PRACTICAL SYNTHESISER COMPONENTS
by Harry Lythall

Introduction

Since I decided to start these synthesiser projects I have been doing quite a few experiments, now it is time to document my results and share with you the practical results of my efforts. I have been basically confirming my theoretical knowledge, and finding a few solutions to my own lack of knowledge. I must, however, point out that the project presented here is the worst case I could think of: a single-modulus synthesizer covering the 88MHz to 108MHz domestic band. Here the modulation can be as low as 10Hz and as high as 50KHz. The complete band is 20MHz (20%) wide so the VCO must have a very high voltage to frequency "gain". The transmitter deviation is also quite high (+/-75KHz).

The practical circuits given in this page can be used in conjunction with the all CMOS simple synthesiser project given in my homepages. If you are seeking for a more stable bugging device then forget this project. The bug would be VERY stable, but has limited concealment potential.

The Oscillator

The Voltage Controlled Oscillator used in synthesisers is, by definition, an unstable oscillator, but we must tame that instability. You may use just about any oscillator you like, as long as the tuned circuit is referenced to an RF ground (the battery or supply rails). The power supply should ALWAYS be stabilised and well decoupled with a few nano-Farads and a few micro-farads, physically connected as close as possible to the oscillating device. Coils should not be allowed to hang loose, this avoids "microphony". A screening tin over the VCO will also help to reduce other sources of disturbances, such as heat and light (diodes with glass bodies are very sensitive to light).

The basic oscillator I have chosen is a modified (inverted) Hartley. This circuit seems to work well from 100KHz to almost 200MHz by suitable choice of L and C. I have made the coil on the PCB to improve mechanical stability.

As you can see, there is nothing particularly clever about the oscillator. It has become one of my most recent standard building blocks and uses a coil etched on the PCB. The transistor can be anything that functions at these sort of frequencies, BSX19 works well, but I have thousands of BC547s in stock and they work well. The coil should be tapped at about 10% to 20% for the emitter of the oscillator transistor. Tapped at about 5% to 10% for the output. Output is about 1mW. I will be posting the PCB foil pattern, but like all things these days, it takes time.

If you wish to cut out all the construction involved with the VCO then you can buy ready-made VCO modules from just about any frequency from a few KHz to several GHz. The cost only starts from about $150 or so.

Novel VCO

Varicap diodes are not readily available here in Sweden (you have to make some allowances for some of these backward countries) so I have been experimenting with other forms of frequency control. One novel method of varying the frequency of the oscillator (to convert it into a VCO) is to use an ordinary switching diode. This works fine if the oscillations are quite strong and are several volts peak-to-peak at the top of the tuned circuit.

As the RF voltage on the top of the LC varies, the diode conducts and switches IN the capacitor 2C, but it only switches in for a part of the RF waveform. The part of the waveform is determined by the DC bias you stuff onto the diode. The drive source should be a low impedance since this method also behaves as a rectifier and charges the capacitor with the peak level of the RF voltage. This needs to be sunk to get a wide frequency control.

If the DC bias (control voltage) is higher then the peak RF waveform then the diode will always be reverse-biased. The diode then acts as a normal "varicap" (Variable Capacitance) diode. This method can give up to a 2:1 frequency ratio (eg. 75MHz to 150MHz). 2C is typically twice the value of the tuning capacitor maximum. If you make it any largerthen it will damp the oscillator and stop it from oscillating before reaching the lower frequencies.

One of the disadvantages of this circuit is that frequency control at low DC voltages also affects oscillator amplitude. But it works, and it is cheap (my kind of circuit).

Conventional VCO

A conventional VCO uses a varicap diode to vary the capacitance across the LC circuit of the oscillator. In this circuit, the VC does little to cause rectification of the RF signal, so the input impedance is quite high. All you will be driving is a reverse-biased diode. This can therefore be fed directly from a high-impedance filter.

C/2 is typically half the value of the tuning capacitance, but this is normally selected to give you the tuning range you want. The tuned circuit capacitor C may be totally removed and C/2 increased in value to give you a very wide range VCO, although this does give rise to other problems and peculiar effects. All-in-all, this is about the simplest and most practical way of using a varicap diode. It is worthwhile noting that the voltage to frequency transfer caracteristics are NOT linear.

A normal varicap diode will need roughly double the DC voltage to half the capacitance. This is not linear so the frequency as a function of the control voltage will not be linear. There are "linear" varicaps on the market, but in reality is is hardly worth the effort and expense. When you therefore calculate the filter, V/F sensitivity should be that part of the characteristic giving the highest frequency change. Luckily, as the capacitance reduces accross the tuned circuit, you only need smaller capacitance changes to give the same frequency change, so this effect does give some form of linearity compensation.

