ASYMMETRIC DIGITAL SUBSCRIBER LINE APPLICATIONS
A Major Qualifying Project Report:
submitted to the Faculty
of the
WORCESTER POLYTECHNIC INSTITUTE
in partial fulfillment of the requirements for the
Degree of Bachelor of Science
by
Ruben Brito
Maxim Liberman
Drew Merrill
Date: 4/26/2001
Approved:
Professor John McNeill, Major Advisor
II
Table Of Contents
Table of Figures............................................................................................................... IV
Abstract.......................................................................................................................... VII
Abstract.......................................................................................................................... VII
1.0 Introduction................................................................................................................. 1
1.1.1 The ADSL Frequency.............................................................................................. 2
1.1.2 Benefits of ADSL.................................................................................................... 3
1.1.3 Workings of ADSL................................................................................................... 5
2.0 Literature Review....................................................................................................... 8
2.1 DMT............................................................................................................................. 8
2.1.1 Quadrature Amplitude Modulation..................................................................... 14
2.2 Transmission Lines ................................................................................................... 19
2.3 Amplifier.................................................................................................................... 20
2.3.1 AD8016.................................................................................................................... 24
2.3.2 Instrumentation Amplifier.................................................................................... 26
3.0 Methodology and Results ......................................................................................... 28
3.1 Amplifier Specifications ........................................................................................... 28
3.1.1 Current Feedback Amplifier ................................................................................ 28
3.1.2 Power Consumption and Voltage Swing ............................................................. 31
3.1.3 High Current Output............................................................................................. 32
3.1.4 Slew Rate................................................................................................................. 33
3.2 Formula Derivation .................................................................................................. 34
III
3.3 Simulation Results .................................................................................................... 49
3.3.1 Line Driver Data .................................................................................................... 52
4.0 Conclusions................................................................................................................ 58
Bibliography.................................................................................................................... 60
Appendix A Datasheets................................................................................................... 61
IV
Table of Figures
FIGURE 1: ADSL FREQUENCY BANDS .............................................................................. 2
FIGURE 2: ADSL VS. MODEM........................................................................................... 3
FIGURE 3: ADSL CONNECTION [HTTP://WWW.ADSL.COM/ADSL_TUTORIAL.HTML] ................... 4
FIGURE 4: ADSL VS. ISDN AND CABLE MODEM............................................................. 5
FIGURE 5: DISTANCE SPECIFICATION............................................................................... 6
FIGURE 6: DOWNSTREAM AND UPSTREAM CHARACTERISTICS ....................................... 9
FIGURE 7: A SINGLE BIN .................................................................................................... 9
FIGURE 8: UPSTREAM AND DOWNSTREAM DATA........................................................... 10
FIGURE 9: DATA TRANSFER RATE .................................................................................. 11
FIGURE 10: RANGE........................................................................................................... 12
FIGURE 11: QUADRATURE AMPLITUDE MODULATION .................................................. 15
FIGURE 12: VECTOR REPRESENTATION OF A QAM SIGNAL ......................................... 15
FIGURE 13: SINC FUNCTION IN THE TIME AND FREQUENCY DOMAIN........................... 16
FIGURE 14: BINARY AND MULTILEVEL SIGNAL SCHEMES............................................. 17
FIGURE 15: SAMPLE OF 16 QAM CONSTELLATIONS.............................................. 18
FIGURE 16: SELECTION GUIDE........................................................................................ 21
FIGURE 17: SLEW RATE LIMITING.................................................................................. 22
FIGURE 18: SLEW RATE DIAGRAM.................................................................................. 22
FIGURE 19: SFDR POWER RATIO................................................................................... 23
FIGURE 20: INSTRUMENTATION AMPLIFIER................................................................... 26
FIGURE 21: INSTRUMENTATION OF AN AMPLIFIER WITH GAIN OF 1............................. 27
FIGURE 22: CURRENT FEEDBACK AMPLIFIER MODEL [MANCINI 1]............................. 29
V
FIGURE 23: FEEDBACK RESISTOR SELECTION ............................................................... 30
FIGURE 24: CURRENT NEGATIVE AMPLIFIER WITH GAIN OF 1.................................... 30
FIGURE 25: CURRENT NEGATIVE AMPLIFIER WITH GAIN OF 1.................................... 31
FIGURE 26: POWER FILTER CIRCUIT.............................................................................. 32
FIGURE 27: VOLTAGE FOLLOWER FOR THE SLEW RATE TEST.................................... 33
FIGURE 28: VOLTAGE FOLLOWER INPUT ....................................................................... 34
FIGURE 29: VOLTAGE FOLLOWER OUTPUT.................................................................... 34
FIGURE 30: SCHEMATIC FOR ANALYSIS ......................................................................... 35
FIGURE 31: SCHEMATIC FOR ANALYSIS.......................................................................... 38
FIGURE 32: EQUIVALENT CIRCUIT FOR A CURRENT FEEDBACK OPAMP....................... 41
FIGURE 33: SINGLE ENDED AMPLIFIER.......................................................................... 42
FIGURE 34: APPARENT RESISTANCE DERIVATION CIRCUIT.......................................... 45
FIGURE 35: GAIN AND RESISTANCE FOR VOLTAGE FEEDBACK AMPLIFIER................... 48
FIGURE 36: GAIN AND RESISTANCE FOR CURRENT FEEDBACK AMPLIFIER................... 48
FIGURE 37: POSSIBLE SOLUTION..................................................................................... 49
FIGURE 38: SIMULATION SCHEMATIC............................................................................. 50
FIGURE 39: INPUT INTO THE INSTRUMENTATION AMPLIFIER........................................ 50
FIGURE 40: VOLTAGE OUTPUT........................................................................................ 51
FIGURE 41: FFT OF THE OUTPUT ................................................................................... 51
FIGURE 42: FREQUENCY RESPONSE PLOT...................................................................... 52
FIGURE 43: LINE DRIVER ................................................................................................ 53
FIGURE 44: SLEW RATE GRAPH...................................................................................... 53
FIGURE 45: VOLTAGE INPUT AND ITS FFT ..................................................................... 54
VI
FIGURE 46: VOLTAGE OUTPUT........................................................................................ 55
FIGURE 47: FFT OF THE INSTRUMENTATION AMPLIFIER............................................. 55
FIGURE 48: VOLTAGE OUTPUT AFTER THE TRANSMISSION LINE .................................. 56
FIGURE 49: FFT AT THE END OF THE TWISTED PAIR..................................................... 56
FIGURE 50: FREQUENCY RESPONSE OF THE LINE DRIVER ............................................ 59
VII
Abstract
In this project we studied the Asymmetric Digital Subscriber Line (ADSL) driver.
Traditionally, a line driver is terminated with a back termination resistance, which
matches the impedance of the line. This impedance matching insures that there are no
reflections traveling across the line, which would disrupt the actual signal being
transmitted. The problem with traditional back termination methods, is that the circuit
will lose half of its voltage, and subsequently, half of the circuits signal power over this
back termination resistor. In order to compensate for this, one would have to in turn
transmit at twice of the planned signal power. This Major Qualifying Project (MQP)
developed an alternative method to matching the impedance of the line with that of the
line driver. Instead of just matching the line with an equal resistance, this design has an
apparent resistance (what the actual line sees) equal to that of the line, but its actual
resistance is much smaller than the lines. This means that less power will be dropped
across the back termination resistance. This new scheme makes the AD8016 thirty
percent more power efficient, which in essence allows the line driver to transmit data
over longer distances, or transmit the original distance, but at higher data rates due to the
increased signal to noise ratio.
1
1.0 Introduction
Asymmetric Digital Subscriber Line (ADSL) is a relatively new technology
promising to provide Internet access with data rates faster than a conventional 56K
modem. In addition to high data rates, it also provides the ability for a customer to both
have Internet access and talk on the phone at the same time. This technology has
tremendous potential; however, currently there is a major drawback. Based upon several
factors, the current range for ADSL service is a two to three mile radius from the
telephone companys central office, which severely limits the number of customers. If
the range of service for this technology were to be increased, then there would be an
increase in the number of customers.
There are several areas of this technology that can be improved in order to
accomplish this goal. Among these areas are the line drivers used to drive the signal
from the central office to the customer, and then from the customer back to the central
office. By increasing the efficiency of these line drivers we can limit the power lost and
as a result, transmit the Discrete Multi Tone (DMT) signal over greater distances
reaching more potential customers. This MQP will examine Analog Devices AD8016
and AD8017 line drivers in an effort to increase their efficiency and thereby increase the
radius of coverage provided by ADSL
1.1 Background
Asymmetric digital subscriber line or ADSL is a technology that lets Internet
users take advantage of the Internet over current telephone lines. The current telephone
network provides each household with copper twisted cable that usually measures 24 or
2
26 AWG. Currently these cables are used primarily for voice communication. The
purpose of using ADSL technology is to utilize the higher frequency over these current
lines to transmit data.
