| Study this lesson and Chapter 3 in your text.  Also read Appendix B at the end of Chapter 3.  Complete hw3. Data Transmission Fundamentals In this lesson 
                            we will study the difference between analog signals and digital signals.  We will study the fundamental building blocks of these signals, and methods of representing these graphically. Also, in this lesson, we 
                            will investigate detrimental influences on signals, like background noise and attenuation.  We will then define the signal to noise ratio and take a look at how to represent this in decibels.  This will 
                            include some practice working with decibel mathematics.   Finally, we will study Shannon's Capacity Theorem.  Using this theorem, we can calculate the maximum theoretical data transmission rate for a given 
                            channel.  Let's get started! Analog and digital SignalsMany people ask, "What is the difference between analog and digital?"  You should be able to answer this question in the next 10 minutes.
 
                            At a basic level, analog an digital mean continuous and discrete.  This is well explained with an example:
                             
                                DigitalThe number of people in a movie theater is some whole number.  There is never a fractional person in the room.  At least we hope not!  When counting, only whole number values are 
                                acceptable.  The number of people in the theater is digital.  Only discrete values can be used to count the people.
 AnalogHere, a number may have any value.  Often there is a range within 
                                which a value may continuously vary.  For example a tree may range from 0.00000000000... feet tall to a perhaps a hundred feet tall.  A tree grows, and passes through all heights on a continuous scale.
 
                                DigitizationWhen we measure the height of a tree or of ourselves, we round off the number.  I'm 5'9" tall.  However, at this very instant, I might be close to 5'9.000354" tall.  But even 
                                that is rounding off.  Measuring analog values requires rounding off.  Perhaps we measure as accurately as our instruments let us.  We are still rounding off.  Maybe we round off for 
                                convenience.  I weigh 200 pounds, rounded off to the nearest 100 pounds!  That sounds reasonable if I round off my height to the nearest foot.  Don't forget your instructor is 6 feet tall and 200 
                                pounds!
 Example:  Step on the bathroom scale in the morning.  Measure your weight (rounded off, of course).  Step on the scale later on in the day.  Has your weight changed?  Yes, because you eat, 
                            drink and sweat.  Your weight varies on a continuous scale between an upper and a lower boundary. Summary:  Analog signals are those that vary on a continuous scale, taking any value between some upper and 
                            some lower boundary.  Digital signals take on only discrete values between an upper and lower boundary. Fundamentals of Signalsconsider an audio test tone.  You could play it over your expensive 
                            sound system to test the quality of your speakers.  You would judge the sound by measuring the frequency of the tone (pitch) and the loudness of the tone (amplitude).  Test equipment could measure these values 
                            and you would compare them to the known quantities of the original signal.
 Try out this test tone which sweeps from 300 Hz to 3000 Hz in 10 seconds.  You'll need something that can play a .wav file (QuickTime for example). Check out these abbreviations:
                             
                                fFrequency is the number of repeating cycles per second.  The unit of measure is hertz, or cycles per second.  Hertz is abbreviated as Hz.
 Example:  Human voice ranges from 
                                about 300 Hz to about 3300 Hz, according to the phone company.  Some people can produce higher frequencies, perhaps up to 5000 Hz.  Humans can hear frequencies from about 50 Hz up to about 15,000 Hz. A
                                Amplitude is the loudness of a sound wave or the voltage of an electrical signal.  In a sine wave, we'll represent amplitude on the Y axis.
 Take a look at the following graph of a sine wave.  The graph shows the amplitude (A, on the Y axis) and time on the X axis.  This sine wave is graphed in the "time domain." That means the X axis represents 
                            time.  The wave repeats twice in one second.  One cycle takes a half second, two cycles take one second.  The maximum amplitude is 2 and the frequency is 2 Hz.
                             If this were alternating electrical current, AC:  
                             As time passes, the value at any instant in time is calculated with the following sine wave equation:
                             So, our example graph would be based on:
                             You can try calculating values at different times using your calculator.
                             
