Looking at a sound spectrum and thinking about possible improvements for the sound. (Or better leave it as it is?)

[Andreas Preuss]

7 hours ago, Don Noon said:

unless you like crackly, jangly sounds.

In a certain way I do like a crackly, jangly sound. And as I said before it’s to me largely a question of ‘how much’. You could almost say how much of red pepper in hot chili beans. When Ray Chen talks about ‘texture’ I think this is exactly what he means, just with a politer expression. (At 11:40)

i think the amount of how much is tolerable depends the frequency response  below the roughness Formant, more or less the bridge hill. I would never aim at something where the bridge hill is lower than the very high frequencies. (Makes me wonder if this is possible at all?)

For a comparison how this actually affects our perception I listened to the Quiztones app. Though it is a bit unrealistic, it gives IMO some ideas.

[Andreas Preuss]

For completeness, here are the graphs of the same violin in the final stage. For my personal taste it sounds excellent right now. A lot of overtones on all notes, and a sound with texture. In the graph it looks like this:

I circled what I am looking at most, the frequencies over 2kHz. If there are enough peaks over 2kHz at approximately the same amplitude its for me a good sign for a crisp and penetrating sound. (red circle)

If there are additionally frequency peaks over 4kHz making a sort of platform it is even better and adds a certain taste to the sound. I think it is better visible on a low resolution graph: 

 

All this represents my personal taste of what I define as an interesting sound. I could repeat this now at least on three instruments including the new concept violin and got each time very positive comments from professional violinists on it. 

I marked what I am guessing is what makes the characteristic of a sound. 

1. how much the yellow line (broadly speaking the bridge hill resonances) is sloped and how far it reaches into high frequencies. If I can, I try to reach the region of 4 Khz. At least I have repeatedly seen that if frequency peaks over 2.5kHz are not well developed, the sound can be still loud, but lacks colors.

2. How much it drops from the first platform to the second platform. (red arrows). I think this defines somehow how much texture (sizzle) is in the sound. The smaller the difference the more texture the sound will have. (unconfirmed hypothesis)

3. How far the second green platform of the roughness formant reaches into high frequencies. At best I think it should roughly cover one octave. (green arrows)

 

[Don Noon]

I wouldn't ignore the energy below 2 kHz, down to maybe 1500 Hz, which is the fundamental of the high E string.

Over 4 kHz is more of a preference thing, for the jangle and crackly overtones.  Probably sweeter sounding with less of it, and as I get older and lose hearing in that range, it doesn't matter all that much anyway.

[Andreas Preuss]

1 hour ago, Don Noon said:

I wouldn't ignore the energy below 2 kHz, down to maybe 1500 Hz, which is the fundamental of the high E string.

?

On the E string thats the last octave  at the end of the fingerboard. Don’t want to say it is unimportant, but before I rather get the D string working well, which is always tricky in the high playing positions.

————————————-

From my unscientific experience working with those graphs, I look at them searching for patterns and constructional methods how to get to the ‘ideal’ pattern. (Who doesn’t?) I would call the pattern above ‘two platform pattern’.  I think it is able to describe how the violin sounds by the parameters I listed and it does describe IMO the general characteristic of the sound much better than the signature modes.

This means in detail how wide and how much sloped each platform is and how they relate to each other. The only thing I find worrisome in the signature mode zone is if there is an amplitude sticking out unproportionaly or the whole ‘mass’ looks too big in relation to the whole mass above. 

In the end, the harmonics of a note decide to a large degree if we like a sound or not.

—————————————

Yes, I noticed that your violins have a silky and smooth texture, definitely not bad. you might become a name for that sound. 

[David Stiles]

3 hours ago, Andreas Preuss said:

you might become a name for that sound. 

'High Noon' sound! 

[Don Noon]

As always, it's a work in process.

My violins started out being just "loud and powerful"... but tonally not so hot.  After several years of working to get a more refined sound, I'm pretty happy... but as alway, looking for (what I think might be) improvements, primarily along the lines Andreas has described, but maybe slightly less jangly / crackly.  It's all trial-and-error and experience, plus whatever can be learned from other makers.

 

[Andreas Preuss]

I was lucky. A violinist friend passed by my shop and made a recording of the violin. In the meantime I didn’t touch anything on the setup.

 

Enjoy!

