IEEE Power Electronics Magazine - June 2020 - 100

White Hot

(continued from page 104)

During the 5-ms delay, the system
frequency transformer. The auxiliaran its shutdown process, includry transformer output was rectified,
ing finishing writes to the hard
filtered, and regulated by an LM7812
linear regulator to power the outputdisk. Of course, the DCOK was imreferenced control circuits.
mediately deasserted if the output
In addition to having the VT103
voltages were ever detected to be
power supply design as a starting
out of regulation.
point, I was surrounded by a very
My design for the H7862C started
experienced team of mentors and
with a two-transistor forward concolleagues. Engineers, such as Jim
verter. To try to minimize the size, I
Gregorich, John Herrmann, Trey
chose the unheard-of switching freBurns, and my supervisor, Dave Berquency of 50 kHz. The switching trantetti, were available for any quessistors were SGS Thomson BUW45
tions I had. Art Parker was a great
NPN power bipolars in a TO-3 case.
department manager. I also worked
The high switching frequency was
with a great technician, Jim Gaudet,
made possible by a base drive circuit
from whom I probably learned more
invented by Jim Gregorich. This drive
than he learned from me. DEC also
circuit was a proportional drive (a
had a great team of component engicurrent transformer from the emitter
neers, including Rao Yedavalli (magto the base that forced the transistor
netics), Charlie LaHaye (capacitors),
to operate with a current gain of
Roger LaChapelle
four). To keep the
(transistors), and
transistor out of satGeorge Checkowski
uration to speed up
For an engineer right
(ICs), who provided a
the turn-off switchout college, I could not
great education on
ing, a Baker clamp
have asked for a
the practical aspects
with two diodes in
better group of folks
of t hese compo series with the base
from whom to learn.
nents. For an engiwas used. This as neer right out colsured that the colleclege, I could not have
tor-emitter voltage
asked for a better group of folks from
was always greater than the base-
whom to learn the practical skills
emitter voltage. This caused addiand knowledge needed for power
tional conduction loss but greatly
decreased the turn-off delay, time,
supply design.
and loss. One of Jim's innovations
The model number of the Profeswas to peak charge a capacitor from
sional 350 power supply was H7862C.
the +12-V output winding and use this
The power outputs were +5 V at 20 A,
voltage and stored energy to drive the
+12 V at 8 A, and -12 V at 1 A. The
turn-off of the main power transistor.
main output power connector was on
This created a consistently large turnthe rear of the power supply. The
off current in the main power transisoverall size was 4 in (10 cm) high,
tor base.
8.25 in (12 cm) wide, and 13 in (33 cm)
The main power transformer and
front to back.
There were two power status sigoutput inductors were custom de nals: ACOK and DCOK. The power
signs. I specified the required elecsupply had to maintain the output
trical and magnetic characteristics,
voltages in regulation for a full
but the parts themselves were de half cycle of ac power loss before
signed by Rao Yedavalli, the magnetdeasserting the ACOK signal. Once
ics component engineer. As I recall,
the ACOK was deasserted, the outthese were or igina lly ma nu facput voltages had to remain in regutured by Merrimack Magnetics, a
lation for at least another 5 ms, and
local company that supplied many of
then the DCOK was deasser ted.
DEC's power supply magnetics. The

100

IEEE POWER ELECTRONICS MAGAZINE

z June 2020

+5-V output rectifier was a dual 60-V,
45-A Schottky diode, and the +12-V
output rectifier was a dual 150-V,
30-A fast-recovery silicon diode.
Both of these rectifiers were in TO-3
packages.
The toughest part of the specification was the output noise. While nominally specified at something like 5%
peak to peak, the real requirement
was that a 5.25-in hard drive had to
work with the power from my supply.
At DEC, no hard disk had been successfully powered from a switchmode power supply. Disk drives,
which, until then, were washingmachine-size cabinets, were powered
from linear regulators operating from
either a mains frequency transformer
or a mains frequency ferroresonant
power supply. While investigating the
VT103 power supply and related
designs, I noticed that the auxiliary
switch on the +12-V output created a
large amount of noise. This was never
going to work with the hard drives. I
decided to go with a coupled inductor on the +5- and +12-V outputs and
to use a weighted-sum regulation
scheme. The control IC was the Silicon General SG3527A.
As best I recall, I learned about
the coupled inductor approach from a
paper written by someone at IBM
circa 1977. I have, from time to time,
sea rched to f i nd t h is reference
but have never found it. As for the
weighted-sum regulation scheme, I
am pretty sure that I created that on
my own. I later learned that this had
been done before, but I don't recall
ever seeing a reference. I stabilized
the control by stabilizing each loop
(+5 V and +12 V) independently. I worried about this for many years, as it
"felt right" but I had not proved that
this guara nteed stability. I was
quite relieved when I saw a paper
at the 1986 Power Electronics Specialists Conference that proved that
the approach I had taken did guarantee stability.
To prove this concept, Jim, my
technician, and I created a basic

IEEE Power Electronics Magazine - June 2020

Table of Contents for the Digital Edition of IEEE Power Electronics Magazine - June 2020