness" is ok. If it's still not clear, I can record playing the scale as I suggested, both on PTQ and on a fine acoustic grand I have access to for comparison.

Yes I have. I tried at least one (the first) preset of all the piano instruments, and for the pianos which I liked most (for other reasons) I tried several of the presets. I even
tried the U4 tacky and the honky-tonk. The latter is waaaaay too bright everywhere, but still has this problem in the bass.

I agree with YvesTH -- The Grotrian Intimate preset, used in your video comparison, has an intentionally somewhat subdued character including in the bass range, so it was probably not the best Pianoteq preset to use for comparison. Try the Bechstein, the Steinways, the YC5 and/or the Petrof Mistral.

Yamaha is known for its rather bright, distinct sound, but your keyboard speakers, like your headphone speakers, are fairly small (the Yamaha's are about 6.5 inches I think), so they might have a somewhat poor bass-frequency response in comparison with the treble, and Yamaha (I'm not sure of this, of course) might have adjusted the internal sounds of the NU1 to compensate for this.

However, regardless of those possibilities, try adjusting Pianoteq's main (pre-effects) Equalizer, to boost the treble frequencies, which will boost the upper frequencies across the keyboard range, including those present in the bass notes.

When I'm playing Pianoteq with my Kawai MP11SE keyboard, I don't notice any "dullness" (muted, more indistinct) sound in the bass. For most presets, the bass sounds appropriate, with plenty of power at higher key-press velocities. Hopefully you will be able to adjust both Pianoteq and your hardware and listening environment to obtain a sound that you like.

Yes.

Thanks you for all of these ideas. I tried them but still not identified the problem. Instead of tediously go through the list of what I've done and you trying to understand if I did it correctly or not, let me instead reframe the problem in a slightly different way, and then provide recording for you (or whoever is interested) to check. I think the recording would be very objective and will help identify the problem.

REFRAME OF THE ISSUE

Suppose I want to play a 7-octave scale, say from lowest C to highest C on a 88-key keyboard (I'd prefer doing B-scale rather than C, but the silent note in the demo force me to do C). I play this scale with Pianoteq, let's say for now on the keyboard (but it could be with an artificially created MIDI file). I play it with constant touch and I achieve constant volume. I obviously do not change the settings or the touch (chosen velocity of playing) while playing. What I experience is that the brightness of the sound increases during the first 3 octaves in unnatural way for an acoustic piano. After the first 3 octaves, the brightness remain constant. I can change the touch and/or the settings as I please, but the lowest notes are too "dull" compared to the midrange and the higher range of the scale.

So in this reframed description, the velocity objections go away, since the scale is played all at (about or exactly) the same velocity. In case of manual playing, in which the velocity is not exactly the same, the velocity changes will change also a volume change (in addition to a tone/brightness change), so it should still be evident and natural, agree? All of your other objections remain, so let me described my proposed test.

PROPOSED TEST

I play the scale with PTQ, with the "NY Steinway D prelude", create a FLAC or WAV and play it from the computer, from a different software than PTQ itself. Of course that file will show the problem in my setup, but nothing else.

I play the scale with the NU1, which can record a WAV of what one plays, so I do that. While doing so, the problem is not evident, but as you say the NU1 can do adjustment on its own sound to match the speaker/amplifier. I take the NU1 WAV, move it on the computer and play with the same WAV player as the one used for playing the PTQ, and compare the two. My preliminary tests indicate that the problem is not there with the NU1 WAV. Of course it's possible that the NU1 saves that WAV with the adjustment on its own sound to match the speaker/amplifier, so this alone tells us something, but not the whole story.

Third, I play the same scale on a fine Mason & Hamlin acoustic grand piano (which is supposed to have a character similar to the NY Steinway D prelude and if anything be even less bright to the point that some call it "dark"). I record the scale on my phone, move the recording on the computer and play it again. My preliminary tests indicate that the problem is not there with the phone-recorded acoustic piano. Of course it's possible that the phone saves this recording with some equalizer by chance cancelling the potential amplifier/speakers problem which we are speculating, so we are not there yet -- but getting close.

