Topic: Assumed "micing" distance in Pianoteq?

Sorry for the long post. Please take it as a sign of an abiding interest in Pianoteq.
After all, the more I play Pianoteq, the more I like it. (The only problem is that I spend more time trying various parameter settings than I should.)

In listening to the sounds, however, I do wonder if what some people may want  when they say they want more wood in some presets is that they want a closer micing perspective, which would include more resonance sounds, etc. The sound is now more of a slightly distant micing, suitable for classical playing, maybe in a chamber music setting.

Of course I understand that the instrument is modeled, not miced. But even so, doesn't a piano vary greatly in timbre and volume depending on where you sit? Is the  assumption in the programming that there is a pure sound generated by all of the physical vibrations and dampings, with the ability to add reverb for a sense of more space? Or is the harmonic content and volume of all the sounds dependent on an assumed position of the listener?

(If the harmonic content was modeled on recordings of each note, the notes were recorded with mics at a given distance, with effects on the harmonics such as the proximity effect or the lack of it, the dispersal of the sound in the air, and the relationship between amplitude and the harmonic content at any given moment, and all that stuff.)

I ask these things just because the perspective does seem a little distant. (I noticed it again on the great new instruments.) And the amplitude seems a little low even when I pound with the hammers set to hard and the reverb off.

Could there be a way to let the user move the piano forward and back? A slider, with all of the parameters adjusted out of sight to a closer sound?

Not a simple thing to create, I know, since it might mean having to do all of the modeling again for the closer sounds. But once you had the specs, the ratios of distance to harmonic content, of amplitude to harmonic content, etc. you might have rough guidelines for future emulations of specific pianos.

A possibility?

Last edited by Jake Johnson (05-09-2007 15:16)

Re: Assumed "micing" distance in Pianoteq?

I would just add my vote for this one. It's absolutely true that a mic (pair) placed inside the piano or 1 m away gives you a compeltely different sound and dynamic range. I would really like to be able to adjust this.

Re: Assumed "micing" distance in Pianoteq?

Further uncalled for, probably unwanted, questions and thoughts on this subject:

1. Is there a relationship between the relative amplitude of partial frequencies and distance, so that the mix of partials changes as you move closer or further away from the piano? My (non-mathematical) thought is that the closer you sit to the piano, the louder the low and middle partials will be in relation to the upper partials, since you are sitting close to the strings, while the further back you move, the more the upper partials will stand out and the more the midrange partials will die. (Not a simple equation, I imagine: Higher partials, we know, die out faster than midrange and low partials, but the relative amplitude of the partials at any given point in time will vary with distance in differing ratios on different ranges on the piano. Will also depend on the resistance of the soundboard and the cabinet.)

2. Regardless, is time a factor in the mixture of harmonics? (And thus in the perceived timbre of a note?) Will faster moving (higher frequency) partials, reflected faster off the soundboard, reach the ear slightly earlier than middle and low frequencies, some of which oscillate the sound board\cabinet, etc, as one way of being heard, which takes time? (These lower frequencies are of course heard from two sources: as the string is struck and again in the soundboard resonance.) And is this slight delay increased as you move further back from a piano? Would a fragment of a millisecond make a huge difference in timbre if some partials reach the ears sequentially instead of simultaneously? (I know that a fragment of a millisecond makes a big difference when creating an attack envelope, or in other sound manipulation, but I've never experimented with or read much about slight variations in harmonic timing as affecting the perceived timbre.)

3. In addition, would the same result be found in the harmonic spectrum of the notes (considered as single units instead of sets of partials, so we are back to speaking of notes as just notes in a basic way) when playing, for example two notes at once or a chord? In other words, if a pianist hits a middle C and a E6 at the same time, with velocities that the pianist hears as creating the same amplitude, will a listener at a distance perceive the amplitude of the middle C as lower, the further he sits back in the hall? (Notice I say that the pianist hits with velocities she perceives to create the same amplitude, not with the same velocity.)  Would the distant listener register the upper note very slightly before the lower note, as well as the upper partials of each note?

Would all of these combine, along with overall amplitude in relation to velocity, hammer amplitude, etc to create a sense of distance or closeness? What are the other factors to consider?

(Sorry to put forward this notion, Philippe, when you've just provided the excellent new presets, but many people like a close perspective. And I'm not sure I've seen a full, precise study on the relationship between harmonic spectra and physical distance from multiple nonlinear interactive parallel and serial oscillators. (Does that describe a piano?). Are there graduate students who need an idea for a thesis and have occasion to sample and compare close and middle distance notes to determine the ratio (nonlinear?) of relative spectral amplitude and timing to distance?