レヂュック.m4a

[Don Noon]

Sounds clear and bright, with the A0 jumping out somewhat on the open D string.  But I guess that's what you get with a Guarneri model, unless you want to go un-Guarneri thin to get the A0 frequency down.

1.7g for a bridge always ends up sounding too harsh for my tastes, but it would enhance the high frequencies.  I usually use 1.8-1.9g... but I also use torrefied wood for everything, which tends to be brighter than "normal" wood (at least immediately off the bench).

[Andreas Preuss]

8 hours ago, Don Noon said:

1.7g for a bridge always ends up sounding too harsh for my tastes,

Did you ever test such a harsh (or rough) sounding setup in a bigger hall? 
My experience is that you still can hear it in the first row but further back not any more. 

[David Beard]

I don't think these charts tell us the critical thing with high frrquency content, which is how much is signal versus noise.

Our 'ear' naturally evaluates this qualitatively.    If there is too much noise, we use adjectives like harsh or ugly.   If there isn't enough high freqeuncy signal in general we see the sound is dull or unclear.   And if there is ample high signal compared to noise we say things like brilliant, clear, articulate, crisp.

 

Using tech instead of the ear is 'off point'.   The measure that matters here is the subjective evaluation of a skilled and developed ear.    And, as it turns out, the ear is also the only tool that can easily evaluate these things.

Maybe some day the tech will become more helpful.  But that seems a long way off.

[Andreas Preuss]

8 hours ago, David Beard said:

I don't think these charts tell us the critical thing with high frrquency content, which is how much is signal versus noise.

Our 'ear' naturally evaluates this qualitatively.    If there is too much noise, we use adjectives like harsh or ugly.   If there isn't enough high freqeuncy signal in general we see the sound is dull or unclear.   And if there is ample high signal compared to noise we say things like brilliant, clear, articulate, crisp.

 

Using tech instead of the ear is 'off point'.   The measure that matters here is the subjective evaluation of a skilled and developed ear.    And, as it turns out, the ear is also the only tool that can easily evaluate these things.

Maybe some day the tech will become more helpful.  But that seems a long way off.

 I do agree to this 50 percent. The human ear as an evaluation tool is very precise in one sense, but lacks the objectivity of sound recorded graphs. Therefore my general approach was always to play and listen to an instrument first and only thereafter look in a sound graph for detectable patterns which might be able to explain something. So if I reverse this and look into a graph now, I see the FFT spectrum as an indicator for what I can expect, but I do not see it as the ultimate truth. 

i tried to explain this above. I have seen the pattern repeating and each time this pattern of high frequencies results in a certain sound timbre. If you get those two platforms, it is in my view an indication for an interesting sound. It’s in end only my own bias, but I am not trying to set up an universally valid sound theory. 
 

Working with graphs a kind of taught me where it is easy to manipulate a graph in a certain direction and where not. And I think the most telling parts of a sound graph are the high frequency bands. 
 

i do not see that an improvement in recording technique is the answer. We need a better understanding how to interpret graphs in a better way, or how to make a better translation of it. This includes as well to look not only at one sound graph but several taken under different conditions. 

[Marty Kasprzyk]

11 hours ago, David Beard said:

I don't think these charts tell us the critical thing with high frrquency content, which is how much is signal versus noise.

Our 'ear' naturally evaluates this qualitatively.    If there is too much noise, we use adjectives like harsh or ugly.   If there isn't enough high freqeuncy signal in general we see the sound is dull or unclear.   And if there is ample high signal compared to noise we say things like brilliant, clear, articulate, crisp.

 

Using tech instead of the ear is 'off point'.   The measure that matters here is the subjective evaluation of a skilled and developed ear.    And, as it turns out, the ear is also the only tool that can easily evaluate these things.

Maybe some day the tech will become more helpful.  But that seems a long way off.

Noise isn't necessarily harsh or ugly.  

The Audacity plot of the sentence: "Whisper something sexy." shows the signal is mostly noise.

[Michael Darnton]

Evaluating a violin from a spectrum snapshot is like a coach trying to figure out why his team lost by looking only at the final score. Surely someone here gets that? Players do.

[David Beard]

1 hour ago, Marty Kasprzyk said:

Noise isn't necessarily harsh or ugly.  

The Audacity plot of the sentence: "Whisper something sexy." shows the signal is mostly noise.

