Group Delay

Group Delay
Ron Hranac
Group Delay © 2005 Cisco Systems, Inc. All rights reserved. Cisco Public 1
The Basics
Bandpass filter equivalent
• Consider the 6 MHz spectrum occupied by an analog

TV channel or digitally modulated signal, the 5-42
MHz upstream spectrum, or any specified bandwidth
or passband as the equivalent of a bandpass filter.
The Basics
Transit time and velocity of propagation
• A signal takes a certain amount of time to pass

through a filter
The transit time through the filter is a function of the filter’s
velocity of propagation (also called velocity factor)
Velocity of propagation is the speed that an
electromagnetic signal travels through some medium,
usually expressed as a percentage of the speed of light in a
vacuum
The Basics
Velocity of propagation versus frequency
• In many instances the velocity of propagation

through a filter varies with frequency
The velocity of propagation may be greater in the center of
the filter’s passband, but slower near the band edges
Passband center
-3 dB point
The Basics
Delay and absolute delay
• The finite time required

for a signal to pass
through a filter—or any
device for that matter—
is called delay
Absolute delay is the
delay a signal
experiences passing
through the device at
some reference
frequency
The Basics
Delay versus frequency
• If delay through a filter is plotted on a graph of

frequency (x-axis) versus time delay (y-axis), the
plot often has a parabola- or bathtub-like shape
Time delay
5
4
3
2
1
0
Frequency
The Basics
Network analyzer plot of Ch. T8 bandpass filter
• The upper trace shows
magnitude versus
frequency: the filter’s
bandpass characteristics.
The x-axis is frequency,
the y-axis is amplitude.
• The lower trace shows
group delay versus
frequency. The x-axis is
frequency, the y-axis is
time. Note the bathtub-
like shape of the curve.
Graphic courtesy of Holtzman, Inc.
The Basics
Phase versus frequency
• If propagation or transit time through a device is

the same at all frequencies, phase is said to be
linear with respect to frequency
If phase changes uniformly with frequency, an output
signal will be identical to the input signal—except that it
will have a time shift because of the uniform delay through
the device
• If propagation or transit time through a device is

different at different frequencies, the result is delay
shift or non-linear phase shift
If phase changes non-linearly with frequency, the output
signal will be distorted
The Basics
Delay and phase distortion
• Delay distortion—also known as phase distortion—

is usually expressed in units of time: milliseconds
(ms), microseconds (µs) or nanoseconds (ns)
relative to a reference frequency
• Phase distortion is related to phase delay
• Phase distortion is measured using a parameter
called envelope delay distortion, or group delay
distortion
The Formal Definition
Group delay is “the derivative of radian phase with

respect to radian frequency. It is equal to the phase
delay for an ideal non-dispersive delay device, but
may differ greatly in actual devices where there is a
ripple in the phase versus frequency characteristic.”
IEEE Standard Dictionary of Electrical and Electronics Terms
The Math (Yikes!)
• In its simplest mathematical representation…

dϕ
GD =
dω
where GD is group delay variation, φ is phase in
radians, and ω is frequency in radians per second
• For the purists, group delay τ also is defined
∂ϕ (ω )
τ (ω ) = −
∂ω
• And yet another definition is
d d
D(ω )∆ − Θ(ω )∆ − ∠H (e jωΤ )
dω dω
The Translation (Whew!)
• If phase versus frequency is non-linear, group

delay exists.
• In a system, network, device or component with no
group delay, all frequencies are transmitted
through the system, network, device or component
in the same amount of time—that is, with equal time
delay.
• If group delay exists, signals at some frequencies
travel faster than signals at other frequencies.
The Analogy
Imagine a group of runners with identical
athletic abilities on a smooth, flat track …
All of the athletes arrive at the finish line

at exactly the same time and with equal
time delay from one end of the track to
Example courtesy of Holtzman, Inc.
the other!
Group Delay: An Analogy
Now let’s substitute a group of RF signals
for the athletes. Here, the “track” is the
equivalent of a filter’s passband.
Frequency
Time
All of the frequencies arrive at the
destination at exactly the same time and
with equal time delay through the filter
passband!
Back to athletes, but now there are some that
have to run in the ditches next to the track.
Some athletes take a little longer

