The new biography
Steve Jobs contains a remarkable claim about the power supply of the Apple II and its designer Rod Holt:
[1]
Instead of a conventional linear power supply, Holt built one like those
used in oscilloscopes. It switched the power on and off not sixty times
per second, but thousands of times; this allowed it to store the power
for far less time, and thus throw off less heat. "That switching power
supply was as revolutionary as the Apple II logic board was," Jobs later
said. "Rod doesn't get a lot of credit for this in the history books
but he should. Every computer now uses switching power supplies, and
they all rip off Rod Holt's design."
I found it amazing to think that computers now use power supplies based
on the Apple II's design, so I did some investigation. It turns out that
Apple's power supply was not revolutionary, either in the concept of
using a switching power supply for computers or in the specific design
of the power supply.
Modern computer power supplies are totally different and do not rip off
anything from Rod Holt's design. It turns out that Steve Jobs was making
his customary claim that everyone is stealing Apple's revolutionary
technology, entirely contrary to the facts.
The history of switching power supplies turns out to be pretty
interesting. While most people view the power supply as a boring metal
box, there's actually a lot of technological development behind it.
There was, in fact, a revolution in power supplies in the late 1960s
through the mid 1970s as switching power supplies took over from simple
but inefficient linear power supplies, but this was a few years before
the Apple II came out in 1977. The credit for this revolution should go
to advances in semiconductor technology, specifically improvements in
switching transistors, and then innovative ICs to control switching
power supplies.
[2]
Some background on power supplies
In a standard desktop computer, the power supply converts AC line
voltage into DC, providing several carefully regulated low voltages at
high currents. Power supplies can be built in a variety of ways, but
linear and switching power supplies are the two techniques relevant to
this discussion. (See the notes for more about obsolete technologies
such as large
mechanical motor-generator systems
[3] and ferroresonant transformers
[4][5].)
A typical linear power supply uses a bulky power transformer to convert
the 120V AC into a low AC voltage, converts this to low voltage DC with a
diode bridge, and then uses a linear regulator to drop the voltage to
the desired level. The linear regulator is an inexpensive easy-to-use
transistor-based component that turns the excess voltage into waste heat
to produce a stable output. Linear power supplies are almost trivial
to design and build.
[6] One big disadvantage however, is they typically waste about 50-65% of the power as heat,
[7]
often requiring large metal heat sinks or fans to get rid of the heat.
The second disadvantage is they are large and heavy. On the plus side,
the components (other than the transformer) in linear power supplies
only need to handle low voltages and the output is very stable and
noise-free.
A switching power supply works on a very different principle: rapidly
turning the power on and off, rather than turning excess power into
heat. In a switching power supply, the AC line input is converted to
high-voltage DC, and then the power supply switches the DC on and off
thousands of times a second, carefully controlling the time of the
switching so the output voltage averages out to the desired value.
Theoretically, no power gets wasted, although in practice the efficiency
will be 80%-90%. Switching power supplies are much more efficient, give
off much less heat, and are much smaller and lighter than linear power
supplies.
The main disadvantage of a switching power supply is it is considerably
more complex than a linear power supply and much harder to design.
[8]
In addition, it is much more demanding on the components, requiring
transistors that can efficiently switch on and off at high speed under
high power. The switches, inductors, and capacitors in a switching
power supply can be arranged in several different arrangements (or
topologies), with names such as Buck, Boost, Flyback, Forward,
Push-Pull, Half Wave, and Full-Wave.
[9]
History of switching power supplies to 1977
Switching power supply principles were known since the 1930s
[6] and were being built out of discrete components in the 1950s.
[10] In 1958, the IBM 704 computer used a primitive vacuum-tube based switching regulator.
[11] The company Pioneer Magnetics started building switching power supplies in 1958
[12] (and decades later made a key innovation in PC power supplies
[13]). General Electric published an early switching power supply design in 1959.
[14] In the 1960s the aerospace industry and NASA
[15]
were the main driving force behind switching power supply development,
since the advantages of small size and high efficiency made up for the
high cost.
[16] For example, NASA used switching supplies for satellites
[17][18] such as Telstar in 1962.
[19]
The computer industry started using switching power supplies in the late 1960s and they steadily grew in popularity.
Examples include the PDP-11/20 minicomputer in 1969,
[20] the Honeywell H316R in 1970,
[21] and Hewlett-Packard's 2100A minicomputer in 1971.
