Attack transients of free reed pipes in comparison to striking reed pipes and diapason pipes

Transcription

1 submitted to: ACUSTICA acta acustica Vol. 8 () 7 Attack transients of free reed pipes in comparison to striking reed pipes and diapason pipes Jonas Braasch Institut für Kommunikationsakustik, Ruhr-Universität Bochum, Germany Christian Ahrens Musikwissenschaftliches Institut, Ruhr-Universität Bochum, Germany Summary Attack transients of three free reed organ stops were measured throughout the whole frequency range of the stop and compared to the measurements of a striking reed stop and a flue stop. The purpose of the measurement was to examine why free reed pipes are often judged to have a sluggish attack. As the results show, the attack transients of free reed pipes differ in a number of parameters: rise time, amplitude and initial delays of the partial tones from striking reed pipes and flue pipes. In addition free reed pipes and striking reed pipes do not start with a chiff which is typical for diapason pipes. The analysis showed that the rise time of free reed pipes is shorter than that of striking reed pipes, but often in the same order as the rise time of the diapason pipes. A psychoacoustical test was conducted and revealed that parameters other than rise time, namely different initial delays of the partials and the presence of the chiff, in the case of the diapason pipe, might lead to the perception that the attack duration of the free reed pipe is longer than that of the diapason pipe. PACS no..7.cd,.7.np. Introduction Free reed organ pipes were introduced at the end of the 8th century and soon became popular in Europe, especially in the German speaking countries. In contrast to striking reed pipes (bound or beating reed pipes), the reed in free reed pipes does not beat against the shallot, an orifice in the reed, when it is played. Instead it swings freely through a perforated oblong plate of brass, such as the one found in reed organs or accordions. Stops with free reed pipes were often named Clarinet, Cor anglais, Bassoon or Aeoline. They were appreciated for their mellow, round and agreeable sound [], whereas the sound of striking reed pipes was often considered to sound somewhat hard, rattling, clanking and having most of the time something nasal []. In the twenties of this century free reed pipes were abandoned with the beginning of the Orgelbewegung, a German movement that tried to restore the ideal of the organ sound during the age of Bach. The promoters of this movement considered free reed pipes too mellow, pappy and sluggish [,, ]. Nowadays, free reed pipes have become more popular again and are sometimes considered when building a new organ, for example in the Bürgersaal zu München (Vleugels, 99) or in the Augustinerkirche (Klais 99). The judgements on free reed pipes, described above, have been made on the subjective, perceptive acoustical impression only. A listening test has not been yet conducted to prove or deny this impression. The aim of this study was to determine the role of physical acoustical properties of free reed pipes that lead to these perceptive judgements on the attack transients. Received 9 May 999, accepted June. Until now only a few acoustical measurements have been made on free reed instruments. One historical investigation has been made on free reed pipes by Weber [?] and Töpfer []. All the investigations that have been made recently are measurements of other free reed instruments or of the reeds of these instruments. In these studies the steady sound properties of free reed pipes, namely the measurement of the sound spectrum, the influence of the blowing pressure on acoustical parameters, especially pitch and the motion of free reeds, were investigated on the reed organ [7, 8], the harmonica [9, ] and the accordion []. The influence of the tube length on the sound was measured on the Asian mouth organ khaen []. Theoretical approaches were made to explain the behavior of free reed pipes. Hilaire et al. measured the aerodynamic excitation of two different free reeds, showing that the amplitude of vibration grows exponentially []. In this investigation an analytical model was accomplished to estimate the pressure which excite the reed motion. Cottingham et al. modeled the variation of frequency with blowing pressure for an American organ reed, based on the linear model of Fletcher for reed instruments [, ]. The free reed pipe measurements in this study were made on the Klarinette 8 stop of the Klais organ built in 998 in the Auditorium maximum of the Ruhr- Universität Bochum, as well as on the Clarinette 8 and Cor anglais 8 stops of the historical Walcker organ in Essen-Werden, built in 9. For comparison purposes the sound of two other stops of the Klais organ, one with striking reed pipes (Krummhorn 8 ), the other one with flue pipes (Principal 8 ) and the sound of an orchestra clarinet were measured. Besides the classical Fourier spectra, log-mag amplitude vs. time diagrams as well as its derivative in time

