I have noticed a fundamental difference between the measurements of loudspeakers and the measurements of other components. When measuring loudspeakers, primarily acoustic measurements are used, whie for electronics, the measurements are exclusively electrical. While the electrical properties of, say, an amplifier are important to it's operation, I would think that the ultimate acoustic effects of any device are more important to the behavior of the device in the listener's system.
It would seem to be valuable to develop a range of "measurement reference" speakers, whose inherent acoustic behavior is well understood. An amplifier, preamp, source, or wire could then be inserted into the system, and a full spectrum of acoustic measurements taken. Software can normalize the results to the inherent behavior of the speakers, and the changes to the sound plotted. Obviously, several reference speakers with different characteristics and using different design methodologies should be used for these tests. After all, the sound energy in the room is what our ear/brain system decodes, not the electrical energy at some arbitrary node in the system.
Also, have you considered the use of long-integration-time narrowband analysis? There is probably considerable infomation that is averaged out of 1/3 octave measurements that could be revealed by FFT software using a 1/8 or 1/16 Hz (or smaller) vernier. As my user id indicates, I spent several years as a US Navy Sonar Technician, and I spent many hours observing the fine-scale structure of sound. Truly a fascinating subject. Sound is an amazingly tricky and intricate thing.
Thank you for making Stereophile happen, and for being available to the community for conversations both fair and foul.
FFT is only applicable for a 'linear' system. As you know
hearing is a non-linear function. The ear is almost linear in
the 20 - 500 Hz range. At 1 kHz and above the ears acts more
like a log detector.
You make a very valid point - about acoustical measurements
and the need for standards and a known reference. The measurning
and correlating what is heard to electrical measurements has yet
to be accepted as possible. Many mesurements are of little value
unless the measurements are 'tied' to a repeatable sonic difference.
I was, ex-Navy myself (Radar: ETR) was pulled into the audio-world
by a good neighbor, and became hooked on the intrigueing nature
of the audio problem: getting better fidelity.
As you mentioned the speaker is where the electrical meets the 'ear'. Until standards are setup and accepted; the measuring
and correlating 'processes' will stay in the realm of 'guessing'
what sounds 'good'.
The question about narrowband measurement was not really related to the question about electrical vs. acoustic measurments. Sorry if this was confusing. The only point of tangency is the larger question of the correlation between measured and perceived behavior of audio systems.
The reason that I proposed narrowband analysis was to investigate some behaviors that get little to no attention, specifically the full harmonic families developed by a component, and the fine-scale frequency stability of the unit under test.
Without going into details, since I'm not sure how much of the techniques are classified (I know the data are, but I don't know about the algorithms), if a component has a slight freqency instability, this shows up on a narrowband trace as a blurring of the discrete frequency components of the signal. If the measurement vernier is narrow enough, the signal can actually be seen wandering around the true frequency in real time.
With respect to harmonic development, if one takes a narrowband picture of the test tone and it''s harmonic family, and then compares it to a narrowband analysis of the same tone as reproduced by the unit under test, changes in the harmonic structure (relative ampiltude of harmonics, the width of the harmonic series, several others) should be observable.
> > I would think that the ultimate acoustic effects of any device are more important to the behavior of the device in the
listener's system. < <
This may be valid for turntables and tube preamps which are microphonic in some cases but certainly not for solid state amplifiers. Make the following experiment: Knock at a solid state
amplifier and measure the output of the amplifier or loudspeaker -
you won't measure anything (at least if the mechanical construction
of the amplifier is decent) so why should someone measure an amplifier
when it is exposed to sone waves generated by a loudspeaker which
sure are hundred times weaker than if you knock at the amplifier ?
I think you misunderstood my question, so let me clarify. What I mean by the acoustic effects of a piece of gear is the changes in the sound that is projected into the room by the speakers. This, obviously, is what we hear. Ears are not oscilloscopes. Since we cnanot hear electrical signals, the measurement of the changes an amplifier makes to the electrical signals passed through it are of limited value if one is trying to make a group of objective measurements which will correlate with subjective listening experience.
I was proposing that the device under test be inserted into a system with known acoustic properties, ie, we know a lot about the sound energy it projects, and then make measurements to see if and how the system's sonic behavior is different with the new device. Over time, this should build up a knowledge base that can be correlated to listening preferences and subjective descriptions of sound quality.
I was not proposing to measure the effect of the sound energy on the component under test, but rather the opposite.
I see. I've heard many people say that distortion measurements
of amplifiers are worthless if they are made with a load resistor
instead of a loudspeaker. As for decent solid-state amplifiers, this
is wrong. I've compared the behaviour of a standard power amplifier
with load resistor and with loudspeaker and in result, the amplifier
behaved in an identical manner - 0.013% THD, the distortion spectrum
was nearly identical in both cases with the deviations being close to the precision limit of the spectrum analyzer.
So I don't think solid state amplifiers with decent specs behave
different with different loudspeakers, excepted in some rare untypical cases (overload, clipping and the like).
Out of curiosity, did you measure the distortion by smapling the electric signal at the amplifier's speaker terminals, or the in-room sound with a microphone? Distortion in a system does not always sum in a simple fasion from node to node. It is posible for the distortion present in the signal going into the speakers to _decrease_ the speaker's distortion. While these measurements are harder to make, I think they could be very informative.
> > It is posible for the distortion present in the signal going
into the speakers to _decrease_ the speaker's distortion. < <
Even if this were true it wouldn't disprove my claim because
with and without speaker the power amplifier behaved in an identical
manner so no loudspeaker of this world can make different sone signals
out of the same electrical input signal.
There is a very good discussion of this, with the relevant mathematics, at:
Give it a look, it is quite possible for a speaker to be "linearized" by the presence of low-order distortions in the electrical signal applied to it's terminals. And again, the distortions in the sound signals arriving at your ear are the only ones that matter, since they're the only ones you can hear.
"While the electrical properties of, say, an amplifier are important to it's operation, I would think that the ultimate acoustic effects of any device are more important to the behavior of the device in the listener's system."
You can't because you're always at the mercy of the speaker/room effects. However, most designers will to some sort of listening tests, they just can't publish any spec that way becaue they are only repeatable for the environment they were measured in originally.
This post is made possible by the generous support of people like you and our sponsors: