Measuring Audio -- Part Two
Dr. Floyd Toole (top) and Errol J. Byers -- photos
originally taken about 30 years ago when this interview was first published.
|
This month, we present the second
part of a roundtable discussion that I conducted about 30 years ago with Dr. Floyd Toole
and Errol J. Byers on the challenges of measuring audio equipment. These two experts were
responsible for the actual testing that lay behind the audio evaluation program of AudioScene
Canada magazine, of which I was one of the editors. For the background to the
discussion, please see "Measuring Audio -- Part One."
We concluded the first installment with a discussion of
measuring speakers.
Ian Masters: Transducers are
admittedly tricky. Is distortion in electronic components any easier to pin down?
Errol Byers: Probably not, although the problems are
different. Fortunately, the levels of distortion in most "quality" electronic
components are low enough that they're not of as serious concern as in some other areas.
This is particularly true of amplifiers. There is room for
improvement in the tape recording process, and, I am sure, in disc recording processes.
But amplifiers are quite good at the moment, I think.
IM: But can you set up a testing
environment that can be repeated in the user's own living room?
EB: No. The principal difference is in the type of
load applied to the amplifier. A speaker is not, in fact, an 8-ohm resistive load like the
sort used for most amplifier testing.
If you want to measure distortion of a 20kHz signal, you
would have a load that is resistive to at least five times that frequency. If you want to
test an amplifier above the audio spectrum, as is becoming popular, then you may want a
load that is resistive up to a megahertz. I have yet to see a speaker that comes anywhere
near being resistive at a megahertz!
Floyd Toole: Or even close to 8 ohms over a wide
frequency range.
EB: That's right. The conditions are not exact, so
again it boils down to a comparison. Unfortunately, in the case of amplifiers at least,
this comparison is not always valid because some amplifiers perform equally well with an
8-ohm resistive load and with a speaker, while another may perform well with the 8-ohm
load but not very well with a speaker.
IM: Is the corollary true -- that some
amps perform well with a speaker, but not so well into the 8-ohm resistive load?
EB: Yes, but I think it's usually true that an
amplifier that performs badly into a resistive load will perform badly into a speaker. The
opposite is not true -- if it performs well into the resistive load, that doesn't
guarantee it will perform well with a speaker.
Another thing to keep in mind is that an amplifier's
distortion is usually very low by comparison with a speaker's distortion level, so it may
not be heard.
IM: Do the new "super amps"
[with more than 100 watts per channel -- a lot of power at the time] represent any
additional problems?
EB: Yes. With the trend to smaller and
less-efficient speakers, and the trend to higher-power amplifiers, protection circuitry
has been added to prevent wiping out a very expensive portion of the amplifier. These can
be voltage limiters, current limiters, or a combination of the two; and, sometimes these
circuits can be falsely activated. Or, occasionally, phase shift between voltage and
current in a reactive load can cause problems with the protective devices.
IM: And these would occur "in the
field," and not just in the laboratory?
FT: It's probably more likely in the field, since
then a speaker will be used, and it has the sort of complex impedance that we are talking
about.
IM: Getting back to more general
topics, how do you approach a piece of equipment for the first time? What is the first
basic measurement?
FT: Well, the device itself, to a considerable
extent, determines the kind of measurements you can make. For example, the harmonic
distortion measurement, where you measure the level of the harmonics (which go upwards in
frequency from the fundamental), requires that the device exhibit usable bandwidth or
frequency response comfortably beyond the frequency at which you are measuring the
distortion.
This is not always the case. For example, in the case of a
tape recorder, we know that few of them perform adequately beyond 20kHz at normal tape
speeds. So, logically, this means that, at any frequency above 10kHz, even the second
harmonic is lost, so a harmonic distortion measurement is of little value. You are then
forced to look for alternative measures of non-linearity at high frequency -- particularly
a measure that will produce the distortion products below the test frequencies. This is
where intermodulation distortion techniques come in.
There is one test that has existed for many years, but is
very little used: the CCIF Test, where two frequencies closely spaced are applied to a
device and the distortion products that turn up below the test frequencies measured --
usually just the difference frequency itself. You can run this pair of frequencies above
the 10kHz limit that I mentioned -- right up, perhaps, even to 20kHz -- and not be
concerned that the device rolls off dramatically beyond that, because what you're
measuring is downwards in frequency from that.
EB: The same thing occurs in FM tuners, where there
has to be a filter to eliminate the 19kHz pilot transmitted along with the FM signal, and
so the frequency response usually falls off above 15kHz. So if you try to measure
distortion much above about 5kHz, you lose the distortion products.
IM: But if the distortion products are
eliminated this way, are they likely to cause problems?
FT: To go back to something I mentioned earlier,
these methods of measuring distortion are really methods of measuring deviations from the
ideal -- the non-linearity of a device. Just as the device can produce intermodulation
products from the two closely spaced frequencies I mentioned, it will produce
intermodulation products from any combination of frequencies that appear in musical
material.
For example, a simple sound like a cymbal clash -- which is
really rather complex, but it's a simple musical instrument, and a very common one --
produces a lot of harmonically and non-harmonically related components in the
high-frequency range, and they tend to be at rather high levels. It is possible with this
kind of sound actually to hear intermodulation components separately in some devices.
You may not always hear these, though, because of a
convenient thing that happens psychoacoustically or psychologically, called
"masking." This, put most simply, means that the music itself may well mask the
distortion products. This is perhaps more true of these high-frequency intermodulation
products than anything else. The cymbals may be banging away, producing considerable
quantities of IM distortion; but if the string section is playing at the same time,
producing a lot of wanted sound down in the middle and lower frequencies, that wanted
sound may in fact mask the unwanted sound.