                                y = 2 sin(2 Pi 2 t) 
                                 
                                    
                                        | t (seconds) | y (amplitude) |  
                                        | 0.00 | 0.00 |  
                                        | 0.125 | 2.00 |  
                                        | 0.250 | 0.00 |  
                                        | 0.375 | -2.00 |  
                                        | 0.500 | 0.00 |  
                                        | 0.625 | 2.00 |  
                                        | 0.750 | 0.00 |  
                                        | 0.875 | -2.00 |  
                                        | 1.000 | 0.00 |  The values in this table will produce the sine wave shown a little earlier in this lesson. HOWEVER, you need to change your calculator from degree mode to radian mode.  Check your 
                            manual.  Most $12 dollar scientific calculators can do this.  You set the mode to radians, plug in the value and hit the sine button.  If your calculator won't do this, then note the following conversion.
                             
                                There are 2 Pi radians per 360 degrees.  Pi is approximately 3.14Example:  Change 45 degrees into radians.
 
                                    = 45 degrees  x  2 Pi radians  /  360 degrees = Pi/4 radians
 = 3.14 / 4 radians, or about 0.785 radians
 Another Example: Change 4.0 radians into degrees.
                                 If we increase the frequency, what happens?
                             
                                y = 2 sin(2Pi 10 t) 
                                 
                                    
                                        | t (seconds) | y (amplitude) |  
                                        | 0.00 | 0.00 |  
                                        | 1/40 | 2.00 |  
                                        | 2/40 | 0.00 |  
                                        | 3/40 | -2.00 |  
                                        | 4/40 = 0.1 | 0.00 |  
                                        | 5/40 | 2.00 |  
                                        | 6/40 | 0.00 |  
                                        | 7/40 | -2.00 |  
                                        | 8/40 = 0.2 | 0.00 |  Here is what the graph will look like.  Note that the cycles per second have increased.  There is one cycle in 0.10 seconds.  Therefore, in one second there are 10 cycles.  That may be obvious, 
                            because the frequency is ten hertz.
                             Another name for cycles per second is hertz, or Hz, as mentioned earlier.  The signal above has a frequency of 10 Hz and a maximum amplitude of 2.0.  Note that in the two examples above, the 
                            frequency remain and maximum amplitude remain constant. We may also graph these examples in the "frequency domain".  That means we make a graph 
                            comparing amplitude on the Y axis and frequency on the X axis.  Our first example above had a frequency of 2.0 Hz and an amplitude of 2.0.  
                             As you can see, an arrow at 2.0 Hz with a height (amplitude) of 2.0 is drawn on the graph.  The arrow is called a frequency spike because it represents a single frequency.  The height of the spike 
                            represents the maximum amplitude of the signal. For our second example the frequency domain would look like this:
                             The spike occurs at f = 10 Hz and the amplitude is 2.0. Example:  Sketch the graph in both the time domain and the frequency domain of the following:  y = 8 sin(2Pi 500 t)
                             
                                Look at the equation and determine by inspection that the frequency is 500 Hz and the maximum amplitude is 8.0.  Now, sketch out your graphs.  They should look like the following two graphs. Frequency Domain
   Time Domain
   The hard way to complete this task is to try it on a calculator.  It is better to note that the 
                                frequency is 500 Hz, so there are 500 cycles in one second.  That also means that one cycle takes one five-hundredth of a second (labeled).
                             From the preceding example, we learned that knowing the time for one cycle is convenient when trying to graph a wave in the time domain.  The time for one cycle is called the period, abbreviated as T.
                             Example: If the frequency of a signal is 10 Hz, what is the period?
                             
                                T = 1 / f  =  1 / 10 Hz  =  1 / 10 cycles per secondT = 0.1 seconds per cycle
 When frequency is constant, like in the following graph, the period can be measured from 
                                any two identical points.  The two points define a single cycle.  
 Building a Square Wave If you turn on two test tones at the same time and they both have identical pitch and loudness, they 
                            will add together.  They have identical pitch (frequency) so their amplitudes add together. ExampleWhat would you hear if both of the following tones were played simultaneously?
 
