Switching Voltage Regulator

Ivan Bercovich Jeffrey Little Taylor Caggiano

2008

ECE 324 – Electronics II – Term Project

Switching Voltage Regulator Group 17

The goal of our group was to design and build a switching voltage regulator while meeting the following specifications: adjustable output voltage from 2V to 10V independent of the load, capable of sourcing 120mA at any voltage in the specified range, maintain the voltage ripple on the output smaller than 50mV, and the efficiency of the circuit must be better than 50% at full load over the entire range. Our final design met and surpassed all the specifications. The efficiency was 50.1% at 2V, and 80% at 10V. Our output had a ripple of 30mV, the design was able to source 145mA at 10V (69 Ω load) and 150mA at 2V (13Ω load).

Table of Contents Table of Figures ............................................................................................................................................. 3 Project Goals ................................................................................................................................................. 4 Achievement of Goals ................................................................................................................................... 6 Project Description........................................................................................................................................ 7 Reference Voltage..................................................................................................................................... 7 Error Amplifier .......................................................................................................................................... 8 Pulse Width Modulation ........................................................................................................................... 8 Switch ........................................................................................................................................................ 9 Filter ........................................................................................................................................................ 11 Feedback ................................................................................................................................................. 11 Schematic ................................................................................................................................................ 12 Debugging ................................................................................................................................................... 12 Parts List and Cost ....................................................................................................................................... 13 Conclusion and Final Thoughts ................................................................................................................... 14 Bibliography ................................................................................................................................................ 15

2|Page

Table of Figures Figure 1: Measure total power used by circuit ............................................................................................. 4 Figure 2: Ripple at the output with: 0.56µF capacitor (left) and 680µF capacitor (right) ........................... 5 Figure 3: Project Specs vs. Actual Performance............................................................................................ 6 Figure 4: Efficiency vs. Vout (120mA) ........................................................................................................... 7 Figure 5: Block diagram of the Switching Voltage Regulator........................................................................ 7 Figure 6: Overlap of Pulse Wave, Triangle Wave, and Error ......................................................................... 9 Figure 7: Sziklai Pair .................................................................................................................................... 10 Figure 8: Simplified block diagram of a feedback loop ............................................................................... 11 Figure 9: PSPICE schematic of the Switching Voltage Regulator ................................................................ 12 Figure 10: Parts List and Cost ...................................................................................................................... 13

3|Page

Project Goals Our switching voltage regulator was to be designed so that it could deliver up to 120 mA of current to the load while maintaining a fixed output voltage, adjustable from 2V to 10V. A ripple was expected to appear at the output, because of the switching, with a frequency equivalent to that of the triangle wave, in our case ~32kHz. The project specifications required us to design a LC filter that would limit the peak to peak magnitude of this ripple to a maximum of 50 mV. This filter capacitor is also responsible for maintaining the voltage across the output when the switch is open, and this along with its filtering capabilities, had to be considered when choosing an appropriate value. In order to justify the construction of a switching voltage regulator, we had to make it more efficient than the average linear voltage regulator; this is why we were required to maintain a minimum efficiency of 50% across the entire range (at full load). Since many linear voltage regulators can be up to 60% efficient, there was an expectation to achieve even more efficiency at higher output voltages. Efficiency is the ratio of the (power dissipated by load) / (power delivered to the entire circuit) and was measured and checked to be greater than 0.5 at full load. We measured the power delivered to the board by measuring the current going into the board’s power strip and multiplying it by the voltage at the power strip. We measured the current going into the board as directed in the project description, seen below in Figure 1.

Figure 1: Measure total power used by circuit

To meet the voltage ripple specification we had to choose the size of the capacitor at the output according to the following equation found in “Soclof, S., Design and Applications of

4|Page

Analog Integrated Circuits”:

. (The frequency was selected to be 32kHz).

Also, we had a limit to the different size inductors we could use. We were given two 2.2mH inductors with our kit, and there were inductors of sizes 0.4mH, and 47mH in the lab. After deciding to use the two 2.2mH inductors in parallel (giving us an inductance of 1.1mH), calculating the size of the capacitor using the above equation (Vout being between 2V and 10V) we get a calculated capacitance of approximately 10µF. To make sure that our calculations were correct, we probed the output of our circuit with the oscilloscope on x10 gain, and made sure that our ripple was less than 50mV. It should be noted that although we calculated the capacitance to be 10µF, this value capacitor was not what we ended up using. The reason for this is that since the capacitor's job was not only to filter out noise, but also to keep the output voltage constant, we had to make sure that the value we picked was able to handle both responsibilities. Please refer to Figure below to see that with our chosen capacitor, we are meeting the ripple specification.