[22][23]
By 1971, companies using switching regulators "read like a 'Who's Who'
of the computer industry: IBM, Honeywell, Univac, DEC, Burroughs, and
RCA, to name a few."
[21]
In 1974, HP used a switching power supply for the 21MX minicomputer,
[24] Data General for the Nova 2/4,
[25] Texas Instruments for the 960B,
[26] and Interdata for their minicomputers.
[27]
In 1975, HP used an off-line switching power supply in the HP2640A display terminal,
[28] Matsushita for their traffic control minicomputer,
[29] and IBM for its typewriter-like Selectric Composer
[29] and for the IBM 5100 portable computer.
[30]
By 1976, Data General was using switching supplies for half their systems, Hitachi and Ferranti were using them,
[29] Hewlett-Packard's 9825A Desktop Computer
[31] and 9815A Calculator
[32] used them, and the decsystem 20
[33]
used a large switching power supply. By 1976, switching power supplies
were showing up in living rooms, powering color television receivers.
[34][35]
Switching power supplies also became popular products for power supply manufacturers starting in the late 1960s.
In 1967, RO Associates introduced the first 20Khz switching power supply product,
[36] which they claim was also the first switching power supply to be commercially successful.
[37] NEMIC started developing standardized switching power supplies in Japan in 1970.
[38]
By 1972, most power supply manufacturers were offering switching power supplies or were about to offer them.
[5][39][40][41][42]
HP sold a line of 300W switching power supplies in 1973,
[43] and a compact 500W switching power supply
[44] and a 110W fanless switching power supply
[45]
in 1975. By 1975, switching power supplies were 8% of the power supply
market and growing rapidly, driven by improved components and the
desire for smaller power supplies for products such as microcomputers.
[46]
Switching power supplies were featured in electronics magazines of this era, both in advertisements and articles.
Electronic Design recommended switching power supplies in 1964 for better efficiency.
[47]
The October 1971 cover of
Electronics World featured a 500W switching power supply and an article "The Switching Regulator Power Supply".
A long article about power supplies in
Computer Design in 1972
discussed switching power supplies in detail and the increasing use of
switching power supplies in computers, although it mentions some
companies were still skeptical about switching power supplies.
[5]
In 1973,
Electronic Engineering featured a detailed article "Switching power supplies: why and how".
[42]
In 1976, the cover of Electronic Design
[48] was titled "Suddenly it's easier to switch" describing the new switching power supply controller ICs,
Electronics ran a long article on switching power supplies,
[29]
Powertec ran two-page ads on the advantages of their switching power
supplies with the catchphrase "The big switch is to switchers",
[49] and
Byte magazine announced Boschert's switching power supplies for microcomputers.
[50]
A key developer of switching power supplies was Robert Boschert, who
quit his job and started building power supplies on his kitchen table in
1970.
[51]
He focused on simplifying switching power supplies to make them
cost-competitive with linear power supplies, and by 1974 he was
producing low-cost power supplies in volume for printers,
[51][52] which was followed by a a low-cost 80W switching power supply in 1976.
[50] By 1977 Boschert Inc had grown to a 650-person company
[51] that producing power supplies for satellites and the F-14 fighter aircraft,
[53] followed by power supplies for companies such as HP
[54]
and Sun.
People often think of the present as a unique time for technology
startups, but Boschert illustrates that kitchen-table startups were
happening even 40 years ago.
The advance of the switching power supply during the 1970s was largely driven by new components.
[55]
The voltage rating of switching transistors was often the limiting factor,
[5]
so the introduction of high voltage, high speed, high power transistors
at a low cost in the late 1960s and early 1970s greatly increased the
popularity of switching power supplies.
[5][6][21][16]
Transistor technology moved so fast that a 500W commercial power supply featured on the cover of
Electronics World in 1971 couldn't have been built with the transistors of just 18 months earlier.
[21]
Once power transistors could handle hundreds of volts, power supplies
could eliminate the heavy 60 Hz power transformer and run "off-line"
directly from line voltage. Faster transistor switching speeds allowed
more efficient and much smaller power supplies.
The introduction of integrated circuits to control switching power
supplies in 1976 is widely viewed as ushering in the age of switching
power supplies by drastically simplifying them.
[10][56]
By the early 1970s, it was clear that a revolution was taking place.
Power supply manufacturer Walt Hirschberg claimed in 1973 that "The
revolution in power supply design now under way will not be complete
until the 60-Hz transformer has been almost entirely replaced."