IM: So this would be a case where you
might be able to measure distortion, but not be able to hear its effects.
FT: Yes, you may not always be aware of it. There
may, however, be isolated instances where, for some reason, there isn't enough masking
sound and you can hear these distortions.
EB: To measure a system, the ideal would be to put a
great collection of music through it, and pick out what is objectionable in the music. But
this would be a practically impossible task, which is why we use the simple measurements.
But there's a vast difference between putting a single tone -- or a pair of tones --
through a system, and using the sort of complex signal you would get from a symphony
orchestra.
FT: There is a strong tendency to make measurements
that are easy to make -- to produce numbers and graphs. We are indeed measuring important
aspects of performance, but we are not necessarily measuring them in the optimum fashion.
Having made those measurements to the best of our abilities, we are at a loss precisely to
relate them to the audible quality of a piece of equipment's performance.
IM: So really, the main function of
measurements of this sort is to relate one piece of equipment to another.
FT: Yes, it's a relative evaluation. We can say
that, in a general sense, lowest distortion is best -- a sort of "motherhood"
statement -- but we do not know what level of distortion is acceptable.
IM: If it's the easy measurements that
are made, what sort of test should be performed that isn't?
EB: Well, a good example is the measurement of
turntable suspensions. When you buy a turntable, or just look at the manufacturer's
specifications or at test reports, you're never given any indication of whether or not you
can set it up in a living room and then dance in the same room. But such a test could be
done with reasonable ease.
IM: By some kind of index? Such as
"Yes, you can drop an eight-pound ball on the floor six feet away" -- that sort
of thing?
EB: Something of that nature. Or you could plot a
frequency response curve of the turntable suspension by putting it on a vibration table.
The immediate difficulty is that you don't know what level of vibration to apply to the
turntable to approximate what you might encounter in a house.
But there is a fair amount of information about building
vibration, so you could certainly get a sample of what frequencies and levels are there in
typical houses. You could then subject the turntable to that sort of vibration, leaving
the pickup on a stationary record, and see what sort of output you get.
Since this would be a new test, there would, of course,
have to be a period in which an interpretation of the results could be worked out. Some
work would have to be done to find out, after you have produced a frequency response
curve, whether it means you can dance in your living room.
Another sort of test, as you suggested, would be a
weight-dropping test -- but you would have to simulate a standard floor. This would really
be a sort of "go/no-go" test -- if the pickup jumped out of the groove, the
suspension is no good; if it didn't, it's all right.
FT: Less severe, but equally annoying, is the
problem of acoustic feedback, which can occur with much less vibration than actual dancing
on the floor. You can get a howling through your system at low -- and sometimes even
medium -- frequencies. This is a function of the mechanical coupling between the phono
cartridge and the wall or floor -- whichever is the contact point for the turntable base.
It is determined by the vibration transmission properties of the turntable suspension --
its frequency response, if you like.
EB: Feedback can have rather drastic results, too.
If you happen to turn up the volume when the turntable's resonant frequency is being fed
through the system, you get a buildup of low frequency, and this eventually is
self-destructive.
IM: Are there any overlooked tests on
tape recorders?
EB: In recording, where there is a high-frequency
bias on the tape, and noise levels are relatively high, there is a great opportunity for
intermodulation distortion between the high musical frequencies and the bias. If this is
combined with a stereo FM tuner that also adds a hit of 19kHz and 38kHz signal from the FM
multiplex system, the opportunity is glorious for IM distortion products.
We don't really know how objectionable these products may
be in a particular situation. But the measurement itself would be quite simple.
IM: Are IM distortion measurements not
generally made on tape recorders?
EB: Sometimes -- but not with the same regularity as
with amplifiers.
FT: The SMPTE test is usually done, isn't it?
EB: Yes. This mixes a 7kHz and a 60Hz signal, and,
in effect, measures the sidebands generated of 7kHz ±60Hz. What we're looking for in tape
recorders is the noise intermodulation products, and these may be over the whole spectrum.
FT: With the relative levels of those two signals
you are, more than anything else, measuring the degree of non-linearity at 60Hz.
EB: Yes, that's true.
FT: This is true of amplifiers as well. Whenever you
see this kind of IM distortion measurement -- and you see it frequently in many
specifications and test reports -- it is really of rather restricted value in assessing
the overall distortion performance of the device.
IM: So this distortion figure is only
one of a family?
FT: Yes. It's a measurement that should be made, but
it's only one measurement, by no means all-encompassing.
IM: So a piece of equipment could look
good from this test, and still be terrible.
FT: Oh, indeed yes. It could have low distortion at
low frequencies, and come out charmingly in this test, but still have incredible amounts
of distortion at high frequencies, and those won't show up at all. You'll just hear it.
IM: What you're saying, then, is that
some valuable tests are being done, but they are too narrow.
FT: Yes. Most of these things have some historical
foundation, and they all began with a specific problem in mind. But now they have come to
have almost universal application; and it is definitely misleading to think that, because
of their universal application, they are the be-all and end-all of measurements.
IM: One area of testing that seems to
be receiving some attention recently is the phono cartridge. Are there any serious
omissions here?
FT: With phono cartridges, the performance of the
cartridge is inseparable from the signal on the disc, the material of the disc, and so on
-- and this, of course, is traceable back to the quality of the recording apparatus.
With cartridges, we are in the strange position of being
able to make measurements that produce reproducible numbers from a particular test
recording; but we are unable actually to say that this is, in fact, an absolute measure of
the distortion of the device -- or even of the recording medium. It is merely a measure of
the performance of the device in that particular groove of that particular record. If you
change the recording, or the source of the recording, or change the disc material, you
change the level of distortion dramatically.
...Ian G. Masters
ian@mastersonaudio.com
|
