Read Chapter 5 in your text and study this lesson. Complete hw 5. Digital and analog transmissions of digital and analog information. In lesson three, you
studied various physical layer media. These media are capable of carrying digital or analog signals that represent either digital or analog information. Often you'll have one type of information and have to
convert it to the other form for transmission as signals. The bulk of communications is moving toward digital transmission of signals on copper and fiber media. In this lesson we'll look at the ways digital and
analog signals are converted to more convenient forms and how improvements can be made to some of these systems. Let's layout the ground work by organizing our options.
|
|
|
|
|
AM and FM radio, CB radio, Audio and Video tape recorder |
|
|
|
Digital voice encoding (PCM), real audio, digital camcorder, PCS cellular phone |
|
|
|
ASK, FSK, PSK, and QPSK modulation, microwave and satellite transmissions |
|
|
|
Manchester encoding for Ethernet, AMI encoding for a T-1 line. |
|
We will discuss the last three rows of the table above. If you are interested in AM and FM, you might take Point-to-point Transmission Systems next fall. Another route would be Radio and
Cellular Systems in the Spring. These courses will dig a bit deeper into how they work. The last row, digital info carried on digital signals is quite popular these days, now that the Internet
is such a pervasive force. We'll spend nearly all the remaining time in this course working under the assumption that we are on pure digital systems for data communications.
In this lesson your task is to learn methods of sending analog signals and digital signals across media and how to convert from one form to another in preparation for transmission on media.
Converting from analog to digital Some of you may have participated in a voice chat session over the Internet, seeing how that has
become somewhat popular over the last year or two. Of course your voice is sent out across the Internet as bits, but it is certainly an analog signal before it gets to the computer. Some sort of
conversion process has to turn your voice into digital bits. The compression waves travel through the air to the microphone which converts the compressions
into an analog electrical current that varies up and down in synch with your voice. The computer will take that signal and convert your voice signal into bits using a digitizer. The bit stream is
stored in memory for later use (or sent out immediately in the case of chat). One popular way of doing this is called Pulse Code Modulation, or PCM. Here is how it is done.
Your voice is measured once every 0.000125 seconds, or 8000 times per second. The digitizer looks at the voltage that comes in from the microphone to determine the amplitude (voltage) of the
signal. The digitizer then assigns a combination of bits to represent that sample. How does the digitizer know what bits to assign?
The digitizer simply divides the total range of possible voltages up into small intervals. It then labels each interval with a unique combination of bits. An extremely simple version is shown
below. The range from -4 V to +4 V is divided into 8 intervals, each 1 volt in size. The eight intervals can be uniquely identified with just three bits. Each sample is simply assigned the bit
combination for the interval that it falls into.
Once the bits are determined, they can be sent out just as any other digital information would transmitted. When the signal is reproduced on for the listeners in the chat room, for example, the
signal value is decoded from bits to a voltage again. The only issue here is that we don't know what the original value was. We only know what interval it was in. For this reason we reproduce
a voltage that is in the center of the interval. This means we have some error. You may think of this as a round off error, but it is usually called "quantization error".
When we reproduce the voice it will be a very close approximation of the original. To improve our quality we can double the number of intervals. That means our error will be half as big.
Double again and our gets smaller still. Obviously the more levels we have the better the quality.
As the number of intervals, called levels, increases we have to use more bits to uniquely identify them. With three bits we had eight levels. With 4 bits we would double that number of intervals
and as a result our intervals would be half as big. How many would we have?
Where L is the number of unique combinations we can make with a given number of bits, and 'n' is the number of bits we need to do that.
Example: Voice signals need about 13 bits to identify enough unique intervals for good voice quality.
a) How many intervals would that require? L = 2n L = 213 L = 8,192 levels
b) If we expect the maximum voltage of the signal could be +3.0 volts, and the minimum voltage could be -3.0 volts, how large would one interval be? range = 3V - (-3V) = 6 volts
The range is divided up into 8,192 levels, so one level would be: interval = 6V / 8,192 = 0.000732 volts
The PCM (pulse code modulation) digitization process described above is a linear system. This means that all the intervals are the same size. There is a more advanced type of voice digitization
called u255 Companded PCM. Companded means 'compressed and expanded' and u255 Companded PCM encoding is often called 'u-law' for short. u255 uses smaller and smaller
intervals as the voltage approaches zero. That way the error will not be very large compared to the value of the signal.
Linear encoding uses 13 bits per voice sample and u255 uses 8 bits per sample. What are the advantages of this? Look into the next example.
Example: What is the data rate for voice using linear and companded encoding if we must send 8000 samples per second to maintain good quality voice reproduction?
