A SIMPLIFIED METHOD FOR THE DETERMINATION OF THE QUALITY OF AN ANECHOIC SPACE AT THE CSIRO NATIONAL MEASUREMENT LABORATORY

Transcription

1 A SIMPLIFIED METHOD FOR THE DETERMINATION OF THE QUALITY OF AN ANECHOIC SPACE AT THE CSIRO NATIONAL MEASUREMENT LABORATORY Bruce H Meldrum and Anthony Thorley TIPP /10/02 CSIRO National Measurement Laboratory Sydney, NSW, 2070, Australia. Bruce.Meldrum@csiro.au 1 INTRODUCTION The testing of acoustic measuring equipment requires a sound field with well specified characteristics and that the sound field be reliably reproducible. The Australian Standards AS parts 1 1 and 2 2 for Sound Level Meters follow international practice described in the IEC Standards IEC and IEC in that the sound field must be of plane progressive waves. This implies an anechoic space and this practice is followed by the majority of other National Standards Laboratories. It is not a trivial exercise to build such a space and then to make the measurements necessary to accredit the space in order to verify that there are no room modes present in the frequency band of intended use. There is little guidance for performance criteria other than AS (ISO ) (Determination of sound power levels in Anechoic and Hemi-Anechoic rooms) Appendix A to use for accreditation. It is perhaps for these reasons that few working anechoic spaces exist in Australia and not all have been through an accreditation examination. During the accreditation of the CSIRO National Measurement Laboratory to the requirements of AS ISO/IEC (General requirements for the competence of testing and calibration laboratories), the Acoustics and Vibration Standards project was inspected as an operating unit within the laboratory infrastructure. As part of this inspection, the procedure used for acoustic testing in the free field, amongst other functions, was verified. In preparation for the verification the Acoustics project embarked on a program to test the performance of an existing anechoic space constructed over 15 years ago that is in regular use for free field calibration activities. As it was not possible to engage in a program of the scope that is normally recognized as a full evaluation, a modified fit for purpose program was adopted. This modified program was carried out with minimum disruption to the normal work flow for the space and has demonstrated with a high degree of confidence that the space is suitable to be used for the purpose intended with few restrictions. This work has been reported internally by Thorley and Meldrum 7 and the space is described below together with the test program and the results. Not reviewed by Acoustics 2002 Scientific Committee

2 2 ACOUSTIC STANDARDS IN AUSTRALIA The CSIRO National Measurement Laboratory is responsible through the National Measurement Act of the Australian Parliament to maintain a system of primary physical standards to which all legal measurements in Australia are referred. The physical standards have their basis in the seven base units which are defined in the International System of units, known as the SI units. These are: the metre, the kilogram, the second, the ampere, the Kelvin, the candela and the mole 3. Units for other quantities are derived by combining the base units. This process is described in many publications such as the monograph by Sandars 8 on Traceable Measurements from the NML technology transfer series. The National Measurement Act (1948, amended 1960) describes the Australian legal units. In an amendment to the regulations in 1984 the acoustical quantities Sound Pressure Level (SPL, db) and Sound Intensity were added to the list. However, the practical use of the definition for SPL relies entirely on the ability to disseminate SPL via a microphone with known sensitivity and this imposes a further subdivision based on intended usage. This subdivision is the use in a pressure field or in the free field. The term pressure field implies a coupler or closed space with dimensions much less than the wavelength of sound at the highest frequency of use such as would be encountered in an acoustic calibrator. In this situation the SPL throughout the coupler is essentially constant for practical purposes. The descriptor free field implies a sound field where the SPL is not constant within the space but will obey a 1/r 2 law as the distance of the observer from the source is varied and where the incident field at the observer is from one direction only. This is a simulation of an outdoor situation where a point source is suspended from a thin acoustically absorbent string many ten s of metres above a soft absorptive surface in a region with no wind or air movement. This is obviously not practical from a control perspective so the process must be moved indoors, to an anechoic space. This then requires a space where there are no echoes from the surrounding enclosure which must then be totally absorbent. Each of these situations requires different types of reference microphones. The methods to set these tests up are described below. 3 REALISATION OF FREE FIELD STANDARDS AT NML The IEC (Measurement Microphones) series of standards describes processes for realising microphone sensitivity at chosen frequencies in both the coupler (Pressure Reciprocity) and free field environment (Free Field Reciprocity). Both use the reciprocity method whereby a class of microphone described in (Laboratory Standard Microphones, Specifications) can operate both as a receiver and a transmitter. Given a set of three microphones (a triad), and when operated in pairs, the sensitivity of the individuals may be determined. The coupler method is readily achieved by either in-house equipment such as has been in use at NML for many years 12 (AIP Handbook of Condenser Microphones) or by modern commercial systems based on the IEC methods described in These methods achieve microphone

