RFCD 101: CDMA Basics Technical data is subject to change Copyright@2003 Agilent Technologies Printed on Dec. 4, 2002 5988-8499ENA Although there are many types of spread spectrum communications systems, this presentation gives an overview of the proposed CDMA cellular system defined by the TIA 45.5 sub-committee. Largely based on the CDMA system developed by Qualcomm, the TIA system uses direct sequence modulation with digital codes to spread its spectrum. Other types of spread spectrum systems use frequency hopping techniques or a combination of frequency hopping and modulation with digital codes (direct sequence). The intent of this paper is to provide insight into the technology of CDMA and to describe some of the operating features of the now standardized TIA CDMA system. The TIA standard documents that define this system are EIA/TIA-95-B (the common air interface), EIA/TIA-98-B (mobile minimum performance standard), and EIA/TIA-97-B (base station minimum performance standard). For PCS system in the United States, the CDMA air interface standard is defined in ANSI J-STD-008. 1
Cellular Access Methods Power Time Power Time Power FDMA Time Frequency CDMA Frequency TDMA Frequency The Problem of Access: The personal communication industry is faced with the problem of an ever increasing number of users sharing the same limited frequency bands. To expand the user base, the industry must find methods to increase capacity without degrading the quality of service. The current analog cellular system uses a complex system of channelization with 30 khz channels, commonly called FDMA (Frequency Division Multiple Access). To maximize system capacity, analog FDMA cellular uses directive antennas (cell sectoring) and complex frequency reuse planning. To further increase system capacity, a digital access method is being implemented called TDMA (Time Division Multiple Access). This system uses the same frequency channelization as FDMA analog and adds a time sharing element. Each channel is shared in time by three users to effectively triple system capacity. CDMA stands for Code Division Multiple Access and uses correlative codes to distinguish one user from another. Frequency divisions are still used, but in a much larger bandwidth (1.25 MHz). In CDMA, a single user's channel consists of a specific frequency combined with a unique code. CDMA also uses sectored cells to increase capacity. One of the major differences in access is that any CDMA frequency can be used in all sectors of all cells. The correlative codes allow each user to operate in the presence of substantial interference. An analogy to this is a crowded cocktail party. Many people are talking at the same time, but you are able to listen to and understand one person at a time. This is because your brain can sort out the voice characteristics and differentiate them from the other talkers. As the party grows larger, each person must talk louder, and the size of the talk zone grows smaller. Thus, the number of conversations is limited by the overall noise interference in the room. CDMA is similar to this cocktail party analogy, but the recognition is based on digital codes. The interference is the sum of all other users on the same CDMA frequency, both from within and outside the home cell and from delayed versions of these signals. It also includes the usual thermal noise and atmospheric disturbances. Delayed signals caused by multipath are separately received and combined in CDMA. This will be discussed in greater detail later in this presentation. 2
CDMA is Also Full Duplex Amplitude US Cellular Channel 384 Reverse Link Forward Link AMPS Amplitude 45 MHz 836.52 MHz 881.52 MHz Reverse Link Forward Link Frequency CDMA 836.52 MHz 45 MHz 881.52 MHz Frequency Traditional cellular system are known as full duplex systems since two channels are used at the same time. This allows completely independent transmission to and from the mobile at the same time. In the North American Cellular system, the forward and reverse link channels are separated by 45 MHz. The EIA/TIA-95-B system is also full duplex and uses 1.25 MHz wide channels for both the forward and reverse directions. For cellular applications in North America, EIA/TIA-95-B CDMA uses the same 45 MHz separation between forward and reverse links that AMPS uses. For PCS applications, J-STD-008 CDMA uses a separation between forward and reverse links of 80 MHz. Other systems throughout the world may use other spacings. 3
Cellular Frequency Reuse Patterns 6 2 5 1 4 7 3 6 2 1 1 1 1 1 1 1 1 1 FDMA Reuse CDMA Reuse One of the major capacity gains with CDMA is due to its frequency reuse efficiency. In order to eliminate interference from adjacent cells, narrowband FM systems must physically separate cells using the same frequency. Complex frequency reuse planning must be done in such a system to maximize capacity while eliminating interference. A typical reuse pattern for analog and TDMA systems employs only one seventh of the available frequencies in any given cell. This could really be called frequency non-reuse. With CDMA, the same frequencies are used in all cells. If using sectored cells, the same frequencies can be used in all sectors of all cells. This is possible because CDMA is designed to decode the proper signal in the presence of high interference. Adjacent cells using the same frequency in CDMA simply cause an apparent increase in the channel background noise. By allowing the use of the same frequencies in every cell, CDMA has six times the capacity of existing analog cellular systems. 4
