MODR RV722

KELVIN-VARLEY RESISTIVE VOLTAGE DIVIDER

'• .
Instruction Manual
.. :"

Part Number 8112 August 1970

\ ~-
\ (
\.
'\·~
\
0 0

0 0

ll•PUT 11£S1STANC[
100,000 OHlolS

\
\
,1

Electro Scientific Industries, Inc.

13900 N.W. Science Park Drive• Porttalld. Oregon 97229 •Telephone: (503) 646·4141

\
.

The following ~ble lists the most recent revision of each page at
the present date of printing: .

PAGE DATE
Title 8/70
A 8/70
i and ii 7/66
1-1and1-2 3/64
1-3 and 1-4 7/66
2-1 and 2-2 7/66
2-3 thru Z-5 3/64
3-1 7/66
3-2 8/70
3-3 thru 3-6 3/64
4-1 and 4-2 7/66
5-1 7/66

CCopy111•1 1964, 1966, 1970 Eloctro Scoo11tilic l11n1t1101, l11c

tit ..... us"'

A
eleli RV 722 8/70
 TAB LE OF CON TEN TS

I INTR ODU CTIO N

1-1
1. l DES CRIP TION 1-3
1. Z SPE CIFI CAT ION S

11 AIDS TO OPE RAT ION

2-1
l .. 1 POW ER LIM ITAT ION S 2-l
2. 2 TEM PER ATU RE AND POW ER EFF ECT S 2-2
2. 3 LOA DIN G EFF ECT S 2-4
2.4 USIN G CER TIFI CAT E COR REC TION S

III CAL IBRA TION

3 -1
3. 1 EQU IPM ENT REQ UIR ED 3-1
3. 2 TRA NSF ER STA NDA RD CAL IBRA TION 3-2
3.3 DET ERM INA TION OF TER MIN AL LINE ARI TY 3-5
3.4 DET ERM INA TION OF END COR REC TION S 3-6
3. 5 CHE CKIN G THE POW ER COE FFIC IEN T
4-1
TV PAR TS LIST
5-1
V APP END IX

i
elsli RV 722 7/66
..i

The following are registered trademarks of Electro Scientific Industries, Inc.:

ESI9 Electro Scientific Industries, Inc.

ESIAC 9 Algebraic Computer
ESIPOT• Potentiome ter
DEKABOX 9 Decade Resistors and CapaCitors
DEKABRID~E 9 Bridge Circuit .
DEKADIAL ~ecade Dials
DEKAMAT IC Automatic Unit
DEKAPOT® Decade Potentiome ters
DEKASTAT 9 Decade Rheostat
DEKAPACI TOR 9 Decade Capacitor
DEKAVIDE R® Decade Voltage Divider
KELVIN KLIPS 9 Four-term inal Clips
KELVIN KLAMPS 9 Four-term inal Clamps
PORTAME TRrc• Portable Measuring Instrument
PVB 9 Potentiome tric Voltmeter Bridge

Application for registratio n has been filed for the following:

DEKA TRAN Decade Transform er

ii
el•li RV 72.Z 7/66
 SECTION

INTRO DUCTI ON

1.1 DESCR IPTION

tory-
The Model RV 722 DEKA VIDER ® Decade Voltag e Divide r is a labora
y circuit for high accu-
standa rd resisti ve divide r. It uses a Kelvin -Varle
atic is illustr ated in
racy, repeat ability , and stabili ty. The circuit schem
Figure 1. lA.

figure 1.1A

input.
Decade dials indicat e the output voltag e in propor tional parts of the
volt-
The output can be read to seven signifi cant digits. The compe nsated
age at the output commo n (termi nal 4) is factory adjuste d to equal the
000, O.
voltage at the output tap (termi nal 3) when the divide r is set to 000,
This provid es high accura cy for low-ou tput voltag e applic ations.

The
The equiva lent circuit of the Model RV 722 is shown in Figure 1. 1 B.
Kelvin -Varle y equiva lent circuit looks like a simple potent iomete r or
the
tapped resisto r, except that ~t has additio nal resista nce in series with
tap.

t
EouT

4
2
figure 1.18

1 - l
elsli RV 7 22 3/64
 The Model RV 7ZZ may be mounted in any standard 19 inch rack~ Electro-
. static shielding is provided by the metal panel and dust cover which are
isolated from the active circuit. High accuracy and resolution of the divider
make it ideally suited for many application s. Dividers, potentiome ters
and ratio sets can be calibrated accurately . Voltage and current measuring
devices can be checked easily. The precision ratio available can be used
for high accuracy impedance compariso ns. The RV 722 i~ suited for· use at
low frequency ac as well as de.

1 - 2
eleli RV 7Z2 3/64
 1.2 SPECIFICATIONS

Lin eari ty: Lin eari ty is the exp ress ion

of the acc ura cy of a divi der in
pro por tion al par ts of inpu t. Two type
s of line arit y are freq uen tly
spec ifie d: abso lute and term inal .
Abs olut e line arit y is the acc ura cy rela
tive to the outp ut at the end -sca le
sett ings of 0 and 1. Tha t is, the divi der
sett ing s. Ter min al line arit y is the acc is defi ned as cor rec t at thes e
ura cy rela tive to the inpu t at the
divi der end term inal s.
Sev eral ES! divi ders hav e sep arat e inpu
t
so that the outp ut com mon may be com and outp ut com mon term inal s
pen sate d for sma ll volt age dro ps
in lead s and swi tche s, mak ing the outp
ut zero whe n the sett ing is zero .
Two term inal -lin eari ty spe cifi cati ons
are give n for thes e div ider s, one
rela tive to the inpu t com mon term inal
and the othe r rela tive to the out-
put com mon term inal .
·
Mid -sca le line arit y and coe ffic ient rati
ngs app ly for sett ings betw een
0.1 and 0.9 . The rati ngs typi call y imp
rov e belo w 0.1 in pro por tion to
the squ are roo t of the sett ing and abo ve
O. 9 in pro por tion to the squ are
roo t of l min us the sett ing.

Abs olut e ~inearity

Init ial: :t:O. 5 ppm at mid -sca le, imp rovi

ng to zero at end sett ing s.
Lon g-T erm : :t:l ppm at mid -sca le, imp
rovi ng to zero at end sett ing s.
Cer tific ate- Cor rect ed: :t:O. 2 ppm at mid
end sett ings . -sca le, imp rovi ng to zero at

Ter min al Lin eari ty (Re lativ e to lnpu t

·Ter min als) : Sam e as abso lute
line arit y exc ept for end volt age dro ps
not exc eed ing O. 5 ppm for 10 kO
divi der or o. 05 ppm for 100 kO div ider
.·
Com pen sate d Ter min al Lin eari ty (Re lativ
Sam e as term inal line arit y exc ept that e to Out put Com mon Ter min al):
volt age dro p at zero sett ing is
com pen sate d to :t:O. 01 ppm for 10 kO divi
der and O. 002 ppm for 100 kO
divi der.
Lin eari ty Coe ffic ient s
Tem per atur e: :l:O. 2 ppm /•c at.m id-s
cale , imp rovi ng to zero at end
sett ing s. .
[Link] er: :t:l ppm /Wa t mid -sca le, imp
rovi ng to zer o at end sett ings .
Swi tch Con tact and Wir ing Res ista nce
Var iatio ns
Init ial: Les s than :t:O. 04 ppm for 10 kO
divi der or :t:O. 004 ppm for
100 kO divi der.
Lon g-T erm : Les s than :t:O. 08 ppm for
10 kO divi der or :t:O. 008 ppm
:£or l 00 kO divi der.
Cal ibra tion Con diti ons : 23• c, low inpu
t pow er.
Num ber of Dec ade s: Sev en.

1 - 3
ala.I i RV 77.'J ? /1..1..
Smalle st Step: O. I ppm.
Input Resist ance: 10 and 100 kO (accur acy :1:0. 005%).
Maxim um Input Power : 2. 5 W continu ous• 5 W interm ittent.
Maxim um Input Voltag e: 220 V rms for 10 kO, 700 V rms for 100
kO.
Breakd own Voltag e! 1000 V peak to case.
Calibr ation Data: Certifi ed test report supplie d with Unit gives calibra
data accura te to ±:0. 2 ppm lineari ty (at time of final inspec tion). Calibrtion
a-
tion presen ted in form suitabl e for interpo lation calibra tion of correc
tion
at any dial setting .
Dimen sions: Width 19 in. (48.25 cm), height 5.25 in. (13.3 cm),
depth 8. 4 in. {21. 3 cm).
Weight : 12. 5 lb (5. 7 kg).

1 - 4
eisli RV 722 7/66
 SECTION II
AIDS TO OPERATION

null-b alance
The Kelvi n-Var ley divide r is design ed prima rily for use in
high imped ance circu its. It canno t be
circui ts or as a voltag e sourc e for
or becau se of the additi onal resist ance in
used as a simpl e variab le resist en the
1. lB). The resist ance ratios betwe
series with the tap (see Figur e
not linear ly relate d to the voltag e
taps of a Kelvi n-Var ley divide r are
ratios .

2.1 POWER LIMITATIONS

n preca utions .
To avoid damag ing the RV 722. it is neces sary to take certai or 220 volts
The input voltag e limita tion of 700 volts for the 100 kQ model
exces sive
for the 10 k!l model will nprma lly protec t the instru ment from
DIAL SETT INGS AND
power dissip ation •. Howe ver, WITH CERT Arn
EXCE SSIVE CURR ENT
CIRCU IT CONN ECTIO NS IT IS POSS IBLE TO DRAW VOLT -
T IF THE INPUT
AND PERM ANEN TLY DAMA GE THE rnsTR UMEN
CE SHOU LD BE
AGE IS MAIN TAINE D. FOR THIS REAS ON THE SOUR
POWE R-LIM ITED TO ONE WATT IF POSS IBLE.
2
This can be done by insert ing a resist or with a value R = !
in series with
r is to fuse
the supply voltag e E. Anoth er metho d of protec ting the divide
the outpu t with a 10 ma fuse.

2.2 TEMPERATURE AND POWER EFFECTS

the RV 722
Figur e 2. 2A illust rates the divisi on of curre nt and power in
or is almos t direct ly
resist ors. The tempe rature rise in each resist
propo rtiona l to the power applie d to it.
CURRENT TEMPERATURE RISE

Figure 2.2A

of the eleven
Uneve n power distri bution result s from the shunti ng of two
decad e (refer to Figur e l. IA). As
resist ors in each decad e by the next
bution , the shunte d resist ors do not
a res'ul t of this uneve n power distri high
the divide r is to be run at fairly
get as hot as the -rest. There fore, if ed for
a short time should be allow
powe r and the first decad e is switch ed,
the redist ributi on of heat in the resist ors.

2 - 1
eleli RV 722 · 7/66
2.3 LOADING EFFECTS

W11en a load is placed across the output of the RV 722, the output voltage
will change. This change in output voltage may or may not have to be con-
sidered in the measurement being made• depending on the degree of accuracy
desired. With high impedance loads the effect of the load may be less than
the effect of the linearity deviation. In this case the change in output voltage
due to output loading can be neglected.

The RV 722 can be represented by an equivalent generator and output resis-

tor in series 1 as shown in Figure 2. 3A. With this equivalent circuit and a
known load resistance, the effect on output voltage can be easily analyzed.
Variations in output voltage can be expressed as LINEARITY DEVIATION
(deviation from nominal output in proportional parts of the input), or as OUT-
PUT DEVIATION (the deviation from nominal output in proportional' parts of
the output). ·
Rour

Figure 2.3A

SE 11-1 T. .
·----- --0
,,
THEVENIN EQUIVALENT CIRCUIT

Figure 2. 3B is a plot of the output resistance of the RV722 as a function of

its dial setting. The resistance values are symmetrical. about a setting of
.500,000,0. For example, the output resistance at a setting of .985,000,0 ·
will be the same as that for a setting of .015,000,0. The resistance corre-
sponding to • 255, 555, 5 can be found in Figure 2. 3B. With the Kelvin-Varley
circuit the maximum output resistance occurs at settings of • 455, 555, 5 and
• 544, 444, 4. The output resistance measurements are made by shorting the
input terminals together and measuring the resistance across the output
terminals.

2 - 2
elsli RV 722 7 /66
 The nomogr a.m in Figure z. 3C is for determi ning the linearit y deviatio n
and output deviatio n when the load resistan ce, output resistan ce and
dial setting are known.

8 Ra. IOMEG
LOAD
RESISTANCE
IN OHMS OUTPUT DEVIATION LINEARITY DEVIATION
-0.001%
IMEG ~IOpprn
2
RouT
IOOk OUTPUT
RESISTANCE
-0.01%
IOk IN OHMS ..... .....,c....,,..
.....~114
2 -tOOppm
,
......
..................&
Ik ..,,,,,. a
,,,.,,,,,,
-0.1°4
IOk 00 -1,000pprn

,,,,..?,.,,, '
,,,.
lie
TO USE:
A. Enter with R1.. oncl RouT• Find OUTPUT DEVIATION :: - I
OUTPUT DEVIATION on ltft scolt l+..!!L
RouT -10°4
100 B. Tronsftr OUTPUT DEVIATION to
riglll teole
LINEARITY DEVIATION : S11. OUTPUT DEVIATION
C. Enter wlil't OUTPUT DEVIATION
2 on right scale and with S. Find
LINEARITY DEVIATION -10%
10

NOMOGR AM FOR OUTPUT DEVIATION AND LINEARITY DEVIATION DUE TO OUTPUT LOADING
Figure 2.3C

2 - 3
elsli RV 722 3/64
2.4 USING CERTIFICATE [Link]

Each RV 722 is factory calibrated before shipment. The calibration cer-

tificate accompanies the . instrument and can be used to increase measure-
ment accuracies. The iertificate includes four separate correction graphs.
These graphs may be u~ed to determine the amount of linearity deviation* at
any dial setting.

A sample of the calibration certificate graphs is shown on the next page.

These graphs show the linearity deviation contributed by the first decade,
the second decade, the third decade and by internal lead and contact resist:..
ances.

Notice that the bottom graph has two lines plotted on it. The line extending
from zero deviation is the plot of the linearity deviation due to internal lead
and contact resistance relative to the output common (terminal 4). The other
line is the same plot relative to the input common (terminal 2). Observe
that with a dial setting of • 567, 890, 0 it is necessary to use the bottom two
graphs. Also, with a dial setting of • 005, 678, 9 neither of the top two graphs
would be used because the error of the first two decades is zero.

EXAMPLE 1:

The linearity deviation cor;rection for a dial reading of • 678, 912, 3

would be; (-0.13 ppm)+ (+0.05 ppm)+ (.00 ppm)+ (-.03 ppm)= -.1 ppm.

EXAMPLE 2:

The linearity deviation correction for a dial [Link] of • 005, 6 78, 9

would be; (-.01 ppm)+ (-.03 ppm)= -.04 ppm. This correction
does not change the dial reading of • 005, 678. 9.

*NOTE: The output voltage of a perfect divider is the product of the

setting and the input voltage;

EOUT = SEIN {for a perfect divider)

Linea.:rity de-viation Ir , indicates how close the output voltage comes to

this perfect value in proportional parts of the input voltage.
EOUT - SEIN EOUT
Linearity Deviation =r : : EIN
=
EIN
- S

2 - 4
elsli RV 722 3/64
 MODEL RV7 22 KELVIN"' VARLEY VOLTAGE DIVIDER

~
I
..
~

l l
I
i.
I

I
I
I
I
0.3 0.4 0.5 0.6 o. 7 0.9
SETTING OF All 7 DIALS

w +0.1
0
<(
0
LU
0
0
z
0
0
w
(/)

. r

0.02 0.03 0.04 0.05 0.06 0.07 [Link] 0.09 0.1

SETTING OF LAST 6 DIALS

w +0. 05 M"'M"1'T'!"M"'f"l'"'"r"'~-ro-~l"M'T"!"l"'l"'!"!"!'!"'l"l"'l"M'"f'!"T''!"T"l"t"l"l"l"'!"'l'l"M"!""T'l'"f"M"l"l"'!"'I',.,....,..,.,.,..,.,.,
............... .....,.,..,.,................. ..

0
<(
0
w
0
Q
a:

0.9
0.1

23 _ •c
TEMPERATURE
2 - 5
eleli RV 722 3/64
 SECTION Ill
CALIBRATION

er, howe ver, to use its

The RV 722 is a highl y stabl e and accu rate divid
shou ld be reca libra ted
shor t-ter m stabi lity to the fulle st adva ntage it
r whic h the divid er is
shor tly befo re being used . The cond ition s unde
re and excit ation volt-
calib rated shou ld dupl icate the ambi ent temp eratu
the effec t of temp era-
age as close ly as is prac tical . This will mini mize
ture diffe renc es and slow drift s in the divid er.

