Comparing Current and Emerging CDMA Forward Link PHYs

Comparing Current and Emerging CDMA Forward Link PHYs
By Marios C. Agathangelou, Racal Instruments Wireless Solutions, CommsDesign.com
November 12, 2003 (12:39 p.m. EST)
URL: http://www.eetimes.com/story/OEG20031112S0026

Like other wireless standards, the CDMA air interface has gone through a major overhaul in the last few years. In order to support 3G class services, the CDMA community have added standards like cdma2000 1X, cdma2000 1xEV-DO, and cdma2000 1x-EV-DV, each delivering a different feature set to 3G networks.

In all of the CDMA standards, Walsh codes are used in the conversion of higher layer data bits into an RF signal that a mobile can process (also known as the forward link). However, as the CDMA standards evolve, the way in which these Walsh codes are used can vary greatly in the forward link physical layer, creating headaches for designers.

In this article, we'll compare the forward link physical layers of cdmaOne, cdma2000 1x, cdma2000 1xEV-DO, and cdma2000 1xEV-DV systems. We'll then show how Walsh coding techniques are used differently in each system.

CDMA System Architecture
In line with many communication systems, CDMA uses a fairly layered architecture. In this article, we'll focus on the forward link physical layer (PHY) of a CDMA system.

The PHY on the forward link is responsible for transferring streams of data bits from the base station to the mobile phone over the air. Common sources of data are computer files, sampled speech, and messages from the signaling layer (otherwise known as layer 3). Signaling messages are essential for the operation of the system because they provide the mobile phones with information such as the frequencies of the channels that the base station is transmitting at, and notification of incoming calls.

Data from various sources, such as a vocoder in the case of speech, a Web server in the case of computer files, or the signaling layer are forwarded to the link access control (LAC) and medium access control (MAC). Broadly speaking, these layer 2 tasks decide how to order and prioritise the data, construct frames, number frames, and forward frames to the PHY.

The PHY on the transmit si de has a very simple function: to convert the frames received from the MAC layer into an RF signal. On the receiver side, it has to extract data bits out of the received RF signal (and noise) and give them to the MAC layer. In this article, we'll be looking solely at the transmit side.

cdmaOne—The First Spec
The first CDMA system, branded cdmaOne, was introduced through the IS-95 standard, which quickly evolved to the IS-95-A standard, both of which were published by the Telecommunications Industry Association (TIA). The forward link of the physical layer has four types of channels: pilot, sync, paging, and traffic.

The pilot channel does not carry any information but is used to aid the receiver in detecting, synchronizing to, and demodulating the received signal. The sync channel transmits a message periodically that carries system time information. Mobile phones use this information to synchronize to the network time. The paging channel transmits further system messages (periodical ly for all mobile phones), but also messages directed to individual mobile phones, such as call requests. The traffic channel is used to carry user information (such as speech or data) to individual mobile phones during a call. It also carries signaling messages (such as requests to change frequency) to individual mobile phones, during a call. A base station always has a pilot channel, and it typically has a sync channel, a few paging channels, and many traffic channels active at the same time.

Figure 1 shows the main functional blocks used to convert traffic channel frames into an RF signal. The other channel types are effectively subsets of this.

