transistor frequency multipliers - vhf … multipliers.pdfusing an active device such as a...

1.Introduction

In a previous article we looked at the useof diode multipliers. These all sufferfrom one serious disadvantage, the con-version loss when multiplying to highorders can be quite high. This makesthem only useful for low orders of multi-plication of perhaps 3 as a practicalmaximum.Using an active device such as a transis-tor can overcome the inherent loss of thesimple diode multiplier, because the tran-sistor can have appreciable gain whenused as a multiplier, which eases thedesign of multiplier chains where highmultiplication is required. If the transis-tor is operated in grounded emitter con-figuration with the input RF drive appliedto the base the impedance is medium anda greater RF voltage swing can be ob-tained from relatively low drive power.

2.Principle of Operation

If a transistor is used as an amplifier for a

signal requiring linear amplification thenit is limited to the output power it canproduce because of the maximum voltageswing on the base or collector. If theinput voltage swing exceeds the base-emitter forward voltage the junctionstarts to behave like a rectifier diode anddistortion occurs. We can simplisticallythink of a multiplier transistor as beingtwo separate items. The first is the diodejunction contained within the base-emit-ter and the second is a very linear voltageamplifier between the base and collector.By overdriving the base-emitter junctionwe generate harmonics of the input fre-quency and these are then amplified in ahighly linear manner by the collector-emitter portion of the transistor.Unfortunately it isn’t as simple as this,but the simplistic idea forms the basis ofall transistor frequency multipliers. Wecould place a multiplier diode across thebase-emitter and couple into the transis-tor base via a capacitor. By generatingharmonics in the external diode these arethen amplified by the linear biased tran-sistor. Some commercial two-way radiomultiplier chains used this technique.Unfortunately the efficiency this achievesis worse than the diode on its own if weinclude the DC power being drawn bythe transistors before and after the diodemultiplier.By utilising the inherent transistor base-emitter diode junction we have a better

John Fielding, ZS5JF

Transistor FrequencyMultipliers

VHF COMMUNICATIONS 4/2006

224

method of generating high levels ofharmonic power. Of course we still haveto filter out all the unwanted harmonicsat the collector and select the desired onewith as little loss as possible. However,by altering the conduction angle of thetransistor we have a method of enhancingcertain harmonics and hence reducing theunwanted ones.The most basic circuit is shown in Fig 1.Although this will work it does have anumber of problems, which we willconsider later.The drive signal to be multiplied isapplied to the base of TR1 via C1. Theinductor L1 across the base to groundholds the base at DC ground and lookslike a high impedance to the input fre-quency. The emitter is also directlygrounded, so for the transistor to conductthe RF voltage across L1 needs to exceedthe base-emitter junction voltage of about0.6V. This means that when the RFvoltage is below the threshold nothinghappens, but as soon as the base voltageexceeds about +0.6V on the positivepeaks of the RF waveform the transistorwill start to turn on. If the RF drive israised even more the transistor will beconducting for a greater portion of theRF envelope and so more collector cur-

rent will flow. The transistor is beingoperated as a Class C amplifier and thewaveform at the collector is similar to asquare wave, which is rich in harmoniccontent. All that is required is a suitableband pass filter to select the wantedharmonic and reject the remainder.On the face of it this seems to be perfect.But one thing to be aware of is that themultiplier is very critical to the RF drivelevel, too little and it will snap off andtoo much and it will saturate and the gainwill fall dramatically. Also the thresholdis very dependent on the base-emitterforward voltage. If the ambient tempera-ture varies over a wide range the base-emitter voltage does the same. At verylow temperatures the voltage required tomake the base-emitter junction conductwill be greater than when the ambienttemperature is high. So although itworks fine in the shack it may not workat all from a contest site on top of amountain in winter!The danger with this simple circuit is thatthe base-emitter junction can be drivenvery negative on the opposite half cycleand small signal bipolar junction transis-tors have a limited reverse base-emitterbreak down voltage, as little as 2V candestroy the junction.

Fig 1: Basicmultiplier circuit.