Figure 2: Ripple at the output with: 0.56µF capacitor (left) and 680µF capacitor (right)

The hardest goal to meet was the efficiency specification. To meet this goal we had to make sure that the ratio of the (power dissipated by load) / (power delivered to the entire circuit) was more than 0.5. We measured the power delivered to the board by measuring the current going into the board’s power strip and multiplying it by the voltage at the power strip. We measured the current going into the board as directed in the project description, seen below in Figure 1.

5|Page

Achievement of Goals Our circuit was able to meet, and extend beyond the limits of the specifications. Below, Table 1, is a chart of the specifications and the experimental data gathered by our group.We went through a lot of debugging to meet our goals, especially the efficiency specification. Also, once we met the specifications we took an Edison like approach to see how far we could push our circuit beyond the specifications: This approach yielded desirable results, since we were able to surpass the specifications. Seen below in Table 1, we were able to adjust our output voltage from 1.85V to 11.5V, and we were also able to source 145mA over the full range. Our efficiency at the high end was also significantly greater than what was required. Specification Output adjustable from 2V to 10V Source 120mA over full voltage range. Output voltage ripple of < 50mV Efficiency > 50%, η=(Power Supplied to the load)/(Total Power Consumed)

Measured

Output adjustable from 1.85V to 11.5V Sources 145mA over full range. 30mV output ripple 50.1% at 2V, 80% @ 10V

Figure 3: Project Specs vs. Actual Performance

It was also mentioned in the lab description that good switching voltage regulators should have “low” dynamic output resistance defined as

. We measured this quantity by leaving

the load constant and varying the output voltage, calculating the current, and then solving the above equation. The result was an output resistance of about Ω, which 2 satisfied the suggestion of building the circuit to have “low” dynamic output resistance. To see how our efficiency varied with output voltage, we tested our circuit in the lab and recorded our findings in Figure 4 in the next page.

6|Page

Figure 4: Efficiency vs. Vout (120mA)

As you can see, we had superior efficiency at high output voltages, and even though we had significantly worse efficiency at low voltages, we still were able to meet the required efficiency level of 50%.

Project Description

Figure 5: Block diagram of the Switching Voltage Regulator

Reference Voltage In order to move through the range of specified output voltages, we had to choose a variable component. After analyzing the schematics and studying the design, we came to the conclusion

7|Page

that the variable component had to be either the feedback gain or the reference voltage. After performing some PSPICE simulations, we decided that changing the reference voltage would yield better final results and further simplify the design. The reference voltage was designated using the board's unregulated 13 Volt supply and Ω a 50k

potentiometer. We chose the

unregulated supply to be 13V because this was the lowest voltage that allowed full functionality while minimizing power consumption (since P=I×V). The potentiometer, used to adjust this voltage, does not have to be 50k and was actually a 2k pot for most of our design and debugging. We originally chose a 2k Ω potentiometer because it allowed us to pick more precise reference voltages. It was not until the efficiency became a concern did we actually put thought into this component. Since power is V2/R, the larger R is, the less power the part will dissipate, so making the Pot 50K Ωs saved a considerable amount of power. This reference voltage is compared to the feedback (Vout) of the circuit and used to determine an error voltage.

Error Amplifier The comparison of the reference voltage to the feedback from the output is what controls our circuit. There is an "error amplifier," a 741 Op Amp, with a linear gain of 50V/V. Originally, we did not realize that we had to limit the gain of the error amp. However, after Prof. Oliaei pointed out that infinite gain across the OpAmp will produce instability and oscillations in our circuit, we chose to build feedback around the OpAmp giving us a gain of 10 V/V. During the debugging process, we changed the value of the OpAmp gain until we found one that seemed to make our circuit work best (settling on 50 V/V). The resistor values chosen to achieve a 50 V/V gain were 50kΩ and 1kΩ. Finally, the output of the error amp, which was the difference of the non-inverting and inverting terminals of the OpAmp multiplied by 50, was fed into the noninverting terminal of a LM339 comparator.