CDMA Capacity Gains Capacity = (Chan BW) X (1) X (1) X (Fr) (Data Rate) (S/N) (Vaf) (1,230,000) (1) (1) CDMA = X X X (0.67) (9,600) (5.01) (.40) CDMA = 42 Calls ( Using 1.5 MHz BW ) Processing Gain AMPS = 1.5 MHz / 30 khz = 50 Channels Capacity = 50 Channels / 7 ( 1/7 Frequency Reuse ) AMPS = 7 Calls ( Using 1.5 MHz BW ) To see how CDMA offers greater capacity, we need to look at its potential in a given bandwidth. Remember that for any cellular system, the capacity can be made arbitrarily large by adding more and more cells. A more realistic approach for a capacity comparison is the number of calls per used bandwidth. Installing CDMA in an existing AMPS analog cellular system requires that a minimum of 1.5 MHz of bandwidth be removed from analog service. While the actual spreading bandwidth of a single CDMA frequency is 1.23 MHz, a total of 1.5 MHz is required to provide guardbands to reduce potential interfere with adjacent analog channels. Additional CDMA frequencies added to the system will only require 1.23 MHz of bandwidth. In this configuration, a single CDMA cell will support 42 telephone calls. This is derived from the equation shown. The processing gain is the ratio of the CDMA final bandwidth divided by the encoded voice data rate. The signal-to-noise ratio required for good voice quality varies greatly with propagation conditions. On average, typical transmission conditions require a signal-to-noise ratio of about 7 db to provide adequate voice quality. Translated into a ratio, the signal must be 5 times stronger than the noise. The parameter V af is the voice activity factor. Since CDMA uses a variable rate voice encoder, V af for CDMA is 0.4. F r is the frequency reuse efficiency and S g is the sectorization gain. For CDMA, F r is 0.67 (in other words almost 70% reuse efficiency). The frequency reuse efficiency is not 1 since the additional interference produced by surrounding cells causes a reduction in capacity. If the CDMA cells use 3-way sectored antennas, S g is about 2.6 (almost 3 times the capacity when using sectorization). Again, the sectorization gain is not 3 due to the increased interference from the surrounding sectored cells. Given the same amount of bandwidth, an AMPS system has a capacity per cell of only about 7 calls. This is because although AMPS would have 50 channels in 1.5 MHz of bandwidth, only one-seventh can be used in any given cell because of interference. Using sectors in analog does not improve the capacity per MHz since interference from adjacent sector still requires a complex frequency reuse plan. Sectorization in analog simply results in physically smaller cells. 5
The CDMA Concept 10 Khz BW 1.23 Mhz BW 1.23 Mhz BW 10 Khz BW Baseband Data 0 f c f c 0 Encoding & Interleaving CDMA Transmitter Walsh Code Spreading CDMA Receiver Walsh Code Correlator Decode & De- Interleaving Baseband Data 9.6 19.2 1228.8 1228.8 19.2 9.6-113 dbm/1.23 Mhz Spurious Signals 1.23 Mhz BW 1.23 Mhz BW f c f c Background Noise External Interference Other Cell Interference Other User Noise Interference Sources f c f c CDMA starts with a narrowband signal, shown here at the full speech data rate of 9600 bps. This is spread with the use of specialized codes to a bandwidth of 1.23 MHz. The ratio of the spread data rate to the initial data rate is called the processing gain. For EIA/TIA-95-B standard CDMA, the processing gain is 21 db (10 log (1228800/9600)). When transmitted, a CDMA signal experiences high levels of interference, dominated by the coded signals of other CDMA users. This takes two forms, interference from other users in the same cell and interference from adjacent cells. The total interference also includes background noise and other spurious signals. When the signal is received, the correlator recovers the desired signal and rejects the interference. The correlators use the processing gain to pull the desired signal out of the noise. Since a signal to noise ratio of about 7 db is required for acceptable voice quality, this leaves 14 db of extra processing gain to extract the desired signal from the noise. This is possible because the interference sources are uncorrelated (orthogonal in the case of the forward link) to the desired signal. 6
What is Correlation? Is a Measure of How Well a Given Signal Matches a Desired Code The Desired Code is Compared to the Given Signal at Various Test times Received Signal Correlation = 1 Correlation = 0 Time Correlation = 0 Correlation = 0 Correlation is key enabling concept for direct sequence CDMA systems. Correlation is a measure of how well a given signal spread with a digital code matches a desired code. In the above example, a digital sequence is received and then compared to the desired code. This comparison takes place over a range of different times. When time aligned, the correlation is 1 indicating that an exact match occurred between the received signal and the desired signal. At other times the correlation is near zero, especially if the digital codes used to spread the waveform are designed properly. It is the fact that we can correlate to signals that enables direct spread CDMA to function. 7
CDMA Paradigm Shift Multiple Users on One Frequency Analog/TDMA Try to Prevent Multiple Users Interface Channel is Defined by Code Analog Systems Defined Channels by Frequency Capacity Limit is Soft aallows Degrading Voice Quality to Temporarily Increase Capacity areduce Surrounding Cell Capacity to Increase a Cell s Capacity Analog CDMA It should be clear at this point that CDMA technology is not intuitive. For most people familiar with FM communication systems, a paradigm shift is needed to properly discuss CDMA. Here are some of the key differences between CDMA and analog FM: Multiple users are on one frequency simultaneously. For a long time, RF engineers have spent a lot of effort trying to keep other people off the same channel. This is a critical issue for Analog FDMA and TDMA systems. Now, with CDMA, we are trying to put many conversations on the same channel. A Channel is defined by the correlative code in addition to the frequency. For a long time, people have thought of channels in terms of their frequency. With CDMA, channels are defined by various digital codes as well as by the frequency. The capacity limit is soft. In analog systems, when all of the available channels are in use, no further calls can be added. Capacity in CDMA can be increased with some degradation of the error rate or voice quality or can be increased in a given cell at the expense of reduced capacity in surrounding cells. 8
CDMA makes use of Diversity Spatial Diversity Making Use of Differences in Position Frequency Diversity Making Use of Differences in Frequency Time Diversity Making Use of Differences in Time Another aspect of CDMA is its use of diversity. CDMA uses three types of diversity: Spatial diversity, Frequency diversity, and Time diversity. CDMA uses the diverse nature of these three properties to enhance its system performance. The next slides will examine in detail how CDMA makes uses of the diverse nature of these three properties to enhance performance. 9
