|Publication number||CA2083303 C|
|Application number||CA 2083303|
|Publication date||30 Mar 1999|
|Filing date||19 Nov 1992|
|Priority date||31 Dec 1991|
|Also published as||CA2083303A1, DE69217891D1, DE69217891T2, EP0550146A1, EP0550146B1, US5367539|
|Publication number||CA 2083303, CA 2083303 C, CA 2083303C, CA-C-2083303, CA2083303 C, CA2083303C|
|Inventors||Terry William Copley|
|Applicant||Terry William Copley, American Telephone And Telegraph Company|
|Export Citation||BiBTeX, EndNote, RefMan|
|Classifications (15), Legal Events (2)|
|External Links: CIPO, Espacenet|
DIGITAL BLOCK PROCESSOR FOR PROCESSING A PLURALlTY OF
TRANSMISSION CHANNELS IN A WTRF.~ F..~S RADIOTELEPHONY SY~ I ~;M
Field of the Invention This invention relates to wil~,less tch~c~ -ic~tion ~ ,t,~ls having a 5 plurality of co.n..~ tinn Ch~nnel~ each having a single FM carrier at a Sepalat~, r~ uency ~ t~ to that ch~nnel It is particularly concPrne~l with the reverse link radio transce;~e. of such a wireless tcleco....~ .-ir~tion system located at a s~tion~ry signal r~di~tion site.
Back~round of the I~ .-tio ~esently t-h-e reverse link signal plocess.. ~ of a wireless telephone co-----~nir~hon system such as a cell site or a microcell require a S~,IJalŗlt; radio tr~scei-/er for each tr~n~mi~sion ch~nnel in service. In a cell site ser~ing up to 30 ch~nnel~ 30 individual tlansc~ ers are required. These ~ scei~ , des~ tP~
radio ch~nnel units (RCUs), are individually expensive and ,e~,lese.lt a major portion 15 of the cell site or microcell cost. Common equip,llent serving a plurality of channels at RF levels may be shared to reduce cost, but SC~ate receiving e4~ipl~ent is l~uil~d at t-h-e IF level for each ch~nnel Summary of the Invention A digital block receiver system, in a cellular/wireless FM
20 radiotPlephony system, receives and heterodynes a block of cellular/wireless receive ch~nnPl~ to a very low IF by analog processing This block IF signal is applied to a precision high speed A/D converter and converted to a di~1'i7f~ time series. A
window function is applied to the digitized time series and a high speed FFr is applied to rl~uency isolate the individual ch~nnelc The active ch~nnelc are 25 yl~)cess~ by a digital signal processor to recover the FM ch~nnel m~ tion.
In particular applic~tion~ teleco.. -.. nic~tion ch~nnpls may encompass analog FM voice signals; analog FM supervisory tones; analog FM ,cign~lling tones;
and digital FSK data mP.ss~ges These ch~nnelc may, in accord with the principles of the invention all be processed as a block.
A comyul~;on~lly efficient method of separating and dem~1ul~hng the signals of the procesced block signal uses the power spectra obtained from the Fourier transform output coefficients to recover the frequency mnd~ tPA carrier of each ch~nnPl In accordance with the invention the in~ neous FM carrier frequency 35 is d~te-...;nPd by colllyu~ing the first moment of the Fourier ~ sroml power spectrum associated with each ch~nnel to be recovered. A power m~king operation may be optionally applied and individual spectra that significantly exceed the mask are assumed to be excessively noise contaminated and are rejected for purposes of the recovery operation. The first moment calculation is repeated without the excessively noisy spectra to achieve an improved estimate of the instantaneous FM carrier 5 frequency.
In accordance with one aspect of the present invention there is provided in a Frequency Division Multiple Access (FDMA) wireless radiotelephone system using FM modulated signals, a method for processing a plurality of FM information signals occurring in substantially adjacent frequency defined channels; comprising the steps 10 of: receiving a block of receive channels as a plurality of analog RF signals to be processed; converting the plurality of analog RF signals into a digitized time series;
windowing the digitized time series into a plurality of defined frequency ranges equal to the number of defined channels; including: defining a window time span havingan effective time duration less than a time duration of 1/4 cycle of a present baseband 15 modulating frequency, and restricting frequency domain sidelobes produced by windowing to a value less than a minimum acceptable post detection signal to noise ratio, generating Fourier coefficients with fast Fourier transforms in response to the windowing in order to identify the information signals of the individual channels of the digitized time series by recovering from the Fourier coefficients instantaneous 20 frequencies from each FM carrier in each individual frequency defined channel;
including: performing at least a first moment calculationthat incorporates powerspectra offset from a channel center frequency by an amount at least equal to a sum of a peak frequency deviation of FM modulation and a highest value of a mod~ ting frequency, and excluding from the first moment calculation power spectra offset from 25 a channel center frequency by an amount that exceeds a sum of a peak frequency, deviation of FM modulation and a highest value of a modulating frequency, recovering the individual voice channels from the instantaneous frequencies of each FM carrier.
