CA2003954C - Method and circuitry for symbol timing and frequency offset estimation in time division multiple access radio systems - Google Patents

Abstract

ABSTRACT OF THE DISCLOSURE
In order to correctly demodulate a received sequential burst of symbols in a time division multiplexed/time division multiple access (TDM/TDMA) portable radio digital telephony communications system, proper timing of the sampling time in each received symbol of the burst is necessary. In addition, in order to compensate for component drift, an estimate of the frequency offset between transmitting and receiving units is also required. A method and circuitry for estimating symbol timing and frequency offset is disclosed in which the IF radio signal is sampled and digitized at a sampling rate which is sixteen times the symbol rate. The digitized samples are processed to obtain phase values. A one symbol delay is introduced and differential phase values derived, a differential phase valve being derived for each of the sixteen sampling times per symbol.
The differential values are collapsed into one quadrant in the phase plane and then expanded back to the full plane. For each of the sixteen sampling times, a separate vector sum is formed of the expanded and collapsed differential phase values over substantially the entire burst. Symbol timing is selected to be the particular one-of-the-sixteen sampling times at which the vector sum has the largest magnitude. Frequency offset is directly determined from the angle in the phase plane of that vector having the largest magnitude.

Description

200:~9S~

~QB~JN12 nF T~IF Il`~VE-lTIO~
The invention relates to the processing of received digitally modulated radio signal transmissions iD a time division multiplexed/time division multiple access (TDM/TDMA) system, such as that which will be used in conjunction with low powerS portable digital telephony, and morc particularly to determiniDg for each burst of received symbols, the optimum symbol time and frequency offset estimation for coherently demodulating the burst Pcople by their vcry nature are highly mobile; no where is this more true thaD in modern day society with its myriad forms of travel At the same time, many 10 people incrcasingly have a need to be able to telephonically communicate with others particularly while they are on rthe go, i e while they are moving However, this need for mobile communications, which e~cisted for quite some time, has remained basically unsatisfied Since telephones traditionally have cords, any mo~ement of the telephone was traditionally limited by the length of its cord 15 For many years, ODIy a veritable handful of telephones actually traveled with their users g These mobile telephones induded aeronautical, marine and other forms of early radio telephones Inasmuch as these mobile telephones were priced well beyond the affordability of the average telephone subscriber, none of these radio telephones ever encountered widespread use Accordingly, for the vast majority of subscribers, a telephone set was 20 installed at each subscriber location and there it remained unless it was reinstalled elsewhere Thus, these subscribers either remained close to their telephone and thus restricted their mobility particularly in the aDticipatioD of receiviDg a telephone call, or intentionally sought out a public or private telephone located along their route of travel whene~er the need arose to place a telephone call Now with iDcreasiDg sophistication of miniaturized electronic techDology and decreasiDg atteDdaDt cost thereof, various vendors provide a number of de~icos (and/or senices) that offer tetherless telephoDy. These devices, e~tplained in more detail below, attempt to free a subscriber from being bound by the ambulatory COnStraiDtS
imposed by e~i-tiDg wireliDe telephoDe sets. In effect, each of these de~icos now permits 30 subscrlben effecti~ely, at least with a cortaiD e~ttent, to talce their blepbone with them, obtaiD o~change ccess, and remain in communication where~er they go These devices include cordless telephones, collular mobile rsdio traDsceivers, public paclcet radio data networl~ traD-coi~rer- aDd radio pagers As a grOWiDg number of consumers perceived the freedom of movemeDt offered by these de~rices, a large demand was created for these .
~ .''.' ;: ' V - ` .~
;~ ' ~
,:

20039~

devices. Moreover and not une~cpectedly, as the prices of these devices coDtiDue to fall due to manufacturiDg economies and technical developments, the demand for these devices correspondingly continues to substantially increase. Specifically, appro~cimately 25 million cordless telephone sets are in use today throughout the UDited States with dcmand for these 5 sets coDtiDuing to rise as the price of cordless telephones with increasing sophisticated has remaiDed within a S100.00 to S200.00 range. ID addition, appro~imately three mi11ion cellular telephone sets are currently in use throughout the United States. As tbe price of various cellular sets falls from more than a S1000.00 which occurred merely a year ago to only a few hundred dollars today, the demand for these sets has increased precipitously. As 10 a result, the number of installed sets has climbed at such an astonishing rate that in certain urban areas, such as New Yorl~, the number of sets in use at pealc times is beginniDg to straiD the capaciq of the e~isting cellular networlc to handle the concomitant call traffic.
While, each of the present tetherless telephonic technologies possesses certain advaDtages, each technology also unfortunately has certain drawbacl~s that 15 significantly restrict its use. In this regard, see, e.g., Co~, "Portable Digital Radio Communications -- An Approach to Tetherless Access", IEEE CommunicatioDs Magazine, Vol. 27. No. 7, July 1989 pages 30-40; and Co~, "Universal Digital Portable Radio Communications", Proceedings of the IEEE, Vol. 75, No. 4, April 1987, pages 436- 4î6.
Specifically, as to cordless telephones, such a telephone consists of two 20 transceivers: base unit and a handset, that collectively form a lovv power duple~ aDalog radio linlc. The base unit is eonnected, qpically by a subscriber to a wireliDe access point in a conventioDal telephoDe networlt in lieu of or as a replacement for a wireline telephone, in order to implement a tetherless substitute for a telephone cord. Once connected, the base UDit appears to the telephone Detworl~ as a coDveDtional telephone. The base unit eoDtaiDs a 25 transmitter and recei~er, and simple eontrol and iDterface apparatus for dialing, acceptiDg riDgiDg, termiDating eails aDd eoupling voiee from the telephone line to the transmitter and from the reeeiver within the base unit to the telephone line. The handset, whieh is truly portable, eoDtains simple coDtrol logle for iDitiating, recei~ring and termiDatiDg ealls with the bue UDit Dd for hrnlng its own traDsmitter on and off. To provide true duple~ operation, 30 separ-te earrier frequeneies are used by the traDsmitters in the base UDit and handset. Since eordless telephoDes operate with very low input power to their transmitter, usually on the order of oDly several milliwans, the handset geDerally utilizes several small rechargeable batterie~ as ib power source. This eDables the haDdset to be made relatively small, lightlveight nd to be eontiDuously used for a relati~dy long period, qpically several hours, 35 before its baneries require rechargiDg. Furthermore, the ~ery low level of power radiated from the haDdset poses essentiaily no biologieal radiation hazard to its user.

Z00~95(~
~ .

