The Comparison between MLS and Swept-sine Impulse Response Measurement Methods
Measuring the impulse response has been one of the backbones for the room acoustics measurement and analysis to get features of a space such as frequency response, energy decay, reflection level, etc. Since the impulse response of an acoustic space is one of the most significant characterizations, the comparison between different measurement techniques is demanded to get more accurate impulse response. To perform the measurement of the impulse responses of a room, two of the most common means are compared: Maximum Length Sequence (MLS) and Swept-sine signals. Although there are more methods used for measuring impulse responses, these two techniques are well established and used widely. These methods have been researched in many papers, but these methods have different behavior and the comparison between them has discussed in a few papers. The accurate impulse response measurement is very critical in room acoustics, because many parameters are given from the measurement. Therefore both methods will be performed to capture the impulse response of a room, analyzed and discussed in this research.
Maximum-Length Sequences (MLS)
MLS method for the acoustical impulse response measurements was designed by Schroeder in 1979 and became one of the measurement standards in 1980s. MLS signal is analogous to artificially generated white noise signal, which also can be used for measuring Impulse response. However using pure white noise requires long testing time to get the accurate measurement since it is non-periodical and has randomly distributed sequence. On the other hand, MLS signal is a periodic and pseudorandom sequence, which has a binary sequence of L = 2n - 1 (n: the order of the sequence).
Since the measurement using MLS method adopts circular cross-correlation to de-convolve the impulse response, the periodic impulse response can be achieved. It makes one of the important problems of using the MLS method, which is time aliasing error. The other major problem of the MLS technique is the occurrence of artificial distortion commonly ‘distortion peaks’. These cause characteristic noise when the measured impulse response is convolved with a dry signal so as to realize the space. On the other hand, according to Policardi, MLS has complete frequency spectrum and provides flat sound spectrum so that it’s easy to estimate the IR frequency response (2011). Also the immunity to signals that are not correlated with the exciting signal is one of the strengths of MLS technique.
MLS technique is on the basis of the assumptions of linear time-invariant (LTI) system, which could cause a problem when the assumptions are not suitable. Especially MLS method is so delicate that it doesn’t perceive non-LTI system, and requires that the exciting signal is minutely synchronized with the digital sampling system to measure the impulse response. The measurement using swept sine was proposed by Farina to overcome the limitations of MLS method. This technique is based on the ideas as follows: According to Farina, by using an exponential time-growing frequency sweep, it is possible to simultaneously de-convolve the linear impulse response of the system and to selectively separate each impulse response corresponding to the harmonic distortion orders considered (2000). Indeed, he succeeded in achieving a sequence of measured impulse responses separated along the time axis after de-convolving the sampled impulse response. Through the FFT process, the linear frequency response and spectrum of the harmonic distortion orders can be achieved.
The theoretical background of both MLS and Swept-sine techniques will not be discussed deeply in this paper, and the above brief descriptions of two methods are derived from Farina (2000), Stan, Embrechts and Archambeau (2002).
Goal & Methodology
There are various techniques to capture impulse response of a room. But the choosing one of the methods depending on the conditions of a room is important. Through this comparative experiment, I expect that the result will show the pros and cons of the MLS and Swept sine methods, and this will be helpful for deciding one of the measurement techniques when a room acoustic measurement is performed.
Figure 1. The disposition of the measurement elements
A comparative test of the two impulse response measurement techniques was performed in a classroom and a studio at NYU Steinhardt Education building. One directional microphone (AKG perception 220), a loudspeaker (APS Klasik) and an audio interface (Focusrite Clarett 2pre) were used for the test. The disposition of the measurement elements (Figure 1.) in two different rooms is based on the test of Stan, Embrechts and Archambeau (2002). The microphone and the loudspeaker were placed in the middle of the room, placed one meter apart from each other at the same height.
The acoustic measurement software ‘ARTA’ was used to generate the MLS signal and the swept sine signal since the program provides many types of signal generator and analyzer, which enables reliable, accurate and fast comparisons between the different techniques. It controls the emission of the different exciting signals through the loudspeaker and records the signal through the microphone simultaneously. And then, the de-convolution process will be performed automatically.
