Jump to content
Maestronet Forums

A new round ... Dünnwald parameters


Lars Silen

Recommended Posts

When adjusting a violin one big problem is the mass if data if one tries to do some "objective" measurements. Look at any reasonable spectrum done at high resolution. You will see a "wood of frequency peaks". How to interpret those peaks is another matter :) .

 

Anders Buen has written some articles about the Dünnwald parameters where three numbers are calculated and those numbers are said to provide some kind of classification of classic Cremonensic instruments (put in one bin) and modern instruments that will go into another fairly distinct bin. Notice that nothing is said about "quality" :) .

 

The idea of this thread is: Has somebody tried to use the Dünnwald parameters as tools for instrument optimization? My experience is that it helps a lot to look at only three numbers instead of tens of peaks. It is also much easier to see when an adjustment will move the instrument being worked on towards some easy to interpret "goal".

 

What have I done:

 

I record a G-major scale played from low G upwards to B on the E-string. Much vibrato is used because Dünnwalds parameters are based on recordings of real music.

 

Next I generate a spectrum (4096 points) and export the result as a plain text file that can be processed further.

 

I run a small program, written in Python, that calculates the Dünnwald parameters for the adjustment round (see below).

 

Based on the parameters it is easy to see how the last modification influenced the instrument. It also feels like the changes go in the correct direction and when the parameters improve the sound of the instrument also improves.

 

An example:

 

I recently started a round of optimization on my own Harding fiddle that I use myself. The initial state was:

 

./Dunn_A.py 00_hard_initial.txt

File to process:  00_hard_initial.txt
Dunnwald parameters for :00_hard_initial.txt
A = 57.9133410465
B = 54.39818555
C = 50.5321468125
D = 48.6583449425
E = 45.1510729073
F = 37.1172910121
L[Db] = -14.525938
ACD - B = -2.91313386481
DE - F = 9.31585192488

 

After seventeen steps of internal sanding the sound had improved very much and so had the Dünnwald parameters:

 

./Dunn_A.py 17_har_btn_uppe_esidan_balans.txt

File to process:  17_har_btn_uppe_esidan_balans.txt
Dunnwald parameters for :17_har_btn_uppe_esidan_balans.txt
A = 57.2790034651
B = 55.5393324333
C = 52.5985449688
D = 51.3234882299
E = 49.7805976623
F = 39.7953643065
L[Db] = -3.600813
ACD - B = -2.3831962358
DE - F = 10.5492311683

 

The final situation can be compared to for example the Stradivarius ”Hellier” från 1679 :

L(dB)     =  -6,0

ACD-B  =  -1,2

DE-F      =  10,3

 

Generally higher values within reason are thought to be better.

 

My original article about adjusting the Harding fiddle (in Swedish) is found here http://larsil2009.wordpress.com/2015/03/19/dunnwaldparametrar-som-hjalp-vid-fioltrimning/

 

Anders Buens article used as the base for the experiment is found here: http://www.google.fi/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CCQQFjAA&url=http%3A%2F%2Fwww.akutek.info%2FPapers%2FAB_Timbre_Parameters.pdf&ei=A7oKVcCUBYzqaKfdgcgE&usg=AFQjCNF4h1UWWcqbdVWNCUE0PxQR-twRsw&bvm=bv.88528373,d.d2s

 

This is how the amplitudes of the frequency ranges used to calculate the Dnwald parameters changed during the optimization process. Notice that the range seems to be roughly 10 dB.

 

dunnwald_raw_data.png?w=450&h=348

 

Comments or ideas?

Link to comment
Share on other sites

My guess is that the most difficult thing is to decide when to stop. There are many possible obstacles:

  • Make the instrument too shrill/sharp --> hurts the ears
  • Make the tone too dark ... not nice.
  • etc.

The ultimate limit is of course when one sand through the plates --> write off.

 

There is another problem: The measured values tend to change during the next days. The change may several dB. This means that one should take small steps and wait/play/listen to get a feeling for if the instrument is ok at the moment or if another improvement step should be taken. The change can be extremely fast.

 

I got my #2 Guarneri back for adjustments (owned by my brother). The sound was nice but there was a slight feeling of "cotton wrapping" to the sound. The measured Dünnwald parameters for this instrument were initially:

 

./Dunn_A.py johan_röd_initial.txt

File to process:  johan_röd_initial.txt
Dunnwald parameters for :johan_röd_initial.txt
A = 56.6736285814
B = 54.1317280833
C = 49.4705171875
D = 49.6919481724
E = 43.3954089735
F = 34.4657867177
L[Db] = -17.400888
ACD - B = -2.63035728086
DE - F = 11.2312983495

 

Especially the L-parameter was very low and it could be heared as "cotton" to the sound and sime very slight "sandiness". Not any really big problem but audible.