Pulse Width Modulation The pulse width modulator consists of three basic parts: the output of the error amplifier, a triangle wave, and a comparator (LM339). The way the comparator works is that the output of it swings to the positive rail (13V) when the non-inverting input is larger than the inverting input, and swings to the negative rail (0V) if the inverting input is higher than the non-inverting 8|Page

input. Therefore, by inputing a DC voltage from the error amp into the non-inverting terminal, and a triangle wave into the inverting terminal, the two are compared, and the output will be a pulse that is high (13V) for the duration of time the error is higher than the triangle wave, and low (0V) Figure 6: Overlap of Pulse Wave, Triangle Wave, and Error

for the time when the error is less than the

triangle wave, see Figure 4 above. This functionality is desirable since the greater the error is, the wider (higher duty cycle) the pulse will be, and vice versa. To create the triangle wave, we built a square wave generator that works by oscillating between its positive and negative rail depending upon the polarity across the capacitor (See diagram below). This square wave is then run through an integrator circuit built with a 741 OpAmp, and the result is a triangle wave. The frequency of the triangle wave is determined by the equation: which was found in an Analog Device's Application note on Pulse Width Modulation [ref].

Switch The switch is what allows the output to be regulated; the state of the switch determines whether or not current should be supplied to the load—higher supplied currents will cause a wider duty cycle, so for an open circuit setup, the pulse width will be negligible. The duty cycle of the pulse determines how long the switch should stay on, delivering current to the output stages of the circuit. At first, the switch was thought of as a trivial part of our design, but turned out to be an engineering challenge. It seemed like a simple NPN BJT would allow for proper functionality, however, this turned out to be one of our biggest design issues. There were two main problems with the single transistor switch: (1) the base current necessary to saturate the transistor was very high, which had adverse effects on the pulse signal; (2) the protection diode (more details below) forced a voltage of -0.7V at the transistor's emitter when the switch was off. Since when

9|Page

the switch is off, the base is at 0V, and with the diode forcing a voltage of -0.7V on the emitter, the transistor would be immediately turned back on since Vbe required to turn on the BJT is exactly 0.7V. First, to deal with the large base current, we realized we had to find a way to increase the current gain, also known as β, of our switch. Some research was done to find some design ideas to achieve the desired current output (without draining too much current from the base), one design option was the Darlington pair. However, putting the transistors in the Darlington configuration gave us a drop of 1.4V between the base and emitter (across the two BJTs). We looked online and found a complementary Darlington pair configuration, better known as Sziklai Pair, seen in the

Figure 7: Sziklai Pair

figure on the left. The Sziklai pair is not only good from the current gain (now β2) and the voltage drop (which is only 0.7v) perspectives; it also has a faster rise time, which makes it more efficient at switching. Although the Sziklai pair deals with the problem related to the base current, we were still looking for a solution to the second problem, given that Vbe for the Sziklai pair is the same as a single NPN BJT. We had two options: we could make our pulse wave (comparator ouput) reach slightly negative voltages when it turned off, effectively turning the switch off even with the 0.7V at the emitter, or we could find a diode that had a lower voltage drop when conducting current. The first diode placed at the emitter of our switch was the 1N914 (0.7V Fwd Voltage Drop) diode, which was available in the lab. This was the diode that presented the problem that when there was 0V at the base of our switch the voltage drop from base to emitter was .7V (0V(-0.7V from diode)) and the switch was still on. This was the main reason for our inefficiencies. A Schottky diode would provide a low forward voltage drop (about 0.2V) and fast switching also allowing us to leave our supply voltage at 13V, eliminating the need for the negative voltage required for the alternate model. The Schottky diode we chose was the 1N5819, which one of our group members obtained from his work on a previous project. This diode provides

10 | P a g e

protection for our open switch and creates a path to ground for the current, which is maintained by the inductor, to be able to keep flowing to the load while the switch is off.

Filter The filter is comprised of a 1.1mH inductor (two 2.2mH inductors in parallel) and a 680µF capacitor. The capacitor is responsible for filtering out the ripple caused by the 32kHz switching frequency as well as maintaining the voltage at the output when the switch is open. As discussed previously the calculated capacitor value to filter out the switching noise was approximately 10µF

. Nevertheless, we realized that the capacitor value would

need to be significantly larger in order to maintain the desired output voltage for the entire period when the switch is off. The efficiency was a major concern and it was important that our circuit could maintain the output for as long as possible in order to keep or switch off (ie not consuming power).

Feedback The feedback for our circuit consisted of nothing more than a connection form the output to the error amplifier. Although the project description hinted towards the idea of using a voltage divider to make Vout = k*Vref (k>1), we saw that we could make our feedback work without this element. Towards the end of the debugging period, we tried several resistor combinations Figure 8: Simplified block diagram of a feedback loop

to test the advantages of a loop gain smaller

than 1. Nevertheless, we did not see any advantages for using this configuration, and we went back to the simpler model with a unity gain.