R = r x n Where R is the data rate (we've seen this eqn. before), r is the symbol rate, and n is the number of bits per symbol R = 8000 samples/sec x 8 bits/sample
R = 64000 bps for companded encoding R = 8000 samples/sec x 13 bits/sample R = 104,000 bps for linear encoding
From the example you should be able to see that we don't need as much bandwidth to send quality voice communications over digital transmission systems if we digitize the voice with companded encoding. Digital Data on Analog Channels OK, now lets see what happens when we convert digital signals in preparation for transmission on
analog channels. We might do this in order to send our digital bit stream out over the 'airwaves' on a wireless LAN. While radio waves are analog, they can carry digital information.
We studied sine waves in an earlier lesson. Now let's take advantage of that knowledge. Combining a high frequency radio wave, a carrier, with the digital signal we can send digital
information over an analog channel. Here are two signals, the carrier wave and the digital information.
By multiplying the value of the carrier wave with the value of the bit stream, we get a 'modulated' carrier wave. A carrier that contains the intelligence of the original bit stream but one that can
travel over the airwaves. Here is what we get.
At the far end of the medium, the receiver is listening to determine if a one or zero was intended. If the signal has not degraded too much, the ones and zeros can be determined exactly.
The above system is a very simple way of sending digital information over an analog channel. This particular method is known as Amplitude Shift Keying, or ASK. As shown above, only one bit is
represented by a single signal. We could use four amplitudes (four unique symbols) and indicate two bits with the transmission of a single signal. That would double our data rate.
Two other methods are also popular. FSK, or Frequency Shift Keying, uses two different frequencies to indicate ones and zeros, for example. Four frequencies would permit us to send
two bits per symbol. PSK, or Phase Shift Keying, uses two different phases. FSK
Before we take a look at PSK, let's consider what phase is. A sine wave usually starts at zero and heads upwards toward its first peak. However, if we change the phase, by adding in some
extra value to the angle like this, sin(2pi f t + x), we can offset this zero point. Adding 180 degrees (or pi radians) allows us to move the sine wave enough that it begins at zero but starts out
moving down. Using these two different signals allows us to send binary ones and zeros, just as we did with FSK and ASK. Take a look.
Summary: ASK = Amplitude Shift Keying. Change the amplitude of the carrier wave to indicate each unique signal.
FSK = Frequency Shift Keying. Change the frequency of the carrier wave to indicate each unique signal.
PSK = Phase Shift Keying. Change the phase of the carrier wave to indicate each unique signal. Each of the methods above can send one bit per signal or more (subject to the signal to noise ratio
limit we studied in lesson three of course). They can also be combined. See if you can guess what the following is:
Looking at the signal there seem to be phase changes and amplitude changes. The frequency remains the same. This is an example of ASK combined with PSK. Two bits can be sent with
each signal. A very popular technique is QPSK, which means Quadrature Phase Shift Keying, where four phases are used. This means that two bits are indicated by a single signal transmission. Digital Information Sent as Digital Signals The most common transmission techniques used these days are digital transmissions on electrical
cables. Simply changing the voltage allows us to send different signals to the opposite end of a cable. Many techniques have been developed to do this. Unipolar NRZ Line Coding
The simplest technique for sending digital data is called Non-return to Zero or NRZ. To indicate a one, turn on the electrical current for the full duration of a one bit. For a zero turn it off. The
duration of a bit is called the bit time. You can determine the amount of time a bit takes with the following equation.
bit time = b = 1/R units for b = 1 / [bits per second] = seconds per bit Where R is the data rate in bits per second and b is the bit time in seconds per bit. b will
be a very small fraction if the bit rate is high.
Here is what this might look like.
In the example shown, the bit time is one second. That would give us a data rate of:
b = 1 / R 1 sec per bit = 1 / R R = 1 / 1 sec per bit = 1 bit per second. Maybe we should shrink the bit time a little. :)
One problem with unipolar NRZ (and all unipolar line codes) is that there is a net flow of electrical current from one end to the other. Current is either positive or zero.
Bipolar NRZ Line Coding Fortunately that's easy to fix. All we have to do is make the current vary from + to - voltage to indicate 1s and 0s.
If we have close to the same number of ones and zeros, then the average current flow will be zero. This means we'll save some power when we run the system. For short distances, like those
within your computer from CPU to hard drive, this isn't really an issue, but for long distances, it can become very important. For longer distances, slightly more sophisticated line codes have been developed. AMI Line Code for T-1 Transmissions Unlike NRZ, Alternate Mark Inversion doesn't fill the entire bit time with a constant voltage. A
1-bit is assigned a positive or negative voltage that takes up 50% of the bit time and it is centered in the bit time. A zero is represented by zero voltage. Every 1-bit has the opposite polarity of the
preceding 1-bit.
Let's make a rule that when you don't know what the preceding 1-bit was, you should start with a positive pulse, as shown in the graphic above.
