Oscillator Basics
Oscillator Basics
Oscillator Basics
Hartley Oscillator
I decided to use the Hartley Oscillator for the simple reason it's my favourite. Recently it was discussed that your favourite oscillator was likely the one which worked best for you and I think that is quite true. So here it is in it's most simplified form.
Colpitts Oscillator
The basic Colpitts oscillator circuit look like this and you will see some similarities.
Figure 2 - schematic of a collpitts oscillator If you consider positive feedback is applied to compensate for the losses in the tuned circuit, the amplifier and feedback circuit create a negative resistor. When Z1 and Z2 are capacitive, the impedance across the capacitors can be estimated from a formula I won't lay on you here because it includesbeta, hie, as well as XC1 and XC2. Suffice to sayit can be shown that the input impedance is a negative resistor in series with C1 and C2. And the frequency is in accordance with:
A frequency change of a few tens of hertz back and forth over a couple of minutes would mean nothing to an entertainment receiver designed for the FM Radio band. Such a drift in an otherwise contest grade receiver designed to receive CW (morse code) would be intolerable. It's a question of relativity.
HARTLEY OSCILLATORS
What are Hartley oscillators?
Hartley oscillator are inductively coupled, variable frequency oscillators where the oscillator may be series or shunt fed. Hartley oscillators have the advantage of having one centre tapped inductor and one tuning capacitor. This arrangement simplifies the construction of a Hartley oscillator circuit. First off let's look at a schematic of a hartley oscillator.
Figure 2 - schematic of a hartley oscillator For this design I'm going to say we will be constructing a general purpose VFO to operate at 5000 5100 Khz no particular reason, pick anything you like. Now I chose a 2N4416A FET purely because I bought a big bag of them years ago and have them on hand. You could use any general purpose JFET you can readily obtain. Note the 2N4416A is a metal can and the case is grounded. 4
The frequency determining components are L1, Ct (a nominal 10 pf trimmer), C1a, C1b, C2, C3, Cv and C4. Note: I have been asked a number of times the function of C4 in this circuit. Capacitor C4 is to reduce the loading on the tuned circuit components. It may be as small as possible consistent with being able to provide sufficient drive to the succeeding buffer amplifier stage. Often the home constructor will often make C4 a trimmer. The other components are bog standard. The two resistors, silicon diode and zener diode need never change, capacitor C5 is about right for this frequency. C6 can be selected to give higher / lower output to the buffer amplifier. Smaller C6 values give lower output and conversely higher values give larger output. The silicon diode I'll explain later, the zener diode is to give a regulated 6.2 volt supply. Now there is NOTHING sacred about my frequency determining capacitor combination O.K.? Too many people look at these kind of circuits and think they must duplicate everything literally, not so. This is just a typical representation. C1 to C3 plus Cv and Ct are just a combination of parallel and some series capacitors all designed to give us a bit of flexibility with the tuning range. Cv could easily be replaced by two back to back tuning diodes. What you need to do to get the circuit to work is to have an inductive reactance for L1 of around about 180 ohms. At 5 Mhz this works out at about 5.7 uH and, if you don't know how I arrived at that figure I seriously recommend you spend some time on my other tutorials such as Basics and LC Filters. The important aspect is that the feedback point from the source of the JFET connects to about 25% of the windings of L1 from the ground end. Now I've depicted an air cored inductor. It could be, just as one example among a great many, 18 - 19 turns of #20 gauge wire on a 25.4 mm (1") diameter form spread evenly over a length of about 25.4 mm (1"). The tap would be at about 4 1/2 turns. Check that out with the formula's I taught you elsewhere. Alternatively, with degraded performance, you could use a T50-6 toroid and wind say 37 turns of #24 wire (5.48 uH) tapping at 9 turns. The AL factor for a T50-6 is 40. Again do the other tutorials if necessary, I'm not going to repeat old work and it's going to be even harder from here on. I'll thoroughly explain new concepts, not the old. So if we are to have our oscillator working at about 5 Mhz, we know the LC is 1013 and if L is say 5.7 uH then total C for resonance (just like LC Filters eh!) is about 177 pF. We want to be able to tune from 5000 to 5100 Khz a tuning ratio of 1.02 which means a capacitance ratio of 1.04 (min to max). Let's fiddle with some numbers! I have a Jackson Bros. air variable capacitor (very Rolls-Royce) which swings from 10.5 pF to 105 pF, a typical 10:1 ratio in air variables. This I will use for Cv. If the total swing is 1.04 (actually 1.0404:1) and Cmax is 177 pF it follows Cmin is 170 pF. A variation of only 7 pF (roughly). Now we're treading on unsafe ground here with such a large variable capacitor. We could: A) rip plates off it to reduce capacitance (don't even think about it) B) go to varactor diodes with a small swing. That's O.K. but performance becomes degraded. C) obtain a smaller air variable with Cmax of say 25 pF.