Marty, how are you ascribing some of that as noise?   High frquency is not automatically noise, as in 'not signal'.

Even 'noise' is 'signal' if it comes from the performers articulation etc.

What is 'noise' instead of being 'signal' is sound radition that does not contribute to a clear image of the 'playing'.

That would be harmonics that are distorted because something is behaving inelastically in the instrument.  Vibrations that jangle around the instruments too long and loose their time relationship to the playing.  Or too independant resonances that get stimulated with too little relation to the playing. Etc.

I'm not aware of any tech other than the ear that can evaluate 'playing image' versus extraneous or detrimental noise??

[David Beard]

5 hours ago, Andreas Preuss said:

 I do agree to this 50 percent. The human ear as an evaluation tool is very precise in one sense, but lacks the objectivity of sound recorded graphs. Therefore my general approach was always to play and listen to an instrument first and only thereafter look in a sound graph for detectable patterns which might be able to explain something. So if I reverse this and look into a graph now, I see the FFT spectrum as an indicator for what I can expect, but I do not see it as the ultimate truth. 

i tried to explain this above. I have seen the pattern repeating and each time this pattern of high frequencies results in a certain sound timbre. If you get those two platforms, it is in my view an indication for an interesting sound. It’s in end only my own bias, but I am not trying to set up an universally valid sound theory. 
 

Working with graphs a kind of taught me where it is easy to manipulate a graph in a certain direction and where not. And I think the most telling parts of a sound graph are the high frequency bands. 
 

i do not see that an improvement in recording technique is the answer. We need a better understanding how to interpret graphs in a better way, or how to make a better translation of it. This includes as well to look not only at one sound graph but several taken under different conditions. 

I also kinda agree with you 50%.

I also don't mind some grit or noise if that is the cost for a full and faithful playing image from the instrument.

But, I believe the ear is the proper main tool to evaluate.

Sure, I can see that a visual graph on occasion might help in analyzing some specfic diffficulty.  But this is limited.  Ultimately it is the musical ear that we must satisfy.

Understanding on a mechanistic level can sometimes help, but it can also at times mislead.   Tech is not the ear.   Regardless of our technicsal understanding, the ear is the final target.  Even when we don't understand why, satisfying skilled ears is the only actually relevant measure.  It is the measure closest to the issue.

[Andreas Preuss]

7 hours ago, David Beard said:

Marty, how are you ascribing some of that as noise?   High frquency is not automatically noise, as in 'not signal'.

Even 'noise' is 'signal' if it comes from the performers articulation etc.

What is 'noise' instead of being 'signal' is sound radition that does not contribute to a clear image of the 'playing'.

That would be harmonics that are distorted because something is behaving inelastically in the instrument.  Vibrations that jangle around the instruments too long and loose their time relationship to the playing.  Or too independant resonances that get stimulated with too little relation to the playing. Etc.

I'm not aware of any tech other than the ear that can evaluate 'playing image' versus extraneous or detrimental noise??

It seems that if you filter out those high frequencies that the listener perceives less noise. This must be the reason why Dunnwald called it noise Formant. 
 

however, and I think I pointed this here out repeatedly, ‘noise’ is a very interesting and IMO important ingredient in the sound of a violin. Balanced dosage is everything. In those terms I think the word ‘noise’ is largely misleading and should be replaced by the word ‘texture’. 

[Andreas Preuss]

8 hours ago, Marty Kasprzyk said:

Noise isn't necessarily harsh or ugly.  

The Audacity plot of the sentence: "Whisper something sexy." shows the signal is mostly noise.

 Spoken words are not intended to ‘sing, anyway. Makes me wonder how the spectrum of Louis Armstrongs singing voice looks like. 

[David Beard]

34 minutes ago, Andreas Preuss said:

It seems that if you filter out those high frequencies that the listener perceives less noise. This must be the reason why Dunnwald called it noise Formant. 
 

however, and I think I pointed this here out repeatedly, ‘noise’ is a very interesting and IMO important ingredient in the sound of a violin. Balanced dosage is everything. In those terms I think the word ‘noise’ is largely misleading and should be replaced by the word ‘texture’. 

I think we're missing each other here.

I'm suggesting that not all high frequency is created equally.

And not all 'noise' is created equal either.

Some is extraneous.  Some is not part of the playing sound, or the articulation.

