Q: What are differences between MLS and Fourier Analyzer?

A: Both types of analyzers use crosscorelation method and signal/spectrum averaging to estimate the impulse/frequency response. The fundamental difference is how crosscorrelation estimation is done; MLS analyzers uses Hadamard transform, while Fourier analyzer uses Fourier transform. The second difference is that MLS analyzers exclusively use the MLS  excitation signal, while Fourier analyzers can use various excitation signals: noise, periodic noise, swept-sine or wideband music signals. ARTA has implemented both measurement methods. It is left to users to decide which type of analyzers will be used for response estimation, but ARTA User Manual suggests the use of Fourier analyzer with swept-sine or periodic pink-noise excitation.

Following papers discuss the problem more deeply:

I. Mateljan, K. Ugrinovic: The Comparison of Room Impulse Response Measuring Systems, Proceedings of the First Congress of Alps Adria Acoustics Association,  Portoroz, Slovenia, 2003, ISBN 961-6238-73-6 .

I. Mateljan: Audio Quality Measurements in Communication Systems, Proceedings of the Second Congress of Alps Adria Acoustics Association, Opatija, Croatia, 2005, ISBN 953-95097-0-X.

Q: Why ARTA uses "dual channel Fourier analyzer" mode and "single channel Fourier analyzer" mode?

A: A dual channel mode is proper mode to estimate frequency/impulse response. Single channel mode always gives biased IR estimation, but if you have a low quality soundcard "single channel mode" can give better S/N at low frequencies.

Single channel mode is obligate if you use the soundcard microphone input channel, as it is, on many soundcards, implemented as mono channel,

Q: Why ARTA is showing impulse response with "zero time" shifted to sample position 300?

A: Modern digital systems often use some kind of finite impulse response (FIR) filtering or equalization. Such kind of filtering introduce predelay (or pre-ringing) in impulse response (IR). The predelay usually uses from 64 to 256 samples. To see effects of digital filter predelay,  ARTA shifts "zero time" to sample position 300.

In the single channel mode, zero time is undefined as PC-soundcard latency is not constant. Then ARTA shifts IR in a way that position od IR maximum is close to sample position 300.

This means that single channel mode IR measurements can't be used for absolute delay ("time of flight") estimation.

Q: When I measure FR in dual channel mode, why I always obtaine the same magnitude response although I change souncard output level?

A:  FR or IR measurement in dual-channel mode is example of  "ratio mode measurement".  It is independent of excitation level as it gives the ratio of measured system output and measured system input voltage.

Spectrum analyser is example of " level mode measurement", where measured value is proportional to excitation level.

In program STEPS you can choose level mode or ratio mode when measuring system response.

Level mode was popular in old swept-sine response measurement, while ratio mode is dominant method in Fourier analyzers.

Q: When measuring Impulse response with periodic noise we always have to wait at least one sequence length to start measurement in steady state condition, while when measuring with swept-sine ARTA start recording of response immediately, and after swept sine excitation stops recording continue for one more sequence length. Why?

A: In both cases ARTA uses Fast Fourier Analysis to calculate IR. A known characteristic of such analysis is that it treats the signal as periodic with period equal to sequence length. When measuring room response the reverberation should be accounted, and as rule of thumb we need sequence length that is larger than reverberation time. 

When measuring with swept-sine ARTA treats it as aperiodic transient signal and must record one sequence more to capture signal reverberation tail.

The recorded sequence of double length is then used in cross-corelation calculation using FFT. The resulting IR is truncated to length equal to excitation sequence length.

Q: I have a calibration file for the microphone FR. Is there possibility to include such a calibration file in ARTA?

A: Starting from version 1.0.1, ARTA has capability to apply frequency response compensation to spectrum and frequency response curves. It is expected that frequency response, i.e. response of microphone  that has to be compensated, is  in ASCII formatted textual file with name extension .MIC. Format of that file is as follows:

a) .MIC file contains lines of text

b) Lines that start with digits or dot character are expected to contains two numeric value separated with spaces or tab characters; first value is frequency in Hz and second value is magnitude of frequency response in dB. After these two values the line can contain any text up to the end of line. Usually, a point in the middle of the frequency range is chosen as reference 0dB value.

c) All other lines are treated as comment

d) You can edit .MIC file with any text editor (i.e. Notepad), but keep in mind that frequencies must be entered in sorted order.

Here you can download a sample of .MIC file for the microphone MB550.

Good source of cheap calibrated microphones is at http://lasip.hifi-selbstbau.de/.

You can define .MIC compensation files for other purposes. I.e. Peter de Jong has contibuted  .MIC files (RIAA_MIC.ZIP) to compensate for phone RIAA response.