[57]
In 1977, an influential power supply book said that "switching
regulators were viewed as in the process of revolutionizing the power
supply industry".
[58]
The Apple II and its power supply
The Apple II personal computer was introduced in 1977. One of its
features was a compact, fanless switching power supply, which provided
38W of power at 5, 12, -5, and -12 volts. Holt's Apple II power supply
uses a very simple design, with an off-line flyback converter topology.
[59]
Steve Jobs said that every computer now rips off Rod Holt's revolutionary design.
[1] But is this design revolutionary? Was it ripped off by every other computer?
As illustrated above, switching power supplies were in use by many
computers by the time the Apple II was released. The design is not
particularly revolutionary, as similar simple off-line flyback
converters were being sold by Boschert
[50][60]
and other companies. In the long term, building the control circuitry
out of discrete components as Apple did was a dead-end technology, since
the future of switching power supplies was in PWM controller ICs.
[2]
It's surprising Apple continued using discrete oscillators in power
supplies even through the Macintosh Classic, since IC controllers were
introduced in 1975.
[48] Apple did switch to IC controllers, for instance in the Performa
[61] and iMacs.
[62]
The power supply that Rod Holt designed for Apple was innovative enough to get a patent,
[63]
so I examined the patent in detail to see if there were any
less-obvious revolutionary features. The patent describes two mechanisms
to protect the power supply against faults. The first (claim 1) is a
mechanism to safely start the oscillator through an AC input. The second
mechanism (claim 8) returns excess energy from the transformer to the
power source (especially if there is no load) through a clamp winding on
the transformer and a diode.
This is the AA11040-B power supply for the Apple II Plus.
[59]
AC power enters, on the left, is filtered, goes through the large
switching transistor to the flyback transformer in the middle, is
rectified by the diodes to the right (on heatsinks), and then is
filtered by the capacitors on the right. The control circuitry is along
the bottom. Photo used by permission from kjfloop, Copyright 2007.
The AC start mechanism was not used by the Apple II,
[59] but was used by the Apple II Plus,
[64] Apple III,
[65] Lisa,
[66] Macintosh,
[67] and Mac 128K through Classic.
[68] I could not find any non-Apple power supplies that use this mechanism,
[69] except for a 1978 TV power supply,
[70] and it became obsoleted by IC controllers, so this mechanism seems to have had no impact on computer power supply design.
The second mechanism in Holt's patent, the clamp winding and diode to
return power in a flyback converter, was used in a variety of power
supplies until the mid-1980s and then disappeared.
Some examples are the Boschert OL25 power supply (1978),
[60]
Apple III (1980),
[65]
Apple's power supply documentation (1982),
[59]
Tandy hard drive (1982),
[71]
Tandy 2000 (1983),
[72][73]
Apple Lisa (1983),
[66]
Apple Macintosh (1984),
[67]
Commodore Model B128 (1984),
[74]
Tandy 6000 (1985),
[75]
and
Mac Plus (1986) to Mac Classic (1990).
[68]
This flyback clamp winding seems to have been popular with Motorola in
the 1980s, appearing in the MC34060 controller IC datasheet,
[76] a 1983 designer's guide
[77] (where the winding was described as common but optional), and a 1984 application note.
[78]
Is this flyback clamp winding the innovation of Holt's that other
companies ripped off? I thought so, until I found a 1976 power supply
book that described this winding in detail,
[35]
which ruined my narrative. (Also note that forward converters (as
opposed to flyback converters) had used this clamp winding dating back
to 1956,
[79][80][81] so applying it to a flyback converter doesn't seem like a huge leap in any case.)
One puzzling aspect of power supply discussion in the book
Steve Jobs[1]
is the statement that the Apple II's power supply is "like those used
in oscilloscopes", since oscilloscopes are just one small use for
switching power supplies. This statement apparently arose because Holt
had previously designed a switching power supply for oscilloscopes,
[82] but there's no other connection between Apple's power supply and oscilloscope power supplies.
The biggest impact of the Apple II on the power supply industry was on
Astec - the Hong Kong company that manufactured the power supply. Before
the Apple II came out, Astec was a little-known manufacturer, selling
switching DC-DC inverters. But by 1982, Astec had become the world's top
switching-powers-supply manufacturer, almost entirely based on Apple's
business, and kept the top spot for a number of years.
[83][84] In 1999, Astec was acquired by Emerson,
[85] which is currently the second-largest power supply company after Delta Electronics.
[86]
A little-known fact about the Apple II power supply is that it was
originally assembled by middle-class California housewives as piecework.