                                tone 1:  f = 1000 Hz, A = 3.0tone 2:  f = 1000 Hz, A = 3.0
 You would hear the following:
                             
                                tone:  f = 1000 Hz, A = 6.0
                             Here is a representation in the time and frequency domains.
                             
                                Time Domain
   Frequency Domain
   The mathematical version of this follows.  Note that the amplitudes end up adding together and you have a single tone at the same frequency.
                             
                                tone 1: y = 3 sin(2Pi 1000 t)tone 2: y = 3 sin(2Pi 1000 t)
 Y = tone 1 + tone 2Y = 3 sin(2Pi 1000 t) + 3 sin(2Pi 1000 t)
 Y = (3 + 3) sin(2Pi 1000 t)
 Y = 6 sin(2Pi 1000 t)
 What happens when you add two signals together and the frequencies are not the same?  At some locations the amplitudes of the signals add together in the time domain.  At other times they will 
                            have different polarity and subtraction will occur.  In the frequency domain, the spikes occur at totally different locations. ExampleAdd the following two sine waves together and sketch the graphs of the time and frequency domains.
 
                                y1 = sin(2Pi t)y2 = (1/3) sin(2Pi 3 t)
 These are graphed separately below:
   The time domain will represent the sum of the two equations:
                             
                                y at time t = y1 + y1 = sin(2Pi t) + 1/3 sin(2Pi 3 t)Plug in a value for t and calculate a single number
 The frequency domain will show the two distinct frequency spikes, one at 3 Hz with an amplitude of 1/3 and one at 1 Hz with an amplitude of 1.0.  here they are:
                             Take a look at page 39 in Data and Computer Communications for similar discussion of adding waves together. 
                            By adding just the right waves together, we can build a square pulse.  It turns out that we need to add successive waves where the amplitude is smaller each time and the frequency is higher each 
                            time.  Take a look at the following equation:
                             
                                y = sin(2Pi f t) + 1/3 sin(2Pi 3f t) + 1/5 sin(2Pi 5f t) + 1/7 sin(2Pi 7f t) + 1/9 sin(2Pi 9f t) + 1/11 sin(2Pi 11f t) + ...
                             The farther we go with this the more square our wave will be.  Take a look at the following.  
   This leads to some important issues:
                             
                                * To create a perfectly square pulse you have to have an infinite number of waves added together. * Each term adds a new and higher frequency to the collection.  A perfectly square pulse 
                                needs infinite bandwidth. * Because wires, fiber and air tend to have restricted bandwidth we have less than perfect signals.  Of course we would need to build equipment that could add up an infinite number 
                                of signals. * Less than perfectly square pulses mean that we have to wait for the pulse to ramp up to its full value, which wastes a small amount of time.  When you waste that tiny amount of time 
                                billions of times a second it adds up.  We eventually end up having to slow our data transmission rate down. * By choosing a wide enough bandwidth we can create a pulse that is just square enough to 
                                support our purposes in transmissions.  Our data rate may be limited, but we'll be able to operate within a small range of frequencies.
                             Page 44 in Stallings' gives an example of trying to build pulses within a limited bandwidth.   We can conclude that a fixed bandwidth puts a limit on our maximum data rate. Transmission ImpairmentsImperfectly square pulses are not the only negative influence on transmissions. Other problems include attenuation, delay distortion, and noise.
 