But some 'noises' are meaningful parts of the playing articulation.

 

Likewise, some high frequncy content is harmonious upper partials of the musical content.   And some is legitimate articulation of the playing.

But, some upper frequency content and some 'noise' content are non contributing, not part of the playing, 'junk noise' and 'junk partials'.   

The problem is our conversations, and the graphs, tend to lump 'junk noise' and 'articulation noise' together.  We shouldn't.  Our ears don't.   

And we tend to lump 'junk partials' together with 'harmonious partials' as all simply 'high frequency'.  But we shouldn't.  And again, our ears don't lump these together.

 

 

[Andreas Preuss]

3 hours ago, David Beard said:

I'm suggesting that not all high frequency is created equally.

And not all 'noise' is created equal either.

Some is extraneous.  Some is not part of the playing sound, or the articulation.

I see, I think I got your point.

So this means that there is the sound created by the strings AND a second overlapping  ‘mechanism’ which can add to that ‘foreign’ frequencies? Long ago, as you know, I suspected apparently wrongly longitudinal waves. 

In any case I find the ‘lump junk’ portion of a sound spectrum most interesting. And I think it is worth the effort to explore the ‘peak jungle’ a bit closer.

[David Beard]

Yeah.

We surely agree that high frrqiencies are important, and insufficeintly understood.

 

[Don Noon]

22 hours ago, Michael Darnton said:

Evaluating a violin from a spectrum snapshot is like a coach trying to figure out why his team lost by looking only at the final score. Surely someone here gets that? Players do.

While I certainly don't go along with the Dunnwald approach of quantifying violin quality via spectral analysis, I definitely think a spectrum is a useful tool to help understand what is happening acoustically... in conjunction with player evaluations.  It is at least one way to keep track of history in a more concrete way, with player evaluations being rather squishy... especially if you're trying to decide how a violin has changed after some mods that take time.

Even the final score of a game is important, to know the difference between your team and the other one.  If you don't know the score, you wouldn't even know if you won or lost.  But the details of the game matter too.

[Michael Darnton]

@Don Noon Thanks for being one person to notice.

Here's one example of the problem: one of the most important features for a player is that when bowing in different places on the string the player expects to get a different tone quality out of the violin. As I understand FFTs, that means that widely varying positions would give different FFTs, because different parts of the spectrum are being chosen by elevating or suppressing them. Thus, no single FFT can account for all of those various voices.

That's why I've often accused modern makers of making the "one-stop organ" version of the violin: they look for The Strad Tone, define it as one single beautiful voice (which it certainly is not), defined by something like an FFT, and then try to recreate a violin that makes that single-voiced FFT. And that is most certainly not what a fine violin is.

Perhaps the most common criticism I hear of modern violins is that they only offer one voice. I have even had a maker tell me that his violin was good because  no matter how you played it, you got that one "good" voice--this was his objective. You couldn't create an ugly sound (by his judgement) even if you wanted.

When I suggested this previously I was assured by someone that no, his violins had that multiple-voice quality. I don't believe it. I see very few violins of any vintage that offer this, and certainly not modern ones. I think that people test their own violins, hear a little of this and convince themselves that it's there, when in reality the difference should be huge--from nasty and gritty to so foggy that it's hard to tell exactly what note is being played. I use the example of HH who does all the movements up and down the string, but no change comes out. I am sure she has convinced herself that something is happening or she wouldn't do that. Certainly students are told right from the start that changes happen. . .  when they are playing violins so bad that there's no chance of this happening for them. Do they learn to "hear" it happening when it's not, figuring that it must be something so small that they can't yet hear it with untrained ears? I think that may be happening. And then they never do learn to hear it. These are the ones who buy new violins and are happy.

Some players have told me the moment they have discovered this with a violin that actually did it, as if a light bulb of possibilities suddenly went on. On the other hand, I suspect that  some players encounter this and discard it immediately as a violin that gives them back things they can't expect and can't control (because they don't understand the possibilities and can't at that moment control them).

I believe that most players and makers are delusional on this issue: they think their violin offers this and that they control it, sort of taking it all on faith. In my career the most difficult customers were the ones who understood this phenomenon, knew how rare it was, and were looking for it. In their budget they knew how hard it was to find, but they kept looking. I did not understand it myself until I had to deal with them.

