Note: Descriptions are shown in the official language in which they were submitted.
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
TESTING OF CATV SYSTEMS
Field of the Invention
This invention relates to the detection of impedance mismatches in
circuits. It is disclosed in the context of a system for detecting impedance
mismatches
in forward and return CATV signal paths, but it is believed to be useful in
other
applications as well.
Background of the Invention
Various techniques for detecting impedance mismatches and other
phenomena in signal paths are known. There are, for example, the techniques
illustrated and described in U.S. Patents: 5,343,286; 5,:323,224; 5,307,140;
5,069,544;
5,066,118; 5,008,545; 4,904,864; 4,893,006; 4,838,690; and, 4,816,669. While
most
of these references disclose their techniques in the context of optical time
domain
reflectometry (OTDR), the concepts disclosed in them are applicable to other
impedance mismatch detecting techniques as well. In some of these references,
reflections from impedance mismatches are employed im one way or another to
determine the existence, and in certain circumstances the locations, of those
mismatches. These references all teach the application of test signals having
configurations calculated to enhance detection of such reflections and
extraction of the
information sought from such reflections. Other techniques have been proposed
for
detecting imperfections in forward and return circuits in CATV systems. There
are,
for example, the systems proposed in Williams, Proofing and Maintaining
Upstream
Cable Plant With Digital Signal Analysis Techniques, May, 1997. However, this
paper
also proposes the application of a test signal having a configuration
calculated to
enhance detection of imperfections in the forward and return paths of CATV
systems.
Most of such systems also require a clear channel for the conduct of the test.
Disclosure of the Invention
According to one aspect of the invention, a method for determining the
transit time to a feature in a digital communication circuit comprises
generating a
quantity of data that is at least quasi-random, transmitting the quantity of
at least
CA 02315759 2000-06-20
WO 99/38023 PCTIUS99/01432
-2-
quasi-random data along the circuit from a transmitting end of the circuit,
recovering
reflections from the circuit adjacent the transmitting end of the circuit,
correlating the -
reflections with the quantity of data to generate a correlation result,
identifying a
reflection peak in the result, and correlating the reflectians with the data
to generate a
correlation result, for identifying a reflection peak in the result, and for
determining a
time delay to the reflection peak.
Illustratively according to this aspect of the invention, the method
comprises a method for determining the location of an impedance mismatch in
the
circuit, the step of recovering reflections from the circuit comprising the
step of
recovering reflections from impedance mismatches in the circuit. The method
further
comprises the step of multiplying the propagation velocity of the data through
the
circuit by the time delay to the reflection peak to determine the round trip
distance to
the impedance mismatch.
Illustratively according to this aspect of the invention, the steps are
1 S repeated, and average time delays over the number of repetitions are
developed.
Illustratively according to this aspect of t'he invention, the step of
developing average time delays over the number of repetitions comprises
summing the
time delays determined by the repetitions, and dividing the sum of the time
delays by
the number of repetitions.
Additionally illustratively according to this aspect of the invention, the
method further comprises the step of passing the data through a digital root
raised
cosine filter.
Further illustratively according to this aspect of the invention, the step
of passing the data through a digital root raised cosine filter comprises the
step of
passing the data through a digital root raised cosine filter with an excess
bandwidth
factor of 20%.
According to another aspect of the invention, an apparatus for
determining the transit time to a feature in a digital communication circuit
comprises a
first device for generating data that is at least quasi-random, a second
device for
coupling the first device to the circuit to transmit the at least quasi-random
data along
the circuit from a transmitting end of the circuit, a third device for
recovering
reflections from the circuit, the third device coupled to the circuit adjacent
the
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
-3-
transmitting end of the circuit, and a fourth device for correlating the
reflections with
the data to generate a correlation result, for identifying a reflection peak
in the result,
and for determining a time delay to the reflection peak.
Illustratively according to this aspect of the invention, the apparatus
comprises an apparatus for determining the location of an impedance mismatch
in the
circuit, the third device recovering reflections from impedance mismatches in
the
circuit, and the fourth device comprising a fourth device for multiplying the
propagation velocity of the data through the circuit by the time delay to
determine the
round trip distance to the impedance mismatch.
Illustratively according to this aspect of the invention, the fourth device
comprises a fourth device for correlating multiple reflections with multiple
strings of
data, and for developing average time delays over the number of repetitions.