(Of course another question, here, is the filtering that mics introduce...Particularly when the factor of distance is included. )

But consider the possibilities:
1. The range of timbres, and uses, for Pianoteq would be almost infinite.
1. Users could choose their distance from the piano easily. Thus there would be fewer worries about there not being enough wood, etc. (Users could still add or subtract wood, or anything else, by using the usual sliders.)
2. You could add Pianoteq to almost any existing mix by choosing a distance similar to that of the other players and then panning it left or right so it was out of the way of the other instruments, or more in the forefront as a solo instrument.  In other words, you could place the piano anywhere on a sound stage.
3. I would stop posting long, bothersome messages such as this one.

Faithfully

GJ

Re: Assumed "micing" distance in Pianoteq?

I agree fully with both of you that this question of mic position is very important, and it is definitely something that we will work on during the next months. It will take some time because the task is not easy, as one can see it from the many aspects that you mention. In the meanwhile, when using the current version, there are some settings that make the piano closer, for example if you turn reverb off and choose loudspeakers output (stereo width may be used also). The “closest” preset is probably C2  bridge, although it is not my preferred. Thank you for your comments which we will have in mind during development.

Re: Assumed "micing" distance in Pianoteq?

@Jake.

Acoustics textbooks are hard work - and psychoacoustics textbooks are not much easier

re 1: In the open air, which is the simplest case, higher frequencies are attenuated at a faster rate, so as you move further away, the sound will become less bright (c.f. cars and roads, communicating with submarines). In a room or hall the situation is much more complicated in regard to the balance of the sound in relation to your position from the piano. The walls of the room not only affect attenuation of different frequencies in a complex way according the construction materials, but the size and shape of the room will cause low frequency standing waves to be excited. It is true that these standing waves are present also at much higher frequencies, but they are most intrusive subjectively and objectively at low frequencies where they are least attenuated by soft furnishings, and will persist for a longer time. Finally, in a room, your position relative to the piano will affect the proportions of sound you hear direct from the piano, and after (modified) reflection by the room. Perception both disinguishes and amalgamates these sound waves - the brain has to tell you (correctly) that there is a piano (tiger) in the vicinity, but also approximately where, and how many.

re 2: I would think that the effects of dispersion in a medium of propagation are small in practice, relative to other factors that contribute to the sound of a piano. Energy does, however, travel at different speeds in different materials, and where those materials have (interesting) resonant properties, it will take different amounts of time for resonances to build up in them, depending also on their shape, the driving frequency, and how those materials are constrained (the strings are clamped at their ends, but the lid is contrained only at one edge). It is these factors which will influence strongly the resulting timbre.

While it is true that a fragment (fraction) of a millisecond can make a difference to a waveform as a graph - bear in mind that we have been (reasonably?) happy listening for years to a medium (CD) which cannot reproduce a square wave of 6kHz correctly.

The time factor of special interest, from the modeling point of view, is not only in the amplitudes of the partials and how these change, but how they ever so slightly change in frequency during the duration of a note. This seems to me to be part of the reason why modeling real musical instruments has only been partly successful to date. It is both difficult to analyze and difficult to treat mathematically, i.e. FFT does not work. (Some appreciation of this factor also distinguishes good hi-fi equipment from the mediocre - in other words, good hi-fi has fewer frequency distortions - but that is another subject).

So, as I see it, it takes a special sort of genius to put all these things, these different (imperfect) equations and (incomplete) facts of acoustics, together mathematically in a way which works subjectively.

To move from fact to speculation - it wouldn't be unreasonable in the first instance to assume, for the purpose of deriving a model, that all the physical elements are equidistant from, and synchronous as regards the listener. Of course, as an assumption, it can be reevaluated once the validity of the model has been established...

Re: Assumed "micing" distance in Pianoteq?

HyperReal: You say that it's reasonable that a model can assume that "all the physical elements are equidistant from, and synchronous as regards the listener." Do you mean that this assumption is understandable as a point of departure for a model? 

If so, I fully agree. These assumptions have for many years been a basis of physical modeling. I didn't mean to insult Philippe and the equations. I hope he understands that my intentions were honorable.

But can we all look forward to a better understanding of the relationship between time and relative harmonic content? And the contribution of the shape, size, and absorptive qualties of the room, as you say.

In a selfish way, however, I do want to learn more about the way distance and harmonic content interact. I want to know if there is precise equation.

Can we think together about the effect of the room? The effect of the room would depend on where the listener sat? And would come in a similar sequence and rate, but with different proportions, depending on her physical position? (And would she hear a different prolongation of harmonics? See the next paragraph.) The person playing the piano would first hear the hammer striking the wires and in less than a millisecond hear the sounds bounced off the soundboard and cabinet, and then hear the sound of the vibrating soundboard and cabinet, and then the (also filtered) reflections off the walls? Slightly further away, the listener would hear things in the same sequence, but at different rates? Or in a different sequence, or with a different sense of the time during which the frequency existed? Would this slightly more distant listener hear the reverberations of the walls at about the same time as the reverberation of the soundboard?

