Loudspeaker & Room characterization  

+++  Test & Measurement  +++  Phantom center  +++  Case studies  +++ 

Phantom stereo-center-speaker

A solid center image is a fundamental requirement for a stereo loudspeaker setup. Without it one cannot expect to hear a spatially detailed phantom scene on program material. The existence and perceptual location of the phantom center can be determined by listening to a sequence of  4-cycle Blackman envelope bursts at different frequencies. Table 2 provides details for sound files with bursts in 1/3rd-octave steps to be used as test signals. 

File 04 is a fast sequence of bursts going down and up in frequency. Frequencies in File 05 go down fast in 100 ms step increments and up again slowly in 500 ms steps, which may change the perceptual locus described by the phantom signal. The sequence ends with a 100 Hz burst followed by a 1000 Hz burst.

With the LX521 in my living room the changing frequencies trace out a vertical line starting out with low frequencies at the bottom. The image for the last three high frequency bursts bends towards the right loudspeaker. When I turn around, with the speakers behind me, the highest frequencies move to the left speaker. This indicates that my right ear dominates and that I have some high frequency hearing loss in the left ear. I cannot say whether or to what degree may brain compensates for this when listening to program material. I must assume that it affects my perception of some elements in an aural scene.

Investigate also how a different placement of the speakers in the room affects the phantom center.

File 06 consists of cosine envelope bursts of 100 ms duration for each of the frequencies, which increase in 1/2-octave steps at 500 ms intervals. The long duration of the bursts defeats the formation of a center image. Burst #18 is again short in duration. Compare its audibility to that of burst #17. I cannot hear #17.

File 07 allows to test for any audible differences between a 4-cycle Blackman burst and a 4-cycle Cosine envelope burst. You can also view the differences by recording the sound file with ARTA.

Table             Download Burst files: 04__Phantom_12800-100-12800Hz_100ms.wav 05__Phantom_12800-100Hz_100ms_100-12800Hz_500ms.wav 06__Phantom_50-12800Hz_100msBurst_500ms.wav Play the files in Repeat mode. Note the location of the stereo center phantom as the burst frequency changes.  07__200HzBlackman_200HzCosine_4cycles.wave Play the file in Repeat mode. See if you can hear a difference between the two burst envelopes.

You can check the solidity of the center image with a voice recording and the left or right shift of the image when moving your position by playing these sound files:

49_female_english.flac
50_male_english.flac
51_female_french.flac
52_male_french.flac
53_female_german.flac
54_ male_german.flac

The files are excerpts from the EBU SQAM CD.

A brief history of burst tests

The BBC used tonebursts with rectangular envelope in the late 1950's. I heard it was Harwood who actually made paper cutouts of bursts at different frequencies and set them up next to each other to observe the burst decays versus frequencies. This idea was picked up later when FFT instrumentation became available to generate windowed Cumulative Spectral Decay plots (Mike Berman, Laurie Fincham) from the impulse response of a loudspeaker. 
Rectangular bursts have many artifacts primarily due to low frequency components in their spectrum. Therefore I used Shaped Tonebursts . I was interested to determine the frequency response of a loudspeaker from the peak of the burst response and to hear the onset of distortion as the burst amplitude is increased. The burst decay showed hidden resonances at a time when I had no access to modern FFT analysis. 

Don Keele used shaped bursts to determine the maximum linear output SPL from the loudspeakers, which he reviewed for AUDIO magazine. Shaped burst signals were included in the ALMA loudspeaker test CD. Don proposed a swept burst signal, going down in frequency continuously. Marshall Kay wrote a BASIC program to generate the signal with an HP arbitrary waveform generator. MLSSA was used to look at the signal for changes in the modulation envelope. Nothing useful came of it. For a long time ARTA did not have oscilloscope capability and could not replace our aging PC/MLSSA setups. But eventually they added it on our urging. I still have problems with the trigger.

More recently Rob Collins picked up the idea of swept burst tests and developed the mathematical expression for programming GoldWave to generate a sequence of 4-cycle Blackman envelope bursts, which step with 50 ms spacing between 1/3rd octave frequencies from 12.8 kHz down to 100 Hz. This sequence became the basis for the phantom test signals above. The mathematical expression is several pages long and not easily modified. I have resorted to the expedient method of generating a catalog of single frequency burst signals with cosine and Blackman envelopes and then to copy and paste what I need along the time axis of a blank GoldWave file. The dead time between bursts is significant perceptually and also for viewing reflection decay in a room. 

Under the heading of Ping Tests I describe the use of burst signals for room acoustic tests, for phantom source localization, for acoustic polarity determination and for driver distortion tests.

**********************************************************************************************************