Subject Violin Acoustics Date Mon Oct 7 2002 15:46 Author Werner Thurnheer (towerlilt@swissonline.ch)

Dear Lars I'm trying out the WinMLS for the violin maker purposes. I checked the frequency resolution of resonance peaks to be at 2,9 Hz. For tunig I realised as yet the resonance indication by brocken tea leavs bouncing at resonance peaks with an accuracy around +/-0.5Hz or less. How to arrive there with WinMLS? I read the hints of pink noise generators and the higher sampling frequency about 200 kS/s. How WinLMS can split the actuator frequency range like pink noise or narrower? How WinLMS can generate a higher sampling rate and support the higher bit resolution? How WinLMS can run a swept sine wave? How WinMLS can generate an impulse to measure the resonance decay in time? What do you recommend to complete WinLMS? What's the price? What does the Violon Maker toolbox? Thank for your answer Best, Werner

Subject violin acoustics Date Tue Oct 8 2002 14:07 Author Lars Morset (morset@winmls.com)

Dear Werner, thank you for your questions. I will try to answer them. >I'm trying out the WinMLS for the violin maker purposes. I >checked the frequency resolution of resonance peaks to be at >2,9 Hz. For tunig I realised as yet the resonance indication >by brocken tea leavs bouncing at resonance peaks with an >accuracy around +/-0.5Hz or less. How to arrive there with >WinMLS? To get a frequency resolution of 1 Hz, you need 1 second of the signal in time domain. If the sampling frequency is 48000 kHz, this is 48000 samples. This can be set many ways in WinMLS, the easiest way is by going to the Measurement Settings Toolbar (found from the View menu) and make sure the sequence order is set to 16. The number to the right of the sequence order should then say 1.365 seconds. If the time window is set to include the whole time signal, you should get a frequency resolution >1 Hz. The frequency resolution can be set as post-processing in WinMLS using the time data window (true frequency resolution), this is done from the time window toolbar. Another way is by setting the length of the FFT larger than the time window, this is done from Plot->General Frequency Domain Settings... Select the 'Use fixed size' radio button. >I read the hints of pink noise generators and the higher >sampling frequency about 200 kS/s. I am curious on why you would need a higher sampling rate? The sampling rate will not influence the frequency resolution if using WinMLS. >How WinLMS can split the actuator frequency range like pink >noise or narrower? You can make your own pre-filters. I will strongly recommend NOT to use random noise (like pink noise). The modern methods (MLS and sine sweeps) are much better. If you need a limited frequency range I would use a sine sweep, in the beta version of WinMLS you can define the start and end frequency. >How WinLMS can generate a higher sampling rate and support >the higher bit resolution? This is sound card dependent. WinMLS can measure up to a sampling rate of 215 kHz and 24 bits using the proper hardware. >How WinLMS can run a swept sine wave? To do this you need to download the beta version. If you have not already download the release version, you may do it from: http://64.41.85.4/downloadWinMLS/WinMLS2000/WinMLS2000_Release3Build010c.exe Install it by double-clicking on the file. Download the beta version from: http://64.41.85.4/downloadWinMLS/WinMLS2000Update/WinMLSbeta4.exe Strongly recommended: If you do not want to overwrite the current version of WinMLS, you should rename it from ‘WinMLS.exe’ to e.g. ‘WinMLSold.exe’ Double-click on the beta version file to unzip it and send us an e-mail with the new serial number. If you are going to try the sine sweep method, make sure that you select "Sine Sweep" (change it from "MLS") in the Measurement Selection Toolbar. Then go to Measurement->Swept Sine Settings... and increase the sweep duration to e.g. 11 seconds. At the very bottom you set the length of the impulse response, this should normaly be set to less than 2 seconds (change it from 4 which is the default). >How WinMLS can generate an impulse to measure the resonance >decay in time? The impulse response is "always" the result using WinMLS (This is true if you use WinMLS to measure systems, but not true if you use WinMLS to analyze signals. In your case you want to analyze the violin, which is a system not a signal). You do not need to measure the impulse response using an impulse, it can be done more accurate using MLS or sine sweeps. This is fundamental, but seems to be the hardest thing for us to explain to non-experts. From the impulse response you can always find the frequency response (using FFT). You can plot waterfall, step response, energy-time-curve, Schroeder curve (which gives the decay range), etc... These are all plot types in WinMLS, and can be selected from the Plot Toolbar. The advantage by using MLS or a sine sweep is that the signal-to-noise ratio is much better than with using an impulse. For your violin measurement I recommend that you use the magnet-coil system we have developed. That will give What do you recommend to complete WinLMS? For your violin measurements I recommend our magnet-coil system to excite the violin. Do you want to excite at the violin bridge? Where do you want to measure the response, on the violin body/bridge or the resulting sound pressure using a microphone? I hope this helped. Best, Lars

