Note: Descriptions are shown in the official language in which they were submitted.
~Z~i79~3
APPARATUS AND ME~IOD ~OR PROVIDING IMPRW ED
RESISTIYE RATIO STABILITY OF A RESISTIVE
DIYIDEB NETWORK
FIELD OF T~ INVEi~ILQ~
The pre3ent Inventlon relates generally to thin and thi¢k fll~
re3i3tor network~ and ln partlcular to improving the re~i3tlve ratlo
tability of reslstor divider networks.
~AÇ~GROUND OF ~ VENTIO~
High preci~ion, electronic measurement and test require resl~tive
divider network~ having high reslstive ratio stabllity. This means that
the ratios of the resistor values should remain as stable as po3aible
~hen the network is ~ubJected to environmental and operational influence~
such as temperature and voltage change~.
In the pa~t, high precision in3truments had to u3e Yery expensi~e~
physically large~ ~lre-wound resistor~ in thelr dl~ider networks. Film
re3i~tor Detworks, ~hile gatisfactory for les~ accurate instrument~, were
8enerallY lncapable Or the required ratio stability unles~ they were
~pecially selected, which made them very expenslve. Ratio stability on
~ the order o~ 0.5 part3 per million (ppm) per degree centlgrade (ppm/C)
`~ for ambie~t temperature changes and 2 ppm for input voltage changes Or
1000 Yolt3 are involved here.
Ratio 3tabllity is arfected by three prlmary factors:
1. the dlfference in the temperature coefriGient of re3istance
(TC~) of the reaistor3 ~aking up the divider ( al30 known aa TCR
tracking);
2. the dlrference in the voltage coerriclent Or re~lqtance (VCR~ Or
the reslstors ~al~o kno~n a~ vcn tracklng); and
3. the dlfference in the temperatures of the re~lator~.
In conalderlnB the affect o~ ~actor 1, TCR 19 de~lned a~:
TCR = R~_~ R ~
~ here R2 and Rl are the reslstance Yalue~ of a sln61e reslstor at
temperatures t2 and tl re~pectively. The TCR may be either positive or
ne8ati~e.
The di~ference3 in the TCR's Or the resi~tQrS comprlsing the
network, or the TCR tracking, have a most signifioant efrect on ratio
stability. In a two re~i~tor netHork, lf the TCR Or both re~istors
comprising the network are identical, th~ ratio of the two will remain
constant as the a~blent temperature changes. If the TCR's Or the two
resistors are not the sa~e, as 19 usually the case~ the ratio
due to TCR effects will change as the ambient temperature changes. While
TCR's may be either positi~e or DegatlYe, ~hich means that the re~istance
may either increase or decrease ~ith lncreasing temperature, the greater
the dif~erence in the TCR'~ Or the two re~lqtor~, the greater will be the
change ~n the ratio or the poorer will be the ratio stability.
In considering the affect nr faotor 2, YCR i9 defined as:
VCR - RZ~~
~j - R1 ( E2 - El )
where: R2 and Rl are the resl~tanoe ~alue~ of a slngle rea~3tor at
applied ~oltages ~2 and E1 respectivel~.
The YCR o~ depoalted film reslstors ls alwaya negatlYe, and for ~el}
designed, properly manuractured, thin fllm re~l~tor~, the VCR 1~
generally qulte low. For example, th~n fil~ réslstora ~ade from 100 to
200 ohms per square material, typieally have VCR'3 in the rar~e Or .001
,..,. ; . ,. ,., ..~..
, . , ', :, ' . ' "~.......... ,~, . .;
, . .
` ~LZ~7943
to .01 ppm~volt. Hence, a 10 ~egoh~ re9i~tor wlll decrease ln ohmlc
~alue by 1 to 10 ppm ~10 to lO0 ohm~) ~hen the voltaKe applled to lt ia
increa~ed 1,000 YDC (e.~. from lOOV to llOOV).
~hen considerin6 resl3tlve divlder net~orkg, the ~oltage change la
ln proportlon to the re~i~tor value~. Hence, for divlders ~lth ratios
greater than 10 to 1, only the VCR Or the higher Yalue resistor i~
~lgniflcant.
