Harpia Acoustics

Measurements

Neglecting measurements in evaluation of loudspeaker quality has become the norm in recent times. We are going to explain this situation.

Only the frequency response and the impedance plot taken into consideration as "measurements".

There is no doubt that frequency response should be as linear as possible. But linear response does not guarantee good - or even acceptable - sound, for example when it is combined with poor impulse response. In practical reality loudspeakers with some moderate peaks and dips often sound much better than others with perfectly linear response curve.

On the other hand, loudspeakers with seriously distorted frequency response (for example -20dB in the middle of the passband) can be discarded without audition.

Directional characteristics cannot be analyzed without the knowledge of room acoustics. In rooms without acoustic treatment or rooms with poorly designed acoustic treatment wide polar pattern may increase the intensity of reflections and degrade sound.

Impedance plot with wide variation and very abrupt changes is an evidence of poorly designed high order filters - such loudspeakers have poor quality sound and present a tough load for amplifiers.

When impedance curve changes gently and its variation is moderate, then there is a chance that the loudspeaker is good, but still there is a possibility that with poorly designed linearization it can present "dead" sound. Generally impedance can be linearized even in worst loudspeakers, usually making them even worse.

Due to existance of several measuring techniques, various magazines can present different sets of measurements for the same speaker..

There are 3 main methods for measurement of frequency response:

1. Anechoic chamber
This is the most refined (and expensive) method.

Advantages:
- possibility to measure at the same distance at which loudspeakers are typically used: 2 - 5 m
- results practically independent of the measurement system (as long as the measurement equipment is of good quality) and the operator

Disadvantages:
- gigantic cost of building a large chamber
- 9 dB difference in the low bass between anechoic measurement and a measurement in the corner of the room where loudspeaker is used. If anechoic measurement shows perfectly linear extension to 30Hz, the same speaker placed in a room corner will have response elevated by 9dB at 30Hz.

2. Nearfield sinusoidal + MLS

Advantages:
- low cost (computer, a suitable sound card, microphone)
- possibility to make measuremets in small rooms.

Disadvantages:
- results depend on the measurement system (errors in software algorithms) and on the operator who joins all the plots (reflex port, bass from the woofer, MLS plot)
- nearfield bass measurement does not show the influence of baffle width on the speaker's radiation, so there is an error, although the plot measured this way indicates more elevated bass in comparison with anechoic measurement.
- plots are merged by the operator who measures reflex port area, cone area and manually adjusts the levels of various plots (this can lead to errors and mistakes).
- software packages have smoothing options, so when the operator has positive attitude towards a speaker he can smooth the plot and if he has negative attitude he can leave the plot more ragged
- measurements are made at distances limited to about 1,5m on the condition that floor between speaker and microphone is damped.

The measurements of Sonus Faber Stradivari Homage published in Stereophile and Audio (Polish) can serve as a good example. The difference in bass range is up to 5dB "in favour" of Stereophile.

3. Third octave noise measurement

This is the most reliable method, unfortunately influenced by acoustic properties of the room used for measurements.
The good thing about this method is that measurment corresponds well with listening experience in the room.
On the negative side, the measured curve is influenced by room acoustics. Measurement made in an acoustically unsuitable room tells very little about the speaker, and exposes room's modes.

Illegible plots (plots too small, poor descriptions, quoting distortion in dB)

Quite often plots published in magazines are so small, that a maginifying glass is necessary to read them.
Presenting measurement results in strange ways is another problem. Such forms of presentation are supposed to make measurements easier to understand for amateurs but in reality make them incomprehensible even for specialists.
Distortion figures are very rarely quoted, and they are practically never published in an easy to understand way. Usually it is necessary to calculate the difference in dB between two plots, and convert those values to percentage points. Most certainly 99,9% of readers are not able to handle this task.

Little attention paid to step response.
This plot is seldom published (Stereophile, German Stereo).

In general the fewer times plot goes up and down the better. But good step response itself is not sufficient. The evidence of this is given by loudspeakers based on wideband transducers. Linear distortion - in other words coloration - is the reason why such speakers are not a good solution.

Summing it up

Evaluation of loudspeakers based on measurements (even complete ones) is very difficult, sometimes even impossible. So maybe it is better to just listen?
On the other hand designing loudspekers which perform well in listening tests without measurements and high quality simulation software is not possible.