CDMA Spatial Diversity Diversity Reception: Multiple Antennas at Base Station aeach Antenna is Affected by Multipath Differently Due to Their Different Location aallows Selection of the Signal Least Affected by Multipath Fading If Diversity Antennas are Good, Why Not Use Base Stations as a Diversity Network? Soft Handoff The concept of diversity reception has been well known for some time. A diversity receiver uses multiple antennas at one reception site. Since these antenna are placed to be a non-integral number of wavelength apart, when one antenna is experiencing a multipath fade it is likely that the other antennas will not be in a fading condition. This leads to receiver designs where the antenna with the best signal is selected to be processed by the receiver. AMPS analog cellular base stations use this type of diversity for improved fading resistance. CDMA also employs diversity reception for base stations. One of the most problematic locations for a cellular phone is in between cells where handoffs occur. If the mobile experiences a deep fade during handoff, a dropped call can result. If diversity reception is useful at a single receiver location, then can using multiple base station be used in a diversity network to help phone during handoffs? The answer is yes: CDMA seeks to overcome the handoff problem by using two or three base stations as a giant diversity system. Using multiple base stations simultaneously talk to the mobile during a handoff is known as a "soft handoff". 10
Spatial Diversity During Soft Handoff MTSO Land Link Vocoder / Selector Base Station 1 Base Station 2 CDMA extends the idea of diversity reception with the concept of soft handoff. In the slide, a mobile CDMA phone has established a call with base station one. As the mobile moves away from base station one and approaches base station two, a device in the phone known as the searcher identifies base station one as a good candidate for soft handoff. The searcher identifies other base stations as good candidates for soft handoff when the received level exceeds the T_add (Threshold for adding a candidate cell for soft handoff) parameter of the system. Once a candidate exceeds the threshold, the phone sends the candidate information to the Mobile Telephone Switching Office (MTSO) via base station one. If the network has available capacity, the MTSO then directs the base stations and mobile to perform a soft handoff. During soft handoff, the mobile listens to the two cells on different codes while the base stations each listen to the same transmission from the mobile. The signals from the base to mobile are treated as multipath signals and are coherently combined at the mobile unit. Each base station sends its received signal via the network to the (MTSO), where a quality decision is made on a frame-by-frame basis, every 20 msec. The MTSO selects the better frame from the two signals returned from the base stations. Thus the two base stations act like a giant antenna diversity system. This helps to overcome the fading problem that occurs between cells where handoffs must take place. As the mobile moves further away from base station one, the searcher in the phone will determine that its power has dropped below the system parameter T_drop. The T_drop information is sent to the MTSO, which then directs the soft handoff be terminated. This allows for smooth handoffs between cells that the user is totally unaware of. Of course there is a price to pay for this clever design: the system uses more capacity for each soft handoff made and there is greatly increased network traffic between CDMA cellsites and the MTSO. 11
CDMA Frequency Diversity Combats Fading, Caused by Multipath Fading Acts like Notch Filter to a Wide Spectrum Signal May Notch only Part of Signal Amplitude 1.23 MHz BW Frequency Frequency diversity is inherent in a spread spectrum system. A fade of the entire signal is less likely than with narrow band systems. Fading is caused by reflected images of an RF signal arriving at the receiver such that the phase of the delayed (reflected) signal is 180 degrees out of phase with the direct RF signal. Since the direct signal and delayed signal are out of phase, they cancel each other causing the amplitude seen by the receiver to be greatly reduced. In the frequency domain, a fade appears as a notch filter that moves across a band. As the user moves, the frequency of the notch changes. The width of the notch is on the order of one over the difference in arrival time of two signals. For a 1 usec delay, the notch will be approximately 1 MHz wide. The TIA CDMA system uses a 1.25 MHz bandwidth, so only those multipaths of time less than 1 usec actually cause the signal to experience a deep fade. In many environments, the multipath signals will arrive at the receiver after a much longer delay. This means that only a narrow portion of the signal is lost. In the display shown, the fade is 200 to 300 khz wide. This results in the complete loss of an analog or TDMA signal but only reduces the power in a portion of a CDMA signal. As the spreading width of a CDMA signal increases, so does its multipath fading resistance. Many spread spectrum systems use a 5 or 10 MHz wide channel to further improve fading resistance. 12
CDMA Time Diversity Rake Receiver to Find and Demodulate Multipath Signals Data is Interleaved Spreads Adjacent Data in time to Improve Error Correction Efficiency Convolutional Encoding Adds Error Correction and Detection Viterbi Decoding Most Likely Path Decoder for Convolutionaly Encoded Data Time diversity is a technique common to most digital transmission systems. The Rake Receiver is used to find and demodulate multipath signals that are time delayed from the main signal. The Rake Receiver will be explained in the next slides. Transmitted signals are spread in time by use of interleaving. Interleaving the data improves the performance of the error correction by spreading errors over time. Errors in the real world during radio transmission usually occur in clumps, so when the data is de-interleaved, the errors are spread over a greater period of time. This allows the error correction to fix the resulting smaller, spread out errors. Forward error correction is also applied to the transmitted data. This is usually done by adding parity bits that allow received errors to be detected and to some extent corrected. Performance of the receiver can be further enhanced by using a maximal likelihood detector. The particular scheme used for CDMA is convolutional encoding in the transmitter with Viterbi decoding using soft decision points in the receiver. 13
