BROADBAND SWEEPING: A NEW APPROACH William Gregory

BROADBAND SWEEPING: A NEW APPROACH William Gregory
BROADBAND SWEEPING:
A NEW APPROACH
William Gregory Kostka
Engineering Manager
Gillcable TV
234 East Gish Road
San Jose, CA 95112
ABSTRACT
An active CATV system frequency
response is typically measured by using
either a high-level or a low-level sweep-but both of these methods have inherent
limitations. Gillcable's on-going gated
mid-level sweep project is developing a
new method which can improve resolution
and readabi 1 ity, while simultaneously
minimizing system interference.
High-level sweeps provide good signal
to noise ratios and are easy to detect,
but they significantly interfere with
existing cable signals. Because of this
interference, the sweep repetition rate is
usually low, thus impairing readability.
While low-level sweep is relatively noninterfering, its low level reduces the
signal to noise ratio thus impairing
resolution. Furthermore, any higher level
cable signal will mask the sweep and also
impair readability.
Gillcable's gated mid-level sweep
project is based on selective spectrum
sampling; it is not an analog sweep, but
a series of samples at arbitrary
frequencies.
The sample points can be set
to avoid any critical cable frequencies.
But beyond just that, to avoid interfering
with television pictures, the signal
source will only generate signals on an
occupied channel during video blanking
intervals.
Since the gated signals only
minimally interfere with existing signals,
sweep level can be set high enough to
deliver good signal to noise ratios for a
high-resolution display.
communication at both the head-end and in
the
field;
while others,
like
simultaneous high-level sweep or low-level
sweep
can
be
continuously
and
automatically combined with regular cable
signals.
The two former cases are not
applicable in large active cable systems.
The first method would require removing 24
hour premium services for extended
periods,
an act certain to provoke
customer complaints, while the second
method would require far too much time to
sweep an 1800 mile plant like Gillcable.
The
l a t t e r method avoids
those
deficiencies. The sweeps run concurrently
with cable programming, so no program
interruptions are required.
Furthermore,
since they are continually available, many
sweep
receivers
can
be
in
use
simultaneously, so a large plant can be
swept in reasonable time.
The high-level sweep is typically run
15 to 20 dB above video carrier level.
This relatively high level delivers a
clear display because of its high signal
to noise ratio.
Also, its level makes it
easy to recover, so sweep receivers are
relatively uncomplicated and hence
inexpensive.
High-level sweep, however,
does have a significant drawback--it
interferes with the existing cable
signals.
As the sweep passes through an
occupied TV channel, it can
• create visible distortions in the
picture,
• prematurely trigger the vertical
sync circuits in the TV causing the
picture to roll,
METHODS FOR MEASURING FREQUENCY RESPONSE
Frequency response, gain variation at
different frequencies, is a critical CATV
distribution system parameter.
Several
methods exist for measuring this response.
Some, like noise insertion or standard
broadband sweeping, require the cable
signals be removed;
others, like slow
sweeping, require technicians in constant
•
cause VCR servos
while recording, or
to
lose
lock
•
fool AFC'd set-top decoders into
"following" the sweep up and then locking
onto the next channel.
The exact symptoms depend on subscriber
terminal equipment, sweep level, and sweep
1985 NCTA Technical Papers-83
speed.
High-level sweeps also affect inband and out-of-band telemetry and system
pilots.
The sweep can be trapped at these
critical frequencies, but in a loaded
system the sweep display soon begins to
have as many holes in it as a golf course.
One other distortion that can occur is
1 oss of accuracy caused by too high a
sweep level driving amplifiers into
compression.
Basically, as long as the
amplifiers are operating in their linear
regions high-level sweep provides a clear
accurate display, but the interference is
significant.
The interference can be minimized by
reducing the time the sweep is actually
present on the system.
The repetition
rate for high-level sweep is typically one
sweep every 5 to 20 seconds.
This low
"rep" rate can be accommodated by using
storage oscilloscopes or patient
technicians.
Incidentally, to further
minimize the effects of the high-level
sweep used at Gill, we have installed
timers that prevent sweep from occuring
during prime-time hours and have installed
radio-controlled- switches that enable
sweep for 10 minutes at a time.
That is
enough time for a technician to adjust a
station, then the sweep automatically
turns off while he is in route to the next
station.
Low-level sweep systems typically run
20 to 40 dB below video carriers.
This
prevents it from being as interfering as
high-level sweep.
Low-level sweep can be
set far enough below video carriers so as
to be virtually non-interfering.
This
reduction in level, however, makes the
sweep signal more difficult to recover, so
the receivers are correspondingly more
complex and expensive. They are narrowband spectrum analyzers that track the
transmitter by locking to a pilot carrier.