After one adjustment step where I sanded the G-side f-hole's edge towards the bridge (total 160 sandings 1.6/100 mm) the next measurement gave the result:

 

./Dunn_A.py johan_röd_+40_med_vibrato.txt

File to process:  johan_röd_+40_med_vibrato.txt
Dunnwald parameters for :johan_röd_+40_med_vibrato.txt
A = 60.8953518372
B = 59.38526285
C = 56.0765789375
D = 57.1677487701
E = 51.0858742848
F = 40.1555862419
L[Db] = -6.858046
ACD - B = -1.44362829444
DE - F = 13.1534942623

 

The cotton was gone completely and the sound was very nice. I told my brother to bring the instrument after say 6 months because I didn't want to unintentionally destroy a state of the instrument that was really nice! After the change the instrument fits the group if Cremonense instruments perfectly from the point of view of the Dünnwald parameters.

Link to comment
Share on other sites

The technical set up is as follows:

 

A Zoom R8 machine is used for the recordings.

A Rode NT1 large membrane mic is used (latest model, very high quality)

The distance from the instrument to the mic is roughly 4 m to get rid of the near field.

Audacity is used to generate the spectrum.

A custom python program is used to calculate the Dünnwald parameters. The program was initially written by my son Sebastian Silen. I did some additional improvements later. The values A ... F are automatically written to a file. It is thus extremely easy to plot the trends.

I use the Grace (xmgrace) program for plotting.

Link to comment
Share on other sites

Intersting. Have you tried computing the parameters for a variety of playing conditions, like no vibrato, consistent narrow/fast and then wide/slow and see how that affects the parameters? I would worry a lot about inconsistencies in the playing affecting the paramaters to the same order of magnitude as changing the plate.

Link to comment
Share on other sites

The idea of this thread is: Has somebody tried to use the Dünnwald parameters as tools for instrument optimization? My experience is that it helps a lot to look at only three numbers instead of tens of peaks. It is also much easier to see when an adjustment will move the instrument being worked on towards some easy to interpret "goal".

 

Yes, I did "optimize" a fiddle vaguely along the line of Dunnwald parameters and entered it in the last VMAAI competition (tone only).  I used a pair of strong magnets, and experimented with various positions until I knocked down the peaks in the "transition hill" (or Dunnwald B range).  My best guess at the resulting parameters:

 

L(dB) = +1, or +26 if you add the 25dB to the difference (A0 was 1dB higher than the highest B peak)

ACD-B = +3 

DE-F = +9

 

The numbers are really, really good.  It even scored reasonably highly with the listening judges.  It scored near the bottom with the playing judge, and, having played it myself, I definitely agree.

 

One of the major problems with the Dunnwald calculations is that you can have a boomy, tubby fiddle that has very good numbers, which points to the underlying basic problem:

 

The difference between good and bad is often spread out across a number of different things.  As soon as you boil down the variables into three numbers, you have cooked out most of the flavor and end up with mush.  It's probably better than nothing, if you don't play and hear reasonably well and/or know someone like that, but it's very limiting, I think.

Link to comment
Share on other sites

Intersting. Have you tried computing the parameters for a variety of playing conditions, like no vibrato, consistent narrow/fast and then wide/slow and see how that affects the parameters? I would worry a lot about inconsistencies in the playing affecting the paramaters to the same order of magnitude as changing the plate.

We have done some experiments. Generally the L-parameter is improved slightly when vibrato is used. We have also done some reference runs in different environments but still too few to say anything definite.

 

I have also discussed the possibility of taking any reasonably good instrument and put it into the hand of a professional allowing one to try to tweak ones own tone to force it closer to some set "goal". The idea is to learn to play the instrument in such a way that it sounds as close to some reference instrument as possible.

Link to comment
Share on other sites

Lars,

 

I think these kind of experiments are only valid for comparing violins in the same environment and recording setup.

This is true to some extent. I know for sure that my room where I work most of the time tends to produce instruments that are close to being too dark. Much work is then needed to brighten the instrument afterwards. On the other hand it feels like an uesful additional tool in the toolbox because the parameters are easy and very fast to calculate.

Link to comment
Share on other sites

Yes, I did "optimize" a fiddle vaguely along the line of Dunnwald parameters and entered it in the last VMAAI competition (tone only).  I used a pair of strong magnets, and experimented with various positions until I knocked down the peaks in the "transition hill" (or Dunnwald B range).  My best guess at the resulting parameters:

 

L(dB) = +1, or +26 if you add the 25dB to the difference (A0 was 1dB higher than the highest B peak)

ACD-B = +3 

DE-F = +9

 

The numbers are really, really good.  It even scored reasonably highly with the listening judges.  It scored near the bottom with the playing judge, and, having played it myself, I definitely agree.

