Jon Risch's Web Site
A NEW CLASS OF IN-BAND MULTITONE TEST SIGNALS
Jon M. Risch, Peavey Electronics Corp., Meridian, MS. USA
Copyright 1998
Presented at the 105th Convention of the Audio Engineering Society
1998 September 26-29th, San Francisco, California as preprint #4803
CONT'D
5.0 Actual Measurement Results
Unless otherwise noted, the electrical tests all had their level adjusted
with a mixer to bring the signal up to the same approximate input level to
the FFT spectrum analyzer, so as to maximize dynamic range and the display
of potential distortion products. The mixer and complete signal playback
system used as the source for the tests were checked to assure that the distortion
components arising from these components did not add to, or limit the measured
performance of any of the tests.
5.1 Electrical, Record/Playback Devices Measured
Several CD players were checked with the new signals, the professional
balanced output player used as the signal source worked well, and a top of
the line consumer CD player had an even lower noise and distortion floor,
see Figure 20. A portable CD player with a line out showed good results
too, see Figure 21, the gray spectrum.
Headphone output jacks on portables were another story, as depicted in
the black graph of Figure 21. This is the same player, the lower curve
out of the line level jack, and the curve showing a higher level of distortion
and noise being from the headphone out jack.
Several CD players measured had problems when the recorded signal had peaks
at digital 0 dBFS, where they exhibited symptoms of clipping/compression.
Figure 22 shows this effect to a moderate degree. Figure 23 depicts
the difference between 0 dBFS recorded levels, and -6 dBFS recorded levels
on another portable player. It can be seen that a reduction of recorded
level by 6 dB has reduced distortion by over 20 dB. Gain has been added
via a mixer to keep the input levels to the FFT analyzer the same, the only
difference is the level recorded on the CD.
A high quality consumer cassette deck was examined via a record/playback
of the test signal, using Dolby B noise reduction. The source was the
high quality CD player shown in Figure 20. Figures 24, 25, 26 and 27
show the response of the cassette deck when the test signals were recorded
at 0 VU levels, accounting for pre-emphasis. Note the presence of distortion
and modulation components at fairly high levels. The high frequency
content of the test signals in Figures 24, 25 and 27 prevents the normal
advantage the Dolby B high frequency noise reduction provides.
A consumer Mini Disc deck was tested, and the results depicted in Figures
28, 29, 30 and 31. In Figure 28, the Phi12r spectral signal, the noise
and distortion in-between the signals reaches levels as high as with the
cassette deck. Note the generally lower levels of distortion products
and noise in between the signal bands compared to the cassette deck with
the split band signals. One of the original split band spectral test
signals did show a high frequency roll-off with the band containing frequencies
at 100 -182 Hz plus 10 - 18.2 kHz. The last tone in the high frequency
band was noticeably attenuated, in contrast to the high quality consumer
cassette deck. See Figure 32.
A digitally based crossover and filter system was tested using the Phi12r
spectral test signal, shown in Figure 33. The test signal passed through
the A to D and the D to A stages without any digital filtering or EQ operations
being performed on the signal. Note the rise in the noise floor of
over 10 dB, and a few very low level products, compared to the source CD
player. Interestingly enough, once a filtering algorithm was
engaged, the noise floor went down some, as seen in Figure 34, yet it is
still higher than the source.
As can be seen from the test measurements, the new signal is effective
in testing electronics, digital systems, and recording systems. I have
no doubt that it would be useful for the testing of various transmission
systems as well, especially transmission systems using companding.
However, access to such systems was limited, and the author was unable to
present measurement results within this paper.
5.2 Electroacoustic Loudspeaker Tests
The new test signal proved highly effective in measuring loudspeakers,
as they tend to have a much higher level of intermodulation and crossmodulation
than electronics and similar systems. Changes in the distortion content
with changes in drive level were easily noted. All loudspeaker measurements
were taken in an anechoic chamber, at 1M, unless otherwise noted.
Figure 35 shows the distortion for a commercial plastic injection molded
15" based two-way system with a 1" throat compression driver (called Speaker
A), at one watt RMS drive levels for the Phi6 test signal. Note the
distortion products clearly visible in the middle/high range. Figure
36 is this same speaker at 10 watts RMS drive level, and the distortion has
risen considerably. Figure 37 depicts this same system with the Phi12r
test signal. The higher density of potential distortion products allows
a clearer picture of where the speaker is running into trouble. Contrast
this with the overlay of the same loudspeaker tested with all the different
original split band spectral signals that were explored (Figures 4 through
9) , in Figure 38. An obvious problem can be seen centered around 1
kHz, and again at 4 kHz. With only one measurement, the Phi12r reveals
substantially the same clues of trouble. The tri-band spectral also
provides a similar level of information with one measurement.
Conventional harmonic distortion plots do not provide much of an indication
that this speaker has a problem. See Figure 39. The levels of
harmonic distortion are not unusually high, and do not predict that this
speaker will sound fuzzy or less clear than another. Further spectral
signal test measurements at 10 watts are shown in Figures 40, 41, and 42.
To provide a reference point, a prototype plastic injection molded 15"
based two way with a 1" compression driver (Speaker B) is compared.
In Figure 43, it can be seen that speaker B has 10, and up to 15 dB less
distortion than speaker A. The harmonic distortion plots for speaker
B are not radically different than those for speaker A, yet the intermodulation
and crossmodulation is substantially lower. Figures 44, 45, and 46
show the comparison with the split band spectral signals.