In accordance with another aspect of the present invention there is provided 30 in a wireless radio telephone system, apparatus connected for receiving voice and data signals in block form from a group of individual radio signals each incorporating a single FM carrier that is distinct in frequency; comprising a receiver assembly connected for processing the group of individual radio signals in the block forrn and , . ~
'' -- - 2a ~ 3 ~
converting the radio signals to an IF frequency; an analog-to-digital converter connected for converting the radio signal in the block form into a digital time series format; a window function processor connected for applying a window function to the digital time series; a fast Fourier transform processor operative and connected for 5 isolating in frequency multiple spectra associated with each individual FM carrier in the block form; an FM detector connected for estim~tin~ an instantaneous frequency of individual FM carrier based on the first moment of power spectra associated with the individual FM carrier and recovering voice and signal information of each FMcarrier; a detection circuit connected for recovering voice and data signals, SAT
10 (supervisory audio tones) and manchester encoded data; a data link interface connected for forrn~tting the voice and data signals for digital tr~n.~mi~cion to a control and processing location.
In accordance with yet another aspect of the present invention there is provided a signal processing an arrangement for processing a plurality of FM
15 information signals occurring in substantially adjacent frequency defined channels;
comprising: an analog processor for accepting a block composite signal cont~ining the plurality of FM information signals from an antenna; an analog-to-digital processor to convert the single block of RF signals into a digitized time series;
conversion circuitry for converting the block of RF signals into a very low frequency 20 block IF signal having its low end frequency coincident in frequency with Fourier coefficient and its high end frequency coincident with another Fourier coefficient and having added Fourier coefficients spaced symmetrically within the frequency range of the very low frequency block IF signal; digital signal processing apparatus including stored program instructions for windowing the digitized time series and generating 25 Fourier coefficients with fast Fourier transforms to identify the individual channels of the digitized time series by recovering instantaneous frequencies from each FM
carrier in each individual frequency defined channel; and the stored instructions in digital signal processing apparatus operative to determine first moments of Fourier transform power spectra associated with each individual frequency defined channel 30 and for identifying individual spectra distorted by noise and interference and repeating the determination of first moments to generate an improved reduced noise - 2b-content associated with individual frequency defined channels; FM detection circuitry for recovering the individual voice channels from the instantaneous frequencies of each FM carrier.
Brief Description of the Drawin~s FIG. 1 is a block schematic of an illustrative wireless communication system embodying the principles of the invention;
FIG. 2 is a block schematic of an illustrative FM digital block receiver embodying the principles of the invention, FIG. 3 is a graph of the adjacent channel frequency selectivity of an unmodulated FM carrier signal such as is processed by the receiver of FIG. 2;
FIG. 4 is a graph illustrating a performance criterion of the FM detection system;
FIG. S is a graph of FM detector nonlinearities produced by the window function sidelobes;
FIG. 6 is a graph of the power spectra of the unmodulated FM carrier of a typical channel;
FIG. 7 is a graph of the power spectra of the modulated FM carrier of a typical channel;
FIG. 8 is a graph of two tone post detection intermodulation products using seven power spectra per channel;
FIG. 9 is a graph of two tone post detection intermodulation products using five power spectra per channel;
FIG. 10 is a graph of a multiple pass FM detection power mask;
FIG. 11 is a graph of single and multipass FM detector sensitivity;
FIG. 12 is a graph of the voice band post detection amplitude response;
FIG. 13 is a diagram showing special case positioning of power spectra in relation to the channel boundaries for two illustrative embodiments;
FIG. 14 is a tabulation showing a first moment computational efficiency between a special case symmetrical analysis versus a general case analysis;
FIG. l S is a diagram showing the relationship between spectra rejected by the FM detection power mask and channel C/N;
.._ FIG. 16 is a flow diagram showing the method for the mllltipqcs FM
Detailed Description An illu~lldli~,f wireless telephone system is shown in the FIG. 1. A
S mobile ~wilching center 101 (MSC) is connect,~A via a trunk or microwave link, to three cell sites 102, 103 and 104. It inte..;onnectc these cell sites 102-104 to a public ched land telephone nf lwol~ 105. Each cell site 102-104 services a defined geo~hical service area. The illustrative cell site 102 is connected to a microcell 106 located within its service area and a picocell (an in building wil~,less system) 10 serving a b~ ing 107 within the service area. These connec~ionc may be le,llf nt~,d using fiber optics, met~ wire, point-to-point microwave links or a colnbinalion of these Conne~!;Qns.
The cell site 102 serves as a macrobase station which provides telephone service to pe.~onal co~ lni~tion devices of pedestrians 111 and mobile radio 15 telephone service to mobile vehicles 112 within its service area. In ~ldition it serves as a control center servicing the picocell of building 107 and microcell 106 (serving the pedcsl ian 113) and inte~onnecling them to the MSC 101.
