Beat the Heat
Voltage regulators are a key technology for managing the heat dissipation in an electronic system. In this article, Jeff takes you through the evolution, science and functionality of voltage regulators. He looks at the heat and efficiency characteristics of voltage regulators, and compares linear versus switching technologies.
Acouple of resistors are often used to divide an input voltage into some lower output voltage. Each resistor drops a portion of the input voltage. This all works fine in a static situation. However, once the input voltage changes, so does the output voltage. That’s not a good voltage regulator. Substituting a Zener diode for the output resistor can help to keep the output voltage relatively stable, but as you know, this only works if the dynamic parameters all remain within a narrow range of conditions. And voltage regulation under varying conditions has always been a challenge.
The transistor made precision voltage regulators possible, but the required circuitry contained many discrete parts. During the 1960s, work began with the integration of transistors into an integrated circuit (IC), which was the combining of discrete parts and their interconnections onto a single substrate. This process enabled analog, digital and mixed-signal circuits to become functional blocks—instead of having to be built from scratch.
The first voltage regulator IC I used was the UA723 (Figure 1), developed by Fairchild Semiconductor, now Texas Instruments. Initially, many ICs were manufactured in the TO-100 metal can, but soon plastic took over as the packaging material of choice. Although this is a linear voltage regulator, it still requires a lot of support circuitry, and improvements quickly led to the familiar 3-terminal linear voltage regulator.
In the simplest case, a common collector amplifier—also known as “emitter follower” (Figure 2)—is used with its base connected directly to the voltage reference, in this case a Zener. The addition of the transistor to our original Zener circuit leads to a second path for current (as opposed to the original Zener circuit, where all current flows through the series resistor). Here, the second path handles the load current, while the Zener merely provides a voltage reference.
The stability of the output voltage can be significantly increased by using a comparator to monitor the difference between the reference voltage and output. Figure 3 is a block diagram of the functions used to create a fixed voltage regulator with only three terminals. This is commonly known as 78xx-style regulator. They come in a variety of packages, such as TO-92, TO-220 and TO-3, as well as SMT (surface mount) styles.
Some styles have a mechanism like a screw hole or an exposed surface to allow them to be attached to a heatsink. A linear voltage regulator provides a load with current at a fixed voltage, by dropping the excess voltage (the difference between the input and output voltage) across itself. This voltage multiplied by the current it supplies to the load can create a lot of heat. This is an issue that many designs face because of the specifications that may be required.
Unless our goal is to produce heat, any heat produced can be considered waste. For a 3-terminal linear regulator, heat is the byproduct of producing a stable regulated voltage. The heat produced is function of the voltage dropped across the regulator and the current supplied to the load (plus any current used by the regulator, itself). Let’s start with that.
The maximum bias current required to operate the typical 78xx linear regulator is around 8mA. This can vary a bit with input voltage, but we can use this for our discussion. Let’s assume we are using a 7805, which has a fixed 5.0V output and a 2V dropout voltage. This means it must have at least 5.0V + 2V or 7.0V minimum input voltage for the circuitry to regulate. With a 7.0V input and a bias current of 8mA, we have 7V × 0.008A or 0.056W of heat being generated.
Now let’s add a load of 100mA. This adds another 0.100A × 2V or 200mW to the quiescent power we calculated above, bringing the total to 256mW. If we are using a 12V input voltage, we now have a drop of 12V – 5V or 7V across the regulator. If we are using a 12V input voltage, we now have a drop of 12V – 5V or 7V across the regulator. The quiescent power becomes 12V × 0.008A = 96mW, and the power due to the load current becomes 7V × 0.100A = 700mW, for a total of 796mW. Now we are approaching 1W. Doesn’t sound like a lot huh? But 1W will give your finger a good burn if it’s from an SOT-233 packaged device (small-outline transistor)—but not so much if it is spread out over the surface of a TO-3.
Package style will affect a device’s ability to release this heat in some way, and is typically categorized as “thermal resistance” or TR (a measure of the ability of a device to dissipate heat from the surface of the die to the ambient air via all paths). I indicated in Figure 4 the thermal resistances for some case styles. Note the differences between RΘJA (junction-to-ambient) and RΘJB (junction-to-board) and RΘJC (junction-to-case). RΘJA is the TR for a device in contact with free air, whereas RΘJB/C is the TR for a device in contact with sufficient board (copper) or added heat sink.
