PA: 27Mhz FM TX Chain Design

Project Begin @2014/7/05

 I love design and soldering.

Build a VCO

@2014/7/08

I had tried the super VXO using a overtone 27.095Mhz crystal, but only 2 KHz deviation could be pulled. and if use frequency doubler/tripler schema, circuit will grow big. 

selecting vackar VCO,  attracting by it's stability. but it proved,  such a high frequency, a roughly design and roughly construction, vackar oscillator still drift, and unsuitable for  serious usage.  but for this toy, it's almost good, after warm up, it will stick on one channel for a long time. (test by my handy transceiver, ICOM-91A )

following is initialize design, the tank inductor use a 10S FCZ style IF can,  pin 4 to pin 6.   

refer to the schematic,  NOTE: this Vackar oscillator is not optimized yet,  bad bias. on this schematic the buffer bias is wrong, there should had a resistor from collector to base. 

BJT: 2sc9018  AM/FM,IF,LO,VHF tuner Vceo=15V,Ic=50mA, Pc=400mW,Hfe=100(1mA), Ft=800Mhz,NF=4dB(STS9018), Pgain=15dB

 Varactor: BB910, 40p at 1V                        

Vackar VCO construction and tune

final construction changed both design and layout, the finished VCO + buffer look like almost a little monster,  with high density component compress in small cube(1cmX1cmX1.5cm ).  

this layout style is very suitable for quick construction and handy device. 

final schematic changed a lot, including frequency tuning, buffer  design. 

Frequency tune：

       * finally testing proven the inductor is  smaller than expectation,  experimentally adding a 22p capacitor.

       * the 10P with  *  mark is not installed.

       * with 1Vpp Modulation signal, deviation is about +- 5Khz. 

buffer change:

       * the final output amplitude about 1.2Vpp

       * with 1k emitter degeneration resistor, the bias current is far away to driven a 50R load

NOTE: the 6.8V zender actually is 5.6V , not fixing in this schematic. (6.8V is too close to supply 7.2V)

Over All:

            * oscillator is acceptable for experimental purposes project

            * Vackar bias is un-optimized.

            * FM deviation almost OK

            * Output 1.2Vpp, 27.450Mhz (unloaded).

            * Buffer could not driven a 50R load.

this is the simple part of this project, but still few things worthy to note:

* the 47k resistor on the electret mic might too low, 68k might better for 7.2V VCC

* 100R  emitter degeneration resistor is added to reduce gain.

* 47k resistor direct connect to VCO's varactor, modulating the VCO and  provide correct DC bias for varactor

* 900k, 3.3k arrange for 9014 bias the collector at  half of supply voltage [ Bias: favorite BJT/JFET bias guide]

before building final PA stage,  several  design problem must be resolved.

            a)  VCO output 1.2Vpp but unable driven a Class C PA, 1.2Vpp not tested on 50R load, it's unloaded output.

            b)  Class A exciter amplifier needed boosting signal to at least 10mW (1Vpeak) to 5 mW(~12dBm) for Driven Class C PA.

 @2015/7/11 item b this is not totally correct,  the criteria is (include impedance matching circuit to BJT base) driven a 2-5ohm resistor loading could  get at least 0.7V peak voltage, and this is just begin make Class C PA working, how much power you get related to the BJT's characteristic, good BJT good power output, much driven signal, much power output.

different type LO  oscillator given different output level,  a powerful oscillator, like pierce oscillator, might strong enough to driven a class C PA directly (extract signal from transformer). this vackar unable driven PA directly.

most of time, the minimal driven level of a Class C amplifier might not  bothering you, but for QRP device, you might want use Class C to get just  hundreds mW output,  could it be driven via another Class C, then form a driven chain?   Class C amplifier need driven signal amplitude at least about  0.7Vpeak to 0.9Vpeak ,  to Driven a  low input impedance Class C device( might  only few ohm),   typical Input impedance match network is deployed, ie.  4:1 impedance ratio ,  that's to say,  voltage transform ratio 2:1, if so, driven level should be 1.4~1.8V peak voltage.

but wait a moment does this mean if amplifier cloud delivery to 50ohm about 13dBm~15dBm power, and  then everything  will be fine? unfortunately i think the answer is no, let discuss this again when design the final stage PA .  let's assume 12dBm to 15dBm is ok for now.