than others to arrive at the finish
line. Their time delay from one end
Example courtesy of Holtzman, Inc. of the track to the other is unequal.
Substitute RF signals for the athletes again. The
“track” is a filter’s passband, the “ditches” are
the filter’s band edges.
Frequency
Time
Time difference
(unequal time delay
through the filter)
• Group delay exists, because some frequencies—

the ones near the band edges—took longer than
others to travel through the filter!
• Now take the dotted line connecting the
frequencies and flip it on its side. The result is the
classic bathtub-shaped group delay curve.
Time difference
(unequal time
delay)
Frequency
The Classic Group Delay “Bathtub Curve”
• In this example,
the group delay
between 25 and 40
MHz is about 300 ns
(3 vertical divisions
at 100 ns each)
• That is, it takes

the 40 MHz signal
300 ns longer to
reach the headend
than the 25 MHz
signal
Time difference
in nanoseconds
Frequency
Adaptive Equalization
• Adaptive equalization is a tool to combat linear distortions
such as group delay
• All DOCSIS cable modems support downstream adaptive
equalization (always on)
• DOCSIS 1.1 and 2.0 cable modems support upstream adaptive
equalization (pre-equalization in transmitted signal, may be
turned on/off by cable operator)
DOCSIS 1.1 modems: 8-tap adaptive equalization
DOCSIS 2.0 modems: 24-tap adaptive equalization
Adaptive equalization is not supported in DOCSIS 1.0 modems
Adaptive Equalization
• Recall the group delay analogy using athletes on a

track:
Adaptive equalization can be thought of as analogous to
delaying the runners in the middle of the track, allowing the
slower runners in the “ditches” to catch up.
This allows all runners to arrive at the finish line at the
same time, with equal time delay.
Phase Versus Frequency
• OK, group delay exists if phase versus frequency is

non-linear
• But just what does that mean?
• Let’s look at an example, using a 100 ft piece of
.500 feeder cable
Back to Basics: Velocity of Propagation
• Hardline coax used for feeder applications has a

velocity of propagation of around 87%
The speed of light in free space or a vacuum is 299,792,458
meters per second, or 983,571,056.43 feet per second—1
foot in about 1.02 ns
In coaxial cable with a velocity of propagation of 87%,
electromagnetic signals travel at a velocity equal to 87% of
the free space value of the speed of light. That works out to
260,819,438.46 meters per second, or 855,706,819.09 feet
per second—1 foot in about 1.17 ns
So, electromagnetic signals will travel 100 ft in a vacuum in
101.67 ns, and through a 100 ft piece of coax in 116.86 ns
Back to Basics: Wavelength and Period
• Wavelength (λ) is the speed of propagation of an
electromagnetic signal divided by its frequency (f)
in hertz (Hz). It is further defined as the distance a
wave travels through some medium in one period.
Period (T) of a cycle (in seconds) = 1/f, where f is frequency
in Hz
One cycle
+
Amplitude
Time or phase
0
Back to Basics: Wavelength Formulas
• In a vacuum, wavelength in feet (λft) =

983,571,056.43/fHz, which is the same as λft =
983.57/fMHz
• In coaxial cable with 87% VP, λft = 855,706,819.09 /fHz
or λft = 855.71/fMHz
Back to Basics: An Example
• For instance, the period of a 10 MHz sine wave is 1/10,000,000 Hz =
1x10-7 second, or 0.1 microsecond. That means a 10 MHz signal takes
0.1 µs to complete one cycle, or 1 second to complete 10,000,000
cycles.
• In a vacuum, the 10 MHz signal travels 98.36 ft in 0.1 µs. This
distance is one wavelength in a vacuum.
• In 87% velocity of propagation coax, the 10 MHz signal travels 85.57
ft in 0.1 µs. This distance is one wavelength in coax.
10 MHz signal
completes one
cycle in 0.1 µs
+
Amplitude
Time or phase
0
-
• If we can calculate a given frequency’s wavelength

