AN920

LESSON: COLPITTS CRYSTAL OSC DESIGN, AN920

I will respond more complete when I have time, but here is some idea of minimum parrallel (load resistances) across crystal that will still allow circuit to oscillate. Freq MHz Res value 20 3k 10 4k 1 27k 0.5 370k (5kHz) 6M The loop gain is a bit more complicated as just the ratio of C1,2 and include data from the crystal, gm of transistor etc. Here we don't know the crystal data unless we measure it or get it from vendor data and have to make some educated guesses. The series internal resistance of crystals go up as the frequency goes down. That is the reason why if we load the crystal too much the losses over the internal resistance will kill the Q of the crystal and the prevent oscillation from happening. At higher frequencies the internal resistance is much lower (~20 Ohm at 20MHz) and can tolerate much lower loads across it. You can see this from the model of 32k crystal LS= 0.007 CS= 3.5e-015 RS= 18000 CO= 1.7e-012 and 15M crystal LS= 0.0075 CS= 1.5e-014 RS= 25 CO= 4.5e-012

Most single stage varactor RF VCO's will also suffer from modulation sensitivity. That is you need to apply more modulation level for the same required deviation as you increase the tuning voltage (higher frequency) to the VCO. Example: you will need much more modulation for the same loudness (listening volume) on 108MHz than at 88MHz. This problem can be fixed by using a dual section VCO. One varactor for tuning the RF and another biased and lightly coupled to the tank to handle the modulation. With this method modulation sensitivity is constant over a wide tuning range. C3 is very small because only a small variation is needed to deviate the carrier by 75kHz needed for wide band FM transmission Figure shows a typical arrangement, with another section added.

Plots shows happy and more happy circuit values. First plot shows osc struggling to start because of low loop gain. This circuit will not work with very low value crystals like 32kHz watch crystals as the crystal will be loaded too much. You need an oscillator circuit with a FET input for this. If you do want to try other values try and keep C2 slightly higher in value than C1. C1=C2 is also ok in most cases.

The caps will set the feedback ratio. If this ratio is too small (as in this case) the circuit will not oscillate as the loop-gain is less than 1. Try swapping the 2 caps around then it should work. As you go higher in frequency the cap values should be reduced. 20 MHz crystal would require values of about 47pF and 68pF. Emitter resistance also have some influence.

If the VCO tuning response is not linear, the center frequency will not be deviated by equal amounts by the modulating signal, if it is a pure sine wave for example. When this signal gets demodulated this will show up as distortion in the recovered sine wave. FM detectors with some of the best linearity is called a pulse-count detector and was invented by Pioneer corp.

http://www.st.com/stonline/products/literature/ds/7559.pdf

Here's plots from swcad using vendor coils from their library with Q of about 100, which is about normal from winding a 100nH coil. Feedback cap is set at 10pF. Now the collector current does not go under zero. Vc is the voltage on the collector, Ic the collector current and VL the voltage over the 75 Ohm load. Ic was meaured by placing a 0.001 Ohm resistor in series with the collector terminal and using the current probe. Model for 2N3904 is still linear SPICE model and not the more accurate non-linear model used before. Note:For some reason (better resolution?) swcad change the scale when the collector voltage and load voltage swing are plotted on the same graph. Load voltage should swing negative and positive because of the coupling cap. When it is plotted alone on a graph, it shows correctly. The p-p swing indicated is still correct.

So, how do we get more power? Well running the osc on 9V gives us almost 8mW drive and higher harmonics, which gives us about 130mW on the power curve. The problem is that now the frequency will drift as the 9v battery gets flat. So no easy answer apart from adding another amplifier stage.

Finally, I biased the amplifier stage into class-C with removing the 47k resistor and adding a RF choke to gnd on the base. I ran a power sweep again. Now we can see that if we can get the drive level up to about 8mW (which it seems we can't) from the simulation data, it is possible to get up to 150mW into the 75 Ohm antenna or load before gain compression for this class starts to come in. The 2nd last plot shows the Vp-p on the collector as the drive level is stepped up. Distortion is clearly visible as the drive level is increased to maximum (10mW) Last plot shows power dissipation in the transistor (Watts) for various drive levels 1 to 10mW.

Next I increased the feedback cap to 22pF. Now the drive level was near 2.6mW (17 times higher), BUT note the rapid rise of harmonics from the osc (green circles) and we are no longer on the linear part of the power curve. The amplifier is now heavily into gain compression. Output of F1 is now 55mW and amp almost in full compression (no increase in input drive will provide higher output) as we have reached the flat part of the curve. Also note how the osc frequency shifted down as expected because of the higher capacitance. Here is also the plot for power in transistor with 47k base resistor connected with input drive stepped 1-10mW.

I did some more simulation to show the effect of increasing the feedback cap on output level and harmonics. First I took just the output stage and performed a RF power sweep on this stage to determine the maximum output power and gain compression. Then I measured the drive level in mW into the final stage from the oscillator. This plots show the drive with a 4.7pF cap to be about 0.15mW to the final stage. Plotting that on the power sweep graph shows the output produces by the amplifier stage to be 11mW at the fundamental. PO1 is the output and PO2 is the input drive

AC load line and fundamental power with 100k base reistor

http://www.electronics-lab.com/forum/index.php?action=dlattach;topic=10866.0;attach=9361;image Hi AG, Are you sure about this? Using a non-linear simulator with vendor non-linear models for the transistor and coils (Q value 200), shows that the collector current never gets to zero, and the Vp-p on the collector is 3.4V giving only about 10mW F1 into the 75 Ohm load. AC load line analysis shows that it always operates in class A. Collector current swings between 16-41mA on the final transistor collector. This is your original circuit with fb cap 4.7pF. With 10pF fb the output is under 30mW still operating class A. Vp-p on collector increase to about 6V. Increasing the base resistor to 100k shows that the collector current starts to go below 0, but the power drops off to 20mW. Have you actually measured these voltages on a scope?