Just to prove I'm a glutton for punishment and if you're still here so are you, we will purely for the mathematical exercise, persevere with the 105 pF variable. What if we eliminate C3 and make C2 = 15 pF NPO then the series combination of C2 and Cv swing 6.176 pF to 13.125 pF, a variation of over 6.9 pF - are you lost? Go back to the other tutorials on capacitance. If our Cmax was 177 pF then 177 - 13.125 = 163.875 and the 177 pF was approximate anyway. I'd make Ct a 10 pF air trimmer (if available, if not, a ceramic or whatever the supplier offers but 10 pF max.). That leaves about 154 pF to make up. How about making C1a and C1b into 3 NPO capacitors of say 2 X 47 pF and 1 X 56 pF all NPO types. In total that comes to less than 177 pF max. but don't forget there are stray capacitance's in the circuit. In the final wash-up you could simply use 3 X 47 pF. Here's what we finish up with
COLPITTS OSCILLATORS
What are colpitts oscillators?
Colpitts oscillators are somewhat similar to the shunt fed Hartley circuit except the Colpitts oscillator, instead of having a tapped inductor, utilises two series capacitors in its LC circuit. With the Colpitts oscillator the connection between these two capacitors is used as the centre tap for the circuit. The basic Colpitts oscillator circuit look like this and you will see some similarities with the Hartley Oscillator..
Figure 1 - schematic of a collpitts oscillator Perhaps the simplest Colpitts oscillator to construct and get running is the "series tuned" version, more often referred to as the "Clapp Oscillator". Because there is no load on the inductor a high "Q" circuit results with a high L/C ratio and of course much less circulating current. This aids drift reduction. Because larger inductances are required, stray inductances do not have as much impact as perhaps in other circuits.
Figure 2 - schematic of a series tuned colpitts or "clapp" oscillator Rather than present designs for specific frequencies for the Colpitts Oscillator we have submitted a schematic which may be "impedance" scaled to any frequency. Simply convert the suggested reactances back to the required inductor and capacitances at your band of interest. 7
Figure 3 - calculations for a series tuned colpitts or "clapp" oscillator Here in our Colpitts oscillator Ctotal-max is the maximum of the parallel combination including the variable capacitor Cv, set at maximum while Ctotal-min is the same combination with Cv set at minimum. Clear on that?
Note that the other series capacitor depicted in line with Cv in figure 2 may or may not be required in your particular application. Anyhoo! your calculations should have yielded a Cmax of 112.3 and Cmin of 103.51. Now that is a pretty tiny swing, so you can see the possible need for a series capacitor with Cv if all you have available is a fairly high value Cv. Let's assume Cv is 5 - 25 pf (a variation of 20 pF) and Ct is a 10 pF trimmer. Where do we stand now with our Colpitts oscillator calculations? Allowing for taking strays into account we will take the full value of Ct of 10 pF, therefore the balance of the parallel combination accounts for Cmax of 102.3 and Cmin of 93.51. The net variation of C still remains 102.3 - 93.51 = 8.79 pF so Cv needs to be seriously reduced. How is largely determined by the same method below (suck and see approach) but I come up with a "starting point" of about 12 pF in series with Cv. Therefore nearing completion of calculations for our Colpitts oscillator, we find this new series combination with Cv produces a net Cv-max 8.1 pF and Cv-min 3.53 pF. This is obviously far too much reduction using the 12 pF capacitor, so try using some higher values like 27 pF! Using 27 pF in our selection gives Cv-max 12.98 pF and Cv-min 4.22 pF and a net variation of 8.76 pF. Kicked a goal! New Cv is (at max) 13 pF and Ct assumed at 10 pF a total of 23 pf from a required maximum of 112.3 leaving about 89 pf to be made up of fixed capacitors. I'd use at least three fixed capacitors, say 33 pF plus 2 X 27 pF.