"
3.1 EQU IPME NT REQUIRED
er with prov ision s
1) A ten-s tep, lOOk input total , resis tive divid 1010
Mod el SR
for inter comp aring the indiv idual resis tors (ESI
Resi stanc e Tran sfer Stan dard , I Ok per step) .
to a IOOk load , powe r-
2) A de sour ce capa ble of deliv ering 300 volts
DC Gene rator s).
limit ed to one watt (ESI Mode l 801 1 803, or 820
ovol t mete r with a
3) A high impe danc e (I mego hm or grea ter) micr
ving one- half
l 00 micr ovol t full scale rang e and capa ble of resol
or 810 DC Dete ctors ).
micr ovol t or less (ESI Mod el 801
eters (ESI Mode l
4) Low resis tanc e lead- comp ensa ting pote ntiom
LC 875B Lead -Com pens ator) .
eters shou ld be isola ted
The de sour ce and lead- comp ensa ting pote ntiom
from grou nd by at least IOll ohm s.

3.2 TRANSFER STA NDA RD CAL IBRA TION

ard and calcu late the
Inter comp are the resis tors of the trans fer stand
linea rity deviat~on ( r)
by the form ula;
1 IOI

· r = TO ~ (An - A Av)
tor and [Link] is the
wher e An is the meas ured devia tion of the nth resis
aver age devia tion of the ten resis tors.
latio n.
Figu re 3. 2A iUus trate s an exam ple of such a calcu
LINEARI TY
RESISTOR
NllMRER
MEASUREO DEVIATION
PPM f~/~~fj,~Cfr.~R.?p~ DEVIATION r PPM

10$ t:.. A. -A1i.v i51s (A.-A.v)

1.4 +0.311 +0.04
I
+0.8 -0.24 +0.01
2
) +&.O +!.H +0.41
+t.2 +1.16 +0.$2
4
+I !6 +0.66
:r. +2.41
+0.!'>2
& -o.4 -1.441
7 +0.4 -0.641 +0.45
+0:11
e -t.41 -- 44 +0.14
ll +1.4 +o.u
10 -0.4 -1.44 0

TOTil.L I A,• +10,4

•"l
0

TS OF
LINEARITY CALCULATED FROM RESISTANCE MEASUREMEN
ARD
ANES I MODEL SR 1010 DECADE RESISTANCE STAND
Figure 3.2A

3 - 1
eleli RV 722 7/66
3.3 DETERMINATION OF ABSOLUTE LINEARITY

The process for determi ning the absolute linearit y of the divider is as follows:

1) Connec t the equipme nt as shown in Figure 3. 3A to determi ne the

linearit y deviatio n of the first decade of the RV 722. ·

0
0

100v
LEAD COMPENSATOR s.------f•lt-----t~COM;:~~ATOR
DlllTl!CTO ft

I;
6Qle~04l000@(~~~°'°~o1,:;: ~

figure 3.3A

2) Adjust the input voltage to 1 OQ volts for first decade. - 10 volts for
other decades .

3) With both divider s set at zero (as shown) adjust the lead compen sator
so the outputs are within O. 1 ppm of each other ( 100 microvo lts = 1 ppm
for first decade - 10 microv olts =
1 ppm for other decades ).

4) With both divider s set at full scale, adjust the lead compen sator so
that the outputs are within 0.1 ppm of each other.

5) Measur e the linearit y differen ce between the two divider s at nomina lly
equal settings . Two microv oltmete r reading s are taken at each setting
(with the exceptio n of the last decade) , one with all the decades to the
right of the first decade set to zero and one with all
the decades to the
right of the first decade set to their maximu m setting. Be sure that
the microv oltmete r reads a positive differen ce when the RV 722 ratio
is greater than the ratio of the transfe r standar d.

6) Correc t the transfe r standar d reading s by subtrac ting the linearit y

deviatib n of the transfe r standar d from these reading s. Divide the
reading s by the appropr iate power of ten (divide the second decade
reading s by ten, the third decade reading s by 100, etc.) to convert
them to linearit y deviatio n contribu tions express ed as a fraction of
full-sca le for the whole divider .

3 - 2.
elsli RV 72.Z 8/70
 7) Plot the measured deviations as shown in Section z. 4.
8). Connect the equipment as shown in Figure 3. 3B •

•
. . '

• .figure 3.38

9) Repeat steps Z [Link] 7 except in step 5 where the two microvolt-

meter readings are taken at each setting, one with all .the decades
to the right of the second decade set to zero and one with all the
decades to the right of the second decade s~t to their maximum
settings.

3 - 3
elsli RV 7ZZ 3/64
I 0) Connect the [Link] as shown in Figure 3. 3C.

figure 3.3C

11) Repeat steps Z through 7 except in step 5 where the two microvolt-
meter readings are taken at each setting, one with all the decades
to the right of the third decade set to zero and one with all the decades .
to the right of the third decade set to their maximum settings.

IZ) The remaining decades may be calibrated in a similar manner, if

desired. However, normally, this data is not used.

3 - 4
elsli RV 7ZZ 3/64
 3.4 DETERMINATION Of END CORRECTION

The proced ure used for determ ining the end cprrec tions for the RV 722
is as follows :

l) Conne ct the equipm ent as shown in Figure 3. 4A and adjust the

input voltage to the RV 72.2 to 300 volts.

2) With all of the dials set to zero, measu re the voltage betwee n the
output tap (termi nal 3) and the input commo n (termi nal 2). This
reading is the low-en d correc tion.

3} Conne ct the equipm ent as shown in Figure 3. 4B. With all of the
dials set to zero, measu re the voltage betwee n the output tap
(termi nal 3) and the output commo n (termi nal 4). This readin g
is the comp_e nsated commo n correc tion.

4) Conne ct the equipm ent as shown in Figure 3. 4C. With all of the
dials set to full scale, measu re the voltage betwee n the output tap
(termi nal 3} and the input (termi nal 1). This readin g is the full-
scale correc tion.

5} Plot these readin gs as illustr ated on the last graph in Section z. 4.

In these measu remen ts a readin g of 3 microv olts equals O. 01 ppm.

figure 3.4A

,?1
figure 3.48 0
[
\

--------~v,...----------.

0
.,. ,;,~ .
.,. .~ ....
....""'''" 0

0
0
-Wl!'.~"-i!'i!llll':~ ~m=t

3 - 5
algli RV 722 3/64
3.5 CHECKING THE POWER COEFFICIENT

The power coefficient of the RV 722 is l ppm per watt (less for low settings);
If no more than 100 volts is applied to the RV 722 (0.1 watt), or if it is not
necessary to utilize the fullest capabilities of the divider, then there is no ·
need to be concerned with the power coefficient. However, if maximum ac-
curacy is necessary in situations where a significant amount of power is
applied to the divider, it is necessary to know the power coefficient of the
unit.

The recommended procedure for checking the power coefficient is to com-

pare the divider at low and high power to a transformer-type divider which
will not change linearity even with wide variations in the input voltage. See
Figure 3. SA for details of the setup. If the same frequency is used at both
low power and high power, then the ac-dc difference will be the same at
both powers and the power coefficient at each setting will be the linearity
change divided by the power change.

Figure [Link]

3 - 6
e!sli RV 7 22 3/ 64
 SECTION IV

PARTS LIST

The followi ng parts list is in alpha- numer ical order. Manuf acture r
of the part is given in a code numbe r accord ing to the Federa l Supply
Code for Manuf acture rs; see list of manuf acture rs below.

Parts manuf acture d by Electr o Scienti fic Indust ries must be ordere d
from the factory . When orderin g parts from the factory , include
the followi ng inform ation:

Model and serial numbe r of the instrum ent

Electr o Scient ific Indust ries part numbe r
Descri ption of part

CODE LIST OF MANU FACTU RERS

11837 ELECT RO SCIEN TIFIC INDUS TRIES

Portla nd, Oregon

ESI
Descri ption Mfr Part No. Qty Used

Cap, Bindin g Post, Black 11837 1170 4

Cap, Bindin g Post, Gold 11837 1172 2

Bindin g Post 11837 1393 6

Dial, 2nd, 4th & 6th 11837 7808 3

Dial, 1st 11837 7819 1

11837 8130 I
Dial, 7th

Dial, 3rd & 5th 11837 8131 2

Dust Cover 11837 1583 1

. Knob, Bar 11837 1266 7

Probe Tip 11837 8145 2

4 - l
elsli RV 722 7/66
. .•
ESI ·
Description Mfr Part No, Qty Used

Switch Assembly, Resis- 11837 '8139 I

tance, 1st Decade*

Switch Assembly, Resis- 11837 8140 I

tance, 2nd Decade*

Switch Assembly, Resis- 11837 8141 1

tance; 3rd Decade*

Switch Assembly, Resis- 11837 8142 3

tance, 4th*, 5th, and 6th
Decades

Switch Assembly, Res is- 11837 8143 1

tance, 7th Decade

Window 11837 7242 3

Window 11837 7243 4

* The first four decades of this device require unique final

adjustment procedures·. Replacement or repair of these
decades or their component parts by other than ESI
factory personnel or authorized repair service voids
warranty.

'
R
IOk 2k [Link] Ik [Link] 'I k [Link] lk lk

--· 4 - 2
elell RV 722 7/66
 SECTION V
APPENDIX

When using the RV 722 in a measuring system, the selection of an appropri-

ate detector is extremely important. ESI Engineerin g Bulletin No. 23, v
"A New Approach to Bridge Sensitivity11 , explains in considerab le detail the
generator and detector requiremen ts of bridge circuits. This material is
directly applicable to the RV 722. In association with EB 23 the families of
source describing points for RV 722 input voltages of 700 volts, 100 volts
and 10 volts have been plotted on the graph below. Any detector whose
curve falls below these families of source describing points will re~olve at
least 0.1 ppm for any of the input voltages indicated (700V, IOOV or IOV).

Other ESI literature that applies especially to the Model RV 722 is:
DC and Low Frequency AC Ratio Measu·rem ents, Engineerin g Bulletin No. 29 .;
Calibration of a Kelvin Varley Voltage Divider, Engineerin g Bulletin No. 24
A Resistance Bridge Made From a Voltase Divider, Engineerin g Bulletin No. 38 ./

)('~
)( J~~
·0
, ,,,a '\.. .ID<
//. )IVv
(/, /:>
)(' ,, f':
~ 'I>~
"/,
,.1'.?
'

"-,,, 5'-> -Oo ' I'\. "''.'\-.;:

"
)(

~ "'
'\..[Link]. /.

~
)<Vv
/>( 'Y< " (1'..o IA ? I
F0'
">-
-
.
x.· "' ' ' '/,,.
'
i-? " ~
.)<~

~~V1
'\. /.
"'"
~ "~
}'

~
)i '
/>(
·.Oo
-"'·~ '·'

)( :'>~ ~~ '\.. /,'

)(
'

-
'\.. /P'; " "~
/' ::>:''-'
/ 0'-..:t
'"
-
"'A ~ ~
-·
" )(~,)

..
-
x ,.... ,,'· ~

x ,) ' " / . i'\ ><".,. '

" '\,/,~
~ " )(~ ,/, 'x/ " v fV ' )o ,.1'. ?,5~

,,,
~o,.
)(

,, ~1k..~" ',
M
II, ..,.. <..'
,)['\

'->lt( xv.
!/I">
" ')( '\.. /

10-ae ~ 'Vi><
[)< "

>< ...,
)( D~

IQ-IT .,,,'X
100
" )(~
,•
'~
' '\.. i/I">

1.-1 "X','f>-', "

ao
........
:1

~'
)(',,~

"[';
• :>. '\..ti<

" ,_/'Si'
. /rx J
100 1
'/I'

1
.,
..__-ba----,-------~
0
.....
I
THEVENIN SOURCE RESISTANCE

5 - 1
elali RV 7ZZ 7/66
 •n•W....,••
AND
ACCESSORIES

e~~les:----------•
usf:r
no.23
Sle cf:r o Sc len f:lf lc l n dAND IS!», OREGO N NOVEMBER 1962
7524 S. W. MACAD AM AVE.• PORTL
REPLACES JANUARY 1961

I
A IEW APPROACH TO IRID8E SEISITIYITY
by
Jack Riley

Prese nted as Pape r Numb er 23-SL -61

on Janu ary 18, 1961 at the

INSTRUMENT SOCIETY of AMERICA

WINTER INSTRUMENT-AUTOMATION CONFERENCE AND EXHIBIT
ST. LOUIS, MISSOURI • JANUARY 17-19, 1961

•cop,,le ltt IH.'l b1 fledro Sclentltc fndwtrlea, '11c.

LITHO IN U.S.A.
 A NEW APPROACH TO BRIDGE SENSITIVITY

Ever since Christie invented the Wheatstone bridge in 1833, people have been
concerned with bridge sensitivity. In 1862, Thomson released a paper describing
his Kelvin double bridge ( 1 ). The sensitivity analysis in his paper is believed to be
the first published. He concluded that the heating effect of the test current was the
limiting factor. This conclusion was overlooked for the next three decades mostly
because of the low quality of the batteries available for powering bridge circuits.
Schwindler, in 1866, Heaviside in 187 3, and Gray in 1881, all attempted bridge
sensitivity analysis. The second and third editions of Maxwell's "Electricity and
Magnetism, 11 published in 1892, include a section on bridge sensitivity. The mod-
ern approach to bridge sensitivity analysis is given by Wenner(l) in Volume 25 of
the "Journal of Research of the National Bureau of Standards. 11

Most of the writings about bridge sensitivity finish with an impressive algebra-
ic expression which must be evaluated for each possible setting of the bridge. This
paper presents a graphical analysis offering several advantages. The calculations
for a large number of bridge settings can be performed rapidly. The behavior of
the ~bridge and its generator are separated from the performance of the detector.
The suitability of various detectors can be checked in a simple manner. Sensitivity
curves for different bridge and detector combinations can be constructed easily, or
sensitivity can be deduced from the bridge and detector plots directly.

The graphical technique for analyzing a Wheatstone bridge is given. A typical

circuit is used as an e::.::ample and the performance for a particular measurement
is followed through all of the steps. The performance curves for a commercially
available bridge are given to show the practical application of the technique.

WHAT IS SENSITIVITY

A definition of sensitivity is need·

ed before a bridge can be analyzed.
Sensitivity is defined as the ratio of
the smallest discernible change of the
measured quantity to the total value of
that quantity. If a 1% change in the
resistance being measured produces a
barely perceptible movement of the
null indicator, the sensitivity is 1%.

GRAPH PAPER
I 10 100
Special graph paper simplifies the '---OHMS )

graphical analysis. The coordinates

of this paper are log power versus log
resistance; with contours of equal volt- FIGURE L GRAPH PAPER
age and equal current drawn on the
diagonals, as shown in Figure 1. With sents the resistance value of a resistor,
this paper Ohm's law and the power the voltage across the resistor, the
equation are solved simultaneously . A current through the resistor, and the
point plotted on this graph paper rep:re- power dissipated by the resistor.

felsW EB-23 11/62

..
ANALYZ ING A BRIDGE

The sensitivi ty analysis technique

can be illustrate d by applying it to the
bridge circuit shown in Figure z. The
E
D
o<: "\.;
SOURCE
REStSTANCE

THEVENIN
LOAO

RESISTANCE
~'>
SOUACE
RESISTANCE°'.>

NORTON
LOAO
l!ESISTANCE

UNKNOWN
RESISTOR

RATIO
RESISTORS

BRIDGE
BALANCE
RESISTOR

LOAO RESISTANCE

FIGURE Z. BRIDGE CIRCUIT

FIGURE 3. SOURCE REPRES ENTA-
compone nts for this bridge, the gener- TION
· ator which will drive it, and the detec-
tor which is used as a null indicator · a constant current approxim ately equal
are chosen, then the analysis is made. to the short-cir cuit current is supplied.
·First the bridge output resistanc e and If the load resistanc e is high, a con-
open-circ uit voltage are found. Their stant voltage approxim ately equal to
effect on a detector is then studied. the open-cir cuit voltage is supplied.
The curve can be predicted on the basis
of either the Thevenin equivalen t gener-
THE GENERA TOR ator consistin g of a zero-imp edance
voltage source in series with a fixed
It is particula rly informat ive to resistor, or the Norton equivalen t cir-
plot the equivalen t Thevenin or Norton cuit consistin g of an infinite impedanc e
generato r performa nce on the graph current source feeding a fixed internal
paper shown in Figure 1. For either resistanc e and load resistanc e in par-
type of equivalen t circuit (and for most allel. The load. curve can be plotted
practical circuits) the open circuit from either form of equation (1).
voltage, the short circuit current and
the source resistanc e all intersect at
a single point. This can be called the
"source describin g point". This sin-
gle point on the graph paper as shown
in Figure 3 serves to complete ly de- )
it R1,.
scribe the output of the generato r in a P • ( IscRs ( · )11
RL+Rs
manner which can be later used to find
the amount of voltage or current it will
actually supply to a bridge.
The open-circ uit voltage and short-
In Figure 3 the power supplied to circuit current will intersect at the val-
the load is plotted versus the load re- ue of internal resistanc e. The power
sistance. If the load resistanc e is low, at this intersect ion point represen ts the

[elsliil EB-23 11/62

 -3-

maxim um amoun t of power which can If the bridge of Figure 2 is mea-

be dissipa ted in the interna l resista nce suring a five-ki lohm resisto r and if re-
of the genera tor. This interse ction sistors C and Dare each one kilohm .
point is enough to comple tely descri be the bridge will be a three-k ilohm load.
a given genera tor. The curve is always The genera tor will supply about 53 volts
the same shape and can be drawn with to this load as shown in Figure 4.
a templa te once the interse ction of open
circui t voltage and short- circuit cur-
rent is found.
INPUT RESIST ANCES
The genera tor chosen for the bridge
shown in Figure 2 consis ts of three 45-
volt batteri es in series with a protec tive The bridge input resista nce is
resisto r of 4. 7 k. The perfor mance of needed for all values of unknow n re-
this genera tor is shown in Figure 4. sistanc e to be measu red. The input
resista nce is found by paralle ling re-
sistors A and C in series with B and D
in series . At balanc e, resista nce in
the detect or circuit has no effect.
Equati on (2) gives the input resista nce.