                                Attenuation:  Signals lose power as they travel over distance.  Radio waves, electrical signals and optical signals all attenuate over distance.  The signal may become so weak that 
                                we cannot determine the message being sent or worse, we might not even know a signal was sent. Delay Distortion:  In most media, such as fiber or copper, different frequencies travel at 
                                different speeds.  At great distance, two simultaneously transmitted signals will arrive at different times.  Our square pulse is built up with different frequencies and that means our 
                                pulse may spread out and round off.  The farther it travels, the more this occurs. Noise:  Noise changes the amplitude and to a lesser degree the frequency of a signal.  
                                There are four kinds of noise.   Thermal Noise is due to the vibration of electrons in conductors.  The vibration of these 
                                electrons creates a small amount of current that varies in a random pattern.  The higher the temperature of the conductor, the higher the thermal noise is.  Stallings gives a short 
                                equation you can use to determine the amount of noise in a certain frequency range. Intermodulation Noise is interference with the intended signal that comes from the by 
                                products of intended signal combinations.  All right, I know you are saying, "huh?".  When two different frequencies are transmitted simultaneously they produce signals that are called 
                                harmonics.  These harmonic signals are sums or differences of the two intended signals.  
                                 Harmonic signals are lower in power, but they might interfere with other intended signals at the new frequency. Crosstalk is the mixing of intended messages and other intended messages.  You might 
                                have notices this on rare occasion in the Public Switched Telephone Network (PSTN).  You can hear, faintly, some other conversation on the line. Impulse Noise
                                 is, for example, electrical sparks or lightning.  Electrical discharge releases a wide spectrum of electromagnetic energy.  These electromagnetic waves interfere with an 
                                intended signal.  You've heard your radio crackle as lighting strikes in the background.  That's a prime example of impulse noise.
                             All of these impairments cause our nice reasonbly-square pulse to degrade.  At the worst, we end up with an unintelligible mess that we may not know is the signal.  
 The graphic above shows the combination of noise, delay distortion, and attenuation.  In Stallings, 
                            figure 3.16 shows the intrusion of noise as a summation of the noise signal and the intended signal.  Take a look at that as well. 
                            While our signal loses power over distance the effects of noise are gradually taking their toll.  Our concern changes from "how square is the pulse?" to "Where is the pulse?" 
                            To determine what was transmitted our signal must be stronger than the noise.  We describe this comparison in a ratio, the signal to noise ratio.
                             
                                S/N = Power of the Received Signal / Power of the NoiseS/N = P_s / P_n
 Where P_s is the received signal power and P_n is the power of the noise.
                             We hope the ratio is greater than one.  In fact the larger it is, the better.  We are not so much concerned with the power level of the noise as we are concerned about its 
                            relative strength in comparison to the signal.  That is why the signal to noise ratio is so important. Example:  What is the signal to noise ratio, S/N, if the signal power is 100 kW and the noise 
                            power is 100 W?
                             
                                S/N = P_s / P_n  = 100 kW / 100 WS/N = 100,000 W / 100 W
 S/N = 1000
 As you can see, the signal to noise ratio has no units because the units in the numerator and denominator must cancel out.  In the above problem we had to change from kilowatts to watts to 
                            get the units to cancel.  This is a good check on your work.  Ask yourself if the units canceled out properly. 
                            Another Example:  Half the power is lost due to attenuation when transmitting a signal over a 2 mile cable.  If the noise power is 100 mW and the absolute minimum signal to noise ratio is 10 to 1 
                            (or just 10), what is the minimum power level that you must transmit at?
                             
                                Draw a diagram to set this up ...
   Tx = Transmit PowerRx = Receive Power
 
                                First determine what the receive power must be:
                                 Half the transmitted power is received ...
                                 DecibelsOne problem encountered when working with transmit and receive signal powers is that huge numbers and tiny fractional numbers can be mixed into the same equation.  For example a satellite 
                            earth station might transmit at 100,000,000 watts and the received signal at the satellite might be as low as 0.000 000 000 001 watts.  The satellite would turn around and rebroadcast that signal 
                            back to earth.  Then what would have?  You would have nothing if the satellite didn't boost the power level during retransmission.
 
                            Scientific notation might be useful in solving such problems, but there is a better way to deal with such large and small numbers.  By using logarithms to reduce the magnitude of a number and 
                            convert multiplication and division operations into addition and subtraction we simplify our work a great deal.  We'll perform this operation by doing our calculation in decibels.  Appendix B at the 
                            end of Chapter 2 in Stallings discusses decibels. Definition of a bell:  A bell is the log_10 of a number.  There are 10 decibels in a bell.  So,
                             The equation above reads like this:  Y in dB equals 10 times log base ten of the original y.  Example:  If y = 345.2, what is the decibel value of y?
                             Another Example:  What is the value of y in dB if y = 437,476,300,000,000,000,000,000
                             
                                y_dB = 10 log(437,476,300,000,000,000,000,000)y_dB = 236.4 dB
 Yet Another Example:  What is the value of y in dB if y = 3.0 x 10^-45
                             Notice in the above examples that very large and very small numbers become common everyday values.  Also notice that we had not units in the examples.  Units are a pesky feature of decibels. 
                            Learn the following rules by heart:
                             
                                1) y_dB = 10 log( y )The same definition explained above
 2) Keep the units.If your number had watts attached to it, you get dBW.
 