Unfortunately, the primary disadvantage of cordless telephones is the;r highly limited service area Because curdless telephones use relatively low transmitter power, these telephones have a ma imum range that varies from typically a few hundred to a thousand feet, which in turn rcsults in a very small service arca A secondary disadvantage S associated with cordless telephones stems from the limited number of available frequencies At present, only a few separate frequencies, typically up to 10 duple~ channels, ha re been allocated by the Federal Communications Commission (FCC) for use by cordless telephones Moreover, early cordless telephones by their very design have been very susceptible to co-channel interference This interference arises by the simultaneous operation of two or more - --10 cordless telephones situated in close pro~imity to each other, such as in an irnmediate neighborhood of a residential area In a very small geographic area with a very low density -of users, 8 reasonable probability e~cists that within this area one or more duple~c pairs will ~ -not be in use at any one time, and, as such, this interference will not occur therein Nevertheless, in an effort to avoid this interference, relatively sophisticated cordless 15 telephones are DOW capable of operating on any one of a number of preprogrammed duple~
pairs with either the user or the telephone itself selecting, manually in the case of the user and automatically by the telephone, the specific pair that is to be used at any one time Unfortunately, if a wfficient number of cordless telephones are in use in a very densely populated area, such as an apartment building, pair selection may not be sufficient to 20 eliminate the espected incidences of co-channcl interference that results from undisciplined `
and uncoordinated duple~ pair assignment and the resulting chaos e~perienced by users situated therein In addition, since cordle~s telephones rely on analog modulation of a duple~ pair, conversations occurring o~rer a cordless telephone are highly ~ulnerable to ea~esdropping Furthermore, a cordless telephone only provides limited protection against 25 unauthorized long distance or message units calls beinB made therethrough While preprogrammed digital or tono accoss codes are being used between indi~idual handset base unit pairs and provide sufficient protection against casual attempts at unauthorized access, these codes aro oot sufflciently sophisticated to succossfully deter a determined orderly Ysault on a cordlas telephone by an unauthorized user Furthermore, while cordles6 30 telepbones pro~ide limited portable radio accoss to a wireline access point, from a network standpoint cordless telephones do not eliminate the need for telephone lines, i e a customer drop, to be run to each subscriber Nonetheless, in spite of these se~ere ser~ice restrictions, cordless telephones aro immonsoly popular for tho froodom, though vory limited, that they furnish to 35 their users.
..-":

,.': - .-.: -20039~

In contrast to thc very limited range provided by cordless telephones, cellular mobile radio systems accommodate wide ranging vehicular subscribers that move at relatively high speeds. These systems utilize a relatively high power ~50 MHz transmitter, typically operating at an input of spproximately .5 watt to several tens of watts, in a mobile 5 unit with a relatively high efficiency antenna to access a wireline telephone network through a fi~ed cell-site (base station). The base station also uses a high power transmitter in conjunction with a tall antenna, typically erected on a tower or tall building, to provide a relatively large coverage area. Due to the e~cpense, typically ranging to $300,000 e~clusive of land and building costs, and the antenna size associated with each base station, the least 10 number of base stations are often used to cover a given area. Nonetheless, this arrangement generally provides a circular service area centered on a base station with a radius of approsimately 5-10 miles therefrom. In use, a cellular radio system that covers a large region often encompassing a city, its suburbs and major access highways typically includes a number of geographically dispersed base stations. The base stations, containing radio 15 receivers and transmitters and interface and control electronics, are connected by trunks to, and eoordinated aDd eontrolled by ODe or more Mobile Telephone Switching Offices(MTSO~) that, in turn, also provide access to the conventional wireliDe telephone network.
All of the dupler radio channels available to the eDtire system are sub-divided into sets of channels. The radio equipment in each base station has the capability of UsiDg chaDnels from 20 one of the ehannel sets. These sets are allocated to the base station in a pattern that ma~imizes the distanee between base statioDs that use the same sets so as to minimize average eo-ehannel interference occurring throughout a service region. One or more ehannels are designated for initial eoordination with the mobile sets during eall setup.
Each mobile (or hand-held) cellular transceiver used iD the system 25 eontains a reeeiver and a transmitter eapable of operating on any duple~ radio ehannel available to the cellular system. Calls can be made to or from any mobile set anywhere withiD the large regioD eovered by a gtoup of base StatiODS. The eontrol electronies in the moblle ansceiver coordinates with a base station on a speeial eall setup ehannel, Identifies itself, aDd thereafter tunes to a ebannel designated by the base station for use during a 30 partieular call. Eaeb duple~ ehannel uses one frequeney for transmission from base~to-mobile and a different frequoney for transmission from mobile-to-base. The signal strength of eails in progress is monitored by the base stations that ean serve those calls. Specifieally, ~vhen the signal strength for a giveD eall drops below a predetermined threshold, typieally due to movement of the eellular subscriber from one eell to another, the MTS(~ eonnected to 35 that base statioD coordinates additional signal strength measurements from other base statioD8 whieh surround the station that is currently handling the call. The MTSO then attempts to switeh ("handoff") the eall to anotber duple~ channel if one of the other base ~ ~

. ;-.

Z0039~;~

stations is receiving a stronger signal than that being received at the base station that is curreDtly handling the call This handoff of calls, totally transparcnt to the cellular subscriber, preserves the quality of the radio circuit as the subscriber moves throughout the servicc region Moreover, calls are handed off from o~e MTSO to another, as the subscriber S transits from one service area into another Inasmuch as frequency usage is coordinated, relatively efficient use is made of the available frequency spectrum while mlnimiziDg the lil~elihood co-channel interference In each different geographic service area within the lJnited States, there are two competiDg cellular systems using different frequencies Though cellular mobile radio systems provide wide range, these 10 systems suffer various drawbac~s First, cellular systems were originally designed for use in motor vehicles whose electrical systems could rcadily provide sufficient power While portable hand-held cellular transceivers do e~ist, they must operate with sufficieDt transmitter input power, typically at least .5 watt, to reliably reach a base station This, in tUrD, requires that a relatively large battery must be used within the portable cellular 15 transceiver However, due to the limits of preseot rechargeable battery technology, the amouDt of time that the portable traDsceiver can be used before it requires recharging is often quite limited Furthermore, the cost of these rechargeable batteries and hewe of the portable transceiver is rather high Moreover, high radiated power levels, such as that which emanate from a mobile or portable cellular aDsceiver, may be sufficient to pose a potential 20 biological r-diatioD hazard to its user Furthcrmore, since cellular systems were not designed to compeDsate for radio attenuation occurring within buildings, these systems are oDly able to provide little, If any~ service within a building Low power portable cellular transcelvers are not operationally compatible with large cell sizes, designed to match the needs of fast moviog vehicular users, and thus often provide poor communication in many 25 areu within these cells. ID additioD, siDcc cellular systems rely on merely frequency modulatiDg a carrier with voice or data, these systems are Iso susceptible to eavesdropping La~tly, from a Detworl~ perspective, cellular systems are quite inefficient Due to the IDCIUSjOD of MTSOs ~vith unl~s conDected to iodividual base statioDs, baclcbaul of cellular trafflc, over ~vlied trunb, ofteD occurs over several miles prior to its enance iDto the 30 ~IrireliDe networl~, thereby resultiDg iD a wasteful overbuild of Detworlt ansport facilities Public paclcet radio data Detworic- preseDtly e~ist to handle iDfrequeDt bursh of digital data betweeD a fi~ed bue statio~ and a Dumber of portable data tr nsceiver~. The fi~ced site has a tr~nsmitter that uses ~everal ten- of watts; while each port Ue d-ta tr DsCeiver usès traD-mitter that operates at a level of several watts. As such, 35 rdiable coverage is provided over a service area that may e~tend several miles in radius from a base statioD IDdividual base statioDs are coDDected by a fi~ed distribution facility to controller that can, iD turn, be coDDected to either a local e~change Detworlc, to handle -'-' ~ :.
.::
. ;~ '.' : .