R1N • D (A+C) (2)

(O+C)

ll:'.
The curve for the input resista nce also
w has a charac teristi c shape and can be
~
CL drawn with a templa te. Two calcula ted
...... values will locate the templa te so that
::;)
CL the comple te range of values can be
~IOOmwl!"-~~--J11,.----t--r--1-~...-~-Jl!'.---~--.v
plotted . If A is low 1 B is also low so
0
the input will look like resisto rs C and
D in paralle l as shown in equatio n ( 3).
IOmwl'--...-..;;,..,.,,.,..i,+_._,~.......,~-.-..,..;.,...,.,.,.;.........,................,+fWI
100 Ill IOOk

LOAD RESISTANCE R1N

A, B«C,O
> (3)

FIGUR E 4. ·GENE RATO R PERFO RM-

ANCE
If A is high, the input resista nce will
When this genera tor is shorte d, short- approa ch the value in equatio n (4).
circui t curren t of about 30 ma will flow
throug h the protec tive resisto r causin g
it to dissipp .te about four watts. The
maxim um output poweI'. from the gener-
A>>C
> A(
I+%
I ) (4)

ator is approx imatel y one watt. It can-

not da1nag e any bridge or detect or co1n-
ponent s which are capabl e of dissipa ting These two asymp totes will interse ct
one watt of power. This form of power when A = C. The resulti ng curve is
protec tion can often be used to advant age. plotted in Figure 5.

!eis[U EB-2.3 11/62

 -4-

at the detector terminals to the input

voltage. When the bridge is balanced,
there will be no voltage at the output
terminals, so the bridge must be un-
balanced by an amount /:i. to find a sig-
nal to deted. A bridge circuit can be
considered as two voltage dividers. At
balance both dividers are at the same
setting so that their output voltages are
equal. Before the bridge is unbalanced,
both of the dividers are set so that the
voltage EA across resistor A (the same
voltage is also across resistor B) is
10010\,--__,,_....._..........,,,,,00~................."""""!'lk!---'--'................,!-:-.................~
related to the input voltage by equation
UNKNOWN RESISTANCE A
(5). When the bridge is unbalanced by

FIGURE 5. INPUT RESISTANCE

A
= -
A+C
(5)
INPUT VOLTAGE
changing the unknown resistor A to
[Link] bridge input voltage can be A(l +A) the voltage across resistor A
found from the generator performance changes to E 1 which is related to the
curve {Figure 4) and the input-resist- input voltage ty
equation (6). The volt-

A(I +A)
(6)
A(i +Ii) +C

age EA across resistor B does not

change• so the difference between it
and E'A is the open-circuit output volt-
age E U which is found from equation
(7). f>ac'foring equation (7) gives equa-

A(I +Ii) A
A+C
(7)
A(l+A)+C

UNKNOWN RESISTANCE A
tion (8). When A is small, as it usually

Eour ACA
FIGURE 6. INPUT VOLTAGE II {8)
ErN
ance curve (Figure 5). The results
are plotted in Figure 6. A few points
are usually adequate for plotting the is, equation (8) simplifies to equation
entire range of values. (9) which is used for most actual cal-
culations. The plot of equation (9) has

TRANSFER FUNCTION

A transfer function can be calcul-

ated to relate the open circuit voltage

lelsliJ EB-Z3 ll/6Z

 -5-

the same shape as the power curve so

that the same templat e can be used for
drawing it. The charact eristic point
for drawing the curve will have an
ordinate of one and an absciss a of the [Link].------1------+-----~---
--i

resistan ce of C. Figure 7 is a plot of

0.01

10·•

...i!,"°!
...iw
FIGURE 8. VALUE S OF A
5k

OUTPU T VOLTA OE

io·•,o.__..._._._..........1u..oo,_...........................,k_.....................,..!-...._.........,..........,_ From Figures 6 1 7 and 8 the out-

UNKNOWN RESISTANCE A put voltage can be found. Figure 9 is
the plot of output voltage for the bridge
of Figure 2. Both the comple te curve
FIGURE 7. TRANS FER FUNCT ION for l "/o unbalan ce and the curve for the
combin ed 1% plus one dial division are
the transfe r function for the bridge of shown.
Figure 2 if C is one kilohm.

~
UNBAL ANCE A !J
~ IOrnvl:'-----.-----T----t----1------1

The amount of unbalan ce A is a

matter of choice. To find sensitiv ity
I
l::
a fixed value is assigne d to A and the
a
~ lmv
detecto r respons e is found. From this
~
511
the sensitiv ity can be calcula ted.
Bridge accurac y or resoluti on values
can be as signed to A to find' the use-
fulness of a detecto r. For the bridge
being analyze d A is chosen as the
specifie d accurac y and is given by
equatio n (10). FIGURE 9. OUTPU T VOLTA GE

!). = :t (1% +I DIAL DIVISION) (10) OUTPU T RESIST ANCE

To comple te the equival ent-gen er-

.0. is plotted in Figure 8. ator circuit for the bridge output the

f;!~[i) E;B ... 23 11/62

 -6-

• output resist~nce must be found. The BRIDGE OUTPUT EQUIVALENT GEN-

output resistance is the parallel com- ERATOR
bination of A and B in series and C and
l> in series. At balance the output re- The output voltage (Figure 9) and
sistance is given by equation (11}. output resistance {Figure 10) can be
combined to determine the output that
the bridge will deliver to any load re-
R - A(D+C) (II) sistance. The output voltage and out-
OUT - (A+C) put resistance meet at the source des-
cribing point of Figure 3. The values
found when measuring a five kilohm
The curve of Ro T can be plotted from resistor have been illustrated in each
two calculated vRues by using the tem- of the grap~s. With this [Link]
plate that plotted the [Link] resistance. the output voltage is 74 millivolts and
When A is low ROUT is given by equa- the output resistance is 1. 7 kilohms.
tion (12}. The curve showing voltage, current
and power supplied by this source to
various load resistances is given in
RouT -->--
A«C
A t,~ + Ac) (12) Figure 11.

When A is high the resistance becomes

C and D in series as shown in equation
( 13). ....
U>

~
~

RouT A»C> c (1 + ~) (13) ~ 10-•llE-----+--+--

ffi
The asymptotes intersect at A = C.
The output resistance for the bridge ~
10· 1 ...,,~~~llTl'o......,...,..~"'""'~~.,-,i..-.,~
of Figure 2 is shown in Figure 10.
100 lk IOk
RESISTANCE IN OHMS

IOk ----...-----,..-----~-----,

FIGURE 11. BRIDGE EQUIVALENT

GENERATOR FOR ONE-
PERCENT UNBALANCE
WHEN MEASURING A
5 k RESISTOR -

Similar curves for each value of

unknown resistance could be drawn.
They would cover most of the paper
C Sk and therefore be useless •. Fortunately
a plot of only the source describing
points contains all of the necessary
1010!:--'--'~..[Link],o!!::::o-'--'-'.....,."'"!,~k__,,_._........"':';!:::__.._.......~
information. This graph is made by
UNi<NOWN RESISTANCE A
plotting the intersection points of the
output voltage and output resistance
for the values of unknown resistance
FIGURE 10. OUTPUT RESISTANCE of interest. Figure 12 is the result

lelslil EB-23 11/62

 -7-

INPUT
RESISTANCE
i==
z

~
w
ff:::::> INPUT
RESISTANCE
0
I-
3 VOLTMETER AMMETER
&i
u
I-
0::
0
:t
~ 10·• .
)(

w
~
~
0 10· 1
>
I-
3
(..)
a::
0 10·•
z
w 10 100 lk
0..
g SOURCE RESISTANCE

FIGURE 12. BRIDGE EQUIVA LENT SOURCE RESISTANCE

GENER ATORS FOR ALL
VALUE S OF UNKNO WN
RESIST ANCE
FIGURE 13. DETEC TOR REPRE -
SENTA TION
The solid curve is for a A of ::I:: { 1% + 1
dial division ). The dotted curve is for
a A of:!: 1% only.
shape as the generat or curve, but it is
inverte d. If a bridge source describ ing
point lies anywhe re on the curve the
THE DETEC TOR detecto r will produce the deflecti on
represe nted by the detecto r describ ing
It is [Link] possibl e, as shown in point for the curve. The detecto r curve
Figure 13, to .,.scribe -one deflecti on can be found experim entally (or calcul-
of the detecto r by a point. The detec- ated) by finding which combin ations of
tor describ ing point represe nts its in- open circuit source voltage (or short
put voltage , current • and resistan ce circuit source current ) and source re-
for one deflecti on. A curve can be sistance will give the indicate d deflec-
drawn which represe nts the locus of tion of the detecto r. Differe nt amounts
generat or describ ing points which of indicati on are represe nted by differ-
would prod,uce this specifie d detecto r ent describ ing points and their corre-
indicati on. The curve .has the same spondin g curves as shown in Figure 14.

l;als[ij EB-23 11/62

 -8-

~I~ I...!__!$-~._!££)
FIGURE 15. BRIDGE AND DETECT OR
n ; MEG PERFOR MANCE
SOURCE RESISTANCE

FIGURE 14. DETECT OR CURVES SENSITIV ITY CURVE

The graph provides not only qual-

BRIDGE PERFOR MANCE itative data• but quantitati ve data as
wen. The bridge output voltage is al-
A generato r describin g point, such most directly proportio nal to A. The
as those plotted in Figure 12 1 .which is differenc e between the value of A which
above the detector curve will produce was. chosen and the value of A which
greater deflectio n than that for which would supply minimum detectabl e sig-
the curve was drawn. A generato r nal can be found by vertical measure -
point below the curve will not produce
the indicated deflectio n. If the detector
curve is drawn for minimum detectabl e
deflectio n (noise level, smallest ob-
servable movemen t 1 etc.) it will sep- l
arate generato r describin g points into
those which can and cannot be detected.
For the bridge of Figure 2 a detector
10%
--+i -
with one-kiloh m input resistanc e and a I
minimum observab le deflectio n of ten I
millivolt s was chosen. In Figure 15
this detector plot has been added to
Figure 12. Thft sensitivi ty is sufficien t
to see rated accuracy up to about 40
0.1 'I'..............._._._..........,__..._-1-J...W..L. Ul...--'-....l-.J..Lu.i.L L--.J..-l....Wu..i.J..!. !
kilohms of unknown resistanc e. If the 10 100 Iii

bridge controls had sufficien t resolutio n UNKNOWN RESISTANCE A

the detector could find one percent dif-

ferences of resistors down to 100 ohms. FIGURE 16. BRIDGE SENSITIV ITY

lelslil EB-23 11/62

"'
..
ment using the voltage scale. By us- in Figure 17. Here a rather typical
ing this technique, the bridge sensitiv- collection of both ac and de detectors
ity curve of Figure 16 was drawn. is shown with their maximum sensitiv-
ity curves superimposed. It can be
The bridge sensitivity and the seen that some detectors are more
bridge specified accuracy are shown sensitive for low impedance sources;
in similar units. Although Figure 16 others for high impedance sources. A
is a more [Link] method of pre- set of describing point curves such as
senting the data, the same information Figure 12., for a particular bridge or
is available from Figure 15. The instrument can be laid over this set of
performance of various detectors can curves to see which detectors have
be demonstrated more easily by using enough sensitivity for all bridge settings
Figure 15. To predict the performance of interest at the required accuracy.
of a new detector , one need only know
its input resistance and its minimum
detectable signal in tert;ns of either CRITICAL DAMPING
current or voltage. From these data
its curve can be drawn with the tern"'. If a low-resistance source is con-
plate. nected to a galvanometer the pointer
movement will be extremely sluggish.
If the source resistance is too high the
DETECTOR COMPARISON needle will swing back and forth a long
time without settling down to a final
The type of detector curve shown indication. To avoid these problems
in Figure 13 can also be used to com- the source resistance can be kept con-
pare one detector with another as shown stant at the critical damping value by
adding series or shunt resh1tance. The
constant source resistance requirement
modifies the detector curves as shown
in Figure 18.

;;
z . ,If'" i 10..
'~ ~'
r 10··· ~--1-~-¥.=:io.o~----:;~-:.tfc;..._-*_-;-:;~~ ra:
"'
a a Kr''
!:: •o"·~~t--+~~-+--?tr-+r*---t f:
~ ::>
fil IO""
ti u

i• ·o··· ...,.....~~.+-*--+;l4Hlr-+r*---t
- 10···1'.""1
•

FIGURE 18. GALVANOMETER WITH

5 k CRITICAL DAMPING

FIGURE 17. DETECTOR COMPARI- The dotted line at the bottom is the
SON signal which would produce a minimum

·lelslil EB-Z3 11/6Z

 -10-

[Link] deflection of one fifth dial divi- bined with the three damping condition s•
sion without the extra series damping result in nine combinat ions of detector
re sis [Link]. performa nce. The resulting curves for
1/5 [Link] division on the galvanom eter
are drawn only for the ranges for which
PERFOR MANCE OF A MULTI-R ANGE they are applicabl e. The optimum
BRIDGE detector control settings and the mea-
suremen t sensitivi ty can be determin ed
An ES! Model 291 A Resistanc e from the curves.
Bridge Circuit has been analyzed to
show the many useful aspects of the For [Link] , measurin g a 10-ohm
bridge-d etector plots. The resulting resistor usii.::1.g the IO-ohm to 100-ohm
curves are shown in Figure 19. range with the damping switch in posi-
tion 1 about 15 times as much voltage
to.°''Y. + t DtA4.. 0tVISION1 UH8ALANC£ ·Will be available as is needed for a O. 2
,t• scale division deflectio n. By changing
to damping position 2 some 40 times as
much voltage will be available as is
needed for O. 2 scale divisions , but the
meter will be very sluggish. The cen-
ter sensitivi ty position would yield only
about 70% of the o. 2 scale division for
a 0. 0 50/o unbalanc e. It would re qui re
about O. 07% unbalanc e to yield O. 2.
scale di vision.

~ '+.,
IOO, I ~I
----y-
IO~
j,
_!2V
CONCLU SION
OUTPUT RESISTANCE
•
A bridge-p erforman ce analysis
FIGURE 19. BRIDGE PERFOR MANCE technique has been presente d. Separ-
ate curves of bridge output and detector
. The bridge curves are for a capabilit y give a visual concept of
:I: (0. 05% + 1 dial division) unbalanc e. bridge performa nce. The technique is
The solid curves are for a 10-volt useful for analyzing bridge circuits,
generato r; the dotted curves represen t for comparin g detector s, and it can
the bridge performa nce with a 300-volt even be extended to such distant fields
supply. Measure ments to 1 megohm as plotting speed-to rque curves for
are with a resistanc e circuit •. From motors and flux-amp ere turn curves
l megohm to 100 1 000 megohms a con- for magnets.
ductance circuit is used. This bridge
has a light .. beam galvanom eter. Three
different meter damping condition s are
provided . Optimum damping is obtain-
ed by placing a resistanc e in series
with the detector when the supply source
has a low resistanc e• connectin g direct- REFERE NCE
ly to the meter. when the source has the
proper damping resistanc e and placing (1) F. Wenner, J. Res. N. B. S. • Vol.
a resistanc e in parallel when the source ZS, August 1940. Reprinte d in N.
has a high resistanc e. There are three B.S. Handbook 77, Vol. 1, Feb. 1,
sensitivi ty condition s and these, com- 1961.

lelslil EB-l3 11/62

 YULIA\fl
• DIVIDERS

.
®~~
JElec'fro Sclen'flflc
no.29
lndusf:r l••------ ----
7524 S. W. MACADAM AVE.• PORTLAND 19, OREGON SEPTEMBER 1962
REPLACES APRIL 1962

I
DC AND LOW FREQUENCY AC
RATIO MEASUREMENTS
BY
DR MERLE L MORGAN
DIRECTOR OF RESEARCH AND DEVELOPMENT

•
$'ifl!!JIDW ®!n@l!!Ji;> @00 $V/i\OOID&m@$ @~ ~~~'if!nOW OOU$0!lmwll~OO'ir
&oigig OINl$'ii'i~~'ii'&'ii'~ IQ)0Vll$0@~
00~\YJ W®m& $~aim

OCop,,1,W 196.2 by llectro Sclentflk lndulftie1, Inc.