This simply indicates that the junction of the device will be hotter than the ambient temperature by “X” degrees for every 1W of energy being dissipated. In the case of the SOT-223, if the ambient temperature is 25°C, then the junction will be 25 + 62.1 = 87.1°C for a device exposed to air, and 25 + 10.7 = 35.7°C for a device properly heat sunk. The datasheet lists the maximum junction temperature for this as 150°C. So, you can see that without a proper heatsink, one more watt will bring this close to the maximum. No one would run a device that close to its maximum rating. As the designer, you must decide how high you wish to run the junction, and design for a heatsink that will keep the junction below that temperature.
You may have designed a power supply using a TO-220 package. It is one of the most widely used package styles for a power supply. The large metal tab, which is at ground potential, has a screw hole that makes it easy to mount to a heat sink. It can be mounted against the PCB with a heat sink sandwiched between the two, or stood up vertically with the heat sink attached. It’s all about surface area and air flow. If you can’t get the heat away from the junction, your device is going to burn up.
You’ll note the second TRs listed in Figure 4 for the various devices are for perfect heat sinks. When you add a heatsink, you are attempting to lower the TR by increasing its ability to dissipate heat. If the current required and the input voltage produce a power that is low enough, you may not need a heatsink. But once that power begins to grow, you will find yourself requiring larger and larger heatsinks. At some point, you may need to add a fan to increase the air flow. Water cooling is being used to increase the ability to transfer heat away from a device. This is one reason data centers are air conditioned.
When designing power supplies that must work in higher temperatures—such as outside during summer—this becomes even more of a problem since the ambient air temperature may initially be 40°C. As the differential temperature drops between your heatsink and the ambient temperature, it becomes more difficult to transfer the heat.
Put your hand on the cable box, the Blu-ray player or other equipment in your home, and you’ll become aware how much heat is being generated by electronics. Heat is one of the largest factors behind equipment failure. We’ve seen that keeping the input voltage down to within 2V above the output voltage will conserve the most energy. The efficiency of your linear regulator is calculated as output power divided by input power. If we keep the input voltage to the minimum 2V over the output voltage, we can see in Table 1 that the efficiency goes down as we regulate to lower voltages.
It’s not uncommon to require multiple voltages in a system. For example, your system might require some regulated 12V in addition to your 3.3V processor. While the a 14V input might get you 85% efficiency for your 12V, the 3.3V would end up being 3.3/14 or only 23% efficient. You can think of low efficiency as wasted heat!
SWITCHERS TO THE RESCUE
Linear voltage regulators are inexpensive and easy to use. However, if you find yourself pulling out your hair because you need bigger heat sinks, more air flow—and as a result, larger packaging—you might consider using a switching regulator. Switching regulators try to match the output power requirement with just the right input power. They do this by interrupting the fixed voltage input. A buck converter is used where the DC output voltage is lower than the DC input voltage. The DC input can be derived from rectified AC or from any DC supply.
The standard linear power supply uses a mains-isolating transformer to drop the AC input voltage to a reasonable level, so that, when rectified, it can sustain the minimum voltage necessary for the linear regulator. The input capacitor is sized to supply the required output current while minimizing discharge between cycles, thus keeping the ripple above the dropout voltage. You can trade off increased input voltage for a smaller input capacitor, but that creates a larger drop and subsequently more heat to dissipate.
The linear power supply requires a steady DC input voltage, whereas the switching regulator just chops the input voltage into a high-frequency, pulsating DC voltage. The regulation comes from controlling of the duty cycle of this chopping action in conjunction with the flywheel circuitry (Figure 5). The switching device has two states, ON (supplying power to the flywheel circuitry) and OFF (idling). During the ON time (Figure 6), the current is impeded by the inductor, because it stores magnetic energy slowly charging the capacitor (the diode is reverse biased). During the OFF time (Figure 7), the inductor becomes the voltage source to the capacitor, because the diode is now forward biased.
