Carving site recommendations for plate tuning

[Marty Kasprzyk]

We should disassemble great sounding violins and measure all of their component parts' mechanical properties (reverse engineering) and then do the same for lousy sounding ones.  This data should be published.  I don't believe it is possible to duplicate the whole assembly's performance unless the individual parts are duplicated.

This hasn't happened in the past and I very much doubt it will happen in the future.  The owners of great instruments don't want lots of new violins made as good as them.  Science and engineering is a threat to them.

 

 

[Violadamore]

34 minutes ago, Marty Kasprzyk said:

We should disassemble great sounding violins and measure all of their component parts' mechanical properties (reverse engineering) and then do the same for lousy sounding ones.  This data should be published.  I don't believe it is possible to duplicate the whole assembly's performance unless the individual parts are duplicated.

This hasn't happened in the past and I very much doubt it will happen in the future.  The owners of great instruments don't want lots of new violins made as good as them.  Science and engineering is a threat to them.

 

 

I believe that the static measurements could be derived from existing tomography techniques, without pulling any fiddles apart, but that cracking the nut of dynamic sound production to produce a generalized design approach is necessary, and we don't have tools to microscopically record every deflection of an assembled violin while it's being played, yet.  I'd like to analyze 3-D schlieren sound field videos for each one, too, and that sort of thing is still developing. 

There's probably enough great violins in public hands to work with. 

[FiddleDoug]

2 hours ago, Violadamore said:

Shoot, if you're working at that level, forget the silly violins, and PM me the blueprints and manual set for a hyperdrive.   

Actually, he has rounded it to 42. It's actually 42.0176543.

[Peter K-G]

Setting "the sound" to the side and focus on objectives to the OP.

As I posted previously.

Once the arching is done, there is very little room for maneuvering when carving out and trying to tune to a specific modes.

Here is two examples of back plates measured when hollowing out, graduating. Even if trying to maneuver (image 2) it always ends up where it will be according to the arching and properties of the wood.

(The last rise in frequencies are after UV box)

 

 

 

 

[Don Noon]

2 hours ago, Violadamore said:

I just pitched what could just as well be an answer to this thread into the "sound under your ear" thread (it all ties together), and estimated a solution would consume $20 billion and around 9 years of development..  You think I underestimated? 

I think you overestimated "solution" .

[HoGo]

5 hours ago, Violadamore said:

Unfortunately, before you can computerize something, as you note, you have to have data to process.  With faces, or cars, or license plates, you have images to compare.  Even if the "top makers" had data on hand to contribute, exactly what would that data have to be?

How do you even define a "good violin", much less dictate a prescription for carving one, first time, every time?  Please explain this to me.  If you are running on the assumption that the job at hand is to recreate Cremonese classics from new materials, then please give me the list of numbers to code to functionally replicate a Stradivarius from a stack of tonewood. 

Most folks keep the notion that math is about exact numbers. That is true for high school math and not at all some modern branches of math. Modern data-mining, AI, genetic algorithms and related branches of science can solve much more problems unimaginable in the past even without exact numbers to crunch like perhaps someCAD modelling and FEM methods.

[David Beard]

Are the most loved historical Strads and Del Gesu identical in sound and response?  No.

So, is there one specific exact ideal for the violin?  At best, maybe there is a 'sweet spot', a range of 'best'.

Seeking a single static numerical ideal for violins is logically flawed from the start.

Second, consider the fundamental premise of plate tuning, that the plate resonant frquencies 'characterize' the resulting instrument.

The frequencies are measureable. So you can write them down.  And they differ from one instrument to another.  So you can say if two fiddles match in their frequencies or don't.  And you can manipulate the frequencies in free plates to some extent.  So, you get to feel like you're really doing something.

But:  

If two fiddles have matching frequencies, do they sound or respond the same?  NO.

Do good fiddles have only certain frequencies or pattern of frequencies?  NO. 

(They just scatter across a broad range of normal.  Which you automatically get by making basically normal plates.)

Are there special frequencies that are seen in good instrument that aren't also seen in bad instruments?  NO.

 

So there's no correlation.  You can aim for whatever reasonable frequencies you choose, but in itself that does not constitute aiming at a good result.

 

 

[Marty Kasprzyk]

1 hour ago, David Beard said:

Are the most loved historical Strads and Del Gesu identical in sound and response?  No.

So, is there one specific exact ideal for the violin?  At best, maybe there is a 'sweet spot', a range of 'best'.

Seeking a single static numerical ideal for violins is logically flawed from the start.

Second, consider the fundamental premise of plate tuning, that the plate resonant frquencies 'characterize' the resulting instrument.