A lazy man design amplifier from a transformer

the pre amplifier stage should boosting signal to around 15dBm. a typical Class A amplifier might have 10-20dB gain, but this depending on design, and transistor selection.  

            VCC = 7.2V ( 2 x Lipo)

            BJT:  2sc9018 Vceo=15V,Ic=50mA, Pc=400mW,Hfe=100(1mA), Ft=800Mhz,NF=4dB(STS9018), Pgain=15dB

            output transformer:   2.45:1

            broadband transformer reactance: 400 ohm

            collector load: 300 ohm

Now estimate max output power,  <EMRFD> or anywhere introduce PA design there is the formula,  Pmax= (Vcc-Vb)/2*Rload,   sm0vpo suggest  bias the Vb at 25% of supply voltage, but our VCC is limited  so select Vb as 1.2V at 1/5 of the VCC,  then max output power is:  (7.2-1.2)*(7.2-1.2)/ (2*300) = 60mW,  this is enough.  this does not mean we will get the 60mW output, just say, it can produced 60mW max power,  actually output power depending on the driven level and the gain, what not easy to estimate.

         Selecting Vb 1.2V

        300 ohm load,  15mW power ====> the RMS current should be Irms=sqrt(Pout / Rload) = 7mA

        peak value is 7*1.414=9mA

this mean  the signal current swing will  vary from -9mA to +9mA,  we should ensure all the time, BJT current is flowing,  keep it in Class A, so choose a little larger one, say, 10mA for bias.  The hfe of 9018 about 100 at 1mA, i guess it will reduce to 80 when current increased to 10mA, so bias pair should at least have 10*(10mA/80) ~= 1.25mA flowing through,  and bias at 20% of supply, then get 3.3k/680R as voltage divider,  use 40R as emitter degeneration resistor to get 10mA bias current. [  add a parallel capacitor  to 40R will increase the gain, decreases linearity ]

First  of all,  i stuck on a  wrong testing method  for a while,  then see a chaos  waveform,  the voltage value measured is also not correctly, actually, very bad, see  photo,  this is the best waveform  i ever get ,  because use long clip connect to output of transformer, then a Coax cable to 50R load.

compare to next frustrate problem i encounter, this is just nothing.  measured output voltage of 50R load is 1V peak. ( load is close attach to oscope, use the bad long clip hook up to IF CAN).   but the output voltage swing directly from the output transformer pin is  about 18Vpp. this is not just wrong testing method, the constructed circuit had something wrong!

after long time debugging,  i notice  the IF CAN actually soldering into board  is not the one i design to use, this IF CAN marked with 053 , but  i designed to use is the 10BC.

 instead of build another amplifier, I  just wind 4T on secondary of the IF CAN(pin 4 to pin 6) to get it work. now the ratio of P:S is around 3:1.

             Change IF CAN turns ratio to 3:1

then the max output power change to  (7.2-1.2)*(7.2-1.2)/(2*50*9)=40mW

this stage gain is hard to estimate,  collector see a load about 450 ohm now, after fix the IF CAN, we had a 33 ohm degenerate resistor (which actually soldering, is 33R not the designed 40R),  considerate the small signal model  re,  which about 2.6 ohm (26/10mA),  seems the voltage gain about : 450/42.6 = 10, it's 20*log(11)= 22dB voltage gain. let's check the real hw: 

an interesting thing is checking the DC voltage swing of collector,  Oscope setting is  0.2V per div, 10:1 probe. also, left one is unloaded, right is loaded.   loading swing is about 8V peak to peak observed on BJT's collector,  unloading swing reach to 13Vpp, negative swing reached almost bias point of  the BJT(base junction).

the final result : 50ohm load connected

                    driven signal 1V peak to peak

                    collector swing: 8V peak to peak

                    output (50 ohm load):  2.5V peak to peak

so the voltage gain actually get about:  20*log(8V/1V) = 18dB, the actually turns ratio of IF CAN is around 8V/2.5V= 3.2:1

estimate result is 20dB, actually get 18dB, it's not very bad  -:). 