2 submitted to: ACUSTICA acta acustica Vol. 8 () Braasch, Ahrens: Attack transients of free reed pipes were chosen to represent the data. While log-mag amplitude vs. time diagrams are commonly used in psychophysical research, its derivative curves have been rarely taken into consideration, namely by Pollard and Jansson [] and Pollard [7]. Pollard and Jansson simply assumed the importance of the slope (derivative) of a sound level, due to the strong connection between dominant slopes and dominant levels. Meanwhile Heil and Irvine have found cortical neurons with a firing rate that is mainly dependent on the dynamic stimulus properties, such as rate of change of peak pressure [8].. Measurement procedure The sounds of the organ pipes were recorded directly on the Klais organ or on the Walcker organ. The tones c and f ] were measured throughout the whole range of each stop. The Microphone (AKG, C-) was placed cm above the outlet of the resonant tube of the pipe. To keep the microphone out of the direct air vent it was placed approximately 8 cm beside the axis of the pipe body. The sound was preamplified (Mackie, -VLZ) and recorded with a DAT-recorder (Sony, TCD-D7). For the measurement on the Klais organ it was possible to synchronize the sound recording with the state of the specific key ( on or off ) of the electric manual. For that purpose the output of the electric panel cabinet V if key is pressed, V else was used to trigger a function generator (Hewlett-Packard, A) that generated a continuous khz sine waveform. The amplitude of the sine wave increased by about a factor of two, when the key was pressed. To achieve synchronization the triggered sine sound of the function generator was recorded simultaneously with the recording of the organ sound on the second channel of the DAT-recorder. The state of the key was not recorded in the case of the Walcker organ. The recordings of the orchestra clarinet were realized in the anechoic chamber of the Institut für Kommunikationsakustik at the Ruhr-Universität Bochum using the same recording setup as stated above. In this case the microphone was placed m in front of the bell of the instrument, cm above it. Different single tones throughout the whole range of the instrument were recorded twice, one time the experienced musician was instructed to play the sound with a soft attack, the second time to play it with a hard attack. In all the plots the absolute level of the sounds were not taken into consideration. Instead the maxima of the plots were scaled to the same value. The recordings were cut and analyzed digitally on a PC using Matlab.. The fast discrete Fourier algorithm (DFT) was applied to estimate the frequency spectrum of the steady state sounds. For the log-mag amplitude vs. time plots the complete signal was first down-sampled to a new sampling frequency, the fundamental frequency of the sound multiplied by the factor 8. Next it was filtered with a bandpass filter bank (Hamming windowed linear-phase FIR- Filters, 8 coefficients, filter width / of the fundamental frequency, each filter was centered on the frequency of the harmonic). The envelopes of the single partials x i (k) are calculated, using the Hilbert transformation y i (k)=ht[x i (k)]. Next, the logarithm of the curve is calculated and multiplied by the factor (a(k) = p log ( x (k) +y (k))). To estimate the derivative curve a polynom of the order is fitted to the log-mag amplitude curve, using the least squares curve fitting algorithm, and derived in time. The attack time of the partials was calculated as the time interval when it reached the ; threshold of the maximum log-mag amplitude up to reaching the ; threshold of the maximum log-mag amplitude. This definition was favored above the commonly used % 9% linear amplitude threshold interval, because it considers the huge dynamic range of human hearing.. Results and Discussion.. Fourier spectra In Fig. the Fourier spectra of different organ pipes (all tone F ] ) are shown. Characteristic for