20039~

voice-band data, or a packet-data network which itself interconnects various computers Multiple users contend for transmission time on typically a single radio channel Data transmissions on the channel are set up in either direction througb bursts of coordinating data, handshaking, that occur betwcen a base station and a portable data transcciver 5 Appropriate controller aDd radio liDk protocols are used to avoid packet collisions Once a data transfer is complete between that base station and a data transceiver, the channel is immediately available for reuse by others Although data bursts are transmitted at selatively high power, each burst is transmitted for only a short duration As such, the average power consumpaon for a portable data transceiver is far less than that associated with a portable 10 cellular transceiver thereby allowing physically smaller interDal batteries to be used with portable data transceivers than those used iD portablc cellular transcoivcrs Nevertheless, the high radiated power levels associated with a portable data transceiver again pose a potential biological radiation hazard to ies user In addition, these networl~s disadvantageously suffer from 1imited digital transmissioD capacity which restricts these networl~s to carrying short ~ -lS data bursts and not voice, and, like cellular systems, e%perience coverage restraints when used witbin buildings In contrast to the tetherless systems discussed above, radio paging ~ystems provide simple unidirectional traDsmission from a fi%ed location to a specifically addressed portable pager, which when received provides an alerting tone and/or a simple 20 test messue Paging sy~tems provide optimized one way communication over a Isrge region through high power transmitter, typically a few Icilowatts, that uses high artennas at multiple ~ites to provide reliable coverage throughout the region Satellite based paging sy6tems are also in operation to provide e%tended service regions Since a pager is merely a recoiver with a small annunciator, its power requirement is very low As such, a pager is 25 quitc small, light weight, reliable, relatively low cost, and can operate for long intervals before it~ batteries need to be recharged or replaced Due to the adv~ntages in ~ize, cost aDd operatiog duration offered by poclcet pagerr, ttempts e~ist in the rt, to impart limited two way communication into paging ~y~tems which aro thomsolvos highly optimized for one-way trafac One ~uch attempt 30 includes incorporation of an ~answer baclc" message through "roverse~ transmission linlcs between tho individual pagers and the fived sites While these attempts have met with great `~
difficulty, the~e attemph neverthele~s indicate that a ~ubstantial demand e%ists for an ine~pen~ e two-way portable truly tetherless telephonic service tbat overcomes the range limitaaon~ usociated with cordle~ telephones and the weight and cost limitations associated 3S with portablo cellular ~y~tems . . . .
. ~ ' ' '', .

2~0~9~

Furthermore, various intelligent nctwork services arc now being ~-offered by the local telephone operating companies in aD attempt to provide wireline subscribers with a certain degree of call mobility when they are away from their own wireline telephones These services include call transfer and call forwarding Both call 5 transfer and call forwarding allow a subscriber to program a local switch, using any pushbutton tclephone, to transfer all subsequently occurring incoming calls that would -otherwise be routed to this subscriber's telephoDe to a telephone associated with a different wireline telephone number that the subscriber desires anywhere in the world either for a given period of time, as in call transfer, or uDtil that subscriber appropriately reprograms 10 the switch with a different forwardil~g Dumber, as in call forwarding In this manner, the subscriber caD, to a certain e~ctent, coDtinually iDstruct the telephone network to follow his or her movements and thereby route his or her incoming calls to a differcnt number in unison with that subscriber's actual route of travel Unfortunately, with these services, the subscriber must manually interact with the networlc and continually enter a new fonvardiDg 15 telephone number(s) coincident with his or her continuing travel such that the network is always cognizant of the current telephone Dumber to which his calls are to be forwarded Thus, a substaDtial overall need e~ists in the art for a truly portable personal communicatioD technology that is designed for pedestrian use and which utilizes small, lightweight and relatively ine~pensive portable transceivers while eliminating, or at 20 least substantially reducing, the performance drawbacl~s associated with the use of currently e~isting tetherless telephonic techDologies in portable communication applications In an attempt to provide this needed technology, the art has turned to low power portable digital telephony In essence, this technology, similar to cellular radiD, uses a fived base unit (hereinafter referred to as a port) and a number of mobile transceivers 25 (hereinafter referred to as portables) that can simultaneously access tbat port on a multiple~ed basis However, in contrast to cellular radio, portable digital telephony uses low power multipleIed radio linl~s that operate on a time division multiple~ed/time division multiple access (TDM/TDMA) basis to provide a number of separate fully duple~ demand assigned digital channels between a port and each of its associated portables Speclfically, 30 each port would transmit ame division multiple~ed (TDM) bit streams on a predefined carrier frequency, with, in turn, each portable that accesses that port responding by transmitang a TDMA burst on a common though different predefined carrier frequency trom that used by the port Quadraturo phasc shift l~eying (QPSK), with an inter-carrier spaclng of 150 to 300 KHz and within a given operating frequency band situated somewhere 35 betweeD appro~imately O.S to 5 GHz would be used by both the port and portables The power used by the transmitter In the portable would range between 5-10 milliwatts or less on average and provide range of several hundred to a thousand feet As such, the resultiDg ' ~ ~ ' ' .

' ~

" ~ v ~

Z0039~

low radiated power would pose essentially no biological radiation hazard to any user. In addition, the port aDtenDa would be relatively small and suitable for mounting on a utility or light pole. With this transmiSsioD range, a port could simultaneously servc typically 20-30 separate locally situated portables. The same TDM chaDDels would be reused at ports that 5 are spaced sufficiently far apart to rcduce co-channel i~terfercnce to an acceptably low level but yet conserve valuable spectrum. To provide access to the wireliDe telephone network, each port would be intcrfaced, typically through a conventional fL~ed distributioD facility, over either a copper or fiber connection to a switching machine at a local central office. The switching machine would be suitably programmed, jD a similar manner as is an MTSO, to 10 controllably and automatically handoff calls from one port to another as subscribers move their portables from port to port.
Due to the very limited traDsmitter power, each portable is aDticipated to be very light-weight, physically small and provide a relatively long operating life between battery recharging or replacement. The cost to a subscriber for a portable is e~pected, 15 through very large scale integrated (VLSI~ circuit implementatioDs, to reside in the range of SloO.00 to S350.00. In addition, each port would rcquire a relatively small electronic paclcage and carry an overall e~pected cost of less than $25,000.00 -- which is far less, by at least an order of magnitude, than that of a current cellular base station. Moreover, the digital data carried OD each channel could be readily encrypted to provide a desired degree 20 of security and privacy against eavesdropp;ng. Furthermore, with this technology, a port antenna, due to its small size, could be readily moved withh a buildiDg to cope with signal attenuation occurring therein. Port spacings would be properly established within the building and frequency reuse would be properly controlled between these ports to provide portable service having an acceptably low level of co-channel interference to a high density 25 of users situated therein.
From a networlc perspective, low power portable digital telephony is e~tremely attractive. At present, appro%imately S50-100 billion is invested by local operating telephone companies in costs associated wlth copper subscriber loops that run from distributioD points to local telephone company demarcation poiDtS on individual customer 30 drops. For a local telephone company, the per-subscriber cost of installing and maintaiDing a ~ub~criber loop 1~ generally greater t the loop end closest to a subscriber than at the far end thereof since the loop end is more dedicated to that subscriber than the far end is. Given the range pro~/ided by portable low power telephony, ports car, be appropriately positioned throughout an area to provide radio linlc based e~change access and thereby substitute 35 Ine~pensive mass produced VLSI circuitry for costly dedicatcd copper loops that would ~ ~ -otherwise emaDate from a distribution facility to an individual subscriber. HeDce, by iDstalling various ports throughout for e~ample a building, significant labor intensive ""- :' , 2~3954 instalLltion iund miuntenance k~lsks ilssociiated with rewiring of telephone drops imd relociltioll of telephone equipment would be eliminnted with subst~mthll savings being advan~ageously reillized in attendimt subscriber costs as people are moved from office to office therein.
Now, with the attractiveness of low power porkable digikll telephony being readily S applurent, its success, in great measure, hinges on achieving satisfactory performimce through the use of TDMA. TDMA, as currently envisioned for use in low power porkable digikll telephony, will utilize time multiplexed 164-bit bursts for communication fTom each of the porkables to an ILssociated port and 180-bit TDM packets for communication from that port to each of these portables. To yield a datil rate of 16 kbits/second, two successive TDM/TDMA time slots are assigned by the port to each porkable in use. Each 10 TDM packet that is trlmsmitted by the port in any one TDM time slot contains 180 bits. Of these bits, the first sixteen bits contain ia predefined framing synchronization piattern, the next three bits are dummy bits, followed by 161 bits in which the first 147 bits conklined therein hold data and the last 14 bits hold a parity sequence. Unfortunately, different propagadon delays between the port ~md its associated portables ~md -timing differences, the latter resulting from clock jitter occurring between the port and these portables, will :
15 both occur. Hence, to prevent different TDMA bursts that are transmitted from different porkables from overlapping in time, a guard time having a 16 bit duration is used in lieu of the frame synchronization pattern in each TDMA burst transmitted by a porkable to the port. The transmitter in the portable remains off during this guaTd time. Accordingly, each TDM packet transmitted from the port to a porkable cont.uins 180 bits with a self-contained synchronization pattern; while each TDMA burst transmitted from a portable 20 to the port contains only 164 bits and no synchronization patterm For quadTature phase shift keying h~lnsmission, the phase of the Intermediate Frequency (IF) calrier is modulated to one-of-four phase angles. separated by 90 degrees in the phase pllune, in accordance with the bit pattern to be trlmsmitted. Each symbol, consisting of plural cycles of the IF carrier at the modulated phase Mgle, tilereby represents two bits in the data stream. Eacll TDM packet 25 transmitted from the port to a portable thus conkains 90 phase modulated symbols Md each TDMA burst transmitted from a pork~lble to the port contains 82 phase modulated symbols.
Although TDMA has been successfully used for quite some time in fixed microwave satellite communicntions, the use of TDMA in the art of low power porkable digi~lll telephony is quite new. In generai, the art has traditionaUy shunned the use of TDMA in such single user applications 30 for a variety of reasons, one of which being the complexity inherent in controUing a TDMA ch.~mnel.