Click here to view Figure 1
Figure 1: Structure of the cdmaOne traffic channel.
Data frames enter the PHY from the MAC layer. A cyclic redundancy check (CRC) is appended to each frame to help the receiver decide whether the decoded data is valid or not. The resulting frame passes through a convolutional encoder. This produces two code symbols for each input data bit, discards the input data bits, thus producing a frame of code symbols.
Traffic frames can have different lengths (because the data rate can vary) and so the code symbols may have to be repeated a number of times thus making all frames the same length. This guarantees that the code symbol rate is 19.2 kSamples/s. The frame is then interleaved such that if noise corrupts some sequential code symbols, these would appear scattered throughout the frame after the receiver has de-interleaved the code symbols. This gives better immunity to burst noise.
The interleaver output symbols are scrambled to maintain privacy. Scrambling is performed by modulo-2 addition (XOR function) of the modulation symbols with a pseudo-random binary sequence with a large repetition period. One of these symbols is replaced by a power control bit every 1.25 ms. The b ase station uses these bits to instruct the mobile phone to increase or decrease its transmitted power.
Each symbol in the frame is then mapped into a +/-1 "signal" value that is then multiplied by the channel gain. Each channel can have a different gain. Gain is a very carefully controlled parameter in CDMA systems. It is set to the minimum possible value that allows the mobile phone to receive the signal with an acceptable number of incorrectly received frames. Channels to mobile phones located further away from a base station tend to have larger gains. The same is true for channels to mobile phones that experience poor channel conditions. The total power transmitted by the base station is determined by the gain of the individual channels.
These "signal" values, running at 19.2 kSamples/s, are coded by an orthogonal Walsh sequence of length 64. Walsh coding effectively replaces each signal value with a sequence of length 64. The 64 sub-divisions of a modulation symbol period are called "chips". Aft er Walsh coding the output rate is 64 times faster than the input rate and therefore the chip-rate is 1.2288 MChips/s.
The rest of the processing is multiplication by a pseudo-random number (PN) sequence to remove the periodic components of the Walsh sequences. The PN sequence consists of complex numbers at +/-1+/-j, running at the chip-rate (1.2288 MChips/s) and repeating every 26.66...ms. This operation turns the signal into quadrature phase-shift keying (QPSK) and spreads it over a 1.25 MHz bandwidth. The baseband signal is filtered and modulated onto a carrier for RF transmission.
cdmaOne Enhancements
Data rates on an IS-95-A traffic channel are limited to 9.6 kbit/s. However, an addendum, called TSB-74, was developed that moved data rates up to 14.4 kbit/s.
To reach the 14.4-kbit/s mark, 1 in 3 symbols is thrown away after convolutional encoding using a process called puncturing. Puncturing reduces the code symbol rate from 28.8 kSamples/s to 19.2 kSamples/s. This sounds bizarre but it works because the number of symbols transmitted is still larger than that of the original data bits.
IS-95-B also provided enhancements to the initial IS-95-A air interface standard. Through IS-95-B, designers could enable a the base station to transmit up to eight traffic channels simultaneously to one mobile phone. This enabled users to receive up to 115.2 kbit/s.
In IS-95-B there are two types of traffic channels: fundamental and supplemental code channels. The fundamental channel is the IS-95-A traffic channel, which carriers voice, data and signaling messages. The supplemental code channels have the same channel structure but carry data only.
Mobile phones in a call with a base station are allocated a fundamental channel first. Additional supplemental code channels are allocated on request (any number between one and seven). Needless to say that although a mobile phone is allowed to receive up to eight channels simultaneously, the total number of code channels per base station (on o ne carrier frequency) remains at 64 (as determined by the Walsh code length).
In IS-95-B, the MAC layer performs the demultiplexing of a fast stream of data bits into fundamental and supplemental code channels (each with a maximum rate of 14.4 kbit/s). The physical layer does not even know that a mobile phone is receiving more than one channel.
cdma2000 1X—The First 3G Step
The cdma2000 air interface standard was introduced through a series of standards published by 3GPP2 (as C.S0001 through to C.S0006) and TIA (as the IS-2000 series). There have been various versions of and addenda to these standards.
cdma2000 allows each base station to use either one or three 1.25 MHz RF bandwidths. This has led to the terms "1X" and "3X", respectively. 3X has not been deployed anywhere yet and will not be considered further here.
cdma2000 1X was a major update to cdmaOne, with an emphasis on even higher data rates. It includes all the channel types of cdmaOne (it is backwards compatible to it) plus another dozen or so new channel types. Of these, the important one is the supplemental channel, which is a new type of traffic channel that is different from the traditional supplemental code channel. The supplemental channel allows data rates of up to 307.2 kbit/s, and a mobile phone could be allocated two of them simultaneously.
The structure of the supplemental channel is similar to the fundamental traffic channel. Notable differences include:

The introduction of turbo codes instead of convolutional codes for higher data rates.