CRYSTAL OSCILLATORS
What are crystal oscillators?
Crystal oscillators are oscillators where the primary frequency determining element is a quartz crystal. Because of the inherent characteristics of the quartz crystal the crystal oscillator may be held to extreme accuracy of frequency stability. Temperature compensation may be applied to crystal oscillators to improve thermal stability of the crystal oscillator. Crystal oscillators are usually, fixed frequency oscillators where stability and accuracy are the primary considerations. For example it is almost impossible to design a stable and accurate LC oscillator for the upper HF and higher frequencies without resorting to some sort of crystal control. Hence the reason for crystal oscillators. The frequency of older FT-243 crystals can be moved upward by crystal grinding. I won't be discussing frequency sythesisers and direct digital synthesis (DDS) here. They are particularly interesting topics to be covered later.
Figure 1 - schematic of a crystal oscillator The transistor could be a general purpose type with an Ft of at least 150 Mhz for HF use. A typical example would be a 2N2222A. The turns ratio on the tuned circuit depicts an anticipated nominal load of 50 ohms. This allows a theoretical 2K5 ohms on the collector. If it is followed by a buffer amplifier (highly recommended) I would simply maintain the typical 7:1 turns ratio. I have included a formula for determining L and C in the tuned circuits of crystal oscillators in case you have forgotten earlier tutorials. Personally I would make L a reactance of around 250 ohms. In this case I'd make C a smaller trimmer in parallel with a standard fixed value.
10
You can use an overtone crystal for the crystal and set L * C for the odd particular multiple of overtone wanted in your crystal oscillators.
11
OSCILLATOR DRIFT
What is oscillator drift?
Oscillator drift can be related directly to frequency stability. Drift is the unwanted and unwarrented change in frequency measured over seconds, minutes or hours. Just how stable should an oscillator be? My own belief is the answer to this is a similar answer to many questions related to electronics, your oscillator or any electronic project should be as state-of-the-art as is possible, consistent with your design goals. While it is possible, given enough resources, financial and otherwise, to construct a state-of-the-art oscillator accurate to within one part in ten to the 14th power for as one example, an amateur radio transceiver, you generally couldn't justify those extreme lengths. On the other hand a similar CW (morse code) transceiver operating at 7 Mhz which drifted around 100 Hz or more every few seconds would be consider very poor and unacceptable. A cheap portable AM radio receiver which drifted plus and minus 50 Hz over seconds would not necessarily be noticeable. On the other hand if such a receiver kept drifting over time to the point it required constant retuning then it also would be totally unacceptable (reminds me, I must look into my entertainment FM receiver).
For our illustation we are saying the only influence on stability is our 82 pF fixed NPO capacitor when our temperature varies around 25 degrees. O.K. what's a 82 pF NPO capacitor. It's a capacitor where it's value theoretically should NOT change with varying temperature. NPO means Negative, Positive temperature co-efficient = 0 (zero). Assume in this example it is slightly defective and in fact exhibits a positive drift of 100 ppm (parts-per-million) per degree temperature variation. Now the temperature after power turn on of our oscillator rises to 28 deg C inside our enclosure. It follows with a temperature rise of 3 degrees our capacitor is going to vary positive 3 X 100 ppm. Applied to a 82 pF capacitor this results in a net variation of +0.0246 pF. This leads to a new value of 82.0246 pF added to our trimmer value we get a new overall capacitance of 90.97028177 pF which when multiplied by our fixed inductor value above leads to a new LC value of 517.0846435 and a new resonance of 6.999053473 Mhz (6999053.473 Hz). Here our frequency has already varied down by 7,000,000 Hz - 6999053.473 Hz = 946.527 Hz! Wow, 946.527 Hz! That's almost a full kilohertz. You've just gone across several CW stations and an AM receiver probably needs retuning. O.K. it's a bit extreme but certainly highlights my point. I deliberately picked the NPO capacitor for good reason. Here I quote from my Philips Capacitor data book, BC06 - 1999 Page 166 "Precision Capacitors NP0" - the temperature co-efficient is zero per degrees kelvin. HOWEVER the tolerance on that temperature co-efficient is plus / minus 30 ppm. Simply put, the tolerance alone is about 30% of the example I used above. And, remember this is stateof-the-art and high quality manufacturing. Why did I pick on the NPO precision capacitor? Because it is the one capacitor LEAST likely to cause you problems. Why did the temperature rise a modest 3 degrees? Simple, after turning the oscillator power on, current flows through all parts of the circuit, power is consumed, POWER = HEAT. A small piece of wire has finite resistance, even a small current flow must by ohms law produce heat losses.
