Understanding Linear Regulator Noise in Hybrid Power Supplies

In many applications, engineers team a switching DC-DC voltage converter (“switching regulator”) with a low-drop-out linear voltage regulator (LDO). Such a hybrid power supply is a good way to maximize battery life while maintaining a noise-free voltage supply suitable for battery-powered products with sensitive analog circuitry.

However, “noise-free” is only a relative term, because even the best LDOs generate some noise. Many engineers believe that the noise at the output of an LDO is simply a function of its Power Supply Rejection Ratio (PSRR). When used in a series with a switching regulator, PSRR is indeed a good measure of how well the LDO deals with the voltage and current ripples in the switching regulator’s output – but PSRR should not be the only criterion upon which the LDO is selected.

The LDO has several sources of internal noise generated from its internal components and, worse still, some of this is frequency dependent. It is this combination of PSRR and internal noise that determines the stability of the LDO’s output.

This article explains how combining a switching regulator with an LDO can be a good solution for connected, wireless portable products, then leads on to explain why it is important that an engineer analyze the LDO’s datasheet to ensure that its overall noise from the power supply remains within the end-product specification.

LDOs and switching regulators compared

Battery-powered portable devices require voltage regulation to ensure that both battery life is maximized and sensitive silicon is fed with the precise and unvarying voltage it demands. In addition, regulation ensures that the power supply can cope with a wide range of loads without overstressing the battery.

Voltage regulation boils down to two choices, an LDO or a switching regulator. LDOs are elegant devices that are relatively inexpensive, compact, and provide “clean” power (but are not completely pure as we will discuss below). Any competent electronics engineer can design a power supply using a commercial modular LDO and just three external passive components. Moreover, there is a huge range of highly reliable LDOs from which to choose. Linear Technology’s venerable LT1084, for example, was developed two decades ago and is still available today.

The downsides of linear regulators are a lack of efficiency across a wide voltage range and restriction to a step-down (or “buck”) configuration.

The basic premise of an LDO is the use of a feedback loop comprising a reference voltage, error amplifier, a field-effect transistor (FET) operating in linear mode, and a resistor divider. While the resistor divider sets the output voltage, the FET enables the LDO to provide a constant output voltage regardless of load. The error circuit ensures that the output is maintained at that desired voltage. Figure 1 shows a simplified schematic of an LDO.

Figure 1: The fundamental elements of an LDO. (Courtesy of Texas Instruments)

The efficiency of an LDO is proportional to the difference between the input and output voltages; the less the difference, the more efficient the device. Power is dissipated across the FET and resistor divider and large voltage differences across the LDO can result in heat build-up, which can present a challenge in compact portables.

There comes a point when the battery voltage drops to a level whereby the LDO can no longer maintain the desired output and “drops out.” This level is known as the drop-out voltage. Higher-spec devices will tolerate a lower drop-out voltage than less-expensive devices. For example, Linear Technology’s LT3070 LDO––which generates a 0.8 to 1.8 V output at up to 5 A from a 0.95 to 3 V input––features a drop-out voltage of just 85 mV. Nonetheless, even the very best LDOs will likely drop out while some capacity still remains in the cell, reducing potential battery lifetime.

Switching regulators were introduced in the 1960s with the main advantage of high efficiency across a wide voltage and load range. Efficiency at low loads was a problem for early devices but has largely been addressed with “pulse skipping” techniques (see the TechZone article “The Advantages of Pulse Frequency Modulation for DC/DC Switching Voltage Converters”).

A switching regulator uses one or two FETs, but unlike in the LDO, these operate––as the product description suggests––in switching mode. When the FET is on and conducting current, the voltage drop across its power path is minimal. When the FET is off and blocking high voltage, there is almost no current through its power path. These characteristics ensure very little power is dissipated in the regulator enhancing efficiency. A contemporary 12 VIN, 3.3 VOUT synchronous­-switching buck regulator such as International Rectifier’s (IR) IR3898 can typically achieve over 90 percent efficiency (compared to 27.5 percent from an LDO operating under the same conditions).