in feet, we can say that a 100 ft piece of .500 coax is
equivalent to a certain number wavelengths at that
frequency!
• From the previous example, it stands to reason that
a 100 ft piece of coax is equivalent to just over one
wavelength at 10 MHz. That is, the 10 MHz signal’s
85.57 ft wavelength in coax is just shy of the 100 ft
overall length of the piece of coax.
• OK, let’s figure out the wavelength in feet for several

frequencies in a vacuum and in our 100 ft piece of coax, using
the previous formulas. Because of the cable’s velocity of
propagation, each frequency’s wavelength in the cable will be
a little less than it is in a vacuum.
Frequency λft in a vacuum λft in coax
1 MHz 983.57 feet 855.71 feet
5 MHz 196.71 feet 171.14 feet
10 MHz 98.36 feet 85.57 feet
30 MHz 32.97 feet 28.52 feet
42 MHz 23.42 feet 20.37 feet
50 MHz 19.67 feet 17.11 feet
65 MHz 15.13 feet 13.16 feet
100 MHz 9.84 feet 8.56 feet
• Next let’s figure out the number of wavelengths for each

frequency in the 100 ft piece of coax
Frequency λft in a vacuum λft in coax Number of λ

in 100 ft of
coax
1 MHz 983.57 feet 855.71 feet 0.12 λ
5 MHz 196.71 feet 171.14 feet 0.58 λ
10 MHz 98.36 feet 85.57 feet 1.17 λ
30 MHz 32.97 feet 28.52 feet 3.51 λ
42 MHz 23.42 feet 20.37 feet 4.91 λ
50 MHz 19.67 feet 17.11 feet 5.84 λ
65 MHz 15.13 feet 13.16 feet 7.60 λ
100 MHz 9.84 feet 8.56 feet 11.69 λ
10 MHz Sine Wave: A Closer Look
One cycle
Amplitude
Time or phase
90°
135° 45°
180° 0°
225° 315°
270°
Amplitude:
0 0°
90°
135° 45°
180° 0°
Amplitude: 45°
0.707
225° 315°
270°
0°
0.0125 µs
90°
135° 45°
90°
Amplitude:
1
180° 0°
45°
225° 315°
270°
0°
0.0250 µs
90°
135° 45°
90°
180° 0°
Amplitude: 45° 135°

0.707
225° 315°
270°
0°
0.0375 µs
90°
135° 45°
90°
180° 0°
45° 135°
225° 315°
Amplitude:
180° 270°
0 0°
0.0500 µs
90°
135° 45°
90°
180° 0°
45° 135°
225° 315°
180° 270°
0°
0.0625 µs
Amplitude: 225°
-0.707
90°
135° 45°
90°
180° 0°
45° 135°
225° 315°
180° 270°
0°
0.0750 µs
225°
Amplitude:
-1
270°
90°
135° 45°
90°
180° 0°
45° 135°
225° 315°
180° 270°
0°
0.0875 µs
Amplitude: 225° 315°

-0.707
270°
90°
135° 45°
90°
180° 360°
45° 135°
225° 315°
Amplitude:
180° 270°
0 0° 360°
0.1000 µs
Period of
one cycle of
225° 315° a 10 MHz
sine wave is
0.1 µs
270°
• Knowing that one wavelength (cycle) of a sine wave equals

360 degrees of phase, we can now figure out the total number
of degrees of phase the 100 ft piece of cable represents at
each frequency
One cycle
90°
Amplitude
0° 180° 360° Time or phase
270°
Frequency λft in a vacuum λft in coax Number of λ Total