	Measurements Neglecting measurements in evaluation of loudspeaker quality has become the norm in recent times. We are going to explain this situation. Only the frequency response and the impedance plot taken into consideration as "measurements". There is no doubt that frequency response should be as linear as possible. But linear response does not guarantee good - or even acceptable - sound, for example when it is combined with poor impulse response. In practical reality loudspeakers with some moderate peaks and dips often sound much better than others with perfectly linear response curve. On the other hand, loudspeakers with seriously distorted frequency response (for example -20dB in the middle of the passband) can be discarded without audition. Directional characteristics cannot be analyzed without the knowledge of room acoustics. In rooms without acoustic treatment or rooms with poorly designed acoustic treatment wide polar pattern may increase the intensity of reflections and degrade sound. Impedance plot with wide variation and very abrupt changes is an evidence of poorly designed high order filters - such loudspeakers have poor quality sound and present a tough load for amplifiers. When impedance curve changes gently and its variation is moderate, then there is a chance that the loudspeaker is good, but still there is a possibility that with poorly designed linearization it can present "dead" sound. Generally impedance can be linearized even in worst loudspeakers, usually making them even worse. Due to existance of several measuring techniques, various magazines can present different sets of measurements for the same speaker.. There are 3 main methods for measurement of frequency response: 1. Anechoic chamber This is the most refined (and expensive) method. Advantages: - possibility to measure at the same distance at which loudspeakers are typically used: 2 - 5 m - results practically independent of the measurement system (as long as the measurement equipment is of good quality) and the operator Disadvantages: - gigantic cost of building a large chamber - 9 dB difference in the low bass between anechoic measurement and a measurement in the corner of the room where loudspeaker is used. If anechoic measurement shows perfectly linear extension to 30Hz, the same speaker placed in a room corner will have response elevated by 9dB at 30Hz. 2. Nearfield sinusoidal + MLS Advantages: - low cost (computer, a suitable sound card, microphone) - possibility to make measuremets in small rooms. Disadvantages: - results depend on the measurement system (errors in software algorithms) and on the operator who joins all the plots (reflex port, bass from the woofer, MLS plot) - nearfield bass measurement does not show the influence of baffle width on the speaker's radiation, so there is an error, although the plot measured this way indicates more elevated bass in comparison with anechoic measurement. - plots are merged by the operator who measures reflex port area, cone area and manually adjusts the levels of various plots (this can lead to errors and mistakes). - software packages have smoothing options, so when the operator has positive attitude towards a speaker he can smooth the plot and if he has negative attitude he can leave the plot more ragged - measurements are made at distances limited to about 1,5m on the condition that floor between speaker and microphone is damped. The measurements of Sonus Faber Stradivari Homage published in Stereophile and Audio (Polish) can serve as a good example. The difference in bass range is up to 5dB "in favour" of Stereophile. 3. Third octave noise measurement This is the most reliable method, unfortunately influenced by acoustic properties of the room used for measurements. The good thing about this method is that measurment corresponds well with listening experience in the room. On the negative side, the measured curve is influenced by room acoustics. Measurement made in an acoustically unsuitable room tells very little about the speaker, and exposes room's modes. Illegible plots (plots too small, poor descriptions, quoting distortion in dB) Quite often plots published in magazines are so small, that a maginifying glass is necessary to read them. Presenting measurement results in strange ways is another problem. Such forms of presentation are supposed to make measurements easier to understand for amateurs but in reality make them incomprehensible even for specialists. Distortion figures are very rarely quoted, and they are practically never published in an easy to understand way. Usually it is necessary to calculate the difference in dB between two plots, and convert those values to percentage points. Most certainly 99,9% of readers are not able to handle this task. Little attention paid to step response. This plot is seldom published (Stereophile, German Stereo). In general the fewer times plot goes up and down the better. But good step response itself is not sufficient. The evidence of this is given by loudspeakers based on wideband transducers. Linear distortion - in other words coloration - is the reason why such speakers are not a good solution. Summing it up Evaluation of loudspeakers based on measurements (even complete ones) is very difficult, sometimes even impossible. So maybe it is better to just listen? On the other hand designing loudspekers which perform well in listening tests without measurements and high quality simulation software is not possible.
© Harpia Acoustics 2007 :: design: igor herzyk