Illustratively according to this aspect of the invention, the fourth device
comprises a fourth device for correlating multiple reflections with multiple
strings of
data to generate multiple correlation results, for identifying multiple
reflection peaks in
the multiple correlation results, for multiplying the propagation velocity of
the data
through the circuit by multiple time delays to the multiple reflection peaks,
for
summing the multiple time delays, and for dividing the sum of the time delays
by the
number of time delays.
Additionally according to this aspect of the invention, the second device
comprises a digital root raised cosine filter.
Further according to this aspect of the invention, the second device
comprises a digital root raised cosine filter with an excess bandwidth factor
of 20%.
brief Descriptions of the Drawing
The invention may best be understood by referring to the following
detailed description and accompanying drawings which illustrate the invention.
In the
drawings:
Fig. 1 illustrates the autocorrelation of a quasi-random sequence of data
having eight possible values (-7, -5, -3, -1, 1, 3, 5 and 7) useful in
understanding the
present invention;
CA 02315759 2000-06-20
WO 99/38023 PCT/US99101432
-4-
Fig. 2 illustrates a filter characteristic useful in understanding the
present invention;
Fig. 3 illustrates the characteristic of the autocorrelation illustrated in
Fig. 1 filtered by a filter having the characteristic illustrated in Fig. 2;
Fig. 4 illustrates plots of the product of a single autocorrelation and the
averaging of two autocorrelations on the same graph to illustrate an aspect of
the
present invention;
Fig. 5 illustrates a partly block and partly schematic diagram of a circuit
for testing the present invention;
Fig. 6 illustrates a plot of power spectral density versus frequency of a
test signal according to the present invention;
Fig. 7 illustrates a plot of power spectral density versus frequency of a
processed test signal according to the present invention;
Fig. 8 illustrates a plot of amplitude versus time offset of averaged
correlations of a processed test signal according to the present invention;
Fig. 9 illustrates a plot of amplitude versus time offset of averaged
correlations of a processed test signal according to the present invention;
and,
Fig. 10 illustrates plots of amplitude versus time offset of the imaginary
parts of the autocorrelations of two test signals according to the present
invention as
well as the average of these two plots.
Detailed Descriptions of Illustrative Embodiment
As CATV systems evolve to carry digital video data, such as HDTV
and digitally compressed standard video signals, a major concern is developing
over
the quality of current cable plants and subscriber circuits to transmit,
receive and
process the high speed digital data reliably. One of the rr~ajor impediments
to such
transmission, reception and processing is believed to be reflections generated
by
impedance mismatches throughout the CATV circuit. The CATV industry appears to
be moving toward the use of 64 symbol quadrature amplitude modulation (64
QAM),
operating at a data rate of 30 megabits per second (30 Mbps), with each symbol
requiring six bits, for a symbol rate of 5 megasymbols per second (SMsps).
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
-S-
It is presently contemplated that receiver implementations of 64 QAM
will include adaptive equalization designed to mitigate the effects of
reflections, among
other imperfections. However, adaptive equalizers have lLimits on their
effectiveness.
For example, if reflections are more than a few symbol tunes away, the
equalizer may
not be capable of compensating for them. Additionally, even if reflections are
within
the time range of the equalizer, if the magnitude of a reflection is too
large, the
equalizer will not converge, that is, it will not adapt in such a way as to
mitigate the
effects of whatever channel impairments) it is attempting to mitigate. Typical
values
for reflection delay and magnitude equalization limits may be in the range of
5 symbol
IO times and -10 dB, respectively. Assuming that the equalizer is capable of
converging,
after it has converged, observation of the equalizer tap values can provide
insight into
the delay and magnitude of the reflections on the line. These defects can then
be
identified and repaired. However, since there can be situations in which a set
top unit
equipped with an adaptive equalizer will have difficulty demodulating data due
to
reflections, a test instrument capable of detecting reflections without the
need to
demodulate data would be extremely useful.
Equipment has long been in use to measure reflections. Such
equipment includes, for example, TDRs, including OTDRs, and frequency sweep
systems. Each of these methods of measuring reflections requires the injection
of a
known signal into the cable system and observation of the cable system for the
effects
of reflections. This method has the disadvantages that it requires a reference
signal
generation instrument, and either interfering with whatever signals are
already on the
system or requiring unused frequency spectrum in the system. The method
according
to the present invention uses the 64 QAM data carrier itself to detect
reflections
without the need to converge an adaptive equalizer or demodulate data. The
invention
makes use of the presence on the 64 QAM data carrier of random or at least
quasi-
random data. This assumption is not unwarranted since rr~ost 64 QAM systems
require
a data randomizer at the transmitter to assure enough signal transitions to
obtain
timing and carrier synchronization.
