System integration of the Flex paging protocol Part 1–System design constraints.
Trademarks: Flex–Motorola. Short for flexible, wide-area, synchronous paging protocol. ProEncore–Motorola. SignalPro–Advanced Signal.
Integration of the Flex paging protocol onto existing paging channels will present system design challenges for systems not yet optimized for the higher data rate of 6,400 bits-per-second (bps), but which are slated for operation at that rate. The most significant constraint facing designers is control of coverage overlap areas. However, until such time that the system may be optimized for 6,400bps, carriers will enjoy several benefits in paging with 3200 4-level Flex over 2400 POCSAG. In many cases, there is good reason to introduce the Flex paging protocol, at the operator’s earliest convenience, onto paging channels currently performing, even marginally, at 2400 POCSAG.
The Flex paging protocol holds the promise of greater error immunity and higher data rates than any other paging encoding protocol in use to date. Greater error immunity is accomplished by using Bose-Chadhuri-Hocquenghem (BCH) coding along with an interleaving process that helps to minimize burst-type errors that often are prevalent in a radio frequency (RF) environment. Increased data rates are accomplished by the modulation technique of 4-level frequency-shift keying (FSK). However, these potential increases come at a price. Greater error immunity requires greater overhead in terms of pre- and post-transmission processing, and increased data rates significantly challenge simulcast system designs in terms of overlap area control. The following information points out prominent design constraints facing Flex integration and offers suggestions for optimizing system performance.
Simulcast delay spread In most paging systems, the biggest challenge facing acceptable 6400 Flex performance is management of simulcast delay spread (SDS) to acceptable levels. (In a multitransmitter simulcast system, SDS error always occurs. The challenge is to optimize the system’s design parameters to keep error within a specified margin.) SDS is often referred to as simulcast error or simulcast misalignment; however, the term delay spread will be used instead because it is more familiar to those with cellular experience, and it is a more descriptive term of the phenomenon.
SDS, for our purposes, is a received misalignment of state transitions caused by signal timing differences in a multitransmitter simulcast system. Stated more simply, the RF signals from two or more transmitters reach the receiver at different times and with relatively equal energy, causing distortion at the transition edges. (The effect of SDS is a function of both the time and energy level differences of the signals available to the receiver. Available signals within a 6dB differential generally are assumed to be close enough in magnitude to have a significant effect on the receiver. When two or more signals are present with levels close enough to affect the receiver, the situation often is referred to as non-capture.) The two causes of this type of delay spread are:
1. The information signal actually leaves individual transmitters at different times. (Implicit in this statement is the concept of bit jitter, which is a consequence of processor speed and part tolerances. Using current technology and good design practices, bit jitter may be held to insignificant amounts.)
2. The propagation paths from the various transmitters to the receiver, differ in length.
The first cause of SDS may be compensated for by proper configuration techniques such as introduced delay for equalization or by synchronization to the National Institute of Standards and Technology’s Universal Coordinated Time (UTC) via the Global Positioning System (GPS). However, the second cause of SDS, a difference in propagation path distances, ultimately is a parameter that must be dealt with in terms of system design and propagation analysis.
SDS in the time domain SDS is noticeable in the time domain by demodulating the RF signal and displaying the waveform with an oscilloscope. Figure 1 on page 14 shows a digital storage oscilloscope (DSO) capture of a demodulated FSK waveform showing SDS. Note that the FSK transition edge is distorted by approximately 150 to 180 microseconds of timing error.
The general rule of thumb for allowable SDS is one-fourth of the symbol width. On that basis, the acceptable error for POCSAG 2400, 3200 2-level Flex, 3200 4-level Flex and 6400 4-Level Flex are shown in Table 1 on page 10.
Note that going from a 2-level FSK to a 4-level FSK modulation scheme, for a given data rate, doubles the symbol width of the transmitted signal and, consequently, the allowable SDS. The symbol width for 4-level modulation is doubled because two data bits are encoded into each symbol instead of one bit encoded per symbol in 2-Level modulation.
