【正文】 3. Notice the absence of overshoot or ringing.Figure 3: Gausssian filter impluse response for BT = and BT = Figure 3 depicts the impulse response of a Gaussian filter for BT = and . BT is related to the filter’s 3dB BW and data rate byHence, for a data rate of kbps and a BT of , the filter’s 3dB cutoff frequency is 2880Hz.Still referring to Figure 3, notice that a bit is spread over approximately 3 bit periods for BT= and two bit periods for BT=. This gives rise to a phenomena called intersymbol interference (ISI). For BT= adjacent symbols or bits will interfere with each other more than for BT=. GMSK with BT=_ is equivalent to MSK. In other words, MSK does not intentionally introduce ISI. Greater ISI allows the spectrum to be more pact, making demodulation more difficult. Hence, spectral pactness is the primary tradeoff in going from MSK to Gaussian premodulation filtered MSK. Figure 4 displays the normalized spectral densities for MSK and GMSK. Notice the reduced sidelobe energy for GMSK. Utlimately, this means channel spacing can be tighter for GMSK when pared to MSK for the same adjacent channel interference.Figure 4: Spectral density for MSK and GMSKPerformance MeasurementsThe performance of a GMSK modem is generally quantified by measurement of the signaltonoise ratio (SNR) versus BER. SNR is related to Eb/N0 byWhereS=signal powerR=data rate in bits per second=noise power spectral density(watts/Hz)=energy per bitRecent StandardsGMSK has been adopted by many wireless data munication protocols. Two of the systems specifying GMSK modulation are Cellular Digital Packet Data (CDPD) and Mobitex.CDPD uses the dead air time on cellular systems by sending data packets on idle cellular voice is transmitted at using a BT of . This high data rate is facilitated by the 30kHz channel spacing of the cellular network and the spectral conservation of GMSK. Voice has priority over data and will interrupt data transmission, forcing the CDPD system to seek a new idle cellular channel. This could prove to be an obstacle to the throughput promised by its data rate when implemented in a highly congested area where dead time is limited.CDPD is being added to the existing cellular infrastructure and therefore promises to offer widespread coverage. The coverage and ease of adaptation appear to be the greatest strengths of the CDPD system. The slowerthanexpected deployment of CDPD has many people anxious and perhaps a bit nervous about its potential.Competition from dedicated data systems such as Mobitex is not insignificant. While Mobitex has a lower data rate than CDPD (8kbps), it is not sharing its channels with cellular voice transmissions. Several subtleties such as this will make it more difficult for end users to select the system best suited to their needs by obscuring the actual throughput potential of the systems. Mobitex’s choice of 8kbps and a BT of afford it a much tighter channel spacing () than CDPD, but the greater intersymbol interference for BT= limits the system’s tolerance to noise and distortion. The narrower channel also limits Mobitex39。s tolerance to frequency offsets between units.Both CDPD and Mobitex employ forward error correction in their packetting of data. Figure 5 shows the typical packet structures of these two systems for parison. Forward error correction (FEC) helps improve the systems39。 throughput when less than ideal channel conditions exist.Figure 5: Typical packet structures for CDPD and MobitexImplementation ConsiderationsThe design of a GMSK modulator/demodulator appears to be a straightforward task. Most textbooks present the modulator as a “simple” Gaussian filter cascaded with a VCO. However, in practice it is generally not that simple. Many of the sections in a typical radio such as the synthesizer, IF filter, power amplfier, etc. have far from ideal behavior. In particular, the synthesizer presents a unique problem for GMSK modulation. Data patterns consisting of several consectutive ones or zeros have a spectral response extending down to near DC. Most frequency synthesizers will not respond to this low frequency signal (a typical synthesizer effectively has a highpass filter characteristic).Two of the most mon modulation methods, which help considerably where the nonideal behavior of the synthesizer is concerned, are Twopoint modulation and Quadrature modulation.Two point modulationTwo point modulation (see Figure 5) circumvents this synthesizer problem by splitting the Gaussian filtered signal。 one portion is directed to the VCO modulation input, the other portion is used to modulate the TCXO is not in the frequency control feedback loop. Hence, the TCXO can be modulated by the low frequency portion of the signal, and its output is effectively summed with the signal modulating the VCO in the synthesizer. The posite signal has a spectral response extending down to DC.Figure 6: Two point modulation radio block diagramI and Q modulationQuadrature (I and Q) modulation can also be effective in eliminating synthesizer shortings. In I and Q modulation, the Gaussian filtered data signal is separated into inphase (I) and quadrature phase (Q) ponents. The modulated RF signal is created by mixing the I and Q ponents up to the frequency of the RF carrier, where they are summed together. The role of the synthesizer has now been reduced to merely changing carrier frequency for channel selection. The key to optimum performance with quadrature modulation is accurate creation of the I and Q ponents.Figure 7: I and Q radio block diagramBaseband I and Q signals can be created by using an allpass phase shifting network. This network must maintain a 90 degree phase relationship between the I and Q signals for all frequencies in the band of interest.DemodulationDemodulation of the GMSK signal requires as much attention to the preserv