Professional Documents
Culture Documents
Basic Analog
Power Supply
Design
by Gerard Fonte
Part 1
I
t is always important to remember that are important to mention.
that the power supply — either for a The first is that 117 VAC (Volts
particular product or as a general Alternating Current) is really an RMS (Root
There is the old saying: piece of test equipment — has the Mean Square) measurement. (Note that I
“You can give a man a fish potential to electrocute the user, start a have seen ordinary household power
and he will eat for a day or fire, or destroy the device it is powering. specified anywhere from 110 VAC to 125
you can teach a man to fish Obviously, these are not good things. For VAC. I just measured mine and found it
that reason, it is critical to approach this to be precisely 120.0 VAC.) An RMS meas-
and he will eat forever.” design conservatively. Provide plenty of urement of a sine wave is much lower than
There are many articles that margin for components. A well-designed the actual peak voltage and represents the
give the reader a specific power supply is one that is never noticed. equivalent DC (Direct Current) voltage
needed to provide the same power. The
design for building a power
Input Power Conversion RMS conversion varies according to the
supply. And it must be made wave shape and for a sine wave the value
clear that there is nothing Figure 1 shows the fundamental is 1.414. This means that the deviation
wrong with these cook-bbook design for a typical analog power around zero volts is actually 169.7 volts
supply. It consists of three main compo- (for my 120 VAC power). The power goes
designs. They often have very nents: Input power conversion and from -169.7 volts to +169.7 volts each
good performance. But they conditioning, rectification and filtering, cycle. Therefore, the peak-to-peak voltage
don’t teach the readers how and regulation. The input power conver- is actually 339.4 volts!
to design a power supply by sion is typically a power transformer This voltage becomes especially
and is the only method considered here. important when adding bypass capaci-
themselves. This two-ppart However, there are a couple of points tors to the main power lines to suppress
article will start from the noise from entering or leaving the
beginning and explain every FIGURE 1. A basic analog power supply power supply (a common situation). If
consists of three parts. The first two are you think the actual voltage is 120 volts,
step necessary to build a discussed in this article and the last in you may use 150 volt capacitors. As you
basic analog power supply. the next installment.
The design will focus on the
ubiquitous “three-tterminal
regulator” and include a
number of enhancements to
the basic design.
72 December 2007
Fonte.qxd 11/1/2007 11:35 AM Page 73
and are expensive. Occasionally, they 30 amps of surge current. With a bridge Filter PIV
Circuit
are used for high-voltage power supplies circuit, there are two diodes carrying Needs Factor Transformer Use
where suitable capacitors are expensive. this so worst-case is 21 amps each Half-Wave Large 2.82 50% (theoretical)
The formula for calculating the filter which is below the 30 amp specifica- Full-Wave Small 1.414 50% (theoretical)
capacitor (C) is quite simple, but you tion (assuming equal current sharing, Bridge Small 1.414 100% (theoretical)
need to know the acceptable peak-to- which is not always the case). But, this TABLE 1. A summary of the characteristics
peak ripple voltage (V), half-cycle time is an extreme example. Generally, a fac- of the various rectifier circuits.
(T), and current drawn (I). The formula is tor of about 10 is used, instead of 21.
C=I*T/V, where C is in microfarads, I is Nevertheless, this current surge is rectifiers and capacitors. (The ripple
in milliamps, T is in milliseconds, and V not something to be ignored. should be about two volts, as shown.)
is in volts. The half-cycle time for 60 Hz Spending a few cents more to use a The bridge rectifier must be able
is 8.3 milliseconds (reference: 1997 three-amp bridge instead of a one- to handle the high initial current surge.
Radio Amateur’s Handbook). amp bridge may be money well spent. A two-amp unit costs about $0.57,
It is clear from the formula that the a four-amp bridge is $0.77, and a six-
filtering requirements are increased for Practical Design amp device is $0.87. It seems obvious
high current and/or low ripple power that spending an additional dime for
supplies. But this is just common sense. We can now put these rules and improved reliability is worthwhile.
An easy-to-remember example is 3,000 principles to use and start to design a Note that the bridge is specified by
microfarads per ampere of current will basic power supply. We will use the 25.2 what the transformer can supply rather
provide about three volts of ripple. You volt transformer as the core of the design. than what the power supply is eventu-
can work various ratios from this exam- Figure 6 can be seen as a composite of ally specified for. This is done in case
ple to provide reasonable estimates of the previous figures but with practical there is an output short. In such a
what you need fairly quickly. part values added. A second pilot light in case, the full current of the transformer
One important consideration is the the secondary indicates its status. It also will be passed through the diodes.
surge of current at turn-on. The filter shows if there is a charge on the capaci- Remember, a power supply failure is a
capacitors act as dead shorts until they tor. With such a large value, this is an bad thing. So design it to be robust.