The demultiplexing of modulation symbols onto the I and Q channel thus halving their rate.

The use of variable-length Walsh codes.

The cdma2000 PHY offers a maximum data rate of 307.2 kbit/s for a single supplemental Channel. After turbo coding, the symbol rate becomes 614.4 kSamples/s. But after demultiplexing onto the I and Q channels, this symbol rate goes back down to 307.2 kSamples/s. With a chip-rate of 1.2288 MChi ps/s, this means that at 307.2 kSamples/s, each modulation symbol can fit only four chips. In comparison cdmaOne could fit 64 chips in each symbol.
cdma2000 allows for the use of variable-length Walsh codes. These can take any length (power of 2) from four chips to 128 chips. Walsh codes of length 4 are used for the fastest rates (modulation symbols at 307.2 kSamples/s) and Walsh codes of length 128 chips are used for the slowest rates (modulation symbols at 9.6 kSample/s). It is noted, however, that transmitting faster rates, by employing shorter Walsh codes, uses more channel capacity. For example, one can simultaneously transmit only 4 channels that use Walsh codes of length 4.
Changes Provided by 1xEV-DO
cdma2000 1xEV-DO (1X evolution—data only), published by 3GPP2 as C.S0024 and by TIA as IS-856, is a complete redesign and is not backwards compatible with existing cdma2000 1X systems. It is used for data calls only, but it allows for handoff between 1X and 1xEV-DO systems and vi ce versa. It can provide maximum data rates of up to 2.4576 Mbit/s while still using only a 1.25 MHz bandwidth. It is primarily a TDMA system, although some aspects of CDMA are used.
The 1xEV-DO forward link supports four channels: pilot, MAC, control, and traffic. These are time-multiplexed within each slot (1.66...ms) as shown in Figure 2.

Click here to view Figure 2
Figure 2: Slot format of the 1xEV-DO forward link.
The pilot channel carries no information but is used to aid in the detection, synchronization, and demodulation of the signal at the receiver end. The MAC channel uses CDMA (it employs Walsh codes of length 64) and carries control information (such as power control bits) to individual mobile phones.
The remaining parts of the slot are used for data, transmitting either the control or traffic channel. The control channel carries c ontrol information, transmitted periodically, broadcast to all mobile phones. The traffic channel carries packets of user data.
The data rate and the packet size determine the number of slots required to transmit a packet. The data rate can change on a packet-by-packet basis and is selected by the mobile phone requesting a packet. The mobile phone decides which rate to request based on channel conditions. If these are good, the requested rate is higher; if they are bad, the rate is lower.
Each traffic channel slot is directed to only one mobile phone. The base station decides which mobile phone to serve in each slot based on the data rate requests from the mobile phones. These effectively describe the channel conditions as seen by the mobile phones every slot. The base station tries to serve mobile phones when their channel conditions are temporarily better than their long-term average. If the base station chooses to serve a mobile phone, it must serve it at the requested data rate.
A distinct fe ature of the 1xEV-DO system is that a base station always transmits at its maximum power. This makes full use of the RF spectrum capacity and implies that the data rate selection is determined exclusively by the channel conditions.
In 1xEV-DO, the traffic channel and the control channel use the same channel structure. This is shown in Figure 3.

