8 violin clips for tone evaluation

[Don Noon]

I think the main thing to be learned from this is how misleading it can be to go by recordings.  These violins do sound vastly different under the ear and in person.

 

Last night I had a warm-up session with my guitarist, and brought along the firewood fiddle (#3) and my "good" violin (#4).  We both agreed that there was no comparison... the firewood fiddle sounded thin on the low strings, and tonally a bit nasal or unpleasant.  Even before I plotted the response (below), I was fairly certain that there was weakness in the low frequencies and excessive strength in the "transition hill".  The transition hill excess shows up to me as a very loud E string in the lower positions, with some notes feeling like a nail in the forehead... loud, but not refined or adding to clarity.

 

This is a comparison of bowed semitone scale response, low resolution.  It is interesting to note that the firewood fiddle's B1+ frequency is a lot lower than the "good" fiddle, yet the low-end response is weaker.

[La Folia]

They look rather similar to me, and they sound rather similar.  One difference I can hear is in the bass notes.  But from the graph, it appears that the sound sample completely misses the problem region.  The reason it isn't audible is because you didn't play it.

 

--although I wonder about the straight lines of the graph, which appear conceal a paucity of data points?  What am I missing here?

 

If the problem is essentially just a couple of notes and a bit of a nasal tone, shouldn't that be easily solved?  Maybe with a more flexible bridge or something?  And the top looked extremely crudely made on the underside.  What would happen if you just finished the job?  I have to wonder if it would sound just like your other violins.

[Don Noon]

A problem with Audacity is that it uses fixed frequency intervals in plotting, so with low resolution, the lower frequency resolution gets really low.  I'd really like it to plot fixed note intervals, but it is what it is.

 

The entire G and D string are relatively thin, in addition to the low E-string excessive power.  These are not things that can be changed significantly by external adjustments, in my experience.  The low end could be beefed up by thinning the top some more, which I may do eventually.  I have tried many times to make targeted changes to the transition hill peaks, with no real notable success that didn't involve leaving the instrument worse off overall.

 

I also don't believe that the crudeness of the workmanship has much to do with the tone.  I have had similarly poorly made tops on this VSO, some were quite good sounding, and they all were different in fairly major ways.  

[La Folia]

Well, you're the expert, but just so we remember, ready to install:  http://www.maestronet.com/forum/index.php?/topic/334867-firewood-fiddle-a-test-of-the-worst/?p=727665

 

[Salve Håkedal]

..

I also don't believe that the crudeness of the workmanship has much to do with the tone..

I've learned a lot from you through this site, Don, but I have to say I'm not with you on this.

(I wonder where I would have been if that statement was true!)

[Bill Merkel]

A problem with Audacity is that it uses fixed frequency intervals in plotting,   

I noticed Adobe Audition is available as a cloud-based app for $19 a month.  Audacity was a low rent clone of it.  Pro Tools is an industry standard too.

 

Is there a low pass filter in Audacity that you have tuned down?  That would allow (perhaps cause) the effective sampling rate to decrease which would in turn give you big divisions in the plot.  I used to use Audition a lot but not for frequency analysis much, but I do seem to remember that the output divisions there on the screen in that mode were a function of the sampling rate the way you'd expect.  I don't remember the exact mathematical relationship any more, but the output divisions in this transformation get smaller linearly as the sampling rate increases.  There's no upper limit on sampling rate and the lower limit is two times one over the Nyquist frequency of your low pass filter or something  For these purposes the slope of the filter isn't important as long as it's well down at the Nyquist frequency.  As best as I can remember whatever above that that gets through will show up as evenly distributed noise that would only bias the whole plot upwards.  If you're just plugging a mic into your computer this is all taken care of and the trick is to have a high sampling rate along with software that takes advantage of that and give you small divisions in the plot.

[Bill Merkel]

 

By the way, if you get a sudden craving for ice cream while listening to this, it may be because it is a tune I hear all the time blaring from the local ice cream truck (Redwing, B part).

