Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards What DPs really lack (compared to the acoustic pianos)

I actually don't understand how Pianoteq works. So is it sampled or modeled? If sampled, why are people always talking about modeling? And if it's actually modeled, ...HOW? Is it some kind of a numerical simulation of a real piano? What details are modeled? I was once thinking how I would get a simulation of a single string and that would be pretty CPU-heavy, solving the wave equation with damping on a string with fixed ends, getting the Fourier spectrum and generate the note from that...but for many notes at the same time, at different velocities? Sounds impossible...

It's not samples (recording), it's sound produced by algorithms (models) in real time. They tried to re-create digitally how a piano behaves. In sampling technology they are digitally recording (sampling) actual piano sounds with microphones and layering them. But in the end it's all digital 1's or O's so don't ever let any one fool you with this silly idea of one being more real than the next. What should matter the most is how does it sound and how does it play.

Pianoteq isn't a purely physical model. They use recordings of real pianos and attempt to reconcile their model and the sound of the piano. It doesn't actually have to recreate everything in real time.

Correct. Because that obviously is the best way to get an accurate representation of what they are trying to create. The sample is only the end point. What matters most is what happens in-between with all those fancy algorithms. The key thing to understand is that with Pianoteq you are NOT playing back those sampled recordings. That's why Pianoteq takes up so little disc space and requires so little computer horsepower.

For clarification (see http://forum.pianoworld.com/ubbthreads.php/topics/1438951/re-ca63-vs-pianoteq.html#Post1438951):

DP actions are already noisy enough and pedals are squaky, so no need to reproduce that lol

Wasn't that exactly what you strenuously claimed for pianoteq?

Pianoteq can't distinguish between the different types of touch discussed in the article. It doesn't know if you slam it into the keybed or not. It gets a note-on message and a velocity irrespective of how the note is played.

Arguing with the radio...

I've read about cases where it is claimed that the manner of playing influences the sound. Hand, arm, and body motion are said to influence the piano.

On the surface this is nonsense. The piano is a insensitive to such. But I'd entertain the notion that the pianist is influenced by such things.

CG: My father used to use the phrase "talking to the wall". I've never before heard "arguing with the radio".

Well, it’s translated from Bulgarian

I think this is the biggest difference between digital and acoustic, and why playing an acoustic is such a different, and way more versatile, expressive and responsive experience. I never realized it when i had zero experience with an acoustic.

Even the most expensive dp can't imitate this... yet. I wonder how it will be realized in the future.

I think people that are for example claiming a great dp is "better" than a decent upright might overlook this fact.

Unless you believe it's possible to produce fff and beyond without slam (and by the same token, believe it's possible to simultaneously produce ppp and below while also slamming), you'd surely have to accept that slamming and velocity are highly correlated? i.e. Pianoteq distinguishes by virtue of key velocity.

Both modelled and sampled approaches can reproduce slamming if developers take the trouble to model/record appropriately.

Pianoteq is a model of a piano, and when you play it you are in fact simulating the pianos behaviour. The model has many parameters; some of them correspond to the physically measurable characteristics of the piano (length of strings, motion of hammers, etc.), and some of them correspond to things that cannot be measured.

Those latter parameters must also be set, in order for the model to be complete and usable. Since their values cannot be determined through measurements, they must be estimated through some other process. That other process is comparison of the simulated sound with sound from the physical instrument. Therefore samples should always be used. The Bechstein DG is not special in this respect. Of course, it would be possible for the Pianoteq team to simply vote for the values of those unmeasurable parameters, by using their ears as the judge. But using samples is quantifiable in a much more precise way, and therefore it is clearly the way to go for anyone wanting to model pianos.

MIDI has a message format call Control Change with 2 values : the Control Change number and the value. (The number 64 is used for the damper pedal).

The Control change #88 is reserved to add precision on the velocity.

We can have the sequence

Control Change #88, 32
Note On C4, 56

Which means C4, velocity 56+32/128 = 56.25

As CyberGene said, when a keyboard send HiRes MIDI event, it doesn’t mean that all 16256 values are generated : this depends of the resolution of the timer used to scan the keyboard sensor. You may have only 1300 on a Piano Phoenix for example.