Figure 2 on page 14 shows a demodulated DSO capture of 3200 4-level Flex, POCSAG 2400 and 6400 4-level Flex. Note the difference in symbol width between the various formats. In particular, the symbol width of 3200 4-level Flex is 30% greater than that of POCSAG 2400 and, therefore, more tolerant of SDS.
SDS in the spatial domain It is often preferable to convert SDS to the spatial domain when analyzing system performance. This is easily accomplished by dividing by the velocity of light. (Electromagnetic waves travel through free space at the speed of light, about 5.4 microseconds per mile.) The allowable SDS amounts (i.e., one-fourth of the symbol width), converted to a distance, for POCSAG 2400, 3200 2-level Flex, 3200 4-level Flex and 6400 4-level Flex have been calculated below:
POCSAG 2400: 104microseconds * (1mi/5.4microseconds) (approx=) 19.3mi
3200 4-L Flex: 156microseconds * (1mi/5.4microseconds) (approx=)28.9mi
3200 2-L and 6400 Flex: 78microseconds * (1mi/5.4microseconds) (approx=) 14.4mi
(Both 3200 2-level and 6400 Flex have the same symbol width and, therefore, the same SDS constraints.) In terms of distance, 3200 4-level Flex is more tolerant of SDS by a 9.6-mile margin over POCSAG 2400, and POCSAG 2400 is more tolerant of SDS by a 4.9-mile margin over either 6400 Flex or 3200 2-level Flex.
A simplistic two-dimensional, two-transmitter overlap zone is shown in Figure 3 on page 16. Let us presume that the convergence region between two transmitters, located at TX-A and TX-B is a non-capture area of maximum width (delta), according to the one-fourth of the symbol width constraint, is about 28.9 miles for 3200 4-level Flex, 19.3 miles for POCSAG 2400 and 14.4 miles for 6400 Flex.
This tightening of the SDS constraint for 6400 Flex may not seem too radical at this point; however, the example given is far too simplistic to adequately characterize the situation encountered in an actual paging system. First of all, real-world antennas are not isotropic, and their propagation characteristics in a simulcast system are highly dependent upon type, gain, relative elevation and orientation. (Isotropic propagation presumes a spherical radiation pattern, equal in magnitude for any propagation path of the same radius from the antenna element.) In addition, the phenomena of reflection, diffraction and scattering need to be taken into account. The topographical environment, both natural and manmade, has a significant effect upon the received signal level.
Figure 4 on page 16 is a more realistic view of a two-dimensional, two-transmitter overlap region. Note that localized pockets of non-capture may occur at random throughout the coverage area because of any number of the reasons previously mentioned. Of particular importance, Figure 4 dismisses any myths that Figure 3 is representative of an actual simulcast paging environment. In Figure 4, X represents a pager location that happens to be in one of a number of non-capture overlap areas created by transmitters TX-A and TX-B, and (delta) represents the SDS between the two transmitters. To calculate the propagation path difference, (delta), a circle of radius r, centered at X and passing through the nearest transmitter location, TX-B, is drawn. On a straight line from the receiver location, X, to TX-A, the (delta) of concern is the line segment from the point of intersection with the circle to the location of TX-A.
Figure 4, in addition, points out good reason for placing transmitter sites within the SDS constraint of adjacent transmission sites (e.g., 14.4 miles for 6400 Flex). That is, if TX-A and TX-B are not separated by more than the spatial SDS constraint, then there will be no receiver non-capture location at which the SDS constraint will be exceeded by those two transmitters. However, in simulcast system design, care must be exercised to ensure that signals from sites farther away than the adjacent sites do not propagate into the coverage area with sufficient signal strength to cause SDS problems. A situation which is often encountered in the field occurs when the receiver is RF-shaded from a nearby transmitter site, by a structure, the local terrain or both, reducing the received signal strength enough to allow SDS to be introduced by a more distant transmitter. Furthermore, the receiver is likely to be in an overlap region of more than two transmitters. When disruptive SDS is present, it is often found that there are several distant sites contributing to the SDS in a cumulative manner.