Maybe the Secret of Stradivari is what little tunes he used to test his instruments.  Maybe some Monteverdi's greatest hits or some early version of O Sole Mio

[La Folia]

A problem with Audacity is that it uses fixed frequency intervals in plotting, so with low resolution, the lower frequency resolution gets really low.  I'd really like it to plot fixed note intervals, but it is what it is.

Obviously, you can judge whether the coarse sampling is hiding anything, but I think 4096 intervals will give a bit better resolution, even in the bass.

[Bill Merkel]

If what you want is a plot of actual notes, you could take a sheet of transparency and write note names on it in the right places and just lay it on top.  

[David Burgess]

Obviously, you can judge whether the coarse sampling is hiding anything, but I think 4096 intervals will give a bit better resolution, even in the bass.

Sometimes, a courser sampling rate will help reveal things that higher resolution (an excess of squiggly lines ) , will  make more difficult to interpret.

[Don Noon]

I also don't believe that the crudeness of the workmanship has much to do with the tone.

 

I've learned a lot from you through this site, Don, but I have to say I'm not with you on this.

(I wonder where I would have been if that statement was true!)

 

The problem with crudeness is that it is often (usually?) associated with extreme variations in things that might matter, such as arching or graduations.  I try to come fairly close to what I think works with those, but don't fret about a bit of lumpiness.  I challenge anyone to come up with some theory or test that shows a significant benefit to ultra-smooth surfacing.  At best, you could get maybe a gram or two lighter... which I don't think is acoustically a big deal.  

 

 

There is now and has always been a problem trying to show quantitatively what the differences are between instruments, so here are some other attempts:

Bowed semitone scales, 4096 resolution

 

And impact spectra:

 

They do kinda look all the same, with just a few little blips here and there of a couple of dB.

And the kinda do sound sorta the same in recordings.

 

But played in-person, both from player and listener perspectives, they are vastly different.

[Don Noon]

Just another observation, from the impact spectra shown above...

 

The firewood fiddle A0 frequency is quite high at 291 Hz, which (for a very large-bodied fiddle, which this is) means abnormally high body stiffness in general (large F-holes could cause this problem, but I didn't make them oversized).  However, the B1+ frequency is 523, which is not high at all.  From the very low measured longitudinal stiffness of the firewood, this was expected.  B1- frequency is rather high, at 449 Hz.  This also is not a surprise, as the crossgrain stiffness of the firewood measured out around normal for good wood.

 

So I think there is an indicated route to improving this fiddle:  thin the top to get more bottom end and get A0 frequency down, but add a stiffer bass bar to keep B1+ frequency from getting too much lower.  This is not just based on the measurements, but in playing, this feels a bit stiff and sounds a bit thin.  I hope I can hold off on this rework until after I get my VSA instruments done.

[Bill Merkel]

^The peak at 900-1000Hz in comparison to the other fiddle is big enough to be audible and I know that around 1000Hz on an equalizer is where honk lives.  It is also in the range of 1st position on the E string which you said was too loud.  If you know how to take that peak away, it wouldn't hurt it too much I would think.

[Marty Kasprzyk]

>>

And the kinda do sound sorta the same in recordings.

 

But played in-person, both from player and listener perspectives, they are vastly different.

Maybe we're not measuring the right things.

[actonern]

At this point my biggest question is whether you (Don) know what a good violin sounds like.

 

Not saying I do, but whether you do?

 

From your recorded and posted "opus" so far, they tend to my year toward "fast", "loud", and "dryish" in tendency...

[Bill Merkel]

 

Maybe we're not measuring the right things.

 

 

I thought about that too.  They might sound more similar live behind a screen.  Hearing is beaucoup tricky.  On the other hand it could be a lot of the similarities in the graphs is really the frequency response of a ragged microphone.

[La Folia]

All I know for sure is that the sound files sound rather similar, and the measurements made from those files are rather similar.  I don't know how or why the sound files differ from what one might hear in person.