 

One of the major problems with the Dunnwald calculations is that you can have a boomy, tubby fiddle that has very good numbers, which points to the underlying basic problem:

 

The difference between good and bad is often spread out across a number of different things.  As soon as you boil down the variables into three numbers, you have cooked out most of the flavor and end up with mush.  It's probably better than nothing, if you don't play and hear reasonably well and/or know someone like that, but it's very limiting, I think.

I think one should be very careful not to cheat when using the parameters ;) . It is fairly easy to improve the L-parameter by making too "narrow" changes specifically to fix a single peak.

 

Regarding the problems. Yes it is obvious that a full description of a good violin isn't possible with only three numbers ... it would be too easy. On the other hand I think that used with caution they can be useful during the trimming stage.

 

It is actually interesting to notice that with some experience it seems possible to look at the raw recording file and quite well guess if the parameters will turn out ok simply from the amplitudes of the different tones in the scale.

 

I obviously look at all the curves because they are all available without any effort.

Link to comment
Share on other sites

I have been hoping that Don would come up with his own improvement to Dünnwald parameters. 

 

Sorry, Mike, but that looks like a dead end to me.

 

I have been looking at response spectra for just about 10 years now, and although it's fairly easy to spot a truly horrible-sounding violin from the spectrum, I don't think it's so clear what's going on when you get to decent instruments.  

 

I rely on playing and listening, with the response plots for record-keeping, confirmation, and analysis.  A violin may have some annoying tonal issues that can't be seen on the plots, and there are often ugly-looking peaks and dips on the plots that are not an issue.. or might be; you have to play and hear it.  Boiling things down to a number or three I don't think is going to get anywhere.

 

And besides, everyone has different ideas about what's good, anyway.  One of the most difficult acoustic tasks has been (and continues to be) trying to "get" what good violinists say about what they look for in sound and playability.  It's all a balance, and while higher Dunnwald numbers might look good, it might mean that there's too much of a good thing... which is bad.

Link to comment
Share on other sites

Sorry, Mike, but that looks like a dead end to me.

 

I have been looking at response spectra for just about 10 years now, and although it's fairly easy to spot a truly horrible-sounding violin from the spectrum, I don't think it's so clear what's going on when you get to decent instruments.  

 

I rely on playing and listening, with the response plots for record-keeping, confirmation, and analysis.  A violin may have some annoying tonal issues that can't be seen on the plots, and there are often ugly-looking peaks and dips on the plots that are not an issue.. or might be; you have to play and hear it.  Boiling things down to a number or three I don't think is going to get anywhere.

 

And besides, everyone has different ideas about what's good, anyway.  One of the most difficult acoustic tasks has been (and continues to be) trying to "get" what good violinists say about what they look for in sound and playability.  It's all a balance, and while higher Dunnwald numbers might look good, it might mean that there's too much of a good thing... which is bad.

You are right. Too bad.  :)

Link to comment
Share on other sites

Sorry, Mike, but that looks like a dead end to me.

 

I have been looking at response spectra for just about 10 years now, and although it's fairly easy to spot a truly horrible-sounding violin from the spectrum, I don't think it's so clear what's going on when you get to decent instruments.  

 

I rely on playing and listening, with the response plots for record-keeping, confirmation, and analysis.  A violin may have some annoying tonal issues that can't be seen on the plots, and there are often ugly-looking peaks and dips on the plots that are not an issue.. or might be; you have to play and hear it.  Boiling things down to a number or three I don't think is going to get anywhere.

 

And besides, everyone has different ideas about what's good, anyway.  One of the most difficult acoustic tasks has been (and continues to be) trying to "get" what good violinists say about what they look for in sound and playability.  It's all a balance, and while higher Dunnwald numbers might look good, it might mean that there's too much of a good thing... which is bad.

This is something I buy fully :) . Playing the instrument and trying to learn what options there are to change the sound is the key. The problem is that the goal, just as Don said, is something fairly subjective. One player likes a fairly bright instrument and another wants it darker. Which one is better?

 

My experience is that one has a fairly large set of options to "improve" an instrument after it is "finished". The changes will influence also basic resonances meaning that creating a good instrument isn't a one shot try but a long process of step wise improvements. The improvements can continue long after the instrument has found a new home but in that case the changes have to be fairly small and negotiated based on the owner's taste and expectations. 

Link to comment
Share on other sites

Lars,

 

I think these kind of experiments are only valid for comparing violins in the same environment and recording setup.

I have been looking into the problem of separating the instrument acoustics from the room acoustics together with my son Sebastian. There are some interesting tools used in AI and in space research that possibly could be used for this purpose. Still too early to tell anything more.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.


×
×
  • Create New...