I just had two real players this week, one a prominent teacher, one a successful chamber musician, tell me the same thing: that it's a waste of time even looking at modern instruments because modern makers don't understand this. We were talking about "great" modern makers--household names--that they both felt were perpetrators of BS smokescreens in dealing their instruments. I hear this all the time. Our shop gets plenty of customers with no interest at all in modern instruments, because they know what's going on here. Deny it all you want that modern instrument fail on this. Consider if maybe you are wrong in denying it. 

Now that's not the only issue under this same umbrella of FFT problems. It's just the best one that doesn't require "golden ears" (an idea disparaged most by those who don't have them!) There are other things as well that FFT's don't show that are equally important.

There literally is no substitute for ears and experience. FFT is the dullest of knives. Further, I believe that following this path of trying to make the violin behave in a certain way actually contributes to failure because instead of opening up the violin, it leads to the maker constraining the violin to give him the FFT of the limited sound he thinks is the "right" one. A violin that is less than the possibilities.

[Marty Kasprzyk]

12 hours ago, Andreas Preuss said:

>

So this means that there is the sound created by the strings AND a second overlapping  ‘mechanism’ which can add to that ‘foreign’ frequencies?

>

 

I disagree.   All the sound comes from the strings.

The violin body makes the input string bowing sounds more audible and it also acts as a filter which makes some output sounds relatively weaker or louder. If it is in good condition It doesn't create any additional sounds that are not already in the bowed string input.

A good violin is a low pass filter which filters out high frequencies with a fall-off starting at about 3000 Hz.  This reduces the amplitudes of the highest harmonics (partials) of many notes.  A note with many adjacent high number harmonics (7:8,  8:9, 9:10 etc.) having high amplitudes causes  a "harsh", "raspy", "buzzy" dissonant unpleasant sound.

But this unpleasant dissonant sound isn't "noise" because it has an exactly repetitive wave form.  Noise on the other hand is a combination of widely random frequencies.  A bowed string produces some noise because many bow hairs are used which slip off of the string at slightly different times so the pure harmonic series (1f, 2f, 3f...nf) has some random noise scattered between them. 

An example of noise is shown below fft of a bowed open E on one of my violins.  This violin might be considered harsh sounding because its frequency response curve's fall-off is too high so the 8th, 9th and 10th harmonics are too strong and produce a dissonant sound--not because there's some noise between some harmonics. 

 

[David Beard]

20 minutes ago, Marty Kasprzyk said:

I disagree.   All the sound comes from the strings.

The violin body makes the input string bowing sounds more audible and it also acts as a filter which makes some output sounds relatively weaker or louder. If it is in good condition It doesn't create any additional sounds that are not already in the bowed string input.

A good violin is a low pass filter which filters out high frequencies with a fall-off starting at about 3000 Hz.  This reduces the amplitudes of the highest harmonics (partials) of many notes.  A note with many adjacent high number harmonics (7:8,  8:9, 9:10 etc.) having high amplitudes causes  a "harsh", "raspy", "buzzy" dissonant unpleasant sound.

But this unpleasant dissonant sound isn't "noise" because it has an exactly repetitive wave form.  Noise on the other hand is a combination of widely random frequencies.  A bowed string produces some noise because many bow hairs are used which slip off of the string at slightly different times so the pure harmonic series (1f, 2f, 3f...nf) has some random noise scattered between them. 

An example of noise is shown below fft of a bowed open E on one of my violins.  This violin might be considered harsh sounding because its frequency response curve's fall-off is too high so the 8th, 9th and 10th harmonics are too strong and produce a dissonant sound--not because there's some noise between some harmonics. 

 

I don't claim to understand the 'how' in this situation.    (In the quote, I believe Andrea was being sarcastic, treasing me about asserting there can be 'junk' components to the sound.)

But, there is a 'what' to be observed.  Certainly, a bad student player can draw a bow poorly and produce a sound which is a messy combo of musical intent, and unwanted gunk.  Similarly, it at least seems as if a bad instrument can stand between a player and the attempt to draw an 'all content' sound.

Certainly, an over stiff instrument can give the impression of introducing lots of 'non-content' noise.  

[Michael Darnton]

@Marty Kasprzyk Your definition of "noise" seems highly personal, not the usual techical definition from acoustics, which, as I look at various sources, is any undesirable sound component, regardless of source or nature.