1.1.1 The ADSL Frequency
Voice communication that is offered by the plain old telephone service (POTS)
operates below 4KHz. Since POTS does not utilize frequencies higher than 4KHz it
makes it ideal to use higher frequencies to transmit Internet data. One of the most
important concerns is to maintain voice communication aspect that the lines were original
intended for. Figure 1 (below) shows the frequency spectrum that ADSL utilizes. ADSL
starts at 30 KHz, which provides 26Khz between the start of ADSL and end of voice
communications. Having a band with such a leeway will assure that there will not be
interference to the current voice communication technology. Also, having this great
leeway will make it easier to design filters to filter out voice frequencies from data
transmitting frequencies.
FIGURE 1: ADSL FREQUENCY BANDS
3
Due to the asymmetric property, ADSL is rapidly becoming extremely popular in
the world of communications for providing highspeed data communication.
Asymmetric transition technique gives user a wide frequency band for downstream data,
and a small frequency band for upstream data. The upstream data is capable of
transmitting 13dBm of power and the downstream is capable of transmitting 20.4dBm of
power. If the average customer uses the Internet for browsing through web pages,
downloading documents, and playing movies, this user requires a much higher
downstream band then an upstream band. The user will be using upstream band to
request information, and then downstream to receive all of the information that was
requested. Almost always the information that is requested will be much greater in size
than the information that was sent. One of the most demanding applications is video.
ADSL is capable of supporting the highest video quality. These super quality videos
utilize anywhere between 3 and 4Mbits/s. ADSL can support transmission rates as fast
as 6Mbits/s using QAM (quadrature amplitude modulation) technology, which will be
discussed later in this document.
1.1.2 Benefits of ADSL
ADSL is very similar to the present modem set up that is available all over the
world. Table 1, below, demonstrates the difference between ADSL and typical modem.
ADSL MODEM
Type Asymmetric Symmetric
Speed
(Current Standard)
2 Mbps 56 Kbps
Direct Connection Yes No
FIGURE 2: ADSL VS. MODEM
4
The main difference between ADSL and modem is symmetry. Since modem is
symmetrical it has the same downstream and upstream speeds. With ADSL the
downstream speed is approximately 10 times faster then upstream. ADSL is also almost
200 times faster then regular modems. In order to see the difference, consider a 10
Megabyte file that is transmitted over modem and then ADSL. Using a 14 Kbps modem
it would take about 95 minutes to download that file. Using a current standard 56K
modem it would take about 25 minutes do download the same file, but using ADSL it
would take less then 10 seconds [Analog Devices ADSL Technology]. There are two
more substantial benefits to ADSL. The first benefit is the direct connection to the
Internet. This means that there is no need to dialing up to Internet service provider (ISP)
and no more busy signals. The second benefit is that a telephone and facsimile can be
used while surfing the web. Since data transmission is on a different frequency then
voice then the Internet and the phone can be used at the same time. As far as the setup is
concerned, ADSL is very similar to modems. Figure 2, below, shows how ADSL is
connected to the Internet.
FIGURE 3: ADSL CONNECTION [HTTP://WWW.ADSL.COM/ADSL_TUTORIAL.HTML]
The only difference between current modems and ADSL is the ADSL modem. The
CORE network in figure 3 represents a high speedswitching network that supports
multiple services from a single network. The current speed standard for this network is
5
3GHz. As one can see ADSL offers enormous amounts of benefits when compared with
other services such as ISDN and cable modems.
Cable modems and ISDN are two of the biggest competitors to ADSL, however
ADSL has more advantages then both of them. Figure 4 shows the difference between
ADSL, cable modem, and ISDN.
ADSL ISDN CABLE MODEM
Type Asymmetric Symmetric Symmetric
Speed
(Current Standard)
2 Mbps 128 Kbps
2.5 Mbps
number of users
Use of phone while
surfing the Web
Yes No Yes
Direct Connection Yes Yes Yes
FIGURE 4: ADSL VS. ISDN AND CABLE MODEM
ADSL is also better then ISDN because ADSL bypasses the telephone switches, which
are already overloaded with enormous number of phone calls [http://www.adsl.com/
adsl_tutorial.html] . Cable modems claim that they have a very fast transmission rate, up
to 35Mbps. The cable modems can have that blazing speed if the customer is the only
one that is surfing in their neighborhood. However, if each cable is supporting multiple
users, the transmission rate will be divided by the number of users. Also, with having
many users utilize the same cable problems such as cross talk arise. With ADSL one user
uses the phone line and can experience fast speeds using the Internet.
1.1.3 Workings of ADSL
The principle of ADSL is very simple; send coded data at a certain frequency.
There are, however, many problems that arise in trying to send so much information at
fast speeds. The process that used to send this data is called Discrete MultiTone or
DMT. DMT are transmitted over phone lines in frequency bands known as bins. Each
6
bin is uniformly spaced every 4.3125kHz. There is 4 kHz of useable bandwidth in every
bin with the additional 312.5Hz being used as a guard band to prevent interbin
interference. Each of the bins contains data that is being transmitted from the user to
central office (CO). There could be as many as 256 different bins. There are many
challenges that arise when trying to transmit DMT over ordinary telephone lines. One
particular challenge is transmitting DMT over certain distances. The distance depends on
the data rate and wire gauge. The following table (Table 3) shows how data rate and wire
gauge affect the distance.
Data Rate (Mbps) Wire Gauge (AWG) Distance (Feet)
1.52 24 18000
1.52 26 15000
6.1 24 12000
6.1 26 9000
FIGURE 5: DISTANCE SPECIFICATION
As one can gather from the table, the distance that DMT can be transmitted is not very
long. In order to transmit longer distance, amplifiers must be used. However, designing
the amplifier is a very difficult task because if there is a slight phase shift, the decoding of
the information in the signal will result in incorrect output. The other challenge is signal
integrity. One way of checking the integrity of the signal is Multitone power ratio
(MTPR). MTPR measures the difference between a DMT band that is empty and a band
that contains QAM signal. QAM is defined as Quadrature Amplitude Modulation (see
figure 11). The QAM signal is a unique signal that occurs within the band. The empty
7
bin might cause distortion in the signal because it doesnt necessarily disappear. This is
the challenge that must be addressed.
8
2.0 Literature Review
In this chapter we will look at the background that is needed to complete this
project. The section will be divided into three parts, which are: DMT, Transmission
lines, and amplifier. The DMT section will describe the signals that are used for ADSL
and the important background information that is needed to understand the workings of
the systems. The transmission line section will focus on how the transmission line
behaves when in the system. The amplifier section will give the insight of the amplifier
and the important characteristics that are needed in order to understand the functions of
the driver.
2.1 DMT
Asymmetric Digital Subscriber Lines (ADSL) utilizes line code in order to
transmit data across copper telephone lines. It had been decided that the best line code
for transmitting data over telephone lines is Discrete Multi Tone (DMT). DMT is a form
of multicarrier modulation as opposed to Quadrature Amplitude Modulation (QAM),
which is a single carrier based modulation. DMT is based in the frequency domain, and
uses multiple narrow band carriers. All of these carriers are transmitted
simultaneously in parallel. Each of these narrow band carriers is known as a sub
channel. There are 256 bins downstream (from the central office to the customer) and 20
bins upstream (from the customer to the central office). A conceptual example of a DMT
signal can be seen in figure 6.
It is the disparity between the number of bins upstream and the number of bins
downstream that makes this form of DSL asymmetric. These subchannels are
9
independently modulated using QAM. The carrier frequency of each of these sub
channels is related to the center frequency of the bin. Both the carrier frequency and
FIGURE 6: DOWNSTREAM AND UPSTREAM CHARACTERISTICS
the center frequency are processed in parallel. A DMT bin consists of 4kHz of useable
bandwidth. In addition to the useable bandwidth, there is a few Hz of bandwidth acting
as a safety buffer on either side of the bin. Figure 7 shows an example of a DMT bin.
FIGURE 7: A SINGLE BIN
F
c
2kHz F
c
F
c
+2kHz
10
Each individual bin of the DMT signal is Quadrature Amplitude Modulated
(QAM). This means that the signal is both amplitude modulated and phase shifted. Each
individual bin can on average transmit 32 kbps. This is because 8kbps can be transmitted
for every hertz of bandwidth that are in a bin. With a bandwidth of 4000Hz, this would
translate to 32kbps. Based on this rate, figure 8 shows the theoretical maximum for
upstream and downstream transmission rates with ADSL.