Maybe you should send an email to "Self" regarding your views, he welcomes any commentary!

It is anybody's right to disagree :) Just remember that these statements are made by persons that are considered world renowned authorities on the subject and have presented numerous studies and papers to bodies like the IEEE and other professional publications. They have backed up their statements with simulation, experiments and measurements in a lab.

From another author: Randy Slone's book Enthusiasm for class A today is often supported by a variety of misunderstandings and myths. The following list provides some major examples. 1. "Crossover distortion in class B is a constant. At low listening levels, it becomes very prominent." This is an understandable assumption, but fortunately it is not true. The absolute level of crossover distortion decreases as the output signal level is decreased, but not in a proportional manner. The principle behind this phenomenon will be discussed later. However, the end result is only a slight increase in THD as the output level is drastically reduced. 2. "Even though class A amplifiers are inefficient, they are easy to design, and the sonic quality will be exceptional even in poorly designed units." In reality, a well-designed class A amplifier is about equal in complexity to a good class B design, and the class A amplifier will virtually always be more expensive. 3. "In comparison to class B designs, class A amplifiers sound better." A well-designed class A amplifier operating under favorable conditions compared to a well-designed class B amplifier operating under the same conditions will not reveal any perceivable differences. Otherwise, the human ear would have to be capable of hearing distortion levels into the hundredths of a percent at 20 kHz. But we must be careful to compare apples with apples. There are many poorly designed class B amplifiers around, and their inferior performance characteristics are readily detectable. CLASS-AB Class AB is not really a class but rather a poor marriage of both class A and class B characteristics. It has been erroneously taught that a class AB amplifier becomes a class B amplifier when all bias is removed from the OPS. This is incorrect. Technically speaking, class B pertains to an OPS wherein the output devices are conducting for one-half (that is, 180 degrees) of the signal cycle. In order to accomplish this action, a small forward bias must be applied to the output devices to overcome their inherent Vbe drop. If this forward bias is removed from a class BOPS, the output devices begin to conduct for less than 180 degrees of the signal cycle, placing the OPS closer to the category of class C than class B. Class AB pertains to a class BOPS that has been overbiased so that each output device conducts for more than 180 degrees of the signal cycle. This was thought to improve crossover distortion through the mental imagery of the output devices sliding through the crossover region in linear class A operation. However, what actually results is a form of crossover distortion referred to as gm doubling. That is, in the crossover region while all of the output devices are conducting simultaneously, their current gain factors are doubling (i.e., summing), creating a severe wobble in the linearity. Both Fourier analysis and distortion analysis prove that this doubling effect causes distortion harmonics as bad as if the output devices were severely underbiased. Class AB operation also causes increased power dissipation in the OPS, decreasing efficiency and reliability. Since class AB operation provides no advantages whatsoever and only serves to degrade linearity and create additional heat problems, it should be entirely dismissed as a "good idea that just didn't work out." CLASS B As defined previously, class B pertains to an OPS wherein the output devices are biased to conduct for 180 degrees of the signal cycle. In times past, class B operation was referred to as push-pull operation (analogous to sourcing-sinking action), but this is a misnomer. Output devices in a class B OPS will source current to a load for a half-cycle, but they do not sink current during the opposite half-cycle: they are cut off. The term push-pull should be confined to class A type stages. At least 99 percent of all audio power amplifiers utilize a class B OPS. I make no apology for the fact that this book is devoted to the methodology and development of class B audio amplifiers since that appears to be the only practical and viable choice at our current technological level.

The quotation and paper refers to audio power amplifiers and not small signal op-amps.

Here is a good explanation about the subtle differences between biased Class-B and AB From the works of "Wang & Chien" It is important to realize this. Many people and even most engineering text books get this part wrong or don't understand the difference.

I made the 1k resistor on the inverting input R3 because you have multiple R2's From writing the nodal equations for the opamp and solving: Without the capacitor across the feedback resistor G(s) = (R2 + R3)/R3 = 48 With the capacitor added G(s) = (R2 + R3 + R3.C1.R2.s)/(R3 + R3.C1.R2.s) Imput impedance 1k, output impedance ~ 24m Ohm

Another problem with feedback is that you need to increase the slew rate BW of the front end gain stages including the feedback network by the factor of feedback (40 to 50 times) to prevent/minimize dynamic intermod distortion. This was the reason so many transistor amplifiers sounded so bad in the early 70's. This situation improved after it was discovered and pointed out by Dr. Otala in his published papers. He suggested that minimum slew rates for power amps with a 30kHz BW should be 100V/us ! Also adding negative feedback to single-ended, differential pair or push-pull output stages using FET's can be very surprizing, as it actually may boosts higher order intermod distortion over the full spectrum.

In a recent published Master

You think? What do you understand under transconductance?

The jury is still out on TIM. Some people believe this distortion to be caused by a combination of other well known distortion types. One of the misconceptions is the square wave output testing into a capacitive load, measuring the overshoot-ringing to evaluate transient response/distortion of the amplifier. It was proved to show nothing about the performance of the amplifier, but only the resonating of the Q-damped output inductor with the load.

The detectable distortion with low order harmonics present is about 1% Crossover distortion can be detected at levels under 0.3%