                                    Example:  Convert 15 MW to a decibel value y_dB = 10 log(15 MW)y_dB = 11.8 dBMW, or
 y_dB = 10log(15,000,000 W)y_dB = 71.8 dBW
 3) dBm is short for dB miliwatts 4) dBm + dB = dBm 5)dBm - dB = dBm 6)dBm - dBm = dBMiliwatts divided by miliwatts gives you no units, they cancel
 
                                7) dBm + dBm gives you dB miliwatts squared.  We won't mess with this so don't let me catch you adding dBm to dBm. 8) dB + or - dB = dBA ratio times a ratio is a ratio.
 
                                9) Undo your dB with y = 10^( y_dB / 10 )The reverse of taking the log will return you to the normal numbers you love.
 Examples: a) 10 dB + 75 dBm = 85 dBmb) 500 dBm - 50 dBm = 450 dB
 c) 34 dBm - 90 dB = -56 dBm
 d) -75 dB - 25 dB = - 100 dB
 e) 45 dBkm + 50 dBkm = 95 dBkm^2
 (I understand square kilometers)
 f) 10 dB + 75 dB - 200 dBm - 7 dB + 38 dBm = - 84 dBm
 g) y_dB = -45 dBm. In absolute numbers
 y = 10^(-45/10) = 0.000032 miliwatts
 h) y_dB = 135.2 dBkm^2.  In absolute numbers
 y = 10^(135.2 / 10) = 33,113,000,000,000 square kilometers.
 
                            OK, so what you say.  Well, let's apply some of those logarithm manipulation equations and use decibels to simplify an equation. Example:  The equation below looks simple enough until you realize ...
                             ... the following:
                             
                                P = 400,000,000,000 Wh1 = 300,000,000 meters
 h2 = 188,000,000,000 meters
 L = 0.00000000000032
 r = 600,000,000 meters
 You can start cranking out the scientific notation, or you can do it in decibels rather quickly.
                             Plug in the equation for y and use the logarithm manipulation rules you reviewed earlier this semester.  You'll get the following:
                             Plug in each value and continue solving ...
                             Use the rules about adding and subtracting units of the same type.
                             You are finished, unless you want to know the number in absolute.  OK, here you go ...
                             Another Example:A satellite earth station transmits a 500 kW signal up to a satellite.  This signal arrives at one 
                            one-millionth the transmit power.  The satellite quadruples the power (4x) and sends it back to a second earth station.  Again, the signal arrives at one one-millionth the power level it was sent at.  
                            What power level is received at the second earth station.
 To solve this we'll follow the signal through the network.  We should solve this in absolute number first, then in decibels.
                             
                                Tx --->1/1,000,000 attenuation ---> Rx at Satellite --->4X boost at satellite ---> Satellite Tx --->
 1/1,000,000 atten. ---> Rx at second earth station
 Summary:Tx = 500,000 W
 Atten. = 1/1,000,000 = 0.000001
 Power boost = 4 times
 
                                Received at Satellite = Tx times AttenuationRx_sat. = 500,000 W x 0.000001 = 0.5 W
 Boost is 4XTx_sat. = 4 x 0.5 W = 2.0 W
 Received at Second Earth Station = Tx times AttenuationRx_E2 = 2.0 W x 0.000001 = 0.000002 W
 Rx_E2 = 2.0 uW
 What we actually did above is:
                             Now in decibels:
                             