225

Another failing with this simple circuit isthat the impedance presented to the driv-ing stage varies wildly as the drive levelvaries. Below the threshold voltage themultiplier appears close to an open cir-cuit, at high drive levels the multiplierappears as low impedance, just like thediode multipliers.

3.Improving the Basic Circuit

The first problem we will address is thetemperature sensitivity. In a linear am-plifier using the common emitter con-figuration we normally include a resistorin series with the emitter to ground toprovide a small bias voltage. This helpsto hold the collector current more stablewith variations in temperature.The second problem is the sensitivity tothe RF drive. We can to some extentreduce this tendency by applying a smallamount of forward bias to the base-emitter junction with a potential dividerof two resistors. By selecting the resistorvalues such that the base-emitter voltage

is just below the conduction point the RFdrive voltage swing is greatly reduced.This translates into less drive powerbeing required and higher input imped-ance. Fig 2 shows the modificationsneeded.The value of the emitter resistor is deter-mined by the order of multiplicationrequired. For high multiplication factorsthe resistor value is higher than for loworder factors. This is to reduce theconduction angle to enhance the higherharmonics. In simple terms we need tomove the conduction more towards ClassC as the harmonic order increases.The value of the base potential dividerinter-reacts with the emitter resistor. Es-sentially what we need to achieve is theDC voltage on the base with no RF driveto be just below the conduction point or avery small collector current flowing, afew hundred μA typically. With thevalues shown in Fig 2 the base voltagewith respect to ground is approx. 0.62V.But the collector current flowing in theemitter resistor raises the emitter voltageabove ground. For a collector current of20mA flowing when the base is driventhe emitter will be approx. 0.2V aboveground, making the effective base-emit-ter voltage about 0.42V. Hence the input

Fig 2: Improvedfrequencymultiplier.


226

RF drive needs to be greater than about0.3V peak to turn the transistor on.The disadvantage of this circuit is theemitter decoupling capacitor; this needsto be a low reactance to not only theinput frequency but also all harmonics.This requires quite a low inductancecapacitor and a SMD chip type is nor-mally the best choice. If leaded capaci-tors are used, for example ceramic disks,the emitter decoupling may not be goodenough for the higher harmonics andsome loss of gain will occur at the higherharmonic products.

4.Coupling Out Of the Collectorthe Required Harmonic

This is where most published designs getit wrong. Although the multiplier stageappears to work satisfactorily often rear-

ranging the circuit a little can make adistinct improvement.In Fig 1 and 2 the collector is connectedto a radio frequency choke with a widebandwidth. The coupling capacitorserves to block the DC supply and alsotransfers the AC signal to the followingband pass filter. In Fig 3 is shown thewrong way of coupling out the harmonicsignal.The collector of TR1 is tapped down theinductor L1. L1 is resonated to therequired frequency by the variable ca-pacitor C5 and this tuned circuit couldthe first portion of a multi-pole band passfilter.This tapped inductor method is com-monly used for amplifiers such as inter-mediate frequency amplifiers where lin-earity is critical. The tapping of theinductor allows a good impedance matchto the collector, the conjugate match.However, in a frequency multiplier this isexactly the opposite of what we require.To understand this logic we need to lookmore closely at how the bipolar junction

Fig 3: Incorrectcollector coupling.


227

transistor works and its internal structure.Fig 4 shows the two major parasiticcapacitances inherent in the bipolar junc-tion transistor. These are the collector tobase (CCB) and collector to emitter(CCE). There is one other which existsbetween the base and emitter but thisdoes not affect the operation, only theinput matching.Normally in a small signal linear ampli-fier application the capacitance is rela-tively constant and can be tuned out.Because the RF voltages appearingacross the junctions are small the capaci-tances are constant. But in a frequencymultiplier with correct collector matchingthe RF voltages are not small, in fact wewant to make them as large as possible,without exceeding the collector-emittersecondary breakdown voltage of the de-vice. The reason is that with a largevoltage swing the parasitic capacitance isnot constant but behaves like a varactordiode. Motorola in an Engineering Bul-letin [1] gave details of how the varactordiode characteristic can be exploited tomake more efficient frequency multipli-ers. This used the MRF629 a 12V/2.5WUHF power transistor as a parametricmultiplier. This exhibited a power gainof over 6dB when driven by a 100mWsignal at 150MHz and developed asmuch as 700mW output at 450MHz