get charged up. The larger the capacitors, important safety consideration. (Note
the greater this surge will be. The bigger that since this is a DC signal, the 1N4004 Conclusion
the transformer, the greater the surge will reverse voltage diode is not needed.)
be. For most low voltage analog power It may be cheaper to use two small- Details are an important considera-
supplies (<50 volts), the transformer er capacitors in parallel than one large tion in designing a power supply. Noting
winding resistance helps somewhat. The one. The working voltage for the capac- the difference between RMS voltage and
25.2 volt, two amp transformer has a itor must be at least 63 volts; 50 volts is peak voltage is critical in determining the
measured secondary resistance of 0.6 not enough margin for the 40 volt proper working voltages for the supply.
ohms. This limits the maximum inrush to peak. A 50 volt unit provides only 25% Additionally, the initial surge current is
42 amps. Additionally, the inductance of margin. This may be fine for a non- something that cannot be ignored.
the transformer reduces this somewhat. critical application, but if the capacitor Next month, we will complete this
However, there is still a large potential fails here, the results can be catastroph- project by adding a three-terminal regula-
current surge at turn-on. ic. A 63 volt capacitor provides about tor. We will design a general-purpose,
The good news is that modern sili- 60% margin while a 100 volt device current-limited, adjustable voltage power
con rectifiers often have huge surge cur- gives 150% margin. For power supply with remote shut-off. Additionally,
rent capabilities. The standard 1N400x supplies, a general rule of thumb is the principles used for this design can be
family of diodes is usually specified with between 50% and 100% margin for the applied to any power supply design. NV
December 2007 75
Fonte2.qxd 11/29/2007 10:54 AM Page 60
Basic Analog
Power Supply
Design
by Gerard Fonte
H
owever, it should be remembered non-inverting input of an op-amp. The
FIGURE 2.
The practical
implementation
of the LM317
adjustable regulator
requires a few FIGURE 3. By
additional substituting the load
components for the adjustment
for optimal resistor, the LM317
performance and becomes a current
reliability. The regulator (the voltage
text describes the changes according
additional parts to the load). This is a
and their function. very useful circuit
that is not often used.
be higher than the
output voltage for proper operation. voltage of adjustable regulators. switching power supply which is
For the standard three-terminal volt- However, we’ll see a cheap and easy typically quite noisy.
age regulators, this voltage is about solution to this problem, later. The diodes are used to protect
1-2.5 V, depending on temperature the regulator against unexpected
and current drawn. (However, for Additional Components voltage reversals. If the input should
proper design purposes the 2.5V Needed be shorted to ground, capacitors on
“drop-out” voltage should always be the output and/or the adjustment pin
employed.) For the fixed voltage regulator, will still maintain their voltage. They
If the desired input-output voltage only an input and output bypass will then discharge backwards through
difference is less than this 2.5 V, capacitor (connected to ground) are the respective pins towards the input.
the circuit will no longer have the required (typically 0.1 to 1.0 μF). This can destroy the device.
headroom to regulate properly These are to stabilize the op-amp If there is no adjustment bypass
and drop out of regulation. (There in the regulator. If the large filter capacitor used (Cbyp), then the Dadj
are other “low-dropout” or LDO capacitor is in close physical proximity can be omitted. If you are sure that
regulators available that can operate (less than 6”) to the regulator input, the device being powered will never
with much less input-output differen- then the input bypass capacitor can have 10 μF or more load capacitance,
tial — some as low as 0.1 V. However, be omitted. then Do can be omitted, as well.
care must be taken to provide proper In theory, the output capacitor
capacitive loading or else they may can also be omitted. But if your circuit Adjustment Calculations
not function properly.) to be powered just happens to have
The third point is that it takes a load capacitance of 500 pF to It is important to remember that
some small amount of time for an 5,000 pF, then the regulator might (for adjustable regulators) the
output change to propagate through oscillate. Using a 1.0 μF output bypass adjustment pin acts as a non-inverting
the op-amp and transistor (loop capacitor forces the op-amp into a feedback point to the internal op-amp.
delay). During this time, the output stable operating mode. It is always a As such, the op-amp will do whatever
is not being properly regulated. good idea to spend a few pennies on it can to maintain the voltage on this
Obviously, the faster the op-amp and this to be sure you have a reliable pin (which is specified to be 1.25 volts
transistor, the shorter this transient power supply. below the output). So if we apply a
will be. However, manufacturing Figure 2 shows the proper design voltage (through a resistor) to this pin,
fast, high-power transistors is not easy for an adjustable regulator. The input the output will go up. If we short this
or cheap. and output bypass capacitors (Ci pin to ground, then the output will fall
Additionally, the faster the transis- and Co) are the same as in the fixed to about 1.25 volts. If a negative 1.25
tor and op-amp circuit is, the more regulator implementation. volts is applied, then the output will
prone to oscillation it is. Having your The two resistors (Rb and Radj) go all the way to zero volts.