The present invention utilizes the autocorrelation properties of random
data. The correlation of two N length sequences x[n] and y[n] of data is
defined as
N-~m~-1
Rx y[m] _ ~ x[n]y[n + m]. ( I )
n=0
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
-6-
Replacing y[n] with x[n], because this is autocorrelation, -
N-~m~-1
RX[m] - ~ x[n]x[n + m]. (2)
n=0
Correlation is a measure of the similarity of two sequences. Autocorrelation
detects
periodic similarities in a single sequence. Since random data should have no
periodicity, autocorrelation of random or at least quasi-random data should be
nearly
zero at all offsets of the autocorrelated signal except zero offset. Fig. 1
illustrates the
autocorrelation of a quasi-random sequence. Each point in the sequence can
assume
one of 23, or 8, possible values. Eight possible data values were chosen for
this plot
because each of the I and Q subchannels of a 64 QAM data stream is made up of
symbols with eight possible values. As Fig. 1 illustrates, the autocorrelation
exhibits
the expected single peak at zero offset and values close to zero everywhere
else,
indicating that the fi~nction is aperiodic. This alone would commend
autocorrelation as
an effective tool in this application except that, in 64 QAM systems, digital
data must
also pass through a bandwidth-limiting filter prior to transmission. This
process keeps
transmitted energy contained in the designated channels of the 64 QAM system.
In
digital CATV, the channel width is 6 MHZ.
A digital root raised cosine filter with an excess bandwidth factor of
20% is emerging as the standard transmit data filter. The characteristic of a
digital
root raised cosine filter with an excess bandwidth factor of 20% is
illustrated in Fig. 2.
While such a filter constrains the bandwidth of the transmitted energy to a 6
MHZ
channel (which permits multiple channels to be transmitted on a single CATV
conductor), it also has the effect of broadening the autocorrelation peak
illustrated in
Fig. 1. Fig. 3 illustrates an enlarged view of the autocorrelation illustrated
in Fig. I
filtered by the filter whose characteristic is illustrated in Fig. 2. As
illustrated in Fig. 3,
the main lobe of the autocorrelation of the filtered data occupies
approximately four
samples on either side of the central peak. This assumes i:our filtered
samples per data
symbol, a common assumption in digital communication systems. Since the data
symbol rate is 5 Msps, the illustrated sample rate is 20 megasamples per
second.
Stated another way, the sample period is 50 nanoseconds. Four samples thus
occupy
200 nsec. In CATV cable coated with polyethylene (PE), the propagation
velocity of
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
_'7-
an electromagnetic disturbance is 2 x 10g meters/sec. Thus, the 200 nsec time
occupied by four samples is equivalent to 40 m travel in the PE cable.
These observations provide a method for locating reflection-producing
discontinuities in the CATV cable. Since a reflection is, theoretically, a
time-delayed
S version of the transmitted signal, with some amount of attenuation owing to
cable loss
and the loss resulting from the reflection impedance mismatch, autocorrelation
of the
transmitted signal with the reflected signal should exhibit peaks not only at
zero offset,
but also at an offset corresponding to the transmission/reflection transit
time. This
secondary peak will be a scaled down version of the zero offset lobe, with the
scaling
factor proportional to the magnitude of the cable and impedance mismatch
losses.
This scaling is the reason why it is important that the autocorrelation
function of the
data should be as close as reasonably possible to zero outside the main lobe.
Otherwise, the reflection peak in the autocorrelation function resulting from
a low
order impedance mismatch might be subject to being obscured by the
autocorrelation
noise. The effects of autocorrelation noise can be reduced further by
averaging a
number of autocorrelations. Fig, 4 illustrates two plots. Both illustrate
autocorrelation of a signal with a -l OdB reflection 150 m away (300 m round
trip).
The broken line plot illustrates the results of a single autocorrelation. The
solid line
plot illustrates the average of ten autocorrelations. As will be appreciated,
the effects
of autocorrelation noise are significantly reduced, and the; reflection peak
in the
autocorrelation function enhanced, by averaging.
