Tuning Linearity

 

Figure 1 (above)

Varactor tuned oscillators are basically non-linear devices. The output frequency versus tuning voltage is an exponential function of the type shown in Figure 1. Linearity at EMF SYSTEMS is determined as follows: A straight line is drawn from the beginning to the end of the tuning curve. Any deviation from this straight line of the actual curve is considered a frequency error and is figured as a percentage of the corresponding frequency on the straight line. For example, in Figure 1, the error is 30 MHz, the corresponding frequency is 1500 MHz; therefore, the linearity error is 2%. Please take note that a best straight line drawn through the actual curve would yield a much better linearity.

Linearizers are used to give a linear voltage versus frequency transfer function. They can also scale the tuning voltage to convenient levels such as 0 to 10 Volts. When a linearizer is used, the linearity is obtained in the same manner, except a straight line is drawn though 0 V to 10 V.

Oscillator Noise AM and FM (phase)

 

Figure 2 (above)

The perfect oscillator produces a signal whose spectrum would consist of a single line of infinitesimal width. No perfect oscillator has yet been discovered, consequently, oscillators have noise spectra in accordance with well established theory. The spectrum of a typical low noise oscillator is shown in Figure 2.

AM noise is generally far enough below FM noise in an oscillator that it is of little concern. FM noise can vary considerably from one manufacturer to another and should not be taken for granted since it can be the limiting factor in applications like narrow band communications links, frequency synthesizers, Doppler radars, etc. It limits the range resolution, sensitivity, and channel spacing of these systems. FM noise may be due to discrete modulation signals such as power line frequencies and mechanical vibrations. These produce discrete sideband noise at the frequency of modulation as shown in Figure 2; they are also called spurious signals. FM noise can also be due to random type modulation caused by thermal vibrations and flicker noise within the device, the Q of the tank circuit, etc. The spectral density plot of these sidebands shows a continuous spectrum over a wide range of frequencies similar to broadband noise, sometimes called "hash." Control of the discrete noise components is as much the task of the Systems Engineer as the oscillator designer. "Pick-up" on the tuning line due to unshielded wires, poor filtering, grounding, etc. will produce noise sidebands. The oscillator should be isolated from digital circuitry and separate power supplies should be isolated from digital circuitry and separate power supplies should be used when possible. Remember that any noise components on the tuning terminal have the same tuning sensitivity per volt as the desired modulation. For example, if a VCO has a 50 MHz/Volt tuning sensitivity, noise of 1 millivolt peak to peak will FM the oscillator 50 KHz. From FM theory the ration of power in the carrier to one of the sidebands due to this noise can be predicted by:

Where fm = frequency of noise in Hz DF = peak deviation in Hz

With this relationship, FM noise can be compared as "so many db below the carrier" any place in the spectrum in a certain measurement bandwidth. A commonly used bandwidth is 1Hz. The actual bandwidth used in making the measurement may be greater than 1Hz; however, it can be converted to 1Hz for comparison.

 

Figure 3 (above)

Figure 3 is a plot of FM noise often supplied by oscillator manufacturers.

    

Figure 4 (above)

Another way of presenting an oscillator's FM noise is to convert the "db below the carrier" to an equivalent frequency deviation that would produce the same noise level in a 1Hz bandwidth at the same distance from the carrier. The deviation is expressed in Hz RMS. For example, FM noise 80db down in a 1Hz band 10KHZ from the carrier can be expressed as a frequency deviation of 1.4 Hz RMS. Equation 1 can be used to convert from one method to the other if peak is used instead of RMS, for DF. Figure 4 is a convenient chart for converting between the two methods.