In this section an example analysis presented on one sound recording of experimental tuning slot pipe #2, introduced in Chapter 6. The setting with slot widthws = 12 mmand slot height hs= 35 mmis chosen here. The recording was performed using two calibrated condenser micro-phones, as it was described in Section 6.2.2. As a preprocessing step, the single-channel record-ings of the calibration signal and the pipe sounds are merged into two-channel files using theCut editortool ofSoundAnalysis, with Channels #1 and #2 representing the microphones located near the mouth and the tuning slot, respectively.
Figure A.1 displays the main window of the software tool. In the top part of the window, in theWave filesandWave cutspanels, recordings can be imported and cut into segments. When a calibration signal is loaded, the amplitudes can be detected in theCalibrationpanel. Once the calibration was successful, the greenCalibratedsignal indicates that subsequent analyses are au-tomatically performed in calibrated mode. In the panel calledFrequency detection, the algorithm for fundamental frequency detection can be customized. The parameters of the analysis are ad-justable in theAnalysis setuppanel. Different settings are stored for stationary, attack and decay analyses. When the analysis is completed, the graphs are shown in a new window, where the user can adjust the properties of the display and store the results. Finally, in theAnalysis results panel, old results can be reopened and compared using various types of plots.
The steady state analysis results are shown in Figure A.2. By plotting the two spectra on top of each other the differences of the levels and the envelopes can immediately be observed. Follow-ing local baseline maxima in the spectra, the frequencies of natural resonanace of the resonator can also be detected. In the frequency axis, normalized units are chosen, which can easily be set
A.3. AN EXAMPLE ANALYSIS 131
Figure A.1.The main window ofSoundAnalysis v2.0
Figure A.2.Stationary analysis results
(a) Partials view (b) Spectrogram view
Figure A.3.Attack analysis results
in the tool, facilitating the comparison of spectra with different fundamental frequencies. The effective loudness values are also automatically calculated, and it can be seen that the measured loudness is about5 dBgreater at the mouth of the pipe than at the tuning slot.
Analysis results of the attack transient are shown in Figure A.3. Data from the mouth mi-crophone (Channel #1) are displayed only; however, the characteristics of the transient are quite similar for both channels. Thanks to the coherent analysis achieved by resampling the signal, the number of FFT window points can be reduced without losing the accuracy of the computed am-plitudes. Thus, the time resolution of the analysis can be increased and a detailed picture on the attack can be obtained. The time scale in both plots of Figure A.3 is given in a non-dimensional scale, the unit being the duration of one period of the steady state sound.
In Figure A.3(a) the time history of the amplitudes of the first five partials is depicted. As it was observed in Figure A.2, the fundamental is the strongest component in the steady state in both channels. Nevertheless, as it is seen in Figure A.3(a), in the attack phase the second and third partials are both stronger than the fundamental at times100 < t < 150. The steady state sound builds up after a hundred periods and from that point the amplitudes of the partials do not change with time.
Figure A.3(b) displays the same attack but in a spectrogram view. By a detailed examination of the spectrogram interesting effects can be found that are unseen in the amplitude history plot of the partials. Such interesting effect is observed near the third partial around the hundredth period in the transient. As it can be seen, the red “line” corresponding to the third partial breaks aroundt= 130. Taking a closer look at the spectrogram, it is also found that the first part of the line (t <130) has a slightly lower frequency than the second part (t >130). This indicates that the third eigenfrequency (that is slightly lower than three times the fundamental) is also excited in the attack transient; however, in the steady state the first eigenfrequency becomes dominant.
The possibility of the detailed analysis of the attack transients of labial pipes is a useful tool for the organ building community, since the properties of the attack greatly affect the percieved sound quality of the pipes. By usingSoundAnalysisorgan builders can assess the effect of various scaling, voicing, and tuning adjustments in an objective manner, supplementing their subjective observations.
Appendix B
INNOScale scaling software
In this appendix the software tool calledINNOScale1is introduced, which was developed dur-ing theINNOSOUND project. The program supports the computer aided design of flue pipe ranks using traditional and novel scaling methods. More details about the software code are found in the technical report [T3] and the conference paper [C10].
B.1 Introduction
In the last two decades, the traditional method of organ pipe scaling has been supplemented by computer aided methods that facilitate the calculation of the main parameters of the pipes. One example known to the author is the tool calledM!by Laukhuff Orgelbau.2 INNOScaleaims at providing a user friendly interface for performing the tasks of the traditional scaling procedure, as well as making innovative methods accessible to the organ building community. The main features of the tool are the following.
• Support of customisable scale definition points and graphical editing of scaling lines: The user can choose the density of the scaling points and manipulate the lines while the soft-ware automatically interpolates the scale for middle points.
• Support of various types of stops: The most often used pipe forms are handled by the tool together with several special configurations, such asaliquots,transitionsormixtureregisters.
• Support of innovative scaling functions: Beside the traditional methodology,INNOScale incorporates novel scaling procedures developed in the frame of theINNOSOUNDproject.
• Compatibility withMicrosoft Excel: All scaling tables and other data can be exported to Excel format for further editing, printing etc.