the free reed pipes: Clarinette 8 (Walcker, Fig. a), Cor anglais 8 (Walcker, Fig. b), Klarinette 8 (Klais, Fig. a) is the almost linear decline of the maximum sound pressure level with the frequency, whereby the decline of the partials of the Cor anglais 8 is slower than the decline of both clarinet stops. In comparison to the Krummhorn 8 (Fig. d, striking reed pipe) all the measured free reed pipes have a stronger decline in harmonics, but they have a softer decline in harmonics than the Principal 8 (Fig. e, flue pipe). All three free reed pipes have in common that the level of each even partial tone in the lower frequency range is much lower than the level of the preceding odd partial tone, a characteristic that is also typical of an orchestra clarinet. This is not necessarily typical for free reeds. In a study of Koopman and Cottingham the sound spectrum of a Wolfinger Viola C reed, as well as the sound spectrum of a Williams Pipe Diapason C ] reed, have their maximum on the second partial tone [7]. Free reed organ pipes have pipe bodies in contrast to the reeds of reed organs. Although the influence of the pipe body of a free reed organ pipe has been described as smaller than the influence of the pipe body of a striking reed [], the specific acoustic measurements for testing the influence of the pipe body on the spectrum of free reed pipes have not been done yet. Some formants exist in the higher frequency range of the frequency spectrum of free reed pipes, for example in the spectrum of the Walcker Clarinette 8 with a center frequency near 9 khz and in the spectrum of the Klais Klarinette 8 with a center frequency near khz... Attack transients: log-mag amplitude vs. time plots Log-mag amplitude vs. time plots for five different pipes, all tone F ], are shown in Fig.. In each graph the single partials are plotted from left to right as indicated.

3 submitted to: A C U S T I C A Braasch, Ahrens: Attack transients of free reed pipes acta acustica Vol. 8 () Clarinette 8 [Walcker] a khz. 7.. khz. 7.. khz. 7.. khz. 7.. khz Cor anglais 8 [Walcker] b c Klarinette d Krummhorn e Principal frequency F Figure. Fourier spectra of different organ pipes ( ] ) from top to bottom: Clarinette 8 (Walcker), Cor anglais 8 (Walcker), Klarinette 8 (Klais), Krummhorn 8 (Klais) and Principal 8 (Klais). For free reed pipes it is typical for the initial delay to be gradually longer for higher harmonics: Clarinette 8 (Walcker, Fig. a), Cor anglais 8 (Walcker, Fig. b), Klarinette 8 (Klais, Fig. c). Another characteristic of the free reed pipe is the step in the attack curve which occurs in most of the partials of the three measured free reed stops, except in the fundamental. On the contrary, the single partials of the Krummhorn 8 (Fig. d) start more simultane- ously. The sound of the Principal 8 (Fig. e) starts with the chiff, a short noise that precedes the harmonic sound. This non harmonic component is damped after a short time by the body of the pipe. The existence of the chiff is considered to be necessary for perceiving polyphonic organ music, especially the music of Bach, correctly [9]. In the case of the Principal 8 the center frequency of the chiff

4 submitted to: ACUSTICA acta acustica Vol. 8 () Braasch, Ahrens: Attack transients of free reed pipes Clarinette 8 [Walcker] Cor Anglais 8 [Walcker] Klarinette 8 [Klais] Krummhorn 8 [Klais] Principal 8 [Klais] amplitude log [] a b c d e... Figure. Attack transients of different organ pipes (F ] ), from left to right: Clarinette 8 (Walcker), Cor anglais 8 (Walcker), Klarinette 8 (Klais), Krummhorn 8 (Klais) and Principal 8 (Klais). The distance between two grid lines is in the ordinate. is close to the third harmonic. A steep slope exists in this case. The measured attack transients for the Krummhorn 8 pipe and the Principal 8 pipe are comparable to the measurements of Angster et al. of a Prinzipal 8 pipe, respectively striking reed pipe [, ]. In Fig. the attack transients of free reed organ pipes of different pitches (all Klarinette 8, Klais) are shown. The pipes ascend from left to right in pitch one octave apart from the previous pipe, starting with the lowest tone C. The outermost right graph shows the pipe of the tone B [ instead of the tone C. The reason for this is that the tone B [ is the pipe with the highest pitch of this stop depending on a free reed. Starting from B the pipes are equipped with striking reeds, while flue pipes are used in this stop from the tone A [ on. Throughout the whole range the gradual