3 pressure sensitivity at a wide range of frequencies from 20 Hz up to 25 khz for smaller LS2P microphones (1/2-inch) with low uncertainties, typically 0.05 db at the mid frequencies rising to perhaps 0.3 db at the highest frequency. The measurement environment is a coupler ranging in volume from 0.3 cc to 10 cc which is readily characterised by dimensional metrology and the corrections given in the IEC standard 9. The associated electrical measurements are readily made with conventional equipment. Whilst the process is complex, pressure reciprocity determinations are rapidly and efficiently made routinely at NML under computer control and recent upgrading of the system with new equipment has allowed the frequency range to be extended to 25 khz. This is highly significant in light of the developments discussed below. The free field reciprocity method in contrast is far from trivial as the signal levels at the receiving microphone are small and this implies both a high standard of measurement environment to achieve workable signal-to-noise ratio and sophisticated signal recovery techniques to work in the ever present low frequency noise from the environment. The measurement space must also be anechoic at the intended frequencies. The need to know the precise distance between the source and receiver microphones in order to know the attenuation for distance brings an added problem, that of the acoustic centre of the microphone. Put simply, the actual centre of the sensing process is not necessarily at the diaphragm, particularly as the wavelength shortens. This requires an iterative process where the separation distance is varied in small increments and a curve fit made to the results to estimate the acoustic centre. Free field reciprocity to establish a set of free field reference microphones as distinct from a set of pressure reference microphones has been carried out at NML in the past 13 but is so demanding of time resources that the practice has been discontinued in light of better methods described below. It may be asked why establish separate free field and reciprocity reference sets? This has traditionally been brought about by the very different frequency response characteristics of the same microphone in a pressure field when compared to a free field environment. A microphone which has a flat response in a pressure field will exhibit a rising response in the free field due largely to the geometrical diffraction around the microphone and its supporting structure. The difference will be negligible at 250 Hz rising to perhaps 0.07 db at 1 khz and as much as 8 to 10 db at or near the microphone resonance which is typically 20 khz for an LS2P (1/2-inch) microphone. A microphone designed to have a flat response in the free field will conversely exhibit a decreasing response with frequency in a pressure field due to the increased damping. Another important reason for concentrating on pressure reciprocity has been to minimize the uncertainty associated with the measurement. An alternative to the free field reciprocity method for the establishment of free field reference microphones is described in the standard IEC where sets of corrections from pressure to free field are given for typical 1-inch microphones up to 16 khz, above which it was not possible to assemble data for 1-inch microphones due the limitations of pressure reciprocity with the larger microphones. This standard has been revised and will be eventually published in the IEC series as part In IEC values are given for the corrections for LS2P (1/2- inch) microphones from 1 khz up to 25 khz which when used with modern pressure reciprocity systems will allow the efficient establishment of a free field reference based on pressure reciprocity. It is this path which has been chosen at NML and the procedure has been described and accredited for the NML quality system under AS ISO/IEC An important part of this procedure is the measurement space which is required to be anechoic at the frequencies of interest.

4 4 CRITERIA FOR ANECHOICITY The best normative guidance for criteria and testing remains Appendix A of the standard AS , which is intended for use in the determination of sound power in anechoic spaces. The parent document ISO is currently under revision with little indication that the criteria in Appendix A will be changed. Under Paragraph A3.1.1, Test Sound Sources, the requirement is that the source shall be omnidirectional with deviations of less than ± 1 db which as a search of the literature described in Thorley and Meldrum 7 shows is difficult to achieve and will require a number of different configurations for each frequency band. The general method for investigating a room s anechoicity is to compare the variation in SPL with the ideal 1/r 2 behavior expected in a free field. This is indicated by a decrease in SPL of 6 db for each doubling of distance from the source. The test would ideally be carried out by positioning the omnidirectional source at the same location as the device under test, in this case a sound power source. To test the 1/r 2 law the microphone is moved on a traverse in a straight line away from the centre of the source along at least 8 different paths. Key paths are the lines from the source to the room corners. It is suggested in the Standard that one-third octave frequencies be used below 125 Hz and above 4 khz with octave spacing within the band 125 Hz to 4 khz. For qualification, Table A1 from the Standard gives values for the maximum allowable deviation from the theoretical levels for several frequency bands. This Table is reproduced below for information. Type of test room Anechoic One-third octave band centre frequency, Hz Allowable differences db 630 ± to 5000 ± ±1.5 Table 1: Allowable deviation from the ideal 1/r 2 law, from Table A1 of AS CONSTRUCTION OF THE NML ANECHOIC ROOM At the time of the design work for the new National Measurement Laboratory building at Lindfield in the early 1970 s, there was no requirement for or provision made for an anechoic space as acoustic standards were not then identified as an immediate need. After the NML moved to the new laboratory it was realized that an anechoic space would be required particularly for free field reciprocity as the need for traceable acoustics standards had become an important issue. There are two primary requirements to be met during the construction of such a space: A. full absorption of all sound waves made within the space, and B. exclusion of all external sounds or vibration to enable a very low noise floor to be established. The practical realization of both requirements dictates that the volume required for such a facility is probably an order of magnitude larger that the eventual volume of the working space with each