. .
:' .' ', zoo~

In this regard, one crucial function required in TDMA for usc iD IOW
power tclephony is the need to determine the optimum time within a symbol interval to sample the signal and decide what the phase angle iD fact is Once this optimum poiDt for symbol timing is determined, all the symbols within a burst caD be demodulated using 5 carrier recovery circuitry and the burst decoded and converted to an analog specch signal A problem inherent in radio telephony systems of this type transmitting between a stationary and moving station is that very minor changes in propagation characteristics cause the phase of the incoming carrier to either the port from the portable or to the portable from the port to change dramatically As a result, the phase relative to zero degrees of each burst 10 recoived by either the port of any portable is unl nown even though a burst may have been recoived just a few milliseconds previously Similarly, symbol timing varies from burst-to-burst Thus, at both the port and each portablel symbol timing must bc dctermined on a burst-by-burst basis One prior art method for determining symbol timing has used headors 15 and/or training sequences combined with phase-locl~ed loops to acquire symbol timing Disadvantageously, bits and spectrum are wasted which is a particular problem since, in a TDM/TDMA portable radio communications system, the bursts need to be Icept short to minimize delay which would otherwise impair speech transmissiom In U.S Pahnt No 4,849,991 issued July 18, 1989 to Hamilton W
20 ArDold and Nelson R Sollenberger, the latter a co-inventor herein, a method and circuitry for determining symbol timing in time division multiple access radio systems is disclosed In that method, each received burst is ~ampled at a rate that is si~teen times the symbol rato nd stored in memory. The digitized samples are processed to obtain phase values and differentiai phaso vaiue~ re derlved by introducing a one symbol delay, a differential phase 25 vaiue being derived for each of the ~i~teen sampliDg times per symbol Tho differential ph-~e vaiue~ are compared with e~pected differential phase values and the absolute differences are ccumulated for each of the si~teen sampling times o~ter substantially the eDtire burst The ~ymbol timlng for the bur~t is theD selected to be the particular one of-tbe-~i~teeD ~ampliDg time~ that re~ult~ lD the minimum ~um of differences Once the symbol 30 timing is determined, the ~amples ~elecbd by the symbol timing are fed forward to coherent demodulation c rrier reco~ery circuitry which demodulates the current burst ~tored in memory.
In ddition to selecting tho optimum one-in si~teen sampling time for ~ymbol timing, the circuitry iD this prior art patent independently estimates frequency offset 3S bet veen the carrier frequendes of the transmined and received signals and feeds forward an estimate of the frequency offset to the carrier recovery circuitry which then compensates for the offret in ib dctection functions ~ ~
: ' :

` ~0039~;~
.

It is the problem of frequency offset that hss stimulated thc prescnt invention Both the port and each portable have oscillators which are likely, as time passes, to drift away from each other The frequency offset bet veen the port oscillator and each portable s oscillator can bo controlled through the use of c~cpensive componeDts with precise S frequency standards In order to Iceep the cost of the portable terminals at reasonable level, howcver, it is desirable to use cost effective components iD each portable aDd to tolcrate a reasonable degrce of frequency offset between the transmitted and received carrier frequencies Symbol timing estimation, however, is estremely sensitive to frequency offset Once the frequency offset esceed somc threshold valuc, the symbol timing estimator is 10 liltely not to select the best sampling time with a concomitant dramatic increase in the bit error rate that causes the entire rccciver to fail An object of the prescnt invention is to provide a method for symbol timing estimation that is significantly less sensitive to frequency offset and thus can tolerate a higher degree of frequency offset without a degradation in performance SUl~M~Y OF T~TE INVFl~TlOt~
In accordance with the present invention, in a TDM/rDMA portable radio communications system, symbol timing and frequency offset are jointly estimated at the port and at each portable for each received burst by selecting the best sampling time using an aigorithm which is signiacantly less sensitive to frequency offset than the prior art ~s in the prior art patent noted hereinabove, the reccived burst is converted to an IF signal and digitized at a sampling rate which is at least four and preferably s~teen times the symbol rate, over substantially the duration of the burst Each digitizod sample is processed to obtain in phase and in quadrature signals which are then converted to a phase value A symbol delay is introduced aod a differewe tal~en to form 25 difforential phases v-lues These difforontial phaso valuos, which would bo e~pectod to be clustored around four dlscrote phasc valuoo equally spaced by 90 degrees in the phase plane due to the QPSK modulation format, re collapsod into ono quadrant In accordance with tho present iDvention the collapsed differential phases values at each sampling instant for each symbol re espanded baclt into four quadrants and the polar values converted to C-rtesian 30 coordioates For each sampling Instant for all symbols over substantially the duration of the burst, a vector sum is formed from the Cartesian coordinates of the espanded differential pha e v iues For u mpling t sisteen times the symbol rate, symbol timin6 is detormined to be the p rticul r one-of-si~teen umpling instant at which the vector sum has the strongest coheroncy nd thu- tho l rgest magnitude.
An estimate of the froquoncy offsot is dircctly determined from the ngle of that resultant largest summation vector and, in particular, the phase iDcrement due to frequency offset in one symbol pedod is equal to one-forth of the vector angle :~ :