increase of the initial delays of the single partials with the frequency is present. The attack times, measured as the ; to ; threshold interval, are shown in Fig. for the three Klais organ pipes. As already stated previously the Krummhorn 8 has the shortest attack time throughout the measured frequency range, except for the tone C ( Hz). The Principal 8 has the longest attack times for the low frequencies, but above C the attack time of the free reed is greater. For the two highest tones B and C of the Klarinette 8 stop the attack time is significantly shorter. The reason for this is usage of bound reed pipes in the Klarinette 8 stop. Stops named clarinet exist for both striking reed pipes as well as for free reed pipes. Soon after the introduction of free reed pipes at the end of the 8th century clarinet stops in Germany were built using this new type of pipe. How- onset ( ) ( ). 8 8 frequency in [Hz] Figure. Onset time of different Klais organ stops: Klarinette 8, + with dashed line; Krummhorn 8, x with dash-dotted line; Principal 8, circles with solid line. Each data point is the mean average out of the onset time of the first six partials. The error bars show the standard deviation. ever, with the beginning of the Orgelbewegung clarinet stops with striking reed pipes took over once more, throughout the country. To clarify why the orchestra clarinet can be imitated equally well using striking reed pipes or free reed pipes, measurements of the attack transients of an orchestra clarinet were compared to the previous free reed pipe data. The results for the tone C are shown in Fig., in the same way in which the organ data in Fig. was presented. When played with a hard attack (left graph), the clarinet has an attack comparable to the attack of a striking reed or-

5 Braasch, Ahrens: Attack transients of free reed pipes submitted to: ACUSTICA acta acustica Vol. 8 () Klarinette 8, tone C Klarinette 8, tone C Klarinette 8, tone C Klarinette 8, tone B b amlitude log [] a b c d... Figure. Attack transients of different pipes of the Klarinette 8 stop (Klais), from left to right: C, C C, B [. The distance between two grid lines is in the ordinate. amplitude log [] clarinet: hard attack clarinet: soft attack the hard attack is about SPL larger than the maximum level of the sound played with a soft attack. Similar to the free reed organ pipes the single partial tones start gradually with an increasing initial delay time for higher frequencies. This agrees with the observations made by Meyer, who also confirmed a gradual start in the partials, when the orchestra clarinet is played with a soft attack []. The reason for the different acoustical properties of the clarinet when played with a hard or a soft attack could be explained by different articulations of the tongue that slaps, in the case of the hard attack strongly against the reed. Backus made the observation that the clarinet reed only touches the mouthpiece when it is played loud, but not when it is played softly []. It seems that the orchestra clarinet, usually classified as a striking reed instrument, reveals these properties only when it is played loudly.... a... Figure. Attack transients of an orchestra clarinet (C ): left, played with a hard attack; right, played with a soft attack. The distance between two grid lines is in the ordinate. gan pipe, like the Krummhorn 8 in Fig. d. The slope of the attack becomes less steep when the clarinet is played with a soft attack, comparable to the attack of a free reed organ pipe, like the Klarinette 8 in Fig. c. It should be noted that both graphs, the one with the hard attack as well as the one with the soft attack, are scaled to the same level, even though the maximum level of the sound played with b.. Attack transients: derivative log-mag amplitude vs. time plots Fig. show the derivative log-mag amplitude vs. time plots of three different Klais organ stops: Klarinette 8, Krummhorn 8, Principal 8 (all tone F ] ). Numerical values for the maximal values of the derivation curves and their position in time are displayed in Tab. I. As can be seen in Fig. in most of the cases the sounds of all the stops start with a little hump. This is likely to be caused by the wind chest rather than by the pipes. It will not be considered in our analysis, neither will we consider the peak in the derivative curve that results from this hump. The level of the hump is very low, probably below hearing threshold anyhow. Unfortunately this is not considered in the derivation vs. time plot.