when configured as a tripler. The writerused this technique when building the1152MHz LO chain for a 2C39A power-mixer SSB transverter for 23cm andobtained nearly 1W output at 384MHz.This used 150mW input at 96MHz withthe circuit configured as a 4x multiplier.The use of idler circuits at the input andoutput is the secret to efficient multipli-cation.This technique is not just applicable tothe MRF629 (which today is obsolete)but even small signal BJT RF transistorexhibit this effect when a sufficientlylarge RF voltage is developed at thecollector. Hence, we need to match thecollector to a much higher impedancethan suggested by the formula for conju-gate matching. This also explains whyjunction field effect transistors (JFETs)are poor frequency multipliers, as they donot exhibit the same varactor effect asBJTs. Fig 5 shows the inherent varactordiodes in BJTIn Fig 6 is shown an example of thecorrect collector matching for a fre-quency multiplier. This uses a two-section top-coupled band pass filter toselect the required harmonic. The Q ofthe tuning components needs to be suffi-ciently high to develop the required RFvoltage swing.

Fig 4: Transistor parasitic capacity. Fig 5: Inherent varactor diodes in aBJT.


228

The input to the multiplier is via acapacitor with a medium impedance atthe operating frequency. The drivesource is a low impedance stage such asan emitter follower and with an outputvoltage swing of at least 1V p-p. Thetuned circuit of L1 and C3 selects therequired harmonic. This develops thenecessary high impedance, and hence thehigher voltage swing, at the collector of

TR1. The coupling into the next sectionof the filter is made via C5, a very lowvalue capacitor to prevent loading of thecollector. This capacitor is also the topcoupling for the band pass filter so itneeds to be a very low value, which suitsthe circuit configuration well. The sec-ond inductor L2 is resonated by C6, thecoupling to the next stage is performedby C7 which is also a low value capacitorbecause we need to prevent loading onthe band pass filter, which would alterthe selectivity response if excessive load-ing occurs. The values of the inductorsand capacitors need to be chosen to suitthe frequency. If more selectivity isrequired it is often better to incorporatethis into the following buffer amplifierbefore the next multiplier stage.Fig 7 shows the typical spectrum at theoutput of the band pass filter. Thiscircuit is the one shown in Fig 6 and usesan MPSH-10 transistor as a tripler. Theinput frequency is 50MHz at a level of5mW in 50Ω and the wanted output at150MHz measures approx. +13dBm(20mW). The worst spurious products

Gig 6: Typicalfrequencymultiplier stage.

Fig 7: Typical spectrum for a BJTmultiplier.


229

are about 50dB down on the 150MHzsignal. This will require further filteringin another band pass filter before it isapplied to the next multiplier or used bythe equipment.One thing not obvious is why a bipolarjunction transistor is the preferred device.Some schematics show junction Fets asmultipliers. This is a poor choice be-cause the Fet does not contain the sametype of varactor junction as the bipolar.Hence, the drive power required is muchgreater to get any significant harmonicgeneration. Fets make good mixersbecause they are used in a “additivemode” with a second carrier, the localoscillator. The local oscillator energycauses the drain-source current to swingup into the saturation region and the non-linearity is greatest under this condition.The Fet is operated as a saturated switchto “chop” the input waveform and causeadditive and subtractive products, one ofwhich is the wanted mixing product. If ajunction Fet and a similar die size bipolarjunction transistor are compared in amultiplier circuit the bipolar will delivermore harmonic output power and requireless drive power. Hence, junction Fetsare not very suitable for straight multi-plier duty.