200~95~

B~l~ DE~C~IpTION OF TEIE~ DRAWI~
FIG 1 is an overall diagram of a low power TDM/TDMA portable digital telephony system that incorporates the teaching of the present invention;
PIG 2 is a block diagram of the demodulating circuitry at cither the 5 port or a portable that converts the receivcd radio traDsmission into a data stream, and which circuitry includes the joint frequency offset and symbol timing estimator of the present iDvention;
FlG 3 shows the differential phase constellation in the presence of a constant frequency offsee for the circuit iD FIG 2;
FIG 4 is a bloclc diagram of the joint frequency offset and symbol timiDg estimator of the present in~ention; and FI(; 5 illustrates a graph of bit error rate performancc Yersus frequency offset for both the prior art and thc present invention To facilitate understanding, identical reference numerals ha~re been 15 used, where appropriate, to designate identical elements that are common to the figures . :, An overall diagram of low power portable TDM/TDMA digital telephony system 5 that iDcorporates the teachings of the present invention is shown in FIG
1 Low power digital portable telephony utilizes a fi~ed base unit (hereinafter referred to as 20 a ~port") and a number of mobile transceivers (each of which is hereinafter referred to as a "portable") Through use of time division multiple access (TDMA), each portable can access the port through a separate demand assigned TDMA channel to carry duple~ communication on a time division multlple~ed (TDM) basis therebetween The power used by the transmitter in each portable would range between 5-10 milliwatts or less on average and 25 provide a range of soveral hundred to a thousand feet betwecn a port and each of its porhbles To accommodate a relatively large service area, several ports are used with individual cslls being successively handed off from port to port as their corresponding callers carry their portables from tbe ~ervice area associated with one port to that of an djacent port An appropriate witch (not shown) wbich is located within a local central end 30 office i~ ~uitably programmed to operate in a ~imilar manner as does a mobile telephone ~witchiDg offlce in order to controllably handoff calls from port to port as the callers transit correspoDdiDg local service areas associated thcrewith Specifically, sy-tem S contains four ports 30, 40, 50 and 70 and re~pective portables 34 nd 36; 42, 44, 46 and 48; 52, 54, 56 and 58; and 72, 74 and 76 The 35 ports themselve~ are connected to the witch located within central office 10 to provide acceu to a wireline telephone networl~. This connection can typically occur in one of two ways either through copper digital lines 16 and 18 for illustratively ports 70 and 50, '''',,',: '~

Z0039~4 . ~

respectively, or Vi8 intermediary copper digital lines 23 and 27 to remote clectronics 20 aDd 25 for illustratively ports 30 and 40, respectively The remote electrooics contain fi~ed distribution and concentration facilities for multiple~iog traffic, in sddition to that provided by ports 30 and 40, onto fiber feeders 12 aDd 14 which, in turn, feed central office 10 The 5 switch located within the central office is connected, through trunk 7, to the wirelioe telephone networlc Each port would transmit time division multiplexed (TDM) bit streams on a predefined carrier frequency using quadrature phase shift Iceying (QPSK) modulation, with ao intcr-carrier spacing of 150 to 300 KHz and withio a given operating frequency band 10 situated somewhere between appro~imately simultaneously serve as maDy as typically 20 separate locally situated portables that each carries digitized speech at a bit rate of 16 Icbits/second Here, ports 30, 40, 50 and 70 respectively serve portables 34 and 36; 42, 44, 46 and 48; 52, 54, S6 and 58; and 72, 74 and 76 The same TDM channels would be reused at different ports that are spaced sufficientJy far apart, such as ports 30 and 70, to reduce 15 co-channel interference to an acceptably low level but yet conserve valuable spectrum However, adjacent ports would be situated sufficiently close together in order to provide an appropriate degree of overlap of the;r respective service areas thereby ensuring no loss of coverage occurs during call handoff Furthermore, each port utilizes a suitable antenna, such as antenna 31 for port 30, to carry its TDM transmissions to its associated portables and 20 receive TDMA bursts therefrom GiveD the carrier frequencies being used, each of these antennas is relatively small and suitable for mounting on utility or light pole Inasmuch as system 5 replaces local copper drops and telephone cords with short range low power radio linl~s, ambulatory callers are provided with completely tetherless acce~s Accordingly, through radio linl~s 33 and 38, i11ustrative callers 35 and 37 25 located within respective residences 63 and 67 are completely free to move within the entire ~enice area provided by system 5, i e that provided by port~ 30, 40, S0 and 70, while advantageou~ly maintaining continuity of their e~isting telephone conversations as well as being able to place calls through other (~non-home") ports u thelr travel progresses Each port continuously transmits on a TDM basis, while portables 30 transmit in burst~ on a TDMA basis to their associated port Two different carrier frequencies are u~ed to c rry communication between each port and a portable onefrequency, frequeDcy fl for port 30, to carry communication from that port to each of its portables nd another frequency, frequency f2 for port 30, to carry communication frorn each of the~e portable~ to thh port Although adjacent ports used different pairs of 35 frequencies, these c rrier frequencies are aiso reused for ports that are spaced sufficiently far apart from each other to consene ~pectrum The spacing is appropriately set to ensure ;
that co-channel interference that might occur at any port will remain at ao scceptably low .. : .

20039~,~

~ 14 level FIG 2 shows a block diagram of a digital dcmodulator 201 that incorporates the joint frequency offset and symbol timing estimator 202 of the present invention This demodulator would be incorporated at both the port and at each portable S unit in a TDM/rDMA portable radio communications system As aforenoted, the port transmits TDM bursts of 90 symbols while the portable transmits TDMA bursts of 82 symbols For purposes of the discussion hereinafter, it will be assumed that the demodulator in PIG 2 represents a demodulator in a portable unit aDd therefore rcceives bursts of 90 symbols transmitted by the port An antenna 203 receives cach radio transmitted burst having a carrier frequency in the range of 5 to 5 GHz Analog receiver 204 amplifies, filters and downconverts the radio frequency signal to a 1 MHz IF signal Analog-to-digital converter 205 digitizes the IF signal by sampling at si~teen times the symbol rate, for this system the latter being 250 KHz The A/D sampling rate is thus 4 MHz Digital front-end circuitry ~5 206 translates the digital signal to baseband which is low pass filtered to obtain in-phase (I) and in-quadrature (Q) signals A ROM in the front-end circuitry 206 servcs as a lool~-up table of arctangents to determine thc phase of each sample from the I and Q signals For each burst of 90 symbols, the output of circuitry 206 thus consists of 1440 (90 x 16) phase samples which are stored in a one burst RAM delay 207 As the phase ~lalues of the over-sampled burst are stored in RAM
delay 207, the burst is simultaneously processed by the circuitry of the present invention In particular, the present in~ention processes each burst to select the particular sampling time, among the skteen sampling times per symbol, that is most lilcely to be closest to desired salDpling instaDt which has the smallest timing error As will be described in detail 25 hercinafter, joint frequency offset and symbol timing estimator 202 processes differential pha o ~alues at each of the possible si~teen sampling times o~er substantially the entire burst to select the optimum oDe ln--i~teen sampling time SiDce there is no ICDOWD a priori reference angle with which to compare the receivod phase outpuh of front end circuitry 206 with, a differential phase 30 clrcult 208 forms dlfferentlal phase values by imparting a one symbol delay (si~teen samples) to the output of circuit 206 and forming a difference between each current sample phs~e and the phase si~teen samples pre~riously For QPSK modulation, it would bee~pected that at the sampling time closest to the optimum time, the differential phase values would be elther 0, 90, 180 or 270 degrees Noise aDd interference, however, will cause 8 35 devlation between the actuai phase vaiues and these e~pectcd phase values over the burst ~Iso, a frequency offset that is constant over tho burst will impart a rotatian to the constellatioD of phase values PlG 3 shows the resultant constellation of e~pected phase ' .
'~.-',, :,:

~ ~ `

200395'~ :

values at the output of differential phase circuit 208 As can be noted, the e~pected differential phase values are separated by 90 degrees in the phase plane but are offset from the a~es by an aDgle ~, ~ representing the phase increment due to frequency offset in one symbol period S In tho aforcnoted prior art patent, the particular sampling instant, amongst the si~teen possibilities, is selectcd by evaluating for each of the possibilities, the tightness of the cluster over most of the burst since, at this optimum time, thc differential phase values are most closely grouped together around the e~pected values In order to more easily evaluate the tightness of the cluster the differential phaso values are collapsed 10 into one quadrant by removing the two most-significant-bits (MSB) of the digital differential phase valucs The tightness of the clusters for each of the si~teen sampling times are evaluated by summing the absolute differences between the collapscd differential phase values and the espected phase value, i e zero degrees Symbol timing is then selected to be that sampling time that yields the minimum sum of differences It has been found, however, 15 that as frequency offset increases, these sum of differences are not likely to be an accurate indicator of the "best" sampling time for symbol timing and, in fact, are likely to point to a sampling time substantially displaced from thc optimum time As a result, when this inaccurate symbol timing i8 used in a carrier recovery circuit, the recovered data is likely to have a high error rate, and in the e~treme, causes the recovery circuit to totally fail The joint frequency offset and symbol timing estimator 202 of the present invention, to be described in detail hereinafter, employs an algorithm that is substantially more robust to frequency offset As a result, a significantly larger frequency offset can be tolerated ~vhile still achieving an acceptable performance level In particular, estimator 202 evaluates the tightness of the clusters of differential phase values for each of 25 the ~istecn sampling in-taDts by esamiDing the coherency of each cluster rather than differeDces between the differential phases and espected values, the latter, in the presence of frequency off~et, not being actually determinable Again, the differontial phaso values at the output of circuit 208 are collapsed Into one quadrant to mal~e the evaluation independent of in which quadrant these 30 pha~e valuos lio Tbo polar coordinate on the UDit circle of each collapsed differential phase value, however, 1~ e~panded by a factor of four into a full plane representation. For each of tho ~isteen rampling time~, a vector sum over substantially the entire burst is formod from tho collap~ed and e~panded cluster~ by separably summing the CartesiaD X coordinates of tho polar collapscd and espanded phase values, and the corresponding Cartesian Y35 coordinates. At the optimum sampling time the constellation will be the most coherent and the ~ummation vector will add most strongly to yield the largest magnitude vector At ~ampling instants offset from the optimum symbol timing, the collapsed and espaDded ., 2003~

differential phase values will not be as tightly clustered together over the burst and the resultant vector sum will not produce a summation vector having as large a magnitude as at the "best" sampling instaDt Frequency offset can be directly determined from this largest S magnitude summation vector by DotiDg that the polar angle of this vector rcpreseDts four times the phase increment due to frequency offset in one symbol period Thus once the largest vector is selected, estimator 202 determines its angle, and thus the frequency offset, using thc arctangent function and its X and Y coordinstes Estimator 202, once having determined which sampling time is "best"
10 for symbol timing, forwards thc indes of that sampling time to a gate 210 The phase values stored in RAM delay 207 at the selected "best" sampling time are ga~ed, one sample per symbol, to coherent carrier recovery circuit 211 The frequeDcy offset on lead 212 is also input to carrier recover circuit 211 so that the latter may compensate for the detected offset An esample of circuitry thst may be used for coherent carrier recovery circuit 211, is the 15 subject of co-pending CanadiaD patent applicatioD of the iDveDtors of the present inventioD, Serial No 596,528, filed April 12, 1989, It is also disclosed in the aforcnoted U S Patent No 4,849,991 The output of circuit 211 is the recovered bit stream which, at each portable unit, is converted by a digital to analog circuit, to a voice signal for reception by the user JoiDt frequeDcy offset and symbol timing estimator 202 vill be 20 described with reference DOW to the block diagram in 171G 4 Estimator 202 includes a control sigDal geDerator which, in responsc to an c~pected start of burst signal on lead 451, generate~ the control signals used by the various circuit blocl~s In the port, the espected start of blocl~ signal has a fised timing relationship to the transmit burst position In the portable units, this signal can be generated using a technique which can derive the 25 appro~imate otart of a bur~t ~uch ao a techDique described in a paper by L Chang and N
Sollenberger eDtitled "The Uoe of Cyclic Blocl~ Codes for Synchronization Recovery iD a TDMA Radio System", Procccd~n~s of ~h~ Th~rt Nord~c S~mlnar on D~g~tal Land Mo~
Rat~oCommun~catlons Copcnhag~n D~nmar~ September 13-15, 1988, pp 1-8 The signalsgeneratcd by coDtrol oignal generator 401 include a ~tart of burst (SOB) sigDal OD lead 402, 30 iD re-poD-e to the first clocl~ cycle in the bur~t, aDd an end of burst signal (EOB) on lead 403, iD respoD~e to the last cloclc cycle in the bur~t A window signal which i9 active for that preset suboot of tho burot during which oymbol timing is processed is also generated on lead 404. ID particular, ODIY tho 64 symbol periods in the ceDter of the 90 symbol burst are employed for ~ymbol timing purpo-es, 13 symbols being clipped from each end of the burst 35 By elimiDatiDg the edge~ of the burst for purposes of symbol timing and frequency offset eotimatioD, he uncertaiDty as to e~actly when a bur~t arrives can be eliminated and only that dah in the ceDter of the burst that is free of noise and interference considerations need be ; ~ -`: 20039~:~4 -17- ;

considered for symbol timing purposes. During the last symbol of the burst, LS Icad 405 is active. As will be noted, during this last symbol period, which follows the active window period, the magnitude of the sum vectors for each of the sistcon sampling instants arc calculated and the sampling timc associated with the largest sum vector is dctermincd.
A free ruDning clock 406 operating at the sampling rate of 4 MHz, equal to sisteeD times the symbol rate, is conDccted to a counter 407 which counts samplc times modulo 16, from the start of each burst and in response to the SOB signal at its clear input.
Differential phase values from differential phase circuit 208 at each OI
10 the sisteen sampling instants per symbol are input to estimator 202 in parallel format over 8 input leads 408. Each sample represents, in polar coordinates, the differential phase angle on the unit circle in the phase plane. As aforenoted, by discarding the two MSB of this input, the differential phase values are collapsed into the first quadrant of the phase plane which has the effect of removing the modulation from the differential phase samples. The 15 resultant collapsed one-quadrant differential phase values are then espanded into the full four-quadrant phase plane by multiplying thesc phase values by a factor of four and then converting them from polar into Cartesian coordinates. Tho collapsed differential phase ~alues on leads 409 are thus input to ROM lool~-up tables 410 and 411 which produce at their outputs, cos4~ and sin4~, respccti~ely, of the input phase aDgle, ~. The outputs of 20 ROM tables 410 and 411 thus represent the X and Y components, respectively, of the espanded collapsed differential phase angles in Cartesian coordinates.
If the sisteen-times sampled symbol is within the prescribed window (the 14th through 77th symbol), then lead 404 at the output of control generator 401 Is active and the X and Y components of the sampled differential phase are gated through AND ;
25 gates 412 and 413"espectively. The outputs of gates 412 and 412 are input to adders 414 and 415, respectively, and the outputs thereof to 16-bin accumulators 416 and 417, respecave1y. Accumulators 416 and 417 are cleared at the boginDing of each burst by an SOB pulse from control signal generator 401. The iDpUts to accumulators 416 and 417 are cycled through each of the sisteen bins in response to the 0-15 count of counter 407 on lead 30 419 which counts modulo 16 at the sampling timing rate of 4 MHz, equal to sisteen times the symbol rate. In accordance wlth the count of counter 407, the values stored in the appropriate blns of accumulators 416 and 417 are fed baclt to the adders 414 and 415, respectively. Adders 414 and 415 thus sum the current X and Y components at the outputs of ~ND gates 412 and 413, respectively, with the corresponding accumulated sums. The ncw 35 ~ums are then input to the accumulators. Each bin in accumulator 416 and 417 contains for the associabd sampling Instant, the sum of X components, ~:X, and thc sum of Y
componenb, ~:Y, respectively, as the burst is processed.

