Phase noise is rapidly becoming the most critical factor in sophisticated radar and communication systems. This is because it is the key parameter defining target acquisition in radars and spectral integrity in communication systems. There are many papers detailing the mathematical derivation of phase noise; few mention the reasons for its importance.
Phase noise is commonly used as a measure of the frequency stability of an oscillator. This noise is inherently different than the general background noise of any electrical system, which is defined as kTB, where k is Boltzmann’s constant, T is the temperature and B is the bandwidth. Phase noise is a secondary effect related to the topology and construction of the oscillator (see Figure 1). The figure shows the output power of an oscillator versus frequency, comparing an ideal oscillator, with power at a single frequency, to an oscillator with phase noise, which shows power across a spectrum of frequencies very close to the desired output. These skirts, as they are called, are always present and are due to thermal noise within the active devices of the oscillator. The power level of the skirts is dependent upon the quality of the oscillator and is measured in dBc/Hz at an offset frequency from the desired signal, typically called the carrier.
Phase noise affects the performance of many microwave systems, seen by considering direct down-conversion receivers and radars. Direct down-conversion is a type of receiver used in microwave communications systems. One benefit of direct down-conversion is the simplicity of the circuit, which is essentially a single mixer driven by a local oscillator (LO) that converts the input RF signal to a baseband or very low frequency signal. This baseband is applied to an analog-to-digital converter for subsequent digital processing. A common expression for this architecture is “RF in, bits out.” One problem with direct down-conversion, though, is that the input RF signal can be very close in frequency to the LO, which makes the conversion process susceptible to phase noise, especially if the signal strength is low. With radar, the problem is similar. Radar systems operate by transmitting a pulse at one frequency, then measuring the frequency shift of the returned pulse. The shift, called the Doppler effect, is related to the velocity of the object being imaged. Objects moving very slowly will generate a return pulse very close in frequency to the transmitted pulse; if the cross section of the object is also small, the power level of this received signal will be very low. This return pulse is converted to baseband to recover the velocity information, and phase noise can obscure the data.
An illustration of the dilemma faced by direct-conversion receivers and radar systems is shown in Figure 2. If the power level of the RF signal falls below the phase noise spectrum of the LO signal, the baseband information is not recoverable, as the signal is in the noise. Reducing the phase noise increases receiver sensitivity. A second example of how phase noise negatively impacts communications systems is shown in Figure 3, which illustrates a multi-carrier orthogonal frequency-division multiplexed (OFDM) signal. If the phase noise of the LO is too high, the noise will be converted into adjacent channels of the baseband data and ruin the integrity of the information.
AMPLIFIERS AND PHASE NOISE
One obvious place to limit phase noise is with the choice of oscillator, by spending considerable time and money to design or procure a low noise oscillator. However, most oscillators do not generate sufficient output power to drive the LO port of a mixer and are followed by an amplifier. For example, an oscillator’s output of +5 dBm needs to be amplified to a level of 15 to 17 dBm to drive the mixer.
Does the amplifier affect the phase noise of the LO signal? In an ideal situation, the answer is “no,” as the amplifier simply raises the desired LO signal and its skirts by the same level. In reality, microwave amplifiers add noise of their own to any signal. All electronic devices exhibit a phenomenon called 1/f or pink noise, which is noise power that is added to the spectrum of the input signal but falls off proportionally to the inverse of the offset frequency. Figure 4 shows the phase noise versus offset frequency of a low noise amplifier (LNA) that covers 10 to 17 GHz. The phase noise of the incoming signal has been cancelled out, so the plot represents only the noise generated by the amplifier. Note that the phase noise falls off linearly on the logarithmic scale with increasing frequency offset, which is characteristic of 1/f noise. If this noise is greater than the phase noise of the input signal, then the amplifier noise will dominate the output noise spectrum. In this example, the low phase noise of the oscillator would be replaced by the higher phase noise of the amplifier, defeating the purpose of the low phase noise oscillator. Figure 5 illustrates that the skirts of the signal at the input of the amplifier are increased after passing through the amplifier.
One obvious question is what can be done to lower the phase noise of amplifiers? The answer lies in device physics. The 1/f noise is caused by random and thermal charge movement in the channel of an active device. The CMD167, with the phase noise shown in Figure 4, is manufactured on a GaAs PHEMT process with a gate length of 0.13 µm. The FET devices on this process typically have a high 1/f corner due to their high electron mobility. GaAs bipolar devices, in comparison, tend to have lower electron mobility, which means a much lower 1/f noise. As they are considerably better for phase noise than their FET counterparts, one solution to lowering additive phase noise is to use a GaAs HBT process.
Custom MMIC used their extensive knowledge of amplifier design techniques to develop a family of low phase noise amplifiers designed on a GaAs HBT process. The three amplifiers in the family span 5 to 40 GHz (see Table 1). Figure 6 shows the phase noise versus offset frequency for one of the amplifiers, the CMD245, which is available in a 4 mm QFN-style package. For comparison, the phase noise of the CMD167 GaAs FET LNA is included. The phase noise of the CMD245 is 15 to 20 dB lower than that of the CMD167.
Other components in addition to oscillators and amplifiers can contribute to phase noise, including frequency multipliers. Many microwave systems utilize a lower frequency oscillator that is multiplied to produce a higher frequency. One common approach for multiplication is to use a harmonically terminated amplifier to generate the required output frequency. Unfortunately, such an approach will add the amplifier’s phase noise to the multiplied signal, which will degrade the phase noise of the original oscillator. A second approach is to use passive multiplication, which has the potential to add minimal phase noise to the multiplied signal.
Custom MMIC has created a family of passive HBT-style frequency doublers that do not add to the phase noise of the input signal (see Table 2).