Common parameters to the both methods are as follows: sampling frequency (44,100 Hz), number of times the signal is generated through the loudspeaker, recording mode (mono), same background noise level, etc. Also sequence length of both different methods is considered (32767: 743.02ms, 262144: 5944.29ms), and initial and final frequency (20 - 20k Hz) and sweep duration (32k: 743.04ms, 256k: 5944.31ms) are considered for swept sine method. Noise levels in the studio and the garage are present 40 dB and 53 dB respectively. And the amplitude scales of impulsive noise and other harmonics are compared and discussed in this paper.
Analysis & Results
The impulse responses were measured and compared in two different places (a recording studio and a garage). The result shows that impulse response obtained by using two different methods has similar, but slightly different characteristics. Each of the exciting signals was generated three times (for both short sequence and long sequence respectively), and the impulse responses were recorded simultaneously. The measured impulse responses on Figure 2 and Figure 3 show the differences between MLS technique and Swept Sine technique when an impulsive noise is present. The amplitude scales are different in same condition of the rooms.
Figure 2. Impulse response obtained in a recording studio using Swept Sine (left) and MLS (right)
Figure 3. Impulse response obtained in a garage using Swept Sine (left) and MLS (right)
Figure 2 and Figure 3 shows the measured impulse responses of two different places. The result obtained from MLS technique shows that its pseudo-random noise ability randomizes the phase of non-necessary signal that is not correlated to the exciting signal, so that any subsidiary noise are distributed and attenuated by averaging technique while the de-convolution process after the impulse response is recorded. These results were similar to those reported by Dunn and Hawksford (1993). On the other hand, the impulse response measured by swept sine signal has many residual noises including impulsive noise after the impulse response is de-convolved. The residual noises correlated with the emitted exciting signal are not excluded appropriately. And the differences between the two measurements in the garage make the bigger difference clearly than the measurements in the studio. The result shows that a measurement using MLS technique would be preferred to Swept-Sine method for measuring impulse response in a kind of noisy environment.
Conclusions & Future Work
By using ARTA software, the comparative experiment between MLS technique and Swept Sine technique was performed in two different places. And the expected results from the both techniques could be achieved in some degree from the tests. Through the MLS technique, more accurate impulse response can be obtained in a noisy place comparatively. However it was difficult to find differences between two different places. As a future work, a measurement of a huge place such as a catholic church or an auditorium, and an anechoic room would be performed. For more objective comparison of two different techniques, signal-to-noise ratios should be optimized for each technique in next experiment. And besides MLS and Swept Sine techniques, other methods (IRS, different types of noise, Time-Stretched Pulses, etc.) for measuring impulse response would be performed and compared. Particularly IRS (Inverse Repeated Sequence) technique is discussed in Lopez and Pauletto’s paper (2012), which is made of two MLS sequences, one sequence is an inverted copy of the other sequence. This technique needs longer time for impulse response measurement, but it shows greater immunity to distortion than MLS. It also needs to be compared with the other methods in various environments.
Stan, G. B., Embrechts, J.J., Archambeau, D. (2002). Comparison of Different Impulse Response Measurement Techniques. Journal of Audio Engineering Society, 50(4), 249-262 Retrieved from http://www.bg.ic.ac.uk/research/g.stan/JAES_Online_version.pdf
Farina, A. (2000). Simultaneous Measurement of Impulse Response and Distortion with a Swept-Sine Technique. Audio Engineering Society, 108th convention. Paper 5093 Retrieved from http://www.aes.org/e-lib/browse.cfm?elib=10211
Dunn, C., Hawksford, M. O. (1993). Distortion Immunity of MLS-Derived Impulse Response Measurements. Journal of Audio Engineering Society, 41(5), 314-335 Retrieved from http://www.aes.org/e-lib/browse.cfm?elib=7004
Policardi, F. (2011). MLS and Sine-Sweep Technique Comparison in Room-Acoustic Measurements. Elektrotehniški vestnik, 78(3), 91-95. Retrieved from https://pdfs.semanticscholar.org/0228/33d45bb300ae8ce00b2298646ff39ee60f1a.pdf
Lopez, M., Pauletto, S. (2012). Acoustic Measurement Methods for Outdoor Sites: A Comparative study. 15th International Conference on Digital Audio Effects (DAFx-12),
UK. Retrieved from https://www.dafx12.york.ac.uk/papers/dafx12_submission_51.pdf