Ha. Got me. Should have been more clear in that statement. There is potentially no difference in how should "sound" because they are both digital representations but how they behave is a totally different matter.

I see. But if I understood you correctly Pianoteq cannot access this feature.

Yes that's what I believe as well. Why re-invent the wheel? They must sample at some point to understand the properties of the instrument they are trying to recreate.

I wouldn't be so sure of that.

The Pianoteq technology

Characteristics of Pianoteq
The piano sound is constructed in real time, responding to how the pianist strikes the keys and interacts with the pedals
It includes the entire complexity of a real piano (hammers, strings, duplex scale, pedals, cabinet)
Continuous velocity from pianissimo to fortissimo, with progressive variation of the timbre: that makes exactly 127 velocities! A sample-based software program would in theory require hundreds of gigabytes for all these velocities
Complex resonances that only a model can reproduce in all its richness:
Sympathetic resonance of all strings, both without and with sustain pedal
Duplex scale (the undamped string parts which come into resonance)
Sympathetic resonances between strings
Damper position effect when key is released (variable overtones damping)
Other special effects like staccato and sound continuation when pressing down the sustain pedal a short time after key release (re-pedalling)
Timbre modification of repeated notes, due to the hammer striking strings which are already in motion instead of being still
Release velocity
Eight types of pedals (that can be assigned to the four UI pedals):
Progressive sustain pedal, allowing the so-called “half pedal”, but also quarter or tenth’s pedals if you want!
Sostenuto pedal, allowing you to hold some notes after release without pressing down the sustain pedal,
Super Sostenuto pedal, where the notes held by the sostenuto can be replayed staccato, which is not possible on a “real” piano,
Harmonic pedal, allowing you to play staccato while maintaining the sustain pedal resonance,
Una corda pedal, also called soft pedal, modifying the sound quality or timbre by shifting the piano action to the right (on grand pianos),
Celeste peda, where a felt strip is interposed between hammers and strings, creating a softer sound. This pedal is usually found in upright pianos,
Rattle pedal, also called bassoon pedal, which equipped certain historical pianos, as for example the Besendorfer from the Kremsegg collection. A piece of parchment comes into contact with the strings to create a buzzing noise resembling the sound of the bassoon,
Lute pedal, where a wooden bar covered with felt is pressed against the strings, shortening the duration of the sound. It can be found in some historical pianos.
Variable lid position
Natural instrument noises including:
Variable action key release noise (varies with note duration and key release velocity if present)
Damper noise at key release (mainly for bass notes)
Sustain pedal noise: pedal velocity dependant “whoosh” produced by the dampers rising altogether from the strings or falling down
Choice of microphone position and multichannel mixing (up to 5 mics, 5 channels)
Microtuning and scala format files import
Various effects including equalizer, keyboard velocity setting, volume, sound dynamics which controls the loudness levels between pianissimo and fortissimo, reverberation with control of reverberation weight, duration and room size, limiter, tremolo.


https://www.pianoteq.com/pianoteq6?s=new

Fingers striking the key or the keys hitting bottom is a complex resonance I am sure Pianoteq took into account.

In searching for information I found this on their website and I agree with it (for those who want to read their take on sampled versus modeled)

Why a sampled piano is insufficient
The very best sampled pianos of today are the result of many hours of careful recordings associated with complex solutions designed to provide a valuable piano sound. We respect the work of these high class competitors who manage to develop sampled based pianos of this quality. However, as is well-known, sampling technology itself has inherent disadvantages.

To give you an understanding of the reasons why we chose to develop Pianoteq we find it necessary to describe the shortcomings of using samples to create a digital piano:

The sampled piano contains static recordings of each note, how it sounded during a particular moment in time. It does not take into account the influence of other strings vibrating, cabinet resonance, pedal interaction and hammer position.
The sampled piano cannot alter the existing piano samples when it comes to parameters such as hammer hardness, unison tuning, cabinet size, overtones spectrum etc.
The sampled piano has several technical limitations such as audible quantization noise and uneven variation of the timbre (from ppp to fff).
Despite many recent attempts to enhance the sampled piano sound by adding convolution reverb and other post processing effects, the technology as such has too many limitations when it comes to achieving a truly vivid and convincing piano sound.

https://www.pianoteq.com/pianoteq6?s=new