in 100 ft of phase in
coax degrees
1 MHz 983.57 feet 855.71 feet 0.12 λ 42.07°
5 MHz 196.71 feet 171.14 feet 0.58 λ 210.35°
10 MHz 98.36 feet 85.57 feet 1.17 λ 420.71°
30 MHz 32.97 feet 28.52 feet 3.51 λ 1262.12°
42 MHz 23.42 feet 20.37 feet 4.91 λ 1766.96°
50 MHz 19.67 feet 17.11 feet 5.84 λ 2103.53°
65 MHz 15.13 feet 13.16 feet 7.60 λ 2734.58°
100 MHz 9.84 feet 8.56 feet 11.69 λ 4207.05°
• Next, we can plot the 100 ft piece of cable’s phase versus
frequency on a graph. In this example, the line is straight—
that is, phase versus frequency is linear.
4000° x
Phase (total degrees)
3000°
x
2000° x
x
1000°
x
x
0°
0 10 20 30 40 50 60 70 80 90 100 110
Frequency (MHz)
• Finally, we can plot the time delay for each frequency through the
100 ft piece of cable. This line is the derivative of radian phase
with respect to radian frequency. It’s flat, because phase versus
frequency is linear—there is no group delay variation!
400
Time delay (nanoseconds)
300
200
100
0°
0 10 20 30 40 50 60 70 80 90 100 110
Frequency (MHz)
• Another way of looking at this is to say that the

cable’s velocity of propagation is the same at all
frequencies!
• In other words, every frequency takes 116.86 ns to
travel from one end of the 100 ft piece of cable to
the other end.
• But what happens if something in the signal path
causes some frequencies to travel a little slower
than other frequencies?
• Take a look at
the phase
versus
frequency
plot on this
screen shot
• Where phase
is not linear
versus
frequency—
that is, where
the sloped
line is not
straight—
group delay
exists Graphic courtesy of Holtzman, Inc.
• As before, the
group delay
between 25 and 40
MHz is about 300 ns
(3 vertical divisions
at 100 ns each)
• It takes the 40 MHz

signal 300 ns
longer to reach the
headend than the
25 MHz signal
Time difference
in nanoseconds
Why Is Group Delay Important?
• Group delay, a linear distortion, causes inter-

symbol interference to digitally modulated signals
• This in turn degrades modulation error ratio (MER)
—the constellation symbol points get “fuzzy”
Q
Average error power
Average symbol
power
I
MER = 10log(average symbol power/average error power)
• This test equipment

screen shot is from a
cable network’s
upstream spectrum in
which in-channel group
delay was negligible—
about 60 to 75 ns peak-
to-peak.
• Unequalized MER is
28.3 dB, well above the
17~20 dB MER failure
threshold for 16-QAM.
Graphic courtesy of Sunrise Telecom
• In this example, in-

channel group delay
was around 270 ns
peak-to-peak.
• Unequalized MER is
about 21 dB, very
close to the 16-QAM
failure threshold. 16-
QAM would not work
on this upstream!
Graphic courtesy of Sunrise Telecom
MATLAB® simulation for 64-QAM—no group delay
MATLAB® simulation for 64-QAM—with group delay
Wrapping Up
• Common sources of group delay in a cable network

AC power coils/chokes (affects 5~10 MHz in the upstream)
Node and amplifier diplex filters (affect frequencies near
the diplex filter cutoff region in the upstream and
downstream)
Band edges and rolloff areas
High-pass filters, data-only filters, step attenuators, taps or
in-line equalizers with filters
Group delay ripple caused by impedance mismatch-related
micro-reflections and amplitude ripple (poor frequency
response)
Wrapping Up
• The Fix?
Use adaptive equalization available in DOCSIS 1.1 and 2.0
modems (not supported in DOCSIS 1.0 modems)
Avoid frequencies where diplex filter group delay is common
Sweep the forward and reverse to ensure frequency response is
flat (set equipment to highest resolution available; use resistive
test points or probe seizure screws to see amplitude ripple)
Identify and repair impedance mismatches that cause micro-
reflections
Use specialized test equipment to characterize and troubleshoot
group delay (group delay can exist even when frequency
response is flat)
References
• Farmer, J., D. Large, W. Ciciora and M. Adams. Modern Cable