Another leading cause is poor power supply regulation and often using too high a voltage can cause problems. Similarly trying to take power from the ocillator will contribute to oscillator drift. Use the best regulated (locally) and well bypassed power supply at the lowest voltage which will ensure reliable operation is a must. Mechanical stability is another consideration. Use rigid larger diameter wires for interconnections where necessary between components.
An air wound inductor of larger diameter wire, firmly fixed to an inert former is probably the best. However the main problem here is that usually such an inductor assumes a large physical size which may create other problems. This was the reason for the introduction of iron powder and ferrite cores and toroids - minimise size. If size is a big consideration then use a toroid of the lowest possible temperature co-efficient consistent with the frequency range of interest. Fix the coil turns with quality "Q" dope or similar. Use the best regulated power supply at the lowest voltage possible Here I would review the various tutorials on power supplies. To minimise oscillator drift I would go for an on board regulator capable of adjustment. You don't need a high powered job, we're only talking milliamps here so something like the TO-92 version of an LM317 would do the trick. The power supply feeding this should ideally, also be well regulated, probably 12V. The trouble is these TO-92 regulators are often hard to come by so you may have to go to a fixed regulator like a 5V or, 8V type. It all depends on what is available to you. At worst you go to a zener diode. In any event experiment BUT regulate. Make sure your oscillator is "well buffered" to minimise oscillator drift Good buffering is absolutely essential to any oscillator to minimise drift. Taking power directly from an oscillator is a sure recipe for problems in many cases. The components in a buffer are relatively cheap. Mechanical stability has an effect on oscillator drift Rigid construction is of course best. Components and wiring capable of mechanical vibration are a source of problems in oscillator drift. Also a source of "microphonics". Just how rigid? Well like all things here, aim for the best but live with what you can achieve in the real world. I know amateur radio operators who regularly drop projects from a height of 300 mm (12") over a table as a "mechanical drift" test. The project usually passes! Shielding and heat sinking I've seen oscillators constructed on a base of heavy sheet brass. Why? Mechanical reasons and to conduct heat away as much as possible. Depending on your circumstances that's probably over the top but you get the idea. I would most certainly shield the "frequency determining components" with something like scrap PCB, thin copper or brass sheet. I've even then insulated it by gluing pieces of styrofoam to it. O.K. sometimes I'm too obsessive. Other solutions to oscillator drift In precision applications I've solved oscillator drift by accepting temperature variation as a fact of life and met it head on! 15
Simple, what's the highest temperature you expect to encounter? Add 10 degrees and insert a stable heater inside your shield set to that temperature. I built, in this case, a precision crystal oscillator operating at 10 Mhz which was the time base for a precision frequency counter. The precision 10 Mhz crystal and associated capacitors were inserted inside an insulated can in which I had placed an "electronic thermometer" along with a current controlled 5 watt power resistor which served as the heater element. This is called a "crystal oven". I used two LM3911 temperature controllers which I think are now obsolete. From memory the temperature was about 45 deg C (113F) but I wouldn't swear to that. Anyway it was in accordance with the manufacturers recommendation as to the "turning point" for that particular crystal and to that particular specification. The whole point is that you can solve problems, if you can seriously justify the effort, by raising and constantly maintaining the frequency determining components at a constant temperature somewhat higher than you expect to encounter. Warm up from a "cold" start can be anything from 30 minutes to several days. YES! Several days! This gives rise to another important point. Usually any oscillator will take a few minutes to temperature stabilise after turn on. If you have just been soldering components then allow at least 15 - 30 minutes before taking oscillator drift measurements. Remember to write your results down in a book after each adjustment. Marvellous what you will learn from this. Especially when swapping low value capacitors of different temperature co-efficients. Record the results comprehensively.