Each of the lllacrocells 102 - 104, microcells and picocells includes onc or more ~ntf nn~ for tr~nsmitting and receiving radio signals. Antenna 115 servcs thc macrobase 102. Micro cell 106 has the ~nlf nn~ 116. The picocell includes in-building ~nt~,nn~c (not shown). Voice and data signals are lece;~cd by one or morc of these ~ntÁnn~c as one or more RF block signals. Some receive ch~nnf ls may bc~ledic~te-l to voice or to data or in the general case a ch~nnel may handle both voicc and data signals.
An analog FM digital block receiver, receiving the reverse link tr~nsmiccion of one of the cell service areas of FIG. 1, is shown in FIG. 2. Such a radio l~ce;~f r may be located at thc central cell control site, microcell loc~tion or thc picocell loc~tion for providing wide area, local area and in-building service. lllc reverse link signals are illtel~ e ptecl by one or more ~n~.-n~c 201 conn~lel1 to thc 30 1.,l~ ecei\dng site and applied to an analog ~locessing circuit 202. The analog pll)ces~;ng circuit 202 is operative to amplify and filter the received block ofchannel signals and heterodyne the block signal to a low inteQn~i~te frequency (IF).
The low IF is ill~oll~lt to limit the required Nyquist sampling rate used by thesubsequent Analog-to-Digital converter 205. The analog circuitry of the block 35 receiver must be highly linear to prevent interm~ n products within the bloclc signal that are strong enough to limit the dynamic range of the receiver to an ,..~
undesuable low level. The lower edge frequency of the IF block signal must be s~lffici~ntly high to avoid the folding of low frequency signal energy into the IF
block signal and thereby limit the receiver's dynamic range.
The Analog-to-Digital converter 205 con~ the IF block signal into a S digital form comrricing a .~ ti7~d time series. A single Analog-to- Digital converter is l~uu~d for each receive diversity path when real demod~ tion is used. In the event that complex de .~od~ tir~n is used, two Analog-to- Digital converters are~uil~d for each diversity path. The Analog-to- Digital converter 205 must have suffici~nt accuracy to avoid the generation of excessive qu~nti7ing noise and 10 in~l...~J..l~tion products that would limit the dynamic range of the block receiver.
The sampling rate must have a sufficiently high frequency to satisfy the Nyquistsampling criteria for the highest frequency included in the IF block signal.
The digiti7«d time series output of the Analog-to-Digital converter 205 is applied to a window function processor 207 to improve selecdvely between the 15 individual channels in the IF block signal and to enh~nre FM detection performancc.
In selecting window criteria the effect of a strong signal in one ch~nnel interfering with a weak signal in an adj~c~nt or nearby ch~nnel must be concidP~red~ An examplc of such a sih~ti~n~ providing acceptable ~lrol..lance, is shown in the graph of FIG.
3. In FIG. 3 with the mea~u~.-.ent condiPonc 304, an l-nmodul~tPcl FM carrier at an 20 IF of 900 KHz is shown having a leakage signal 301 averaged over a ch~nn~l cent~l~d at 960 KHz that is 78 dB down from the 900KHz signal. For a ch~nncl cen~.~d at 990KHz, the leakage signal 302 is down 81 db.
The windowed signals are applied to a fast Fourier transform (~
processor 209. The ~1~-1 209 processes the windowed signals to isolate in frequency 25 the individual chqnnelc within the received IF block signal. The derived FFT
pa,~ll~,t~ are ~li---i~et in size (points), span (time) and execution rate to enhancc FM det~o~tiQn by the FM detector apl,a.~tus 211.
The FM dÁtection apparatus 211 is illustratively a stored program controlled processor such as a DSP (digital signal p~ucessor). It is pro~u--l-,cd to 30 recover the FM modlll~ting signal from the FFT output power spectra derivcd f~m the IF block signal. FM detectiQn is based on the fact that the first mom~nt of thc ~-1 power spectra associated with a particular channel provides a highly accuratc esfim~t~ of the FM carrier inct~nt~neous frequency for that c~ nel~ under mod-ll~tion and ~lleas~lllent conditions to be specified herein below and explaincd 35 with reference to the flow graph of FIG. 16. Because the insl~n~neous frequcncy is directly pluyullional to the amplitude of the modnl~ting waveform, the result is a highly accurate FM detect on process.
~ lrul~tion of the first nlo~ t is pelru -l~d, in respnn~e to stored control instrucdons, or by hardwired logic circuitry, for all ch~nn~ls of interest that are include~ in the IF block signal. Calculadon of the ~ 1 is peno ii~ ~lly ex~cllt~
S over ~ucces~;~.e windows of the .~cei~c. block predetecdon digitized time series, and a post lleteclion b~seb~n-i dme series for each ch~nnel of interest is thereby pro~uce~
Several cc.n-lition~ must be s~ti~fie~ to achieve a highly accula~e and s~n~itive FM detection process. First, in order to achieve an acceptable degree of 10 post detection linearity and voice band amplitude fl~tress, detection con-litiQn~ are selected that provide an effective time span of the window pleceding the FFT
ap~z,atýs that is less than 1/4 cycle of the highest b~ebal--l m~l~ tin~ frequency present that pl~luces peak deviation conditions. Here the err~i~., dme span is the window width at which 50% of the window weightin~ is obtained.