The frequencies are measureable. So you can write them down.  And they differ from one instrument to another.  So you can say if two fiddles match in their frequencies or don't.  And you can manipulate the frequencies in free plates to some extent.  So, you get to feel like you're really doing something.

But:  

If two fiddles have matching frequencies, do they sound or respond the same?  NO.

Do good fiddles have only certain frequencies or pattern of frequencies?  NO. 

(They just scatter across a broad range of normal.  Which you automatically get by making basically normal plates.)

Are there special frequencies that are seen in good instrument that aren't also seen in bad instruments?  NO.

 

So there's no correlation.  You can aim for whatever reasonable frequencies you choose, but in itself that does not constitute aiming at a good result.

 

 

You would have better results if you duplicated the plate weights in addition to duplicating their resonance frequencies.

[Davide Sora]

4 minutes ago, Marty Kasprzyk said:

You would have better results if you duplicated the plate weights in addition to duplicating their resonance frequencies.

Perhaps, but if the wood is not appropriate the risk of failure is the same, or maybe worse.

[Violadamore]

4 hours ago, HoGo said:

Most folks keep the notion that math is about exact numbers. That is true for high school math and not at all some modern branches of math. Modern data-mining, AI, genetic algorithms and related branches of science can solve much more problems unimaginable in the past even without exact numbers to crunch like perhaps someCAD modelling and FEM methods.

You aren't discussing this with "most folks".  I'm trying to tell you that a reliable conceptual construct leading to a general solution of the case does not yet exist.  You can't analyze any data, precise or fuzzy, without an understood model based on general equations to fit it to.

Anyway, the folks here will require particular solutions (i.e., the previously disdained exact numbers) to guide their cutting.  This is neither the "social sciences", nor pure research.  Cutting off the wrong wood because the solution offered is trendily fuzzy (or based on false assumptions) is not an option.  They're better off so far by using their traditional practices.

 

 

[Don Noon]

2 hours ago, Marty Kasprzyk said:

You would have better results if you duplicated the plate weights in addition to duplicating their resonance frequencies.

2 hours ago, Davide Sora said:

Perhaps, but if the wood is not appropriate the risk of failure is the same, or maybe worse.

If you started with the same wood properties, maybe it would work OK.  If the wood is has different properties, they you'd have to contort the arching and/or graduations to try to match a few free plate frequenciies.

But, even the idea of starting with the same wood properties is an impossible task, if you want to try to match all relevant acoustic properties at all places in the plate.  I'm now working on a second viola plate, external arching roughed out on CNC, same-tree set... yet the weights and taptones are coming out a bit different.  Better or worse, who knows.

 

[Marty Kasprzyk]

If you are trying to duplicate a good instrument my point is this: if you match one property (tap tone 5 of the top plate is an example)  instead of nothing you will get better results.  If you match two things (tap tone of the top plate and its weight) you will get better results than just matching one thing.  If you match three things it's better than matching just two things and so on.    The more you match things the more likely you are to closely duplicate it.  

There's lots of parts all of which have an influence that should be duplicated.  It's not complicated-- just very tedious and time consuming.

Why bother?  Skip all this testing stuff and just make more instruments.

Even a dumb squirrel  finds a nut sometimes.

 

[David Beard]

8 hours ago, Marty Kasprzyk said:

If you are trying to duplicate a good instrument my point is this: if you match one property (tap tone 5 of the top plate is an example)  instead of nothing you will get better results.  If you match two things (tap tone of the top plate and its weight) you will get better results than just matching one thing.  If you match three things it's better than matching just two things and so on.    The more you match things the more likely you are to closely duplicate it.  

There's lots of parts all of which have an influence that should be duplicated.  It's not complicated-- just very tedious and time consuming.

Why bother?  Skip all this testing stuff and just make more instruments.

Even a dumb squirrel  finds a nut sometimes.

 

So if I match the varnish color I will be closer to matching the sound?

No.  Not everything that is 'matchable' is also relevant to tonal result.

 

Tap tones are 'matchable'.  You can match one or several, may be all?

What I'm saying is that the assumption they are relevant to tonal and response outcome is actually false.  As in 'the tap tones are not actually relevant to tonal result'.

The violin is not a xylophone.  We don't play violins by tapping or otherwise passively exciting the natural resonance frequencies.  That is the false assumption.

[Don Noon]

4 hours ago, Marty Kasprzyk said:

If you are trying to duplicate a good instrument my point is this: if you match one property (tap tone 5 of the top plate is an example)  instead of nothing you will get better results.  If you match two things (tap tone of the top plate and its weight) you will get better results than just matching one thing.  If you match three things it's better than matching just two things and so on.    The more you match things the more likely you are to closely duplicate it.  