2003~S~

1D order to determine the sampling timc yielding the largest sum vector, the Icngth of the vector resulting from the sum of the X and Y components is calculated Thc length of this vector is equal to the square root of the sum of thc squarc of the sum of the X components and the square of the sum of the Y componcnts Therefore S (X~2 + (f~ y)2 is calculated for each of the si~cteen sampline iDstaDts, and the sampling instaDt sssociated with that largest vector then determined The outputs of accumulators 416 and 417 which cycle through their 16 bins in rcsponse to the counter of couDter 407, are input to ROM squariDg tables 420 and 421, respectively The output of ROM 420 is thus (~X)2 and the output of ROM 421 is (~ y)2. These outputs are summed by adder 42210 SiDce the outputs of ROMs 420 aDd 421 and thus adder 422 are cha~ging during the window period from the 14th through the 77th symbol, detcrminatioD of the largest sum is aftcr the last symbol in the vvindo~v, and in particular during the last symbol of the burst During this last symbol period, for the si~tecD sampling iDstants, the last symbol LS output of generator 401 is active, which permits, as dctailed hereinafter, the detcrmiDation of the --15 sampling instant at which the largest vector of leDgth (~ X)2 + (~ y)2 occursThc output of digital adder 422 is input to subtracter 423 and a register 424 which holds the magnitude of the biggest vector as the accumulators 416 and 417 are cycled througb their 16 bins Register 424 is cleared at the begiDniDg of thc burst by an SOB pulse and is loaded with the output of adder 422 oDly if its load input, connected 20 to the output 425 of AND gate 428, is activc Register 424 can thus only be loaded at one of si~teen sample instants during the last symbol pcriod The value stored in register 424 is also present on leads 426 which form the second input to subtracter 423 Subtracter 423 forms the difference between the output of adder 422 and the value stored h register 424 and produces aD active output on lead 427 only wheD this differeDce is positive `;;
During the last symbol period, at the first sample instant when the count of CoUDter 407 is ZerQ, the 2X values and the ~ Y values in the first bins in accumulators 416 and 417, respectivoly, are squared and summed aDd read into register 424 since, llt that iDstaDt, the output of subtracter 423 aDd thus the output 425 of AND gate 428 sre active At the ucond sample instant, when the CouDt of counter 407 is 1, the second bins 30 of accumulators 416 aDd 417 are squared and summed If the sum at the output of adder 422 Is Iarger than the sum stored in register 424 at the first sampling instant, theD the output of subtracter 423 and output 425 remain active and this new value is loaded in register 424 replaciDg the value loaded at the first sampling instant If the sum at the output of adder 422 is, however, smaller than the sum already stored iD register 424, then the output of 35 wbtracter Is inactive, thereby deactivating output 425 and preventing the new sum at the output of adder 422 from being loaded , ''- '-''-.
. -:,:
. . -:': .- , :' : ' Z003~

The output of AND gate 428 is also coDnected to the load input of a timi~g register 430, the input of which is connected to counter 407. When a vector magnitude is loaded iDto register 424, the associated count of counter 407 is simultaneously loaded into register 430. Thus as counter 407 cycles through its sisteen counting positions, 5 register 424 holds the magnitude of thc largest vector sum so far and timing register 430 holds the associated sampling instant at which that largest vector occurred. At the end of the sisteen sampling instants in the last symbol period, the desired largest vector magnitude is held in register 424 and its associated sampling instant is held in register 430. It is this sampling instant that is selected for symbol timing of the entire burst being held in the RAM
10 delay 207 in FIG. 2.
At the end of the burst and in response to an end-of-burst EOB signal from generator 401, the symbol timing sampling instant in register 430 is loaded in register :
431 which holds the iDdes of the sampling time used to demodulate the previous burst. As counter 407 continues to count modulo 16, its count is compared by comparator 432 with the 15 new symbol timing in register 431. When equal, during each symbol period, comparator 432 generates a pulse on its output 209 (also noted in ~IG. 2). With reference to FIG. 2, this pulse on lead 209 is input to gate 210 which gates the appropriate 1-out-of-16 sampled phase value out of RAM delay 207 to the coherent carrier recovery circuit 211 for data recovery.
In order to compensate for frequency offset between the port and the portable unlt, reco~rery circuit 211 also requires an estimatc of the offset. Frequency offset estimation is simultaneously performed with symbol timing determination. As previously noted, an estimate of frequency offset is determined from the angle that the largest ~ummation vector malces in the phase plane. Thus oncc the largest vector is determined, its 25 nglo can bc found from Its abscissa and ordinate components. With referencc again to PIG. 4, the sis MSB leads of the 8 parallel output leads of accumulators 416 and 417 are Input to rctangent ROM 440. ROM 440 ~erves as a loolc-up tablo to determine rCtan(4~Y ~X2-. Tho output of ROM 440 is input to frequency register 441 which is loaded ODIy when the output 425 of AND gato 428 is active. At tbe end of the si~teen sampling 30 In~tants in tho last bit interval, therefore, the angle stored In register 441 is the angle corrospondlng to tho ~ummation voctor having tho largest magnitude and is equal to the phase increment due to frcquency offset in a onc symbol period. At the end of the burst, nd in rosponu to n EOB pulse, the angle stored in register 441 is loaded into current frequency register 442 for proccsslDg the burst stored in RAM delay 207 to be input to the 35 carrier rocovery circuit 211 (in FIG. 2). This offset estimation is thus provided over leads 212 to the coherent carrier recovery circuit 211.
' "`' '"':

Z0039~

FIG. S shows the bit error ratc performance as a function of frequency offset using the method of the present iDVentiOn and the method of the prior art for a 10 dB
signal-to-noisc (SNR) ratio for the parameters of the system described, i.e. symbol rate of 250 KHz, sampling at 4 MHz, and a 1 MHz IF signal. As can be noted, the prior art S method cannot tolerate a frequency offset greater than 10 KHz while the mcthod of the present invention has essentially no degradation until 27 KHz.
The above-described embodiment is illustrative of the principlcs of the present invention. Other embodiments could be devised by those skilled in the art without departiDg from the spirit and scope of the present invention.
-: -. , .' ,., ':

:, ' ' :.': ,,'. , :, ~"
:' '., ~:', '; '. ',:" , '.' ' ,., ~: , , .. ,, :
.'; ,. :-.
:; .
.~
~ . .~' ,

Claims (30)

1. A method of determining symbol timing in the processing of a digital phase modulated radio signal transmission consisting of sequential bursts of symbols, comprising the steps of:
converting an Intermediate Frequency (IF) radio signal from a modulated sinusoidinto phase values after sampling said IF radio signal at a plurality of sampling times per symbol at a sampling rate which is a multiple of the symbol rate by at least a factor of four;
obtaining differential phase values at each sampling time from said phase valuesby introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values;
collapsing said differential phase values into one quadrant in the phase plane;
expanding said collapsed differential phase values into a full 360 degree polar phase plane; and obtaining symbol timing by selecting from amongst the set of sampling times the particular sampling time at which the vector sum of the collapsed and expanded differential phase values, substantially over an entire burst, has the largest magnitude.

2. The method of claim 1 further comprising the step of:
obtaining a frequency offset estimate from the angle made in the phase plane by the vector sum with the largest magnitude.

3. The method of claim 1 wherein said sampling rate is a factor of sixteen timessaid symbol rate.

4. The method of claim 1 further comprising the step of:
storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing is being determined.

5. The method of claim 4 further comprising the step of:
selecting for demodulation the stored phase values at the determined symbol timing.