LIT~O ":°'
U.S.A.
 DC AND LOW FREQU ENCY AC

RATIO MEASU REMEN TS

by
Dr. Merle L. Morgan

The :resistiv e voltage , current and impeda nce ratio measur ing devices discus-
sed include simple tapped resistan ce strings , decade divider s, attenua tors,
bridges
and ratio sets.

Transfo rmer devices include simple fixed voltage and current transfo rmers,
decade ratio transfo rmers, and transfo rmer type bridges .

Accura cy, stabilit y and traceab ility of calibrat ion are conside red.

Various types of de and ac detecto rs are reviewe d, and charact eristics of sen-
sitivity and impeda nce matchin g, respons e speed and dampin g, noise limitati ons,
and spuriou s signal rejectio n are conside red.

A discuss ion of techniq ues and pitfalls in ratio measur ement include s ground-
ing, shieldin g, de and ac guardin g, and the effects of ambien t temper ature, power
dissipa tion, and voltage and current levels.

VOLTA GE RATIO If both voltage sources have a ve

high impeda nce, no current can be ·
Any voltage ratio measur ement drawn from either one. In this case:
implies the existen ce of two voltage s an auxiliar y voltage source i e 0 , can:
to be compar ed. We could connect used to supply power to the divider &
voltmet ers to each voltage and divide shown at the right in the figure. The
one reading by the other. But, since voltage is adjusted so that no currerr
we want only the ratio between them, is drawn from either voltage sourcei
the measur ement of the actual voltage s when the ratio is being measur ed. :
will only cause extra work. More im-
portant , it will introdu< ;;e addition al
sources of error in the ratio measur e-
ment.. This is true even if we use a
high accurac y digital voltme ter. There-
fore, when we want to compar e two
voltage s, it is better to use a calibra ted
adjustab le divider .

Typical circuits for such compar ... •o

isons are shown in Figure 1. Each
circuit contain s an adjustab le voltage
divider , the tap setting being repre-
sented by t\le letter S. FIGURE 1. VOLTA GE RATIOS
For the circuits on the left a null
detecto r is observe d while the voltage The voltage ratio to be measur ed
divider is adjuste d until its tap is the is usually produce d by a divider , a
same voltage as. the circuit to which it transfo rmer, an operatio nal amplifi er
is connect ed. or somethi ng of the sort.
.
DIVIDE R RATIO The divider setting S indicate s the
ratio of a part of the voltage to the
If we replace the two voltage s in total. The bridge is calibrat ed to in-
Figure l with another divider , we ob- dicate the ratio between the two parts
tain the circuit of Figure Z. When the into which the total voltage is divided ,
detecto r indicate s a null, the voltage s which is the ratio of the impeda nces.
on both divider taps are equal, so the Both resisto r and transfo rmer circuits
ratios are equal. can be calibra ted to indicate either
type of ratio.

It is also po:f>'!sible to make capac-

itor voltage divi<::e1 ·s. These are par-
ticularl y useful because they are very
accurat e at higher frequen cies. Sirice
FIGURE z. DIVIDE R COMPA RISON we are discuss ing de and low frequen cy
ratio work, we shall concent rate on
resistiv e and transfo rmer type divider s
BRIDGE RATIO with the underst anding that at these
frequen cies, capacit or divider s can be
Each divider may consist of two used in a manner quite similar to re-
impeda nces connect ed in series. This sistor divider s. For example a capac-
is the circuit of the impeda nce bridge itance bridge usually consist s of two
of Figure 3. capacit ors which make one divider and
two resisto rs or a transfo rmer which
make the other.

FIXED TAP DIVIDE RS

In practic al instrum ents voltage

FIGURE 3. IMPEDA NCE BRIDGE dividers take many forms. The sim-
plest is a string of two or more resist-
ors connect ed in series with termina ls
DIVIDE R AND BRIDGE SETTIN G at their junction s. In Figure 5 we see
such a circuit.
If we concent rate on one half of
either of these circuits , as in Figure
4, we see the essenti al differen ce be- iJiil!l MOOEl SR 1010

tween the pair of dividers and the bridge.

RESISTOR TRANSFORMER

DIVIDER FIGURE 5. TAPPE D DIVIDE R

Here two termina ls are provide d

at each junction so that we can use four-
termina l measur ing techniq ues to elim-
BRIDGE inate the effects of wire lead resistan ces.
Figure 6 shows a photogr aph of this unit.

This type of divider is extreme ly

FIGURE 4. DIVIDE R AND BRIDGE useful for high accurac y standar diza-
RATIOS tion of resisto rs and determi nation of

lelslil EB-Z.9 9/62

.
-3-

resistanc e ratios o Its use in calibrat- DECADE DIVIDER S

ing resistive voltage dividers will be
discusse d later. By far the most popular type of
voltage divider for high accuracy ratio
measure ment is the decade divider,
either in its resistive or transform er
form. Several typical decade dividers
are shown in Figure 9.

FIGURE 6. RESISTA NCE TRANSF ER

STANDA RD
Figure 7 shows another type of
voltage divider- .. a voltage calibrato r,
which is used to compare any of several FIGURE 9. VOLTAG E DIVIDER S
With these dividers a voltage can be
split into a fraction which is indicated in
decimal form. The best dividers at the
present state of the art have seven or
more adjustabl e decades. Figure 10 shows
a set of calibratio n curves for a very
high accuracy resistive decade divider.
+.I
0
z
w 0

E
a.
FIGURE 7. VOLTAG E CALIBRA TOR Q.

~
nominal voltage values with the voltage z
of a standard cell. The circuit of this Q
I- kJ
instrume nt is shown in Figure 8. :::>
m ~+.I
... +1ooovO---.-•
j • .,,..o---~ ~ ! 0
...
i;,;i;;m MODEL SV·1948
t -
''la 8 Cl)
!!:
IA.
-.1

~ -.2
~
>
LrJ
....!I

Q -.4
0
....>- +.2
ii:
<( l!::
FIG URE 8. ~+.I
CALIBRA TOR CIRCUIT LU
z
When th,e adjustabl e divider is set :J ~ 0
0

to the certified value of the standard

cell and the input voltage is adjusted
8-.1
w
Cl)
-.2
so that the detector in series with the 0 !I 4 5 6 7
standard cell shows a null, the output DIAL SETT ING
voltage is very accurate ly equal to the FIGURE 10. CALIBRA TION OF A
nominal value selected. RESISTIV E DIVIDER

lelslil EB-29 9/62

.. -4-

The best transfo rmer decade di- When a commo n core provide s magnet ic
viders produce similar accurac ies (see coupling between decades , only one
Figure 11) for low ... frequen cy ac mea ... switch contact per decade is necessa ry.
sureme nts. Extra winding s or extra switch contact s
and conduct ive coupling paths may be
used to improv e the accurac y in prac-
tical designs ..

EQUIVA LENT CIRCUI T

The equival ent circuit of any decade

transfo rmer can be represe nted as shown
in Figure 13.

EQUIVALENT
CIRCUIT
FIGURE 11. CALIBR ATION OF A
TRANS FORME R DIVIDE R
FIGURE 13. RATIO TRANS FORME R

The ideal transfo rmer in the

. TRANS FORME R DECAD E DIVIDE RS equival ent circuit is a device which
has infinite open circuit impeda nce,
A decade transfo rmer can be re- zero short circuit impeda nce and is
present ed by the equival ent circuit of used to specify the transfo rmer ratio.
. The ·actual input and output impeda nces
Figure 12..
can be represe nted separat ely as shown
in the figure. The transfo rmer ratio,
S', is very nearly equal to the turns
ratio. The quantity 5 1 is actually a
CIRCUIT vector quantity which can be calibra ted.
[Link]
It has a magnitu de which may be slightly
differen t from the nomina l value (the
turns ratio) and there may be a slight
phase shift between the input and output
FIGURE 12. DECAD E TRANS FORME R voltage s.

Here each decade is represe nted The calibrat ion certific ate for a
as a transfo rmer winding which is nor- decade transfo rmer gives the deviatio n
mally divided into ten parts. Some- of the ratio S' from its nomina l value
times eleven or twelve decade steps in both magnitu de and phase. The cert-
are provide d to allow extra position s ified ratio will be obtained if no load
at either end. Each knob control s the current is drawn from the output. If
setting of one of these decade s. load current is drawn we must deter-
mine whether its effect can be neglect ed.
Adjacen t df!cades may be coupled
by either of two method s. The decades
may be indepen dent transfo rmers with INPUT IMPEDA NCE
conduct ive couplin g from two taps to
both ends of the followin g winding • or The input impeda nce curves of a typ-
there may be winding s for two or more ical high accurac y decade transfo rmer
decades on one transfo rmer core. standar d are shown in Figure 14.

[els!il EB-29 9/62

 -5-

DECADE TRANSFORME R APPLICA-

TIONS
There are many practical applica-
tions for decade transformer voltage
dividers. The ratios which they pro-
duce can be used in various voltage

~·UJ5
!• and impedance comparison circuits •
•1
1I 'g.t:.~.,::11.~~.. ~,..,
0

! •• 1
. . . '1tALIM' tdl<.Ull •Ot; Mll!IMl1 IJliliifD.!W(l •
Ndl~tKt ""-onc~u DIVIDER CALIBRATION
N1U1 ~ VUIWi NGUINCY

A decade transformer is frequently

FIGURE 14. INPUT IMPEDANCE used to calibrate another decade trans-
former or for ac calibration of a re-
The input impedance varies with sistive decade divider. These divider
frequency. It is inductive at lower fre- calibration circuits are shown in Fig-
quencies, passes through resonance, ures 16 and 17.
and becomes capacitive at higher fre-
quencies. The transformer ratio S'
remains quite accurate over this whole
range of frequencies. The effect of
this input impedance resonance is to
s. II
minimize the current which the trans-
former draws from the voltage source;
the accuracy is not affected. FIGURE 16. TRANSFORME R DIVIDER
CALIBRATION

OUTPUT IMPEDANCE In these circuits, phase compen-

sating capacitors are used to adjust
. The output impedance curves for · for any phase difference between the
the same transformer are shown in two dividers. If the divider on the
Figure 15. right has low reactive output impedance,
such as a transformer, resistance is
added in series with its tap as shown in
Figure 16, so that its output resistance
will be much larger than its reactance.
A capacitor is connected from the de-
tector end of the resistor to either end
of the divider, or a differential capac-
t t"l,..-...,-.,,..,.-,..--.. itor is connected to both ends of the
• divider as shown. This capacitance
!i .i..·_,j.....i.;,+.-"'1j'lloi
!l can be used to draw a small amount of
1~ _..,•.,.,...,"""'l•!l.,..,,"""'!-,............
'------·--COM s ,.~.00 current through the resistor. This
iOUlVAUHf (dKUl'f fOli OUIPVi _..-OAJetl
i . fHAWD(f IN ..[Link]
causes a small quadrature voltage drop
~ t. . MNC• VIHUlfftfQV4Htl

across the resistor. If this voltage is

FIGURE 15. OUTPUT IMPEDANCE extremely small compared to the total
voltage across the dividers, it can be
Note he're that at low frequencies considered to be quite accurately in
the output impedance is· simply the de quadrature and will have a negligible
:resistance of the winding. At higher effect on the magnitude calibration.
frequencies, the leakage reactance If a larger difference in phase is to be
causes sorne inductance to be added to corrected, careful attention must be
the output impedance. paid to see that the current through the

[els[!! EB-29 9/62

 -6-

resistan ce is exactly at right angles to

the applied voltage . Any phase differ-
. ence from 90° in the current through
the resisto r will cause a voltage drop
in phase with the transfo rmer voltage STEP•DOWN VOLTAGE RATIO
being calibra ted, and will affect the
magnitu de ratio measur ement.

Usually very sensitiv e detecto rs

must be operate d with one of their two
termina ls grounde d to prevent stray
STEP-UP VOLTAGE RATIO
pickup at their input termina ls. The
tap of the transfo rmer at the left is FIGURE 18. TRANS FORME R RATIO
grounde d so that this end of the detec- MEASU REMEN T
tor can be grounde d.
provide an auxiliar y supply such as
another divider to drive the lower volt-
age winding of the transfo rmer being
tested. This voltage must be adjuste d
so that the voltage on the higher voltage
winding is exactly the same as that ap-
plied to the standar d divider . The stand-
FIGURE 17. RESIST IVE DIVIDE R ard divider is then adjusted to measur e
CALIBR ATION . the relative voltage applied to the low
voltage winding s. The auxiliar y divider
The calibrat ion of a resistiv e does not need to be calibrat ed, but it
.divider is very similar except that the does have to have sufficie nt resoluti on
divider output is already resistiv e in to adjust the output voltage accurat ely.
nature, the ref ore the capacit ors can Some phase shift adjustm ent is likely
be directly connect ed across either or to be require d on both the standar d and
both halves of the divider as shown in the auxiliar y divider . Techniq ues sim-
Figure 17. In this case the added ca- ilar to those in Figures 16 and 17 should
pacitanc e combin es with the interna l be used; they are omitted from Figure
capacita nce across the two halves of 18 for simplic ity.
the divider to make a capacit ance di-
vider, At null this capacita nce divider
has the same ratio as the resistan ce IMPEDA NCE RATIO
divider. This techniq ue minimi zes
errors resultin g from interna l and A decade transfo rmer can be used
connect ion capacit ances in the resistiv e with a capacita nce standar d to make a
divider being tested.

TRANS FORME R RATIO

Figure 18 shows how to calibrat e

a transfo rmer voltage ratio accurat ely
using a decade standar d divider . The FIGURE 19. CAPA CIT ANGE BRIDGE
calibrat ion circuit for any step-do wn
ratio is essenti ally the same as that for
calibrat ion of another decade divider . capacita nce bridge of extreme accurac y
To calibrat e a step-up ratio without over a very wide capacita nce range as
loading effects, it is necessa ry to shown in Figure 19. Similar circuits

lelslil EB-29 9/62

 -7-

can be used for other impedanc e com- is required . The resistive decade
parisons .. divider circuit usually used as a lab-
oratory standard for high precision
work is shown in Figure 23. This is
SYNCHR O AND RESOLV ER TESTS known as the Kelvin-V arley circuit.

CIRCUIT
DIAGRAM
FIGURE 20. SYNCHR O TEST

Synchro and resolver compone nts

and systems can be checked to accur-
acies of one second or ·better. Figure FIGURE 23. KELVIN- VARLEY
ZO shows one of the synchro test circuits. CIRCUIT
In this circuit, each decade contains
CURREN T RATIO eleven resistors instead of ten. The
resistanc e to the right of each decade
is made equal to the value of two resist-
ors in the decade. When this resistanc e
to the right is connecte d across any two
adjacent resistors of the decade, the
resistanc e across the combinat ion will
FIGURE 21. equal that of one resistor of the decade.
CURREN T DIVISION
The end-to-en d resistanc e of the decade
The reversed transform er divider will be that of ten equal resistors 1
makes an accurate and useful current rather than eleven.
divider, as shown in Figure 2.1.
If it is more convenie nt to use an
taGH lt!iQiUftON interpola ting decade whose total resist ...
Gfloff ...1TOI lOW AC(Ul.A(l Dl\ltOfl
[Link] is higher than two resistors of the
preceding decade, a shunt resistor can
be connecte d across the interpola ting
decade to reduce the combined value to
FIGURE 2.2. ACCURA TE CURREN T that of the two shunted re sis tors. The
decades at the right end of Figure 23
DIVISION WITH LOAD
use this shunting techniqu e.
The circuit of Figure 22 can be
used to make accurate current division
even with high load impedan ces. EQUIVA LENT CIRCUIT

The equivalen t circuit of the Kelvin-

RESISTIV E DECADE DIVIDER S Va:rley divider is shown in Figure 24.
A Kelvin-V arley divider equivalen t cir-
Decade transform ers [Link] probably
the most accurate referenc e dividers EQUIVALENT IN~,.,,1.,--o OUT
CIRCUIT
available for the middle audio frequenc - COM
ies. However , to achieve the ultimate FIGURE 24. KELVIN- VARLEY
in their accuracy the frequenc y range DIVIDER
is somewha t limited. For lower fre ..
quencies and for de a resistive divider cuit looks exactly like a simple tapped

le!sl(J EB .. 29 9/62
 ... g ...

resistor except that it has additional re- , VOLTAGE DIVISION

sistance inserted in series with the tap.
HIGH R£SOlU1!0-h
NUll lOW ACCURACY
nc._10_•. , . . . - - - - . OIYtDI•
--..-Dl..,AAv_mi_R_..oc.....