(Improved Waveform) Conventional VCO

If you are making a wide-range VCO that is to be used in a radio transmitter, then it is important to make the VCO as clean as possible. The basic problem with using a single varicap is that the RF waveform also changes the varicap's capacitance during the cycle. This gives rise to distortion, especially with wide-range VCOs. The sine-waveform begins to look more like a sawtooth-waveform, causing high even harmonics. One solution is to use two diodes and balancing resistors.

The RF waveform still affects the capacitance, but by using two varicaps it affects each diode in the opposite manner - two wrongs making a right! The two unmarked resistors would be typically 1-megohm each. Reducing them would give a better ballance, but would significantly reduce the load on the preceeding filter stage. You should also have a small capacitor between the cathode of the top diode and ground; typically 5C.

Modulating A Synthesiser - Basic

In all the synthesiser projects I have shown so-far, all I have shown is how to generate a signal. This is OK if you are only interested in morse code, but most people are interested in speech (or music) - how unreasonable! You can simply stuff your audio signal into the varicap control voltage, but this leads to other problems. If you want it to work on a fixed frequency then there is little problem, as long as the control voltage is around the same level.

Here you see the amplifer/audio source driving the modulator, the output has DC blocking with the 1uf electrolytic cap and a 100K series resistor to prevent loading of the RF on the tuned circuit.

Unfortunately the 1uf also adds to the loop-filter time-constant and can even make the whole synthesiser loop go hunting up and down the band in search of the elusive "PLL lock" condition. All it will succeed in doing is continuously over correcting. Anyone who has ever seen a novice fly a new model aeroplane (or flown with Dan-Air airlines to Mallorca) will know the effect.

Modulating A Synthesiser - Improved

Another problem with a simple varicap modulator is that the capacitance is not linear with the voltage. Audio modulation is usually of such a low amplitude that the minute part of the diode's used looks very close to linear, but we also have a control voltage to the same diode. That is what "cocks it all up".

When the control voltage is low then small voltages give a large capacitance change (high deviation level), but if the control voltage is high then the deviation will fall off dramatically. A modulation change of 10:1 is typical, so it would not be possible to predict the modulation depth at any given control voltage setting. What we need is a diode operated at a fixed point in it's characteristic curve, so we need yet another varicap diode.

Here we have two varicaps. R2 and VC1 are the usual control voltage from the synthesiser's loop filter, but R1 and VC2 are added for modulation. The diode VC2 is biased at a fixed point, in this case by the output of the AF amplifier that does not have a DC blocking capacitor. VC2 will therefore always be operated at the same point in it's caracteristic and will give a reasonably constant capacitance change under all conditions. Note that since modulation frequency variations are a lot smaller than the control voltage frequency variations, the capacitor C/10 is typically only 10% of the tuning capacitor, probably even a lot less than this. C/10 should therefore be a high tolerance type.

The standing DC on the modulating diode cathode should be rather high, typically 8 to 10 volts where the curve of the capacitance is a little closer to linear. The AF signal should also be kept low so that only a small part of the waveform is used. If not then severe distortion could occur.

Here we see the effect of modulating at too high a signal level. Remember that it is all just a compromise, since too low a signal voltage would reduce the signal to noise ratio. If you intend broadcasting rap music then don't worry, no-one will ever notice either noise or distortion.

Loop Filters

There are many books written around loop filters and you could make a career just by understanding the various forms of electronic filters. At this point I must state that PLL filtering is a big hole in my knowledge, but I have tried to approach the subject objectively and developed the following description that seems to work, and works well (for me).

I am using a CD4046 PLL phase detector and have an output that is normally open-circuit (Tri-State) with occasional blips to +5v or Gnd. These blips I filtered using a short time-constant to restrict them to less than 0.25v DC change on the capacitor C1 under normal "in lock" conditions (from practical observations with an oscilloscope). This voltage I chose because it is less than that required to make a silicon diode conduct.

Now I use a "blooming long" time-constant, R2/C2. This time-constant was chosen to be 5 times longer than the period time of the lowest modulating frequency (10Hz). If this time-constant were shorter then the modulation itself would become an error voltage and the PLL would remove it! R2a is the time-constant resistor, but R2b allows a little ripple to pass, thus damping the loop. This damping is needed to stop the whole loop from oscillating.

The two diodes short out the 100K time-constant if the loop needs to change by a large amount, for example, when changing frequency. In this event the R1/(C1+C2) is the new shortened time-constant. If the filter only allowed the VCO to change by 150KHz every second, then it would take the synthesiser a couple of minutes to retune from 88MHz to 108MHz. With the "speed-up" diodes it will take only about 3 seconds to change from one end of the band to the other (20MHz worst case). This is quite reasonable for a broadcast band transmitter.