Time wi3e, the erfect of VCR on the ab~olute ~alue of a rilm
re~i~tor i9 e33entlally ln~tantaneou3 ~hile the effect of TCR on the
ab~olute value of a r~lm re3i~tor depend~ on the thermal tlme con~tant of
the re3i3tor. Typically 90~ Or the temperature rise is complete in le~3
than one minute. The oombined effect Or VCR and TCR on the re3i3tor
value is called po~er coerficient of reqistance or PCR, and i~ the
algebraio sum Or the change in reslstance of a resl~tlve element due to
lt~ VCR and an increase ln app}ied ~oltage (always negative~ and the
change ln re~l~tance Or the ~ame resi3tlve element due to It~ TC~ and the
self heating cau3ed by the 3ame lncreage in spplled volta8e (may be
either positlve or re~atlve). The combined effect (PCR) can cause the
re~lstor Yalue to elther increa3e, decrease, or in rare case~, e~en
remain con3tant.
In con3idering the affect of factor 3, the relati~e temperature Or
the two reqi3tors depend3 upon three parameter~:
x* 1. the power dlsslpated per unlt ar~a by each re~istor;
2. the di~tance bet~een the two re~lstor~; and
3. tha thermal corductlYlty of the aub-qtrate.
Con3ider rlr3t parameter nu~ber 1~ the power disslpatlon per unlt
area, The power dlsgipated by each re~lqtor lh the network i~ a glven
and is dlrectly proportional to the ohmlc ~alue of each resi~tor. Ir the
... ' ;''' " ". " .: ,' '.'.'`' . ,'. '''','' ' '. , ' ' . . ,, ' :', ".
area Or the network wa~ inflnitely large, the èemperature rl3e Or all
aections, and hence the temperature dirference ~etween 3eCtiO119 ~ would be
e~sentlally zero. Ir the area of the net~ork was lnfinltelY amall, the
temperature rlse would be ~ery hlgh, but becau9e all the resl3tor3
occupled the 3ame 3pace, there would be no temperature di~ference between
sections. Ob~iou31y both cgge3 are impractlcal; but serve a3 theoretlcal
boundaries. Actual net~orks probably average 1/2n x 1" x .025~ thick
wlth the individual resi~tlve element~ placed aide-by-~lde on the
sub~trate. The hl~her ~he total power dis~ipated and the greater the
difference in power di~ipated by the lndividual resi3tors, the greater
wi]l be the difference in temperature be~ween the~reaistors.
Skipplng parameter 2 for a moment, consider parameter number 3 next ~
the thermal conducti~lty of the aubgtrate. The conductl~ity Or most
materlala in common use today - steatite, glasa, alumina, etc. is
~ relatively poor campared to copper. Hence, there uill always be a
~ difference in average temperature or high and low dis~lpative sectlons.
; This leaves onlg parameter number 2, the distance between the
re3iators on the sub~trate. Unfortunately, regardles~ of ho~ close
together the resi3tors are placed 1~ the conventional side-by-~ide
corriguration, there ~ill alwayg be a dlfference ~r the average
temperature of hlgh and low power dlgs~pative sectlons; thls again
coming back to the lmperfeot thermal conductivity Or the ~ubatrate.
Hence, eYen ir the TCR's Or the two reolstors are identlcal, th~
ratios wlll ~tlll change ~hen the appl~ed voltage is increased (unle~s
the TCR of each reslstor 19 zero, ~hlch is virtually impossible).
Further, regardl~s Or the method u~ed to deposlt the resi~tive materlal
on the ~ubatrate, there ls al~ay3 some random varlat~or, in the metallur8y
Or the film. Hence~ the TCR Or the resultin8 metal rllm~ from one edge
- ~LZ~ 3
Or the ~ubstrate to the oppoaite edge tend3 to vary smoothly, although
not neressarily linearly, ~ith dlstance from the reference edee. Nence,
it i9 virkually impossible to haYe req1~tor~ ~ith identlcal TCR's.
There has been a long felt need for a fil~ resis~or dlYider net~ork
where the TCR difrerence o~ the individual reqlstors approaches zero over
the operational temperature range and ~he temperature difference of the
lndlvidual resistor3 approaches zero o~er the operational voltage range.