INPUT RESISTANCE

The input resistance is measured

from the input to the commor1 terminal
with the output terminal disconnected.
The input resistance is constant. FIGURE 26. ACCURATE VOLTAGE
DIVISION WITH LOAD
OUTPUT RESIST ANGE A voltage divider can be used to
provide a calibrated, adjustable volt-
The output resistance is measured age reference from an accurate fixed
from the output terminal to the common input voltage. If a burden current must
terminal with the input shorted. It be supplied without disturbing the volt-
varies with the setting of the divider. age calibration an auxiliary divider can
Equivalent circuit a and curves of output be used as shown in Figure 26.
resistance are shown in Figure 25. Note

CURRENT DIVISION

Although we have talked about the

resistive divider as a voltage divider
it is also applicable as a current divider
as shown in Figure 27.

I '" p. DEKAVIOER

LOAD
FIGURE 25. OUTPUT RESISTANCE
RlOAO
that the large smooth parabola is the
output resistance which would be shown
by a simple tapped resistor. The scal-
.ff.
Max I.,.~ RMAJ.
IN
_
lour -SXl1,. (
R
Ill+
R111
R
LOAD
)

lops added on top of this are the effect

of the later decades. You will note that · FIGURE 27. RESISTIVE CURRENT
there are 10 sma.11 parabolas on top of DIVIDER
the big one for the second decade effect.
Ttie third decade adds ten small parab- As a current divider, the constant
olas on each of these, etc. The thick- input resistance of the Kelvin-Varley
ness of the line on each small parabola divider gives it an advantage over a
represents the extent of these additional decade transformer for small currents
variations. into non-zero impedance loads. If the
load resistance and the input current
are kept constant, the load current will
KELVIN-VARL EY APPLICATION S be accurately proportional to the di-
vider setting• no matter how high the
Some typical Kelvin-Varley applica- load resistance. The load current at
tions are shownir12E'igur es Z6 through 30. full scale must, of course, either be

(;J;;ill EB-29 9/62

 .. 9..

measured or calculated, unless the THINGS TO WATCH OUT FOR

load resistance is essentially zero.
To get the best accuracy out of any
very precise divider we must take var-
ious precaution s. The first thing to
consider is the effect of stray series
and shunt impedance s.

END RESISTANC E
FIGURE za. ACCURAT E CURRENT
DIVISION WITH LOAD Figure 31 shows how we can com-
pare two resistive voltage dividers
If the load resistance is not known, or without errors being introduced by the
is not constant, an auxiliary divider resistances of the leads and terminals
can be used in the circ'!-lit of Figure ZS at the ends of the dividers. The detec-
to make the divider indicate directly tor is connected between the two divid-
the ratio of load current to input cur-
rent.
11,
,.---... ..............
............. .,. 0£T
IMPEDANC E RATIO ~---"
~

A calibrated divider ratio can be

used for measuring impedance values
by comparison with a known standard
as shown in Figure Z9~ -----{~f}-----

"•

FIGURE 31. END RESISTAN CE

CORRECTI ON
z, s
rz=T=S ers at their respective tap settings,
FIGURE Z9. IMPEDANC E RA TIO S and SB. The detector will usually
MEASUREM ENT .ntt read zero when both dividers are
set to either zero or full scale because
of different lead and contact resistances
SYNCHRO AND RESOLVER TESTS in the two dividers. We can purposely
introduce resistances joining the ends
Synchro and resolver bridges are' of the two dividers and supply the cur-
calibrated by divider compariso ns. A rent through taps on these auxiliary
tapped resistors. We can adjust these
llUll
taps to set the voltages at the ends of

e@:ll3 S=t+ Sft111t8-30'l

the dividers equal.

Note that we have a choice. We can

adjust the potentiome ters so that the
terminals of the dividers are set at the
FIGURE 30. SYNCHRO TEST same voltage , or we can adjust the di-
vide rs so that their outputs at their
synchro bridge can be made of a re- · maximum and minimum settings are
sistive divider and a pair of load bal- exactly equal. Which adjustment we
ancing resistors. choose will depend upon the application

lelslil EB-Z9 9/6Z

 -10-

we intend to make of our divider . Nor- both divider s are of equal impedan ce
mally for greates t precisio n the divider s and they are being used at very low set-
. should be calibra ted relative to the high- tings, the ground should be placed at the
est and lowest settings . To do this we commo n (bottom ) end. It is usually
should adjust R 0 with the detecto r con- much easier to isolate a generat or (a
nected to the taps of the divider s and all battery or an output transfo rmer) from
their dials set to minimu m, and R1. ground than it is to isolate the input of
with all dials at maximu m. An inde- the sensitiv e detecto r. The generat or
pendent measur ement of end resistan ce and detecto r can often be intercha nged
can be made to find the correct ions to to minimi ze leakage problem s.
be applied when the voltage output is to
be referre d to the termina ls.
EXTER NAL GUARD DRIVE·

SHUNT IMPEDA NCE

Shunt impeda nces, both resistiv e .

and capacit ive, always exist in any
device. In divider s of the accurac y we
are conside ring we need to be some-
what careful about de leakage paths.
For ac divider s even more care should FIGURE 33. GUARD ING
be taket'L with the effects of stray capac-
itance. An equival ent circuit for the Figure 33 shows what to do if the
princip al leakage impeda nces is shown selectio n of groundi ng point is not suf-
in Figure 32. ficient to maintai n the desired accurac y.
He re a third uncalib rated divider with
sufficie ntly high resoluti on is connect ed
to drive the cases of the two divider s
to the potentia l of the taps of the divid-
ers, as indicate d by an auxiliar y de-
tector connect ion. When a null balance
is obtained between the two divider
taps there will be no voltage differen ce
FIGURE 32. LEAKA GE IMPEDA NCES between either divider tap and its case.
The effect of all six leakage impeda nces
shown above will have been remove d
from the measur ement circuit - - they
GROUN D CONNE CTION are either across zero voltage or their
current s are being supplied by the
When compar ing divider s we have guard drive divider .
four choices of where in the divider
compar ison circuit we connect ground.
By "ground 11 we mean the cases of the NOT A CALIBR ATED RESIST OR
divider s. Whiche ver. ground point we
select, the two adjacen t parts of the In using a Kelvin- Varley divider ,
divider s will be shunted by leakage re- note that it cannot be used as a simple
sistance and capacit ance. If one divider variable resisto r. The resistan ce be-
has a much lower input resistan ce than tween the output and common termina ls
the other, the ground should be connect - will not corresp ond accurat ely to the
ed to the tap of the low resistan ce di- dial setting because of the added vary-
vider, so that the shunt impeda nces ing output resistan ce of the later
will all be placed across the two por- decades . Howeve r, there are many
tions of the low resistan ce divider . If circuits , includin g bridge circuits , in

lelslil EB-Z9 9/62

 -11 ...

which the variabl e resistan ce can be

placed at a point where it does not enter
into the calibrat ion of the circuit.

WHAT POWER DOES

:
i
·•
·o!--1--+----.~
When we use divider s we always ! --::p~n~-~·!!:•::I
must be careful to operate at temper -
ature, voltage , current , power, etc •• ::· ......~"-7..-:[Link]
·•'at--11-+-+- +--t
which does not cause errors. The •Ut-+-+--f-- i--1
~[Link]........-[Link]~ _I

Kelvin- Varley circuit has a particu lar 0 '..,.,:;. ,jf._,H •II

peculia rity as shown in Figure 34.

FIGURE 35. POWER SHIFT
CURRENT TEMPERATURE RISE

million with an input of O. 2 watt but it

increas ed to eight parts per million
with 5 watts of input.

BE CAREF UL

Genera l precaut ions in any mea-

sureme nt a1: the same. Always ob ..
serve any limitati ons such as voltage
or current coeffici ent, the effects of
FIGURE 34. CURRE NT AND TEMPE R- power and ambien t temper ature, the
ATURE DISTRI BUTION effects of the humidit y, pressur e,
electrol ytic potentia l, leakage paths
This figure shows how the current and with moistur e content , magnet ic fields,
power divide in the resisto rs from one electro static fields or high voltage
decade to the next. The temper ature corona. Thermo electric potentia ls
rise is almost directly proport ional to should be watched for, measur ed and
the applied power. The uneven power correct ed for. Care should be taken
distribu tion results from the shunting in precisio n measur ements that the
of two of the 11 resisto rs in one de- initial conditio ns of calibrat ion• or the
cade by the next decade. These two conditio ns for which the unit was de-
resisto rs do not get as hot as the rest signed, are duplica ted as accurat ely
in the string. Thus, in a Kelvin- Var- as possible in the laborat ory. If not,
ley circuit, it is not enough to match the particu lar effect not duplica ted
resisto r temper ature coeffici ents.. If· should be measur ed and correct ions
the divider ,is to be run at a. fairly high made if they are needed.
power, they must have very close to a
zero temper ature coeffici ent because
they will be heated unequal ly by the DETEC TORS
input power. If we were worried sim...
ply about ambien t temper ature• then So far, we have shown many cir-
temper ature coeffici ent matchin g would cuits for divider applicat ions and' di-
be adequat e. The effects of input power vider compar isons with a simple box
on a statisti cal sample of typical high marked DET to represe1 1t a sensitiv e
quality divider s is shown in Figure 35. detecto r for the small unbalan ce sig-
Divider s are calibra ted and often used nals. The selectio n of an appropr iate
at low power. The mean :t: one standar d detecto r is of utmost importa nce for
deviatio n did not exceed three parts per high accurac y measur ement work.

f;[S[i] EB-29 9/62

 -12-

part icul arly thos e usin g mec han ical or

·Ac DET ECT ORS soli d stat e cho ppe rs. The de sign al is
swit ched on and off or reve rsed in
For ac mea sure men ts the dete ctor
pola rity by the cho ppe r to con vert it to
is near ly alwa ys tune d and exhi bits a
an ac squa re wav e sign al, whic h can
fair ly high Q. This is need ed to elim in-
then be amp lifie d by a low nois e ac
ate pick -up of stra y freq uenc ies othe r
amp lifie r. The se cho ppe r unit s have
than mea sure men t freq uenc y. This
inde ed prov ed to be exce llen t dete ctor s,
tuni ng reje cts hum and harm onic s whi ch
achi evin g dete ctio n sens itivi ties very
may be gene rate d if the elem ents in the
nea r the ther mal nois e limi t but they
mea suri ng devi ce are not quit e line ar.
too have som e prob lem s. Stra y ac
For exam ple in mea suri ng iron -cor e
sign als can caus e spur ious resp onse s
indu ctor s or capa cito rs havi ng volt age
and erro neo us answ ers so care mus t
coef ficie nts• harm onic sign als are of-
be take n in the c;[Link] uit to excl ude stra y
ten gen erat ed whic h do not bala nce at
ac pick up. Cho pper dete ctor s are be-
the sam e sett ings as the fund ame ntal .
ing desi gned to be less sens itive to
Als o, a narr ow band widt h dete ctor
stra y ac sign als than thos e whic h wer e
redu ces the ther mal nois e from the cir-
avai labl e a few yea rs ago. Cho pper
cuit to whic h the dete ctor is conn ecte d.
type dete ctor s capa ble of sign als nea r
theo retic al nois e leve l are avai labl e at
subs tant ially less cost than an equ iva-
DC DET ECT ORS lent galv ano met er.
For de sign al dete ctio n, two basi c
dete ctor type s are com mon ly used ,
DET ECT OR MAT CHI NG
galv ano met ers, and mod ulat ors (cho p-
per or othe r) whic h supp ly ac amp lifie rs. The matc hing of dete ctor sens itiv-
The pres ent stat e of the art leav es a ity and inpu t imp edan ce to the requ ire-
deci sion betw een the two dev ices som e- men ts of a brid ge or divi der circ uit
wha t arbi trar y. Both galv ano met ers invo lves mor e than mee ts the eye.
and chop per type dete ctor s are capa ble
of very exce llen t resu lts, but both re-
quir e a good man y prec auti ons in thei r EQU IVA LEN T SOU RCE S
use .
Loo king at Figu re 36 we can see
GAL VAN OME TER S that any divi der com pari son or brid ge
Gal van ome ters have been used for
de dete ctio n for man y yea rs, but they
are som ewh at inco nven ient to use
s.
larg ely beca use of [Link] hani cal prob lem
A very sens itive galv ano met er is ex- Ol>[N CIRCUIT VOL TAG(
DIVIDER
trem ely susc epti ble to vibr atio n and
{ SOURCE IMPEDANCE

othe r mec han ical [Link] trba nces . Re-

S>IOAT Ct~CU!T CURRENT

cent ly, how ever , a very exce llen t gal-

van ome ter has been imm erse d in oil to
imp rove its mec han ical stab ility and a
phot o elec tric unit used to grea tly am-
plify its mot ion. IMPEDANCE BRIDGE
FIG URE 36. OUT PUT CIR CUI T
ELE CTR ONI C DET ECT ORS circ uit can be look ed at from its dete ct-
or term inal s as an open circ uit volt age
The pop ular tren d in rece nt yea rs
s, in seri es with a sour ce imp edan ce ... - a
has been tow ard mod ulat or type dete ctor

l~Jslil EB- 29 9/62

 -13 ...

Theven in equival ent circuit; or as a it represe nts) is connect ed to any load

short circuit current in paralle l with resistan ce, the amount of power can
the same impeda nce - - its Norton be plotted versus load resistan ce to
equival ent circuit.. In this equival ent produce the curve shown in the figure.
:represe ntation the impedan ce is essen- This load curve always has exactly the
tially constan t as long as the circuit is same shape - - it can be drawn by
anywhe re near the null balance . The means of a templat e. The curve shows
voltage or current in the equival ent the well-kn own impeda nce matchin g
circuit goes to zero at true null and laws - - the maximu m power into the
rises to some level wh-en the circuit is load is obtaine d when the load resist-
unbalan ced a small an::[Link]. It is nec- ance equals the source resistan ce, and
essary to be able to see the voltage or this maximu m power will be one fourth
current produce d by the amount of the product of the open circuit voltage
unbalan ce corresp onding to the accur- and the short circuit current .
acy or resoluti on of the divider s or
bridge elemen ts involve- d. The open circuit voltage , the short
circuit current • and the source resist-
ance all interse ct at a single point.
SOURC E DESCR IBING POINT This can be called the 11 source describ -
ing point". This single point on the
It is particu larly informa tive to graph paper serves to comple tely spec-
plot any such equival ent Theven in or ify the source in a manner which later
Norton generat or on a special kind of can be used to find the amount of voltage
graph paper shown in Figure 37. Here, or current actually supplied to any load.

O
SOURCE
RESISTANCE LOAO '"\,, !---...---- -. DETEC TOR GRAPH S
lsc
SOURCE LOAD
Eot RESISTANCE RESISTANCE RESISTANCE
Now let us turn our attentio n to the
detecto r. It is also possibl e, as shown
THEVENIN NORTON INPUT

INPUT
RESISTANCE

VOLTMETER
~ AMMETER

~
i= $4"
!i.;----t~-"""'*~~~_:::;;...oc;~~--i'--~-;l'.o

~
L0.6.0 RESISTANCE :! t,--~---+--
g
FIGURE 37. SOURC E REPRE SENTA -
TION

resistan ce is shown along the bottom

of the graph paper, power vertica lly,
SOURCE RESISTANCE
and the voltage and current lines are
the diagona ls. When either type of FIGURE 38. DETEC TOR
equival ent circuit (or the real source REPRES ENTAT ION

l!alsm EB-29 9/62

 -14-

in Figu re 38, to desc ribe any spec ified for a spec ified unb alan ce. The outp ut
amo unt of defl ecti on of a dete ctor by a. desc ribi ng poin ts for all setti ngs ca:n
· poin t. The dete ctor desc ribin g poin t ther efor e be pres ente d in a grap h as
repr esen ts its inpu t resi stan ce, volt - show n in Figu re 40.
age and curr ent for this defl ecti on. We
can also draw a curv e whi ch repr esen ts
the locu s of sour ce desc ribi ng poin ts
whic h wou ld prod uce the spec ified de-
tect or indi cati on. The sam e tem plat e
can be used to draw both the sour ce and
dete ctor curv es.

Diff eren t amo unts of defl ecti on are

repr esen ted by diff eren t curv es, as
illu stra ted in Figu re 39. If we plot the
desc ribi ng poin t for a sour ce on such a
set of curv es for any dete ctor we can
imm edia tely read from the curv es the
amo unt of met er defl ecti on whi ch will
be prod uced .