SecQnd the window that precedes the FFT appa.~lus also must exhibit frequency domain sidelobes that are lower than the minimllm acceptable post detection signal-to-noise ratio.
Third, in order to further obtain acceptable uncQ.~l~nsated linearity in the detected signal the first moment c~l~ ul~tion must inco,~late power spectra that 20 are offset, relative to the ch~nnel center frequency, by an amount at least equal to the sum of the peak r~u~ cy deviation and the highest mod~ ting rl~uency present that ~,r~luces peak deviation conditions.
Fourth, in order to obtain a highly sensidve FM detection process, the first ..n"-ent cal~nl~tiQn must exclude power spectra that are offset, reladve to the 25 channel center Ll~uency, by an amount that exceeds the sum of the peak frequency deviatdon and the highest modlll~ting frequency present that produces peak deviation crln-litions. For enh~nce~ sensitivity the first .~-o...el-~ calculation must in ~ldi~ion dyn~mic~lly reject those power spectra that are excessively distorted by noise.
The first of the above con-lition~, defining the window effecdve time 30 span, places an upper bound on the highest baseband m(Ylnl~tion rl~uency presen that produces peak deviation condidons. For the exisdng North Alll~,.ican AnalogFM FDMA mobile telephone standard ( herein after design~ted STANDARD) this frequency is 3KHz (maximum voice band frequency). The rationale for this first con-lition is gr~phic~lly shown in the FIG. 4 which defines the effecdve window 35 ' dme span 401 and the true and measured peak value 402 and 403 of a modul~ting sinewave. In the c~lcul~tion of the first Illolllent, the inst~lln~nÁc,u~ frequency of the 20833~3 . ,.--~
FM carrier is observed through the time window. If this in~l ~n~AneouC frequencyvaries during the window span the first mom~ont calculadon is an a~pl~ox;~ iQn of the average in~!An~qn~us rr~uency. As shown in FIG. 4, if the effective window 401 is cenL~.~d at a time coll.,s~ondillg to the true peak 402 of the waveform, the S in~n~neous r ~ue~ (ordinate point) e ,t;n~A~e~ will be less than the true rl~ uency and a non- linea~ will occur. This col-lpl~s;,i~.c effect is S~ hical in both dme and h-~ nP,ous frequency deviadon and tll,.~,fol~ produces odd h ~....onic s of the illu~tl~live sine wave shown. The 1/4 cycle criterion establishes a threshold which if excPeded would cause a rapid reduction in FM detection linearity.
The graph of FIG. 5 demonstrates the rational for the second condition restricting the window sidelobes. As the FM carrier deviates in frequency in ~ nse to the modulAting signal, large deviations may cause power spectra to fall in the sidp1~be structure of the Fourier tl~c~rolllled window. If the «~uency sidelobes of the window function are too high, the derived il~ n~neouc carrier 15 rl~uel c~ may be in si~nifi~Ant error. The error will appear as a noise like ripple 501 and 502 such as is shown in the FIG. 5. The fine structure of this noise like ripple 501 and 502 is a function of the m~l~llAting waveform and the sidelobe structure of the window. In the case of the existing STANDARD, applicable to cellular radiotelephone ~y~l~ms, the first sidelobe must be n~in~Ained at least 65 dB down in 20 order to achie~e acceptable selectivity between a~ljacent ch~nnel~ as illustrated in FIG. 3. This sidelobe level is more than adequate to satisfy the second con-lition~
To further achieve a high degree of post detection linearity, the third con-lition requires that the first moment c~lculation be in accord with Carson's rule which spcÁ;~es the ,..inin~.~--- bandwidth needed for s~qtisfactory FM tr~n~miSsion The linearity errors treated by con~litionc one, two and three are partially correlated within the measured inst~nlAneoL,s frequency obtained in the first ~u..~--,n~ calc~ tion To achieve enhancecl linearity, the partial correlation ~luL~Ily can be exploited by means of an enor con~lxllcq~;on metho~l The in~t~nt~nÁou~
frequency eJl;---~t~l error as a function of the measured in~t~nt~neous frequency can 30 be ob~lned under typical modulation conditions. The esdmated error can be stored in a DSP table and used to partially correct the measured values.
To achieve a high degree of FM detection sensitivity, the fourth condition restricts the use of power spectra that fall beyond the minimllm bandwidth established by Carson's rule. For e~hqnned detection sensitivity, excessively noisy 35 power spectra that fall within this minimllm bandwidth are dyn~mi~-ally excluded from the first moment c~lcul~ions by means of a mllltip~cs power masking process.
~ 2083303 The r~tinn~le for the fourth con~litiQn and a ~cs.~ ;on of the power m~Q~in~
process is provided in the specific~tinn herein below ~t co...l.~nie~l by FIGS. 6 through 11 inclusive.