A 2-way nearfield studio monitor prototype using a 6 1/2" woofer,
and a 1" titanium dome tweeter is tested at 1W RMS in Figures 47, 48, and
49. The speakers are measured at 1/2M to increase the signal to noise
ratio, and attempt to show the very low spectral contamination of the unit.
The input level is consistent with intended usage, and it should be noted
that a 1W RMS signal level has peaks of close to 100W with these multitone
spectral signals! This is excellent performance for a loudspeaker system
by any standard, and provides an interesting contrast with the sound reinforcement
speakers measured above (Speakers A and B).
Individual drivers can be tested using these signals. An inexpensive
stamped frame woofer is tested in Figures 50 and 51. The test signal
is input to the woofer full range. Measurements of a compression driver
tweeter in Figures 52 and 53 are filtered at 1 kHz at a 12 dB/octave rate
to protect the driver, but the low frequency components of the signal still
have their effect on the total distortion, and increase the presence of
low frequency intermodulation products.
5.3 Electrical Current Tests on Bi-Wired Speaker
Cables
One of the other original motivations for investigating this type of test
signal was the desire to be able to objectively explore the performance of
the various forms of audio cables: line level interconnects, speaker cables,
microphone cables, etc. In that respect, it is a bit of a disappointment,
but one interesting measurement result did develop. For many years,
a debate has raged regarding any possible benefit of using two separate cables
to wire a speaker with electrically separated crossover sections, or bi-wiring
as it is known.
Theory indicates that in any advantage is to be had, it would have to manifest
as a reduction in intermodulation between the low frequency currents and the
high frequency currents in the cable. Separating the two ranges should
provide a measurable benefit.
The signals developed and refined in this paper are highly sensitive to
intermodulation and crossmodulation distortion products, so what better test
signal than a multitone spectral?
A wide-band current probe was used to measure the current flowing through
the speaker cable/s. First the single speaker wire was measured, Figure
55. These are distortion products of the electrical current signal
through the cable, and can be seen to be about 45 to 55 dB down from the
primary tone’s level’s. Then the current transformer was placed on
the tweeter cable of a bi-wiring arrangement, Figure 56. Note the reduction
of low frequency currents in the first spectral tone band, and the reduction
of distortion products by 20 dB or more through the entire midrange.
Figure 57 depicts the current flowing through the woofer side of the bi-wire
cables. Some reduction in distortion products is present in the high
frequency regions, starting as low as 1.5 kHz, and reaching reductions of
10 dB. Figure 58 is the full range single cable overlaid with the tweeter
bi-wire cable, and the reduction in distortion products is readily discernible.
Tests on different single speaker cables have so far been inconclusive,
with inconsistent differences of only 2-4 dB at the limits of the test system
resolution, which are not definitive enough to be able to draw any conclusions
about. Interesting measurement and data, nonetheless.
6.0 Conclusion
An improved method of selecting frequencies for a multitone test signal
has been presented , one which allows the distortion products from the test
signal to avoid being covered up by the primary signal tones or other distortion
products. Measurements taken using the improved method have indicated
that it is a highly effective test signal, and shows promise for use in testing
not only electroacoustic systems, but audio electronic, recording and transmission
systems. Standardization and selection of such a multitone signal could
prove useful to all audio engineers for use with virtually all aspects of
audio system and component testing, and may lead to a better correlation of
objective measurements with subjective listening test results.
7.0 Acknowledgements
I would like to thank Bill Whitlock of Jensen Transformers for originally
sending me the information about the spectral contamination test signal developed
by Sokolich and Jensen. I thank Charles Hughes for his help with this
manuscript and for discussing some of the concepts and ideas with me, especially
the split band spectral signals, and Peavey Electronics Corporation for encouraging
me to present this paper. Thanks also to my wife Patricia, without
whose patience and understanding I could never have finished this.
8.0 References
(1) Robert R. Cordell, "A Fully In-Band Multitone Test for
Transient Intermodulation Distortion", JAES Vol. 29 Issue 7/8.
(2) Deane Jensen and Gary Sokolich, "Spectral Contamination Measurement",
AES Preprint #2725
(3) Mix magazine, May 1998, p. 142, review of the Benson StudioStat
8.2, Lab analysis, by Mike Klasco and June 1998, p. 174, review of the KRK
Systems Expose Series E-8, Lab Analysis by Jack Hidley.
(4) Stephen Boyd, "Multitone Signals with Low Crest Factor", IEEE
Trans. on Circuits and Systems, Vol. 33, #19, Oct. 1986
Mathias Friese, "Multitone Signals with Low Crest Factor", IEEE Trans.
on Communications, Vol. 43, #10, Oct. 1997
Larry J. Greenstein & Patrick J. Fitzgerald, "Phasing Multitone Signals
to Minimize Peak Factors", IEEE Trans. on Communications, Vol. 29, #7, July
1981
(5) Richard C. Cabot, "Performance Assessment of Reduced Bit Rate
Codecs", presented at the AES Conference "Managing the Bit Budget", London,
May 1994. Available as a reprint from Audio Precision upon request.
Appendix A - Equipment used in analysis of spectral test signals
CD player, primary signal source from digitally generated CD-R's:
Marantz PMD 321
Mixer: Peavey RSM 1662
Microphone: ACO 7012
Mic Preamp/Meter: Larson Davis Model 800B
FFT Analysis, HD plots, Frequency Response Plots: TEF 20, SLX software,
FFT 8192 point mode
Power Amplifier: Peavey CS-1200
Wide-band Current Transformer: Pearson Electronics Model 411
Copyright Notice
All website content Copyright Jon M. Risch, 1997 - 2004
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.
----> E-MAIL ME ! <-----
Back to home page