FIG URE 40. BRI DGE PER FOR MA NCE

~ ~"'
'• o't~
~-..fl:.q;: Such a plot can be used to dete r-
.,,~or,,.::;
min e the perf orm ance of the brid ge in
com bina tion with any dete ctor . A plot
of the max imu m sens itivi ty curv e for a
5 10·"
0

"'0
... 10-''
'
#"'
'o'I,
sam ple dete ctor is incl uded in Figu re
40; sinc e all the brid ge curv es lie abov e
the dete ctor curv e, this dete ctor will
allo w the brid ge to achi eve full accu racy
lx!5
ove r its who le rang e. If a who le set of
"'
- 10·•)
dete ctor curv es, as in Figu re 39, is
" o<I'
ii)
§ 10·~
',..._ plot ted on tran sluc ent pap er and laid
ove r Figu re 40, the actu al defl ecti on
~+
>
s can be read from the curv es for any
~.l-}'11
10·!1&
<..>
0:
u ,4o ~"£.:,> sett ing.
z 10 ..
'q,
~
g "'• The curv es show n in Figu re 40 are
for a brid ge unb alan ce equa l to its rate d
accu racy ; a sim ilar set of curv es can be
draw n for any othe r spec ified amo unt of
unb alan ce - - for exam ple, the reso lutio n
of the brid ge.

FIG URE 39. DET ECT OR CUR VES DET ECT OR COM PAR ISO N

Such dete ctor curv es can be used

PER FOR MA NCE GRA PHS to com pare one dete ctor with ano ther
as show n in Figu re 41. Her e a rath er
At any give n setti ng of a brid ge or
typi cal coll ecti on of both ac and de de-
othe r null -typ e circ uit a sing le desc rib ... tect or max imu m sens itivi ty curv es are
ing poin t repr esen ts the outp ut sign al

l~lsU] EB· Z9 9/62

.. -15-

/ ~ 4Z. The dotted line at the bottom shows

the signal which would produce a mini-
./' mum visible deflection of one fifth dial
division if the extra series resistance

''*" ~~
were removed.

'~
":'1-°"
;-~~
~

~ '
,,,'ttc:;11
"~c....

.p
'·
,:tO
I:
iil.. B•o"'i.--~;..._~
z
...o
.p
1 t0~·~~~'"""'1~~~~~~~~ ~ ~•

.
"'-----V'---"'00;=../
~ •
SOURCE RESISTANCE

,,H-i~~"'*"'~~. . . . .""*'~~""""+ffl*'"""'*
10·•· FIGURE 4Z.. GALVANO METER WITH
CRITICAL DAMPING

Along the dotted portion of the curve,

FIGURE 41. DETECTOR COMPARI- however, the galvanome ter would be
SON so highly overdampe d that it would
'take a very long time, perhaps many
shown. It can be seen that some detec- seconds, for the meter to reach the
tors are more sensitive for low imped- indicated deflection.
[Link] sources, others for high impedance
sources. Any bridge describing point
graph can be laid over such a collection DETECTOR SENSITIVIT Y TESTS
of detector curves to see which detectors
have enough sensitivity for the bridge A test circuit which will determine
settings and accuracies of interest. the actual input signal versus source
impedance necessary to cause a dis-
cernable indication is shown in Figure
GALVANO METER 43. The ac test circuit shows an ad-

If a low-resista nce source is con-

nected to a galvanome ter it will be ex-
tremely overdampe d or sluggish. If
the source resistance is too high, the
. galvanome ter will be underdamp ed,
and the needle will swing back and
forth a long time befo~e settling down FIGURE 43. AC TEST CIRCUIT
to a final indication, To avoid these
problems the source resistance can be justable voltage source with an adjust-
kept constant by adding series or ·shunt able impedance , usually a resistance,
resistance. If this is done the detector connected in series. To obtain the
curves are modified as shown in Figure detector threshold curve, set Zs at each .

lelslil EB-29 9/62

 -16-

·particular value and find the smallest extension of this technique is applicable
value of e for which the detector will to ac divider calibration •
. show a chg_nge when the generator is
switched on and off. It is best to have We hear a lot these days about
one operator observe the detector and traceability of units to the National
another one throw the switch on and off Bureau of Standards. In case of a di-
at random, to avoid prejudice and to vider, however, there is no unit to
make a statistical sampling of what per- trace to the National Bureau of Stand-
centage of the time the observer can ards. It is not necessary to trace ratio
correctly call the operation of the switch. measurements to any outside source
For each impedance Z one point on the since they are dimensionless quantities
curve will be obtained ~the intersection and no units of measurement are in-
of Zd and e ). If the detector noise var- volved. What we must do is to compare
ies with soJrce impedance this curve one fraction of the voltage in a divider
will not have the "standard template" to the other fraction, and this is some-
shape shown in the preceding figures. thing that we can do ourselves with
adequate equipment. On the other hand,
The de test circuit shown in Figu;re many laboratories may find it conven-
44 is the same except that here we may ient to trace their measurements to a
add a time delay, a resistor and capac- divider which has been calibrated either
itor, so that when the switch is thrown by the National Bureau of Standards or
by some other laboratory having ade-
quate intercomparison facilities.

For highest precision work, how-

ever, it is often advisable to do your
own ratio calibrating. The accuracy of
ratio calibration which can be achieved
may exceed the long term stability of
FIGURE 44. DC TEST CIRCUIT any certified standard divider. With
resistive dividers, a new calibration at
the voltage will rise and fall slowly, the same working temperature is more
perhaps with a time constant of one reliable than a "certified value" found
second or longer. This will be useful even a few hours before.
if we expect to use the detector with a
source that may have transient signals
which can obscure the signal for a cor•
responding period. This time delay
also aids in finding out whether one can
detect slowly changing input signals.

DIVIDER CALIBRATION AND

TRACEABILfrY

Anyone in a well equipped labora-

~ory can calibrate a voltage divider.
He can use comparison techniques to
ietermine the "a'likeness" of nominally FIGURE 45. TRACEABILITY CONSOLE
:qual impedance elements, then connect
hem in combinations to give predictable Figure 45 shows a measurement
roltage ratios. The resistance transfer console designed for making highly
!tandard shown in Figures 5 and 6 is accurate, traceable [Link] of
ideally suited for this application. An resistance, capacitance and ratio at de

l~'.s[il EB-29 9/62

 .. i1-

and low frequencies. The unit on the A Kelvin-Varley divider can be

left is a precisionb guarded, four- accurately calibrated by measuring the
terminal resistance measuring system» linearity deviation of each decade re-
which can trace resistance measure- lative to a ten-step reference standard
ments to certified Thomas-patter n one- divider. These measurements can then
ohm resistors• The center unit is a be tabulated or plotted so that the devi-
precision capacitance bridge which ation at any dial setting can be calcul-
calibrates :relative to certified standard ated simply by adding the contributions
capacitors. The right-hand panel is for from each decade.
ratio calibration - - its standard di-
viders can either be certified by NBS This technique can be shown by
or other outside laboratories, or di- calibrating the simple two-decade di-
rectly calibrated through the use of vider shown in Figure 46. For each of
intercomparis on techniques using the
bridges and standards in the units to
the left. · IN

KELVIN- VARLEY CALIBRATION

S, OUT

The setting of a decade divider is

the nominal ratio of the output voltage
to the input v·oltage. The accuracy of COM
the divider at any setting is expressed
as the deviation of the actual ratio from
r·Iml!.-s-~
this nominal value.. This deviation can Eu'll0 10

be expressed as a fraction of either the

input voltage or the output voltage. FIGURE 46. TWO DECADES

LINEARITY DEVIATION (r) is the the ten settings of the first decade we
diffe re nee between the actual output shall make two measurements of out-
and the nominal output expressed put linearity deviation - - one with the
as a fraction of the input voltage: second decade set at zero, and one with
it set at maximum. If the output divider
were a perfectly linear interpolator
between each of these pairs of output
voltages, the linearity deviation (f) of
LINEARITY is the maximum mag- the whole divider could be specified
nitude of the linearity deviation. graphically by plotting the measured
points and joining each pair by a
OUTPUT DEVIATION (A) is the dif- straight line, as shown in Figure 4 7.
ference between the actual output
and the nominal output expressed
as a fraction of the nominal output
voltage:

SETTING s-
OUTPUT ACCURACY is the maxi-
mum magnitude of the output FIGURE 47. ACCURACY OF FIRST
deviation. DECADE

t;J;IIJ EB-29 9/62

 -18-

We can make a sepa rate calib ratio n of fer Stan dard (Fig ures 5 and 6) can serv e
the linea rity devia tion of the outpu t as an exce ption ally accu rate refer ence
stand ard divid er, since its tap ratio s
• deca de relat ive to its own full scale
rang e - - that is, relat ive to its out- can be calcu lated from resis tanc e inter -
put at [Link] and ten. In the two- deca de com paris on meas urem ents to a linea rity
divid er, the rang e of this deca de is of bette r than 2 parts in 10 milli on.
one- tenth the rang e of the whol e di-
vide r, there fore we can divid e each of A high -imp edan ce mic rovo ltme te r
these readi ngs by ten to expr ess their is used to read the diffe rence betw een
contr ibuti on to the linea rity devia tion the Kelv in-V arley outpu t and the stand -
of the whol e divid er. Thes e cont ribu- ard divid er tap. The lead- comp ensa ting
tions are adde d to those from the first pote ntiom eters show n in Figu re 49 are
deca de to obtai n the devia tion of the adju sted so that the micr ov·o ltme ter
whol e divid er at any settin g• as show n indic ates no volta ge diffe rence betw een
grap hical ly in. Figu re 48. the stand ard and Kelv in-V arley divid ers
when both are set at their maxi mum or
their zero setti ngs. At each settin g of
the deca de unde r test, two micr ovol t-
mete r read ings are taken - - one with
all the deca des to the right set at zero ,
and one with all deca des to the right
set at maxi mum . The ratio of each
readi ng to the appli ed volta ge repr e-
ACC URA CY OF TWO sents the devia tion of the Kelv in-V a'rley
FIGU RE 48.
DEC ADE S from the stand ard, expr esse d as a
fract ion of the full- scale rang e of the
deca de unde r test. Thes e read ings are
For a divid er with more deca des, how-
then corre cted for the calib ratio n of
ever , it is simp ler to plot sepa ratel y
the stand ard and divid ed by the appr o-
the contr ibuti on to linea rity devia tion
priat e powe r of ten to conv ert them to
from each deca de and then to add the
linea rity devia tion contr ibuti ons ex-
readi ngs at parti cula r setti ngs as re- pres sed as a fract ion of full- scale for
quire d. the whol e divid er.
The prac tical calib ratio n of Kelv in-
Varl ey divid ers is acco mpli shed with
the circu it of Figu re 49. The deca de
. .....
~·
,, ~:o
-010
.... .....
-OMO "
+-<0.00r.,IJ
l)~,00
'"'i)OOO,t()
[Link].~

-o~,o •(l.~60 -OOOl'l.••O

H~ir-1 •a~'
""' -~: -0.•4 -00"' -000&,0 ~000,e<'l

. ...-·· -·...
.~
"
·-°r:'
-010
+Q,~'le +O~.),ll

+1>006,0
+O~.to
... 0000,eio
-0(1()(1,0•·

_,,........,
~
".
... ~:!i~ -~:r
-OH
+o_~
+0010
+-O.Q0.11,0

"
+oOJOt,I) •OOOC.,to
~p(I

.." .... .....+~l +C):'f'Q(lrf, +G~ +o~~l,O +c)~,C() -tt.000,[Link]

_,,_
-·-
+O~t,O tt),®!),00
"'
~
.. ·~a' .. ~fl .o:~ +o.~oo,o +o.~.io

. .,.... ....
otl.<I 40 UI •0.0l! ...0.l>Hl,0 .0.000,tlO
CONNECTION FOii CAl.18RATION Of INOIVIOUAL DECADES -o..~J:(l
... ~, "'~ll +o~ +0.01!•,0
~ +O.O!O "
+[Link],O ...0.000.'10

FIGU RE 49. KEL VIN- VAR LEY

" ...
~ ....
"
·~l°
•OOt
+o.~ +0.~,0
•[Link].O
+o.~.l'O
HJOOO.l'O
-[Link],•on

CAL IBRA TION ~

. ... . ~o ·~i .. ::~
•o~~J,Q +0~,IO -0.000.,oeo

.... ....
•001 +0021,0 •0000,llO

to be calib rated is conn ected in para l- ~

&If_...,...,.
-~· ..
l!looll t ....
.,,.
,._Hff
~:- ~·~:ti •o~,.,I)

tw1 llWI "f"I fl flkf -

~ o.000,oo
·--
-o.(1(11(1,0t -a:.ooo,1:ro

(l>(j, .... _ _ , d - 119<1

lel with a ten-s tep stand ard divid er ....."1111..... , .. •••"""• o .,. fu "'• ...,ol*d ..,... . . .c1uotM ., ow1
Mthott 1• 004'11,ffUH

throu gh low- resis tanc e lead -com pen-

L!NlAAllY 0£VIATION MEASUREM ENTS Of"
ES< MODtl. RV·622 D(M.AV10€A" OE CA OE VOl TAG£
••
OIVIO(R

satin g pote ntiom eters . The figur e

show s the conn ectio n for calib ratin g FIG URE 50. LINE ARIT Y DEV IATI ON
the secon d deca de. In this circu it, the TAB LE
ESI Mode l SR 1010 Resi stanc e Tran s-

f~Jsjij EB-Z 9 9/6Z

 -19-

The contributio ns to linearity de- interpolatio n along a line segment. In

viation from the various decades can Figure 51, for example, at a setting of
be tabulated as shown in Figure 50, or O. 543, 212, the first decade graph reads
plotted as pairs of points joined by +0.89 at 0.543,212, the second decade
straight lines as in Figure 51. Note reads +0.06 at 0.043,212, the third
decade reads +0.05 at 0.003,212. The
linearity deviation contributed by the
last three decades is so small that it
has been considered negligible and has
not been plotted. Adding these readings
gives an overall linearity deviation of
+l. 00 ppm at this setting, or an actual
ratio of O. 543, 213, 00. This calculation
calibrates the whole divider as an in-
terpolator between its zero and full-
scale output values •
......".,T
••t llCA8I L.. . . . . tlWat•

If we want to calibrate the divider

relative to the voltage at its input ter-
minals, we can measure the two end
" ..........
i•
-..--.--.--.........-...-.-...,......,
•·"-"'-1--ffii"~F=-l'..........+-I
voltage drops with the microvoltm eter
and plot their effect as shown in the top
ii-·· graph in Figure 51. This shows an
d -·••J-.:1::-d:::-:~==-~=-=..:rl additional contributio n to linearity de-
'NT•• .. 0.01•0.H•OM•O.M • 0 . H - W . M t f t - - viation of -0.0Z. ppm at 0.543,ZlZ,
making the total deviation relative to
FIGURE 51. LINEARITY DEVIATION the input terminals +o. 98 ppm at this
GRAPH setting, or an actual ratio of 0. 543, Z.l Z.,
98.

that the first and last reading for each

decade has been made equal to zero by
the adjustment of the lead-comp ensat-
ing potentiome ters, so that the plot for CONCLUSI ON
the decade will represent the deviation
from perfect interpolatio n between the Voltage ratio devices provide us
highest and lowest outputs that can be with equipment for high accuracy volt-
obtained from that decade and the de-' age, current, and impedance compari-
cades to the right of it. The deviation sons. The present state of measure-
at the ends of this range is included in. ment science and commercia lly avail-
the [Link] of the next decade to the able materials have made possible both
left. Thus; an important feature of. resistive and transforme r dividers of
this method is that the calibration in- demonstrab le one ppm and better per-
cludes 'the effect of all lead resistances formance. Thus laboratorie s through-
and reproducib le contact resistance s. out the world can calibrate their equip-
ment relative to high accuracy national
In reading such calibration graphs, standards by means of accurate ratio
the settings of all dials to the right of equipment and a minimum number of
a given decade must be included in the stable transportab le standards.

lelslil EB-29 9/62

 . . zo-

MORE ABOUT VOLTAG E DIVIDER S of NBS-C, Vol~ 666, No. 1. Jan.-

March 1962.,

(l) A New Approach to Bridge Sensi- (5) Calibrati on of A Kelvin-V arley

tivity, Jack C. Riley. ISA Winter Voltage Divider, M. L. Morgan
Instrume nt-Autom ation Confe re nee and J. C. Riley. IRE Trans. on
and Exhibit, paper 23-SL-61 , 19bl. Instrume ntation, Conferen ce paper
Available as ESI Engineer ing Bul- 5. 3., Vol. I-9, No. 2, Sept. 1960.
letin No. 23. Available as ESI Engineer ing Bul-
letin No. 24.
(2) Bridge Performa nce Graphs. ESI (6) Voltage Divider Calibrati on. ESI
Design Ideas, Vol. 2, No. 1. p_~signideas, Vol. l.·No. 1.

{3) Calibrati on of a Decade Trans- (7) A Voltage Divider Standard , L. C.

former Having One ppm Linearity , Fryer. AIEE Trans. on Commun -
N. E. Morrison . AIEE Summer ications and Electron ics, paper 62-
and Pacific General Meeting and 43, May 1962.
Air Transpor tation Conferen ce,
paper 59-997, June 29, 1959. (8) The Effects of Output Loading on
Available as ES! Engineer ing Bul- Resistive Voltage Dividers . ESI
letin No. 20. Engineer ing Bulletin, No. 10-.-

(4) Voltage Ratio Measure ments With (9) Low Input Resistanc e Voltage Di-
a Transfor mer Capacita nce Bridge, viders. ES! Engineer ing Bulletin
T. L. Zapf. Journal of Research No. 15.

lelslil EB-2.9 9/62

 BRIDGES
AND
I
. '-.!
1 ACCESSORIES

no.38

A RESISTANCE BRIDGE llADE FROll [Link] DIVIDER

•Copyri9lif 1963 lty Electro Scientilic lnd111trie1, Inc.