The power spectra for a 30KHz ch~nnel cente~d at an IF of 900 KHz is S shown graphically in FIG. 6 for the ll~;as~.llent con(litionQ- specified 611. Seven power spectra, 601 - 607, are aQsoci~te~1 with the ch~nnel These spectra, 601 - 607, are spaced 7.5 KHz apart and are symmetrir~l about the ch~nnel center frequency 610. This graph applies to an ul-.nsxl~ ted FM carrier. Five of the spectra, 601 -605, fall on main lobe 609 of the Fourier transformed window while two of the 10 spectra points, 606 - 607, fall at the first null points. In the case of FIG. 6 the es~"at~, of the inQ-t~n~nneous frequency is subst~nti~lly exact. Condition four is, however, violated by the two spectra, 606 - 607, in FIG. 6 that are placed 7.5 KHz outside the ch~nnel boundary. The effect of these two spectra 606 - 607, on FM
detection sensitivity and linearity is subse~luently described herein below.
FIG. 7 illustrates the same measu,~",enl con~litions 711 shown in FIG. 6 except that the FM carrier is mod~ t~l at a low rate and the in~t~nt~neous L~llency is shifted to a point 12 KHz below the center frequency. As before the power spectra 701 - 707 are fixed in frequency, but now the points have shifted along the Fourier sr~"", of the window as it follows the movement of the mrY~ t~l FM carrier.
20 Two of the spectra, 706 -707, in FIG. 7, shift into the window sidelobe structure but the error is negligible because the sidelobes are very low as s~ecificd under con~iti two. However, the first moment c~lcul~tion does exhibit a slight error because the llulllber of spectra, 701 - 707, associated with the ch~nnel have been limited to seven. Hence the first ~ ..nent calculation of the in~ eous frequency will be 25 slightly higher than the actual instantaneous frequency, but this error is so slight that the modul~ted signal may be subst~nti~lly recovered.
The rel~tion~hip between the FM detection linearity and the use of spectra beyond the FM carrier peak frequency deviation is shown graphically in the FIGS. 8 and 9 r~p,ese..~;ng in both figures inte...-o~lul~tion products, 801, 802, and 30 901, 902, resulting from a two tone test. In both FIGS. 8 and 9 the test conditions are in compliance with the first, second and third con-litions and with the conditions specifi~d 803 in FIG. 8. FIG. 9 is in compliance with con-lition four, while FIG. 8 is not. FIG. 8 shows two voice band tones and ~soci~te~ intermr ~ tion products using seven power spectra and FIG. 9 shows the voice band response using five 35 ' power spectra. It is readily appale,lt from conlp~ison of these graphs that the FM
detection linearity is highly sensitive tO the use of power spectra at or beyond the ~ 20833~3 chAnnel edges in accordance with con~itiQn three. It is also appal~.lt from FIG. 9, that when five power spectra are used in compliance with all four of the con~litinne~
acceptable teleconl~ ration~ quality linearity will be achieved.
As inflirAted herein above, with respect to the fourth con-lition, the first S ~--o~nt c~lcl)lAtion can be repeated to Con'l;l~lt~ 8 multi-pass operation. A single pass operation is ope ~ali~,e, but only when full FM detectirJn sellsilivily is not l~uil~id. In ~itllAti~n~ l~uiling a higher degree of sensitivity, the first n~o...e~-t calculation is lepe~3 one or more times. In the m~lltirle pass operation the power spectra are dynAmil~Ally tested to determine if they are excessively collupled by 10 noise or in~.rc.ence. If such is the case they are excluded from first moment calclllAtinn~.
The first pass FM detection process produces an esdmate of the FM
carrier il~sl~nl~neo~ls frequency based on the first ..-om~nt of the power spectra A~sociAted with the chAnnel If no additional passes are ~.«ol~ d this first pass15 e;.limal~ becom-~s the final estimAte For p~,l«olll.ing additional passes, a power spectra mask is generated based on the inSl~ tA~leous carrier frequency estimate of the previous pass and thc local mean power of the spectra located adjAcent in frequency to the previous estimate.
For a very low bA~eb~n-l modulAtior- frequency or very low frequency deviation conditi-nn~> the power spectra will fall on the main lobe contour 710 or in the sidelobe region as shown in the FIG. 7. This contou~ 710 is the Fourier transform of the window filn-tion Under differing modlllAtion con~litiQn~ the contour 710 may be spread in width, but will continue to follow the move,llel1t of the FM carricr is 25 accordance with the amplitude of the modulA~ing waveform.
This conloul 710 can bc simulA~ed under mAximum width or sprcading con~litiQns in a noise free en~i~n,ncnt. These con~iitiQn~ include (1) the modulA~ing ~d~efolm is a ~ine~. d~, that produces the maximum allowed peak frequency deviation, (2) the m-dlllAting waveform is a sinewave with rl~uen~;y at the 30 m;lAi,llw~ value pe ...;~;d for those signals that produce the maximum allowcd pc~c frequency deviation, and (3) the FM carrier is experiencing its ma~imu,ll ratc of change of frequency. For a sinewave modulAting signal the m~ci"lu.ll rate of changc occurs at zero crossings of the sinewave.