3 hours ago, David Beard said:

What I'm saying is that they a relevant to tonal and response outcome is actually false.  As in 'the tap tones are not actually relevant to tonal result'.

I am likewise of the opinion that taptones and many other common measurements... as long as they are not wildly out of normal bounds... are not particularly relevant to the important character of the instrument.  Spending too much time obsessing over them I don't think would be productive, and can distract from giving attention to things that may actually be more important.

 

[Peter K-G]

7 hours ago, Marty Kasprzyk said:

If you are trying to duplicate a good instrument my point is this: if you match one property (tap tone 5 of the top plate is an example)  instead of nothing you will get better results.  If you match two things (tap tone of the top plate and its weight) you will get better results than just matching one thing.  If you match three things it's better than matching just two things and so on.    The more you match things the more likely you are to closely duplicate it.  

There's lots of parts all of which have an influence that should be duplicated.  It's not complicated-- just very tedious and time consuming.

Why bother?  Skip all this testing stuff and just make more instruments.

Even a dumb squirrel  finds a nut sometimes.

 

Like!

This is basic logical thinking.

[bungling_amateur]

7 hours ago, David Beard said:

So if I match the varnish color I will be closer to matching the sound?

No.  Not everything that is 'matchable' is also relevant to tonal result.

 

Tap tones are 'matchable'.  You can match one or several, may be all?

What I'm saying is that they a relevant to tonal and response outcome is actually false.  As in 'the tap tones are not actually relevant to tonal result'.

The violin is not a xylophone.  We don't play violins by tapping or otherwise passively exciting the natural resonance frequencies.  That is the false assumption.

Maybe different pigment loads have subtly different densities and mechanical properties, affecting the sound in a very subtle way?

I think there are two questions. First, whether the aim is to duplicate a specific exemplar, or whether the aim is to make a "good" final product (depending on the definition of "good").

If the aim is to duplicate an exemplar then there must be some kind of hierarchy of influences. Gross shape is fairly high up that hierarchy; arching might be considered part of that. Materials choice is also high up, and then we get stuck because each piece of wood is unique. Varnish colour and tap tones may have a different position in this hierarchy. the law of diminishing returns suggests that some of these aspects are worth putting more effort into than others.

I'm actually kind of fascinated by the general and wider scene of making (and purchasing and using) "replicas" of historic objects which concentrate on replicating incidental aspects and neglect gross structural or functional elements. A lot of this must be connected to marketing I think.

[HoGo]

10 hours ago, Violadamore said:

You aren't discussing this with "most folks".  I'm trying to tell you that a reliable conceptual construct leading to a general solution of the case does not yet exist.  You can't analyze any data, precise or fuzzy, without an understood model based on general equations to fit it to.

Anyway, the folks here will require particular solutions (i.e., the previously disdained exact numbers) to guide their cutting.  This is neither the "social sciences", nor pure research.  Cutting off the wrong wood because the solution offered is trendily fuzzy (or based on false assumptions) is not an option.  They're better off so far by using their traditional practices.

Perhaps I was not clear enough. You don't always need general equations or precise model, actually this is what we cannot have as behavior of violin is way too complicated to model it. You can work more in line of machine learning/ black box analysis way  (usinf AI or other methods) that can make prediction, it actually simulates what human does during learning, even we do learn it without any hard exact input.

What I propose is maker (perhaps high quality company would be better candidate with bigger output) takes care during making and measures and records everything about the materials and process of each single instrument (densities, grain widths, speed of sound, modes, tap tones, model size, size of f holes, arch shape and height, thicknesses etc... The more the better) in a VERY consistent way. That would be pretty much what many luthiers do anyway.  The resulting violin should be evaluated for quality and assigned some marks (volume, sonority ratings etc..) Some of the input data would be exact, some could be just mere descriptions clear to the maker (arch shape could be described in few parameters, like fullness from 1 to 5, 5 being very roundish etc...). Perhaps some parameters could be just graded like low-medium-high (numbered 1-2-3) but if possible more precision. Good maker can, after he makes and judges his instruments (must work consistently), see what works (high ratings of result) and what does not (low ratings) and that's exactly what the propely programmed ML/AI system can do (there are no eqautions or physical models required, just the data and consistent method). WIth enough instrument the system can show which variables are irrelevant and which important so some redundancy at start is needed to get clear view of what's going on.

The result would be that you take a piece of spruce and maple and measure the physical properties (density, speed of sound etc.) and the system can predict the outcome with some probability using this wood and suggest best combination of other properties like arching, thicknesses etc to get best result (or desired result - soloist versus quartet instrument...) or even suggest what combination of wood properties will give you the best possible outcome.