The reduction in level has also had an
effect on display resolution. Since the
sweep signal is below the level of cable
signals, the sweep is masked by the cable
signals and some of their side-bands.
In
a fully loaded cable system, much of the
sweep is simply not visible.
Also, the
corresponding signal to noise ratio of the
recovered sweep signal suffers from the
reduced sweep level;
the display is often
an ambiguous 2 dB wide trace.
This effect
is accentuated by amplifier cascade
length, but then so are system frequency
response problems.
So in long cascades
where the most careful adjustments are
needed, low-level sweep provides the least
resolution.
In short, both high- and low-level
sweep methods leave room for improvements.
84-1985 NCTA Technical Papers
MID-LEVEL SWEEP PROJECT
The purpose of this paper is to
describe an on-going research and
development project at Gillcable. We had
as our goal, an improved CATV sweep
system.
We wanted a sweep system that was
non-interfering, yet was continually
present, and of course, required no
technicians at the head-end.
The sweep
level was to reflect the typical video
carrier levels on the cable:
low enough
to avoid non-linearities from the
amplifiers, yet high enough to provide
a clear, unambiguous sweep display. The
receivers had to be easy-to-use, accurate,
reliable, and inexpensive.
Since they
were field equipment, they also had to be
compact, 1 ightweight, rugged, and battery
operated.
With these goals in mind, we proposed
a sweep system uniquely suited to our CATV
environment. We refer to it as the "Midlevel Sweep System." Mid-level refers to
the RF carrier level, which is set at
video carrier level.
This puts it midway
between high- and low-level sweep;
15 to
20 dB below high-level and 20 to 40 dB
above low-level.
This RF level will
provide a clear display even in long
cascades while avoiding any non-linear
amplifier distortions from using too high
a level. The term "sweep", however, is a
misnomer.
Instead of using an analog
sweep that would pass through all possible
frequencies, this proposed system would
instead,
use a switched carrier at
discrete frequencies.
The frequencies
could be selected somewhat arbitrarily as
long as they were sufficiently closely
spaced to provide enough resolution to
assure response anomalies were not missed.
But more
importantly,
critical
frequencies, like system pilots and
telemetry carriers could be bypassed
entirely.
Interference at those critical
frequencies, then, would not be a problem.
Obviously, we couldn't bypass all
occupied TV channels, and interference
caused by inserting an additional carrier
into an occupied video channel is a major
problem. There is simply no way to do it
without creating some beat in the picture.
However,
during the horizontal and
vertical blanking intervals in an NTSC
television signal, the TV receiver is
blanked;
the picture tube is cut off, and
the screen is black. We proposed to use
these blanked intervals to hide any
interfering beats our additional carrier
created.
Additionally, video side-band
energy should be minimal during those
intervals, so we would have a relatively
clean spectrum
signals.
in
which
to
insert
our
The transmitter and receivers would
be synchronized by an auxiliary data
channel. Both timing and frequency data
would insure the receivers could find
these very short pulses as they are moved
through the cable spectrum.
TRANSMITTER DESIGN CONSIDERATIONS
The transmitter control section needs
to be intelligent enough to decide when a
channel is occupied, even when the video
is scrambled. It also needs some form of
memory wherein to store the critical
frequencies to be skipped.
For these
reasons, a microprocessor will be used to
implement the programmable control
section.
Figure 1 shows a simplified
block diagram of the transmitter.
T~AHSII! TTfR
HEAD-END
channel.
This data stream contains the
necessary frequency and timing information
for the receivers.
The carrier frequency
of the FSK telemetry modulator is
considered one of
the c r i t i c a l
frequencies.
RECEIVER DESIGN CONSIDERATIONS
The receiver uses a microprocessor
for coordinating received telemetry with
synthesizer control, as well as for
interpreting and preparing the received RF
data information for the display.
A
simp 1 ified block diagram of the receiver
is shown in figure 2.
RECEIVER BlOCk DIAGRA"
BlOCK DIAGRAII
CABlE
Figure 2.
Figure 1.
Simplified block diagram of
transmitter.
One of the unique parts of the
transmitter is the "look-ahead" frequency
agile video demodulator.
This demod is
steered to the channel to be sampled next.
The video is detected and the sync
information stripped off.
The sync is
used for timing of the inserted carriers;
no energy will be added to an occupied
channel until either horizontal or
vertical blanking is occuring.
The carriers are created by steering
a synthesized upconverter to the proper
location, then at the proper time, adding
the local oscillator to its input.
Pulse
rise- and fall-times are restricted to one
microsecond by the 350 KHz band-pass
filters.
