Messe: Viscount Physis Piano

[Ashville.Guru]

It's become clear to me that the fatal flaw in these modeled pianos is that you can't make a mathematical model for the sound of wood, and that's why there's a sterility to their sound.

And the 'Sweeping Overgeneralization of the Year' Award goes to...

 

Just kidding. There's a huge difference between saying "X is the flaw observed in all modeled pianos produced so far" and "Flaw X is inherent to the mathematical modeling process". This is a classic example of what's called an Inductive Fallacy.

 

And that's still assuming your observation is correct. There are a gazillion different settings possible with 15 parameters. It's possible that some of those would sound 'woody' enough for your taste. Then again, maybe the settings can't add wood to the sound.

 

This is a subtle point that is easily missed - evaluating a modeled piano demands a change in approach.

• With a sampled piano, the sound is more or less fixed, so you just listen, you like it, don't like it, point out flaws, whatever.
• With a modeled piano, there is no "the sound", so whatever you are listening to is just one out of a gazillion possible sounds. The only sensible way of evaluating it is to play with the settings/parameters, for a reasonable amount of time.

To give an analogy, imagine there's a clueless kid who's randomly turned knobs on a classic analog synth. You just happen to walk in and hear one sound. And leave immediately concluding that analog synths sound like crap...

 

BTW, didn't we have a recent thread about how the point of modular synths is lost in listening to most amateur YouTube videos?

 

I'm not saying that modeling is free from the 'woodless' flaw. Maybe no-ones figured out how to model woodiness. Maybe they have, but you need to wade through the gazillion settings to find it.

 

Doesn't make the above generalization any less silly.

 

[AnotherScott]

I don't know whether or not it's been done (and I don't know how much this corresponds to "woodiness"), but it does seem that one could model the "enclosure" element of a piano... essentially a kind of reverb, i.e. based on the size and shape of the "room" the strings are housed in, and how reflective its surfaces are at various frequencies, and its elements of resonance. Then they could model, not just the sound of a wooden enclosure, but also things like this...

 

http://www.glass-pianos.com/glass-piano.html

[Dana.]

And the 'Sweeping Overgeneralization of the Year' Award goes to...

 

Just kidding. There's a huge difference between saying "X is the flaw observed in all modeled pianos produced so far" and "Flaw X is inherent to the mathematical modeling process". This is a classic example of what's called an Inductive Fallacy.

Meh. I don't care enough to continue debating. I'm glad you've found AnotherScott, though, so you can continue discussing all the minutae. You two are perfect for each other. Have fun!

[mauriziodececco]

My point was simply that a lot of things can be called "physical modeling" and it can be very abstract. The "physical model" of a tuning fork is a sine wave with exponential decaying amplitude. And that's a pretty damn good model!

 

And there's the in-between world where samples are processed to generate abstracted properties files, such as with TruePianos and evidently Kronos EPs, if I understand what I read correctly.

 

Well, if you take a strict technical and scientific standpoint, what you says it is not true (if i remember well from my Ircam years).

 

The alternative is beween physical modelling and what you may call sound modelling. A traditional synthetiser do sound modelling.

 

Lets make an example: take the tuning fork.

Matematically you can study the behaviour of the tuning fork by writing a set of differential equations based on physical caracteristics of the fork; as you said, this is an abstraction of the actual, physical tuning fork; in general,

we (21st century human beings) think that this kind of abstract matematical model is good enough to predict how a real tuning fork will behave.

 

We now by experience and by solving the matematical equations know that the tuning fork will produce a sine wave with a exponentially decaying amplitude.

 

Let's says we want to synthetise this sound: a sound modelling approach will use some kind of synthetizer to reproduce the sound. For example, you would an analog synth, using a sine wave oscillator, a VCA and an exponential envelope generator.

 

The point here is that we are working on the sound; we study the sound characteristic, and we reproduce the sound we want using any means we have at our disposal.

 

A physical modelling synthetizer use another approach: let's study the physical properties of the object producing the sound, i.e. let's take an abstract matematical model (physic, as a science, it is all about writing these matematical models), of the physical object producing the sound, and let it run on a computer somehow, emulating the physical behaviour of the physical object, in the hope that this emulation will be close enough to the physical object behaviour to produce the same sound (or a sound close enough) of the one produced by the physical object.

 

So, here, we do not even need to know the sound that will be produced; we are interested in the physical properties of the object producing the sound.