Would the amplitude decay of each harmonic of both be about the same, although the pianist continues to hear the vibrating strings and soundboard and cabinet at a higher amplitude long after the room reverberations?  In other words, I worry about the other variable that I mentioned--how the harmonics evolve in time. Is the decay time of each perceived harmonic also affected by distance? (Probably, again, in a nonlinear way?) In other words, if our patient piano player struck middle C, would she first hear a different set of harmonic amplitudes (as per all of the above discussion) and then hear a different sustain length and decay rate of each harmonic from the listener who is six feet away? (High frequencies decay faster.) Here, I'm thinking of a listener who is close enough to register the multiple piano oscillators more than the dampened reverberations off the four sides of the room.

Philippe: My apologies, again, for my attempts at understanding how a piano makes a sound and how it is heard. I hope that these concerns are in some way productive.

Re: Assumed "micing" distance in Pianoteq?

Hi
With absolutely no technical knowledge about acoustics, and just going on personal taste, I have felt something similar to Jake's post, and that one of the things I liked about v1 was the dry up front sound, which for me is very useful in rhythm section/pop things. So much so that i've kept v1 just for the early M1 presets.
I noticed with v2.x, that even though it was a great improvement, the added realism that the lid open/closed feature added also meant that the totally dry close sound had gone.

After writing to this forum Guilliame pointed me to C2 Bridge preset which has that close characteristic, and it fulfills that need very well.

Don't get me wrong, I still love this instrument. more each revision.

I know it's not very purist but it would be great if there was an option for that dry/close studio sound ( or more presets like 'bridge' maybe?) along with the incredibly realistic sounds that it has right now. I'm not sure how it could be implemented...maybe a few more presets?

cheers jeff

Re: Assumed "micing" distance in Pianoteq?

Philippe: My apologies, again, for my attempts at understanding how a piano makes a sound and how it is heard. I hope that these concerns are in some way productive.

Jake: you do not need to apologize, as all comments posted here really help us during development, giving us knowledge of what musicians expect (there is btw material there for several years of development - which is good).

Re: Assumed "micing" distance in Pianoteq?

"The time factor of special interest, from the modeling point of view, is not only in the amplitudes of the partials and how these change, but how they ever so slightly change in frequency during the duration of a note."

Hyper-Real:

Very curious to learn more about frequency changes in the shifting harmonic content. Raises many questions: Is there a predictable pattern, or a regular deviation from the fundamental of each harmonic? Does each harmonic shift to the same degree? At the same rate? At the same rate but starting at a different time? How much do the all of the shifts depend on the velocity whap? And there is probably a large variation from note to note, given the different string numbers and widths. Is there any discussion of this on line? (Links that you might provide?)

Hm... So each harmonic shifts in both amplitude and pitch over time, and may arise (and shift in pitch?) at very slightly different times, and shifts differently in all three or four respects from note to note. And each factor probably depends partly (and nonlinearly) on velocity and the distance from the listener and the quantity of vodka the listener drank the night before. Not a set of equations that I can easily imagine.

Many thanks.

Re: Assumed "micing" distance in Pianoteq?

Hi Jake,

Again I can only refer you to technical treatments of acoustics. Despite the fact that music is today ubiquitous (iPods, etc), and noise pollution is worse than ever, acoustics is not a subject which can be summarized easily. Compared to say electricity, where Ohm's Law and Kirchoff's Laws will take you a long way, both the measurements and the mathematics of acoustics are messy and lacking in generality in comparison.

It's well known that the timbre of a musical instrument comprises overtones whose frequencies are close to integral multiples of the fundamental frequency, which is perceived as the pitch of the note. It is this which distinguishes a worked piece of spruce called a violin as a musical instrument from a worked piece of spruce called a kitchen table. But there is a curious category of objects with intermediate properties in their timbres, which are also used to make music, albeit of a particular kind - bells. While the familiar musical instruments broadly speaking have timbres whose overtones remain static in frequency over time, the overtones of bells are neither at integral multiples of the fundamental, nor do they remain remotely constant over the duration of the sound. (Yet do they not make some of the most beguiling sounds one can hear?)

Closer acousic analysis shows that the sound of "musical instruments" (as just defined) has variability in the vibration somewhat like bells. This is not their main mode of vibration - a flautist for example is taught how _not_ to overblow a flute. When the flautist has learnt the right technique, the sound approximates a steady sine wave. (Until the flautist is presented with a modern work to play). Instruments that are mechanically and acoustically more complex than a flute tend naturally to show smaller or larger deviations in the regularity of the waveforms.