power supply oscillate is not a good are required to choose the output Generally, a voltage divider is
thing. Most three-terminal regulators voltage. Typically, Rp is set at 240 used to pass some of the output
have a transient response (closed-loop ohms to provide a proper current voltage back into the regulator to set
delay time) of about 30 μs. (typically about 50 μA) into the the output voltage (as shown in Figure
The last point concerns adjustable feedback of the op-amp. Radj is 2). The equation is Vout = Vref (1 +
voltage regulators (Figure 1 shows a used to vary the output to the Rp/Radj) + (Iadj x Radj). Vref is 1.25 V
fixed-voltage regulator). Without desired value. The capacitor connect- as noted above. The second term (Iadj
getting into a long technical discus- ed to the adjustment pin (Cbyp) x Radj) is the error correction for the
sion, it is difficult to provide a stable is used to improve ripple/noise 1.25 V reference. Generally this is
and cheap reference voltage below rejection. This is especially useful if quite small and is often ignored. If this
1.2 V. This limits the minimum output the raw DC input is coming from a is ignored and the typical value of 240
January 2008 61
Fonte2.qxd 11/29/2007 10:55 AM Page 62
FIGURE 4. This is a nice bench power 32 volts, there will be a three volt
supply that provides 0-30 volts at drop which means that the regulator
one amp. It’s also current limited to must dissipate just three watts while
help eliminate unnecessary smoke supplying 32 W to the load. This is
generation in your circuits. Meters to
monitor the voltage and current can 91% efficient.
be added, if desired. It is clear that keeping the input
voltage as close to the output as
Ω is used for Rp, then the equation possible results in better efficiency
becomes Vout = 1.25 (1 + 240/ and less heat. (Note that with an input
Radj). Usually, this is accurate to voltage of 35 V and a current require-
within a few percent. ment of one amp the total power used
A curious thing happens will always be 35 W. The more power
when we replace Radj with a the load uses, the less the regulator
power load instead of a resistor must dissipate.)
(see Figure 3). The output voltage Often times, multiple voltages are
varies with the load. We see that needed for a circuit. There is nothing
this makes sense because the wrong with placing a five volt
power load variation is acting as a regulator after a 10 V one. In this way,
variable resistor. The result is that the voltage drop for the five volt
the current to the load remains regulator is shared between two
fixed but the voltage varies. In devices. However, if the two voltages
other words, the voltage regulator have significantly different current
has changed into a current regula- requirements, this approach may not
tor. Instead of a constant voltage, be optimal. (You can use Ohm’s Law
we have a constant current. to calculate the effective resistance
It is easily seen that the maxi- of the regulator given the desired
mum current allowed depends voltage and current. The power is then
upon Rp. If Rp is made 1.25 Ω, then calculated to be the resistance times
up to a full amp of current can be the current squared.)
supplied. Note that this current is
limited by the 1.5 A maximum Applying the Theory
output of the regulator and the
maximum output voltage obtainable We can now combine the infor-
(which depends upon the input mation from the two articles to build a
voltage). This current regulation is practical power supply. We’ll start
a very useful feature (of these with a basic supply and then discuss
regulators) that is not often used. how to change it to add features that
may be useful. Figure 4 provides a
Power Dissipation practical circuit for an adjustable,
current limited and voltage regulated,
Basically, the regulator acts like one amp power supply that goes from
a variable resistor in series with the zero volts to about 30 volts. (This
power supply. The amount of power particular circuit has proven to be very
dissipated is simply the voltage useful and adaptable in a number of
difference between the input and high-reliability designs for my clients,
output pins times the current usually with fixed current limit and
drawn. This creates some interesting fixed output voltage.)
situations. The raw DC circuit was described
For example, suppose you in detail last month and will not be
provide 35 V to the input of an discussed further except to note a
adjustable regulator and expect to few minor changes. The power trans-
draw one amp at five volts and former is now specified as a 24 volt,
also at 32 V. At five volts, there center-tapped, 1.0 to 1.5 amp device
will be a 30 V drop across the instead of 25.2 volts at two amps. The
regulator. With an amp of current, reason for this change is three-fold.
the regulator will have to dissipate The first is that some regulators
30 W while providing five watts to are limited to 35 V (most others
the load. It’s only 14% efficient. At are rated to 40 V and there is a high
62 January 2008
Fonte2.qxd 11/29/2007 10:55 AM Page 63
voltage version — LM-317H — that three watt resistor (rather than a be no heat problem. However, if the
can go to 60 V). Note that this is wo watt device) is chosen to be power supply needs to provide a high
technically the voltage differential conservative. The worst case power current at a low voltage, the switch is
between input and output, not requirement of the variable resistor used to connect to the center-tap and
from input to ground. If the output is occurs when it is equal to the fixed reduces the input voltage by 50%.
maintained above ground, then the resistor or 1.0 Ω. Thus, only about 18 volts is applied to
input can be higher. In this instance, there will be the rest of the power supply.