16
17
Figure 2 - schematic of a varactor tuned hartley oscillator Now I've left Ct and C1 a/b all in the circuit. In this application of a frequency synthesiser they are unlikely to be necessary. To tune from 1.8 to 2.0 Mhz which is a frequency swing of 2 / 1.8 = 1.111 which when squared means we need a capacitance ratio of 1.234 to 1 That means the ratio of minimum combined capacitance in the circuit to maximum combined capacitance in the circuit must change by 1.234 to 1. Looking back at the tutorial on oscillators I said the inductor should have a reactance of about 180 ohms. So around the frequency of interest I expect an inductor of about 15 uH to be used for L1 in Fig 2. You should be used to calculating LC numbers by now but L X C at 1.8 Mhz = 7818 and at 2 Mhz it works out about 6332. Dividing both by our 15 uH inductor we get a Cmin of 422 pF and Cmax of 520 pf. Which incidentally if you check 520 / 422 = 1.232:1 So the variation of C is 520 - 422 = 98pF swing.
18
For synthesisers or any voltage tuning you should have the largest voltage swing possible. This minimises the effects of noise voltage on the tuning voltage. My BB112 diode can be operated ideally from 1V to 8V. That means we can tune the 200 Khz (2.0 - 1.8) with a variation of 8 - 1 = 7 volts. It follows 7v / 200 Khz = 35 uV/Hz. If our noise level is below this then the tuning can't be varied or fm'ed by noise. At 8V my diode exhibits a capacitance of around 28 pF while at 1V the capacitance is about 500 pF.
Diodes Back-To-Back
You will note I have two diodes back-to-back in series in Fig 2. Although this in effect divides total varactor diode capacitance by two it eliminates the nasty effect of the rf present in the tank circuit driving a single diode into conduction on peaks which will increase the bias voltage, this also gives rise to harmonics. It follows that my varactor diode capacitance now swings a net approximate 14 pF up to 250 pF when the bias voltage is varied from 8 volts down to 1 volt. You can of course go below 1V for higher capacitance but I tend to be conservative and generally do not go below 1V very much. Now we have a net swing of 250 - 14 or 236 pF. You will recall above I said "the variation of C is 520 422 = 98pF swing" so how do I reduce a 236 pF swing down to a 98 pF swing? Look at capacitor C2 which is in series with both varactor diodes, does this not reduce the net capacitance?
figure 2 - (that is Ct can varied from say 5 to 25 pF) - and we allow the combination C1 a/b to be a total of around 390 pF we have obviously achieved our goal. Is this not cool? For our inductor L1 I would use a toroid although if you have access to a variable inductor you could use it. An air cored inductor most likely would be too large to consider. Suitable toroids of the Amidon / Micrometals type would at 2 Mhz be the T50-2 type which would require about 55 turns of #26 wire or even the T68-2 type requiring about 51 turns of #24 wire. Both gauges mentioned are those which will conveniently fit around the core. No matter your frequency range of interest the basic principles outlined above will more or less still apply.
Figure 3 - schematic of deriving varactor diode tuning voltage The 10K pot is your 5 or 10 turn "quality" potentiometer for tuning, the upper and lower trim pots (set and forget) allow you to adjust the voltage range of your choice that your tuning potentiometer will see. Again use "quality" trimpots. The 100K resistor and the two 0.1 uF capacitors are further filtering. Obviously there is considerable interaction between the trimpots and potentiometer so expect a lot of juggling back and forth. If you wished, in some applications, both trimpots could be replaced by fixed resistors. It is simply a matter of using ohms law.
20