Transmission
Rate Per
Bin
Number
of
Bins
Transmission
Rate Per
Stream
Upstream 32 kbps 20 640 kbps
Downstream 32 kbps 256 8.2 Mbps
FIGURE 8: UPSTREAM AND DOWNSTREAM DATA
Utilizing these transmission rates, ADSL can currently transmit up to 18,000 ft using
DMT, but it has been proven that DMT still had the rate adaptation to reach beyond
18,000 feet [Baines 15]. From this information it can be seen that the equipment used in
the current system limits the transmission distance. Figure 9 shows several of the
performance specs for DMT systems. There are several advantages to using DMT as the
line code for ADSL.
These are:
DMT is based upon an international standard
Its rate adaptation enables ADSL to be provided over longer ranges
DMT signals are able to adapt to the conditions of the line that they are being
transmitted on
DMT signals are unaffected by line impulses and radio frequency interference
DMT signals have a low power spectral density and are power efficient
11
Performance Specs. For a DMT based System
Maximum Distance 18,000 ft.
Data Transfer Rate (Downstream) 8 Mbps
Data Transfer Rate (Upstream) 768 kbps
FIGURE 9: DATA TRANSFER RATE
The first advantage is that DMT is based upon an international standard. This
standard is known as T1.413 and has been adopted by both American Standards National
Institute (ANSI), European Telecommunications Standards Institute (ETSI), and by the
International Telecommunications unit (ITU). This adoption of T1.413 as an
international standard occurred in 1995. As a result of this international standard, many
private companies within the telecommunications industry have develop developed DMT
technology. In fact, there are a number of applications techniques given within the ANSI
standard for DMT. A major benefit that resulted from the adoption of this international
standard is having modems, based on DMT, will be interoperable. Anyone who would
own a DMT modem, based upon this international standard, would be able to use it
anywhere in the world.
Another feature and benefit to using DMTs is that of rate adaptation. Unlike
other line codes, DMT transitions in steps of 32 kbps. This transition begins at 64 kbps
and travels beyond 8 Mbps. DMTs small transition steps enable ADSL to have a greater
range than line codes that have larger transition steps. Figure 10 gives an example of this.
12
FIGURE 10: RANGE
Using figure 10 as an example, the first system can only provide data rates on the order of
1 or 0, with nothing in between. The second system has a maximum data rate that is the
same as the first, except that it increments in steps that are a quarter the size of the first
system. As a result, the range of the second system is four times the range of the first
system.
With DMT, rural areas can be provided with ADSL. The data rate for ADSL in
some rural areas is as slow as 64 kbps, but when looking at the alternative of no DSL, this
is a considerable improvement. If DMT had larger steps, such as 300 kbps, in an area that
could only get 64 kbps, ADSL would be unavailable.
One of DMTs advantages is that transmitters designed to use DMT signals are
able to adapt to the conditions of the telephone line that it is using. Conditions that are
considered are changing line impedances, noise in the line caused by radio frequency
interference and impulses in the line (which is caused by a phone ringing, or a phone that
is taken off the hook). At any time, the DMT transmitter provides the optimum signal
based on the current given line conditions.
13
Line impulses cause by ringing telephones pose a threat to the DMT signal
because they are very small in the time domain, however in the frequency domain, they
have a very broad band. This broad band can cover multiple DMT sub channels. DMT
avoids interference from impulses by having their DMT signals longer (in the time
domain) than the line impulse.
The second major source of noise for ADSL is radio frequency noise (RFI).
Because both ADSL and AM radio use the 1MHZ band, some of the AM radio signal
leaks into phone lines and poses a threat to the ADSL transmission. DMT deals with
this problem, by avoiding this interference, and applying its signal energy in areas that
will not be affected by this AM signal. DMT can avoid this AM interference, because it
is predictable due to the nature of AM radio broadcasting. Another technique that is used
to protect against interference is error correction coding. This is accomplished by adding
a small percentage of information that is carefully coded. Even this small percentage of
information is able increase communication channel efficiency. Methods of error coding
are mentioned in T1.413.
Another benefit of DMT based systems is its low spectral power density. There
are Power Spectral Density (PSD) requirements in order to prevent different systems
from interfering with one another. For each of these systems, there is a PSD mask.
This mask gives the PSD that a system must operate under in order to not interfere with
other systems. The following are two examples of PSD requirements. On older ADSL
systems, there was an option for a high frequency boost of 26dBm to enable a longer
operating radius. Due to new PSD requirements, this option has been banned. In
14
addition, because of VDSL (Very High Speed Digital Subscriber Line), the rolloff above
1.1 MHZ has been required to be steeper.
DMT is considered to have a low Power Spectral Density (PSD) which means
that DMT systems are very power efficient when compared with other systems. The
amount of power used in the driver is directly dependent on the PSD requirements; i.e.
the amount of power that is put into each frequency band. Due to the fact that DMT
systems are matched to the line they are transmitting on, and they can avoid signal
interference, there is a lower power requirement per frequency band, thus requiring a
lower PSD. As one can see, there are many benefits to using DMT as the line code for
ADSL, whether it is internationally recognized its ability to deal with interference, or its
high power efficiency. One problem, which DMT cannot compensate for, is the power
loss in the line driver. Due to impedance matching, 50 percent of a signals power is lost
across the line driver before it is sent through the transmission line. The goal of this
project is to reduce the power loss by 30 percent.
2.1.1 Quadrature Amplitude Modulation
As stated earlier, each individual bin is quadrature amplitude modulated.
Quadrature Amplitude Modulation (QAM) consists of a signal that is both amplitude
modulated and phase shifted. A block diagram giving the basic theory of QAM is shown
in figure 11.
15
FIGURE 11: QUADRATURE AMPLITUDE MODULATION
The signal consists of two components shifted out of phase by 90 degrees. It is this 90
degree phase shift that makes these two components orthogonal. In fact, the output of the
above system can be treated as a vector. I(t) is the real component and thus on the xaxis
and Q(t) is the imaginary component and as a result, on the yaxis. In addition, I(t)
corresponds to a cosine wave while Q(t) corresponds to a sine wave. Figure 12 shows the
axis representation for a QAM modulated signal.
FIGURE 12: VECTOR REPRESENTATION OF A QAM SIGNAL
16
From this vector, we have the magnitudes of both the real and imaginary components of
the signal as well as the real phase angle of the output of I(t)+Q(t).
The carrier waveform used for QAM modulation is the sinc function, which is:
The reason the sinc function is used, is because in the frequency domain, it is a perfect
rectangle thus maximizing the useable bandwidth. A picture of the sinc function in both
the time and frequency domain can be seen in figure 13. The equations for the sinc
function in the time domain and its transform into the frequency domain can be seen
below.
FIGURE 13: SINC FUNCTION IN THE TIME AND FREQUENCY DOMAIN
(2)
(1)
17
As one can see, the period of the sinc function in the time domain gives us the
bandwidth of the rect function in the frequency domain. Thus, for a rect function
with a bandwidth of 4kHz, we would need a sinc function with a period of 250
microseconds.
The QAM modulation used in DMT is done digitally. The modulated signal is
what is known as a multilevel signal that is sampled at 8 kbps for every Hz frequency.
As was stated earlier, this is where the 32 kbps comes from for a rect function with a
4kHz bandwidth. A multilevel signal is similar to a binary signal except there are more
than two reference voltages. Figure 14 shows examples of schemes for both binary
signals and multilevel signals.
Binary Multilevel Signal
Voltage
Reference (V)
Binary
Representation
Voltage
Reference (V)
Binary
Representation
3 11 0 0
1 10
1 01 1 1
3 00
FIGURE 14: BINARY AND MULTILEVEL SIGNAL SCHEMES
Using the above scheme as shown in Figure 14 for a multilevel signal, if the binary word
1110 were transmitted, its real component I(t) would be 11 which corresponds to 3V and
its imaginary component Q(t) would be 10 which corresponds to 1V. The phase angle for
this signal would be 18.43 degrees above the xaxis and the magnitude of the signal
would be 3.16V. Figure 15 shows the above example.
18
Waveforms for digital signals can be represented as an orthogonal series with N number
of terms using equation 3:
Where w
k
represents the digital data to be transmitted and
k
(t) where k= 1,2N are the
N orthogonal functions that modulate the waveform with the w
k
digital data. N
represents the number of dimensions required to describe the waveform. In the case of
the multilevel signal listed in table 2, there are N=4 dimensions.
FIGURE 15: SAMPLE OF 16 QAM CONSTELLATIONS
The plot of the signal vector in figure 15 is what is known as a signal
constellation. This is a plot of a set of binary words and their corresponding voltages.
Depending upon the number of bits per word and words per quadrant, there are different
QAM schemes. By knowing the number of bits per word, we can tell what QAM scheme
is being used. The formula for determining this is
2
Number of Bits per Word
(3)
Q(t)
I(t) I(t)
Q(t)
1110 0010
1110 0001
3V 3V
1V
1V
19
For example, if there are three bits per word (2
3
), then QAM 8 is being used. A further
example shows that if there are four bits per word (2
4
), then QAM 16 is being
implemented. The scheme used in table 2 and implemented in figure 6 is a version of
QAM 16 because it uses four bits per word. It is even possible for there to be 8 bits per
word resulting in QAM 256.