Click here to view Figure 3
Figure 3: Structure of the 1xEV-DO traffic channel.
Depending on the data rate, between one and four packets from the MAC layer are bundled together. CRC is added to the packet, which is then turbo encoded and scrambled. It is interleaved and modulated into symbols using one of three types of modulation: QPSK, 8-level PSK (8-PSK), and 16-level quadrature amplitude modulation (16-QAM). The modulation type used is requested by the mobile phone along with the request of the data rate . 16-QAM is used for the higher data rates, while QPSK is used for lower data rates.
The modulation symbols are repeated as many times as necessary to fill up the data portions in the slot(s) over which the packet is transmitted. For the higher data rates, the modulation symbols do not all fit in the data portions even once and some are discarded. This serial stream of modulation symbols is demultiplexed onto 16 channels. Each of these is Walsh coded with each one of the 16 Walsh codes of length 16. The outputs of the Walsh coders are added to produce one output stream. For every 16 input modulation symbols, 16 chips are produced. The chips are transmitted during the data portion of the slot. They are spread with a PN sequence, baseband filtered, and modulated onto an RF carrier.
1xEV-DV—Merging Data and Voice
cdma2000 1xEV-DV (1X evolution—data and voice), was introduced by the publication of the C.S0001-C to C.S0006-C series of standards by 3GPP2. This enhancement to the cdma2 000 standards is known as "Release C". It is backwards compatible to 1X and has two extra channels on the forward link that are used for the transfer of packet data: the packet data channel (PDCH) and the packet data control channel (PDCCH). The PDCH can transmit data rates up to 3.0912 Mbit/s, while the fundamental channel can simultaneously carry voice.
The operation of the PDCH is similar, but not identical, to the traffic channel of 1xEV-DO. The base station serves either one or two mobile phones in any one slot (1.25 ms). The mobile phones to be served are determined by the base station, using information on channel conditions that the mobile phones report every slot. The base station also determines the data rate at which a mobile phone is served.
The channel structure of the PDCH is shown in Figure 4. Input packets (from the MAC layer) are appended with a CRC, and are then turbo encoded, interleaved, and scrambled. The resulting packet is stored so that parts of it, called subpackets, can later be transmitted non-contiguously.

Click here to view Figure 4
Figure 4: Structure of the 1xEV-DV packet data channel.
After each subpacket is transmitted, the mobile phone acknowledges whether it has been able to fully decode the entire packet. If not, the base station sends another subpacket from the same packet. If the extra information is still not enough to decode the packet, the mobile phone asks for another subpacket, and so on, until either the packet is decoded or certain timeouts are reached.
While waiting for the acknowledgement to arrive, the base station can transmit subpackets of other packets either to the same or to other mobile phones. This way the full channel capacity can be utilized. Subpackets are transmitted over one, two, or four consecutive slots.
The subpacket symbols are modulated into QPSK, 8-PSK, or 16-QAM modulati on symbols. These are then demultiplexed onto channels that use different Walsh codes.
The number of channels varies from 1 to 28 and the length of the Walsh codes is 32. The Walsh codes used in the PDCH are selected such that they do not conflict with those of other active channels, such as pilot, sync, and paging.
The modulation type, the number of Walsh channels, and the number of slots over which a subpacket is transmitted effectively determine the transmitted data rate. For each subpacket, the base station selects a combination based on the channel conditions, the size of the packet, and the number of available Walsh codes. The remainder of the channel processing is channel gain, PN spreading, baseband filtering and RF modulation.
The base station uses the PDCCH to transmit the PDCH configuration (packet size, number of Walsh codes, target mobile phone, etc.). The PDCCH is broadcast simultaneously with the PDCH and carries the configuration for each subpacket. It has a fairly simple struct ure and is quite similar to the other 1X channels.
Wrap Up
There is no doubt that the CDMA family of standards will be important in the development of data-enabled cellular systems. But, the challenge will be effectively implementing these standards in system designs. Clearly, the forward link architecture will be one area where some design challenges occur. However, through the article above, designers can get a grasp of the key design elements needed to build a successful forward link architecture for cdma2000 1X, cdma2000 1xEV-DO, and cdma2000 1xEV-DV systems.
About the Author
Marios C. Agathangelou is a systems engineer at Racal Instruments Wireless Solutions, which is an Aeroflex Company. He received a BEng in Information Systems Engineering from Imperial College, London, and an MSc in Electrical Engineering from Stanford University.