[ not considering the transformer's Q, which will result  reduce the gain, and the 2.5uH primary as RFC is not reach 4*450 ohm impedance, as suggest by EMRFD.]

class A amplifier's  input impedance is complex than what i think.  first thing is the small signal bjt model not very suitable for amplifier work at 27Mhz and had some Power output.

small signal model assumption:

• the collector current is approximately equal to the emitter current (i.e.,β>>1)

• the nonideality factor n is equal to one, and

• room temperature operation;

this amplifier's input impedance is reduce when frequency increase, and same time β dropped.  β will not >>1. by short time surfing WWW, found the β v.s  fT  is really not very simple. here not intend dig into too much, EMRFD chapter 2 talk about this  a little bit.  

let's play with  ltspice estimate the output impedance:

result: ( refer to later measurement by scope on real hw)

1.  estimated by  β, Zin= 140 ohm

2.  estimated by  insert resistor, result : 220 ohm

3. estimated by AC analysis the Vin/Iin, result 120 ohm

to include the bias resistor, 680ohm,  from 1) get 140 ohm, so Zin total is : 140*680/(140+680) =~116, almost same as 3). 

one important thing is β decreased very quick, even a few Ghz Ft BJT also drop to almost only few when reach to 100Mhz. small signal modeled this by a input capacitance, cπ.

       take the decreased β into account  input impedance still estimateable by small signal model.

beyond HF, depending BJT you use, the β almost 1 for some BJT  (this is might why some time you assume input impedance is just the Re, but ignore the  cπ reactance)

LTspice Model of 2sc9018:

.MODEL SS9018G NPN (Is=3f xcjc=0.1 Xti=3 Vaf=75 Bf=102 Ne=2 Ise=0.8p Ikf=250m Xtb=1.5 Br=0.3 Nc=2 Isc=0.8p Ikr=10m Rc=3.5 Cjc=1.8p Mjc=.15 Vjc=.8 Fc=.5 Cje=1.5p Mje=0.5 Vje=.8 Tr=2n Tf=110p Itf=13m Vtf=1.7 Xtf=3 Rb=60 Vceo=15 Icrating=50m mfg=Fairchild)

.model 2SC9018 ako: SS9018G NPN Vceo=15 Icrating=50m

calculated input impedance high than LT model, i believe this major because this is |Z|, instead of real part of Z.  this 300 ohm include the 680 bias, which indicate the bjt input impedance |Z| might be 600ohm. the input impedance measurement method is not very accurate, this is not VNA.

Final thought

        *  Lt spice  does help you get some idea about what the real circuit will be

        *  input impedance estimation by LT spice or by insert a resistor before amplifier,  just get a |Z| value, the real part of Z (complex impedance) could still very small, don't neglect the angle of the Z, in the result of simulated by LTspice.

        *  the real part of |Z|, will pull very high volume peak current from driven stage.

        *  SS9018 beta drop to about 10 (estimate value by small signal model) at 30Mhz.