Two new terms are introduced above, the pulse width and the duty cycle. Pulse width is just a measure of the time of one pulse. The percentage of the bit time used by the pulse is called the
Duty Cycle. AMI has a duty cycle of 50%. Notice how the voltage returns to zero before the end of every bit time.
One advantage of this system is that if an error occurs (a bit changes during transmission) we'll notice this. If we keep track of the polarities, we'll discover erred bits because the alternating
one's pattern will be broken. Sometimes a one is called a mark. For AMI, if two succeeding marks have the same polarity, it is called a violation. That is, it violates the pattern of alternating
marks. Now you can see why we have the name Alternate Mark Inversion. The polarity of each mark is
inverted from the previous mark. That is, the polarities alternate for the marks. AMI on a T-1 transmission system uses voltages of +3 and -3 volts for the marks and zero volts for a zero.
One problem encountered with AMI is that a long string of zeros may throw our count off. We lose track of where the bit boundaries are. For this reason a separate wire pair from the
transmitter to the receiver is provided. A square-wave signal is sent from one end to the other to provide a "clocking signal" or "timing signal".
Each time the clocking signal changes voltage from positive to negative, the voltage on the data line is measured to determine what symbol was sent.
Another way to avoid losing track of bit boundaries within AMI line codes is a process called bit substitution. We'll discuss this later in this lesson. Manchester Line Code
Where AMI is weak in keeping track of bit time boundaries, Manchester line codes are quite strong. Manchester has lots of polarity changes which help the receiver keep track of the signal.
No separate wire pair is needed to maintain timing. The receiver just keys off the polarity changes (at least one per bit time) in the data signal.
Look at the following diagram to see what signals are sent to represent 1s and 0s.
There is always a transition in the middle of the bit time. A transition from positive to negative
indicates a 0 and a transition from negative to positive indicates a 1. If needed, a transition occurs at the start of the bit time to prepare the voltage for its next transition. Take a look at an example
of a Manchester encoded bit stream.
Line Code Summary
Line Code |
Symbol Representation |
Duty Cycle |
Disadvantages |
Advantages |
Unipolar NRZ |
1 is Positive V 0 is 0 volts |
100% |
Net flow of current Timing problems with long strings of ones or zeros |
Very easy to do, simple |
Bipolar NRZ |
1 is positive V 0 is negative V |
100% |
Timing problems with long strings of ones and zeros |
Very simple. |
AMI |
1 is + or - V 0 is 0 volts |
50% |
Timing problems with long strings of zeros |
No net current flow Can detect some errors |
Manchester |
1 is - to + transition 0 is + to - transition |
100% |
More transitions require more bandwidth |
Keeps timing very well, doesn't require separate timing wire pair. |
More on Clocking The receiver of a digital signal must measure pulses at the right time. There are two ways to know
when to do this. One is to keep track of the transitions between the pulses. These mark the boundaries of the pulse. As long as transitions are frequent, the equipment should be able to keep
track of where the center of a pulse is, and therefore, when the voltage of a pulse should be measured. The other way to do this is with a clocking signal sent over an additional wire pair from the
transmitter to the receiver. The only fear about this method is having a clock rate that is slightly different than the actual signal rate. If there is a difference, then the receiver will measure the bit
slightly earlier or later than it should. The measurement will "drift" away from the center as succeeding bits arrive. Eventually, the receiver will measure the wrong bit! This is called a timing error.
Bit Substitution for T-Systems AMI suffers from poor timing because of long strings of zeros. One solution would be to use a
Manchester line code, but that requires a higher signaling rate, which means more bandwidth. Manchester is more costly to implement as a result. What are we to do? Bit substitution is the answer. Recall that AMI can detect a pulse that violates the alternating 1s rule. We can take advantage of this. On the transmit end, before we send the signals on the line, we can substitute a known
combination of bits in for a sequence of consecutive zeros. If the substituted bits violate the alternating 1s rule, then the receiver can identify them. The receiver compares the bit pattern
where the violation occurs to see if the bits match the substitution rules. If there the pattern of bits matches a substitution, then the receiver sets the bits back to zeros and hands the information up
to the next layer. In practice how does this work? First let's look at a way to write AMI line coding that is easier
than sketching a pulse train. We can indicate a mark (a 1) with either a + or a - and we can indicate a zero with a 0. Then we just write the bits down and convert them to "+/- code." Take
a look at the following bits and the "+/- code" that goes with them.
Here are some rules that will clear up a few ambiguous situations.
1. Always make the first mark positive. 2. Every mark has the opposite polarity of the mark before it.
Example: Convert the following bit stream to +/- code: 10111100010
Write the bits out then write the +/- code below, alternating the pluses and minuses. bits: 1 0 1 1 1 1 0 0 0 1 0 +/-: + 0 - + - + 0 0 0 - 0 <-- answer
Another Example: Decode the following +/- code: +0-+-+-+0-+
Now that we can write plus and minus code, we are ready to start substituting bits in for strings of zeros. We'll study four ways to do this, B8ZS, B6ZS, B3ZS, and HDB3.