This is derived from the
approximation:
Bandwidth
= 0.35 I rise-time.
Without the f i l t e r , the effective
bandwidth caused by switching the carrier
on and off at microsecond rates, would
extend beyond the channel being tested.
sent
Synchronizing signals are continually
downstream on the auxi 1 iary data
Simplified block diagram
of receiver.
The auxiliary data channel is
demodulated and the frequency and timing
information made available to the
microprocessor. Using this information,
the local oscillator is steered to the
proper frequency, then at the proper time,
the pulse is detected.
Its level is
captured and converted to a digital value.
These digital values are then decoded and
used to generate the display.
The digitization of the RF signal
provides the means to great flexibility in
receiver design.
Individual receiver
flatness can be compensated for digitally.
By connecting the receiver directly to the
transmitter, any frequency response
anomalies can be recorded and stored in
non-volatile RAM.
Then by compensating
the readings by those factors,
the
displayed response will reflect only cable
system response, not receiver response.
Other techniques like digital averaging
can be used to help readability in long
cascades.
Incidentally, bypassing critical
frequencies with this system will not
cause the display to have holes in it.
The points in between the sampled points
will be approximated by using a quadratic
approximation generated by the nearest
three sampled points.
1985 NCTA Technical Papers-85
INITIAL TESTING PHASE
Our early days were spent trying to
discover how much we could abuse an NTSC
video signal and not cause perceptible
interference.
We wanted to use the
horizontal sync pulses for timing because
they were contained in the horifontal
blanking interval, see figure 3 , and
their high frequency, if completely
utilized, would allow the entire spectrum
to be scanned in a very short time.
If we
were to sample a 50 to 300 MHz cable
spectrum every megahertz it would take 251
samples.
The horizontal line rate for
NTSC is about 15750 Hz.
If we could
manage to use every line for one sample
then we could scan the entire band in less
than a sixtieth of a second.
We could
then have a display refresh rate of better
than 60 Hz as indicated by table 1.
HOII!ZO~TAL
BAND
SAMPLE
FIIEQ.
IIATE
TOTAL
SAMPLES
SAMPLE
RATE
DISPLAYRATE
REFRE~
50-300 MHZ
1
~z
2S1
1S750 HZ
&2.8 HZ
50-300 MHZ
2
~z
126
1S750 HZ
125.0 HZ
VERTICAL RATE
SAMPLE
FREO.
BAND
TOTAL
SAMPLES
SAMPLE
RATE
DISPLAY
RATE
REFRE~
50-loo
~z
1
~z
2S1
60 HZ
~.2
50-Joo
~z
z
~z
12&
60 HZ
2.1 SECS.
SECS.
Table 1. Minimum screen refresh rates
for typical frequency intervals during
either vertical or horizontal blanking.
Actual display refresh rates would be
higher, as there is spectrum where there
are no video signals to wait for, some
critical frequencies would be bypassed,
and adjacent video channels' horizontal
sync phase relationships would tend to
reduce the effective horizontal rate when
moving from channel to channel.
Figure 3. Diagram showing horizontal
blanking and sync relationship.
Keeping up with these rates
to be difficult.
Synthesizers
able to slew to a new frequency
within one horizontal line,
86-1985 NCTA Technical Papers
was going
had to be
and lock
about 63
microseconds, and the auxiliary data
channel had to have a very-high data rate.
This turned out to be impractical, not
because of equipment limitations, but
because
of
television
receiver
complications.
TV receivers suffered from three
types of visible interference, even though
the beats were invisible.
The exact
symptoms varied from model to model
depending on what type of video AGC was
used;
what type of horizontal AFC was
used; and what type of detector was used.
Video AGC's were of two varieties:
•
Sync peak detectors with
filters, and
•
low-pass
Sync gated circuits.
By adding extra energy during the
horizontal sync period, that sync pulse
had a higher amplitude than adjacent sync
pulses.
This extra amplitude had no
visible affect on the AGC circuits that
used sync peak detectors and low-pass
filtering.
But the gated AGC circuits
behaved differently. These gated circuits
derived AGC voltage on a line by line
basis, so the AGC voltage applied to the
following line of video was almost
entirely determined by the preceeding sync
p u 1 s e' s
amp 1 it u de.
By art if i c a 1 1 y
increasing one sync pulse's amplitude, the
gain was reduced for the entire following
1 ine of video.
This reduced the contrast
for that line and, depending on program
video content, the effect varied from
nearly imperceptible to very obvious.
Horizontal AFC circuits were also
affected by the timing of these pulses.