5.Multiplier Stability

In general, frequency multipliers usingBJTs are inherently very stable becausethe input and output are tuned to differentfrequencies. Unlike a tuned amplifierwhere the input and output are the same,the only thing that can make a frequencymultiplier unstable is either poor layoutor the internal parametric components ofthe transistor. If the drive level ismarginal the transistor can exhibit erraticoperation. If the input drive varies overtoo great a range the multiplier can snapon or off as the level varies. Hence, it isnecessary to have some reserve drive inhand. A good minimum is about 3dBabove that required to ensure reliablemultiplication. When tuning up a multiplier stage thepower output should vary smoothly asthe input and output are aligned. If themultiplier snaps on and off as the tuningis altered it is usually a sign that the inputdrive level is marginal. A good check ofthis is to measure the base voltage of theBJT with full input drive. If the drive isadequate the base voltage should be avoltage slightly negative with respect toground. A value of -0.25V to -0.75V isnormally about right. When the inputdrive level is more than required the toppotential divider resistor in the base biaschain can sometimes be omitted and thedrive develops a negative base voltageacross the base to ground resistor. Incases such as this it is prudent to connecta reverse biased diode across the base toground to clamp the base reverse voltageto less than 0.7V to protect the base-emitter junction.Fig 8 shows the spectrum for a multiplierstage where the input drive level hasbeen deliberately reduced to show theeffect that instability causes.

Fig 8: Unstable multiplier spectrum.


230

6.Optimum Supply Voltage

In all the years of designing frequencymultipliers using BJTs it has been foundthat there is a certain range of supplyvoltage where the optimum performanceoccurs. It is not quite clear why this is sobut is probably to do with the inherentparametric varactor diodes in the BJT.However, from many experiments theoptimum supply voltage on the collector-emitter junction is about 7 to 9V, the socalled “sweet-spot”. With voltages farfrom these limits the efficiency suffers,with too low a voltage the power outputfalls away and with too high a voltage theefficiency also falls off. Hence, whendesigning a multiplier chain using BJTs itis often best to use a supply voltage of 8to 9V. This is convenient as a 3-terminalregulator can be supplied by the nominal12V-vehicle supply.If the drive level into the next stage is toohigh a simple cure is to starve the inter-stage buffer amplifier transistor orMMIC following the multiplier by in-

creasing the collector feed resistor valueto limit the collector voltage.

7.Choice of Transistor Type

As strange as it may seem we do notneed to use a “super-transistor” for anefficient multiplier. Often the choice of atransistor is dictated by the frequency ofoperation and hence the ft as an amplifierneeds to be quite a bit higher than theoperating frequency. This however is fora transistor operated as a linear amplifier,when used as a highly non-linear multi-plier we can often get by with less exoticdevices.Form experiments made on a wide selec-tion of transistor types it has becomeapparent that even lowly devices such asthe 2N2222 with an ft of only 100MHzwork perfectly well up to a few hundredMHz when operated as a doubler ortripler stage. For the frequencies higherthan this a move to a popular type suchas the MPSH-10 or the BFR92A or

Fig 9: Test pointadded to multiplierstage.


231

BFR93A in the SMD SOT-23 packagework well even as high as 2GHz, whichthe manufacturers data does not suggest.Of far more importance is the drive leveland the filtering components used toselect the wanted harmonic. If the filterhas excessive loss then the advantage ofusing an exotic transistor is wasted.

8.Test Points

When aligning a multiplier chain it ishelpful to be able to break into the chainat critical places to attach test equipmentto see what is going on. Some designersmake all the multiplier stages to termi-

nate in 50Ω so that normal test equip-ment can be connected. This not onlyintroduces extra loss but also adds unnec-essary components to the circuit. We cansafely assume that the base input imped-ance of a typical BJT when driven at theoptimum level is about 200Ω to 500Ω.Therefore the drive power is not going tobe too high; perhaps +10 to +13dBm in50Ω when transformed up will givesufficient base-emitter voltage swing.Probing a high impedance portion of thecircuit with a RF AC voltmeter oftencauses detuning because of the probecapacity. A simpler method is to incor-porate some test points where a 50Ωinstrument can be attached withoutbreaking the chain. Often two resistorscan achieve this and the 50Ω equipmentcan be attached as and when requiredwithout upsetting the system. This con-

Fig 10: Push-Pulldoubler.


232

cept is illustrated in Fig 9. The tworesistors are a high value one to tap intothe high impedance point and a secondtermination resistor to suit the test equip-ment. A coaxial cable can be connectedand taken to a spectrum analyser to seethe spectrum. The amplitude will not becorrect but the spectrum should still bethe same.