Q: Why ARTA FR compensation does not use phase data?

A: ARTA is primarily used in acoustical measurements where system response is measured with microphone. FR compensation curve for microphone is usually obtained by comparing response with response of reference microphone.

There is no guarantee that any reference microphone has correct phase data, as there are just a few laboratories in the whole world that are capable to estimate microphone phase response, and it holds only for selected microphone design.

FR compensation is applied in windows that shows FR response magnitude, but it is not used in Impulse response time analysis (step response, ETC, energy decay, group delay and reverberation estimation). Keep in mind that in this case the analyzed response contains contribution of microphone response. For top class measurements we recommend use of IEC Class I microphones.

In classical acoustical measurements (spectrum, band analysis, SPL) we do not need phase information.

In electro-acoustical measurements we need correct phase response  when designing loudspeaker crossover. A further analysis will show that we do not need FR compensation phase data to get correct crossover design.

Let analyze two-way system response:

D1 = FR of first driver without FR compensation
D2 = FR of second driver without FR compensation
C1 = FR of first crossover filter
C2 = FR second crossover filter
FRC = FR compensation

The total response H, assuming both responses are measured with a same microphone, is:

  H = D1 FRC C1  +  D2 FRC C2

  H = FRC  (D1 C1 + D2 C2)

Three conclusions are important:

1) The phase of FR compensation does not change the magnitude of total response.

2) It is the same if we apply FR compensation on each driver’s response or on total response calculated without compensation.  In CAD systems that do a crossover response optimization it is better to apply compensation on each driver response as it is easier to define crossover target function, as maximally flat in pass-band.

3) The FR of all drivers should be measured with same microphone.

Q: When measuring loudspeaker impedance with LIMP I get a noisy measurement results. What is the problem?

A: When LIMP uses fast FFT method, with pink noise or multitone excitation, to estimate the loudspeaker impedance, the method is highly susceptible to environmental noise (remember that loudspeaker acts as microphone for environmental noise).

To eliminate the influence of noise:

a) Use the output channel or external amplifier which can drive the loudspeaker through a serial reference resistor in the range 10-100 ohms,
b) Adjust levels on soundcard input channels to be above -10dBFS.
c) Monitor the possible soundcard input channel overflow (input level 0dBFS)

If you use the soundcard line out channel for impedance measurements, then reference resistor must be at least 600 ohms. In that case it is normal to have "noisy" measurement results. You can improve measurement if you have quiet measurement environment and if you apply a large number of averaging.

It is highly recommended to use soundcard headphone output instead of line output and use serial resistor of 47 ohms.

Q: I have a soundcard that works in 16-bit and 24-bit modes, but in both cases I measure the same noise floor in spectrum analyzer mode. What I made wrong?

A: Nothing is wrong with your measurement. Many souncards that are declared as 24-bit soundcards have effective noise floor higher than some high quality 16-bit soundcards. In that case use soundcard in 16-bit mode to get more processing power.

Q: I have a soundcard that works in 24-bit mode but it not function properly in 32-bit mode.  Is my soundcard driver defective?

A: Probably your soundcard driver is OK. The 32-bit mode is alternative way of transferring 24-bit recorded data to computer, and not all soundcard drivers support this mode.

Q: My soundcard works properly in 24-bit and 32-bit mode. Shall I use 24-bit or 32-bit mode?

A: In this case it is recommended to use 32-bit mode as it uses less processing power.

Q: I've measured loudspeaker FR with ARTA in free space (outdoors).  I've got high noise at low frequencies. Help me solve this problem.

A: Measurements outdoors, in free field,  are usually affected with wind and possible traffic noise. In both cases you have large level of noise and distortion in measurement results.  For measurement of frequency response outdoors use STEPS, but if you must use ARTA, to get impulse response, use microphone wind protector and dual channel measurement with averaging. 

Q: I have tried to measure the frequency response with ARTA and impedance with LIMP using sampling rate 96kHz. I was hoping that I shall get response up to 46kHz, instead I got noise like response at frequencies above 22 kHz. What is wrong with my measurement setup?

A: This problem usually arises on some soundcards when cut-off frequency of the aliasing filter is fixed to 22 kHz. How is that possible? Some soundcards can do sample rate conversion by software and by hardware (i.e. Soundblasater Audigy ZS and almost all soundcards with Firewire interface). That soundcards have their own control panel (mixer) for adjustment of the hardware sampling rate. In cases when soundcard setup in ARTA and soundcard hardware setup have different sampling rates, the common sampling rate it is being adjusted in software. The problem is that this conversion does not change the cut-off frequency of the antialiasing filter. It is always close to the half the sampling rate that is fixed by hardware control-panel setup.