How to choose the best power solution to improve RF signal chain phase noise performance

Today’s radio frequency (RF) systems are becoming increasingly complex. This high level of complexity requires optimal performance across all system metrics, such as tight links and noise budgets. Ensuring the correct design of the entire signal chain is critical. However, one part of the signal chain that is often overlooked is the DC power supply. It plays an important role in the system, but it can also have a negative impact. An important metric for RF systems is phase noise, which can be degraded depending on the power solution selected. This article examines the impact of power supply design on RF amplifier phase noise. Our test data proves that choosing the right power module can improve phase noise by 10 dB, which is key to optimizing RF signal chain performance.

What is phase noise?

Phase noise is the noise in a signal that is introduced as a result of unintended advances or lags when the signal arrives at the receiving end of a system. Just as amplitude noise is an offset or deviation from the nominal amplitude of a signal, phase noise is an offset or deviation from the nominal phase of a signal.

An ideal oscillator outputs a sine wave, as shown in Equation 1:

This sine wave is perfectly periodic, and the Fourier transform of Videal(t) is represented as an impulse function of the output waveform frequency. A more realistic representation of the oscillator output includes random fluctuations in phase (and amplitude), as shown in Equation 2:

This waveform includes some random process φ(t), which causes the signal phase to shift to some extent. This phase shift causes the Fourier transform of the non-ideal clock output to look more similar to Figure 1.

Figure 1. Phase noise of a nonideal sine wave.

Since the phase is slightly shifted, there are now multiple frequency components in the signal. Therefore, the signal is spread out around the center frequency.

Causes and Contributions of Phase Noise

The DC power solution of the signal chain is a significant contributor to phase noise, but it is often overlooked. Any noise or ripple on the power rails that power the signal chain can be coupled internally. This results in increased phase noise, which can hide critical frequency components in the transmitted bandwidth or introduce spurious offsets from the carrier. These spurs are particularly difficult to deal with because they are close to the carrier and pose great challenges to the filter due to the strict transition band requirements.

Figure 2. Noise in the power rail and its effect on the RF carrier signal.

Phase noise can be caused by many different factors. There are three main sources: white noise, shot noise, and 1/f, or flicker noise. White noise is caused by the random thermal motion of free electrons when current is passed through it. It is similar to shot noise in that it is caused by the random nature of the current. Unlike white noise and shot noise, flicker noise varies with frequency. It originates from defects in the semiconductor lattice structure and is also random in nature. Flicker noise decreases with increasing frequency; therefore, a low 1/f corner frequency is very useful. A typical phase noise curve can be approximated as regions with a slope of 1/fX, where x = 0 corresponds to the white noise region (slope = 0 dB/decade) and x = 1 corresponds to the flicker phase noise region (slope = –20 dB/decade). The regions with x = 2, 3, and 4 are closer to the carrier frequency.

Power Solutions

Figure 3. Power supply topology in an RF signal chain.

In an RF signal chain, ensuring proper biasing and powering amplifiers can be challenging, especially when the drain voltage is also used as an output port. There are many types of power solutions and topologies available on the market. The specific power solution you need depends on your application and system requirements. This experiment uses low dropout (LDO) linear regulators and step-down switching regulators to collect data, as shown in Figure 3. Step-down switching regulators are a typical solution to large voltage drops, with high efficiency and low operating temperature. Switching power supplies can step down higher voltages (such as 12 V) to more commonly used chip-level voltages (such as 3.3 V and 1.8 V). However, they can introduce severe switching noise or ripple to the output voltage, resulting in significant performance degradation. LDO regulators can also step down these voltages with lower noise; however, their power consumption is mainly in the form of heat. Using an LDO regulator is a good choice when the difference between the input voltage and the output voltage is small, but when the junction-to-ambient thermal resistance θJA exceeds 30°C/W, the high current drawn from FPGAs and ASICs causes the performance of LDO regulators to degrade rapidly.

Test Setup

Three different ADI power products were used in this experiment: LTM8063, LTM4626, and LT3045. Table 1 summarizes some of the data sheet specifications of the power solutions used.

The input signal was swept at 100 MHz, 200 MHz, 500 MHz, and from 1 GHz to 10 GHz. The phase noise was analyzed by selecting frequency offsets from 10 Hz to 30 MHz. The test setup is shown in Figure 4. The input RF signal was generated internally by a Rohde & Schwarz FSWP50 phase noise analyzer. This oscillator has excellent performance and was used because it clearly shows the additional phase noise or modulation spurs caused by the power supply.

Figure 4. Simplified block diagram of the test setup used in this experiment.

Two ADI amplifier products are used to represent a block in an RF signal chain.

result

Figure 5 compares the phase noise response of the PA when powered by the LTM8063 and the bench supply. It can be seen that the PA performance degrades slightly beyond the 1/f frequency. The PA consumes more supply current and the observed phase noise increases by approximately 2 dB to 4 dB.

Figure 5. (a) Performance of the HMC8411 and ADPA9002 at 2 GHz, and (b) phase noise response of the ADPA9002 powered by the bench and LTM8063 at two different input frequencies.

Figure 6 shows the phase noise response of the HMC8411 at input frequencies of 2 GHz and 8 GHz. The response follows closely, with the common-mode phase noise/frequency relationship given by Equation 3:

Figure 6. Phase noise response of the HMC8411 and LTM8063, showing phase noise/frequency.

This relationship shows that the phase noise increases by approximately 6 dB for every doubling of the input frequency. It can be seen that a 4× increase in frequency results in an approximately 12 dB increase in phase noise at a 10 Hz to 100 Hz frequency offset.

Figure 7 shows the phase noise response of the HMC8411 at 100 MHz and 10 GHz when powered by the LTM8063 and by a bench supply. The bench supply phase noise response is used as a benchmark to judge the performance of certain power solutions. The LTM8063 has excellent performance across a wide range of frequencies, with only an approximately 2 dB increase in broadband noise floor compared to the bench supply.

Figure 7. Phase noise response of the HMC8411 powered by a test bench and LTM8063 at two different input frequencies.

Generally, a high current module (such as LTM4626) is used as the main power supply so that the power distribution network can step down the voltage according to the requirements of each circuit module. From Figure 8, it can be seen that the phase noise performance of LTM8063 and LTM4626, which is a cascaded LT3045 ultra-low noise LDO regulator, is similar. If the voltage and current output provided by LTM8063 can meet the design requirements, the power supply solution can save a lot of cost and board space.

Figure 8. Phase noise response of the HMC8411 with various power solutions. fc = 5 GHz.

As can be seen in Figure 9a, switching power supplies can exhibit significantly different behaviors at different frequency bands. For power LNA phase noise below 5 kHz, the LTM8063 and LTM4626 have negligible effects on it, which is similar, but above 5 kHz, the performance between the two is very different. The LTM4626 is designed and optimized for high-end digital products. These devices usually require high efficiency and fast transient response, so their power supplies may have characteristics such as very low passive impedance, fast switching edge rate, high control loop gain and bandwidth. These characteristics will produce perturbations of several millivolts in the output voltage. Although these perturbations are insignificant in digital systems, they will degrade the performance of signal chain products. Despite this, using the LTM4626, there are no obvious spurs in the output spectrum when the SFDR is 102.7 dB, as shown in Figure 9b. However, the LTM8063 is designed for low noise (EMI and output) and its performance is optimized in signal chain applications. It has very good low-frequency stability, small output perturbations, and less noise on the switching fundamental and its harmonics.