9.Using MMIC Amplifiers asMultipliers

There have been several applicationnotes and other articles published con-cerning the use of wide band microwave

monolithic integrated circuit amplifiers(MMICs) as frequency multipliers. Thechoice of devices is somewhat limitedbut several writers have reported goodresults. However, because of the internalbiasing components of typical MMICsthese devices do not really lend them-selves to this application. Whereas it istrue that any transistor can be over drivento force it into a non-linear operatingportion this seems not so easy with thetypical MMIC. As the typical MMIC isdesigned to interface with 50Ω termina-tions intuitively we can see that wecannot run the collector into the highimpedance necessary to induce the varac-tor diode effect.Where MMICs can be made to work iswith an external shunt diode across theinput, with DC blocking capacitors toprevent the bias being upset, to generate

Fig 11: Push-Pushtripler.


233

the harmonic energy and then to use theMMIC as a conventional amplifier toraise the level. However, when you runthe calculations of the total power re-quired the efficiency still ends up beingquite low and a cheaper BJT will prob-ably out perform it at far less cost.MMICs are good as inter-stage bufferamplifiers but not so good as multipliers.

10.Special Multipliers

Where we need additional suppression ofcertain harmonics we can utilise some ofthe ideas found in diode multipliers. Fig10 shows a push-pull doubler using twoBJTs. The circuit suppresses the oddorder harmonics by about 30dB andleaves just the even order harmonics.The variable resistor VR1 allows thecircuit to be balanced by adjusting theemitter currents of TR1 and TR2. Thisprovides the best null to the unwantedproducts. For optimum performance adual-transistor in a single package offersthe best results.The wideband transformers T1 and T2

need to be wound for the best balanceand this limits the higher frequency per-formance.This circuit can also be modified to forma push-push tripler where the predomi-nant harmonics are the odd order ones.This is shown in Fig 11.

11.Multiplication Order

In many cases we need to multiply by afixed number and this is clearly definedby the starting and finishing frequencies.For example a receive-converter for23cm may require a 24x-multiplier chainto arrive at the final LO injection fre-quency. A 24x-multiplier can be ar-ranged in several different ways. Weneed to multiply by 24, which can bebroken down into doublers, triplers orquadruplers (2x, 3x or 4x). Choosingwhich multiplier factor to use and wherein the chain is often not simple. As ageneral rule it is often better to make thefirst multiplier stage with the greatestmultiplication factor, so in this case wemight choose the first multiplier stage tobe a 4x. Against this must be weighedthe difficulty of filtering out the harmon-ics close to the wanted multiplier prod-uct.As an example, consider the 23cmreceive-converter for an IF of 144MHz.The starting frequency is 24x less thanthe LO injection, which in this case is1152MHz. This determines the crystalfrequency as being 48MHz. This is agood choice as the 3x product is 144MHzand if a test port is provided on the firstmultiplier then a small amount of the144MHz signal can be used to set theband edge on the 144MHz tunable IF. Ifthe 144MHz receiver is accurately cali-brated in frequency readout the receivercan be used to zero beat the crystal

Fig 12: Wide band sweep of LOsignal.


234

oscillator exactly onto 144MHz. Hence,in this case the first multiplier needs to bea tripler. However, consider the case where thefiltering of the first multiplier is insuffi-cient to reject all the 48MHz spacedproducts. If these are allowed to passthrough the subsequent multipliers thenwe will end up with 48MHz productsspaced either side of the final LO fre-quency. The 1152MHz product is not ofconcern but the image at 1008MHz(288MHz away) will cause degradationin the receiver performance.Hence, it is better to first double the48MHz to 96MHz, band pass filter theresult as well as possible to reduce the48MHz products to negligible levels, andthen proceed from there. No integermultiplication factor using 96MHz canproduce a signal at 1008MHz so this is abetter plan.Assuming we choose the first multiplieras a doubler, this leaves a total of 12xfurther multiplication to arrive at the finalLO injection. Filtering at 1152MHz willrequire a fairly good filter and to ease thedesign of the filter we need to place theunwanted harmonic products as far awayas possible. We could choose the finalmultiplier to be a doubler and hence theinjection frequency into the final multi-plier would then be 576MHz. If thischoice is made then we need to assess ifit is possible to arrive at 576MHz from96MHz. The answer is we can bymultiplying by 6, this would be a 2x anda 3x multiplier in either order.If we chose the more traditional route ofmultiplying by 3 from 384MHz to arriveat 1152MHz then we need a total multi-plication of 4x from 96MHz; this couldbe 2x & 2x in two separate stages or a 4xin one stage. Again we need to assess thedifficulty of making a band pass filter at384MHz that can reject the 96MHzspaced products with sufficient attenua-tion. The safest method would be thedual 2x approach as the filtering of a192MHz spaced product with a 384MHz