5 outer dimension being at least double that of the inner usable space. This increase in dimension arises from: A. the thickness of the inner absorptive lining required for effective wide band absorption, and B. the need to have a massive vibration isolated outer box to stop all re-radiated sound arising from direct vibration through the ground or re-radiated sound via the shell. For a well constructed anechoic chamber, as a rough guide, the lowest frequency at which the behavior is still anechoic (cutoff frequency) is approximately the frequency at which the wavelength of sound λ at that frequency is equal to the shortest dimension of the space. The wavelength λ at an air temperature of 21.5 C for two frequencies of interest here is: at f = 31.5 Hz, λ = m at f = 125 Hz, λ = 2.76 m. It can be seen from these figures that a space required to be anechoic to 31.5 Hz where λ = m will be a very large structure with outer dimensions of approximately 20 m per side. Arising from the limitation of being forced to fit the NML facility within the floor to ceiling height of 3.5 m of the existing rooms and with no prospect of breaching the upper ceiling, the structure had to be a compromise. The final dimensions are listed in Table 2. Dimension Inner Outer Length 2.8m 4.6m Width 2.2m 4.0m Height 2.5m 3.5m (Floor to Ceiling) Table 2: Inner and Outer dimensions of the NML anechoic chamber Based on the rule of thumb above it would be expected that the NML space had a cutoff frequency between 125 Hz and 160 Hz, ie λ = 2.76 m to 2.16 m. We shall explore this in more detail below. The conventional construction method for such a chamber uses pyramid shaped wedges of mechanically stable acoustically absorbent material with the apex pointing into the space so that sound waves are progressively absorbed within the valley between the pyramids. Such wedges may be up to several metres long which can significantly reduce the usable inner volume. The space constraints of the construction of the NML chamber did not allow the luxury of long faced wedges so the designers chose to construct the absorbent surfaces from graded density cotton blanket with the least dense material on the innermost surfaces. This material is available in a number of densities depending on intended use, from the cotton wool known from medical use to the thermal insulation in clothing to the damping material used in loudspeaker enclosures. The typical thickness of the total combined layers is 500 to 600 mm with a 50 to 100 mm airspace behind. It was not possible to construct a massive sealed outer box suitable for vibration isolation so a floor to ceiling masonry box with walls of brick was constructed with light angle iron supports for the suspension of the blankets attached to the walls facing inwards. Further support for the large surfaces of the blankets was provided in the form of open mesh fish net. An acoustically transparent mesh floor (trampoline) of tensioned stainless steel 1.5 mm cable at 50 mm spacing