                                Summary:Tx = 10log(500,000) = 57 dBW
 Att. = 10log(0.000001) = -60 dB
 boost = 10log(4) = 6 dB
 Rx_dB = 10log(Tx_E1 x Att. x 4 x Att.)Rx_dB = 10log(Tx) + 10log(Att.) + 10log(4) + 10log(Att.)
 Rx_dB = 57 dBW + (-60) + 6 + (-60)
 Rx_dB = -57 dBW
 Rx_abs. = 10(-57/10) = 0.000002 W
                             Another Example:What power must a signal be transmitted at the attenuation on the channel is -60 dB, the minimum signal to noise ratio is 130 dB and the noise level at the receiver is 3.0 x 10^-7 W?
 To solve this problem you have to work backwards.  First figure out what the minimum received power has to be so that it meets the signal to noise ratio requirement.
                             
                                S/N_dB = 130 dB = 10log(Rx / N) = 10log(Rx) - 10log(N)130 dB = 10log(Rx) - 10log(3 x 10^-7 W)
 Rx_dB = 130 dB + 10log(3 x 10^-7 W)
 Rx_dB = 130 dB + (-65 dBW)
 Rx_dB = 65 dBW
 Next work backwards to determine the transmit signal strength.  You lose 60 dB on the link, so...
                             If you want to, you can convert this back to an absolute number. Try this problem on your own:  On a 10 mile cable the power level falls 50% per mile.  This is 
                            equivalent to -3 dB/mile, and at 10 miles that's 30 dB lost.  What is the signal to noise ratio if a 100 mW signal is transmitted and the noise level at the receiver is 300 nW?  Calculate your answer in dB. 
                              Channel Capacity We've already seen that there is a relationship between bandwidth and data capacity.  We also 
                            know that our signal has to be stronger than the noise for us to detect and understand it.  How are these issues related? 
                            Fortunately (or unfortunately for the math-challenged) a man named Claud Shannon determined a way to calculate the maximum theoretical data rate given the bandwidth and the signal to noise 
                            ratio of the channel.  Here it is.
                             
                                C = 2W log_2( 1 + S/N )1/2
                                 
                                    Where:C = maximum theoretical data rate, bps or bits per second
 W is the range of frequencies available, bandwidth, in Hz
 S/N is the signal to noise ratio (not in decibels)
 Example:  Your telephone line allows signals between 300 Hz and 3300 Hz.  All other frequencies 
                            are filtered out.  Your modem knows this an limits itself to this range.  If the signal to noise ratio on the public switched telephone network (PSTN) is 250 to 1, what is the maximum data rate you 
                            can get with your modem?
                             
                                W = 3300 Hz - 300 Hz = 3000 HzS/N = 250 (given)
 So,
 C = 2Wlog_2(1+S/N)1/2 C = 2(3000)log_2(1+250)1/2
 C = 6000(1/2)[log_10(251)/log_10(2)]
 C = 3000(7.97)
 C = 23,900 bps
 Another Example: What is the S/N ratio on your phone line if your modem tells you it's operating at 53,000 bps and the bandwidth is 3000 Hz?
                             
                                W = 3000 Hz (given)C = 53,000 bps (given)
 C = 2Wlog_2(1+S/N)1/2 53,000 = (1/2)(2)(3000)log_2(1+S/N)
 53,000/3000 = log_2(1+S/N)
 17.67 = log_2(1+S/N)
 217.67 = (1+S/N)
 S/N = 217.67 -1
 S/N = 208,062
 Yet Another Example:  How much bandwidth do you need to get 1 Mbps when the S/N ratio is 200,000 to 1?
                             
                                C = 1,000,000 bps (given)S/N = 200,000 (given)
 W = ?
 C = 2Wlog_2(1+S/N)1/2 1,000,000 = Wlog_2(1+200,000)
 W = 1,000,000 / log_2(200,001)
 W = 56,787 Hz
 How many times greater is this than what the phone company currently gives you?
                             
                                You currently get 3000 Hz. 56,787 Hz = y(3000 Hz)y = 56,787 / 3000
 y = 18.9 times greater.
 Shannon's equation determines the maximum theoretical data rate, but it doesn't tell us how to achieve that data rate.  We have to choose a number of signals to send per second and the number 
                            of bits that each signal represents.  Then we may attempt to transfer data.
                             