Under the STANDARD the peak frequency deviation is +/- 12 KHz and 35 maximum voice band frequency is 3 KHz. A parabolic-like mask 1002 in conformity with the preceding con-litions is illustrated in the FIG. 10. In the illustrative example of FIG. 10 the maximum value 1001 of the power mask 1002 is centered on the estimate 1003 of the instantaneous frequency obtained from the previous pass, and the average power of the spectra located adjacent to the estimate of the instantaneous frequency obtained from the previous pass. Four spectra, 1011 -5 1014,are used to obtain the average power and seven spectra, 1011 -1017, are associated with the channel. In the illustrative example the FM carrier to noise ratio is +5 dB; the signal is highly corrupted by noise.
An empirically derived threshold 1022 is established above the mask 1002 and individual spectra, 1016 and 1017, exceeding this value are rejected in10 recalculations of the first moment. In the illustrative example of FIG. 10, this threshold 1022 is set 13 dB above the mask 1002 in order to counter the effect of noise corruption of the mask position and to assure that rejected spectra, 1016 and 1017, are significantly corrupted by noise. The two highly corrupted spectra, 1016 and 1017, located at and just below the lower edge of the channel are rejected and do 15 not bias the second pass first moment calculation. Hence the estimate of the FM
carrier instantaneous frequency increases in accuracy at designated point 1018 and the FM detection process is now considerably more sensitive.
A typical voice band signal to noise ratio (S/N) versus IF carrier to noise ratio (C/N) characteristic is shown in the graph of FIG. 11, for conditions 20 corresponding to the STANDARD. In the seven spectra per channel example the two pass FM detection process is 5 dB more sensitive than the one pass process at C/N =
+ 10 dB. The use of two passes and five spectra, in compliance with all four conditions, results in the most sensitive FM detection process characteristic 1101 illustrated.
Post detection amplitude response, as a function voice band modulation frequency, is illustrated in the graph of FIG. 12. The response characteristics for the supervisory audio tone (SAT) and si~n~ling tone (ST) signals used in the STANDARD are shown by points, 1211 and 1212, on the curves, 1201 and 1202, corresponding to peak FM deviations at 2 KHz and 8 KHz, respectively.
The amplitude roll off of FIG. 12 is caused by the same compression effects described under conditions one, two and three. The compression is a source of non-linearity, but it also affects the amplitude of the fundamental mo~ ting frequency. As shown in FIG. 12, the effect of condition one is the (lomin~nt source of compression and amplitude roll off above a baseband modulation frequency of approximately 3 KHz. Above the 8.5 KHz threshold 1215, the 1/4 cycle rule of condition one is violated; the amplitude response is approximately 3 dB down at this ~' .
frequency. The in~lir,q~te~l mea~,u~.l.cnt con~litionQ 1205 apply to FIG. 12.
The bqQ-eban-l sa-m--pling rate is established by the ~~l exec-lhon rate.
This rate is ~lete- mi~e~1 by the Nyquist sampling criterion and must be high enough to avoid aliqsing of the bq-Q-eb~nd mod~ ting signal. Under the STANDARD, 5 ~qn-~hea~r enrodefl blank and burst FSK data messvqges are sent from the mobile to the block receiver. These mPssages have a symbol width of 50 microsecon-1s and at least one b~seb~ld sample is ~uil~,d per symbol for proper deco~ing If the ch-q-nnel sample time is phase locked to the symbol phase, an FFT exec~ltion rate of 20 KHz is s~qtiQfqrt~ry. However since the individual mobiles ~ sn~l as~nchlonously with 10 respect to each other, a sampling rate of at least 30 KHz is required and the FFT
execution rate must satisfy this re4uilc;.l-ent.
While the FFT must operate for all the diversity paths only the selecte~
path must be processed at the full execution rate required to decode the ~f~nchester encoded data mess~ges. For non-selected diversity paths, the ~ l execution rate 15 must only satisfy the need to n~aaurt receive signal strength amplitude or quality at a rate conQiQtent with airwave telecon ~ r~tion system fading statistics. For the STANDARD a non- selected diversity path FFT execution rate of less than S KHz isadequate and the ~ l processing load can be sized accordingly.
The STANDARD uses receive signal strength as the basis for diversity 20 path selection. However, with multi-pass FM detection, a channel quality mea~u~ ent can be established based on the percentage of spectra rejected by thepower mask. Rejection statistics can be used to estimate the c~l~n~e1 C/N
characterisdcs as shown in FIG. 15 for the mea~u~e,llent conditionQ- 1501 in~ic~t~A
In this figure, the power mask rejection st~tiQtics for a threshold of + 7 25 dB are shown for the two mod~ tion conditions that bound the accuracy of the C/N
estimqt~. At a given C/N, rejected spectra are minimi7e~ as per characteristic 1511, for an ~ ...od~ tell FM carrier. Rejected spectra, as per characteristic 1512, are m~cimi7ed when the FM carrier is modulated at the highest mo~ tiQn frequency that produces the maximum allowed peak deviation and at an amplitude that 30 produces the peak deviation. For the conditions of FIG. 15 the C/N can be estimated to an accuracy of appr~ ely +/- 2 dB in the low C/N region of interest.