Why Interleaving Works Original Data Frame 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 TX RX Errors/Time 1 2 3 4 5 6 7 8 910 11 12 13 14 15 16 Errors/Time 1 2 3 4 5 6 7 8 910 11 12 13 14 15 16 Interleaved Data Frame 1 2 3 4 1 5 9 13 5 6 7 8 2 6 10 14 9 10 11 12 3 7 11 15 13 14 15 16 4 8 12 16 TX RX Errors/Time 1 5 9 13 2 6 10 14 37 11 15 4 8 12 16 Errors/Time 1 2 3 4 5 6 7 8 910 11 12 13 14 15 16 This slide graphically demonstrates why interleaving data improves error correction performance of data transmission systems. In the top example, data is sequentially read out of a buffer than goes by rows. No interleaving is employed. The data is read and transmitted in numerical order. During transmission, data blocks 5 through 8 are corrupted by some interference. When the data is received, blocks 5 through 8 are lost and the error correction is insufficient to recover such a large block of lost data. In the lower example, the same data is first interleaved using a simple pattern of reading the rows into columns. The interleaved data is then read out by row and transmitted. During transmission the data in the same four blocks is corrupted by interference. However, the blocks that were lost are no longer sequential. Blocks 2, 6, 10, and 14 were lost. When the interleaved signal is received, the receiver reverses the interleaving process to restore the data to its original sequential pattern. Notice what happens after de-interleaving: the lost blocks are now spread in time resulting in small, isolated error locations. Now the limited error correction built into the signal can correct the errors. Interleaving makes the most use of the error correction built into a data transmission system. 14
The Rake Receiver Amplitude Time Frequency Instead of trying to overpower or correct multipath problems, CDMA takes advantage of the multipath to improve reception quality in fading conditions. CDMA does this by using multiple correlating receivers and assigning them to the strongest signals. This is possible because the CDMA mobile is synchronized to the serving base station. The mobile's receiver can distinguish direct signals from multipath signals because the reflected multipaths signals arrive later than the direct signals. Special circuits called searchers are also used to look for alternate multipaths and for neighboring base station signals. The searchers slide around in time until they find a strong correlation with their assigned code. Once a strong signal is located at a particular time offset, the searcher assigns a receiver element to demodulate that signal. The mobile receiver uses three receiving elements, and the base station uses four. This multiple correlator system is called a rake receiver. As conditions change the searchers rapidly reassign the rake receivers to handle new reception conditions. Instead of trying to overpower or correct multipath problems, CDMA takes advantage of the multipath to improve reception quality in fading conditions. CDMA does this by using multiple correlating receivers and assigning them to the strongest signals. This is possible because the CDMA mobile is synchronized to the serving base station. The mobile's receiver can distinguish direct signals from multipath signals because the reflected multipaths signals arrive later than the direct signals. Special circuits called searchers are also used to look for alternate multipaths and for neighboring base station signals. The searchers slide around in time until they find a strong correlation with their assigned code. Once a strong signal is located at a particular time offset, the searcher assigns a receiver element to demodulate that signal. The mobile receiver uses three receiving elements, and the base station uses four. This multiple correlator system is called a rake receiver. As conditions change the searchers rapidly reassign the rake receivers to handle new reception conditions. 15
Rake Receiver Design Antenna T 0 T 1 T 2 T 3 T 4 Delay Taps W 0 W 1 W 2 W 3 W 4 Tap Weights Output The design of a rake receiver can be visualized as a series of time delayed correlator taps fed from a common antenna. If each correlator tap is delayed to match the arrival of a particular transmitted signal, then the outputs of each tap can be recombined in phase. Once an RF signal with a particular travel time is locked onto by the correlator tap, an estimate of the gain or loss experienced by that signal must be made. The weighting of the taps perform this gain normalization function. Once adjusted, the outputs of each finger of the rake can be combined to form a better version of the transmitted signal. Notice that this description visually matches the analogy of a common garden rake with each tap forming a tine of the rake, hence the name rake receiver. Another form of time diversity occurs in the base station when transmitting at reduced data rates. When transmitting at a reduced data rate (more detail will be presented on this later), the base station repeats the data resulting in full rate transmission. The base station also reduces the transmitted power when it operates at reduced data rates. This added redundancy in the transmitted signal results in less interference (power is lowered) and improves the CDMA mobile's station receiver performance during high levels of interference. 16
Synchronization All Direct Sequence, Spread Spectrum Systems Should be Accurately Synchronized for Efficient searching Finding the Desired Code Becomes Difficult without Synchronization In order for any direct sequence, spread spectrum radio system to operate efficiently, all mobiles and base stations should be precisely synchronized. If they are not synchronized, it becomes difficult to search for the codes used to identify individual radio signals. Precise synchronization also leads to other benefits. It will allow such future services as precise location reporting for emergency or travel usage. 17