It is the conditions of the system that determine the best QAM scheme to be used.
In the case of this MQP, the system is a telephone line. The more noise that is present in
a particular bin, the fewer the bits per word and number of words per quadrant can be
used. In this case QAM 4 or QAM 8 would be acceptable. If almost no noise is present,
then QAM 256 would be acceptable if it is available.
As one can see, it is the combination of a multicarrier signal with a single carrier
modulator that makes DMT such an effective line code. Because of this, DMT has both a
long duration in the time domain and a large bandwidth in the frequency domain. If one
were to use just a single carrier signal such as QAM, it would span the time domain, but
would be very short in the frequency domain.
2.2 Transmission Lines
One aspect of this MQP deals with sending signals into a system and then
receiving signals from that system. That system is a telephone line. Therefore it is
important to understand some fundamentals of transmission lines.
A transmission line can be modeled as a circuit with characteristic impedance
related to its inductance and capacitance. Inductance is due to both EM interference as
well as currents traveling through the telephone wires. Capacitance is a result of the
wires insulation. Reflections are dependent upon a transmission lines impedance and
20
method of termination. There are three types of termination for a transmission line.
There can either be an open circuit termination, a short circuit termination, or a
termination with a load. For a more detailed understanding of transmission lines and
transmission line theory, please refer to Ulaby's Fundamentals of Applied
Electromagnetics.
2.3 Amplifier
The heart of the line driver is the amplifier. Choosing an amplifier is not a trivial
task. The amplifier characteristics will determine the bandwidth, power, and noise that
the signal in the circuit will have when amplified through the lines. If the amplifier is not
capable of meeting the specification that are needed, then the entire circuit will not
function properly.
Since the signal is being sent through long transmission lines, which are
susceptible to interference from noise, there is a need for a high current output amplifier
that has excellent Spurious Free Dynamic Range (SFDR) in order to minimize the noise
that is sent into the transmission lines. The amplifier must also have high output voltage
swing for reasons described later on in this section. Also, it would be extremely
beneficial to have an amplifier that has current feedback. The benefits of current feedback
will be discussed later in this chapter. In figure 16, below, we looked 3 different
amplifiers that Analog Devices manufactures. The Amplifiers are: AD8016, AD8017,
and AD8018. These amplifiers were chosen because they belong to the High Output
Current Drive family.
21
Part
Number
Current
Feedback
Supply
Voltage (V)
BW@ Gain
of 1 (MHz)
Slew Rate
(V/ s)
SFDR
(dBc)
Noise
(nV/ Hz)
I
out
(mA)
AD8016 Yes 12 320 1000 77 2.8 600
AD8017 Yes 5 160 1500 76 1.9 220
AD8018 Yes 5 120 300 N/A 4.5 400
FIGURE 16: SELECTION GUIDE
To choose the best amplifier we looked at the specifications in the table. The first
thing that we considered was current feedback. Current feedback amplifiers are used for
high gain and highspeed applications. One of this opamp's advantages is that its
closedloop gain can be changed, when used in a feedback application, without
significantly affecting its loop gain (Johns and Martin 1997). Having this kind of
amplifier is extremely beneficial because there is no longer a worry that raising or
lowering the amplifiers gain will tremendously affect the bandwidth. The current
feedback amplifier will be discussed further in the later part of this chapter.
The next spec that we looked at was the supply voltage. Since we are sending the
signal over long distances and the lines will attenuate the signal, having a large voltage
swing will assure that the signal will be able to travel further. In this case it is clear that
the AD8016 has the largest (24 Volts) voltage span.
The third specification that we considered was bandwidth. The bandwidth of
ADSL is 1.1 MHz, so at the first look all of the amplifiers seem perfect, however, the
bandwidth is measured at gain of 1. Even though this is a current feedback amplifier and
the bandwidth is not affected by the gain, as much as a voltage feedback amplifier would
be, if the choice exists we should still utilize the opamp that has the highest bandwidth.
This will also help limit distortion and help with SFDR.
22
Next specification that we considered was slew rate. Slew rate is defined as
change in voltage divided by the change in time (slope). Figure 17, demonstrates slew
rate limiting.
FIGURE 17: SLEW RATE LIMITING
Slew rate is an extremely important characteristic that must be considered when
designing amplifiers. Slew rate is especially important when dealing with high frequency.
To better understand how slew rate can affect an amplifier we can consider the following
example. An amplifier with a gain of 1 with a input square wave at 1 MHz and slew rate
of 6 V/s. Figure 18 shows what the input (sine wave) and output would look like.
FIGURE 18: SLEW RATE DIAGRAM
Input Step Waveform Slew Rate Limited Output
23
As one can see the slew rate can cause tremendous problems. If we were running a
digital system and a digital high would be any voltage over 1 volts then the system would
never hit a logic high even though by the looking at the input one would imagine that
there would not be any problems with having a logic high. In choosing the amplifier the
AD8017 clearly has the best slew rate specification.
The next specification that we will consider is the Spurious Free Dynamic Range
(SFDR). SFDR is simply "the ratio of the fullscale input or output signal to the highest
harmonic or spurious input/output noise component amplitude. Essentially, this is an
indication of how far it is possible to go below the fullscale input signal without hitting
noise or distortion [Analog Devices].
FIGURE 19: SFDR POWER RATIO
Figure 19, demonstrates SFDR. The figure shows a signal that has a data bin eliminated
at the 554.3 kHz frequency. In this case the SFDR is 75dBc. It is extremely important
that this specification is very good because when the data gets to the receiving end and
there is a small bin at that particular frequency then the receiver will interpret it as
information and the it might give the output a wrong set of data. The AD8016 has the
best SFDR in this class of amplifiers.
24
The second to last thing that we will look at is noise. The noise is given by
nV/ Hz. The noise is simply the distortion that will occur on the transmission wave.
The circuit can usually tolerate the noise but if the noise becomes in excess then that data
might be defective. The noise is dependent on frequency and each resistor can add to the
noise. The resistors have a voltage noise of 4kTBR, where k is the Boltzmans constant,
T is temperature, B is bandwidth, and R is resistance. Therefore if the noise gets too
large we can adjust it by varying those parameters. For the noise specification the
AD8017 has the best specification of 1.9 nV/ Hz.
The last specification that we will look at is current output. Since we are using a
transformer to deliver the signal to the transmission line we need a high a current output
to have an excellent signal at the output of the transformer. We know that to get the best
signal we need a high voltage swing. We also know that in the transformer the power
input and the power output are going to be equal.
P
input
=(V
input
)(I
inpu
t)
P
output
=(V
output
)(I
output
)
P
input
=P
output
(V
input
)(I
inpu
t)=(V
output
)(I
output
)
From the equation above it can clearly be seen that if there is a high output current from
the amplifier to the input of the transformer, one can have a high voltage at the output of
the transformer feeding transmission line. As one can see having a high current is very
advantageous and AD8016 by far has the highest current output.
2.3.1 AD8016
The AD8016 is a high output current driver that we will be using to figure out the
best possible scenario to transmit DMT signal over existing telephone lines. This
25
amplifier gives some of important characteristics that we are looking in order to transmit
the signal. These characteristics are:
High output current
Low power dissipation
High output voltage swing
Flexible powerdown
Lower distortion
Having these characteristics the AD8016 is a perfect match for the ADSL line driver task.
From the table 41 we saw that AD8016 had the highest specification that we need in
most categories. The only two categories that it did not have the best specifications is
noise and slew rate. Since we only running at a maximum of 1.1MHz, the slew rate that
AD8016 is rated for is more then perfect. The noise aspect we can regulate by either
adjusting components that create noise or setting up filters to get rid of the noise.
The AD8016 is also a transimpedance (currentfeedback) opamp. The most
important aspect of this kind of an amplifier is bandwidth remains relatively constant at
gain changes. This is a great option over voltage feedback amplifiers because there is no
worry that the bandwidth when trying to achieve large gain. However, the disadvantage
to having a current feedback amplifier is that it will have a bit more noise. Typically the
current has more noise then the voltage and the amplifier will amplify that voltage.
26
2.3.2 Instrumentation Amplifier
We will be using two AD8016 amplifiers to create an instrumentation amplifier
circuit to transmit the signals across the transmission line. The instrumentation amplifier
FIGURE 20: INSTRUMENTATION AMPLIFIER
provides a few very important characteristics that we will need to use in order to transmit
the data across a transmission line. The characteristics that it provides are:
Rejects Large Interference Signal
Provides Differential Input/Output
This instrumentation amplifier circuit can be found in figure 20. This circuit is the main
component of the line driver.