The effects of SDS are not limited to areas of low received signal strength. In fact, this condition is often present in areas with more than adequate signal level (e.g., -70 dBm). In general, the problem is one of too much signal level present from distant sites, and the solution resides in compressing and controlling the propagation pattern to avoid exceeding the SDS constraint. This does not mean that reducing transmitter power is the answer. In fact, 6400 Flex requires an additional signal-to-noise margin of about 3dB over 3200 4-level Flex to maintain a desired bit error rate. The key to overcoming SDS problems, by compressing and controlling signal levels, is in the selection of antenna types and gains, transmitting antenna elevations, transmitter output power levels and transmitter site placement. In short, adequate 6400 Flex performance requires proper system propagation analysis and design. Unfortunately, most current paging systems have evolved from a slower data rate design without proper redesign for the increased data rates. In many cases, the increased signal-to-noise requirements were met by the method of replacing existing antennas with ones of higher gain. (For each doubling of the data rate, the link margin requirement increases by 3dB to maintain a given bit error rate.)
Referring again to Figure 3, the propagation pattern is shown in the horizontal plane only. The problem encountered with higher gain antennas is that the gain increase is accomplished by a narrowing of the vertical beamwidth pattern. This narrowing of the beamwidth directs the signal strength toward the horizon at the cost of reducing it near the transmitter sites. This is precisely the type of situation which promotes SDS problems.
Because SDS is a function of multiple parameters, there is no single design remedy that covers all situations. The simplest design case is that of flat terrain in a rural environment, and the complexity of the design solution increases greatly as the terrain conditions change and as the environment becomes primarily urban. However, there are several design guidelines outlined in part 2 of this article that are pertinent, in general, to providing 6400 Flex. Also outlined in part 2 are some straight-forward diagnostic tools for determining whether or not SDS is problematic in a given area and, in instances, for determining the distant, offending sites.
Zero-beating and deviation offsets Zero-beating, which is actually a phase cancellation condition, is presumed to occur in an RF simulcast environment periodic at the rate of 1/f(sub o), where 1/f(sub o) is the total frequency difference between canceling signals. The cancellations occur with less frequency but for longer duration as f(sub o) becomes smaller. Deviation offsets are often used in an attempt to alleviate, or more precisely, to manage the phenomenon of phase cancellation. By programming a different deviation offset into the transmitters, a cancellation occurrence can be controlled and the negative effects minimized. (A deviation offset increases or decreases the actual deviation level by the offset amount. For example, a +10Hz offset programmed into an exciter set to deviate at ±4,500Hz would result in actual deviation levels of -4,490Hz and +4,510Hz.) Difficult to observe in the field, zero-beating is most likely to occur in weak overlap regions of only two transmitters because the probability of phase cancellation is greatly reduced as more signals are added and/or as the carrier-to-interference margin increases. In addition, because of the increased accuracy and stability of later-generation exciters, zero-beating has become an issue of greater concern than it was with older, less-accurate equipment.
According to laboratory results in a stationary environment at low signal levels, POCSAG performance shows the most improvement with relatively large offsets (i.e., 200Hz to 1000Hz), whereas 6400 Flex shows optimum improvement with smaller offsets (i.e., less than 200Hz). Initial laboratory testing of 6400 Flex offsets points to multiples of 30Hz as producing the best results (e.g., 30Hz, 60Hz, 90Hz and 120Hz) with multiples and sub-multiples of 200Hz producing less desirable results. The reason for the reduced performance relates to the data rate and block structure of 6400 Flex. Using a 32-bit-by-32-bit block structure along with an interleaving process results in a bit of information associated with a particular pager being transmitted every 5 milliseconds at a rate of 200Hz. Table 2 at the left shows the relationship among deviation offset, cancellation occurrence and the frequency of a possible bit collision with a phase cancellation in a particular pager’s information word.
The implication is that certain offsets exist that produce degradation because of their timing alignment with the block structure, and the results of a two-transmitter overlap simulation show a correspondence to that notion.