6 allows careful access and a massive door of lightweight concrete with rubber seals and locking catches is mounted on a roll out frame suspended from the ceiling. A pneumatic actuator with safety interlocks provides the opening and closing motion. Mounted in an upper corner of the space is a small video camera and spotlight used to read instrument displays if an interface is not available from the device under test. A pair of laser diode (lecture) pointers using external power provides an intersection point in the centre of the chamber to which position of all devices is referenced. This point is situated 109 cm from the sound source which is positioned in a corner facing along a diagonal. The normal source is a 250 mm coaxial transducer incorporating both a roll surround low frequency unit and a high frequency unit mounted coaxially within the low frequency magnet and radiating through a horn in the centre of the woofer cone (P-Audio model BM-10CXA Blue Monster ). The source uses an infinite baffle enclosure consisting of a 1m length of thick walled 250 mm PVC pipe mounted vertically in a corner of the chamber. The lower end is sealed and a 90 bend with a mounting place allows the transducer to face out along a diagonal 1 m above the floor. A roll of acoustic absorber is placed in the enclosure tube to inhibit any Helmholtz resonances. The crossover between the low and high frequency units at 1 khz is accomplished electrically with a computer-controlled switch. The calibration and test process is controlled from a PC using a GPIB bus. A Stanford Research SR830 lock-in amplifier is used to provide both the energizing signal which is amplified with a Bruel & Kjaer type 2706 power amplifier and a signal recovery channel. The detection capability of the lock-in provides stable and accurate signal recovery in the presence of low frequency noise which abounds in air conditioned buildings where doors are opened and closed. A Bruel & Kjaer type inch free field microphone is used as the reference, its response is established by 3 independent means and stored in the control system as a data file. Before use, the chamber is calibrated with the reference microphone in position and the control system learns the drive level at each 1/3-octave frequency to produce a constant SPL at the reference point. The reference microphone is then replaced by the test object and the learnt levels replayed. An SPL of 74 db is generally used as this is sufficiently above the noise floor and avoids excessive heating in the source which may cause drift. In normal use the absolute level is not of prime concern as all measurements are made relative to 1 khz (or 250 Hz) and the measurement system then produces a frequency response at 1/3 octave spacing. 6 EXPERIMENTAL RESULTS For the test of performance to the 1/r 2 law, a series of measurements using the system described above was made over a range of source to microphone distances from 64 cm to 184 cm at approximately 10 cm spacing and over the standard frequency range from 31.5 Hz to 20 khz. The microphone was moved along a diagonal away from the source. The standard test control program was modified so that the measurement set made at the standard distance of 109 cm became the reference levels to which the levels at each frequency for each distance were referred. The initial response of the microphone in its standard position (d = 109 cm) has been subtracted from all subsequent responses so that Figure 1 shows the relative SPL at each distance without

7 any influence from the microphone response Each curve corresponds to a frequency response at a particular position but displaced in amplitude by the 1/r 2 law. 8 6 Relative Sound Pressure Level (db) Room Length λ /4 Room Length λ Microphone/Speaker Separation (d) d = 109 cm (standard) d = 64 cm d = 129 cm d = 79 cm d = 139 cm d = 89 cm d = 154 cm d = 99 cm d = 169 cm d = 119 cm d = 184 cm Frequency (Hz) Figure 1: Plots of sound pressure level (db re: 0 at d = 109 cm) as a function of frequency for the values of d listed. The microphone response at the standard position of d = 109 cm is subtracted from all responses. It can be seen that there is in general a steady increase or decrease in sound pressure level as the microphone is moved toward or away from the speaker. In an ideal free field, one would expect a decrease in sound pressure level of 6 db for every doubling in distance. Clearly this uniform behavior does not apply at the lower frequencies where it is obvious that the behaviour of the chamber deviates from that of an ideal free field. Recall that the internal chamber dimensions are length 2.8 m, width 2.2 m and height 2.5 m. Typically, the chamber temperature is 21.5 ºC and at this temperature a frequency of 125 Hz corresponds to a wavelength (λ) equal to the length of the room. Similarly, at a frequency of 31.5 Hz the room length is equal to λ/4. These frequencies are shown in Figure 1. This data may be represented as a plot of level vs. distance for each frequency as shown in Figure 2 below. The general method of investigating a room s anechoicity is to compare the variation in SPL with the ideal 1/r 2 behaviour. From Figure 2 it can be readily seen that the greatest deviation from ideal 1/r 2 behaviour occurs at the low frequencies, most noticeably at frequencies below 160 Hz; this trend was also evident in Figure 1. Ideal 1/r 2 free field behaviour is also plotted in Figure 2. To further aid the comparison, the deviation from an ideal free field can be obtained by subtracting the ideal free field value (at the appropriate d) from each of the recorded data points. The resulting deviations can be plotted and compared to the allowable tolerances according to AS The deviation from ideal free field behaviour is shown in Figure 3 below, together with the allowable tolerances. It is interesting to note the behaviour of the room for frequencies below 125 Hz where the deviation from the 1/r 2 law indicates an almost constant SPL at any of the measurement points within the distance range.

8 Relative Sound Pressure Level (db) ideal f (Hz) Microphone/Speaker Separation d (cm) Figure 2 Plot of sound pressure level (db re: 0 at d = 109 cm) as a function of d for each of the 1/3 octave frequencies in the range 31.5 Hz to 20,000 Hz. The ideal 1/r 2 free field behaviour is also shown. Deviation (db) ALLOWABLE TOLERANCES For f 630 Hz or f 6300 Hz = ± 1.5 db For 800 Hz f 5000 Hz = ± 1.0 db f (Hz) Microphone/Speaker Separation d (cm) Figure 3 Deviation of NML anechoic chamber sound pressure level from ideal 1 /r 2 free field behaviour. The deviation in db is plotted as a function of d for each of the 1 / 3 octave frequencies in the range 31.5 Hz to 20,000 Hz. The sound pressure levels are relative to 0 db at d = 109 cm. Also shown are the allowable tolerances according to AS