This may sound like Sci-fi but similar systems have been used in some large technological companies (they have the cash) for decades, not only internet-oriented but also bio-medical, chemical...

Of course all this is going to cost money and there isn't probably enough cash flow in violin making to fund this and top makers don't make as many instruments. Big companies (like Yamaha) could do this but I think the advantage is not as high for them when they can cheaply mass produce in China (they don't need to offer superior product while they are cheaper than similar comparable). Perhaps companies like Collings or Taylor could step in this direction.

 

 

[Marty Kasprzyk]

10 hours ago, David Beard said:

So if I match the varnish color I will be closer to matching the sound?

No.  Not everything that is 'matchable' is also relevant to tonal result.

 

Tap tones are 'matchable'.  You can match one or several, may be all?

What I'm saying is that the assumption they are relevant to tonal and response outcome is actually false.  As in 'the tap tones are not actually relevant to tonal result'.

The violin is not a xylophone.  We don't play violins by tapping or otherwise passively exciting the natural resonance frequencies.  That is the false assumption.

You're right, a violin is not like a xylophone with separate individual bars tuned to certain vibration frequencies.

 In a violin body all the vibrating parts are glued together.

[David Beard]

8 hours ago, Marty Kasprzyk said:

You're right, a violin is not like a xylophone with separate individual bars tuned to certain vibration frequencies.

 In a violin body all the vibrating parts are glued together.

The big difference is it's a driven system.

Only the stopped and unstopped strings are excited to their natural resonance frequencies.

The rest is driven by the input from the strings.  This means the resonances are pushed off their natural frequencies and can only set up standing waves at frequencies that follow and conform to the driving input.

In such a driven response, how well the resonances bend to comply is more important.  Their natural frequencies they bend away from become rather less specifically important.

 

The violin is also not a steel drum.

 

[Violadamore]

1 hour ago, David Beard said:

The big difference is it's a driven system.

Only the stopped and unstopped strings are excited to their natural resonance frequencies.

The rest is driven by the input from the strings.  This means the resonances are pushed off their natural frequencies and can only set up standing waves at frequencies that folliw and conform .to the driving inout.

In such a driven response, how well the resonances bend to comply is more important.  Their natural frequencies they bend away from become rather less specifically important.

 

The violin is also not a steel drum.

 

It is, however, one of the most bizarre "step-down transformers" ever invented. 

Funny you mention the drum.  If you examine some of the most primal Asian string instruments, you might wonder if the original idea came from a bored somebody idly playing around with a bow (or two) and a tom-tom. 

[violguy]

I can hardly wait until Italian archaeoligists discover a hidden cache of Tony's spreadsheets!

[Marty Kasprzyk]

20 hours ago, David Beard said:

The big difference is it's a driven system.

Only the stopped and unstopped strings are excited to their natural resonance frequencies.

The rest is driven by the input from the strings.  This means the resonances are pushed off their natural frequencies and can only set up standing waves at frequencies that follow and conform to the driving input.

In such a driven response, how well the resonances bend to comply is more important.  Their natural frequencies they bend away from become rather less specifically important.

 

The violin is also not a steel drum.

 

A violin is indeed like a steel drum--if you hit it, it makes a sound.  

The frequency response curve of the sound coming from a hammer impact to the bridge is the same as when the strings are bowed in a glissando.  Both methods excite the various body resonances the same way.

The attached paper describes this more thoroughly.

AilinZhang_JimWoodhouse_JASAJournal.pdf

[ctanzio]

20 hours ago, David Beard said:

The rest is driven by the input from the strings.  This means the resonances are pushed off their natural frequencies and can only set up standing waves at frequencies that follow and conform to the driving input.

This is a fundamental truth of driven systems that is difficult for many people to comprehend.

To try and state it more clearly: when you play a note on the violin, the standing waves that develop in the body have the same frequencies of the vibrating string, not the natural frequencies of the violin body.

 

[Carl Stross]

19 hours ago, Violadamore said:

It is, however, one of the most bizarre "step-down transformers" ever invented. 

Funny you mention the drum.  If you examine some of the most primal Asian string instruments, you might wonder if the original idea came from a bored somebody idly playing around with a bow (or two) and a tom-tom. 

Why don't you just ask ? You know there is somebody here who knows everything :

 

[Carl Stross]

21 hours ago, David Beard said:

The rest is driven by the input from the strings.  This means the resonances are pushed off their natural frequencies and can only set up standing waves at frequencies that follow and conform to the driving input.

Even theoretically, only in some specific cases. Violin does not seem to me to be one of those. Not even close.