band pass filter is likely to be easier.This also simplifies the filter design andit is likely to have less insertion loss thanthe narrower filter previously required.In some cases commercial helical filterscan be obtained aligned for the requiredfrequency which allows a more compactdesign. Companies such as Toko andTemwell list suitable filters in theirstandard product range that occupy verylittle board area and these have adequateattenuation to the unwanted products.Unfortunately not all amateurs have ac-cess to a spectrum analyser to ascertain ifthe multiplier chain filtering is adequate.Many constructors have little more than aRF diode probe and perhaps an absorp-tion wave-meter or milli-watt meter toalign the multiplier chain. Although itmay appear to work satisfactorily it isonly when it is examined on a spectrumanalyser that the truth emerges, some-times causing a big surprise! Unwantedproducts at the final LO of a receive-converter only cause the operator a prob-lem, but if the same dirty LO is also usedto mix up in a transmit converter then weare potentially causing spectral pollutionto other users. This is not only irrespon-sible behaviour but probably in contra-vention of our license conditions. Thesespurious signals may interfere with sensi-tive receivers used by other services,perhaps radio astronomy, civil or militarycommunication services and should beavoided for obvious reasons!One factor which is misunderstood bymost amateurs is how clean the LOsignal needs to be. If the signal is beingused to drive a mixer to down-convert orup-convert then the harmonic content ofthe final signal is not important, in factgiven the correct phase relationship a bitof 2nd and 3rd harmonic content can bebeneficial to the following mixer. Thefollowing mixer will regenerate the har-monics of the input signal so there is noneed to suppress the 2nd and higherharmonics of the LO signal. What is offar greater importance is to eliminate asfar as possible the lower multiplier prod-


235

ucts especially the ones closest to thewanted LO frequency. To illustrate the problem take a look atFig 12. This is the tripler circuit shownin Fig 7 but with the spectrum analysersweep increased to cover up to 1GHz.Although the products close to thewanted 150MHz are acceptable the farout products around 700MHz are verybad. They are barely 30dB down and aredue to a poor layout where the harmonicenergy was able to hop over the filter andget picked up in the output amplifierstage. A contributory factor is the topol-ogy of the band pass filter; it is a top-coupled design and hence tends to actlike a high pass filter to the higherharmonics. If the spectrum analysersweep had not been opened up we wouldbe blissfully unaware of how bad thespectrum was. With the writer’s spec-trum analyser covering from 10kHz to12.4GHz it is possible to see significantlevels of harmonic energy all the way upto nearly 3GHz. This is with a transistor“officially spec’ed” only to an ft of650MHz.If significant levels of spurious signals atthe LO exist then these are potentialinterference generators for frequenciesfar removed from the wanted channelfrequency. Most microwave receiverfront ends do not have very high selectiv-ity because of the nature of LNA design;few LNAs have anything like the re-quired rejection characteristics becausethis would destroy the low noise figurerequired. If the unwanted signal is ofsufficient amplitude to get past the frontend filtering and into the mixer thenthese will appear as “birdies” or “sprogs”when the receiver is tuned across theband. If you are really unlucky then youmay find that your weak signal DXchannel is completely obliterated by astrong signal far from the wanted fre-quency. Running some simple mixingspurious calculations can highlight poten-tial problem areas. The choice of the 1st IF and hence the