6 submitted to: ACUSTICA acta acustica Vol. 8 () Braasch, Ahrens: Attack transients of free reed pipes 7 derivation of amplitude log [/s] a derivation of amplitude log [/s] c Klarinette 8 [Klais]... Principal 8 [Klais]... derivation of amplitude log [/s] b Krummhorn 8 [Klais]... Figure. Derivative log-mag amplitude vs. time plots of different Klais organ pipes (F ] ). The distance between two grid lines /s in the ordinate. Another characteristic which occurs in all the partials except the fundamental is the characteristic step in the attack curve of free reed pipes as already previously mentioned. In the derivative curve it can be found in the in the form of the double peaks. The comparison with free reed pipes of different pitches show that the peaks are usually further apart for pipes with a higher fundamental frequency and smear to one broad peak for pipes with a lower fundamental frequency. A gradual increase in the initial delays of the single partials with the frequency is also observable for the maximum rate of the log-mag amplitude change. In general, the derivation curve of the Krummhorn 8 has two peaks, too, but in comparison to the Klarinette 8 the derivative curve is narrower and higher with a mean maximum peak of /s averaged over the first six partials. The mean maximum of rate change for the Klarinette 8 is 7 /s. Kl Kr Pr p 9/7 /7 /99 /77 /8 / p 8/9 /88 /78 /79 -/- -/- p /99 /8 /8 /9 /8 9/7 p /79 9/9 /8 /8 /79 /9 p /8 /97 7/78 / / / p /88 9/9 /79 /7 /8 9/7 Table I. Maximal change in the log-mag amplitude of the first six partials p-p for three Klais organ pipes (F ] ): Klarinette 8, Krummhorn 8 and Principal 8. The first value, left of each slash, is the time in ms at which the peak occurs, the second, right of each slash is the peak value in /s. In the cases of double peaked curves the second data pairs are displayed one row below. The mean maximum of rate of change for the Principal 8 estimated over the first six partials is 8 /s. The highest rate of change of the log-mag amplitude occurs in the third partial with 8 /s. This value is extraordinary high and also exceeds the values measured by Pollard and Jansson [] and Pollard [7]. We already noted in Section. that this is caused by the presence of the chiff. Recent publications show evidence that dynamic attributes of sound, like the rate of change of peak pressure, could play a key role in the perception of attack transients [8]. The comparable broad and low peaks of the rate of change of the log-mag amplitude in the case of the Klarinette 8 could be one explanation for the perception of a fairly long onset time. The high rate of change of the logmag amplitude that can be found in the third partial of the Principal 8, might be an explanation of why the chiff is important for perceiving a stronger attack. Here we reach the field of speculation, because up to now it has not been fully clear how the dynamic attributes influence perception, rather it has just been clarified that and in certain cases even illustrated how dynamic attributes are coded in neural responses.. Psychoacoustical Validation A psychoacoutic test was conducted to test which of the acoustical properties of the free reed pipes that were measured and described in the last section lead to the particular perception. In contrary to the previous investigations concerning the perceptual attack [, ] we decided to use synthesized sounds instead of recordings of musical instruments. This allowed us to change single parameters, e.g. onset delay and attack times, that often correlate in natural instruments and so to determine their special role. Ten different sounds were synthesized containing one or more attributes of a certain type of organ pipe: free reed pipe, striking reed pipe or flue pipe. The sounds were presented in a psychoacoustic pair method comparison test to

7 8 Braasch, Ahrens: Attack transients of free reed pipes the listeners who had to detect the sound with the shorter perceptual attack duration in each pair. submitted to: ACUSTICA acta acustica Vol. 8 ().. Method and procedure All test sounds were generated digitally on a PC using Matlab. (sampling frequency. khz). The fundamental tone with a frequency of Hz and the first six overtones were generated using sine functions. For the attack and decay envelope cos -ramps were used. The ramp for the decay was set to ms for all the partial tones and stimuli. The value of the attack rise time was variable between the stimuli, but constant for all the partials within one sound. The following values were tested: ms, ms and ms. In Tab. II the parameters of the stimuli are listed. The rise time is given in the first column labeled rt. The total length of each sound was constantly ms, while the amplitude of each partial was set individually. The values of the relative amplitudes are listed in Tab. II in the columns a ; a7 according to the order of the partial. Three