As with LDOs, there is a huge range of switching power supplies available from major suppliers such as Freescale, Maxim, ON Semiconductor, and Texas Instruments (TI).

The downsides of a switching regulator are size (although power density is superior to LDOs), cost, design complexity, and noise. The last two items are related because much of the design complexity is related to inductor selection and the design of the input- and output-filtering circuitry required to ensure the peak-to-peak value of ripple is restricted. Judicious choice of inductor(s) and other passive components can limit output voltage- and current-ripple, but some electromagnetic interference (EMI) is inevitable (see the TechZone article “Capacitor Selection is Key to Good Voltage Regulator Design”).

The design challenge is exacerbated by the modern preference for high-frequency switching power supplies. By operating at a higher frequency, the power supply can utilize smaller inductor(s) reducing its size and cost. However, high-frequency operation makes the EMI problems more difficult to solve (see the TechZone article “Design Trade-offs when Selecting a High-Frequency Switching Regulator”).

EMI generated by a switching regulator can upset other components, particularly in a portable product where devices are densely packed and printed-circuit-board (PCB) traces are short. Worse yet––bearing in mind there is a growing cohort of design engineers keen to incorporate an RF chip into their latest product to endow it with wireless connectivity­­––stray EMI increases the difficulty of designing robust RF circuitry.

A third option

Rather than accept the compromises introduced by selecting either an LDO or a switching regulator, many enterprising engineers have created hybrid power-supply topologies by linking the output of a switching regulator directly to the input of an LDO.

The switching regulator accepts a wide range of input voltages and efficiently regulates to a higher, lower, or inverted supply. Also, if the output of the switching device (i.e., input to the LDO) is set such that it is only slightly greater than the required output of the power supply, the LDO can continuously operate in its most efficient range.

The key premise behind such a hybrid supply is that the LDO filters the switching regulator’s ripple-affected regulated output, eliminating potential EMI issues and obviating the requirement to spend long hours refining the PCB design and calculating the inductor and capacitor values for output-filter circuits (as would be the case if the switching regulator was used on its own). Other benefits include a power supply with improved stability, greater accuracy, faster transient response, and lowered output impedance (see the TechZone article “Hybrid Power Supplies Deliver Noise Free Voltages for Sensitive Circuitry”).

Unfortunately, things are a little more complicated than just selecting two regulators and linking them together.

Controlling noise

A measure of the LDO’s ability to smooth the switching supply’s voltage- and current-ripple is the PSRR. PSRR quantifies how well the LDO filters ripple from its input (in this case the switching regulator’s output) across a wide frequency range, and is expressed in decibels (dB). (The engineer should note that the PSRR response varies according to ripple frequency. It is important to ensure that the switching regulator’s operational frequency matches, or is at least is close to, the LDO’s best PSRR frequency response.)

Many engineers conclude that, provided the PSRR is good, then all will be well with the hybrid power supply. That’s not the case because, as described above, LDOs include voltage references, FETs, resistors, and other external circuitry that introduce additional (and not insignificant) noise independently of PSRR (see Figure 2). An LDO with a good PSRR but high self-induced noise would be a poor choice for generating a clean voltage from a hybrid supply. Consequently, it’s important the engineer checks both PSRR and internal-noise parameters on the shortlisted device’s datasheets.

Figure 2: The noise from a linear regulator comprises external noise, internal noise, and that left over after rejection. (Courtesy of Texas Instruments)

The simplified block diagram in Figure 1 shows the elements that generate the primary noise in an LDO, specifically, the temperature-compensated voltage reference (“bandgap”), the resistor divider, the input stage of the op amp and, to a lesser degree, the FET.

The bandgap is the dominant source of noise. One way to decrease this noise is to lower the bandwidth of the error amplifier inside the LDO. This does have the downside of reducing the transient response of the device. Alternatively, this noise can be eliminated by adding a large low-pass filter (LPF)––comprising a large internal resistor and an external capacitor––at the bandgap output so that little of the noise makes it into the gain stage (this same filter also improves the PSRR).