Reverse Link Power Control Maximum System Capacity is Achieved if: All Mobiles are Power Controlled to the Minimum Power for Acceptable Signal Quality As a Result, all Mobiles are Received at About Equal Power at the Base Station Independent of Their Location Two Types of Control Open Loop Power Control Closed Loop Power Control Open & Closed Loop Power Control are Always Both Active One of the fundamental enabling technologies of CDMA is power control. Since the limiting factor for CDMA system capacity is the total interference, controlling the power of each mobile is critical to achieve maximum capacity. CDMA mobiles are power controlled to the minimum power that provides acceptable quality for the given conditions. As a result, each mobile's signal arrives at the base station at approximately equal levels. In this way, the interference from one unit to another is held to a minimum. Two forms of power control are used for the reverse link: open loop power control, and closed loop power control. Despite what seems logical, both open and closed loop power control are active at the same time once a traffic channel is established. Both are constantly active and controlling the power of the phone according to their respective control algorithms. 18
Open Loop Power Control Assumes Loss is Similar on Forward and Reverse Paths Receive Power + Transmit Power = -73 All Powers in dbm Example: For a Received Power of -85 dbm Transmit Power = (-73) - (- 85) Transmit Power = +12 dbm Provides an Estimate of Reverse TX Power for Given Propagation Conditions Open loop power control is based on the similarity of the loss in the forward path to the loss in the reverse path (forward refers to the base-to-mobile link, while reverse refers to the mobile-to-base link). Open loop control sets the sum of transmit power and receive power to a constant, nominally -73, if both powers are in dbm. A reduction in signal level at the receive antenna will result in an increase in signal power from the transmitter. For example, assume the received power from the base station is -85 dbm. This is the total energy received in the 1.23 MHz receiver bandwidth. It includes the composite signal from the serving base station as well as from other nearby base stations on the same frequency. The open loop transmit power setting for a received power of -85 dbm would be +12 dbm. Thus open loop power control adjusts the transmit power of the phone to match the propagation conditions that the phone is experiencing at any given time. By the TIA/EIA-98 standard specification, the open loop power control slew rate is limited to roughly match the slew rate of closed loop power control directed by the base station. This eliminates the possibility of open loop power control suddenly transmitting excessive power in response to a receiver signal level dropout. 19
Closed Loop Power Control Directed by Base Station Updated Every 1.25 msec Commands Mobile to Change TX Power in +/- 1 db Step Size Fine Tunes Open Loop Power Estimate Power Control Bits are Punctured over the Encoded Voice Data Puncture Period is Two 19.2 Symbol Periods = 104.2 usec Closed loop power control is used to allow the power from the mobile unit to deviate from the nominal as set by open loop control. This is done with a form of delta modulator. The base station monitors the power received from each mobile station and commands the mobile to either raise power or lower power by a fixed step of 1 db. This process is repeated 800 times per second, or every 1.25 msec. The power control data sent to the mobile from the base station is added to the data stream by replacing the encoded voice data. This processes in called "puncturing", since the power control data is written into the data stream by over writing the encoded voice data. The power control data occupies 103.6 micro-seconds of each 1.25 millisecond of data transmitted by the base station. Because the mobile's power is controlled to be no more than is needed to maintain the link at the base station, a CDMA mobile typically transmits much less power than an analog phone. The base station monitors the received signal quality 800 times per second and directs the mobile to raise or lower its power until the received signal quality is just adequate. This operating point varies with propagation conditions, the number of users, and the density and loading of the surrounding cells. Analog cellular phones need to transmit enough power to maintain a link even in the presence of a fade. Most of the time, analog phones transmit excess power. CDMA radios are controlled in real time and kept at a power level to just maintain a quality transmission based on the changing RF environment. This has the benefit of longer battery life and smaller, lower cost amplifier design. If recent health concerns over cellular phone radiation are determined to have some basis in fact, CDMA will be preferred because of its much lower RF output power. 20
CDMA Variable Rate Speech Coder DSP Analyzes 20 Millisecond Blocks of Speech for Activity Selects Encoding Rate Based on Activity: ahigh Activity Full Data Rate Encoding (9600 bps) asome Activity Half Data Rate Encoding (4800 bps) alow Activity Quarter Data Rate Encoding (2400 bps) ano Activity 1/8 Data Rate Encoding (1200 bps) How Does This Improve Capacity? Mobile Transmits in Bursts of 1.25 ms System Capacity Increases by 1/Voice Activity Factor CDMA takes advantage of quiet times during speech to raise capacity. A variable rate vocoder is used; for the original vocoder the channel is a 9,600 bps when the user is talking. When the user pauses, or is listening, the data rate drops to only 1,200 bps. Data rates of 2,400 and 4,800 bps are also used, though not as often as the other two. The CDG 14.4 vocoder is similar with the four channel rates running at 14,400, 7,200, 3,600, and 1,800 bps. The data rate is based on speech activity and a decision as to the appropriate rate is made every 20 msec. Normal telephone speech has approximately a 40% activity factor. The mobile station lowers its data rate by turning off its transmitter when the vocoder is operating at less than 9,600 bps. Thus CDMA mobiles also operate in a TDMA mode (pulsing) when the vocoder determines that the transmission rate required for a given frame is less than full rate. At 1,200 bps, the duty cycle is only one-eighth of the full data rate. The choice of time for this duty cycling is stochastic based on a pseudo random algorithm. This has the affect of randomizing the transmission times of each mobile. When averaged over many users, the average transmitted power is lowered. Lowering the transmit power at the mobile reduces the level of interference for all other users. This increases the capacity of CDMA by nearly a factor of two. 21