 

So, if we take this strict definition (that is clearly not the definition used currently by the various marketing departement), things like decomposing the sound in various components (hammer, string, transient, etc) and recomposing the sound would not be physical modelling; at my time, it would be called (maybe) frequency domain additive resynthesis.

But it is still based on the sound.

 

On the contrary, AFAIK, many of the analogue synthetiser emulator software could be considered physical modelling based, because they use a matematical model of the electronic circuit of the synthetizer, and not a model of the sound the synth produce (an electrical circuit is still a physical object; physical modelling do not require mecanical modelling).

 

OK, all this just to be precise and pendantic (:-<). At the end, as you say, the only thing that count is how the beast sound. Who care what is inside, when you play it.

 

Maurizio

 

[Mark Schmieder]

Can't make a mathematical model for the sound of wood?

 

I do that as part of my day job!

[Ashville.Guru]

Can't make a mathematical model for the sound of wood?

 

I do that as part of my day job!

Sounds interesting. Could you elaborate a bit more?

 

 

[Ashville.Guru]

So, if we take this strict definition (that is clearly not the definition used currently by the various marketing departement), things like decomposing the sound in various components (hammer, string, transient, etc) and recomposing the sound would not be physical modelling; at my time, it would be called (maybe) frequency domain additive resynthesis.

But how do you know that the Physis (or any other modeled piano, hardware or software) is using this decomposing/recomposing approach? For all we know, it could be using actual differential equations, modeling the physical objects from scratch. Numerically solving the equations using FEM. If so, it would make sense to 'expose' some of the more intuitive, meaningful parameters, such as hammer hardness, string length, etc.

 

The point is, it may well be using 'hardcore' physical modeling, in the truest sense of the term. We'll perhaps never know, given that the algorithms are proprietary.

 

BTW, welcome to the forum, Maurizio!

 

[JeffLearman]

From what I've heard, what "physical modeling" approaches use is more like:

 

What is the effect of this physical property on the sound ... let's model that effect.

 

They're not reconstructing the sound of the piano from discrete models of all the physical components. Rather, they model the effect that each of a number physical components has on the sound.

 

In any case, the proof is in the pudding. I've heard modeled Rhodes that I think are cool instruments, but I haven't heard one that sounds or works for me like a Rhodes. I prefer sampled Rhodes. My favorite soft piano happens to be semi-modeled (it's a model based on reduction of sample data).

 

Regardless, digital pianos keep getting better, which makes me a happy guy.

[Tobias Åslund]

There's now a couple of demos to listen to on the site

[mauriziodececco]

A couple of answers to a couple of posters:

 

 

From what I've heard, what "physical modeling" approaches use is more like:

What is the effect of this physical property on the sound ... let's model that effect.

 

Well, not really, at least not ten years ago in the scientific community :->.

 

The above definition apply essentially to all kind of synthetizer; for example, a classical

analogue synthetizer (oscillator/filter) correspond 100% to this model, and it what actually

initially designed with this model in mind: there is a sound source (the guitar string, for example) and

a physical body modelling the sound produced by the string (reproduced by the filter).

 

The original idea in physical modeling was to focus on the physical process, and reproduce it, and to

discover what sound is produced; of course, if you dig enough, you'll find that at least in some case

the approach tend to converge; for example, if the mathematical solution to a physical model is a

sine, an sinusoidal oscillator is both a physical model and a sound model; when the situation is more

complex, the two tend to diverge.

 

 

So, if we take this strict definition (that is clearly not the definition used currently by the various marketing departement), things like decomposing the sound in various components (hammer, string, transient, etc) and recomposing the sound would not be physical modelling; at my time, it would be called (maybe) frequency domain additive resynthesis.

But how do you know that the Physis (or any other modeled piano, hardware or software) is using this decomposing/recomposing approach? For all we know, it could be using actual differential equations, modeling the physical objects from scratch. Numerically solving the equations using FEM. If so, it would make sense to 'expose' some of the more intuitive, meaningful parameters, such as hammer hardness, string length, etc.

 

The point is, it may well be using 'hardcore' physical modeling, in the truest sense of the term. We'll perhaps never know, given that the algorithms are proprietary.

 

Sorry, i was just making an example, referring to other post in the discussion; of course i know nothing at all about the Viscount; but their research lab is in my parent town, i should try to discover something :->.

 

BTW, welcome to the forum, Maurizio!

 

Thanks !