Class C amplifier need driven signal amplitude at least about  0.7Vpeak to 0.9Vpeak ,  to Driven a  low input impedance Class C device( might  only few ohm),   typical Input impedance match network is deployed, ie.  4:1 impedance ratio ,  that's to say,  voltage transform ratio 2:1, if so, driven level should be 1.4~1.8V peak voltage.

question is,  does this mean only if drive stage could delivery to 50ohm about 13dBm~15dBm power, and  then everything  will be fine?  unfortunately i think the answer is no.  the 1.4V~1.8V peak's really mean is when connect the drive stage to PA,  the BJT base conjunction should get at least 0.7V peak voltage driven level ,  experimentally,  0.7V is not enough for some BJT, suggest try from 0.9V.  The simulation let you get ideas about the driven level requirements,  under 0.9V driven level, 2N2222 given x2 times voltage gain,  x3@1.xV driven level.

in circuit, unless you use signal generator, you can not ensure the final stage GET calculated drive level. for example want driven level 0.9Vpeak, if pre amplifier output just is 0.9Vpeak to 50ohm load,  this still can not ensure final stage get 0.9V driven level. because final stage input impedance is hard to certainly estimated, PA might pull more or little power from pre amplifier compare to  a 50ohm load, real result even related to more early stage in the chain.

Continue the smoking design,  drive stage output 1.25Vpeak, 12dBm powr, is this enough?  i think this is depending on arrangement of the PA.  if some impedance match network added to lower driven signal's source impedance, i.e, 2:1 turns ratio,  driven signal definitely insufficient.

Just check what we get. 

 50R load: 5.5Vpeak, 25.2dBm, 330mW

driven level: around 1.2Vpeak

Rbias=200ohm

 Rbias, based on experimental,  this resistor provide negative swing of signal a reasonable return path load,  at the positive swing of driven signal this resistor set a bias impedance loading, flatting the BJT input impedance which is fairly nonlinear. and same time bias the BJT to zero voltage.  EMRFD note this is wide band load, and give a stable impedance to driven stage.

during adjust Rbias, from tens ohm to 200 ohm the driven signal level will vary from tens mV to 1.4Vpeak, out power begin to increase from 0.8V driven level , which  is identical with the LTspice simulation result.  What if no Rbias? experiment show the final PA might be in an unstable state, there are low frequency parasitic oscillation.

while adjust Rbias,  driven amplifier will see a vary load impedance, from very low increased gradually.  Rbias  provide a stable impedance, but lower than BJT's input impedance while it is not turn on,  so present a low load to driven stage, result a low gain (around Rc/Re),  the output power from pre amplifier  is then un-sufficient to driven final Class C amplifier.  while Rbias reached 200 ohm the driven signal is enough to produce a 6Vpeak output to 50 ohm load, and fairly stable. 

I don't think this mean the Class C input impedance is not low, just because too small Rbias stick the input impedance to a stable low impedance,   reduce the amplifier's gain while BJT not turn on, result  BJT never could be turn on under low Rbias. 

while PA generated 360mW (6Vpeak) output power,  transmitter pull  additional 100mA current, at 7.7V, this is about  770mW. only generate 360mW power, this mean the efficiency only 50%,  not a so bad PA, but neither efficiency .

under 50% efficiency,  the 2N2222  160mW power,  need a small heat sink, otherwise it's will became heat and unstable, pull more current from driven stage, leading to a chaos response.   a widely used solution by ham is parallel two 2N2222,  i did, also add heat sink, and glue the 2 2N2222 together. another benefit of doing this is PA now became easy driven, only 100 ohm Rbias is enough to driven PA to output 360mW power.

the reason of low efficiency is unmatched loading V.S the output impedance.  360mW power under 7.2V Vcc, perfect loading should be  RL=VCC2/2*Pout = 72ohm,  the output low pass filter should alter to match this.

Final thinking

@2014/7/13

* Oscillator output need evaluated before design, which i skipped. 

* driven stage  RFC is not suitable, it had  too small XL.

* Final PA output impedance mismatch, result low efficiency.

*boosting 20~50mW to around 300mW, use Class B PA might a good choice, which require no 0.9V driven level limitation.

even though this design had many pitfall,  I'm still very happy to get this design working, the PA output power exceed my expectation actually.  from VCO to PA, the whole system need arrange properly to get good result. 

by this little project, many aspect of Class C PA been thinking carefully,  being experimentally understand.

PA to Antenna 

this small article became too long and this topic worth another one, PA: TX chain to Antenna.