LITHO IN U.S.A.
 A RESISTANCE BRIDGE MADE FROM A VOLTAGE DIVIDER

HOW TO MAKE A BRIDGE SIMPLIFIED BALANCE EQUATION

The high resolution and linearity {WHEN EQUATION (2) IS SATISFIED)
accuracy of a resistive voltage divider
can be used to make direct reading re-
sistance measurements .* The circuit
·shown in Figure 1 will permit the mea-
Ru: 1R1N~+ ::J (3)

surement of the unknown resistor Ru.

The balance equations for finding the A divider can be substituted for RA
value of Ru are given in Equations 1 and and Re as shown in Figure 2. The di-
3. Equation 2. is a restriction on the vider setting lo can be used for range
fixed resistances of the bridge which changing or for setting up multiplying
gives a zero divider reading for a zero factors for strain guage1:1, resistance
value of unknown resistance. Equation thermometers , temperature coefficient
3 gives the value of the unknown resist- tests, etc. The value of Rs will have to
ance when Equation Z has been satisfied be changed each time that lo is changed.
and the divider setting is adjusted to The value of Rs can be found by shorting
give a null on the detector. the unknown terminals, settings to zero,
&0 to the desired ratio, and balancing
R1N the resulting bridge with R8 •

Ru Rs
UNKNOWN
RESISTOR
' Rs
DET

RA Re

FIGURE I. MEASUREMENT CIRCUIT

BALANCE EQUATION
FIGURE 2. CIRCUIT USING TWO
(I) DIVIDERS

Ru - UNKNOWN RESISTOR TO BE MEASURED

The balance equations can be rewritten
RA, R8 , Rs - FIXED RESISTORS in terms of the divider setting So as
R.,,. - FIXED INPUT RESISTANCE Of DIVIDER shown in Equations 4 11 5, and 6 ..
s - DIVIDER SETTING (O<s<I)

BALANCE EQUATION
RESTRICTION ON FIXED RESISTORS
{FOR s:O WHElll ft:O)
(4)
R11. Rs = Ra R1111 (2)

*This circuit was suggested. by Charles B .. Newcombe of Lockheed Missile& Space Co.

[e]slil EB-38 6/63

 -2-

.
[ :

• RESTRICTION EQUATION RESTRICTION EQUATION

(8)
(S)

SIMPLIFIED BALANCE EQUATION

SIMPLIFIED BALANCE EQUATION
(WHEN EQUATION (5) IS SATISFIED)
Ru: sR1N"+
~ ~)a ~ I+ RI +R s )
tN
(9)
R1N . RP
Ru= s - (6)
. So

The circuit. of Figure 4 results in

Equations 10, 11 , and 12. This circuit,
HIGH RESISTANC E CIRCUITS by varying R0 , changes the full scale
The value of Ru at full scale divider value for either high or low resistance
setting Cs= I) must be higher than RrN as measureme nts without altering the zero
shown by Equations 3 and 6. Shunting setting.
circuits can be used to reach low re•
sistance values. Each has certain
advantages and disadvanta ges.

LOW RESISTANC E CIRCUITS Ru Rs

_ The circuit of Figure 3 results in I
Equations 7, 8, and 9. This circuit, ·
by varying Rp, changes the full scale Ro
DET
value without altering the zero [Link]-
ment.
Ra

Ru FIGURE 4 CIRCUIT FOR DEVIATION

ADJUSTMENT
Rs

BALANCE EQUATION
DET

Re

RESTRICTION EQUATION
FIGURE 3. CIRCUIT FOR INDEPEND- (11)
ENT, FULL SCALE SETTING

SIMPLIFIED BALANCE EQUATION

(12)

lelslil EB-38 6/63

 -3-

The circuit of Figure 5 results in CIRCUIT IMPROV EMENTS

Equation s 13. 14. and 15. This circuit, The usefulne ss of the circuit can
by varying R:s. changes the zero setting be extended by several modifica tions.
without altering the total resistanc e The range can be changed by changing
range. R1111 • the ratio RA/R 8 or both. Four-
terminal connectio ns can be used to
improve low resistanc e accuracy by
Ro reducing the effects of lead and contact
resistanc e. Fixed amounts of Ru(zero
resistanc e for example) can be added
Rs or subtracte d from the measurem ent
circuit by changing Rs•

HOW THE CIRCUIT WORKS

DET A simple geometri cal construct ion
can be used to show how the bridge
operates .
Re

E 0

FIGURE 5. CIRCUIT FOR INDEPEND-

ENT ZERO SETTING

BALANCE EQUATION

RESTRICTION EQUATION
Re 8 I
- =-
RA Rs Ro
-
+ Ru~ (14)

SIMPLIFIED 'BALANCE EQUATION

Ru= sR10 ~+ ::) C+~) (15)

FIGURE 6. VOLTAGE DIAGRAM

Figure 6 is a voltage diagram showing

the voltages across each of the resist-
Note that an addition to· Ru such as lead ors in the circuit. When Ru equals zero
and contact resistanc es can be compen- the voltages across R1N and Re are equal.
sated by changing Rs without changing When Ru is added in series with Ra<l and
the ratios in the simplifie d balance Rs the voltage divides in proportio n to
equation. the resistanc e. This can be shown

leJs[il EB-38 6/63

.... -4-

geome tricall y by projec ting the resist- here only as a sugges tion of som.e of
- ance sum onto the input voltage graph. the possib le uses of the divide r as a
To get a balanc e, the divide r setting bridge elemen t. With precis ion com-
needs to be moved . The diagra m shows ponent s the circui t of Figure 7 will be
that the voltage gene rated by the setting compa rable to a good Wheat stone bridge
times R,N is propor tional to the voltage withou t adjust ment. If zero and full
genera ted by the unknow n resista nce. scale setting contro ls are added a high..;
ly accura te compa rison bridge will
result. The four-te rminal connec tions
shown will reduce lead and contac t re-
A BRIDG E FOR MEASU RING FROM 10 sistanc e effects when the divide r is at
MICRO HMS TO l 00 MEGO HMS low setting s. By connec ting termin al
The divide r bridge circuit can be 1 and l togeth er and 3 and 4 togeth er,
used to make a practic al measu ring a two-te rminal bridge can be made
system for a wide range of resista nce which will have its full scale accura cy
values . A practic al circuit is given unaffe cted by change s in zero resista nce ..

Ru
4
50k
N
_.....,.._......i DET 1----'~ ~
>
fl:
40k 0
RANGE
-------------i I
----- --r-- 1
[Link] 1 1
RuMAX

I I I o 1000.
I l.0050k I I
9k I I o Ik
I
I I I
10.52611
~!Ok
IOOk 200k
goon «>IOOk
[Link]
o I MEG
1.0IOlk
0 o IOMEG
!00£1 [Link] o IOOMEG

FIGURE 7. BRIDGE CIRCUIT

![Link]] EB-38 6/63

 -5-

BRIDGE OPERATING INSTRUCTIONS ~ mlJ ~

Connect the test leads to the unknown

resistor to be measure d ..

Balance the bridge with the divider

dials.

Multiply the divider setting by the For higher accuracy measure the zero
value of Ru1ux for the range used to resistanc e of the leads and subtract
find the measure d resistanc e value from the measure d value.
of Ru•

CALIBRA TION FOR PRECISIO N MEASUR EMENT

Short the test leads together.

Set the divider to zero.

Adjust R5 for a detector null. This calibrate s the zero divider setting.

Connect a certified resistor to the Any certified resistor within the bridge
unknown terminal s. range having a value within a factor-of -
ten of the resistor to be measure d.

Set the range and divider dials to the

certified value of the certified resistor.

Adjust RP for a null. This calibrate s full scale without dis-

turbing the zero adjustme nt.

On the high ranges a calibratio n adjust-

ment RP can be added by changing the
ratio RA/Re•

r;;f$1Il EB-38 6/63

•

WARRANTY OF QUALITY

Electro Scientific Industries, Inc., warrants its products to be free from defects in material and work-
manship. Rigorous quality control permits the following standard warranties:
1. Two years for components and instruments utilizing passive circuitry. One year
on repairs of out-of-warranty items.
2. One year on components and instruments utilizing active circuitry as identified
in the price list. Six months on repair of out-of-warranty items.

During the in-warranty periods, we will service or, at our option, replace any device that fails in
normal use to meet its published specifications. Batteries, tubes and relays that have given normal
service are excepted. Special systems will have warranty periods as listed in their quotation.

DISCLAIMER OF IMPLIED WARRANTIES

THE FOREGOING WARRANTY OF ESI® IS IN LIEU OF All OTHER WARRANTIES, EXPRESS

OR IMPLIED. ESI SPECIFICALLY DISCLAIMS ANY IMPLIED WARRANTIES OR MERCHANT-
ABILITY OR FITNESS FOR A PARTICULAR PURPOSE. In no event will ES! be liable for special
or consequential damages. Purchaser's sole and exclusive remedy in the event any item fails to comply
with the foregoing ex press warranty of ESI shall be to return the item to ESI, shipping charges prepaid,
and at the option of ESI obtain a replacement item or a refund of the purchase price.

~ ...... Clilllllliii...,,P""'
•""""!..**---
........ ....;
Electro Scientific Industries. Ire
13900 N.W. Science Park Drive• Portland, Oregon 97229• Telephone: (503) 641-4141•Telex:15-1246
 VOLTAGE
DIVIDERS

®~~~
Scien -tific lndusf:rie•-----------~
no.24
7524 S. W. MACADAM AVE. •PORTLAND 19, OREGON AUGUST 1960

CALIBRA TION OF A KELVIN VARLEY

VOLTAG E DIVIDER

by
:M:erle L. :M:organ
and
Jack C. Riley

Presente d as
Conferen ce Paper 5. 3
on June 24, 1960 at the

1960 CONFERENCE ON STANDARDS AND

ELECTRONIC MEASUREMENTS

© Copyright 196~ by Electro Scientific Industries; Inc.

LITHO IN U.S. A.
 CALIBRATION OF A KELVIN-VARLEY Sl'.Al'IDARD DIVIDER At first it might eeem desirable to have a calibra-
By tion made at each possible aettillg. The divider shown in
Merle L. Morgan and Jack c. Riley Figure 2, however, has one million one hundred thousand
Of poesible settings. If meaeuremente could be mde at the

• Electro Scientific Induatriee rate of one a eecond it would take about two months to com-
plete the calibrations and the results w~mld fill a volume
of booke comparable in size t.o an encyclopedia. There has
to be another way. I t would be preferable if each decade
could be calibrated individually and the resulting devia-
tions combined to give the deviation at any desired setting.
The lin"l!rity deviation of a mnltiple decade Kelvin- We are going to tell one way that this can be accomplished.
Varley voltage divider C!lll be calibrated by comparing it
decade by deoade with a ten-step standard divider. The CALIBRATING A TWO-STEP DIVIDER
etandard dirtder can be calibrated by precieel;y measuring
its resietors and calculating its linearity. The basie for We will start with a two decade divider. We are going
both at these teobniquee is derived mathematically. The to measure the linearity of the interpolating divider firet;
procedure for measuring HCh decade and the method of·· com- next the linearity of the first divider, and then we will
bining the contributions of.' all the decades to find the ehow how to combine the results to give the linearity at
linearity deviation of a given setting will be presented. any desired setting. Linearity measurements are made in
The expected [Link] of the meaeuremente will be analyzed. terms of' linearity deviation r. Linearity deviation is the
Contributions of resistor [Link], resistor and contact difference between the ratio of output voltage to input
stability, lead reeietance, temperature variations,· analyt- voltage and the setting. The setting is the nominal ratio
ical eimplificatione, and pover dieeipation will be die- of' out;mt voltage to i..'111Ut voltage.
cueeed.

INI'RODJCTION

A multiple decade Kelvin-Varley voltage divider ie

often used ae a standard for calibrating other voltage di-
Videre. It is therefore necessary that we be able to cali-
brate this standard. fr1E1N

GEN
.
w
0
;;
£ou1

0.5
.
0
0 ;;
0
w •
.."' I
2

"
~
0
>
~~

- - Figure 3 Measuring the Interpolatinp; Divider

Figure 1 Kelvin-Varley Circuit
Figure 3 shows the circuit connections for measuring
Figure l shows the circuit diagrem for a Kelvin-Varley the linearity deviation of a ten-step unknown interpolating
voltage divider. The first decade consists of eleven resis- divider. We assume that a perfect ten-step divider is
tors. At any given setting two of these resistors are available for mking the measurements. The potentiometers
shunted by an interpolating d1v1der with an input [Link] at top and bottom are provided for adjusting the zero end
equal to the value of the two resistors that it shunts. full eoale settings of· both dividers to agree. The result·
Thus, the input to the first decade looks like ten equal is shown in the voltage diagram at the right where the zero
resistore in series so that the interpolating divider will and full scale eettings of both dividers are equal. The
work over a range of one tenth of the input. The second, generator voltage is adjusted eo that the high impedance
third, etc. decades can also be Kelvin-Varley circuits. detector will read linearity deviation [Link] in conven-
The [Link] decade, however, consiets of ten resietanoes in ient ecale divisions. The detector is moved from point to
aeries with eleven tape brought out. It can be set to point and the linearity deviations are recorded.
values of one, two, three, etc. up to ten. It is only
through the uee of this tenth setting that the divider can
be set to :f'ull scale. The full scale dial reading will be
99• • • 9l'EN. Thie feature also means that there w:tl 1- he two r ACTUAL OUTPUT VOLTAGE _SETTING
wa:ye of reaching any setting ending in zero. (Except zero and INPUT VOLTAGE
full scale)

SETTING s-

Figure 4 Linearity Dev1ation of the Interpo~tinc Divider

In Figure 4 we have a graph showing typical results.

Notice here that the "zero" elld "one" settinge show no
linearity deViation. The end correcting potentiom&tere
Figure 2 ESI Model RV 622 DekaVider® Decade Voltage Divider were adjusted to assure this.

EB-24 8/60
 Figure 7 shove a two decade setting S where:

•..
~
s = Nominal EOut = Sa +
Em
Sc
10
..,"' "'a
> The 11nearit7 deviation r iB derived 1n relation to the

•
~
0
~ setting. The voltage d1sgram shove hov any desired output
!' u
voltage lllll1 be obtained b7 moving the two decades.
GEN ~
fbElN Sb• OTEN

raE1N

r-~-s
•••
i lb-s.·~

t Eb-Ea· (i& • rb-r•)• ..

Figure 5 Measuring the Interpolated Divider ii------------------~

In Figure 5 the first decade ie being calibrated. Here
twent7 -euremente are neceeeal'1 because each setting ex- [:ks<\
cept zero and cme can be reached 1n two different va7s. The r -x•r,+(s,+r,)(;5•r,-r,)-,si;-~
-surements are made by first setting the dividers to zero
and adjusting the bottcmi potenticmeter for agreement with .__,__.,
r-r.(1+sc) +rbsc•R;rc•li:(rb-r•)
the perfect ten-step divider then moving the settings to LINEAR INTERPOLATION ~ SMALL
full scale and adjusting the top potenticmeter. The voltage
diagram at the right shows how the perfect ten-step divider
is used for f1nd1ng the l1nearit7 deviations at both ends of Figure 8 Combi;ling Linearity Deviation [Link]
the 1nterpola~ing divider for each setting of the first di-
vider. Part of this voltage diagram is enlarged in Figure 8
to show how the l1nearit7 deviation at aD1 setting can be
expreseed 1n terms of the measurements previouel;y made on
each decade. Here, we restate the voltages in terms of
settings and -B1lred linearity dev1stione. Next we find
the ratio of output voitage to input voltase and subtract
f'rcllll it the settings. The setting S is the ncm1nal ratio
of output voltage to input voltage. 'l'his difference is the
linearity deviation for the divider. The linearit7 devia-
tion is found to be a linear interpolation between the de-
viaticme found at the ends of the interpolating divider--
plus l/10 of the deviation of the interpolating divider.
SETTING s- ~e is alsb a V91'1 [Link] tei'lll which is the product of two
11nearit7 deviations. Both of these deviations will be in
the order of ten parts per million typ2,call;y so that their
product will be onl;y a fw parts 1n loJ.-0 and can therefore
Figure 6 Corrtribution
of the Interpolated Divider to be ignored.
Lineerit7 Deviation

Figure 6 is a plot of the results of these measure-

ments. Here a straight line is drawn between the readings
obtained with the interpolating divider at zero and at full
scale. 'l'his is dcllle because a linear divider ca:mected be-
tween these two voltages and set at a value BOlll8Vhere be-
tween would have the l1nearit7 deviation shown b7 this , _ _.,.v,.,I~f AT SETTING S
straight line at the desired eett:lng. We have now c--
pleted the l1nearit7 meaeuramnte of each decade separately s
and we need to know how these can be combined to find the
linearity de'viation at a given estt:lng.