The gain for this amplifier can be found from the following equations:
V
1
=+V
in
V
2
=V
in
+V
O
=+V
in
+R
1
((V
1
V
2
)/ R
2
)
V
O
=V
in
+R
3
((V
2
V
1
)/ R
2
)
Full Swing=(V
1
V
2
)(1+((R
1
+R
3
)/R
2
)
R2
R3
R1
27
As one can see from equations above the signal that will be amplified is the difference in
voltage over R2. Figure 21, below, shows how a circuit with R
1
=R
2
=R
3
=1000 will
behave:
FIGURE 21: INSTRUMENTATION OF AN AMPLIFIER WITH GAIN OF 1
One can clearly see that instrumentation amplifier can give us the differential signal that
we are looking for as well as the large voltage swing. This set up will be the heart of the
line driver.
28
3.0 Methodology and Results
In this section of the results we will look at the steps that were taken in order to
complete this projects and the results that these steps provided. This section is broken up
into three parts, which are: amplifier specifications, formula derivational and simulation
results. The amplifier specification sections deals with the proper way of using the
amplifier, which is AD8016 in this case. In the formula derivation section we will
discuss the derivation of the formula, gain, and apparent resistance, for AD8016 driver
and the results that these formulas provided. Finally, we will look at simulation results
and see how close the simulation matched the formula results. The actual results will be
described in the conclusion section.
3.1 Amplifier Specifications
This section will be devoted to looking at the specification of the amplifier and
the results received from simulations and analysis. This section is divided into three
parts. The first part is on the current feedback specification of the analysis and how it
benefited the design. The second part will discuss the incredibly high current output of
the amplifier. The third part of this section will be devoted to other important aspects of
this particular amplifier such as proper power, high voltage swing, and high slew rate.
The signal transmitting aspects of the amplifier will be discussed in the test procedure
section. All of these aspects are key to excellent amplifier design.
3.1.1 Current Feedback Amplifier
The most central aspect of this amplifier is that it is operated using current
feedback, also known as transimpedance amplifier. Current feedback amplifiers "are a
29
special breed of operational amplifiers that exploit clever circuitry topologies in
conjunction with highspeed complementary bipolar process to achieve extremely fast
dynamics, such as gainbandwidth products in the gigahertz range and settling times to
0.1% in the 10ns range [Franco 3]. In this type of an amplifier the bandwidth and gain
are not linearly proportional, therefore this will be very effective because there will be
very little concern for gain while looking at high bandwidths of about 1.1MHz.
The current feedback amplifier that will be modeled is seen below figure 22.
FIGURE 22: CURRENT FEEDBACK AMPLIFIER MODEL [Mancini 1]
The amplifier functions very similar to the voltage feedback amplifier. The amplifier
will try to force the inverting input to follow the noninverting input. This is done by the
use of the voltage buffer. The difference in voltage at an instance of time will create a
current (I) because of the resistor (Z
B
). The Z
B
for the AD8016 is approximately 25
ohms. This current (I) will be multiplied by the transimpedance of the amplifier to
produce a gain at the output.
In an open loop configuration the current will be multiplied by the Z, which will
yield an enormous voltage at the output, similar to the voltage feedback amplifier. When
there is a closed feedback loop the amplifier will act just like the voltage feedback
30
amplifier and adjust itself to appropriate voltage as set by the ratio of feedback and input
resistors. As opposed to the voltage feedback amplifier, the resistors for the current
feedback AD8016 must be carefully chosen. The feedback resistor for the AD8016 is
responsible for the stability of the amplifier. Figure 23 shows the resistor values that
were specified by Analog Devices in order to make the AD8016 stable.
Gain R
Feedback
R
Input
+1 1000 Infinite
1 500 500
+2 650 650
+5 750 187
+10 1000 111
FIGURE 23: FEEDBACK RESISTOR SELECTION
Choosing the appropriate resistor values also will assure a better Multitone Power Ratio.
Prior to applying the solution to a specific design, the amplifier was tested in a
simple inverting gain amplifier scheme. The schematic for this amplifier can be seen in
figure 24. As one can see the gain of the amplifier is 1 therefore the resistor values have
been chosen from figure 23.
FIGURE 24: CURRENT NEGATIVE AMPLIFIER WITH GAIN OF 1
The results from this simulation can be seen in figure 25. As we expected from the
literature the
31
FIGURE 25: CURRENT NEGATIVE AMPLIFIER WITH GAIN OF 1
The circuit behaves similarly to a voltage feedback amplifier. One thing that must be
considered when the voltages are different between the inputs of the amplifier the current
that is flowing will by multiplied by the transimpedance of the amplifier (Z).
In summary, the current feedback amplifier will be used because the gain
bandwidth product is very high, typically in the gigahertz range. Also, the gain
bandwidth product is not linear, as opposed to voltage feedback amplifier; therefore a
high frequency signal can be transmitted and not be concerned about the gain bandwidth
product limitation. The amplifier behaves similarly to the voltage amplifier therefore the
same analysis technique will be used to derive the gain and apparent resistance equations.
3.1.2 Power Consumption and Voltage Swing
In order to power up the AD8016 we will be using 12 Volt rails. Since the board
will be operating at high frequencies, the signal to the amplifier must be clean of noise or
voltage spikes that may occur on the power lines. In order to accomplish this task, the
circuit recommended by Analog Devices was used. (Figure 26)
32
FIGURE 26: POWER FILTER CIRCUIT
By utilizing this circuit, the power that is being transmitted into the amplifier was clean
and did not have any voltage spikes.
It was very important to have very high voltage rails to produce a high voltage
swing. In this case the voltage swing can be as high as 10 volts (AD8016 spec sheet).
Having a large voltage swing at the output will clarify the transmission signal. One of
the tests is the signal to noise ratio to establish the quality of the signal having a large
signal will create better results.
3.1.3 High Current Output
The next step was to make sure that this amplifier could provide a high current.
The high current is extremely important because a higher current will cause a higher
voltage swing over the load resistor of 50 ohms. The circuit being used can be seen in
figure 24. Since we want the voltage swing as high as possible, the amplifier must
source enough current.
In order to make sure everything will work properly the worstcase scenario will
be analyzed. The AD8016 can source 600mA of current as per the Analog Devices spec
sheet. The highest voltage swing possible will be 10 volts, and the load resistance will
be approximately 50 ohms. Using this information the load is found to draw 200mA of
33
current. The feedback loops have very high resistances and will draw some current, but
not significant enough limit the amplifier. In looking at the simulation the results showed
that the amplifier can provide 323mA of current to the load, in the circuit design, when
the amplifier is at its maximum, which is sufficient for the 200mA that is expected.
3.1.4 Slew Rate
The last thing that we looked at in the amplifier that does not deal with signal
quality is the slew rate. The slew rate is important because it will confirm that the
amplifier is responsive enough to the signal. In order to verify the slew rate, it was
measured with the circuit in figure 27. The circuit is just a simple voltage follower, but it
will satisfy the need to check slew rate.
FIGURE 27: VOLTAGE FOLLOWER FOR THE SLEW RATE TEST
In order to check the slew rate a 10V peaktopeak wave was transmitted at 560 kilohertz.
The reasoning behind using a 560kilohertz will be described in the test procedures. The
input of the amplifier can be seen in figure 28 and the output of the amplifier can be seen
in figure 29 below.
34
FIGURE 28: VOLTAGE FOLLOWER INPUT
FIGURE 29: VOLTAGE FOLLOWER OUTPUT
In the plots above each block represents 2 Volts per division. In order to measure the
slew rate, the rising slope of the output was observed. From the output the slew rate was
measured to be approximately 650 V/s. This slew rate is excellent and this shows that
the amplifier does meet and exceed the demands that are needed for the task.
3.2 Formula Derivation
In this section, the formulas for the transfer function and the apparent output
resistance of our solution will be derived. The derivations will be completed based on a
T
V
35
single ended amplifier and then using symmetry we can combine two single ended
amplifiers to demonstrate the differential solution. Figure 30 below shows the schematic
for single ended version of the solution.
FIGURE 30: SCHEMATIC FOR ANALYSIS
In order to analyze this schematic, we first used a voltage feedback operational amplifier,
because there are assumptions that can be made with these types of amplifiers that will
simplify the initial analysis. The assumptions we made are as follows:
1. The input impedance of both terminals of the opamp approaches infinity
2. The voltage at both inputs of the opamp are approximately equal (V
+
= V

)
Using a UA741 (Texas Instruments) opamp, the above assumptions were proven to be
accurate for analysis.
The first step in the analysis is to perform nodal analysis at nodes V
N
and V
OUT
.
At node V
N
i
1
(The current through R
1
)
= i
1
(The current through R
2
)
(Due to the infinite
input impedance at the negative terminal)
36
And at node V
OUT:
i
S
= i
3
+ i
4
Now, we use KVL in order to account for all of the voltage drops between V
N
(At V

) and
V
OUT
. This is seen in the following equation.