Television Technology: Video, Voice and Data Communications, 2nd
Ed., Morgan Kaufmann Publishers; 2004
• Freeman, R. Telecommunications Transmission Handbook, 4th Ed.,
John Wiley & Sons; 1998
• Hranac, R. “Group delay” Communications Technology, January
1999
www.ct-magazine.com/archives/ct/0199/ct0199e.htm
• Williams, T. “Tackling Upstream Data Impairments, Part 1”
Communications Technology, November 2003
www.ct-magazine.com/archives/ct/1103/1103_upstreamdata.html
References
• Williams, T. “Tackling Upstream Data Impairments, Part 2”

Communications Technology, December 2003
www.ct-magazine.com/archives/ct/1203/1203_upstreamdata2.html
• Hranac, R. “Microreflections and 16-QAM” Communications
Technology, March 2004
www.ct-magazine.com/archives/ct/0304/0304_broadband.html
• Hranac, R. “Linear Distortions, Part 1” Communications Technology,
July 2005
www.ct-magazine.com/archives/ct/0705/0705_lineardistortions.htm
• Hranac, R. “Linear Distortions, Part 2” Communications Technology,
August 2005
www.ct-magazine.com/archives/ct/0805/0805_lineardistortions.htm
Q and A

Group Delay

Uploaded by

Copyright:

Available Formats

Group Delay

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Group Delay

Uploaded by

Copyright:

Available Formats

Group Delay

• Consider the 6 MHz spectrum occupied by an analog

• A signal takes a certain amount of time to pass

• In many instances the velocity of propagation

• The finite time required

• If delay through a filter is plotted on a graph of

Graphic courtesy of Holtzman, Inc.

• If propagation or transit time through a device is

• If propagation or transit time through a device is

• Delay distortion—also known as phase distortion—

Group delay is “the derivative of radian phase with

IEEE Standard Dictionary of Electrical and Electronics Terms

• In its simplest mathematical representation…

• If phase versus frequency is non-linear, group

All of the athletes arrive at the finish line

Some athletes take a little longer

• Group delay exists, because some frequencies—

• That is, it takes

• Recall the group delay analogy using athletes on a

• OK, group delay exists if phase versus frequency is

• Hardline coax used for feeder applications has a

• In a vacuum, wavelength in feet (λft) =

• If we can calculate a given frequency’s wavelength

• OK, let’s figure out the wavelength in feet for several

5 MHz 196.71 feet 171.14 feet

10 MHz 98.36 feet 85.57 feet

30 MHz 32.97 feet 28.52 feet

42 MHz 23.42 feet 20.37 feet

50 MHz 19.67 feet 17.11 feet

65 MHz 15.13 feet 13.16 feet

100 MHz 9.84 feet 8.56 feet

• Next let’s figure out the number of wavelengths for each

Frequency λft in a vacuum λft in coax Number of λ

5 MHz 196.71 feet 171.14 feet 0.58 λ

10 MHz 98.36 feet 85.57 feet 1.17 λ

30 MHz 32.97 feet 28.52 feet 3.51 λ

42 MHz 23.42 feet 20.37 feet 4.91 λ

50 MHz 19.67 feet 17.11 feet 5.84 λ

65 MHz 15.13 feet 13.16 feet 7.60 λ

100 MHz 9.84 feet 8.56 feet 11.69 λ

Amplitude: 45° 135°

Amplitude: 225° 315°

• Knowing that one wavelength (cycle) of a sine wave equals

0° 180° 360° Time or phase

Frequency λft in a vacuum λft in coax Number of λ Total

5 MHz 196.71 feet 171.14 feet 0.58 λ 210.35°

10 MHz 98.36 feet 85.57 feet 1.17 λ 420.71°

30 MHz 32.97 feet 28.52 feet 3.51 λ 1262.12°

42 MHz 23.42 feet 20.37 feet 4.91 λ 1766.96°

50 MHz 19.67 feet 17.11 feet 5.84 λ 2103.53°

65 MHz 15.13 feet 13.16 feet 7.60 λ 2734.58°

100 MHz 9.84 feet 8.56 feet 11.69 λ 4207.05°

• Another way of looking at this is to say that the

• It takes the 40 MHz

Graphic courtesy of Holtzman, Inc.

• Group delay, a linear distortion, causes inter-

Average error power

• This test equipment

Graphic courtesy of Sunrise Telecom

• In this example, in-