[Don Noon]

^The peak at 900-1000Hz in comparison to the other fiddle is big enough to be audible and I know that around 1000Hz on an equalizer is where honk lives.  It is also in the range of 1st position on the E string which you said was too loud.  If you know how to take that peak away, it wouldn't hurt it too much I would think.

 

That is something I have been fighting with for years now.  Once the fiddle is built, that peak has proven to be a nearly immovable object.  It seems to be mostly a product of arching and wood properties, as far as I can tell.

 

At this point my biggest question is whether you (Don) know what a good violin sounds like.

 

To which I ask, what is "good", and how consistent are other people in judging goodness?

 

As an attempt to see how that works, I looked at the judges' scoring from the last VMAAI competition and compared their scores to mine (this was all blind testing for everyone).

I took each judge's scores, and found the R^2 correlation factor for the other judges' scores to theirs.

 

Judge 1 and 2 are in the best agreement; judge 3's scores bear essentially no relation to the others.  I'm inbetween.

Since there is no true standard of what's good or not, the only thing you can do is say that perhaps judge 1 and 2 liked the same things, and judge 3 liked something very different.  Or that there's so much randomness in this that you can't tell much of anything.

 

I like to think that I can hear what's going on fairly well... but I know that my opinion of "good" might be different from others (and why I try to get other opinions).  And I am aware that my primary goal has been to make powerful, projecting soloist violins, perhaps at the expense of some tonal refinement.  Still a work in progress.

 

All I know for sure is that the sound files sound rather similar, and the measurements made from those files are rather similar.  I don't know how or why the sound files differ from what one might hear in person.

 

Hearing is beaucoup tricky.

 

Although the plots look rather similar, there ARE differences... they just don't look like much.  But they do seem to reflect the differences that I hear in person.

 

Hearing is very tricky... and extremely sophisticated when it comes to directional detection.  I suspect that when a mono recording is made (like these test clips), you lose tons of information about where the room reflections are coming from, and it all mooshes together in a soup, and sounds mostly like the room.  In person, the ear/brain can filter out the reflections from the direct sound.

[Bill Merkel]

^Yes.  My experience during my brief stint as a recordist is that the particular mic and the instrument and the mic position and to a lesser extent the room all fuse into a single instrument in a recording and that is what you hear in the recording.  You can get calibration mics that are practically flat and for not a lot of cash but they sound awful in a recording.  That is why all normal mics have some purposeful coloration built into them.  Extremely interesting! It might be worth it to try one of those calibration mics in your experiments.  

 

One time I recorded a live performance using just a lapel mic of a model newsmen wear just laying on a table feeding a Sony minidisc recorder.  Amazingly it all turned out to be a magical combination and it is a highly thought of recording in a certain sphere...   And sometimes you can do everything "right" and it will be hard to salvage.

[go_oa]

My experience helping customers is that a small duet can help a lot in comparing  two similar instruments.

[Addie]

At this point my biggest question is whether you (Don) know what a good violin sounds like.

 

Not saying I do, but whether you do?

 

From your recorded and posted "opus" so far, they tend to my year toward "fast", "loud", and "dryish" in tendency...

Let he who has won tone awards cast the first (s)tone...

Recordings are recordings.

[Evan Smith]

Dons fiddles stink

[funnyope]

I heard promise in 4, 6, and 7. I think 4 and 7 had the most potential. 2 and 3 sounded too wimpy in high registers. 5 was similar to 4 or 6, can't remember which, but inferior. I think it is your average student fiddle.

I must admit, however, that I listened on my cell with a low volume, so let me see if I can listen again later on my computer.

Edited July 12, 2016 by Kenmore M

[Addie]

Dons fiddles stink

You can smell them over the Internet? You are a god!!!

[Don Noon]

Some of my earlier over-baked ones did smell like a forest fire.  Evan has been there to smell them in person at VMAAI competitions.  His always smell like varnish, since they were usually varnished in the back of the van (or whatever you  call that thing) an hour or two before arriving.