OUT S S N
V R i R i V
2 1
We know that i
S
= i
3
+ i
4
,
( )
OUT S N
V R i i R i V +
4 3 2 1
The next step in this analysis is to develop equations for i
1
, i
3
, and i
4
( )
1
1
R
V V
i
N IN
3
3
) (
R
V V
i
N OUT
L
OUT
R
V
i
4
Plugging these back into the KVL equation, we get:
OUT S
L
OUT
S
N OUT N IN
N
V R
R
V
R
R
V V
R
R
V V
V
3
2
1
) ( ) (
We then expand the equation by getting rid of all brackets:
OUT
L
S
OUT
S
N
S
OUT N IN N
V
R
R
V
R
R
V
R
R
V
R
R
V
R
R
V V + +
3 3 1
2
1
2
We then recombine using different like terms to get the equation:
1
]
1
+ +
1
]
1
+ +
3 1
2
3 1
2
1 1
R
R
R
R
V
R
R
V
R
R
R
R
V
S
L
S
OUT IN
S
N
(6)
(7)
(8)
(10)
(9)
(11)
(12)
(13)
(14)
37
By using a voltage divider, we know that:
1
]
1
3 4
4
R R
R
V V
OUT N
This can now be plugged into the previous equation to get:
1
]
1
+ +
1
]
1
+ +
1
]
1
+
3 1
2
3 1
2
3 4
4
1 1
R
R
R
R
V
R
R
V
R
R
R
R
R R
R
V
S
L
S
OUT IN
S
OUT
By rearranging the equation, we now get:
1
2
3 3 1
2
3 4
4
1 1
R
R
V
R
R
R
R
R
R
R
R
R R
R
V
IN
S
L
S S
OUT
1
]
1
1
]
1
+ +
1
]
1
+ +
1
]
1
+
This results in our transfer function:
1
]
1
1
]
1
+ +
1
]
1
+ +
1
]
1
3 3 1
2
3 4
4
1
2
1 1
R
R
R
R
R
R
R
R
R R
R
R
R
V
V
S
L
S S
IN
OUT
The second part of formula derivation consists of finding the apparent output impedance
of the circuit. In order to accomplish this, we detach the Load Resistance R
L
and in its
place, we attach a voltage V
X
, such that if we find the amount of current it draws (i
X
), we
can find the apparent output impedance by dividing V
X
by i
X
(Ohms law) In addition,
V
in
is shorted, because it is an independent voltage source. Figure 31 shows the
schematic that was used to find the apparent output impedance.
(15)
(16)
(17)
(18)
38
FIGURE 31: SCHEMATIC FOR ANALYSIS
The first step is to perform nodal analysis at V
x
i
X
= i
1
+ i
2
We then develop the equations for each of the currents listed in the circuit.(With the
exception of i
X
)
3
1
R
V V
i
N X
S
X
R
V V
i
1
2
2
1
3
R
V V
i
N
Now, we plug the equations for the individual currents into the nodal equation for node
V
X
3
1
3 3 3
1 1
3
1 1
R
V
R
V
R R
V
R
V
R
V
R
V
R
V
R
V V
R
V V
i
N
S S
X
N X
S S
X
S
X N X
x
1
]
1
+ +
Again, by using voltage division, we know that:
1
]
1
3 4
4
R R
R
V V
X N
(19)
(20)
(21)
(22)
39
We can now plug this back into the equation for i
x
:
1
]
1
+
1
]
1
+
3 4
4
3
1
3
1 1
R R
R
R
V
R
V
R R
V i
X
S S
X x
In order to complete this analysis, we need to find an expression for V
1
in terms of V
x
.
In order to do this, we create an expression using KVL that travels from V
1
through R
2
and R
1
all the way to ground.
0
1 3 2 3 1
R i R i V
Therefore
( )
2 1 3 1 3 2 3 1
R R i R i R i V + +
By plugging in the equation for i
3
we get
2
1 2
2
1 2
1
2
1 2
1 1
) ( ) ( ) (
) (
R
R R
V
R
R R
V
R
R R
V V V
N N
+
+
By substituting for V
N
we get
2
1 2
3 4
4
2
1 2
1 1
) ( ) (
R
R R
R R
R
V
R
R R
V V
X
+
1
]
1
+
Now that everything is either in terms of V
1
or V
X
, we can solve for V
1
1
]
1
+
+
1
]
1
1
) (
) (
2
1 2
2
1 2
3 4
4
1
R
R R
R
R R
R R
R
V
V
X
In an attempt to make this expression less awkward, we have simplified it. By
multiplying both the top and the bottom of the equation by R
2
we get the final result for
V
1:
X
V
R
R
R R
R R
V
1
4
4 3
1 2
1
) (
) (
+
+
(23)
(24)
(25)
(26)
(27)
(28)
(29)
40
We then plug this back into the equation for i
X
:
1
]
1
1
]
1
+
+
1
]
1
+
3 4
4
3 1
4
4 3
1 2
3
) (
) ( 1 1
R R
R
R
V
R
R
R R
R R
R
V
R R
V i
X
S
X
S
X x
By separating this equation into i
X
and V
X
we get:
1
]
1
1
]
1
1
]
1
+
+
1
]
1
+
3 4
4
3 1
4
4 3
1 2
3
1
) (
) ( 1 1 1
R R
R
R R
R
R R
R R
R R R
V i
S S
X x
Now by using Ohms law, we get the apparent output resistance:
1
]
1
1
]
1
1
]
1
+
+
1
]
1
+
3 4
4
3 1
4
4 3
1 2
3
1
) (
) ( 1 1 1
1
R R
R
R R
R
R R
R R
R R R
i
V
R
S S
X
X
X
By both testing these results on an actual voltage feedback operational amplifier and
simulating the circuit in PSPICE, we were able to conclude that equations for gain and
apparent output resistance are accurate for a voltage feedback opamp.
Although these equations were accurate for a voltage feedback opamp, the AD
8016 is a current feedback opamp. With a current feedback opamp, the impedance at
the inverting input does not approach infinity; in fact the input resistance to this terminal
is very small (approximately 25 ohms). This negates the first assumption that was made
when we performed analysis on the voltage feedback opamp. The low input resistance at
V

allows a current I
N
to flow into the negative terminal. The output voltage of the 8016
is dependent upon this I
N
multiplied by a transimpedance G. G by nature is very large
(approximately 100,000) this can be summed up in the following equation.
G I V
N OUT
Figure 32 shows an equivalent circuit for a current feedback operational amplifier.
(30)
(31)
(32)
41
FIGURE 32: EQUIVALENT CIRCUIT FOR A CURRENT FEEDBACK OPAMP
The second step of our analysis was to derive the equations using the model for
the current feedback operational amplifier. There are a few extra items that need to be
taken into account when analyzing a current feedback operational amplifier. Figure 32
shows the equivalent circuit that is used in the analysis of a current feedback opamp.
From this model, it can be seen that current flows into the positive terminal of the opamp
and that current flows out of the negative terminal of the opamp. This is different from
our assumptions that were made for the voltage feedback opamp.
In addition, it is the current I
N
when multiplied by the transimpedance G that
gives us output voltage V
OUT
. By nature, G is very large. As a result, I
N
must be very
small in order to avoid hitting the power supply rails. The final difference that is to be
noted here is that the input resistance into the positive terminal of the opamp is not
infinity. It is in fact 50k. The input resistance of the negative terminal of the op amp
can be neglected. During the course of our analysis, we will refer to figure 33.
42
FIGURE 33: SINGLE ENDED AMPLIFIER
We begin the analysis by writing equations for the currents given in Figure 33
1
1
R
V V
i
N in
,
2
1
2
R
V V
i
in
,
z
V
i
1
3
,
3
4
R
V V
i
N out
,
L
out
R
V
i
5
,
S
out
S
R
V V
i
1
Now, using KCL, we generate node equations involving each of the currents.
i
2
=i
1
+i
3
and i
S
=i
4
+i
5
and
(33)
(34)
(35)
(36)
(37)
(38)
(39) (40)
43
The next step is to write an expression for V
N
.
1
1
]
1
3
R R
R
V V
eq
eq
OUT N
Where
in
in
eq
R R
R R
R
+
4
4
Where R
in
is the input resistance of the positive terminal.
