IF filters are placed at 5.0 MHz and have nearly vertical skirts. The RF ary vertical skirts and 5.5- 10.0 MHz, nearly a l The RF bandwidth is Ist nanow enough(4.5 MHz) that no two signals can exist within the passband to cause mixing. The image frequencies would lie in the range 15.5-20.0 MHz and are also outside the filter range. In a practical de- of the would be an even smaller portion of the theoretical 1-octave bandwidth 100105 150 1.1.5 Couversion Gain and noise Since any practical circuit will generate excess noise, miters are no exception. Each type of mixer ciruit will therefore have its own noise fig If the combination of this noise figure and any bosses in the RF filte ahead of the mixer are low enough, no amplification is required nor even desired before mixing. Amplifiers would inevitably have some nonlinear ies and would increase the signal input level to the mixer so that other mixing products could appear. When considering a mixer noise figure should be remembered that two frequencies could contribute to noise out put at the IF frequency, the desired frequency and its image. removal of by placing a filte ways sufficient. If the mixer sees" a resistive impedance at the image frequency, thermal noise will be added. The filter should therefore appear as a short circuit at the image frequency nne 7
gain must be set the overall noise figure to the feedback in this mowe the linearity and the resulting loss of gain will actually be well came mxer may produce an IF signal that is either higher or lower in amplitude than the rF input signal. The relative size would be indicated by consideration,as any deficiency can always be made up for with other amplifiers. The only real problem involves setting the overall noise figure of the receiver. With no RF stage and a mixer with a conversion loss, the frt stage of the IF must have a very low noise figure 1.1.6 RF Amplifiers Once the desired range of input signal levels to the mixer has be chosen, the RF amplifier can be designed (or eliminated )as re Its gain should be just sufficient: to bring the weakest signal from the tenna up to whatever level is needed to overide noise generatad in the amplifier and mixer. The total noise factor of the receiver will be givea by F=F1+ F2-1 where, Fi=noise factor of the first stage F2=noise factor of the seoond stage G=power gain fram input to seoond. stage For the arrangement shown in Fig 1.4, the RF amplifier has a ga of 12 dB (8: 1)and the mixer has a loas of 4 dB (0.398: 1). The noise figure of the RF stage is 2.0 dB(1. 585)and of the IF amplifier is 2.5 dB(1.778). The overall noise figure would then be 1.829(or262dB) 8
The RF amplifier has therefore provided enough gain so that the overall noise figure of 2.62 dB is only 0. 62 db higher than that of the RF amplifier itself. Higher gain would provide little overall improvement and would simply cause more problems with the mixer Gaipl2 dB Loss4 dR (1.585 E. 1. 4 Nose o the first three stages of a rainer After the gain and noise figure are set, the nent requirement is the filtering associated with the RF amplifiers. Part of this will depend on the ber amd part on the amplifier itself. The minimum filter requirements needed to complement the perfect square-law mixer were discussed in Sec tion 1. 1. 4. If the mixer also has a significant third -order compoment, sev eral new frequencies could and up in the If passbend. These would be (1)IF= RF+ 20SC, RF= 16 or 26 MHz (2)IF= 2RF oSC, RF= 2.75 or 7.75 MHz (3)=3RF,RF=1.66MHz The examples shown alongside assume an IFof 5.0 MHz and the lo- cal oscillator at 10.5 MHz, a situation taken from Fig. 1. 3. The first new frequencies(1)at 16 MHz and 26 MHz would be outside the passband of the minimum 1-octave filters(5.5-10 MHz )and, also, if balanced mixers are used, the mixer would not function at even harmonics of the oscillator. This frequency then does not present any problem. The second pair(2)represents signals at 2. 75 or 7.75 MHz. The latter lies right in the middle of the 1-octave filter range, so that if the mixer has significant third-order distortion, added fltering would be needed. The one choice is alf-octave filters, the other is narrow-band, continuously tunable filters (with their tracking problems). Other spurious signals can be created by monics produced within the RF amplifier itself; such is the case with quency(3)where a third hanmonic created by amplifier nonlinearities
could pass straight through the mixer. The 1.667-MHz input can easily be eliminated with flter ahead of the rF amplifier. The total spurious frequency Roblem therefore depends to a great tent on the linearity of the RF stage, on filters before and after this stage and on the mixer itself, The big problem involves gain control. To main the mrer input, the gain of the RF stage may need to be reduced at some time. For automatic gain control, the amplifi er must have a nonlinear transfer characteristic so that a change of bias produces a change in gain. The resulting secand-order nonlinearities could then produce spurious signals, which would case mixing products to appear within the IF passband, particularly since the filters ahead of the RF stage are usually minimal. If AGC is used, good RF filtering is a better approach is to make the amplifiers very linear, even by go. ing to the extremes of balanced amplifiers with negative feedback. Gain control can then be manual-either tuming a potentiometer or switching in resistive pads, or automatic if linear devices are used The idea of a linear, two-terminal device that will not distort a sign et can change its resistance with a voltage change may, seem strange. A small incandescent lamp is one example. If a Dc voltage is applied across the lamp and slowly changed, the curent flowing into the bulb will not change linearly with the applied voltage. As the filament heats up, its re- sistance will increase. Any rapid voltage changes, however, will cause linear cument changes, since the thermal time constant of the filament will be long enough to hold the has been used for automatic level control of good-quality audio for years. The moden equivalent of the lamp is the pin diode frequencies the device acts like a diode, but at higher frequence like a variable resistor, since the lifetime of its charge carriers is quite long(up to 500 ns). Above about 10 MH( depending on the particular diode, a linear attenuator can be made that can be varied with a Dc control voltage
1.1.7 IF Amplifiers The intermediate frequency section of the fnst mixer and the final detector circuits. It must: (1)Provide a high amount of gain, 60-100 dB, and reduce en strong signals are present 2)Filter out all unwanted signals outside the passband (3)Limit amplitude variations in the case of FM signals, thereby determining the FM capture ratio (4)Limit the amplitude of noise pulses in the case of AM and med without destroying the noise figure set by the receivers fromt end and without introducing distortion products within the desired passband For the majority of receivers, a total gain of at least 20 dB will exist in the RF and mixer stages, go the IF noise figure is ususlly bot signifi cant.For the few cases where no rF stage is used and the mixer operates with a coversion los, the IF noise figure will be very important. Any Losses in the F filters ahead of the amplifying stages must be considered; for if an RF gain of 15 dB, a mixer loss of 6 dB, and an IF filter loss of 9 dB occur, the IF sigmal level will be right back to where it was at the anteana teminals. The IF noise figure would then be very important Attention to noise figure itself is not sufficient, as the total noise bandwidth must also be considered. One part of this has already been pointed out; the image frequency fom the mixer will add thermal noise in addition to the possibility of interfering signals. The total noise bandwidth could then be twice as wide as the IF flter bandwidth. The other noise problem can occur whenever separate filters and amplifiers are used 如面1b:邮如幽打如