Mobile Power Bursting Each Frame is Divided into 16 Power Control Groups Each Power Control Group Contains 1536 Chips (represents 12 encoded voice bits) Average Power is Lowered 3 db for Each Lower Data Rate CDMA Frame = 20 ms Full Rate Half Rate Quarter Rate Eighth Rate Each 20 millisecond frame in EIA/TIA-95-B CDMA is divided into sixteen "power control groups". When the mobile transmits, each power control group contains 1536 data symbols (chips) at a rate of 1.2288 Mbps. When the voice coder moves to a lower date rate, the CDMA mobile bursts its output by only sending the appropriate number of power control groups. For example, at quarter rate, only four of the sixteen power control groups are transmitted. Remember that the exact location of the transmitted groups is randomized to spread the transmitted power over time. For each lowering of the data rate, the average transmitted power is reduced by 3 db. 22
Base Station Variable Rate Vocoder Base Stations Do Not Pulse TX Channels How Does the Base Station Handle Variable Rate Vocoding? Repeats Data Bits When Transmitting at Reduced Rates Repeating Data Adds 3 db Coding Gain Lowers the TX Power 3dB for Each Lower Rate The base station uses a slightly different scheme when the vocoder moves to lower rates. First, EIA/TIA-95-B CDMA base stations do not pulse their transmissions. Rather, base stations repeat the same bit patterns as many times as needed to get back to the full rate of 9,600 bps. So, if the vocoder selects a frame to be half-rate, the data bits are sent twice to fill the entire frame. The transmit power is then adjusted down by 3 db, since repeating the data twice adds three db more processing gain to the signal (21 db + 3 db = 24 db for a half rate frame). Adjusting the gain down maintains the approximate same signal to noise ratio that existed for a full rate frame. Quarter and one-eighth rate frames repeat the data four and eight times to fill each frame, and are lowered in power by 6 and 9 db respectively. This allows more capacity on the forward link since frames operating below full rate are transmitted with lower power, which reduces the total interference. 23
Forward Link Traffic Channel Physical Layer Vocoded Speech Data 20 msec blocks Convolutional Encoder Interleaver 19.2 Long Code Scrambling 9.6 14.4 1/2 Rate 3/4 Rate 19.2 Long Code 19.2 19.2 Power Control Puncturing 800 bps Walsh Coder 19.2 P.C. Mux 800 bps 19.2 1.228 8 Walsh Code Mbps Generator 1.228 8 Mbps 1.2288 Mbps I Short Code FIR Short Code Scrambler FIR Q Short Code I Q 1.2288 Mbps The next section will follow the digitally encoded voice data through the encoding process for a forward link traffic channel. Voice data at 9600 bps or 14400 bps (full rate) is first passed through a convolutional encoder, which doubles the data rate for the 9600 bps case or increases it by 1.33 times for the 14400 bps case. It is then interleaved, a process that has no effect on the rate, but does introduce time delays in the final reconstruction of the signal. The interleaving processes increases the effectiveness of the convolutional encoder. A long code is XOR'ed with the data, which is a voice privacy function and not needed for channelization. The closed loop power control data is then punctured into the data stream using the long code to determine the exact location of the power control bits. CDMA then applies a 64 bit Walsh code which is uniquely assigned to a base to mobile link to form one channel. This sets a physical limit of 64 channels on the forward link. If the coded voice data is a zero, the Walsh sequence is output; if the data is a one, the logical not of the Walsh code is sent. The Walsh coding yields a data rate increase of 64 times. The data is then split into I and Q channels, and scrambled with short codes. The final signals are passed through a low pass filter, and eventually sent to an I/Q modulator. We will now take a closer look at each of the steps the base station takes to create the final transmitted CDMA signal. 24
CDMA Vocoders Vocoders Convert Voice to/from Analog Using Data Compression There are Three CDMA Vocoders: IS-96A Variable Rate (8 maximum) CDG Variable Rate (13 maximum) EVRC Variable Rate (improved 8 ) Each has Different Voice Quality: IS-96A - moderate quality EVCR - near toll quality CDG - toll quality All digital communication systems use various processes to convert analog voice signals to and from digital form. Long distance telephone systems have used 8-bit PCM (Pulse Code Modulation) for many years to provide high quality voice transmission. Most PCM systems sample with eight bit resolution to convert voice into a digital data stream of 64. In recent years, ADPCM (Adaptive Delta PCM) has become a popular alternative to straight PCM since it provides essentially the same voice quality as PCM while using only 32. This allows more voice channels to be sent on the digital network with quality loss. CDMA now has three vocoder standards for converting voice to digital form while providing a high degree of data compression. The original vocoder as defined in IS-96A is a variable rate vocoder with a maximum rate of approximately 8. This is quite an improvement over PCM or ADPCM encoders (four to eight times more efficient). However, because of the variable rate nature of this vocoder, the average bit rate is under 4! The new CDG (CDMA Development Group) vocoder is also a variable rate vocoder but uses a maximum data rate of 13 to provide essentially toll quality voice. The most recent addition is the EVRC (Enhanced Variable Rate Coder) that retains the maximum data rate of 8 but yields voice quality just slight less than the CGD 13 vocoder. 25