The measure of c~nnel quality can be used as the basis for selecting the best diversity path based on the most favorable CIN ratio rather than carrier strength alone. This method will result in better diversity selection performance under co-35 channel or adjacent channel interference con-lition~ and will create a new p~ameter to assist in cellular mobile handoff decisions.
In ~d-lition, the mea~.,.ent of ch~nn~l C/N in the se1Prte~l di~
path can be used to mitigate the il~lclr~"~.lce or impulsive noise that is heard during mnltir~th prop~g~ti~n in~luced Rayleigh fading con~1ition~. The use of e~
C/N is benefici~l when the FM carrier drops to a level near or below the level of S channel noise or ulle~r~ ce. For Rayleigh fades that do not fall as far as the noise or int~,rwe.lce floor, impulsive noise may be induce~l as the result of a 180 degree phase rotation that typically occurs at the cusp or null of the fade. The 180 degree phase rotation produces impulsive noise with a m~gnitude that is highly correlated with the first derivative of the FM carrier amplitude. The inst~ neous carrier 10 amplitude is n~a~ed at the FFT execution rate as a part of the mllltip~es FM
detection process 1608 (as shown in FIG. 16). A combination of spectra rejectionst~tictil~s and carrier amplitude first derivative mea~u,.,.nents can be used to identify when impulsive noise is present and ~ ,~«o-~ when mitig~tion metho l~ should be applied. Mitig~tion can consist of a full or partial muting during the normally brief 15 periods when il.t~,r."~;nce or impulsive noise is present. As an ~ltern~tive to muting, a low level of subjec~ively pleasing white background noise may be inserted during the critical part of the fade.
The equation for computing the first mom~nt of the windowed power spectra is;
~ (Fo + OFFSETk) P~c INST FREQ EST = k= 1 k= N
~ Pk k=
P k = Ik + Qk and where:
k is an index number for individual spectra;
N is the total number of power spectra ~soci~ed with a specific channel;
Pk is a specific power spectra 20833~3 ".
Ik is a real compol1ent of the Fourier coeffirient;
Qk is the ima~n~r cGIll~onent of the Fourier coefficient;
Fo is the l.,re,~.-ce frequency for the first ...u...- n( calrulatiQn;
Offsetk is the dirL~nce bel~.een the r ~uellcy of Pk and Fo~
In general the process of the first .. ~.. e.~ calculadon is e.-co.. l~csed by the following process steps. (1) Initially form all the Pk = I2k + Q2k values for k = 1 to N. This l~Ui~S 2N muldples and N adds. (2) Form all Fo + Offsetk values for k= 1 to N. This requires N adds and subtracts. (3) Form all (Fo + Offsetk) Pk values -for k = 1 to N. This requires N muldples. (4) Form the dividend ~u~ ;Qn This 0 lC~ui~S N - 1 adds or subtracts. (5) Form the divisor ~un..~ ion This requires N - 1 adds. (6) Form the quotient using 1 divide. These pl~cesses comprise 4N - 2 adds or subtracts, 3N muldples and one division.
In order to enh~nce the process efficiency of the first n.o~n~ t calculation the selected power spectra, 1301 -1305, or 1311 - 1314, may be specially 15 placed with respect to the channel edge frequency boundaries 1321 and 1322. The power spectra, as shown in the diagram of FIG. 13, are evenly spaced and placed sylll.lle~ically with respect to the center frequency of the ch~nnel If N is odd a power spectra 1303 is placed at the channel center frequency 1323. In addition, the power spectra placement relative to the channel boundaries is the same for all 20 channels of interest in the received block signal.
For the STANDARD, channels are of equal width and all ch~nnel frequency centers are evenly spaced from one another. The ~-l produces evenly spaced power spectra; the desired spectra spacing is established by the ratio of the Analog-to-Digital sample rate to the FFT size in points. The pl~t~en~Ánt of power 25 spectra that is the same for all channels is ~ccomrlished for these con-lition~ if the ratio of ch~nl-~l center spacing to power spectra sp~cing is an integer. The pl~l~ement of power spectra ~yll~n~-h;~lly with respect to the ch~nnel center is .q,Áco...rlished under the prece~ing con~litio~s by selecting a block receiver IF such that any channel center 1323 coincides with a power spectra 1303 (N odd) or is exactly between two 30 spectra, 1312 and 1313, (N even).