When inserting these pulses in the
horizontal sync period, we had some
discretion in pulse position relative to
sync. S orne types of set AFC's ignored the
extra energy when the pulse was started
coincident with the start of sync, while
others ignored it when it was exactly
centered in the horizontal sync.
The
visible effect of not being properly timed
was a slight pulling of the horizontal
oscillator which was visible as a
discontinuity in vertical lines, followed
by a curved line as the oscillator
regained its original phase.
No matter
where we chose to start the pulse, we were
guaranteed to visibly affect some of the
TV sets.
TV sets with product detectors also
exhibited one other sensitivity to these
additional carriers that diode detector
sets did not. When the additional carrier
was removed,
the detector started
oscillating.
The duration of these
oscillations were both frequency and level
sensitive.
At our chosen levels,
the
oscillation lasted through color burst,
back-porch, and into active video, about 8
to 10 microseconds. This was visible as
disturbance in one line at the left margin
of the picture.
Short of reducing the RF level to 20
dB below video carriers or less, there was
no way to insert these pulses into the
horizontal blanking interval undetected.
Since we had set out to have a higher
level than that for improved resolution,
we then focused our attention on the
vertical interval (see figure 4 1 ).
... ~,;;o;<..;:-,;;:,~-''----.""..,,,,.... ...,, .....
Figure 4.
Vertical sync intervals for
fields 1 and 2.
By using the vertical interval, we
effectively hid all of tl:e above mentioned
video artifacts.
Single 1 ine AGC affects
in gated sync circuits occured during the
time the entire line was blanked, hence
the effect was invisible.
By restricting
ourselves to lines 4 through 9, any affect
we had on the horizontal oscillator has
until line 21, where active video starts,
to recover. This was 1 ong enough for all
the sets tested.
Lastly, since the screen
is blanked the detector oscillations are
not visible as video disturbances. This
is not to say that any of these problems
quit occuring, I only indicate that their
effects cannot be viewed on a normal TV
set when they occur in the vertical
interval.
We have in fact inserted
additional carriers into the vertical
interval at RF levels 20 dB higher than
the video carrier with no visible
interference.
CONCLUSION
Tests are s t i l l in progress to
determine if there are any unforseen and
as yet undiscovered complications that
would prevent this from being a viable
method. All the evidence so far indicates
that signals carefully inserted into the
vertical interval are ignored by normally
operating television receivers and VCR's.
Nevertheless, we are trying to find if any
combinations of timing, level or frequency
can cause a VCR to break servo lock or
cause interference to a TV receiver.
Further tests are needed to clarify
just how close the samples need to be in
order to assure we will not skip typical
cable frequency response problems.
Certainly, one megahertz is close enough
as experience indicates that response
anomalies seldom affect less than six
megahertz.
But six megahertz is probably
not close enough for the occasional
problem that affects only a small portion
of the cable spectrum.
Sampling at two
megahertz intervals now seems to be the
optimum compromise between speed,
synthesizer design, and resolution.
Still to be decided is the final form
of the receiver display.
Among the
options are flat panel
displays,
especially the electro-luminescent types.
But the lowest cost display is still an XY display using portable oscilloscopes
like the Tektronix 323's or Leader LB0308'S.
As a work-in-progress report, I can
say that the initial concepts and designs
have been worked out and patents applied
for.
Continuing work will certainly bring
the new sweep system out of the lab and
into the field by the end of the fiscal
year.
Initial results are encouraging and
indicate that we will be able to have a
non-interfering, high-resolution CATV
sweep system in use at Gillcable then.
ACKNOWLEDGEMENTS
The relatively slow 60 Hz rate of the
vertical
interval had concommitant
simplifications.
Synthesizer design was
much simpler;
we no longer had to slew
and 1 ock in 1 ess than 63 microseconds.
Also, the auxiliary data channel could now
carry much more data between samples at a
lower
data
rate
for
better
synchronization.
The lower throughput
also permitted the use of common,
inexpensive microprocessors to reduce
cost.
But as Table 1 indicated, the
display refresh rate had fallen to one
sweep every four seconds.
The display
requirements had become slightly more
complicated, screen refresh had to be
independent of incoming data.
I would like to acknowledge those at
Gillcable who have helped this project
along.
In particular, David Large was
responsible for the initial concepts, and
Rich Wayman was responsible for much of
the data gathering and transmitter design.
There was also the sweep crew who
continually reminded us of what real cable
problems were like outside of the lab.
REFERENCE
1.
Figures 3 and 4 were adapted from
"TELEVISION OPERATIONAL MEASUREMENTS,
Video and RF for NTSC Systems", Tektronix,
1984, p. 9, 11, 12.
1985 NCTA Technical Papers-87
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement