Final Test Signal Versions
In addition to working out this spacing relationship, I also came up
with a spreadsheet that allows easy determination of the source of a particular
distortion, which when used in conjunction with the technique mentioned
above, allows you to exactly determine which two (or more) frequencies
were the source for that particular distortion product. Final Test Signal
Versions Here are the the final versions of the test signals I developed:
Phi 6 Spectral: 100, 261.8, 685.4, 1794.4, 4697.9, 12299 Hz. Individual
level of each tone is at -15.6 dB.
Phi 12 Revised Spectral: 100, 122.0, 261.8, 348.2, 685.4, 987.0, 1794.0,
2870.4, 4697.9, 6765.0, 12299, and 16358 Hz. Individual level of each tone
is at -21.6 dB.
Phi Low-High Split Band Spectral: 100, 116.18, 134.98, 156.80, 182.19
and 4697.9, 5927.8, 7479.7, 9437.9, 11909 Hz. Individual level of each
tone is at -20 dB.
Phi Low-Mid Split Band Spectral: 100, 116.18, 134.98, 156.80, 182.19
and 986.99, 1245.4, 1571.4, 1982.8, 2502.0 Hz. Individual level of each
tone is at -20 dB.
Phi Tri-Band Spectral: 100, 116.18, 134.98, 156.80, and 986.99, 1245.4,
1571.4, 1982.8, and 6764.9, 7618.5, 8579.8, 9662.5 Hz. Individual level
of each tone is at -21.6 dB.
For my tests, I generated the tones in the digital domain, burned them
to a CD-R, and used a squeaky clean CD player as the test signal source.
Analysis was performed on a 16 bit FFT based system, using 8K point FFT's.
This results in 4096 real data points within the audio band, resulting
in an approximately 5 Hz bandwidth for each FFT frequency bin. Other instrumentation
for FFT spectrum analysis is available that is capable of more than 16
bits, with resolution up to 64k FFT's, which makes the FFT bin width about
0.6Hz wide. Some of these cost a fortune, and others as little as a current
PC with a studio grade sound card and the appropriate software.
During the course of testing, it was found that CD players had measurable
amounts of IM distortion, and all measured different using these test signals.
So much for all CDP's are alike. I used a CDP that had near source signal
levels of output distortion so as not to compromise the dynamic range of
my measurements. To my knowledge, no one has published anything about these
IM distortion differences, probably because they do not tend to show up
with traditional two-tone IM tests as well. One player showed lower SMPTE
IM and harmonic distortion levels, yet had significantly higher levels
of Phi Spectral IM distortion than another player. Many other measurements
were made as well, a very high quality cassette deck was compared to a
mini-disc deck; digital electronics were tested and compared, raw loudspeakers
and loudspeaker systems were tested and compared. Each one of these audio
components reacted in their own specific fashion to the test signal, and
in most cases, levels of the traditional IM and harmonic distortions were
inconclusively low between the various component test pairings. Sometimes
one unit would look significantly better using traditional measurements,
yet end up measuring much worse using the multitone. The correlation between
which unit sounded good, and which didn't followed the Phi Spectral measurements
almost exactly.
Of course, there was one other audio component that was measured: speaker
cables.
Check out other Spectrals
A quick note: one of the current spectral contamination measurements
is currently being used at Mix magazine for studio monitor reviews.
Check it out at : http://www.mixmag.com/ and click on the "Tannoy Reveal
2-Way Reference Monitors" link, and go down to Fig. 4. Note the number
of tones at even frequency intervals.
One other thing about my version of Phi Spectral signals. They have
a very high crest factor, on the same order as the amount of individual
tone reduction factor, or a little less. That is about 15 to 21 dB of crest
factor. This is also more like real music, instead of a single sine wave
with it's 3 dB crest factor.
Testing Speaker Cables with the PHI Spectral
Now that I have covered a little bit about the test signal itself,
I will go into the actual cable tests. Explanation of Measurements In measuring
cables, I hypothesized that for bi-wired speaker cables, the big difference
would be in the current flow, as the impedance’s of the crossover would
create a preferential situation with regard to the flow of current in the
woofer cable vs. the tweeter cable. To check this, I used a Pearson Electronics
Model 411 current sensor (http://www.pearsonelectronics.com/) to measure
and analyze the current flow in the speaker cables. No other method of
measurement will yield the same information, as a series resistor to convert
the current to a voltage will introduce it's own resistance, tending to
swamp any speaker cable differences out. Voltage measurements will not
provide the same information, unless you are using a resistor load, without
crossovers and actual speakers involved. I initially tested a single 12
gauge zip cord against my cross-connected 89259 in bi-wire mode, that is
what was published in my AES preprint. The deadline for that paper was
June 15th, since then, I have tested various combinations of gauges, various
retail audiophile cables, all of which have been tested in bi-wire and
single cable modes. While I have found differences between single speaker
cables of 3 and 4 dB between the various cables in single wire mode, which
I believe to be repeatable and consistent, but these small changes would
no doubt be questioned endlessly by certain parties, so I make no definite
claims for these results, other than to relate them for future study. Perhaps
a 20 or 24 bit analyzer with the appropriate resolution can quantify these
cable differences more clearly. Gauge did not strongly influence the distortion
levels, the difference due to smaller and larger gauges were well within
the measurement error limits, while the differences between audiophile
cables and zip cord were generally higher, although not above a level beyond
any doubt, as I have indicated above.
Analysis
When I obtained the distortion plots for the bi-wire vs. single wire
conditions, I was careful to analyze the results, so as to fully understand
what was going on. At first, I felt that the reductions might be due to
the simple aspect of the crossover impedance’s, and that perhaps the distortion
reductions followed the crossover curves. Close examination revealed that
not only does the distortion plot NOT exactly follow the curve of the crossover
current flow (this can be examined by looking at the web site, page7 vs.
page 8, and comparing specific frequencies. I posted a set of example frequencies
and discrepancies earlier in response to concerns about the crossover roll
off being the only valid effect here.) but for the Phi Spectral signal,
the crossover roll off DOES NOT DIRECTLY AFFECT THE DISTORTION PRODUCTS
AT THAT PARTICULAR FREQUENCY because they are not generated by any frequencies
at that location in the frequency domain, but by the bundle of tones at
the low end and at the high end! The distortion products would not be affected
by the current flow (or voltage) frequency response at that point on the
graph!!!! In order to make this perfectly clear, let me give a concrete
example. For instance, the distortion spike at just above 1 kHz in Fig.
Z on page8 at the web site. It is about 25 to 27 dB higher for the single
cable than it is for the bi-wired tweeter cable. Now even if we completely
ignore the fact that this distortion product derives from a pair of original
tones that are either in the 100-200 Hz band, or in the 4.7 to 11 kHz band,
and pretend that it is tied precisely to that point in the frequency domain,
there is something inconsistent with the accusations made: the current
flow for just the tweeter cable is only down 15 dB from the peak current
flow, not 25 dB If we look at the current flow for the single wire cable,
it is at a level that is only 4 dB higher, not 25 dB. I will post this
graph, to be called Fig. W, on my web site at page9.
CONT'D
All applicable copyright laws apply, all rights reserved, except transmission by USENET and like facilities granted. Any use or inclusion in print or other media are specifically prohibited. The informational content is not warrantied in any way or form, and any use of said content are at the reader's own risk, the author shall not be held responsible in any way for any damages or injuries arising from the content of this web site. Common safety practices are encouraged at all times. Do not fold, spindle or mutilate.