R1/C1, R2/C2 will not have removed the reference frequency ripple from the loop control voltage. R1/C1 is too short and R2/C2 has that blooming damping in circuit! The filtering so-far will allow the complete synthesiser to lock on to the desired frequency, BUT with a reference frequency of 1KHz there will be a lot of 1KHz modulation on the VCO. This 1KHz is loud and unacceptable, so I have added R3/C3 and R4/C4 to filter out the 1KHz ripple I had on my VHF synthesiser. It worked fine, but if you have a higher reference frequency then you don't need so much filtering. Perhaps even a little inductance or even a little passive 1KHz notch filter would give a perfect result.

At this stage I must point out that a 100MHz synthesizer with a 1KHz reference frequency is pushing your luck a bit far. The 88MHz to 108MHz FM band has a channel spacing of 50KHz, so a dual modulus synthesiser with a 50KHz reference frequency is all that is required. The filter would then be very simple indeed. An amateur radio transmitter for 145MHz would only need a 25KHz (12.5KHz) reference frequency so the ripple is still outside the audio bandwidth and high enough to make filtering a simple task. The time constants would then be a lot shorter and the re-tune time can be reduced to just a few milliseconds.

Loop Filter - Further Experimentation

If it was necessary to have a very short time-constant in the loop filter, for example with frequency hopping applications, then the R2/C2 time constant must be so short that your lower audio frequencies are not filtered out. This makes part of your modulation become an error signal which the loop will correct. The loop will therefore remove your lower modulating frequencies!

One way out of this situation is to devide your modulation signal into the high-frequency components and the low-frequency components. Modulate in the normal way with the high frequencies, but apply your low frequencies as frequency modulation to the PLL reference oscillator. In this way, the LACK of low modulating frequencies will be seen as an error and the loop will add them! Sorry, no circuits, I will leave this to your own imagination and expertise.

Coupling Filter To VCO

The Op-Amp at the end in the previous circuit is just a voltage level converter. My varicap diode needs 1v to 15 volts to give a decent control range, but the CMOS/TTL circuit will only give me 0v - 5v. This little circuit will do the job quite nicely.

Most varicap diodes need up to 28vDC to get the full range so you could add or improve this as you wish. The resistor values are not at all critical, as long as the ratio is right - 2:1, 3:1, etc.

I have already shown you that a varicap diode is not linear when you consider the frequency versus voltage. This means that the loop gain will vary, (voltage to frequency conversion) depending upon the DC level of the loop voltage. Since we are presently mucking about with the varicap driving voltage, let me take this opportunity to try to correct the non-linearity. With low DC control voltage we have a high frequency change, but with a high control voltage we have a low frequency change. What we therefore need is a DC amplifier with the opposite DC transfer function.

This does the trick, perhaps a little too much, depending upon the actual components you use in your own synthesiser. The resistor Rx can be adjusted to give you the highest voltage level you need, but as the output from the Op-Amp rises, then more diodes will conduct, switching in more parallel resistance in the DC feedback circuit. As long as the input voltage is less than 0.7vDC, the output of the Op-Amp will vary from 0 - 0.9v DC since there is just one 470K resistor in the feedback loop. But if the input is between 0.8vDC and 1.6v DC, then the output of the Op-Amp will lie somewhere between 0.9v and 1.7v. As the DC control voltage rises, the DC amplification also rises at an ever increasing rate.

The 100K resistor will give the Op-Amp a gain of 2 with 5v DC input, but a gain of only 1.2 at 1v DC input. This gives us an output voltage swing of 0v DC to 10v DC. 150K would give up to 15v. Select the resistor for the maximum voltage you need to drive your varicap diode. You will, of course, need at least this voltage to power the Op-Amp.

Remember too, diodes do not switch instantly, they conduct slowly between 0.62v and about 0.72v. This "softness" helps to smooth the curve out so there are no well defined gain switchover points. Now, our voltage to frequency translation will be more constant over the range of 0-5v from the loop filter, so the overall loop gain will become more constant, making the filter damping correct over the whole VCO frequency range.

If you decide to muck around with the feedback circuit, then it is quite important NOT to add any capacitance to the Op-Amp feedback circuit, otherwise you could destroy the properties of the loop filter, or to be more precise, the damping that is preventing the loop from self oscillation.

Please note that this method is only really required for synthesisers that have a very wide frequency range, such as the 20% range of our 88MHz to 108MHz broadcast bands synthesiser. If you were constructing an amateur radio 144MHz to 146MHz synthesiser then this technique may be totally unnecessary. The total frequency range is only about 1.5% of the total range of the 2-metre band. This is what ICOM and YAESU do in many of their VHF FM transcievers - Most ham operators feel that their rig is some form of "FM deviation standard" because it was so expensive. It is still a compromise, and it DOES vary across the band!

Best regards from Harry Lythall - SM0VPO, Lunda, Sweden

Return to INFO page