Were it not for the effects of modulation of the random data onto the
quadrature (I and Q) RF carrier, this analysis would be complete, having
demonstrated
the benefit of using autocorrelation to ascertain impedance mismatch
reflections and
the like in CATV systems. However, with an RF carrier, the reflection is not
only a
time-delayed and attenuated copy of the transmitted signal, but is also phase
rotated an
amount determined by the number of wavelengths modulo 360° of the
reflection. The
phase rotation causes the reflected signal's I and Q components to be
misaligned with
the I and Q components of the transmitted signal. Reference is here made to
the
definition of the autocorrelation of a complex signal:
N-~m~-1
Rx[m) _ ~ x[n]x*[n + m) (3)
n=0
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
_g-
where x*[n) denotes the complex conjugate of x[n]. Hence, since QAM transmits
independent data streams in the I and Q subchannels, autocorrelation of
baseband -
QAM results in a sequence whose real-valued part has a peak at zero offset
twice the
height of the peak in the autocorrelation of either the I or Q subchannel
alone. The
imaginary-valued part of the autocorrelation results from the cross-
correlation of the I
data and the Q data. Since these data sequences are independent and have zero
means,
their cross-correlation has approximately zero value everywhere. When this
baseband
I and Q subchannel information is modulated onto an RF carrier, and assuming
an
arbitrary phase rotation ~ of the reflection,
N-~m~-'
RX[m) _ ~ x[n]x*[n + m], where x[n] = I[n) + jQ[n) (4)
n=0
N-~ml-1
- ~ (I[n) +JQ[n)) (I[n + mil -JQ[n + m])e'~
n=0
N-~m~-1
- ~ (I[n] + jQ[n)) (I[n + m) - jQ[m + m]) (cosh + jsinc~) (6)
n=0
N-~m~-1
- ~ (I[n)I[n + m) - jI[n)Q[n + m) + jI[n + m)Q[n) + Q[n]Q[n + m]) (cosh +
jsin~l)
This simplifies to
RX[m) - (RI[m) -~- RQ[m)) (cosh + jsin~). (8)
To determine the magnitude of this function, it can be multiplied by its
complex
conjugate, (RI[m) + RQ[m]) (cosc~ - jsin~). Doing so yields
~~[m] ~ - RUm) + ~[m)~
Since the autocorrelation of each of the separate data streams will have the
requisite
narrow main lobe and low sideband noise, as previously demonstrated, their sum
will
also. Therefore, the autocorrelation of the transmitted 64 QAM data stream
with its
reflection from impedance mismatches in the CATV circuit can reliably by used
to
determine the distances to these impedance mismatches. The following
experiments
were conducted to demonstrate the reliability of this technque.
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
-9-
With reference to Fig. 5, an arbitrary waveform generator 20 such as,
for example, a Stanford Research Systems model DS345 AWG, having an output -
sample rate of 40 MHZ was programmed to output 16,100 samples of 64 QAM data
repeatedly. This corresponds to about 2000 symbols of 64 QAM data at 30 Mbps.
The data was modulated on a 4 MHZ carrier. This signal was then placed on a
reflection-generating cable system 22, and a digital oscilloscope 24 such as,
for
example, a Gage model CS 1012 oscilloscope sampling a1; 20 Msamples/sec., was
used
to monitor the reflections. The test circuit included SOS~~-7552 impedance
matching
transformers 26, 28, two-way 7552 splitters 30, 32, 34, and a length 36, for
example,
150.1 m, of test cable, such as, for example, Belden 9231 or 9266 video cable,
a
terminal length of, for example, 6.9 m of Channel Master RG-59 video cable
terminated by a 2 dB pad. The data recovered by the digital oscilloscope 24
was then
transferred to the Matlab program running on a PC 40 for analysis according to
the
present invention.
Several experiments using different lengths 36 of test cable and
terminations were run to determine the accuracy and resolution of the
reflection delay
and amplitude measurement. Fig. 6 illustrates a typical power spectrum of
collected
test data, including the transmitted signal and the reflection. As illustrated
in Fig. 6,
the signal width is less than 6 MHZ and is centered at 4 MHZ. The power
spectrum
extends from DC to the 10 MHZ Nyquist limit. The first step is to recover the
I and Q
data. This is accomplished by complex downconversion to a center frequency,
ideally
0 Hz, but in any event much less than the symbol rate. Frequencies in the
range of 50
KHz are acceptable. As previously demonstrated, phase offset does not
adversely
affect the measurement either. After complex downconversion, the resulting
signal is
low pass filtered to remove the component at twice the original Garner
frequency. The
result of this filtering is illustrated in Fig. 7. Since the signal is now
complex valued,
all frequencies from DC to the sampling frequency must tie illustrated to
illustrate the
signal's spectral density. Once the near-baseband I and Q signals have been
recovered,
a complex autocorrelation, as described above, can be performed.