However, for a general-purpose about 0.4 W of power through it. This So, if five volts at one amp is
power supply you must assume the means that a one watt potentiometer needed, then only 13 W of power
output will be shorted to ground from is adequate, but a two watt unit will must be dissipated by the regulator as
time to time. This means that the input be used to be very conservative. heat (instead of about 30 W). This
voltage must be limited to about 37 V Remember, it’s better to spend a little allows the standard TO-220 style reg-
because there will be about 1.4 V bit extra on the power supply than to ulator to be used within the specified
dropped by the bridge diode and spend a lot on ruined circuits. limits. (If you want to get fancy,
an additional 2.5 V dropped by the The voltage regulator section is you could monitor the adjustment
current regulator. simply the design from the beginning voltage and use a separate circuit to
Also, there is no reason to provide of this article. The only change is to automatically switch between the
two amps of primary DC current if the use a variable resistor (R6) instead of a transformer windings.)
maximum output is rated at one amp. fixed resistor (Radj).
(Note that the rating is less than the Heatsinking is an important issue. Fixing the Minimum
absolute maximum the circuit can pro- The voltage regulator must be able to 1.2 Volt Problem
vide.) The third reason is for regulator dissipate up to 32 W of power, worst
heat control. This is also the reason case. This creates a significant Most people want a power supply
for the added switch which will be problem. The typical TO-220 case is to go all the way to ground. It turns
discussed in more detail below. limited to 15 W as per the spec sheet. out that there is an easy way to
The main filter capacitor has been One poor solution is to reduce the accomplish this: just add two diodes
reduced from 10,000 μF to 4,700 μF maximum power rating to about 0.5 A in series with the output. Each diode
because the power supply current for lower voltages. This is changing the drops the voltage by about 0.7 V
specification has been reduced from specs to match the design instead of (depending on load). So, two diodes
two amps to one amp. designing to match the specs. There is will reduce the output by 1.4 V. In
The adjustable current regula- another approach that requires a bit actual operation, only about 1.0 V is
tor is just three parts: an LM-317, a of creative engineering that will be dropped at no load. This makes the
fixed resistor, and a variable resistor. detailed below. minimum output no load voltage 0.2
The fixed resistor is necessary to limit Curiously, the current regulator is V. This is virtually zero.
the current through the resistor string. very efficient. The voltage drop is If you really want to, you can
Obviously, as the resistance gets typically 2.5 V (worst case) so the add another diode in series to drop
lower, more current flows. With a 1.25 maximum power dissipated at one the output further, but it seems
V differential between the output and amp is 2.5 W. A simple and inexpen- unnecessary. The resistor to ground is
adjustment pin, the one ohm fixed sive clip-on heatsink is sufficient here. included to drain off any leakage
resistor limits the current to 1.25 A. through the diodes. Without the
If this resistor was eliminated and Fixing the Heat Problem resistor, the output can float up to the
the adjustable resistor was set to 0.1 Ω voltage applied to the diodes.
for example, the current would be As noted, there is too much heat It should be noted that you can
(theoretically) 12.5 A (using Ohm’s to be dissipated at low voltages. The pull the output of the regulator to
Law). This would either destroy the IC problem comes from the large voltage ground if you apply a negative voltage
or cause it to shut down from thermal drop from the input to the output of to the adjustment pin. This is some-
overload. Neither is good. the regulator. The solution is to use what complicated if a negative voltage
Also, the power through that a center-tapped transformer and is not easily available (as in this case).
small resistance would be 15.6 W switch between taps. Center-tapped The addition of two diodes is a quick
(power is equal to current squared transformers are common and and easy fix.
times resistance). This would burn generally cost no more than non
the variable resistor. By forcing a center-tapped transformers. Minimum Current Discussion
minimum of 1.0 Ω in the circuit, When the switch is connected to
the current is limited to a maximum the top end of the transformer, the full As noted on the schematic
1.25 A and the variable resistor is voltage is applied to the rest of the shown in Figure 4, the minimum
protected. circuit and the full range of voltages is current limit is about 5 mA. For
The worst case power dissipated possible. As long as the current draw virtually all applications, this is not a
by the fixed resistor is 1.25 W. A is kept below about 0.5 A, there will problem. Very few components will
January 2008 63
Fonte2.qxd 11/29/2007 10:59 AM Page 64