9600 bps Frame Mixed Mode Bit CDMA Frame Formats 192 bits in a ms frame 171 12 8 266 12 8 Information Bits CRC Encoder Tail Bits 14400 bps Frame 1-bit Reserved Mixed Mode Bit 288 bits in a ms frame Information Bits CRC Encoder Tail Bits 4800 bps Frame Mixed Mode Bit 96 bits in a ms frame 79 8 8 Information Bits CRC Encoder Tail Bits 7200 bps Frame 1-bit Reserved Mixed Mode Bit 144 bits in a ms frame 124 10 8 Information Bits CRC Encoder Tail Bits 2400 bps Frame Mixed Mode Bit 48 bits in a ms frame 39 Information Bits 8 Encoder Tail Bits 3600 bps Frame 1-bit Reserved Mixed Mode Bit 72 bits in a ms frame 54 Information Bits 8 8 CRC Encoder Tail Bits 1200 bps Frame Mixed Mode Bit 24 bits in a ms frame 15 Information Bits 8 Encoder Tail Bits 1800 bps Frame 1-bit Reserved Mixed Mode Bit 36 bits in a ms frame 20 Information Bits 6 8 CRC Encoder Tail Bits Once the analog voice is compressed by one of the vocoder processes, some additional data is added to produce a frame. Each frame in CDMA is 20 milliseconds regardless of the data rate used. This slide shows all of the possible frame configurations for both the 8 and 13 vocoders. In the case of the 8 vocoder running at full rate, each frame consists of a mixed mode bit, 171 vocoder bits, 12 bits of CRC, and 8 encoder tail bits. The result is a frame of 192 bits which produces a continuous date rate of 9,600 bps ( 192 bits x 50 frames/sec = 9,600 bps). The mixed mode bit indicates if the frame is pure channel data or if it contains at least some signaling. The CRC bits allow the mobile to verify that it has correctly decoded a frame. The encoder tail bits provide 8 zeroes in a row to flush the contents of the convolutional coder in preparation for processing the next frame of data. For the CDG 13 coder, the frame is still 20 milliseconds, but a total of 288 bits are sent to produce a data rate of 14,400 bps ( 288 bits x 50 frames/sec = 14,400 bps). The 14,400 bps channel has the mixed mode bit, CRC bits, and encoder tail bits like the 9,600 bps channel. However, the 14,400 bps channel includes a CRC for all data rates while the 9,600 bps channel only provides a CRC check for full and half rate frames. The 14,400 bps channel also includes 1 bit that is reserved in the forward link but is used by the mobile in the reverse link to indicate a frame erasure (the CRC check did not pass). This assists the base station in performing forward link power control efficiently. 26
Forward Error Protection Uses Half-Rate Convolutional Encoder Outputs Two Bits of Encoded Data for Every Input Bit Data Out 9600 bps Data In 9600 bps D D D D D D D D Data Out 9600 bps Unlike many digital cellular systems, CDMA provides powerful error correction to all voice data bits. This is desirable in CDMA since the idea is to increase the occupied bandwidth (spread the data). The forward link uses a half-rate convolutional encoder to provide error correction capabilities. This type of encoder accepts incoming serial data and outputs encoded data derived from a series of delay taps and summing nodes. A half-rate encoder produces two output symbols for every symbol input. For the CDMA forward link, the half-rate encoder produces two 9,600 bps serial data streams when driven by a single 9,600 bps data stream. These two 9,600 serial data streams are combined at a higher rate to produce a single 19,200 bps data stream. The resulting redundancy in the digital data after convolutional encoding imparts powerful error correction capability to the TIA CDMA cellular system. 27
14.4 Traffic Channel Forward Link Modifications Replaces 8 Vocoder with a 13 Vocoder (both Variable Rate) Effects: Provides Toll Quality Speech Uses a 3/4 Rate Encoder Reduces Processing Gain 1.76 db Results in Reduced Capacity or Smaller Cell Sizes Vocoded Speech Data Convolutional Encoder 3/4 rate 14.4 19.2 20 msec blocks In an effort to provide CDMA with even greater voice quality, the CDG (CDMA Development Group) has proposed and implemented a new vocoder. This new vocoder uses and improved, higher data rate of approximately 13 to digitized voice signals. After adding bits used to support the traffic channel, the final traffic channel data rate with this new vocoder is 14.4. To accommodate this new vocoder with the least impact to the existing 9.6 traffic channel structure, the CDG simply modified the convolutional encoder rate from a one-half rate to a threequarter rate encoder. Thus, the output from the convolutional encoder is still the same 19.2 used in the original CDMA system. No other changes are required in the coding structure which simplifies the implementation of this new voice quality mode. Testing has shown that the improvements of the 14.4 vocoder result in voice quality that is the equivalent of good land-line long distance telephony! Obviously, this level of voice quality will be a distinct marketing advantage for CDMA in the highly competitive cellular and PCS markets. However, the voice quality improvement does not come for free. By reducing the level of error correction provided in the convolutional encoder, the overall processing gain is reduced. In this case the overall processing gain is lowered from 21.07 db to 19.31 db. The result of lower processing gain is that something must give: either the capacity is reduced or the cell sizes must be reduced. The capacity loss for this reduction in processing gain is 1/3 (run the formula on slide #4 for 14,400 bps and you will see the result). Both of these choices have negative effects for operators: the cost to support the same number of uses will increase due to the need to install more cell sites. CDMA network operators will have to balance the benefits of this new vocoder against the costs of implementing it. One possible solution is the EVRC (Enhanced Variable Rate Coder). This new 8 coder promises to produce voice quality equal to the 13 coder without losing processing gain. The TIA committee is in the process of standardizing the 13 coder and is working on selecting the EVRC vocoder design. 28
CDMA System Time How Does CDMA Achieve Synchronization for Efficient searching? ause GPS Satellite System Base Stations Use GPS Time via Satellite Receivers as a Common Time Reference 11 10 9 8 7 12 6 1 2 3 4 5 GPS Clock Drives the Long Code Generator As mentioned earlier, both mobiles and base station in direct sequence CDMA must be synchronized. In the IS-95A system, synchronization is based on the Global Positioning Satellite system time. Each CDMA base station incorporates a GPS receiver to provide exact system timing information for the cell. The base station then sends this information to each mobile via a special channel. In this manner, all radios in the system can maintain near perfect synchronization. Most designs also include atomic clocks to provide a backup timing reference in the event that an insufficient number of GPS satellite can be received. These are capable of maintaining synchronization for up to several hours. The GPS clock used for CDMA system time is then used to drive the long code pseudo-random sequence generator. 29