Now using KVL, we will write an expression for V
out
starting at V
N
.
out S S N
V R i R i V
2 2
Now substituting in the KCL equations we get the following:
out S N
V R i i R i i V + + ) ( ) (
5 4 2 3 1
Finally we substitute in our equations for the individual currents:
out S
L
out N out N in
N
V R
R
V
R
V V
R
z
V
R
V V
V
1
]
1
1
]
1
3
2
1
1
In order to simplify this expression, we multiply it out and then we recombine the like
terms
out out
L
S
N
S
out
S
N in N
V V
R
R
V
R
R
V
R
R
V
z
R
V
R
R
V
R
R
V + +
3 3
1
2
1
2
1
2
By recombining using like terms, we get the following:
1
]
1
+ +
1
]
1
+ + +
L
S S
out
S
N in
R
R
R
R
V V
z
R
R
R
R
R
V V
R
R
3
1
2
3 1
2
1
2
1 1
Using equivalent expression for V
N
we get:
1
]
1
+ +
1
]
1
+ +
1
1
]
1
+
+
L
S S
out
S
eq
eq
OUT in
R
R
R
R
V V
z
R
R
R
R
R
R R
R
V V
R
R
3
1
2
3 1
2
3 1
2
1 1
Now almost everything is in terms of V
in
and V
out
except for V
1
. We need to come up
with an equivalent expression that uses either V
out
or V
in
.
S S out
R i V V +
1
(41)
(42)
(43)
(44)
(45)
(46)
(47)
(48)
(49)
44
Substituting using KCL
S out
R i i V V ) (
5 4 1
+ +
Now we plug in the expressions for the individual currents
S
L
out N out
out
R
R
V
R
V V
V V
1
]
1
+
3
1
multiplying out and recombining we get:
N
S
L
S S
out
V
R
R
R
R
R
R
V V
3 3
1
1
1
]
1
+ +
Which becomes:
1
1
]
1
1
]
1
+ +
3 3 3
1
1
R R
R
R
R
V
R
R
R
R
V V
eq
eq
S
OUT
L
S S
out
We now plug this back into the main equation
1
]
1
+ +
1
1
]
1
1
1
]
1
1
]
1
+ +
1
]
1
+ +
1
1
]
1
+
+
L
S S
out
eq
eq
S
L
S S
OUT
S
eq
eq
OUT in
R
R
R
R
V
R R
R
R
R
R
R
R
R
V
z
R
R
R
R
R
R R
R
V V
R
R
3 3 3 3
2
3 1
2
3 1
2
1 1 1
We now multiply out again and recombine using like terms
1
1
]
1
1
]
1
+ +
1
1
]
1
1
1
]
1
1
1
]
1
1
]
1
+ + +
1
]
1
+ +
3 1
2
3 3 3 3
2
3 1
2
1 1 1
R
R
R
R
R R
R
R R
R
R
R
R
R
R
R
z
R
R
R
R
R
V V
R
R
S
eq
eq
eq
eq
S
L
S S
L
S S
out in
We now rearrange the expression one last time in order to get V
out
over V
in
1
1
]
1
1
]
1
+ +
1
1
]
1
1
1
]
1
1
1
]
1
1
]
1
+ + +
1
]
1
+ +
3 1
2
3 3 3 3
2
3
1
2
1 1 1
R
R
R
R
R R
R
R R
R
R
R
R
R
R
R
z
R
R
R
R
R
R
R
V
V
S
eq
eq
eq
eq
S
L
S S
L
S S
in
out
The final step in the analysis of the 8016 was to derive an equation for apparent
resistance. The equation for apparent resistance with the current feedback amplifier is
(50)
(51)
(52)
(53)
(54)
(55)
(56)
45
similar to the analysis for the voltage feedback amplifier. Figure 34 shows the schematic
used for the 8016s apparent resistance analysis.
FIGURE 34: APPARENT RESISTANCE DERIVATION CIRCUIT
We begin the analysis by using KCL.
2 1
i i i
x
+
5 4 3
i i i
We then create expressions for each of the currents. We will use only i
1
and i
2
initially.
S
x
R
V V
i
1
1
3
2
R
V V
i
N x
Now, we plug these into the KCL equation for i
x
.
3
1
R
V V
R
V V
i
N x
S
x
x
+
(57)
(58)
(59)
(60)
(61)
46
We also can create an equation for V
N
in terms of V
x
.
1
1
]
1
eq
eq
X N
R R
R
V V
3
Now, we multiply out all of the terms in the expression for i
x
and add in the new
expression for V
N
.
After this is accomplished, we get the following equation.
( )
S eq
eq
S
X x
R
V
R R R
R
R R
V i
1
3 3 3
1 1
1
1
]
1
+
+
We now have to derive an equation for V
1
. In order to accomplish this, we need to derive
the KCL equation for i
3
, because i
3
times the transimpedance is equal to V
1
. The KCL
equation for i
3
is as follows.
5 4 3
i i i
We now create individual expressions for i
4
and i
5
.
2
1
4
R
V V
i
N
1
5
R
V
i
N
As was stated earlier, we know that
3 1
i z V
We now combine the above equations to get.
1 2
1
3
R
V
R
V V
i
N N
(62)
(63)
(64)
(65)
(66)
(67)
(68)
47
We then plug this into equation for V
1
.
1
]
1
1 2
1
1
R
V
R
V V
z V
N N
Now we multiply out and recombine using like term. In addition, we will use an alternate
expression for V
N
. After we do that, we get a final expression for V
1
1
]
1
1
]
1
+
1
1
]
1
2
1 2 3
1
1
1 1
R
z
R R R R
R
V z
V
eq
eq
X
We now plug the new expression for V
1
back into the expression for i
x
.
( )
1
]
1
1
]
1
+
1
1
]
1
+
1
1
]
1
+
+
2
1 2 3
3 3 3
1
1 1
1 1
R
z
R R R R
R
R
V z
R R R
R
R R
V i
eq
eq
S
X
eq
eq
S
X x
We now divide V
X
by i
X
in order to get the apparent resistance.
( )
1
1
1
1
1
]
1
1
]
1
1
]
1
+
1
1
]
1
+
+
1
1
]
1
+
+
2
1 2 3
3 3 3
1
1 1
1 1
1
R
z
R R R R
R
R
z
R R R
R
R R
i
V
R
eq
eq
S eq
eq
S
X
X
apparent
Figure 35 and 36 summarizes these results.
(69)
(70)
(71)
(72)
48
Voltage Feedback Amplifier
Gain
1
]
1
1
]
1
+ +
1
]
1
+ +
1
]
1
+
3 3 1
2
3 4
4
1
2
1 1
R
R
R
R
R
R
R
R
R R
R
R
R
S
L
S S
Apparent Resistance
1
]
1
1
]
1
1
]
1
+
+
1
]
1
+
3 4
4
3 1
4
4 3
1 2
3
1
) (
) ( 1 1 1
1
R R
R
R R
R
R R
R R
R R R
S S
FIGURE 35: GAIN AND RESISTANCE FOR VOLTAGE FEEDBACK AMPLIFIER
Current Feedback Amplifier
Gain
1
1
]
1
1
]
1
+ +
1
1
]
1
1
1
]
1
1
1
]
1
1
]
1
+ + +
1
]
1
+ +
3 1
2
3 3 3 3
2
3
1
2
1 1 1
R
R
R
R
R R
R
R R
R
R
R
R
R
R
R
z
R
R
R
R
R
R
R
S
eq
eq
eq
eq
S
L
S S
L
S S
Apparent
Resistance
( )
1
1
1
1
1
]
1
1
]
1
1
]
1
+
1
1
]
1
+
+
1
1
]
1
+
+
2
1 2 3
3 3 3
1
1 1
1 1
1
R
z
R R R R
R
R
z
R R R
R
R R
eq
eq
S eq
eq
S
FIGURE 36: GAIN AND RESISTANCE FOR CURRENT FEEDBACK AMPLIFIER
It can be noted that with a current feedback amplifier, as the transimpedance goes to
infinity, its behavior (Gain, Apparent Resistance) are the same as that for a voltage
feedback amplifier. This can be seen in Figure 37.
49
FIGURE 37: POSSIBLE SOLUTION
The Current Feedback column on the left side of the table is using current feedback
equations with a transimpedance (z) of 100,000. The a in both current feedback columns
is simply the summation of R
3
and R
eq
to aid in copying the equations into Microsoft
excel. The middle two columns are for the voltage feedback amplifier and the final two
columns are for the current feedback amplifier with the transimpedance approaching
infinity. Note how its gain and apparent resistance are exactly that of the voltage
feedback amplifier.
3.3 Simulation Results
In order for us to predict how the circuit is going to behave, we conducted several
simulations to evaluate the circuits performance. PSpice Schematics will be used to
simulate the circuit. The circuit that will be simulated can be found in figure 38, below.
