Text preview for : dd_cpem02.pdf part of Fluke dd cpem02 Fluke appnotes dd_cpem02.pdf
Back to : dd_cpem02.pdf | Home
THIN FILM DESIGNS FOR 1000V AC RANGE RESISTORS
David Deaver
Fluke Corporation
POB 9090
Everett, WA 98206
Abstract resistor. The most effective shields provide decreasing
field strength longitudinally along the resistor resulting
Alternatives to the traditional coaxial design of precision
in no current flow in the capacitance to the shield.
ac range resistors have resulted in performance
breakthroughs, reducing both ac-dc difference and
These coaxial resistors usually have very high thermal
settling times. Application of thin film resistor
resistance to ambient resulting in very long settling
technology and innovative thermal design has made
times. When used with a 5mA TVC, the 1000V range
these advancements possible.
resistor would dissipate 5W, easily resulting in 50-100
degC temperature rise.
Introduction
In the past decade, national measurement institutes Thin Film 1000V Range Resistor Designs
(NMIs) have focused on resolving their differences in the
For the thin film designs, dimensional stability is much
vicinity of 1000V, 100kHz. These differences used to be
better allowing smaller enclosures. In addition, since the
considerable.
resistors are planar, a rectangular enclosure works nearly
as well as the coaxial design making mounting and
Accredited by NVLAP in 1995, the Fluke Primary
reducing thermal resistance much easier.
Standards Lab sought an additional accreditation by the
German DKD in 1997. They accepted most of the claims
already approved by NVLAP with standards traceable to
either the NIST or the PTB. However, the claims made
792A c d 5790A
for 1000V at higher frequencies could only be accepted i k e a
if standards were calibrated at the accrediting body's g b
associated NMi (NIST or PTB) because of the
international differences. Eventually, claims enlarged to
include an uncertainty component due to the f j
international differences between NIST and the PTB, h
were accepted by both NVLAP and the DKD regardless
of whether the standards are traceable to NIST or the Fig. 1 Construction Details of 792A and 5790A 1000V Resistors
PTB. By the 1999 joint assessment, the international a Aluminum enclosure
differences had been reduced enough that the b Al203 (alumina) substrate
international difference uncertainty was no longer c Thin film resistor network
d Glass cover
required. e Glass frit
f BeO (792A) or AlN (5790A) heat substrate
Two examples of high voltage design are considered, the g Flexible adhesive
1000V range resistor in the Fluke 792A and the 1000V h Copper mount and heat conductor
i Heat conductor mounting bracket
divider in the Fluke 5790A. j Heat conductive elastomeric gasket
k Driven shield (792A only)
Coaxial 1000V Range Resistor Designs Figure 1 shows a simplified diagram of the construction
Because large values are required for high voltage range techniques used for the 792A and 5790A 1000V range
resistors, the capacitance to nearby mounting structures resistors. Both use a thin film resistor network deposited
greatly affect their frequency response. They have on an alumina substrate. The thermal resistance of this
traditionally been mounted in cylindrical enclosures. To substrate is much higher than the underlying BeO
the first order, any increase in capacitance caused by a (beryllium oxide) or AlN (aluminum nitride) heat sink
lateral displacement of the resistor toward one side of the substrate. Alumina is used because of its superior surface
enclosure is offset by a reduction in capacitance as the smoothness and compatibility with the thin film and
resistor moves away from the opposite side. A driven glass frit materials. Using a patented process (U.S. Patent
shield is often used to flatten the ac response of the number 4,803,457), the temperature coefficient of the
thin film resistor can be trimmed to less than a part in An Alternative to Shields
6
10 .
7KH $ UDQJH UHVLVWRU LV D N WR
divider followed by an amplifier to provide higher input
The alumina is bonded to the heat conductive substrate
impedance than a "TVC compatible" range resistor. The
with a thin layer of a flexible adhesive. The heat
higher resistance provided even more of a challenge to
dissipated by the resistor has only to travel through 0.1
flatten the frequency response.
mm of alumina before reaching the heat conductive
substrate, which has very low thermal resistance. The
Instead of a driven shield, it was decided to use the
792A uses strips of copper soldered to the heat substrate
dimensional stability of the thin film substrate to design
and then clamped to the exterior case to conduct the heat
interstitial capacitances into the serpentine pattern of the
from the resistor. The 5790 clamps the heat substrate
network. Current flowing out of the network to the case
between the two halves of the enclosure using thermally
would, to a first order, be supplied by currents from the
conductive gaskets. Both techniques result in thermal
interstitial capacitances. The 5790A divider is shown in
resistances from the thin film resistor to the case of less
Figure 3. The interstitial spacing is reduced considerably
than 1 deg C / Watt.
as compared to the 792A resistor.
The low thermal resistance combined with the very low
temperature coefficient result in very fast settling times
compared to the coaxial designs.
AC Performance
Making the cases smaller increases the capacitance to the
resistor resulting in increasing ac-dc difference as the
frequency increases. One technique of compensating for
this capacitive coupling is through the use of driven
shields; that is, a shield connected to the input side of the
resistor, which is positioned between the case and the
resistor. The 792A uses a driven shield as shown (Figure Fig. 3 5790A 1000V Range Divider
2) affixed to the glass lid of the left resistor element.
Using the driven shield around one of two series Figure 4 shows the large effect of the enclosure
connected resistive elements was analyzed at NIST [1]. capacitance. The output peaks considerably without the
In the prototype resistors, the shield was trimmed for enclosure but has a much flatter response when mounted
best flatness. In the production version, a trim capacitor in its aluminum case.
was added to the low voltage side of the range resistor to
allow the flatness to be adjusted. This simple shield 30%
technique does not yield the optimum flatness, however. 25%
Considerable work was done [2] to design a driven shield
Relative Deviatio n
20%
for a thin film 1000V range resistor, which would
15%
provide an optimum distribution of the electrostatic field Out
to reduce current flowing to the shield or the case. 10%
5%
0%
10000 100000 In 1000000
-5%
Frequency (Hz)
Fig. 4 5790A 1000V Range Resistor In and Out of its Enclosure
Relative to a 792A 1000V Range Resistor
Fig. 2 792A 1000V Range Resistor
AC Stability study, this resistor will soon be replaced. Also, because
of the apparent drift of Working Standard 5, the data
For a 1000V range resistor to be useful, its ac-dc presented for units under test (UUTs) will be relative
difference must also be very repeatable and stable. Two only to Working Standard 14.
factors make it difficult to use the calibration certificates
issued by the Fluke Primary Standards Lab to evaluate UUTs Relative to Working Standard 14
the stability of the ac-dc difference of the range resistors.
The first is the stability of Fluke's reference standard. Figure 6 shows a plot of the performance of forty units
Figure 5 shows the 1000V, 100kHz performance based which have been calibrated more than once in the
on the NIST assigned values of ac-dc difference The Primary Standards Lab relative to their first calibration.
shifts represent not only the actual shifts in its ac-dc All these data are presented without working standard
difference, but the shifts in the NIST standards and their ac-dc difference corrections applied.
assigned values. Because considerable work has been
done to reduce the differences between national labs at
All UUTs Relative to Initial Measurement
1000V, 100kHz over the past decade, these shifts in the 80.0
assigned values by national labs are significant.
60.0
Secondly, because working standards are used to
Shift in AC Response (