OF ~ 3YE'Il~
An advantage of the pre~ent invention i9i to provide a resl~tl~e
divider network whereln the TCR difference of the indlvidual resistors
approaches zero over the operational temperature range and the
temperature difrerence Or the individual re~istors approache~ zero over
the operational voltage range.
In the present invention, each of the elements of the re~tstive
; dividers are formed into a plurality o~ subelements spaced acros3 the
sub~trate and lnteroonnecSed to form discrete re31stor elements. The
subelements Or th~ re~i3ti~e element dlssipating the lo~er power are
lnterleaved between the subelements Or the resi~tive element dlssipatir~
the higher power. ~y havlnB a provision for resistiYe trl~mlng Or each
subelement, it 19 pos~ible to obtain desired absolute values and ratios.
The present inventlon ~urther provlde~i an improved resistor diYider
network havin8 an improvement in TCR traoking typically in exces3 Or
l.OOQ~.
me abo~e and addltional advantage~ of the present invention will
become apparent to those skllled ~n ~he art from a reading Or the
folIowing detailed descriptlon when taken in con~unction wlth the
accompanylng dra~ings.
~LZ6~79~3
Fl~. 1 is the electrlcal clrcu~t 3chematic repre~enting a typical
two re~istor dlvider netwnrk;
Flg~ 2 i3 a plan ~lew of-a pr~or art re~l~tor dlvider network; and
Fig. 3 19 a plan ~lew of the reqi~tor divider netuork of the pre~ent
lnvention.
DESC~IP~IpN 0~ la~ pREFER~EP EMBODIM~N~
Referring now to Fi8. 1, therein is shown a 3chematic representation
of a re3i~tor di~ider network 10. The network 10 has an input terminal
12 connectible to a flrst resi3tive element 11l. The fir~t resistive
element t4 is connectible to an output termlnal 16 and to a second
resi~tive element 18 which i~ conneotible ln turn to a ground terminal
20.
Aa would be evldent to tho3e ~kllled ln the art, the terminal-~ are
referred to for clarlty only ~ince divider net~orks may be integral ~ith
other circuit componentg and thu3 may not have terminal~c Similarly, the
lnput, output, and ground termlnology i9 for clarity only 31nce multiple
resistor network3 could ha~e mul~lple lnput3, outputs, and grounds.
Further, the elements are con~idered noonnectlble" because they may not
be connected to a terminal when manufactured a3 a complete net~ork; it is
pos3ible that all oor~ection3 ~ay be made apart from the network it3elf.
In preci3ion reslstive divlders, it i5 often neces~ary that tl) the
resi3tance ratios change b~ no more than 0~5 ppm~C over the operating
temperature range , and ~2) by no more than 2.0 ppm over the operatlng
~olta8e raneeO In the pa~t, such strin~ent.requlrement~ could only be
2679~3
met through the u~ of seYeral carefully matched, wlrewound reslstora
which ~ere very expen3i~e and phy31cally quit~ large.
Referring now to Fig. 2, thereln 1_ shown a prior art, fllm
re~lst~ve dlvider network whlch had resistance ratlo stability a~quate
for some, but not all, appllcatlon~ formerly requiring ~irewound
resi3tor3.
The film resi3tor 22 include3 a Qubstrake 24. The ~ubstrate 24
could be glas9 or other material but is by preference alumina (A1203)
with a thermal conduotivl~y about 5% that Or copper. Typically, it i9
3/4" long by 1/2" ~ide and .025~ thick.
Deposited on the substrate 24 by sputtering, ~acuum depo~ltlon or
other means ls a resiQtive material ~hlch in subsequent manufacturing
operations is coated ~lth a photoresistive material, photo masked,
exposed to ultra vlolet llght, then chemically etched to remoYe the
un~anted material leaving the desirad resistlve elements 26 and 28 in a
serpentine pattern whlch may be ~o flne as to be considered rectangular
as a ~hole to a viewer. The serpentine pattern conventionally also
lncludes 3hunts whlch are ~aporlzed away with a laser during a ~laser
trimmlng~ operatlon to establish the exa¢t resi3tanoe and ratlo ~alues,
as well kno~n to those skilled ln the art. The reslstlYe materlal almost
universally used 19 Nichrome (a registered trademark of Drlver ~arris
Company) alloy wbioh contalns 60% nickel, 24% iron, 16S chromium, and
o.ls carbon. Due to ~arlations ln metallurgy Or the film deposited onto
the substrate, the TCR Or the resultlng metal film ~aries from one edge
Or the substrate to the oppoglte edge. The TCR tends to Ysry ~moothly,
althuugh not necessarlly llnearly, with diatance across tbe substrate.