The resistor divider network is another contributor to the overall noise of the LDO. This noise is thermal in nature and thus increases as the temperature rises. The resistor divider is tied to the input of the LDO’s differential amplifier, amplifying the noise by an amount proportional to the closed-loop gain of the regulator.

The differential amplifier itself forms the final significant source of noise generated by the LDO. The amplifier is typically designed so that the input stage has a large amount of gain. Consequently, any noise coming from devices in the signal path located after the input stage is attenuated by the gain of the input stage, upon referral back to the input. There is nothing outside of the internal circuitry that can be done to reduce the noise from this source.

All of the primary noise sources (bandgap, resistor divider, and op amp input stage) are connected to the input of the differential amplifier and thus are not attenuated by any internal gain. Somewhat surprisingly, the power-pass FET, which usually takes up at least half of the total die area of an LDO, does not contribute much noise because of its lack of gain.

The procedure for estimating the output noise from a given LDO is first to refer each noise contributor to the op amp input. For example, to find the noise from the pass FET the engineer needs first to divide its noise contribution by the open-loop gain that exists between it and the op amp input. This gain is typically quite large, ensuring that the noise contribution from the pass FET is usually negligible.

Many engineers choose not to design an LDO from scratch, instead preferring to select from the wide range of proven modular solutions available. Nonetheless, it is still very important to ascertain how much noise the modular LDO will add to the hybrid-power supply.

While the manufacturer of the LDO is often tempted to tout the PSRR as the key performance figure of noise rejection, by carefully reading the small print of the datasheet the engineer will find information about how much noise is generated by the device itself over a relevant frequency range.

This information is usually indicated in two ways. The first is total (Integrated) output noise (in μVRMS), which is an RMS value of the spectral-noise density integrated over a finite-frequency range. The second is illustrated in the form of a spectral-noise-density curve (in μV/√Hz), which is a plot of noise density vs. frequency. Figure 3 shows this information from the datasheet of Texas Instruments’ TPS717xx series of linear regulators. The TPS71733, for example, provides a 3.3 V, 150 mA output from an input of up to 6.5 V and has a noise rating of 30 μVRMS over a 100 Hz to 100 kHz range.

Figure 3: PSRR, total output noise and spectral-noise-density curve from the datasheet for TI’s TPS717xx LDO.

Since a single number specifies the total-output-noise voltage, it is very useful for comparison purposes. However, when noise specifications of different LDOs are compared, it is important that the comparison is made over the same frequency range and at the same output voltage and current values.

Low-noise LDOs

There are some very-low-noise LDOs on the market. For example, Linear Technology’s LT3090––which offers a 0 to 32 V output at up to 600 mA from a 1.5 to 36 V supply––has a noise level of 18 μVRMS (see Figure 4).

Figure 4: Linear Technology’s low-noise LT3090 LDO.

Maxim’s MAX8510 can produce up to ten voltage outputs, ranging from 1.5 to 4.5 V, at up to 120 mA from an input voltage of up to 6 V. The noise rating of the device is 11 μVRMS over a 100 Hz to 100 kHz range.

For its part, Fairchild Semiconductor offers the FAN25800, which is specified with a noise rating of 8 μVRMS over a 10 Hz to 100 kHz range. The LDO operates from an input of up to 5.5 V and offers a 2.8 V output at up to 250 mA.

Both PSRR and LDO noise are important specifications to be taken into account when selecting an LDO for a hybrid-power supply. When comparing devices, it is important to choose a product with a good PSRR, as well as considering the frequency range over which the associated switching regulator will operate, to ensure it coincides with the optimum PSRR of the LDO. However, it is equally important to consider how much noise the LDO itself generates. There is little point in selecting a device that smoothes the voltage and current ripples from the switching regulator if the output supply is then corrupted by the LDO’s own internally-generated noise.

For more information on the parts discussed in this article, use the links provided to access product pages on the Hotenda website.

Further reading

  1. LDO Noise Demystified,” Sanjay Pithadia, Texas instruments Application Report SLAA412, June 2009.
  2. Understanding noise in linear regulators,” John C. Teel, Texas Instruments Analog Applications Journal, 2005.