The first moment calculation compu~ti~n~l efficiency is ~ignific~ntly enh~nced for the prece~ing arrangement of spectra and ch~nnel~ . The ca~ tiQn process collll,.ises the steps of; (1) Form all Pk = I2k + Q2k for k = 1 to N. This requires 2N multiplies and N adds. (2) Because the power spectra are unifolmly 35 spaced in frequency, the Fo + Offsetk telms need not be com~llt~ Fo is set at the channel center and Fo + Offsetk is norm~li7e~1 by the unifo~lll spacing to a ~083303 ", ., predeten..ined integer. (no c~lc-llAtion needed for this step) (3) The (Integerk)Pk product is replaced by adds or subtracts in the next step. (4) The dividend ~ .. Al;on is formed by adding the power spectra co,..,;,ponding to norm~li7e~ offsets +/- 1 once; offsets +/- 2 twice, offsets +/- 3 thrice, etc. for all power spectra ~ssoci~ted 5 with the ch~nn~l This .~-lihes the following adds or subtracts (N2 5)/4; for N odd; N >= 3 (N2 -2)/2; for N even; N>= 2 (S) Form the divisor ~un..nAIiOn This .~lui,~s N - 1 adds or subtracts.
and finally (6) form the quotient using 1 divide.
For N odd this totals;
(N + 8N - 9)/4 adds or subtracts, 2N multiples and 1 divide.
For N even this totals;
(N + 4N - 4)/2 adds or subtracts, 2N multiples and 1 divide. The advantages achieved in co..lplJlational efflciency are ~ull~ ed in the table of FIG.
The overall methodology or flow process for the FM detection process is illustrated in the flow diagram of FIG. 16. In block 1601 the ch~nnel of interest is selected and in subsequent block 1602 the first moment calculation is ~.«o..,.cd in accord with the above description. In the block 1603 the estimate of instantaneous 20 fic~uency is limited to the maximum value known to have been trAn~mitte~ Thc nu~llb~,, of passes completed is co"~p~ed, as per block 1604, to the number of passcs required. The nu~lber required may be fixed or may adaptively be a function of thc recent history of the estim~te~l C/N as described with reference to FIG. 15. If onc or more additional passes are required, the process flow proceeds to block 1606, and the 25 l~uil~d mask processing is initiated. if the number of passes required is satisfied the in~ t~ eous frequency is stored in l.,e,.,o..y, as per block 1605, and the ncxt channel to be processed is selected.
For the mask p.ocessing of block 1606, an index is computed for thc spectra based on the preceding estimate of the inst~nt~neQus frequency. This indc 30 is used to access a table of mask and threshold values that have been p.~sto.~,d in a nlu.y device. A count of remaining spectra is m~int~ined in block 1607 and instrucdons of 1606 are invoked repeatedly until all spectra ~soci~te-l with thechannel have been plucessed. A local mean power value is col-l~uted in block 160~
for those spectra nearest the preceding estimate of the in~t~nt~neous frequency. Thc 35 sum of the mask and threshold values relative to the local mean power is compu(ed for each spectra. In block 1609, each spectra is cO,l,~ d with the sum of the mask and threshold values as adjusted in accord with the local mean power. Spectra are marked as accepted or rejected is accordance with this comparison in the step ofblock 1609. The process flow proceeds to block 1602 and the process is continued.
Following the FM detection step, the diversity path with the best quality 5 is selected, in selection circuitry 213, based on an estimate of the carrier to noise ratio for each diversity path. The C/N is estimated from the power spectra rejection rate experienced as a part of the multipass detection process. For purposes of establishing the highest quality diversity path, the power mask threshold need not be set at the same level as used for optimum FM detection.
Following diversity selection, the digitized time series highest quality path for channels of interest in the block is subjected to the expansion step associated with channel companding, in the expand - de- emphasis circuitry 215. The expanded signal is conventionally processed, in a detection circuit 217, to detect the individual baseband modulating signals that may be present. Voice processing, in the de-15 emphasis circuitry 215 and the detection circuitry 217, includes digital low pass filtering and decimation to shape the voice band amplitude response in accordance with the de-emphasis function and to remove signal and noise content above the voice band. For the STANDARD, conventional signal processing techniques are used to identify the specific SAT frequencies that may be present, and to identify the 20 presence or absence of ST and decode Manchester encoded data messages. Because of the asynchronous characteristic of these encoded data messages, over sampling is required and interpolation of two samples that may fall within a symbol will be necessary.
The information content of decoded data messages, SAT, ST as well as 25 digitized voice is form~tte(l for transmission by means of a digital data link interface 219 to a control location such as a cell site or macrobase 221. The interface will be modular and capable of supporting metallic, fiber, microwave or a combination ofthese transmission media. For cases where the block receiver is located at a cell site or macrobase, the interface is directly established with the digital bus structure of the 30 site.
|International Classification||H04B7/24, H04Q7/38, H04B7/26, H04B1/26, H04J3/16|
|Cooperative Classification||H04B7/2621, H04B1/0003, H04B1/001, H04B1/26, H04B1/0017|
|European Classification||H04B1/00D2C, H04B1/00D, H04B1/26, H04B7/26F, H04B1/00D2F|