The flr~t re~istlve element Z6 ls the hlgher power dls~ipatlng element of
the t~D.
,, ' , . .
-- ~ zt~7~
The rlrat resl9tlYe element 26 is connectlble to an input termlnal
and to an output ter~lnal 32 whlch la al~o connectible to thc second
resisti~e element 2~. The second resl~ti~e element 2a i9 further
connectible to a ground termlnal 34.
~ hile th~ e~ficiency Or heat ~ransfer between the t~o re31stlve
ele~enta 26 and 28 increases a~ tha size of the elements decrea~e and the
space between them decreases, there i3 a llmit c,n ize imposed by a-
number of faceors.
Referrlng no~ to Fig. 3, therelr. 1~ show~ an i~proved re~istor
divider net~ork 40. The network 40 la dispoYed un a qubstrate 42. The
sub~trate 42, compared to a conventional ~imilar re~istor divider
network, is ~omewhat larger being 1-1/2~ long by 5Jô~ uide and .025
thlck. Deposited on the substrate 42 is a resl~tive material which 1~
dlvided up into a number of portlon3. One portlon deflnes a fir~t
res~stive element which is made up Or rlrst reslstive aubelementY 46
through 50. Interleaved bet~een the ~irgt resi~tiYe 3ubelements 46
through 50 is a second resist~e element which la made up of second
re3istive subelements 52 through 55.
Tbe flr~t reqlstive subelements 46 through 50 are interconnected to
form the equivalent Or the rlr9t resl~tive element 26 of Flg. 2 and,
similarly, the second resistive subelements 52 through 55 are
interconnected to form the equlv~lent of the second reslstive element 28
of Fig. 2
The rirst reslstive s~belement 46 on one -qlde Or the substrate 1
connectlble to an input terminal 56 and the la~t re~istlve subelement 50
on tbe other ~ide is oonnectible to an output terminal 58. ~he rir~t
. ..
resi3ti~e subelement 52 on one side 19 connectible to a ground termlnal
60 while the last re~i~.tlve ~ubelement 55 i3 co~nectible to the output
* terminal 58.
Ey compari,~g the reql~tor dlvlder network3 of Flg. 2 and Flg. 3, lt
may be 3een that ~iubstantlal i~pro~ement in re3istive ratlo ~itabill'cy car,
' be achiPved Ju3t in the TCR tracking.
i For e~ampla, asqume that the TCR of the resistive material Yarle3
from 10.00 ppl/C at the le M edge Or the 3ubstrate to 6.00 ppm/C at the
r`.'ght ed8e of the cubstrate for both Figure3 2 and 3.
For the nst~ork 22, the approximate TCR of' the right hand edge v~
element 26 will be 6.50 ppm/C. It become3 apparent then that the average
' TCR ~or the resistive element 26 19 ~10.00 ~ 6.50)/2 or ~'.25 ppm/C.
~'lth the le~t side Or the second resistiYe element 28 approximately 6.40
ppmfC the average TCR of the 3econd resi~tlve element 26 is (6.40
6.00)~2 or 6.20 pp-~'C. It then becomeq apparent that the dirference
between the TCR'3, or TCR trackir~, i9 (8.25 - 6.20) or 2.05 ppm~C.