Subject reliability of impulse response Date Thu Nov 21 2002 14:25 Author Werner Thurnheer (towerlilt@swissonline.ch)

Dear Lars My triggered impulse resonse mesaurements are not reliable. Although the repeated frequency response curves of a small bar are somewhat caracteristic, the peaks shifts in frequency from measurement to measurement. Is it wrong to hang up the bars for measurement? Is the impulse (given by a 0.5 g wood ball as pendulum) to strong or not enough precise placed? – (Bevor learning to know WinMLS I could not finish an earlier experiment with small stripes of wood because I could not mesure the amplitude enough accuracily.) The mesured bars are 280x30x3mm, 270x30x9mm, 227x30x19mm, 185x19x15mm of spruce, maple and rosewood. Is there a proved measurement arrangement? Best regards, Werner

Subject Repeatable measurements? Date Wed Dec 18 2002 02:16 Author Lars Morset (morset@winmls.com)

Dear Werner, interesting question, thanks! First of all, the impulse response in itself is reliable if the system you are measuring (in this case a wood bar) is 1. Time invariant - not changing as a function of time. 2. Linear - if you double the amplitude of your input signal the output amplitude is also doubled and superposition is valid. A wooden bar will in this case be according to these requirements and therefore the impulse response is valid. The reason you do not get repeatable measurements is probably because when you hit it, you are not able to reproduce the hit (e.g. the direction). Please see a table that gives the repeatability on page 2 of the article ISMA2001_MorsetArticleFinal.doc found at http://www.nvo.com/winmls/nss-folder/articles Reproducability should be discussed in the article above, please let me know if you have further questions. For your setup, hanging the bar, I would recommend that you use our magnet-coil system to get a reliable excitation. The way the bar is suspended is also important, you can experiment by changing the suspension and see what effect this has.

Subject Equipment for Violin acoustics Date Mon Jan 13 2003 11:09 Author Werner Thurnheer (towerlilt@swissonline.ch)

Dear Lars Thanks for the extended opportunity to evaluate WinMLS beta, which expired last month. I want to work with the newest Beta Version of WinMLS. Not being sure to have installed all properly I would reinstall the WinMLS Personal Latest Beta version. Are there programm reductions in front of the Professional version? For exiting the violin-family instrument at the bridge (measuring imput admittance)you recommended your magnet-coil system. For measuring with 'line in'-connection you recommended your preamplifier and microphon. Concerning this three tools can you give some informations about the technical features, the price and the ordering? Best, Werner

Subject Glissandi-excitation Date Fri Feb 28 2003 15:43 Author Anders Buen (buen@bs-akustikk.no)

Hi Lars Henrik! In an article in CASJ you campare impulse, glissandi, and MLS excitation on violins. How do you use WINMLS2000 for measurements of glissandi? Is it by just extending the calculation window over the glissando run and calculate? The method should be usable for analyzing sound fraom great instruments in recordings. What is your opinion about that? Yesterday we were able to measure impulse responses on two Guarnerius DelGesu instruments in Olso Konserthus. Several interested makers there.. Till soon, Anders

Subject Correction for impulse excitation Date Fri Mar 14 2003 10:11 Author Anders Buen (buen@bs-akustikk.no)

I use my finger for excitation of the violins I measure. The fingernail gives some highfrequency resonse, but most of the energy in the impulse is probably in the low frequency region. That gives an underestiamtion of the high frequency response of the violins. Do you have an idea how to compensate for the frequency dependent excitaion signal in a simple way? Would it improve the frequency response to use an compensation from measuring impulses directly by excitating the microphone several times and use the average frequency response of that signal to correct the measured responses of the violins? Any comments? Kind regards Anders Buen http://home.online.no/~an-buen

Click here to return to discussion list