Long Code Generation Long Code Output Modulo-2 Addition User Assigned Long Code Mask 42 bits 42 41 5 4 3 2 1 Long Code Generator The Long Code is generated using a 42 bit linear feedback shift register. The pseudo-random data pattern that it generates repeats every 41.43 days! This is the master clock and is synchronized in all CDMA radios. The GPS receiver's clock output drives the long code's shift register. In order to provide voice privacy, a user specific 42 bit mask is AND'ed with the output of the long code generator to create a unique long code. The large size of the user mask allows for a very large number of unique codes (about 4.29 billion since only the bottom 32 bits are used for the public mask). This is enough codes to ensure that each user can have their own unique scrambling code. There are two types of masks: the public mask and the private mask. The user's public mask is derived from the CDMA mobile's ESN. This provides nominal voice privacy. To provide truly secure voice scrambling, the private mask is formed from a combination of ESN, a random seed, and cryptological processing algorithms. This cryptological processing is the same authentication process that was developed for the AMPS analog system (using, of course, CDMA channels to transmit the information). CDMA authentication is based on an "A" key that is programmed into the phone and is known by the network. The network can then challenge the phone and then compare the result returned from the phone to verify that the phone is legitimate. The phone uses its internal "A" key to process the challenge and return a valid result. 30
Long Code Scrambling User s Long Code Mask is Applied to the Long Code Masked Long Code is Decimated Down to 19.2 Decimated Long Code is XOR ed with Voice Data Bits Scrambles the Data to Provide Voice Security Encoded Voice Data Long Code Generator 19.2 XOR 1.2288 Mbps 19.2 Long Code Decimator 19.2 In the forward link the long code is used to scramble the voice data to provide some measure of security. However, the long code is not used to spread the signal bandwidth in the forward channel. To accomplish the scrambling without increasing the data rate, the long code is decimated down to a lower rate after the user's unique long code mask is applied. The decimation is accomplished by essentially using every sixty-fourth bit out of the long code data stream. A sixty four times decimation reduces the data rate of the long code from 1.2288 Mbps to 19.2. At this point the long code data rate matches that of the encoded voice data it is exclusive OR'ed with. 31
Closed Loop Power Control Puncturing Long Code is Decimated Down to 800 bps Decimated Long Code Controls the Puncture Location Power Control Bits Replace Voice Data Voice Data is Recovered by the Mobile s Viterbi Decoder Long Code Scrambled Voice Data Long Code Decimated Data 19.2 19.2 Closed Loop Power Control Bits 800 bps P. C. Mux 800 bps Long Code Decimator 19.2 Once the data has been scrambled with the user specific long code, the closed loop power control data is then punctured into the data stream. Remember that the power control bits are sent every 1.25 milli-seconds - once in every power control group (a CDMA frame is 20 milliseconds with each frame having 16 1.25 milli-second power control groups). In each 1.25 millisecond power control group there are 24 modulation symbols of data (the data stream at this point is 19.2 so each of the 24 symbols has a period of 52.08 micro-seconds). The power control bits are placed somewhere in the first 16 modulation symbols in each power control group. The exact location of the power control bits are determined by decimating the long code down to a rate of 800 bps and then using the data to point to one of the modulation symbol locations. For a 9.6 voice channel, two modulation symbols are punctured allowing the power control data to be sent twice. For 14.4 voice channels, only a single modulation symbol is punctured with the power control bit. Since the power control bits replace the encoded voice data, holes (missing data) are introduced into the data stream from the receiver's point of view. These holes are accepted and the system uses the Viterbi decoder in the receiver to restore the data lost by puncturing. The recovery of the missing data uses some of the available processing gain in the system. This results in a loss of capacity, but the loss has been accounted for in the system's design. Another way to think of this is that slightly more power is required to maintain the link because of the missing data introduced by the power control puncturing. The power control data is only sent once in the 14.4 case since the reduced processing gain results in higher power being transmitted from the base station to maintain an acceptable signal to noise ratio. The higher power results in a much lower symbol error rate and the need to send the power control data twice is eliminated. 32
Walsh Codes W = 2n W n W W W n n n W = 0 1 W = 0 0 2 W = 4 0 1 0 0 0 0 0 1 0 1 0 0 1 1 0 1 1 0 An important feature of the forward link is the use of Walsh codes. These codes have the desirable characteristic of being orthogonal to each other and to the logical NOT of each other. Walsh code sets are generated by the Hadamard matrix expansion shown below: W 2n = W n W n W n W n The variable, n, in this expansion must always be a power of two. This is seeded with the one by one matrix: W 1 = 0 Each higher order Walsh code set is created by placing the entire set into the first three matrix positions and then by placing an inverted set into the lower right hand matrix position. Do not confuse this matrix with some type of matrix math operation. It is simply a place holder to allow the creation of orthogonal code sets of every increasing size. EIA/TIA-95-B standard CDMA uses Walsh code set 64. This set has 64 unique codes with each code having 64 bits. Notice that for each Walsh code set, the first code is composed entirely of zero data bits. Two functions are said to be orthogonal if the cross-correlation coefficient between the two functions is zero. The cross-correlation coefficient for generalized time variant functions is: z ij = 1 T 0 T f i ( t) fj( t) dt 33