 

Maurizio

 

 

[Charleston]

The 1st in a tutorial series:

 

[video:youtube]

[Tobias Åslund]

Interesting to see in the video when explaining physical modeling there's a screenshot from the PD (or perhaps MAX) software. I guess they used this during the development of the algorithms.

 

Now - I'm still interested in the price and availability of these pianos! GAS

[Al Coda]

Interesting to see in the video when explaining physical modeling there's a screenshot from the PD (or perhaps MAX) software. I guess they used this during the development of the algorithms.

 

Now - I'm still interested in the price and availability of these pianos! GAS

 

Me too !

 

Technically, they use 6 TI OMAP dual core processors.

clonk

Only OMAP 4 and 5 are dual core using ARM Cortex A9 / A15 and Cortex M3 / M4 for DSP offload and support up to 8GB of RAM.

Originally we find OMAP in cellphones and tablet PCs, but the last 2 generations seem to be very powerfull.

OS might be Linux, Windows Embedded (or Android).

There´s a link to OpenMAX IL in the Wiki.

http://en.wikipedia.org/wiki/OMAP

 

A.C.

[Mark Schmieder]

Well, several months later, and many demos added to the original base (most of them NOT at the vendor's website), I am finding enough consistency to the overall timbre as being extremely bright, to feel that this piano is not for me personally.

 

I revisited this tonight and searched for more demos, because I have a line on a Roland RD700NX for half price and may pick it up tomorrow. I wanted to be sure.

[Ashville.Guru]

New video. In Italian, but you get an idea of what's going on:

 

[video:youtube]

[Al Coda]

New video. In Italian, but you get an idea of what's going on:

 

 

Cool, thx !

Very clear what´s going on !

Can´t wait for the demonstration of electric pianos.

 

A.C.

[non ce futuro]

Are there any news regarding price and availability? I just made a quick search but found nothing...

 

[Al Coda]

Are there any news regarding price and availability? I just made a quick search but found nothing...

 

My impression is, it´s still vaporware, but in constant developement.

IMO, this 2nd vid demonstrates the advanced modelling parameters for the acoustic piano model they now shortly developed.

Can be their upcoming vids will demonstrate one modelling engine after the other and ´til it´s finished, which might not be the case before Winter NAMM and Musikmesse 2013,- if at all.

 

I don´t expect availability very soon.

 

Nonetheless, it´s interesting observating progress.

 

A.C.

[Jaime Chimuelo]

PhysMods seem to evolve. For example PianoTeq sucked until vrs. 3.0 then it started getting better and now PTeq 4.1 os my favotie Piano, Beats the several sampled libraries. And my favorite part is I can edit each note, not that it';s necessary but I prefer certian small sections be detuned like the lowest ranges, and slight detuning in the octave area above 440 and it sounds so real. No need for some masking reverb etc.

It's got 5 mics you can move around to get the exact placement you want and a very slight amount of one of my DSP verbs and I am extremely happy.

I hope these guys keep developing it as it;s something that takes a while.

FWIW the newest Rhodes on the PTeq 4.1 is decent, but the Wurlitzer is spot on, so to have a really wqell tuned nice Steinway, a slightly detuned Upright Chickering Grand, and a Wurlitzer all with sostenuto and harmonic pedal, and total control of harmonic resonance inside of the lid, I can't see anything beating it unless it's some really dedicated expensive hardware unit.

Even then, I'd rather keep the controller and save 5 grand or whatever the added expense.

Played the V-Piano and it was so over hyped and average at best.

[Al Coda]

PhysMods seem to evolve.

 

New site up ...

clonk

 

All specs, manuals in different languages available as also firmware download.

Musicstore has ordered and expects delivery march 13/2013.

Musicstore

 

That´s the price for the wooden keys version (H1) incl. onboard speaker system.

Price in germany is EUR 3.999.- incl. 19% VAT.

 

German price for the stage piano version w/o speaker system and w/ standard weighted keys is EUR 3.499.- incl 19% VAT.

 

No new sounddemos of electric pianos and sample-engine up to now, but will come soon I think.

They will be at Musikmesse w/ this thing too.

It´s no vaporware anymore.

 

For a controller, I appreciate it accepts 3 standard momentary ft. switches & 2 MIDI CC pedals as well as the triple-pedal unit.

 

A.C.

[MalH]

Can't make a mathematical model for the sound of wood?

 

I do that as part of my day job!