9 Figure 3 again demonstrates that it is the low frequencies that give rise to the greatest deviations from ideal free field behaviour. The data from Figure 3 is replotted in Figure 4 with the omission of all responses for frequencies below 160 Hz. This allows an expansion of the SPL scale to better view any irregularities. 2.0 Deviation (db) Allowable tolerance for f 630 Hz or f 6300 Hz = ± 1.5 db Allowable tolerance for 800 Hz f 5000 Hz = ± 1.0 db f (Hz) Microphone/Speaker Separation d (cm) Figure 4 Deviation of NML anechoic chamber sound pressure level from ideal 1/r 2 free field behaviour. The deviation in db is plotted as a function of d for each of the 1/3 octave frequencies in the range 160 Hz to 20,000 Hz. The sound pressure levels are relative to 0 db at d = 109 cm. Also shown are the allowable tolerances according to AS Figure 4 shows that for frequencies over 125 Hz the chamber behaviour generally lies within the free field limits as set by AS , the only notable exception being the result at d = 64 cm at a frequency of 12,500 Hz. This deviation occurs close to the source at d < 80 cm and armed with this knowledge the chamber is not used in that region. 7 USE BELOW CUTOFF FREQUENCY The chamber is routinely used below the cutoff frequency of 125 Hz down to a limit of 31.5 Hz which is imposed by the limitations of the source. The diffraction effects in the free field on common test items becomes negligible below 250 Hz and the measurement below 125 Hz in the chamber takes place essentially in a pressure field. The advantage is that the complete microphone including the back vent is exposed to the sound field unlike the situation in a coupler. The critical consideration in this case is that the reference and test microphones are accurately positioned so that the acoustic centre at the reference frequency of 1 khz is identical. Above cutoff frequency changes in position cancel as the 1 /r 2 effect is the same at all frequencies but below cutoff the sound field is constant and positioning errors becomes apparent due to the

10 measurement generally being referenced to 1 khz. The laser positioning system overcomes this problem with calculable and reasonable uncertainties. 8 CONCLUSION It is evident from Figures 1 to 4 that for the range of measurements conducted, the NML anechoic chamber performs to the published criteria as an approximation to a free field over a reasonable range of frequencies and distances around the standard distance of d = 109 cm with the existing source. The chamber cutoff is between 125 Hz and 160 Hz as predicted. Clearly within approximately ± 20 cm of the standard position of 109 cm the deviations in recorded sound pressure level from ideal 1 /r 2 behaviour fall within the tolerances from AS for frequencies at or above 160 Hz. For frequencies below 160 Hz the sound field behaves as a pressure field with essentially constant SPL in the region explored. With careful positioning of the reference and test objects to locate the acoustic centre and by making provision in the calculated uncertainties, the chamber is routinely used from 31.5 Hz to 20 khz. REFERENCES 1. Australian Standard Sound level meters, Part 1: Non-integrating, AS Australian Standard Sound level meters, Part 2: Integrating-Averaging, AS International Standard Sound level meters, CEI/IEC V International Standard Integrating-averaging Sound level meters, CEI/IEC V Australian Standard Acoustics-Determination of sound power levels of noise sources, Part 6- Precision methods for anechoic and hemi-anechoic rooms, AS Australian Standard General requirements for the competence of testing and calibration laboratories, AS ISO/IEC Anthony Thorley and Bruce Meldrum, March 2001, Investigation of the Acoustical Properties of the NML Anechoic Chamber, National Measurement Laboratory report TIPP Glenda Sandars, March 2002, Traceable Measurements, Monograph 3: NML Technology Transfer Series, TIPP International Standard Measurement microphones Part 2: Primary method for pressure calibration of laboratory standard microphones by the reciprocity technique, IEC International Standard Measurement microphones Part 3: Primary method for free-field calibration of laboratory standard microphones by the reciprocity technique, IEC International Standard Measurement microphones Part 1: Specifications for laboratory standard microphones, IEC AIP Condensor microphone handbook, Chapter 12 Reciprocity calibration at the CSIRO National Measurement laboratory. 13. D. L. H. Gibbings and A. V. Gibson, 1984, Free-Field Reciprocity Calibration of Capacitor Microphones at Frequencies from khz to Hz, Metrologia 17, 7-15 (1981). 14. International Standard, Values for the difference between free-field and pressure sensitivity levels for one-inch standard condensor microphones, IEC Draft International Standard, Measurement microphones Part 7: Values for the difference between free-field and pressure sensitivity levels for laboratory standard microphones, IEC