With a DC voltage on VIN, if the duty cycle is 50%, then the Vout will be VIN × duty cycle / 100% or 0.5VIN. In theory, you can see that by varying the duty cycle, we can change VOUT from 0V to VIN. We just need some way to adjust this based on VOUT. Figure 8 is a functional diagram of how this is accomplished. A reference voltage and some part of VOUT are combined to create an error signal. The control logic determines a duty cycle to apply to switching circuitry. Note here that the upper FET will be ON during this phase to apply VIN. The lower FET takes the place of the diode and would be OFF during this phase. When the upper FET is turned OFF, the lower FET is turned ON to complete the path for the collapsing magnetic field’s current.
You’ll find many switching regulators that contain at least the control portion of the circuit, but then require external FETs, inductors, capacitors and other support parts. All of this can take up a fair amount of real estate. I’ve been designing some newer products using these switching regulators, especially when the input voltage is high and heat is a factor. When a system that has been running cool for years gets a few more peripherals added to it, the additional current required can be the straw that broke the camel’s back, and the extra heat becomes an issue. You can replace the linear regulator with a 3-terminal switcher to cool things right down again.
THE 3-TERMINAL SVR
Almost every semiconductor manufacturer now carries some form of 3-terminal switching voltage regulator (SVR). It’s the form factor that makes these devices unique (Figure 9). Whether you stand them up or lay them down, they won’t require any more real estate than the linear regulator they are replacing.
Of course, there are some drawbacks. When you see the word “switching,” you probably think noise. And yes, the ripple frequency is based on the clocking of the switcher. If there are particular frequencies used in your design—let’s say 40kHz because you’re doing some IR stuff—then you will want to choose an SVR that avoids that switching frequency (or a multiple of it), so as to cause a minimum of interference. Let’s compare a few of these devices to the 78xx LVR, to see how they size up. Table 2 shows my interpreted datasheet comparisons between LVR (first item—the STMicroelectronics part) and SVR (remaining items).
The first thing to notice here is the maximum peak-to-peak (pk-pk) ripple of 50µV for the linear regulator versus values in the millivolt range for all the SVR device. Other than the price, the specs are similar. Remember: the cheaper linear price does not include the heatsink that might be necessary—and which can cost much more than the device itself—not to mention any other air flow requirements.
You’ll find a ton of additional information on manufacturers’ datasheets about how you can improve the ripple/line/load regulation, if necessary, with external inductors/capacitors, if you need a cleaner output. You may find that the external discretes you already are using for the linear regulator may be all you need.
I found the efficiency of using the SVR can consistently reduce the power required from your power input circuitry. Whether it’s your own transformer and rectification or a plug-in Wall Wart, it’s a pleasant surprise to find the equipment is no longer throwing off wasted heat. The biggest improvement is for circuits that require multiple voltages, for which using an SVR on at least the lower voltage output can have the largest impact. Table 3 is a general comparison of the linear versus switching regulator.
I know you’ll find these helpful in improving existing equipment. And they can be part of your bag of tricks, potentially useful when looking at options for new circuitry. I doubt these will kill the linear regulator, but they can be helpful when you have a heat problem.
One of my latest designs requires only 100mA at 5V. It must be able to operate over a 12VDC to 24VDC range. Using a linear regulator here requires getting rid of almost 2W of heat. The enclosure size and weight requirements would be compromised if I used a linear regulator. The use of these small switching regulators saved this project from becoming a boat anchor. These 3-terminal switching regulators have essentially eliminated the worry over having to get rid of excessive heat. I encourage you to try them and add them to your own bag of tricks.
As always, too much to do, too little time. Oh yeah, and remember to wash your hands!
Figure 5 and Figure 7
CUI | www.cui.com
Monolithic Power Systems | www.monolithicpower.com
Murata Power Solutions | www.murata-ps.com
RECOM | www.recom-power.com
STMicroelectronics | www.st.com
Texas Instruments | www.ti.com
Traco Power | www.tracopower.com
Würth Elektronik | www.we-online.com
XP Power | www.xppower.com
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • NOVEMBER 2020 #364 – Get a PDF of the issueSponsor this Article
Jeff Bachiochi (pronounced BAH-key-AH-key) has been writing for Circuit Cellar since 1988. His background includes product design and manufacturing. You can reach him at: email@example.com or at: www.imaginethatnow.com.