Lookl~s at the lmproved dlvider network 40, and a sumi~g a nQn~
llnear but alway. decreasin~ TCR as ~e move from left t~ rlght, the
aver~ge TCR's ror the YariouC reiistlve subelements of the rirst
re~lative element typically might be: 10.U0 for subéle~ent 46; 8.50 for
3ubelement 47; 8.00 ror subelement 48; 7.00 for subelement 49; and 6.20
for 3ubelement 50. The average TCR for the rirst re~istive element would
then be (10~00 1 8.50 ~ ~.00 ~ 7.90 + 6.20)/5 or 7.94 ppm/C. For th-~
second re~l~ti~e element, the lnterpolated ~olues of TC~'s oS the
subelement3 uould be 9.25 for subelement 52; 8.25 for subelement 53; 7.50
~or aubelement 54; and 6.60 for subelement 55. Ihua~ the a~era~e TCR rOr
the second resl3tiYe element wou}d be (9~25 ~ 8.25 ~ 7.50 ~ 6.60~/~ or
7.90 ppmfC. me dlfference between the average TCR~ ror the re31stor
. . 9
~79~3
~et~ork ~0 or TC~ trackin8 would then be (7.94 7~90) or 0.04 ppm~c.
Ihe lmproveme~t ln TC~ tracklng of ~he new, interleaved de~1gn over
the old prior art deglgn ia 2.05 ppm/C - .04 ppm/C = 51 ~o 1. Thus,
the preferred embodiment 3ho~g approximately a ~JlQOS lmpro~ement.
In reYlewlng the above, it becomes evldent that as the number of
subelement3 lncrease3, the ratio TC lmproves, approachlng 0. ~urther, if
the Yariatlon in TCR i9 linear from one g1de of the ~ubstrate to the
other, the TC tracking Or a ~o resii3tor divider ~111 always be perfect
(zero), regardless of the number Or subelements used, provldlng that the
total ~umber ls al~ays an odd number.
Still further, it is eYlden~ that the prlor art sufrers from uneven
temperature dii~ribution acroa~ the resiatlve elemants because
temperature is a direct runction Or the power dissipated per unit area by
each resistive element. ~y lnterleavlng 9ubelement3 of the hi6h and low
power diissipative elementg and by plaoing the i~ubelements of the
resistive element dig9ipating the higher power at the sides o~ the
substrate 9c that they erfeotively isurround the gubelementY of tha
~; reslstlYe element di~3ipating the lower po~er, a mucb ~ore unifor~
temperature may be achieved acroas th~ i~ubstrat~.
In order to evaluate the present inYention, 17 samples Or a divide-
by-64 (64:1), thin rilm reslstor network were fabricated using 8
sputtered Nlchrome rilm on a hi~h purlty alumina subatrate.
The fiYe subelement~ ~6 through 50 ~ere lai~er trimmed to be Or
approxlmate equal value and the 9ubelements 52 through 55 ~ere alao
trimmed to be of approximate equal va~ue. The total resistance o~ thc
~ubele~ent3 46 through 50 wa9 9.84500 me~ohm ~ 0.1%. Ihe total
resistance or ths ~ubelements 5Z through 55 wa3 0.15619 megohm ~ 0
~ 79~3
The nominal divislon ratlo ~as ~ = ~ (l.a. 64:1)
.15619
Throueh aortware inst~uction3, a computer controlled laqer trimmer
brou~ht thla ratio to ~lthln ~ .05~ of nominal.
The TC~'s of th~ two setA of subelements ~ere te~ted over the
temperature range Or 18C to 5~C and the dlfference tabulated ror each
network. The abqolute TCR's ranged ~rom approximately 1 ppm~C to 6
ppm/C, and the ratio TC~'s ranged from 0.02 ppm~C to 0.30 ppm/C with
an average o~ 0~12 ppm/C which was four ~im~9 bette~ than the target
~oal of 0.50 pp~/C.
The PC~'3 (power coefricient of ratio) ranged from 0.09 ppm to 0.44
ppm rOr a 1000 volt change in input voltage (100 volts to tlO0 volt~
The average ratlo PC was 0.27 ppm which was el~ht ti~e~ better than the
target goal of 2.0 ppm.
~ y comparison, typical perrormanoe flgure~ for prior art network 22
would be TC tracking of 2 ppm and PC of ratio of 10 ppm.
As many pos~ible embodiments may be made from the pre~ent invention
wlthout departin~ from the scope thereof~ it i9 to be understood that all
matter~ ~et ~orth herein ar~ sho~n ln the accompanying drawing~ as to be
interpreted ln an illustratlve and not a llmltlng ~en~e.
'