[83]
As demand grew, however, power supply construction was transferred to
Astec, even though it cost $7 a unit more. Astec was building 30,000
Apple power supplies monthly by 1983.
[83]
Power supplies post-Apple
In 1981, the IBM PC was launched, which would have lasting impact on
computer power supply designs. The power supply for the original IBM
5150 PC was produced by Astec and Zenith.
[83] This 63.5W power supply used a flyback design controlled by a NE5560 power supply controller IC.
[87]
I will compare the IBM 5150 PC power supply with the Apple II power
supply in detail to show their commonalities and differences. They are
both off-line flyback power supplies with multiple outputs, but that's
about all they have in common. Even though the PC power supply uses an
IC controller and the Apple II uses discrete components, the PC power
supply uses approximately twice as many components as the Apple II power
supply. While the Apple II power supply uses a variable frequency
oscillator built out of transistors, the PC power supply uses a
fixed-frequency PWM oscillator provided by the NE5560 controller IC. The
PC uses optoisolators to provide voltage feedback to the controller,
while the Apple II uses a small transformer. The Apple II drives the
power transistor directly, while the PC uses a drive transformer. The PC
checks all four power outputs against lower and upper voltage limits to
make sure the power is good, and shuts down the controller if any
voltages are out of spec. The Apple II instead uses a SCR crowbar across
the 12V output if that voltage is too high. While the PC flyback
transformer has a single primary winding, the Apple II uses an extra
primary clamp winding to return power, as well as an another primary
winding for feedback. The PC provides linear regulation on the 12V and
-5V supplies, while the Apple II doesn't. The PC uses a fan, while the
Apple II famously doesn't. It's clear that the IBM 5150 power supply
does not "rip off" the Apple II power supply design, as they have almost
nothing in common. And later power supply designs became even more
different.
The IBM PC AT power supply became a de facto standard for computer power
supplies. In 1995, Intel introduced the ATX motherboard specification,
[88]
and the ATX power supply (along with variants) has become the standard
for desktop computer power supplies, with components and designs often
targeted specifically at the ATX market.
[89]
Computer power systems became more complicated with the introduction of
the voltage regulator module (VRM) in 1995 for the Pentium Pro, which
required lower voltage at higher current than the power supply could
provide directly. To supply this power, Intel introduced the VRM - a
DC-DC switching regulator installed next to the processor that reduces
the 12 volts from the power supply to the low voltage used by the
processor.
[90]
(If you overclock your computer, it is the VRM that lets you raise the
voltage.) In addition, graphics cards can have their own VRM to power a
high-performance graphics chip. A fast processor can require 130 watts
from the VRM. Comparing this to the half watt of power used by the
Apple II's 6502 processor
[91]
shows the huge growth in power consumption by modern processors. A
modern processor chip alone can use more than twice the power of the
whole IBM 5150 PC or three times that of the Apple II.
The amazing growth of the computer industry has caused the power
consumption of computers to be a cause for environmental concern,
resulting in initiatives and regulations to make power supplies more
efficient.
[92] In the US, Energy Star and 80 PLUS certification
[93]
pushes manufacturers to manufacture more efficient "green" power
supplies.
These power supplies squeeze out more efficiency through a variety of
techniques: more efficient standby power, more efficient startup
circuits, resonant circuits (also known as soft-switching and ZCT or
ZVT) that reduce power losses in the switching transistors by ensuring
that no power is flowing through them when they turn off, and "active
clamp" circuits to replace switching diodes with more efficient
transistor circuits.
[94]
Improvements in MOSFET transistor and high-voltage silicon rectifier
technology in the past decade has also led to efficiency improvements.
[92]
Power supplies can use the AC line power more efficiently through the technique of power factor correction (PFC).
[95]
Active power factor correction adds another switching circuit in front
of the main power supply circuit. A special PFC controller IC switches
this at a frequency of up to 250kHZ, carefully extracting a smooth
amount of power from the power supply to produce high-voltage DC, which
is then fed into a regular switching power supply circuit.
[13][96]
PFC also illustrates how power supplies have turned into a commodity
with razor-thin margins, where a dollar is a lot of money. Active power
factor correction is considered a feature of high-end power supplies,
but its actual cost is only about $1.50.
[97]
Many different controller chips, designs, and topologies have been used
for IBM PC power supplies over the years, both to support different
power levels, and to take advantage of new technologies.
[98] Controller chips such as the NE5560 and SG3524 were popular in early IBM PCs.