Current Feedback Voltage Feedback Current Feedback
R1 600 R1 600 R1 600
R2 500 R2 500 R2 500
R3 40000 R3 40000 R3 40000
Rs 15 Rs 15 Rs 15
RL 50 RL 50 RL 50
Req 25000 Req 25000 Req 25000
z 100000 z 1E+102
a 65000 a 65000
Gain1 1.385186475 Gain 1.400318842 Gain1 1.400318842
R_apparent 51.44752653 Rapparent 50.82978539 R_apparent 50.82978539
50
FIGURE 38: SIMULATION SCHEMATIC
As one can see the circuit is an instrumentation amplifier. Each individual amplifier
(AD8016) has a gain of 5. The sources voltages are 2 V peak to peak with a frequency of
750KHz. The first thing that we will look at is the voltage input. We will use 2 sine
waves 180 degrees out of phase. This plot can be seen in figure 39.
FIGURE 39: INPUT INTO THE INSTRUMENTATION AMPLIFIER
51
We will now consider the output voltage, FFT of the output voltage and the frequency
response of the circuit. The output voltage can be found in Figure 40. The most
important thing that we must make sure is that the voltage is not being clipped and we are
getting a nice voltage swing.
FIGURE 40: VOLTAGE OUTPUT
As one can see we get a nice sine wave out with a huge voltage swing. Now will look at
the FFT of the output sine wave. The FFT graph can be seen in figure 41.
FIGURE 41: FFT OF THE OUTPUT
52
As one can clearly see there is a large spike at 750kHz. This spike is expected because
output signal frequency is 750kHz. The final component to be observed is frequency
response. The frequency response should not have attenuation as it approach higher
frequencies. The spec sheet shows that there should be no attenuation at the frequency
range being used. The frequency response graph can be seen in figure 42.
FIGURE 42: FREQUENCY RESPONSE PLOT
The frequency response shows no attenuation as the frequency increases. Theoretically
the line driver that is seen in figure 38 works very well.
3.3.1 Line Driver Data
We will use the line driver that was provided by Analog Devices, Inc to run tests. The
line driver schematic can be found in figure 43. The line driver was powered with a +12
Volts to the +VT and 12 Volts to VT with 125mA current limit. The resistors that
control the gain (R17, R18, R19) were all 1000 Ohms, therefore creating a gain of 1 for
the AD8016 amplifiers. The slew rate was the first area to check. A 1 V peaktopeak
.
53
FIGURE 43: LINE DRIVER
square wave was inputted and then looked at the output. The output can be seen in
figure 44. The slew rate was calculated to be 100V/s. This is a very good slew rate
and will not affect our application.
FIGURE 44: SLEW RATE GRAPH
54
The next thing we did is inputted a signal into the circuit. The input signal and its FFT
can be seen in figure 45, below.
FIGURE 45: VOLTAGE INPUT AND ITS FFT
As one can see from the plot above the sine wave signal passes into one input and the
second output is grounded. At this point in time there is not a differential signal that we
can use therefore we grounded the +V input to ground and V input to a 1V peaktopeak
sine wave. The sine wave is oscillating at 750kHz. The FFT had a very large noise
floor. This noise floor on input will create a huge problem because the receiving end of
the line driver may interpret this noise as false information. In order to have a clean
signal we need to design a filter that will filter the input signal. If the signal is not clean it
will be amplified even more in the circuit to cause problems.
The next thing that we will look at is the output of the instrumentation amplifier.
The output can be found in figure 46.
55
FIGURE 46: VOLTAGE OUTPUT
From observing the voltage output the circuit looks good. The next thing that we looked
at is the FFT of the outputs, figure 47 below shows us those FFT outputs.
FIGURE 47: FFT OF THE INSTRUMENTATION AMPLIFIER
The final step taken was to wrap a 1.5 foot of twisted pair cable around a power cable to
view the noise form on the signal. The twisted pair cable was not terminated properly.
The voltage output after the transmission line can be seen in figure 48.
Vout
+Vout
Vout
+Vout
56
FIGURE 48: VOLTAGE OUTPUT AFTER THE TRANSMISSION LINE
One can see that there is some small noise at the peaks of the sine waves. Since we are
looking at a 750kHz wave, we wont see 60Hz noise very clearly. The last thing that we
will look at is the FFT at the end of the transmission line, and once again the noise floor
is a little bit larger. This FFT can be seen in figure 49.
FIGURE 49: FFT AT THE END OF THE TWISTED PAIR
One again it can be seen that the noise floor is getting larger. Now that we know and
understand how this circuit functions and how the signal is transmitted we can now attack
two problems that stand in the way of transmitting good DMT. The first problem
eliminating the noise floor at the input and keeping the noise floor low throughout the
Vout
+Vout
+Vout
Vout
57
system. The second problem is terminating the transmission line properly without
loosing significant amount of power.
58
4.0 Conclusions
The goal of the MQP was to increase the power efficiency of the AD8016 line
driver. By increasing the power efficiency, we would be able to transmit signal at a
higher signal to noise ration. This allows us to either a) transmit the signal greater
distances or b) transmit data at higher speeds. A combination of both could also be
achieved through optimization.
To increase the power efficiency literally means that given the same power
supplies as before, we have increased the power supplied to the load. In our MQP, this
was accomplished by reducing the size of the back termination resistor from 50 ohms to
15 ohms. In traditional line driver circuits, the impedance of the back termination
resistance matched that of the line. This scheme made sure that no reflections occurred,
but at the same time, the signal would lose half of the power over that back termination
resistance. By reducing the size of the back termination resistance, we in turn reduce the
power lost across this resistor, which means that more power is transferred to the load, in
this case it was about 30 percent more. When more power is transferred to the load, we
are able to transmit signal greater distances than before.
With traditionally back terminated lined drivers, the physical transmission limit
for ADSL was 18,000 feet from the central office to the end user, with an average
distance of 12,000 feet. By increasing the power supplied to the load, we will be able to
transmit much farther distances, possibly up to 24,000 feet.
The other benefit of transmitting more power to the load is that a higher signal to
noise ratio (SNR) will result. If we have a higher signal to noise ratio, we will be able to
59
transmit data over the same distance, but at higher speeds. Using Shannons channel
capacity theorem we can see this.
,
_
+
N
S
B C 1
C = Data Rate (bits / second)
B = Bandwidth (Hz)
S = Signal Power (Watts)
N = Noise Power (Watts)
In our system, the bandwidth is fixed so the only variable we have to manipulate in order
to get a faster data rate is the signal to noise ratio. As the reader can see, by transmitting
at a higher signal too noise ratio, it will be possible to increase the data rate. It is to be
noted that the signal to noise ration used in Shannons channel capacity theorem is a ratio
of Watts/Watts and not the log of the actual signal to noise ration. The log of the actual
SNR can however be converted to the above form and used in Shannons channel
capacity theorem. With our schematic, we were able to get a SNR of 80dB. This means
that the noise floor was 80 decibels down from the peak of the magnitude of the signal.
Figure 50 shows this.
FIGURE 50: FREQUENCY RESPONSE OF THE LINE DRIVER
(72)
60
BIBLIOGRAPHY
Reiter, HansErhard. " HighSpeed Remote Access Becomes Reality." Online.
http://www.dslforum.org/PressRoom/backup/adsl/www/PressRoom/editorial6_97.html (7
Sept. 2000)
"Electronics Tutorial" Online. http://williamsonlabs.com/home.htm
http://www.stanford.edu/class/ee379c/ (click on Class Reader, then on Chapter 4)
B.P. Lathi, Signal Processing & Linear Systems Carmichaels: Berkeley Cambridge Press,
1998
Leon, W, Il Couch Digital and Analog Communication Systems Prentice Hall: 1998
"Analog Devices" Online. http://www.analog.com/publications/whitepapers/products
/32bit_wa.html
Mancini, Ron. "Current Feedback Amplifier Analysis and Compensation." Online.
http://wwws.ti.com/sc/psheets/sloa021a/sloa021a.pdf (15 Jan. 2001)
Bob Day and Scott Wurcer. " Optimizing Line driver Designs for Maximum Power;
"Bridging" ADSL Line driver Challenges" Online. www.analog.com (1 Nov. 2000)
"Initial, acronyms and symbols." Online. httt://www.cs.tut.fi/tlt/stuff/adsl/node4.html (10
Sept. 2000)
Baines, Rupert. DMT and CAP Line codes for ADSL 1997
Martin Jackson and Stefan Knight. "At work With ADSL: More Than Bandwidth."
Online. http://www.dslforum.org/PressRoom/editorial4_98.html
" ADSL Application Note." Online. http://www.dslforum.org/app_notes.html
"Analog Devices" ADSL Technology." Online. http://www.analog.com/publications/
whitepapers/products/back_adsl/ (8 Sept. 2000)
American National Standards Institute, ANSI T1.3131998, Network and Customer
Instalation InterfacesAsymmetric Digital Subscriber Line (ADSL) Metallic Interface,
1998.
"ADSL" Online. http://www.adsl.com/adsl_tutorial.html
Franco, Sergio. Analytical Foundation of CurrentFeedback Amplifiers Online.
http://www.analog.com/support/standard_linear/seminar_material/highspeed/1.pdf
61
Appendix A Datasheets