                                Note: We may attempt to exceed the theoretical maximum, but we will encounter an unacceptable number of errors.
                             Comparing Attempted Data Rate with Theoretical Maximum The attempted data rate is the number of signals we send per second times the number of bits each 
                            that each signal represents.  From now on, we'll refer to a single signal as a symbol.  The collection of unique symbols is the symbol set.  
                             
                                R = r n Where R is the attempted data rate in bits per second, r is the rate we send symbols (unique signals) and n is the number of bits represented by a symbol.
                             Example: What data rate does your modem attempt when it sends 2400 symbols per second and each symbol represents 8 bits.
                             
                                R = r nR = 2400 symbols/sec  x 8 bits/symbol
 R = 19,200 bits per second
 Another Example: At what rate does your modem sends signals if the data rate is 53,000 bps and each symbol represents 9 bits.
                             
                                R = r n53,000 = r (9)
 r = 53,000 / 9
 r = 5888.9 symbols per second
 Why describe this as the "attempted" data rate?  While we may design a system at a certain data rate, that doesn't mean we meet the three physical limits discussed earlier. Keeping all other things the same ...
                             
                                * If we try to represent too many bits with a symbol, we will encounter an unacceptable error rate. * If we try to send too many symbols per second, we will encounter an unacceptable error rate.
                             Shannon determined two other theoretical limits.  These two limits determine the maximum number of symbols per second and the maximum number of bits per symbol that we can have.
                             These are limits on what we attempt to accomplish.  Please notice that C is the theoretical limit and R is what we attempt.  Therefore, we must meet the following restriction as well.
                             Example:  The signal to noise ratio is 250 to 1, the bandwidth is 3000 Hz.  These are typical values for the PSTN.  What are the three limits?
                             
                                #1r < 2W
 r < 2(3000)
 r < 6000 symbols per second
 #2n < log_2(1+S/N)1/2
 n < log_2(1+250)1/2
 n <
                                 (1/2)log_2(251)
 n < 4.0 bits per symbol
 #3R < C
 R < 2Wlog_2(1+S/N)1/2
 R < 6000(4)
 R < 24,000 bits per second
 Another Example:  Your network has a bandwidth of 10,000 Hz and the signal to noise ratio is 300 to 1.  You attempt to send 26,000 symbols per second with each symbol representing 3 bits.   
                            (a) What data rate are you attempting?
                             
                                R = r nR = 26,000 x 3
 R = 78,000 bits per second
 (b) Do you pass any of the three physical limits?
                             
                                #1r < 2W
 r < 2(10,000)
 r < 20,000 symbols per second
 r = 26,000, too high, FAIL
 #2n < log_2(1+S/N)1/2
 n < log_2(1+300)1/2
 n < (1/2)log_2(301)
 n < 4.1 bits per symbol
 n = 3 which is less than or equal to 4.1,  PASS
 #3R < C
 R < 2Wlog_2(1+S/N)1/2
 R <
                                 20,000(4)
 R < 80,000 bits per second
 R = 78,000 bps which is less than or equal to 80,000, PASS
 (c) Will this link function without a significant number of errors?
                             (d) What must be done to solve the problem?
                             
                                The number of symbols sent per second must be reduced to, at most, 20,000 symbols per second. Doing so will reduce your data rate to:
                                 
                                    R = 20,000 x 3 = 60,000 bits per second
                                 However, we can use up to 4 bits per symbol, so lets boost n up to 4.  That will bring us up to 80,000 bps, the maximum.
                             Summary:  There are three limits that we must meet in order to avoid a significant number of errors on our channel.  
                             From these three limits we may conclude:
                             
                                * Bandwidth limits the symbol rate* Signal to Noise ratio limits the number of bits per symbols
 * Together, these two limits set the limit on our total data rate
 * If we exceed any of these limits, we will encounter an unacceptable number of errors.
 Back to the top of the page. |