Figure 5 and Figure 6 on page 22 are graphs of data taken with a two-transmitter overlap in a static, low-signal-strength environment with various offsets introduced. The data taken in this experiment are presented only with the intent to show trends related to various offset schemes and should not be considered absolute. (Jack Gleeson, engineering manager at Motorola Subscriber Development Engineering, has been involved with extensive studies of deviation offset schemes and their effects upon paging probability in various simulated RF environments. The data presented here show correlation to the static environment findings at Motorola.) Actual performance is a function of many variables, including signal strength, fade distribution and the receiver’s demodulation technique. The results presented in Figures 5 and 6 are based upon tests run with 50 pages sent to a ProEncore pager on each of the four Flex phases (i.e., phases A, B, C and D). Test pages were encoded with a Glenayre GL3000-ES paging terminal and transmitted via two Glenayre GL-C2000 transmitter controllers and two Glenayre Gold Series exciters. Decoding was accomplished with an Advanced Signal SignalPro data and signal analysis tool. Transmitted pages were 15-character numeric, and the graphed results are a percentage of received perfect pages.
Figure 5 shows minimum paging reception at the 0Hz deviation offset setting, with slightly reduced performance at the 40Hz and 50Hz settings. Phase A and phase C performance appeared to be comparable to phase B and phase D with small deviation offsets.
Paging reception minimums are seen at 200Hz and 300Hz in Figure 6. Also, it is apparent from Figure 6 that phase-B and phase-D paging probability drops more rapidly than that of phase A and phase C as deviation offsets are increased. Therefore, large POCSAG-type offsets should be avoided for 4-level Flex.
Presuming that the pager uses a discriminator demodulation technique, a possible reason for this condition may be explained by considering the probability density functions (PDF) of the symbol FSK states (or levels) and the way the pager information is encoded into each state. Figure 7 above shows a PDF for each of the four FSK levels at the discriminator output.
For the purpose of example, a normal distribution has been assumed for each PDF. Decision lines are shown at the point of intersection between adjacent PDFs which simply identify that there are three regions of overlap for which a decision may need to be made as to which FSK symbol has been received.
Because of the interleaving process and block structure of 6400 Flex, phase-A and phase-B bits are grouped together, and phase-C and phase-D bits are grouped together in two bit pairs. Phase-A and phase-C bits are always the most significant bits and phase-B and phase-D bits are always the least significant bits of the two-bit pairs.
Referring to Figure 7, note that of the three regions of overlap where the decision lines are shown, the regions between 00 and 01 and between 11 and 10 (denoted with red dashed decision lines) are dependent only upon the least significant bit. That is, for example, a 00 symbol could be incorrectly decoded as a 01 symbol, yet, since the most significant big is the same for either symbol, only the least significant bit would be in error. The region between 01 and 11 (denoted with a blue dashed decision line) is dependent only upon the most significant bit. Thus, it would appear that the phase-B and phase-D pagers have a greater likelihood of bit errors than the phase-A and phase-C pagers because there are two decision areas which affect the least significant bit and only one decision area that affects the most significant bit. Thus, as the deviation offsets get larger and the actual deviation level shifts closer to the decision line region, the performance of the phase-B and phase-D pagers fall off more rapidly.
Deviation: POCSAG vs. Flex POCSAG pagers are optimized for 4,500Hz deviation, and Flex pagers are optimized for 4,800Hz. Thus, the question arises as to which deviation setting should be used on a system that is incapable of adjusting deviation automatically on a protocol basis. Deviating at a different value than prescribed changes the modulation index of the signal, which in turn affects the bit energy-to-noise spectral density (E(sub b)/N(sub o)) requirements of the receiver. To know the true effects of a certain deviation setting on a particular pager, one must know the demodulation technique used for that pager (e.g.; matched filter or discriminator.). According to both NEC America and Motorola, Flex pagers should show no significant sensitivity loss with a signal deviation of 4,500Hz reduced from the optimum 4,800Hz. Nevertheless, the ability to change deviation on a protocol basis allows the system to be optimized for both POCSAG and Flex and should be an infrastructure provider’s design goal for the near term.
Part 2, which concludes the series, covers system design recommendations.
Back issues no older than two years are available from customer service. Call customer service at 800-441-0294.
Article photocopies are unavailable from the publisher, but they can be ordered from UMI Information Store, 800-248-0360; 415-433-5500 ext. 282; Fax 415-433-0100.