6. A method for obtaining a symbol timing estimate in the processing of n digital phase modulated radio signal consisting of sequential bursts of symbols, comprising the steps of:
digitizing an Intermediate Frequency (IF) signal at a plurality of sampling times per symbol at a sampling rate which is a multiple of the symbol timing rate by at least a factor of four:
processing the digitized signal to obtain in-phase and in-quadrature signals;

converting said in-phase and in-quadrature signals to phase values at each sampling time;
obtaining differential phase values from said phase values by introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values;
collapsing said differential phase values into values between 0 and 90 degrees in the phase plane;
expanding said collapsed differential phase values into values between 0 and 360degrees in the phase plane;
forming the vector sums in the phase plane of the collapsed and expanded differential phase values separately for each sampling time and substantially over the entire burst; and obtaining symbol timing by selecting from amongst the set of sampling times the particular sampling time at which the vector sum of the collapsed and expanded differential phase values has the largest magnitude.

7. The method of claim 6 further comprising the step of:
obtaining a frequency offset estimate from the angle made in the phase plane by the vector sum having the largest magnitude.

8. The method of claim 6 wherein said sampling rate is a factor of sixteen timesthe symbol rate.

9. The method of claim 6 further comprising the step of:
storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing is being determined.

10. The method of claim 9 further comprising the step of:
selecting for demodulation the stored phase values at the determined symbol timing.

11. A method of determining symbol timing in the processing of a digital phase modulated radio signal transmission consisting of sequential bursts of symbols,comprising the steps of:
converting an Intermediate Frequency (IF) radio signal from a modulated sinusoidinto phase values after sampling said IF radio signal at a plurality of sampling times per symbol at a rate which is a multiple of the symbol rate by at least a factor of four;
obtaining differential phase values at each sampling time from said phase valuesby introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values;

collapsing said differential phase values into polar values between 0 and 90 degrees in the phase plane;
expanding said collapsed differential polar phase values into polar values between 0 and 360 degrees in the phase plane;
converting the polar values of said expanded and collapsed differential phase values into Cartesian coordinates;
summing separately for each sampling time and substantially over the entire burst the X Cartesian coordinates of the expanded and collapsed differential phase values, and summing separately for each sampling time and substantially over the entire burst the Y Cartesian coordinates of the expanded and collapsed differential phase values;
for each sampling time squaring the sum of the X coordinates and squaring the sum of the Y coordinates, for each sampling time adding the square of the sum of the X coordinates and thesquare of the sum of the Y coordinates; and determining at which sampling time the sum of the square of the sum of the X
coordinates and the square of the the sum of the Y coordinates has the largest magnitude and using that sampling time as the time for symbol timing.

12. The method of claim 11 further comprising the step of:
calculating for each sampling time an angle equal to one-fourth the arctangent(sum of the Y components, sum of the X components) and determining the estimate of frequency offset from the angle calculated at the sampling time chosen for symbol timing.

13. The method of claim 11 wherein said sampling rate is a factor of sixteen times said symbol rate.

14. The method of claim 11 further comprising the step of:
storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing is being determined.

15. The method of claim 14 further comprising the step of:
selecting for demodulation the stored phase values at the determined symbol timing.

16. A circuit for providing a symbol timing estimate in the processing of a digital phase modulated radio signal transmission consisting of sequential bursts of symbols, comprising:
means for digitizing an Intermediate Frequency (IF) signal at a plurality of sampling times per symbol at a sampling rate which is a multiple of the symbol rate by at least a factor of four;

means for processing the digitized signal to obtain in-phase and in-quadrature signals;
means for converting the in-phase and in-quadrature signals to phase values at each sampling time;
delay and differencing means for introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values to form differential phase values;
means for collapsing said differential phase values into values between 0 and 90degrees in the phase plane;
means for expanding said collapsed differential phase values into values between0 and 360 degrees in the phase plane;
means for forming the vector sums in the phase plane of the collapsed and expanded differential phase values separately for each sampling time and substantially over the entire burst:
means for determining the vector sum with the largest magnitude: and means for obtaining symbol timing by selecting amongst the set of sampling timesthe sampling time associated with the vector sum having the largest magnitude.

17. The circuit in claim 16 further comprising means for obtaining an estimate of frequency offset from the angle made in the phase plane by the vector sum having the largest magnitude.

18. The circuit in claim 16 wherein said sampling rate is a factor of sixteen times the symbol rate.

19. The circuit in claim 16 further comprising means for storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing is being obtained.

20. The circuit in claim 19 further comprising means for selecting for demodulation the stored phase values at the determined symbol timing.

21. A circuit for providing a symbol timing estimate in the processing of a digital phase modulated radio signal transmission consisting of sequential bursts of symbols, comprising:
means for converting an Intermediate Frequency (IF) radio signal from a modulated sinusoid into phase values at a plurality of sampling times per symbol at a sampling rate which is a multiple of the symbol rate by at least a factor of four;
delay and differencing means for introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values;

means for collapsing said differential phase values into one quadrant in the phase plane;
means for expanding said collapsed differential phase values into a full 360 degree polar phase plane;
means for forming the vector sums in the phase plane of the collapsed and expanded differential phase values separately for each sampling time and substantially over the entire burst;
means for determining the vector sum with the largest magnitude; and means for obtaining symbol timing by selecting amongst the set of sampling timesthe sampling time associated with the vector sum having the largest magnitude.

22. The circuit in claim 21 further comprising means for obtaining an estimate of frequency offset from the made made in the phase plane by the vector sum having the largest magnitude.

23. The circuit in claim 21 wherein said sampling rate is a factor of sixteen times the symbol rate.

24. The circuit in claim 21 further comprising means for storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing estimate is being obtained.

25. The circuit in claim 24 further comprising means for selecting for demodulation the stored phase values at the determined symbol timing.

26. A circuit for determining symbol timing in the processing of a digital phasemodulated radio signal transmission consisting of sequential bursts of symbols, comprising;
means for converting an Intermediate Frequency (IF) radio signal from a modulated sinusoid into phase values after sampling said IF radio signal at a plurality of sampling times per symbol at a rate which is a multiple of the symbol rate by at least A factor of four;
delay and differencing means for introducing at least a one symbol delay and taking differences between delayed phase values and corresponding present phase values;
means for collapsing said differential phase values into polar phase values between 0 and 90 degrees in the phase plane;
means for expanding said collapsed differential phase values into polar values between 0 and 360 degrees in the phase plane;
means for converting the polar values of said expanded and collapsed differential phase values into Cartesian coordinates;

first summing means for summing separately for each sampling time and substantially over the entire burst the X Cutesian coordinates of the expanded and collapsed differential phase values [.SIGMA.X];
second summing means for summing separately for each sampling time and substantially over the entire burst the Y Cartesian coordinates of the expanded and collapsed differential phase values [.SIGMA.Y];
first squaring means for squaring for each sampling time the sum of the X
coordinates [(.SIGMA.X)2];
second squaring means for squaring for each sampling time the sum of the Y
coordinates [(.SIGMA.y)2];
adding means for forming a vector sum for each sampling time by adding the square of the sum of the X coordinates and the square of the sum of the Y coordinates [(.SIGMA.X)2+(.SIGMA.Y)2];
means for determining which vector sum has the largest magnitude;
means for determining symbol timing as that sample time associated with the vector sum with the largest magnitude.

27. The circuit in claim 26 further comprising means for calculating for each sampling time the angle equal to and determining the estimate of frequency offset from the angle calculated at the sampling time chosen for symbol timing.

28. The circuit in claim 26 wherein said sampling rate is a factor of sixteen times said symbol rate.

29. The circuit in claim 26 further comprising means for storing the converted phase values at each sampling time of said sampled IF radio signal while said symbol timing estimate is being determined.

30. The circuit in claim 29 further comprising means for selecting for demodulation the stored phase values at the determined symbol timing.