Binary 8-Zeros Substitution B8ZS is one technique used on T-1 lines with AMI encoding to eliminate long strings of zeros.
The rules are fairly simple. When all the bits of a channel (8 bits) are zeros, we check the nearest preceding mark and then substitute according to the following table:
Preceding Mark |
Substitute these bits |
+ |
000+-0-+ |
- |
000-+0+- |
Examples: Use B8ZS on the following:
a) 10000001 00000000 00000010 Convert this to plus and minus code: +000000- 00000000 000000+0
^^^^^^^^ The middle channel is all zeros. It follows a negative pulse. The
table says we substitute 000-+0+- for the eight zeros. Our answer is ... +000000- 000-+0+- 000000+0 b) 00000000 00000000 00000000
This is already in plus minus code, right? Note that all three channels are all zeros. We have to substitute for each channel. When you don't know what the preceding mark was, assume it was a
positive pulse. OK, now we can look into the table for the first substitution. We get 000+-0-+. Make this substitution. 000+-0-+ 0000000 00000000
Now check the next substitution. Yes, the next substitution will check the preceding pulse after we made a substitution there. It follows a plus and so will the third substitution. We get the
following answer. 000+-0-+ 000+-0-+ 000+-0-+
Unfortunately we have to keep track of where the channel boundaries are when using B8ZS. That's inconvenient and has led to other solutions which just look for a consecutive number of
zeros in a row at any location. In other words, the following substitution methods ignore channel boundaries. Binary 6-Zeros Substitution
B6ZS is nearly identical to B8ZS. Ignoring channel boundaries this time, we will substitute for each consecutive string of six zeros that we find. Use the following table for B6ZS.
Preceding Mark |
Substitution |
+ |
0+-0-+ |
- |
0-+0+- |
Example: Use B6ZS on the following bit stream, 1000000000100000001
High Density Bipolar 3-zeros HDB3 is slightly more complicated than B6ZS, you have to count the number of marks since the
beginning or since the last substitution was completed. Then use this table:
HDB3 |
Number of Marks is |
Preceding Mark |
ODD |
EVEN |
+ |
000+ |
-00- |
- |
000- |
+00+ |
I don't know why this is isn't HDB4 when four zeros are substituted at a time. Let's try an example anyway.
Example: For the following bit stream, 1000000001000, substitute using HDB3.
+/- code: + 0 0 0 0 0 0 0 0 - 0 0 0 subs: 0 0 0 + - 0 0 - final: + 0 0 0 + - 0 0 - - 0 0 0
Checking the table for the first substitution, look under the column for ODD because we had one mark since the beginning. The mark is positive so the substitution is 000+.
After completing this substitution, count the number of marks since the substitution. You get zero. We will treat zero as an even number, so look under the EVEN column across
from the +. You'll find the substitution is -00-.
Binary 3-Zeros Substitution B3ZS is our last substitution method. There are many more bit substitution methods, but these are
reasonably common ones. B3ZS is used on AMI line codes implemented on T-3 transmission systems. It is just like HDB3, but you substitute for a string of just 3 zeros when you find them.
The higher signaling rate on a T-3 system makes even a string of just 3 zeros susceptible to timing errors.
B3ZS |
Number of Marks is |
Preceding Mark |
ODD |
EVEN |
+ |
00+ |
-0- |
- |
00- |
+0+ |
Example: Use B3ZS on the following: 10000000 01000011 000
Decoding Bit Substitution In practice, both ends of a transmission link know which substitution method is being used. To
decode a substitution, the receiver first searches for violations of the alternating 1s rule. Matching the bit pattern with the appropriate substitution will indicate if a violation is a substitution or an
error. Each substitution will be replaced with zeros, returning the bit stream to its original condition.
Examples: Decode the following plus minus code. Unlike the receiver, you don't know which substitution method was used.
+/- code: +000+-00 -+00+00- +-00-000 Solution 1. Look for and mark the violations. +/- code: +000+-00 -+00+00- +-00-000
^ ^ ^ ^
2. Compare the bit patterns to those of substitution methods you know.
Can't be B3ZS because there are three zeros left in some parts of this bit stream. Those would have been substituted for using B3ZS. This doesn't match the B8ZS or B6ZS patterns.
The only other method you've studied is HDB3, so that must be it.
3. Mark (or underline) the groups of 4 bit substitutions and make sure they match the HDB3 substitution table. Yes, this is HDB3. +/- code: +000+-00 -+00+00- +-00-000
^^^^*** *^^^^ **** 4. Replace the substitutions with zeros. No subs: +0000000 0000000- +0000000 5. You are not done yet. Change this code back to ones and zeros. final: 10000000 00000001 10000000
|