different combinations of amplitudes were tested, one with a relatively slow decay of the amplitude versus frequency: a, ; a, /; a, /; a, /; a, /; a, /7; a7, /, one with a relatively fast decay versus frequency of the amplitude: a, /; a,;a, /; a, /; a, /; a, /; a7, /8 (In this case the maximal amplitude is on the second partial), and a third combination which is identical to the first combination except that all the even partials are muted (marked with in Tab. II). An initial delay could be set separately for each partial. Several combinations were tested. The values in milliseconds are listed in Tab. II in the columns d ; d7 for the single partials 7. In all the cases all partials were in phase. The initial delays were set to corresponding values. Despite the different initial delays, all the partials end synchronously at the end of the sound. In the case of sound stimulus ten a chiff was simulated preceding the sound. It was generated out of a broadband noise with a flat frequency spectrum that was passband filtered with corner frequencies of 8 Hz and Hz using a FIR-Filter (Hamming windowed linear-phase FIR-Filter, 8 coefficients). The chiff was placed at the absolute beginning of the sound and had an overall length of ms. The rise time was ms, the fall time ms (cos -ramps). Finally, the calculated sounds were stored digitally on harddisk. The test was conducted computer controlled in the anechoic chamber of the Institut für Kommunikationsakustik using a PC. The sounds were filtered with the head-related transfer functions of a dummy head (TDT, PD; Head Acoustics) before they were presented to the listener via a headphone (STAX, Lambda Pro). All levels of the signals were set to approximately SPL. Eighteen listeners ( male, female), aged between and years, took part in the experiment. In a single session all the possible combinations between the different sound stimuli were presented twice, the second time in reverse order. Altogether, 9 pairs were presented pseudorandomly in one session. All the listeners except two participated in two sessions. The task of the listener was to determine if the first or the second of the presented sounds had a faster attack. After the presentation of each pair of two stimuli the listener had to type the answer or into a terminal (Sharp, PC-E). The answers were recorded automatically by the computer, together with the identities of the presented sounds. Before each test five test stimuli were conducted, chosen at random. The test stimuli were not recorded... Results The results of the experiments are listed in Tab. III. The id s of the sounds are listed in their order of rank from left to right in the second row. The number of cases in which each stimulus was considered to have the faster attack within the presented pair is displayed in the bottom row. Each stimulus was compared times to another stimulus ( listeners 9 different combinations repetitions + listeners 9 different combinations repetitions= ). The main observations could be proved statistically with a binomial-test. As nullhypothesis was assumed that the listeners did not prefer one of the two compared stimuli. In this case the listener would judge that one of the two sounds had a faster attack by random with a probability of %. In the binomial test those 8 responses were taken into consideration in which the tested pair was compared ( listeners repetitions + listeners repetitions= 8). In all the tested cases the nullhypothesis could be rejected on a significance level of =:. After each of the following statements the id s of the compared sound stimuli and the probability of the nullhypothesis are given in brackets: A steeper attack slope causes the listeners to judge the sound as having a shorter attack duration (/, p = : ;7 ; /, p =: ; ). These results are in agreement with the results of Gordon []. Gordon stated that the slope of the rise function is also a key factor for the perceptual attack time (PAT). On average the chiff improved the perceived attack of a sound (/9, p = :9). A closer examination of the data reveals that the listeners split into two groups when judging sound number ten with the chiff. It was placed within the first three ranks for nine of the 8 listeners, six listeners placed it in the middle ranks five to seven, while the remaining three listeners placed it in rank nine. Thirteen listeners judged that sound ten had a faster attack than the identical sound without a chiff (sound nine), two listeners judged sound ten and sound nine as having the same rank and the three listeners who placed sound ten on the ninth rank perceived sound nine as having a faster attack. After the test the listeners were asked explicitly about their perception of the chiff. Most listeners reported that the chiff gave the sound a stronger or faster attack, but some listeners stated that the presence of the chiff slowed the attack of the sound.

