3.2 THE FCC PART 15 RULES
3.2.3 Direct Sequence versus Frequency Hopping
A comparison of direct-sequence and frequency hopping spread-spectrum radio techniques for commercial applications must take into account many issues, some addressed here.
3.2.3.1 Conversion of Narrowband Radios
Conceptually, frequency hopping is easier to understand than direct- sequence spreading. In addition, most radio manufacturers can easily con- vert their narrowband radios to spread-spectrum radios in which the center frequency hops around according to a fixed pattern, or code. However, hop- ping around the transmitting radio carrier frequency is not enough: Now, the receiving radio must be synchronized to the hopping pattern of the trans- mitting radio. Synchronization is one of the most problematical operations in frequency-hopping radios.
3.2.3.2 Cost of Development and Products
The cost of development and the cost of the final product are important con- siderations in selecting a spread-spectrum technique. For most commercial applications, a fast frequency-hopping radio is not cost-effective. In fact, hop- ping rates of commercial radios are generally very slow compared to the data rates, typically having several hundreds of bits or more transmitted during a hop duration. Similarly, direct-sequence implementations for commercial applications with large processing gain require costly high-speed circuits which are generally impractical. Thus, the processing gain for commercial direct-sequence radios is usually limited to less than 20 dB to avoid having to use high-speed circuits.
With lower processing gain (fewer chips per bit), direct-sequence radios for commercial applications can use digital matched fibers instead of ser- ial correlators. Matched filters have the advantage of being able to achieve faster acquisition times than serial correlators or frequency-hopping radios.
In general, however, direct-sequence radios with matched filters demand a greater development effort than other types of spread-spectrum radios. The final cost of direct-sequence spread-spectrum radios with digital matched filters, however, is no greater than that for serial correlator type direct- sequence or frequency-hopping radios. In fact, as shown in the currently marketed cordless telephones, this type of direct-sequence spread-spectrum radio with digital matched filters can yield low-cost consumer electronics products.
3.2.3.3 Performance
Without interference from other radios and in free space, both direct- sequence and frequency-hopping spread-spectrum radios would give the same performance—in theory. In practice, however, the performance depends on the design of the radio and propagation conditions, and the real- ized performance varies greatly.
With large multipath delays (delays greater than a chip time), direct- sequence spread-spectrum radios can be more robust since they can better overcome the effects of multipath. Commercial slow frequency-hopping
radios behave as narrowband radios. Thus, for higher data rates, the impact of multipath tends to degrade frequency-hopping radios more than direct- sequence radios.
The issue of mutual interference resulting from many spread-spectrum radios independently operating in a given area is very interesting. In a sys- tem in which many such spread-spectrum radios co-exist in the same area, it is difficult to compare the performances of these two types of radios, which depend largely on the application.
With frequency-hopping radio, for each hop, radio performance is “good”
or “bad,” depending on whether any other similar radio signal hops into the same band at the same time. The probability of this occurring depends on the number of hopping radios in the area and the availability of non-over- lapping hopping bands.
Multipath conditions are likely to be different in each hopped frequency, even without interference, so performance during each hop can vary greatly, owing to different multipath conditions. Thus, frequency-hopping radios result in a channel that is time-varying, with changes in channel conditions occurring at each hop transition, somewhat like a fading channel in which fades occur at discrete times. As in any fading channel, coding with inter- leaving can dramatically improve performance (see Part 1, Chapter 3, Sections 3.7 and 3.8). Unfortunately, it is impractical to interleave data when the hop rate is slow, as is true for most commercial frequency-hopping sys- tems.
Thus, the major difference between the two types of spread-spectrum radio is that frequency-hopping radios experience occasional strong bursty errors, while direct-sequence radios experience continuous but lower-level random errors. In direct-sequence radios, errors can be considered to be scat- tered randomly over time; in frequency hopping, errors are distributed in clusters.
For packetized data systems such as wireless LAN applications, fre- quency-hopping radios have an advantage. During a hop, frequency-hop- ping radios may transmit one or more packets. The higher-layer LAN protocol typically checks for errors in a packet, and if they exist, requests retransmission of the packet. In this type of system, the number of errors in a packet is not important—only the fact that an error has occurred is important. Thus, if the average number of packets with errors is the main performance criterion, it is preferable to have errors arrive in clusters, as in frequency-hopping radios.
The receiving slow frequency-hopping radio, however, generally takes longer to synchronize to the transmitting radio than corresponding direct- sequence radios with matched filters. For packetized data, this can be a major problem owing to the time needed for resynchronization for each packet or group of packets.
For all practical purposes, the FCC rules limit frequency-hopping radios to a maximum data rate of 1 Mbps since their instantaneous bandwidth is a
maximum of 1 MHz. Direct-sequence radios can have higher data rates when sufficient bandwidth is available. For the 5.7-GHz band, several manufac- turers offer commercial direct-sequence spread-spectrum radios capable of a wireless LAN burst rate of at least 5 Mbps and full-duplex operation of 2 Mbps.
Voice applications are more tolerant of errors but require full-duplex operation. In voice applications, the acquisition time is not as critical, but unlike applications with packetized data, strong bursts of errors are more bothersome.
In addition, owing to lower data rates (typically 32 kbps or less) for voice than for most data applications, each direct-sequence radio channel requires less bandwidth. For cordless telephones, e.g., each voice channel typically requires 2 MHz of bandwidth. For the 902—928 MHz ISM band, this means that 13 non-overlapping direct-sequence voice channels are available, with cordless telephones “smart enough” to select the clearest channel among them. This somewhat resembles “intelligent” frequency hopping in that the radio selects the clearest channel, hops to it, and stays there. With frequency hopping, which indiscriminately hops over the total band, the voice channel may be hit with occasional high bursts or errors owing to multipath and inter- ference in different hop bands.
Digital video conferencing has the most demanding requirements in both types of radio, requiring higher data rates and full-duplex operation.
Probably the most important functionality to be compared between direct-sequence and frequency-hopping spread-spectrum radios is the poten- tial overall capacity achievable in areas with many such radios. In Section 3.5, we address this question for a high-density network of many synchro- nized cells.