LO frequency can have a dramatic effecton the image and other spurious responsecharacteristics. If the 23cm band is usedas an example then the choice of 28MHzas the 1st IF is not a good choice. Forthis we need a LO at 1268MHz and theimage will then be at 1268-28MHz =1240MHz which is right in the middle ofthe aeronautical DME and radio altimeterband worldwide! Typical radio altim-eters are downward looking radar sys-tems fitted to most aircraft and theseradiate up to 10kW of pulsed power.DME systems are used for distancemeasuring and the aircraft radar interro-gates a land-based transponder, this thensends a reply pulse delayed by a knownamount. By measuring the time delay theaircraft can calculate the exact distance tothe ground transponder. A typical air-craft DME transmitter is about 300W andthe ground station anything up to 10kW.The writer recently had to do a signifi-cant system revision for a radio telescopereceiver for the Hydrogen Line at1420MHz that is under development. Itwas originally proposed to use a1280MHz 1st LO and the 1st IF was at140MHz. Only recently it became appar-ent that a significant amount of L-bandmicrowave signals on or near the imagefrequency of 1140MHz from aircraftDME transponders existed with high sig-nal levels caused severe interference andthis forced a change in 1st LO and 1st IF.The 1420MHz receiver has only 40dB ofimage rejection because of the nature ofthe LNA and front-end filter design toobtain the required sensitivity. The mini-mum discernible signal of this receiverwith the LNA noise figure of 0.18dB isless than –146dBm, which is about 20dBbetter than a good 23cm receive con-verter. In this case it was found thatchanging from a “Low-Side” injectionscheme to a “High-Side” injection placedthe image frequency at about 1700MHzwhich is also a protected portion ofspectrum reserved for SETI research andthe 1st IF then occurred at 152MHzwhich is another protected piece of spec-


236

trum for Radio Astronomy. The 1st LOthen occurred at approx. 1572MHz.

12.Special Local OscillatorTechniques

During the upgrade to the writer’s 23cmtransverter an unusual approach wasused. The original design used a 5thovertone crystal at 96MHz, but this suf-fered from excessive warm-up drift andother problems. Because the desire wasto make as little change as possible it wasdecided to make a new 96MHz oscillatormodule that then fed into the existingmodules. This new module used a12MHz TCXO, which was multiplied upto 96MHz, a factor of 8x. The firstmultiplier used a 2x and the second a 4x.After filtering and amplification the96MHz signal was fed to the originaloscillator multiplier chain. A test port onthe second multiplier provided a smallamount of 144MHz harmonic energy tocalibrate the 144MHz transceiver.Because the first multiplier output spec-trum contained products spaced 12MHzand 24MHz either side of the wantedfrequency the filtering required was quitesevere. If these products are not removedat the source then it practically impossi-ble to eliminate them further down themultiplier chain. The aim was to attenu-ate these close-in products by at least70dB, in the end a figure of over 80dBwas achieved by careful attention to theearly band pass filters. The 96MHz finaloutput was clean; the worst spuriousproduct was -86dBc, which was consid-ered sufficient. Filtering, shielding andsupply line bypassing achieved a com-pact and clean module suitable to drivethe existing multiplier chain. One of theother problems with the 96MHz crystaloscillator was its susceptibility to strong

RF fields when transmitting 250W on23cm. This caused FM when using SSBor a chirp when CW was used. Withadequate shielding and supply line by-passing these problems disappeared withthe new oscillator module.It must be appreciated that the filteringinductors need to be well shielded, thebest option is pre-wound canned coils orhelical filters such as those made byToko. If air wound inductors are usedthe coils radiate spurious energy and thiscan hop over shielding plates and effec-tively bypass the filter. The RF voltageat the collector of a multiplier stage canbe many volts, and these make good“mini-transmitters” unless the RF field iscontained in shielded compartments.Grounding, layout and shielding are allvital parts of any oscillator chain designand construction.

13.References

[1] EB-70A “Frequency multiplicationsimplified by internal faraday shields inMRF629”, Motorola Semiconductors


237

transistor frequency multipliers - vhf … multipliers.pdfusing an active device such as a...

Documents