8 submitted to: ACUSTICA acta acustica Vol. 8 () Braasch, Ahrens: Attack transients of free reed pipes 9 ID: rt a a a a a a a7 d d d d d d d7 Chiff. / / / / /7 / -. / / / / /7 / -. / / / / /7 / -. - / - / - / / / / / / / / - / - / / - / - / / / / / / / ,.8-9. / / / / / / / / / / / / x Table II. Parameters for the ten sound stimuli that were tested in the psychoacoustic validation experiment. rt is the rise time of the attack envelope, a ; a7 the relative amplitudes of the partial tones 7 and d ; d7 the initial delay times of the partial tones 7 in milliseconds. The column chiff is marked with an x if the sound is preceded by a noise chiff. rank id Table III. Results of the psychoacoustic test. In the second row labeled id the sound stimuli are listed according to the rank estimated by all eighteen listeners. In the bottom row the total number of judgements is given for the attack when this sound was perceived as the faster one within the presented pair. Gordon has already reported that the listeners in his psychoacoustic experiment split into two groups when judging the perceptual attack time of a flute sound containing a chiff []. The initial delay of the single partials increased the perceived attack duration if the initial delay time increased with the order of the harmonic (/, p = : ; ). But it also increased, if the sound started with the second harmonic (/9, p =:). A difference in perception could also be shown for the reversal of the delay times between the first and second harmonic (8/9, p =:). Grey had already indicated that the presence of synchronity in the transients of the single partials plays an important role in categorizing different sounds after he analysed the results of a pair method comparison test using the method of multidimensional scaling []. The attack is perceived longer, if all the even harmonics are distinguished (/, p = :) or if the levels of the higher harmonics are reduced (/, p = :8 ; ). however, start more simultaneously, whereas the sound of the Principal starts with a chiff that has its center frequency near or at the third partial. The analysis of the logmag amplitude derivation curves of the single harmonics show that the rate of change of the log-mag amplitude curves of the Klarinette 8 is broader and not as high as the rate of change of the log-mag amplitude curves of the Krummhorn 8 and the Principal 8. In addition the two peaked structure is much more distinct for the Klarinette 8. In the case of the Principal 8 the chiff causes the highest rate of change of the log-mag amplitude measured for all three organ pipes. The comparison of the two different types of reed pipes with an orchestra clarinet shows that the attack of the free reed pipes has similar acoustical properties to the attack of an orchestra clarinet played with a soft attack, whereas the attack of the striking reed pipes has similar properties to an orchestra clarinet played with a hard attack.. Conclusion The analysis of the attack transients of three different free reed organ stops: Clarinette 8 (Walcker), Cor anglais 8 (Walcker) and Klarinette 8 (Klais) shows that the rise time of free reed pipes exceeds the rise time of the measured striking reed pipes (Krummhorn 8, Klais), but that they are for frequencies below Hz lower or in the same order as the measured rise time of the flue pipes (Principal 8, Klais). The initial delays of the single partials differ for all three types of organ pipes: free reed pipes have gradually increasing initial delays of the single partials, starting with a delay that increases nearly monotonic to the order of the harmonic. The partials of the Krummhorn, The results of the psychoacoustical pair method comparison test show that free reed organ pipes have quite a number of characteristics that lead to a longer perceptual attack duration compared to the perceptual attack duration of striking reed and flue pipes: the longer rise time of the single harmonics, and therefore a broader and flatter derivation curve of the log-mag amplitude leads to a slower perceptual attack. The same observation can be made if the initial delays of the single partials increase gradually with the frequency. The listeners tend to judge attacks as being slower if the amplitude of the harmonics is decreased with frequency. On average the presence of a chiff causes a slower perceptual attack, although for a small group of listeners the chiff has the opposite effect.