These features of musical instruments can be demonstrated quite easily. Modern electronic tuners work by sampling the sound, extracting a pitch by analysis, and displaying the result to a fraction of a cent. (100 cents per semitone). If the tuner has not been "slugged" you can easily see when tuning say a guitar that different strengths of plucking the string result in a different pitch being produced, and that the pitch also varies as the sound dies away.

The other clear demonstration comes from synthesis. While it is quite easy to synthesize something which resembles a flute sound and differs from a saxophone sound, it has proven quite difficult to synthesize a convincing sound.

Clearly somewhere along the line from waveform analysis, through modelling, to synthesis something has been missed. If it was evident what the missing ingredient(s) were, then we should be blessed with fine emulations of every known instrument. Yet even Pianoteq, from reading reviews and comments, does not convince some people, although it is hugely superior to any other piano simulator I have heard.

It was merely idle curiosity on my part whether Pianoteq was modelling some aspects of time-dependent frequency variations in the sound, in view of its seeming to be a subtle but important part of the sound of a musical instrument, yet not something that is treated well by the common mathematical techniques of signal processing. It may, of course, be something that the Pianoteq team would prefer to remain commercially confidential.

Re: Assumed "micing" distance in Pianoteq?

I read these requests & think , that people forgot real piano sounds , when they try to say about piano micking , ... pianoteq is not sampling piano , it generate sound itself , for what needs microphones modelling if U hear straight good piano sound , or ears are acomodate to Mp3 quality , or to "micked & adapterized" sound . Most people request is , - it means " how to make piano sound bad'er..." . I understand that most of U say "No! U not Understand us" ... I understood more than need to understand , for this purpose what U asking , need to make in pianoteq , Wav impulse responce read section & U will have many things of mike positions , Eq's, acoustisc modelling ... etc .
For future better to thing , how to do multi channel (surround) output, with adjusting piano position in acoustic feald, for this amazig software. Sorry for english .

Re: Assumed "micing" distance in Pianoteq?

Voxa:

I may have chosen my words too quickly. When I spoke of "micing distance," I only meant the assumed distance of the sound from the listener, which is related to what you suggest when you request the ability to adjust the position in the stereo field. I did not want to request a reduction of harmonic accuracy through mic filtering or brightening.

Some questions, though:

1. At what distance from the piano is the listener within this stereo field?

2. How does the distance affect the impression of stereo? As we move closer to a piano, which has strings that lie on each side of our head if we sit near or lie under it, we of course hear the "separation" of notes more clearly--we become more aware of the location of the strings (left and right and step by step both ways) because we are closer to the first in the sequence of oscillators--the strings, and the sound of each string set get less distinctly placed in space as we move further away. And even when sitting at the piano, sounds from both sides of the instrument reach both ears at different rates and with different filtering. I can't begin to know how the developers could calculate the interaction of varying  harmonic content x the (nonlinear) effects of distance as they affect the perceived stereo placement.

And come to think of it, do we hear the overtones in a single note more clearly when close up than at a distance, too?

(Regardless, I'm not sure that mics can be left out of the question. Aren't they always in the chain for examining complex sound data-- a necessary, but coloring, tool for both gathering data and testing theories about spectral changes? Mics gather the raw data on which a theory is based or proven, even if they are the mics attached to a hardware spectrograph instead of to a wave file recorded through a computer sound card. Fourier was thinking of idealized, non-percussive instruments and had no apparent interest in the problems of air impedance and dispersion over varying distances or any way of registering minute changes in relative amplitude or frequency or duration over time  He was thinking of the rate of vibrations of the string or other oscillator. (Time and distance were not on his mind--he assumed that the sounds simply existed, as if a note was an egg he could pick up and roll around on his palm.) 

I don't mean to sound argumentative or to walk you through things you already know. I'd like to know more about stereo placement, too, and your post made me think more about the way distance and time affect the sound.

Last edited by Jake Johnson (19-09-2007 23:38)

Re: Assumed "micing" distance in Pianoteq?

I just want to add that even if I'm completely amazed by the quality and the musicality of PTQ, I still don't have the same feeling when I'm playing on my real grand (a 1949 Gunther), which is probably partially due to the fact that I'm listening to myself playing thru headphones or speakers. So, the "mic placement" issue is really an important one for me, because it would also have an influence on the "speakers" placement... Why stereo? How stereo? Why not 5.1 ? 2.1? Of course, the vibrations of the instrument itself going thru the keyboard when you play on the "real thing" also have an influence on this missing feeling with PTQ, but let's go and search further and further, so we could one day get as close as possible of the feeling you get when playing on a real piano...