This test will remove lots of uncertainties from the issue:

• comparing brightness within the scale of the single instrument (rather than single notes from one instrument to the other) will make the possible velocity difference a moot point, as long as also the volume during the scale stays the same

• comparing various recordings obtained differently from the same reproducing audio chain (computer, amplifier, speakers) will remove many of the unknows there

• comparing the various recordings on a different audio chain (different computer, Onkyo amplifier and PolkAudio floorstanding speakers that I have in another room) will provide invaluable insight: if it's the ampli/speakers to be "wrong", it's highly unlikely that a different set will show exactly the same issue

• sharing these recordings with you guys can provide different ears and different opinions on the sounds

For example, on the last bullet, some of you could say: "no it's the NU1 and the acoustic that are too bright in the bass for these recordings, PTQ is the only one that's fine".

Before I invest into the time for doing this test comprehensively (as I said I did it some quick-and-dirty already), does anybody any suggestions on how to make it better? For example, I plan to play myself (rather than making an artificial MIDI at exact constant velocity) because that will be what I am forced to do on the acoustic and so it'd be apple-to-apple. In addition, that's how I plan to play PTQ, not with artificial MIDI files. I think the minor velocity differences which my manual playing will create should be automatically evident in the small differences of volume, and could actually help, rather than hinder, the investigation. Do you agree? Do you have other thoughts on what to test? I could use other virtual instrument, for example the Chateaux Grand from https://sites.google.com/site/soundfonts4u/ -- would be useful to have additional instruments, or would just complicate the comparison? How about different PTQ presets?

Thanks a lot, and happy Thankgiving to the ones in North America!

@dv,

Have you found other VSTs that sound acceptable through the speakers on your Yamaha or through your headphones?

I have my Pianoteq playing through some monitor speakers (Emotiva Stealth8) facing the ceiling, with smaller speakers, dialed-down but facing me (Emotiva AirMotiv4s), and an 8" subwoofer.  Very convincing.  But, when I brought my laptop and external soundcard setup to my father's house and played it through his Yamaha Clavinova CVP609, it sounded terrible. 

The only thing that I could figure is that Yamaha 'tunes' their internally-produced sound with an equalizer to make the most of their built-in speaker package.  As such, those speakers are by no means 'flat' in their responses, so that feeding them Pianoteq (or any recorded music) sounds strange, lacking bass, with exaggerated midrange, etc.

Now, how this influences your headphones, I don't know.  I have Beyerdynamic 880 headphones that sound quite good with Pianoteq.  I feed both my monitor speakers and my headphones with a Steinberg UR22 mkII.

Thank you for your patience and suggestions. It's very unfortunate that one has to get a Ph.D. in sound engineering to be able to use this software. Especially because one of the big plusses of PTQ compared to the competition is that its user interface does not look like a sound engineering desk, but like a piano!

In any case, I set the volume of the laptop to the lowest that it could go while I would still hear anything from the piano (while putting the volume to the latter at maximum). The problem stay identical.

I then went to https://www.modartt.com/steingraeber and played the first Rach's Prelude. The bass sounds growling and bright like it should, no problems at all. Both low laptop volume (and super high piano volume) as well as mid volume on the piano and "crank up" on the laptop. So I am not jumping in this dB rabbit hole: the computer/interface/amplifier/speakers seem to be totally fine. However, even in that recording, the mid-upper registers are too bright (listen for example to the part between 1'30" and 2"). For that particular piece of music is fine, but for others I would like to have the same bass as this one, but a mellower mid-upper -- but I guess I can live with it.

So I decided to selected the same instrument and tried to reproduce the same sound, to see what it took. The issue seem to be a combination between the velocity curve and the dynamics. Even tiny, visually imperceptible differences in the velocity curve, make a HUGE difference in timbre. On the other hand, even with the dynamics at the maximum, the difference in volume between the softest I can play (sub-25 velocity) and the heaviest (over-100) is a far cry from the difference in volume for similar difference in touch for an acoustic piano. Same with either piano speakers or headphone. The latter part could be sound interface limitation, so I will try different hw and I am not too worried about it (plus, it's something I can live with).

On the other hand, I guess I have now some clues on things to adjust to see if I can make it sound "right". So thanks everybody very much for their suggestions, you all certainly nudged me to try some things which I would not have.

My last question is if Modartt publishes good MIDI files with the corresponding sound recordings to guide my further attempts. I know there are some in the shared files on this forum, but I'd like something more authoritative. Why? Well, because I was using the velocity curve suggestion from this website for the NU1, and that goes in the direction OPPOSITE of what I think is needed, so I am now wary of taking something "at random"