11 | P a g e

Schematic

Figure 9: PSPICE schematic of the Switching Voltage Regulator

Debugging The majority of the debugging was due to things that had been overlooked or thought to be simple, not misconceptions. The problem with the power consumption due to the lack of switching was a major issue that actually took us right up to the end to figure out. Also, our initial design did not provide the required output current. The research of different BJT configurations was done in order to provide a larger current gain. Other changes that we made to make sure that we would meet the efficiency specification were that we rose to the size of the potentiometer from 2kΩ to 50kΩ and we lowere d the voltage on the board from 15V to 13V. One of the most frustrating problems that we had to deal with was early in the construction process. We could not figure out why our output voltage wouldn't go above 5.1V. We looked back at our design and everything seemed fine, and we couldn't figure out what the problem was. We thought the diode may have been broken, so we got a new one out of the 1N914 drawer in the lab. After switching out the diode everything was the same, so we figured that

12 | P a g e

wasn't the problem. It turns out that a bag of 5.1V Zener diodes had been mistakenly put in the 1N914 drawer, which is why our output wouldn't go above the voltage drop associated with the diode. This was a frustrating problem that set us back a couple of lab periods, but once we corrected the error, things ran more smoothly.

Parts List and Cost (Prices not included in the Lab description were found in DigiKey; since the price ranges depend on quantity, we used the price corresponding to hundreds of parts) Part

Price

Usage

Potentiometer

$4.00

Reference Voltage

741 OpAmp 2 x 50kΩ Resistor 2 x 1kΩ Resistor

$0.22 $0.04 $0.04

50 V/V Linear Amplifier 51 V/V Linear Amplifier 52 V/V Linear Amplifier

Capacitor 6 x Varied Resistor 741 OpAmp 1/2 339 Comparator

$0.10 $0.12 $0.22 $0.14

Triangle Wave Triangle Wave Triangle Wave Triangle Wave

NPN BJT PNP BJT Shottky Diode 1/2 339 Comparator

$0.20 $0.20 $0.10 $0.14

Switch Switch Switch Switch

2 x 2.2mH inductor 680 µF capacitor

$1.74 $0.10

Filter Filter

$7.36

Total

Figure 10: Parts List and Cost

As you can see, the majority of the cost lies in the potentiometer that is used to adjust Vref. After seeing approximately how much our designed circuit cost, we thought it would be interesting to look online and see how much a chip manµFacturer charges for a switching voltage regulator. Searching through Analog Devices' website, we found the ADP1111, which is a switching regulator with an adjustable output from about 3V to 12V, it has a maximum switching current of 1.5A, and its switching frequency was 70kHz. The cost of the chip is $2.06, which is slightly cheaper than our circuit if we could have found a more cost effective way to vary the reference voltage.

13 | P a g e

Conclusion and Final Thoughts Although this project might seem simple at first glance, it is quite hard to meet all the specifications and acquire a comprehensive understanding of the components "under the hood". Unlike a project involving mostly digital components, our voltage regulator was entirely analog. This means that it is a lot harder to judge how well the product is working or if it is working at all. Additionally, in an analog design, the engineers have to be a lot more creative when trying to debug the design or when tuning the component values in order to gain better performance. This project was a great experience, because by the time we were done, we were able to describe every single aspect of it from the logical and from the physical points of view. We also acquired a comprehensive understanding, not only about the steady state behavior, but also regarding the transient response. The night before the presentation, the three of us sat in M5 and discussed possible questions that might come up during the demo; we were impressed with the degree of our understanding and the quality of each other's responses. If we had more time to work on this project, we all agree that further improvements in power consumption could be made, especially on the low end, around 2v. This power loss comes from the high voltage across the Sziklai pair and the significant current going through the switch at the same time. To solve this problem, we could have made a variable Vcc on top of the Szicklai pair, which is proportional to the output voltage (just enough to maximize the current flow). Because in our design the reference voltage is similar to the output voltage, we could have powered the switch with the reference voltage summed to a fixed DC value (by making use of a summer).

14 | P a g e

Bibliography Beards, H. (1987). Analog and Digital Electronics. Prentice Hall. Savant, R. &. (1991). Electronic Design. Benjamin Cummings. Smith, A. S. (2004). Microelectronic Circuits (5th ed.). Oxford University Press. Soclof, S. (1991). Design and Applications of Analog Integrated Circuits. Prentice Hall.

15 | P a g e