 

Often in physically modelled pianos, only the hammer and strings are modelled. The soundboard and case are often applied using convolution with an impulse response. That way, the woodiness is applied as a filter.

 

It's easy enough to tell if this is the case. Just see how much soundboard control there is in the model.

[Tobias Åslund]

There's now an audio player on the site, showcasing a broad selection of pianos and other sounds of the Physis. They do sound good to my ear - I really like to try one out in person!

 

http://www.physispiano.com/audio.php

[Al Coda]

There's now an audio player on the site, showcasing a broad selection of pianos and other sounds of the Physis. They do sound good to my ear - I really like to try one out in person!

 

http://www.physispiano.com/audio.php

 

Some Rhodes sounds sound strange to me in the upper half of the keyboard.

I wonder if that´s a matter of preset programming or if the "modell" is not accurate enough.

No Wurli and Clav models demoed since the website w/ audio player is up and that´s since 2-3 weeks.

Thomann & Musicstore have Physis Piano advertised and available already, but for that price, I´d expect much better e-pianos.

The keyboard section is the weakest point IMO, the organs sound horrible.

 

It now has strong competitors,- not in the physical modelling department but as a performance keyboard instrument.

Casio PX5S and Studiologic NUMA Concert.

I was surprised about the quality of zawinulish Rhodes in NUMA Concert and I like it´s black and not too wide because it has the wheels on top of surface and offers a MIDI CC pedal input.

The Physis Piano should definitely be user tested in regards of programmability and results.

 

A.C.

[Tusker]

Can't make a mathematical model for the sound of wood?

 

I do that as part of my day job!

 

I've heard a number of really good wooden sonorities in physical models. At the risk of confusing Dana's point, (to me) a lot of modeled pianos have a metallic sheen, in addition to the wooden resonances they are modeling. This is analogous to the "glassy sheen" many of us notice in virtual analogue synths, some of which model filter resonance and overdriven VCA's quite beautifully.

 

Sometimes, the sonic signature of a particular technology is not visible until we stand back from it. Sometimes it's a matter of years.

 

I am glad we are living in a exciting period in digital music history, when physical modeling is doable.

 

Jerry

[kanker.]

The some of the piano sounds are pretty impressive. The Rhodes are terrible.

[Josh Paxton]

The some of the piano sounds are pretty impressive. The Rhodes are terrible.

 

 

This. Some of the Rhodes sounded weirdly like they were trying to be Wurlis, but no actual Wurli models.

[Tobias Åslund]

My local store got a Physis H1 today! I only tried it for 20 min and it was on the top tier of a wobbly Spider stand... but anyway - it feels and sounds really good, although I'm not blown away.

 

I told the manager he need to put it on a proper stand - and he will, so I'll probably do a proper test next week. Will probably post a mini review after that.

[Al Coda]

... so I'll probably do a proper test next week. Will probably post a mini review after that.

 

That would be great.

Please try tweaking the electric pianos and tell us about the results.

I liked some of the acoustic pianos from the audio demos but was somewhat shocked about the Rhodes.

I´d be interested to hear something about clav and wurli sounds too.

Since we have options in lower price ranges like Casio PX5S and NUMA Concert, Physis Piano should really sound excellent to be a competitior and worth to buy.

 

A.C.

[Tobias Åslund]

I was quite unimpressed by the Clavis, but the Rhodes and Wurlis felt nice.

 

I'm already quite sure people won't think the Physis feels and sounds $2.000 better than a Casio PX5s, but IMHO it's an interesting and very musical digital piano.

[Al Coda]

I was quite unimpressed by the Clavis, but the Rhodes and Wurlis felt nice.

 

I'm already quite sure people won't think the Physis feels and sounds $2.000 better than a Casio PX5s, but IMHO it's an interesting and very musical digital piano.

 

Were electric piano sounds (except clavi) significantly better that the audio demos made believe ?

 

How good was the action ?

I have the impression, most so called "hi-end actions" now are basicly all Fatar TP40GH / TP40GH wood, slightly modified by the manufacturer (I think even the Lachnit is ...).

What was/is your impression ? Did you try fast repetitons of a single note ?

 

I still like that kind of customizable touch screen UI and how many ft.-switches and MIDI CC pedals can be connected which is worth some extra bucks IMO.

 

What is your impression in regards of UI and build quality ?

 

If you cannot comment on that now,- please keep us informed.

 

A.C.