had been replaced. Big surprise! Board #A4 had been
completely redesigned by HP. No more three ROMS,
just one big one. Hopefully, it won’t die like the others.
Figures B and C in the sidebar show photos of HP’s board
The final chapter to this story is better than the
beginning. The fellow who sold me the original nonworking unit gracefully refunded my $99 and told me to
keep it as a door stop. All in all, it turned out for the better.
The Figure 7 cert for the 39 year old black beauty was
1.01928 volts. However, some Weston cells were known
to lose a small amount of voltage over the years; about 30
µV/year: - 30 µV x 39 years = -1560 µV (1.56 mV). Quite a
With my newly acquired and calibrated HP3456A in
hand, Figure 8 showed the actual loss and it was much,
much less: 1.019280V - 1.019165V = 115 µV, i.e., a 0.115
mV loss in 39 years. Amazing! Maybe the lower loss was
because of the brand-new condition of the cell and benign
storage in a warehouse all that time. Who knows?
Meanwhile, time marched on and new technologies
popped up to challenge the Weston cell. In this case, it was
the Josephson junction which proved to be 1,000 times
more accurate and stable.
In 1962, a graduate student at Cambridge University
named Brian Josephson derived a series of equations that
postulated that two superconducting electrodes separated
by a thin layer of insulation would create a special junction,
A few millivolts was not a very useful calibration
source, so developers fabricated an integrated circuit type
structure that had an array of thousands of tiny junctions
all in series, all adding up. It took 20,208 tiny junctions to
produce one volt and almost 300,000 to make 10 volts.
Figure 9 shows a microscopic view of an early one volt
version. If you look closely, you can see the serpentine
array of junctions, although the whole chip was only about
Figure 10 is a view of a typical complicated laboratory
setup using several cylindrical liquid helium dewars
(pronounced do’-ers) to cool the integrated arrays to four
degrees above absolute zero and provide the microwave
power to the junctions. Figure 11 is a very simplified
diagram of the arrangement of the junctions, showing the
flow of the precise microwave signal.
Typical frequencies are around 75 GHz. Also for
simplicity, the current-bias wires that determine the
operating point of the array and polarity of the DC outputs
are not shown.
FIGURE 9. This highly magnified view of a superconducting
Josephson junction array has 20,208 tiny junctions that
generate 1.000000 volts. Photo courtesy of NIST.
FIGURE B. Old HP3456A (serial #18467) with three defunct
FIGURE C. Newer HP3456A (serial #19178) with redesigned board
and a single ROM.
HP redesigned their HP3456A voltmeter to replace three defective ROMs with one larger memory chip.
September/October 2018 57