It should be remembered that the AWG 20 output was a continuous
repetition of 16,100 samples at 40 MHZ which represented about 2000 64 QAM
symbols. Since the digital oscilloscope 24 samples at 20 MHZ, it will recover
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/01432
-10-
16,100/2, or 8050, unique samples before the stream repeats. By identifying
the
transmit finite impulse response filter startup and ending transients in the
recovered -
data, an 8050 point unique received data sequence of 201)0 symbols is used.
2000/K
complex autocorrelations are then performed on blocks of K symbols out of the
2000
total symbols received. The results from these autocorrelations are then
averaged and
the magnitude of the average of the complex autocorrelations is computed as
described
in equation 9. The result of this autocorrelation is illustrated in Fig. 8.
From Fig. 8,
one significant reflection at approximately 1.5 p.sec. is immediately
recognizable. To
identify others, the scale of Fig. 8 is expanded in Fig. 9.
Fig. 9 illustrates three plots, two in broken lines and one in solid lines.
The broken line plot with the lowest amplitude sidelobes was generated using
the test
apparatus illustrated in Fig. S in which the length of test cable included
150.1 m of
Belden 9231 video cable, 6.9 m of Channel Master RG-59 video cable and a 2 dB
pad
terminating the RG-59 cable. This plot illustrates a fairly well-defined
reflection at
1.566 psec., corresponding to termination of the Channel Master RG-59 cable,
and an
amplitude about 3.6 dB lower than the other broken line plot, the plot
generated from
the same lengths of cable with the 2 dB terminal pad removed. The solid line
plot was
generated using the test apparatus illustrated in Fig. 5 with the 150.1 m
length of 9231
cable only. The data illustrated in Fig. 9 agree remarkably well with the
physical test
setup. The propagation velocity in the 9231 cable is 2 x :108 m/sec. The 1.566
usec.
offset thus corresponds to a round trip distance of 313.2 m compared with the
314 m
(150. lm + 6.9 m) actual distance. The 9231 cable itself generated a
reflection sidelobe
peak at 1.497 sec., corresponding to a distance of 299.4 m versus an actual
round
trip distance of 300.2 m. The presence of the 2 dB pad termination is also
evident in
the difference in the amplitudes of the sidelobes of the two broken line
plots.
It should be noted that the width of the main lobe in the illustrated
embodiment will obscure any reflections within a couple of hundred nsec.,
corresponding to a distance of about 30 m or so. However, the main lobe
component
is entirely in the real component of the autocorrelation response. The
imaginary
component is simply 2RIQ[m), which is zero when there is. no reflection. When
there is
a reflection, there will be a phase angle dependent response in the imaginary
part of the
autocorrelation. Fig. 10 illustrates this effect. As illustrated there, a
significant
CA 02315759 2000-06-20
WO 99/38023 PCT/US99/Oi 432
-11-
response can be seen quite close to zero offset in the imaginary component.
However,
the absolute magnitude of the response is dependent upon the phase angle of
the -
carner to the reflection. It is likely that a scheme repeating this test at
multiple carrier
frequencies could extrapolate an absolute amplitude, since changing the
carrier
frequency will also change the phase angle to the reflection.
It should further be noted that, although the illustrative embodiment
was disclosed generally in the context of testing of the forward path from the
CATV
plant toward the subscriber apparatus, the same principles can be applied
directly to
testing of the return path from the subscriber apparatus toward the CATV plant
in
two-way systems. In the return path, modulation is typically quadrature phase
shift
keying or 16 QAM. The delay resolution of the measurement will not be
sensitive to
the modulation format. It is, however, proportional to the modulation
bandwidth. The
wider the bandwidth, the finer the resolution and the closer to the source
reflections
can be detected. As with 64 QAM measurements, the data must still be
randomized or
at least quasi-randomized. Again, however, this is generally a standard
feature of
digital communications systems.