9 7 Braasch, Ahrens: Attack transients of free reed pipes Acknowledgement The authors wish to thank Jens Blauert (Institut für Kommunikationsakustik) for his generous support, Philipp Klais, Heinz Günther Habbig and Horst Sandner (Klais Orgelbau) for their help and cooperation during the organ measurements. We would like to acknowledge the helpful comments of the Associate Editor d. Murray Campbell, and two anonymous reviewers on an earlier version of this manuscript. submitted to: ACUSTICA acta acustica Vol. 8 () [] J. M. Grey: Multidimensional perceptual scaling of musical timbres. J. Acoust. Soc. Am. (977) [] J. W. Gordon: The perceptual attack time of musical tones. J. Acoust. Soc. Am. 8 (987) 88. References [] K. Lehr: Die moderne Orgel in wissenschaftlicher Beleuchtung. Verlag Bernhard Friedrich Voigt, Leipzig, 9. [] W. Hackel, W. Topp: Ein Orgelreisebericht aus dem Jahre 87. Ars Organi (99). [] W. Adelung: Einführung in den Orgelbau. Verlag Breitkopf und Härtel, Leipzig, 9. [] W. Ellerhorst: Handbuch der Orgelkunde. Verlag Benziger & Co. AG., Einsiedeln, 9. [] C. Mahrenholz: Die Orgelregister, Ihre Geschichte und ihr Bau, reprint of the second edition. Verlag Bärenreiter, Kassel etc., 98. [] J. G. Töpfer: Die Theorie und Praxis des Orgelbaues. Weimar 888/Reprint Den Haag, 97. [7] P. D. Koopman, J. P. Cottingham: Acoustical properties of free reeds. ROS Bulletin (99) 7. [8] J. P. Cottingham, C. J. Lilly, C. H. Reed: The motion of airdriven free reeds. J. Acoust. Soc. Am. (999). [9] R. B. Johnston: Pitch control in harmonica playing. Acoustics Australia (987) 9 7. [] H. Bahnson, J. Antaki, Q. Beery: Acoustical and physical dynamics of the diatonic harmonica. J. Acoust. Soc. Am. (998). [] R. Caussé,???: Acoustics of the accordion??? Proceedings of the é Congres Français d Acoustique, Ecole Polytechnique Fédérale de Lausane, France,.?????? [] J. P. Cottingham, C. A. Fetzer: Acoustics of the khaen. Proceedings of the International Symposium on Musical Acoustics, Leavenworth, Washington, USA, [] A. O. S. Hilaire, T. A. Wilson, G. S. Beavers: Aerodynamic excitation of the harmonium reed. Journal of Fluid Mechanics 9 (97) 8 8. [] J. P. Cottingham, C. H. Reed, M. Busha: Variation of frequency with blowing pressure for an air-driven free reed. J. Acoust. Soc. Am. (999) 9. [] N. Fletcher: Excitation mechanisms in woodwind and brass instruments. Acustica (979) 7. [] H. F. Pollard, E. V. Jansson: Analysis and assessment of musical starting transients. Acustica (98) 9. [7] H. F. Pollard: Feature analysis of musical sounds. Acustica (988). [8] P. Heil, D. R. F. Irvine: On determinants of first-spike latency in auditory cortex. J. Neurophysiol. 77 (997). [9] W. Lottermoser: Orgeln, Kirchen und Akustik: Die akustischen Grundlagen der Orgel. Verlag Erwin Bochinsky, Frankfurt am Main, 98. [] J. Angster, A. Miklós: Documentation of the sound of a historical pipe organ. Applied Acoustics (99) 8. [] J. Angster, J. Angster, A. Miklós: Akustische Messungen und Untersuchungen an Orgelpfeifen. Acta Organologica (997) 7. [] J. Meyer: Akustik und musikalische Aufführungspraxis. Verlag Erwin Bochinsky, Frankfurt/M, 99. [] J. Bachus: Vibrations of the reed and the air column in the clarinet. J. Acoust. Soc. Am. (9) 8 89.