[99] The TL494 chip became very popular in a half-bridge configuration,
[99] the most popular design in the 1990s.
[100] The UC3842 series was also popular for forward converter configurations.
[99]
The push for higher efficiency has made double forward converters more popular,
[101] and power factor correction (PFC) has made the CM6800 controller very popular,
[102]
since the one chip controls both circuits. Recently, forward converters
that generate only 12V have become more common, using DC-DC converters
to produce very stable 3.3V and 5V outputs.
[94]
More detailed information on modern power supplies is available from many sources.
[103][104][98][105]
This typical 150W XT power supply uses the popular half-bridge design.
The AC input filtering is on the right. To the left of this is the
control/driver ciruit: the TL494 IC at the top controls the small yellow
drive transfomer below, which drives the two switching transistors on
the heatsinks below. To the left of this is the larger yellow main
transformer, with the secondary diodes and regulator on the heatsinks,
and output filtering to the left. This half-bridge power supply design
is totally different from the Apple II's flyback design. Photo copyright
larrymoencurly, used with permission.
Modern computers contain a surprising collection of switching power
supplies and regulators. A modern power supply can contain a switching
PFC circuit, a switching flyback power supply for standby power, a
switching forward converter to generate 12 volts, a switching DC-DC
converter to generate 5 volts, and a switching DC-DC converter to
generate 3.3 volts,
[94]
so the ATX power supply can be considered five different switching
power supplies in one box. In addition, the motherboard has a switching
VRM regulator to power the processor, and the graphics card has another
VRM, for a total of seven switching supplies in a typical desktop
computer.
The technology of switching power supplies continues to advance. One
development is digital control and digital power management.
[106]
Instead of using analog control circuits, digital controller chips
digitize the control inputs and use software algorithms to control the
outputs. Thus, designing the power supply controller becomes as much a
matter of programming as of hardware design. Digital power management
lets power supplies communicate with the rest of the system for higher
efficiency and logging. While these digital technologies are largely
used for servers now, I expect they will trickle down to desktop
computers eventually.
To summarize, the original IBM 5150 PC power supply was different in
almost every way from the Apple II power supply, except both were
flyback power supplies. More recent power supplies don't even have that
in common with the Apple II. It's absurd to claim that power supplies
are ripping off Apple's design.
Famous switching power supply designers
Steve Jobs said that Rod Holt should be better known for designing the
Apple II's power supply: "Rod doesn't get a lot of credit for this in
the history books but he should."
[1]
But even at best, power supply designers aren't famous outside a very
small community. Robert Boschert was inducted into Electronic Design's
Electronic Engineering Hall of Fame in 2009 for his power supply work.
[51]
Robert Mammano got Power Electronics Technology's lifetime achievement
award in 2005 for starting the PWM controller IC industry.
[10]
Rudy Severns got Power Electronics Technology's lifetime achievement
award in 2008 for his innovations in switching power supplies.
[107] But none of these people are even Wikipedia-famous. Other major innovators in the field get even less attention.
[108] I repeatedly came across the work of Elliot Josephson, who designed satellite power systems in the early 1960s
[18], has a bunch of power supply patents including the Tandy 6000
[75], and even has his patent number printed on the Apple II Plus and Osborne 1 power supply circuit boards
[59], but he appears to be entirely unrecognized.
The irony in Steve Jobs' comment that Rod Holt doesn't get a lot of credit is that Rod Holt's work is described in
dozens of books and articles about Apple, from
Revenge of the Nerds in 1982
[109] to 2011's best-selling Steve Jobs biography, which makes Rod Holt easily the most famous power supply designer ever.
Conclusion
Power supplies aren't the boring metal boxes that most people think;
they have a lot of interesting history, driven largely by the
improvements in transistors that made switching power supplies practical
for computers in the early 1970s. More recently, efficiency standards
such as 80 PLUS have forced power supplies to become more efficient,
resulting in new designs. The Apple II sold a huge number of switching
power supplies, but its power supply design was a technological dead end
that was not "ripped off" by other computers.
If you're interested in power supplies, you might also like my article Tiny, cheap, and dangerous: Inside a (fake) iPhone charger.
The DMM comes with a temperature probe, battery, alligator clips, and
test probes. Note that the test probes have very short metal tips,
unlike the long tips on most probes. The alligator clip probes are a
nice addition. My biggest complaint with the DMM is the temperature
probes connections are soldered with no strain relief so I worry the
wires will break off.