Figure 9 Linearity Deviation of a Two Decade Divider

Figure 9 shows graph1call;y how this is done. Two steps

of the first dirtder are ahovn. The linearity deviations of
Sc OUT the end mea81lrements are connected by a straight line. One
tenth of the linearity deviation measurements of the inter-
polating divider are then added to the values along the
straight line. These points then show the linearity devia-
tion of the divider for each possible setting. There are
llO possible settings of this two decade divider. We can
find the linearity deviation for aJ11 cme of these settings
by having made onl;y 27 measurements. If we measure a three
r • !l!Y.t.
EIN
- s,, - !s.
IO decade divider, the procedure is uact:q the same except
that both of the ·iast two stages are the interpolating di-
vider and tbe7 are aet to 00 and to 9J.'D at each setting of
the first decade. The same procedure can be continued for
as lllllD1 decades as are desired.
Figure 7 Tvo Decade Voltage Divider

l!!l!!lil EB- 24 8/ 60
CALIBRATING A "PERFECT" IlIVIDER

Row do we phyeice.l)Jr make these measurements? First

we need the "perfect" ten-step divider. We calibrate a ten-
etep divider by resistance comparison. Thie gets us within
:O.l ppm of knowing its linearity. Next we use this divid-
er to set an adjustable ten-step div1der within :k 0.2 ppm of
"perf'ect" --we settle for this.

Figure 13 Four-Terminal Connections

Figure 13 ehowe the way to make these four-terminal

connections eo the resistance between center terminals can
be measured accurate)Jr. These center terminals are the o:oee
used for voltage divider tape.

Figure 10 ES! Model SR 1010 Decade [Link] Standard

EO<JT
r·~-s
Figure 10 ehowe the ES! Model SR-1010 DECADE RESISTANCE
STANDARD which ie calibrated e.e a voltage divider.

ElN - actuo! !rtput vo!lo!jje

Eou; - oetu.:11 output ... otta!jje
R 11 - resistant• of !he l'!lh 1esislor in ohm~

%
1 A., - rasislonc• from COM 10 our in [Link]

~Rn - lolal input resistance in ohm$

- selling •n l)tr unil

Figure 14 Linearity Deviation as a Function of Reeietance

The linearity dev1ation equation is rewritten in terms

of resistance values as shown in Figure 14.

Figure 11 Resistor Configuration

Thie unit consists of twelve resistors perm!Ulently

cOIJllected in series as shown in Figure 11.

RIO= 1000 OHMS

CIRCUIT FOR REFERENCE VOLTAGE DIVIDER.

Figure 15 ES! Model 242 Resistance Measuring System

Figure 12 ES! Model SR 1010 Decade Reeietance Standard
Circuit Diagram Absolute resistance meaSUT.,mente of the accuracy needed
ere considered imprectical--but resietance comparieone to
Extra terminals are provided ae shown in Figure 12 eo better than one ppm are easy [Link] the ES! Model 242 Resist-
that four-teminal measurements of the individual resistors ance bridge ehown in Figure 15.
can be made.

EB-24 8.
 ,,
DIFFERENCE FROM
RNEJ~~TJ>:
§
MEASURED DEVIATION
PPM THE AVERAGE PPM DEVI
!OS 6, A~ -AAv ;\;
I 1.4 +0.'56 +0.04
2 +o.e -0.24 +0.01
Rn - resistance of the nth resistor in ohms 3 +$.0 +,,.96 +0.41

Rs - resisfonce of the standard resistor in ohms

•• +2.2
+2.4
+!.16
+1 !&
+0.$2
+0.66
6 -0.4 ...;.1.4'4 +0.52
1 +0.4' -O.t4 +0.4~
l:.n - per unit deviation of the unknown from the stondard
•• -t.4
+1.4
-3 44
+o $6
+0.11
+n 14
10 -0.4 -1,44 0

TOTAL. iA,.-+10,4 0
Figure 16 Dei'izl1D8 Resistance Deviation
LINEARITY CALCULATED FROM RESISTANCE MEASUREMENTS
All of the divider resistors are [Link] to a single OF AN ES! MODEL SR-1010 DECADE R!SISTANCE STANDARD
standard resistor. The resulting values of the deviation
from the standard (defined 1n Figure 16) are recorded. Figure 19 Semple Caloulatione of Linearity Deviation

Figure 19 is a. table of :measurements and [Link]

which were mde for calibrating a divider.

~ 1 r.1 (1•A,,) ~~A,,-<~1 6,,

r - -,,---•·
)Y. ,~,("A•) 10 • f., A,
I ··(""·)
R~ 1:(1+0,,} D

"'' : ~-~--OOUT
s
·.'!11•""

Figure 17 Linearity Deviation 1n Terms of Resistance

Deviation
Figure 20 [Link] Decade Voltage Divider
In Figure 17 the equation for the linearity deviation
is converted into en eqllation in terms of the resistance de- In Figure 20 we eee the adJustable voltage divider
viations. Note that the resistance of the [Link] resistor which can be set to within 2/10 of e. part per :million by
conveniently cancels out of the equation for linearity devi- using our previously calibrated divider [Link] e. [Link]
ation. The value of 411.v is so much smaller than one that it meter for lllB8suring [Link] deviation.
can be removed from the denominator to give the fornrale
actually used for calculation.

STEP 3

I 10s
l: (t:.n-l:.•v)
r .. -10 n:I .__...
STEP I

STEP 2

Step .I. Calculate the overage deviation.

I 10
l:.Av " jQ .~, l:.n
Step 2. Subtract the average from each deviation to gel the
difference from the overage.
Step 3. Sum these differences from the average and divide
by 10 to gel the linearity deviations al each top.

Figure 18 Calculation of Linearity Deviation from

Resistance Deviation Measurements Figure 21 [Link] a Voltage Divider Decade

Figure 21 is a production set-up for meaeuring the de-

viation of one decade of e. six decade divider. In the fore-
Figure 18 shows the process for actually calculating ground 1e the divider being tested. In the background we
the linearity deviation at each step on our ten-step divid- see the adjustable divider which ie assumed to be perfeot.
er. First the resistance deviation values are SU!llllled to The panel directly behind contains the two end correcting
[Link] the average. Then the average is subtracted from the potentiometers for setting the ends equal, and a voltage ad-
individual readings and the linearity deviation is found by justing control for setting the sensitivity so that the
taking one tenth of the sum of these deviations from the
linearity deviation can be read directly from a meter.
average.

EB-2.4 8/60
 Figure 24 shove a graphical representation of' the

.. . ..... ......... _
-
........
DECADE linearity deviation of' the f'irat three decades of' a a1x de-

,:.
+IU

x -
+u
:-
-
.....'lo
~
-0.10
~:o "':"" .......
~
cade divider. A sample setting 1e given to show how to f'ind
and combine the contributions of each decade. Figure 24
also shows the end linearity deviation due to lead and con-

- ..... -- -..- -
4f.O
ll' ::!: % tact resistance at the input. These end linearit.y deviation
"':' ~ """=''°
_.. values are found by first setting the divider to zero and
!:.
.....
4;\"
-0.10 % ... meaeuring the voltage between the common and output tape •
~ ~ ~ +DDJ:'-"' Then the divider ii! set to f'ull scale and a meaeuremen1; ill'
A ~

. .,,. .... ..... -·.. -

-0.U

...........
... ..... ....... -
:. +:.1 llllde of the voltage difference between the input and
•'1;1'9 ~ ·~·.o +G.<\':''°
+I.I output tape. Thie end linearity contribution is also
.
.. ....... - ..... .. -· -
:. , +<ti' ::!: ..[Link]..•.o .._
~ ... [Link] interpolated to f'ind ita contribution at [Link] sett-
ing. External circuit lead and [Link] [Link] t'lrope must
:. -o.p

x - ....... ... ..
~~ +<ti' ~
........
~

..._,
also be added f'or accurate measurement. End correcting yo-
"""lll''° tentianetera can be ueed With the divider to el1m1nate all
r:.' -::"°

..... -... .
~ ~ ..- lead and contact contributions f'rom the 11188eurementa.
,......,.,_
..
:. ....
.£ +ii• +Of.I" ~ ~,IO
Now we :have accomplished. what we aet out to do. We have

··-= -
-0.010 +[Link].0

-:.4 +o.o.J,1,0 ...._, f'ound a technique f'or calibrating a voltage divider. We have
: ·~
~'g.
+of:" ""o.o:""
~·
~ ~
also f'ound how to use the calibration to correct an;y setting
···-=:-:::::::=.:-o.':.':-.::==
LINEARITY OEVlATION MEASUREMENTS OF AN
=::: ~.= of the divider. Ae usual there are et1ll a few waede in the
garden to be pulled, We bad to drop a f'ew amll terms to
ESI MODEL RV-622 OEKAVIDEA• DECADE VOLTAGE
DIVIDER
simplify the calculations that we have ueed but we were
careful to ese that they are well beyond the measurement ac-
Figure 22 Measurements on a Tn>ical Six Decade Voltage curacy with which we are dealing. There are a f'ew other
Divider problems which are not quite ao easily diamiseed, however.
The f'irst of theee is reeil!tor stability with time. Thil'
Figure 22 shows the 101 [Link] deviation measure- can only be shown b;r repeated meaeurl!llllenta over a prolonged
ments on a specific divider. These measuremente have al- period, Typically we bavs f'ound stabilities of a f'ew ppm per
ready been J11Ultiplied by the appropriate powers of ten so year. Unite [Link] become more stable with age. The next
that they represent the contribution to the linearity devia- wor'r7 is temperature stability. By careful wire selection,
tion at any given setting. The valuee ebovn are 1n ppm of l()(ll/, inspection of' the temperature coef'ficient of' individual
input [Link]. reeiatora and such precautions ae winding all of the reeie-
tora 1n each of' the f'irst f'ew decades f'rom the aame spool of

' •• .i•-•11•-·•
t•-•I 11·'. J
wire, the effects of' ambient temperature can be minimized.
Linearity temperature coef'f'icienta of' lese than two ppm per
degree centigrade are typical.
•'§~---.,
c ~H l
: ! i

-....... ·-- -
... ...,.,,. ....... .....
FIRST
..,... . ...
-....... .--·...... ...
-....... -.. ....- -·-
......... ,.,,_
""""" ,IFTM
,
'"''"
EFFECT OF FOLL POWER

....-
DECAO<

.._..... ._.

---.......... ·- -- - - --
""""""
.-.. [Link] .........
.._.
_._,.,, +o.9ti,toH
CURRENT TEMPERATURE RISE

CALCULATING THE LINEARITY DEVIATION AT A SAMPLE SETTING Of 0.543,212

Figare 23 Interpolation Calculation From a Set of Tn>ical

L1nearity Deviation Measurl!llllente

Figure 23 shows the interpolation equation and the

calculations neceasar,y f'or f'[Link] the linearity deviation
at a aampl.e setting f'l'Olll the -a~a on the individual
daoadea,

Figure 25 Temperature Distribution in a Irelvin-Varley

Voltage Divider

-~
.. o.• Max1mma power, however, gives ue a different tempera-
AllT [Link] LllillMITT KVIATIOlll
ture probll!llll. Here, the :ll:l41vidual resistors are not heat-
ed equally. Figure 25 shows the current distribution
throl!gh the f'1rat two decades of' a Irelv1n-Var1'y voltage di-
vider. The reeiatance values are also shown. The chart on
the right shows the temperature distribution among the re-
fl u...-...---...---...---...---...---...---...---...---.,.-,
eiatora. B7 taking the temperature rise of' the 1JJdiT1dual
reaiatora of' the f'1rat decade aa At we f'1nd tbat the two re-
r ·l--4--i--ffii"l'""""'"1'"'-1"""""-1-1 aiatora which are bridged by the interpolating divider only
11~-· reach a temperature of' At/4. The 1Dd1.v1dual resistors of
!I ~...!-:d.:-r.1.:-.~..:-:0bt::-:::!:-::!~ the second decade reach a value of' At/20 except f'or the two
'SOTAl.I -[Link]+G.n+..-+O.• •0.N ... UllllMHTYDl¥Ulnlll that are bridged which ~ reach a temperature d1f'f'erence
of At/&J. The temperature 1'[Link] an the third and later decks
Figare 24 Graphical Representation of Linearity is so Bllllll it can be ignored. But on the f'iret decade the
Dietribution Contributions dif'f'erance in temperature rise between the bridged rea1etore
and the rest of' the [Link] is significant. Therefore,
the resistor temperature coe:f'f'icienta must be low 1n addi-
tion to being matched.

EB-24 8/60
 Figure 28 illustrates a different problem. As the se-
cond decade setting 1e changed its interpolating divider is
connected across slightly different resistance values. As a
result the resistance presented to the first decade will
change slightly. This will result in changing the voltage
available at the tap points on the first decade. The effect
on the first decade is the eame ae though the resistance of
the third decade had been changed by an amount equal to the
variations in the resistances of the secon:l. decade. In
Figure 28 we approached the problem from this view point.
A variation of Am in the resistance of the third decade will
result 1n half this variation in the resistance seen by the
second decade. It will imke a change of Am/20 in the re-
sistance of the second decade. This will make a change of
Am/4o in the resistance seen by the first decade and will
change the total resistance of the first decade by Am/400.

Figure 26 Voltage Divider Resistance Decades

Figure 26 ehows that the reeistore in the first decade
have been md.e much [Link] than thoee in the following de-
cades to reduce the effects of temperature riee with applied R(~-s,)
power. The temperature rise on these res1etors for e five
watt divider input ie about 10° c. but remember that two of
these resistors have a temperature rise of only about ~o C.
There are lll!UlY variables involved in the effects of power on
linearity. We take all of the precautions which we can and
then see etatistica~ what happens.

r()"" 4.6~ s( + tLl,..,\ :a";,(<-0)\(~\

!+-)
\' 406) 400 /

Figure 29 Effect of One Resistor in a Ten-Step Voltage

Divider

-i•f--ll"tC!FttofOlUl•lfl;t-
Figure 29 shows the calculation for the effects of
0 Cl ll• CJ
o<YIOl:"llUTI ...
OB ,(;
this resistance on the accuracy of the settings of the
Figure 27 What Happens to Linearity Deviation When Full first decade, In the expressions for the linearity devia-
Power ie Applied tions the term (l + Am/400) can be set equal to one without
an.T appreciable effect ,in th~ accuracy of the result.
Since linearity deviations of the first decade are the
onl,y ones 1118ter1a~ affected by power changes we can make
meaeuremente on thie first decade to fiDd. wliat actually hap-
pens. Here we have average linearity deviations of ten di-
viders for each setting, and the RMS differences (one sample
deviation) from this average. In Figure 27 these values
have been plotted for the negligible temperature change
which resulte from 2/10 watt input and for the results when
five watts are supplied to the input terminals.

EFFECI' OF [Link]:NG LATER DECADES

CHANGES TOTAL
' ,,.
1C>iANGES TOTAL
Am

.r/RST
400
SECDl,jC l
/
"20
'!'HIRtl
OECAOE
R,
OECAOE
., DECAOE

SETTING s-

Figure 30 How the Linearity Changes When the Following

Decade ie Switched

In Figure 30 we have a graphical representation of the

amount of linearity deviation experienced by the first de-
cade because of a Am variation in the second decade resist-
ors. We see that the linearity deviation in the first de-
cade is always lees than l/400 of the resistance deviation
Figure 28 Effect of Resistance Variations Dile to of the second decade. Thie would be l/l6o,ooo of the re-
Switching Later Decades sistance variation of the third decade and so on.

EB-24 8/60
EFFECT OF CON.l'ACT RESISTANCE VARIATION

z
o
EQUIVALENT
llJ
I
CONTACT

0.05,....-----~----~~----
2
RESISTANCE 1N M1LLIOHMS
I
3

~
~l 01------+. - - - # - - - 1 - - - - - - - <

r
>

0 -----,J,-''- ----.L_ _ _ __,_

[Link]!:-
2
EQUIVALENT CONTACT RESISTANCE IN MILLIOHMS

Figure 31 Contribution of Contact Resietance Variations To

Linearity Deviation For a 10 Kilohm Divider
Another possible source of trouble is contact resist-
ance variation. To minimize this problem all switch con-
tacts of the Model RV 622 are doubled. Measurements of
large nu:mbers of contacts and a statistical analysis of the
results have revealed the resulte shown in Figure 31. A
linearity deviation of lees than ll:0.025 ppm can be expected
from contact resistance variations on a 10 kilohm ESI [Link]
RV 622 Six Decade Voltage divider.

Now the voltage divider is calibrated and we know what

precautions are necessary to use it for part-per-mill ion
measurements.

EB-24 8/60
 Kelvin-Varley
Voltage Divider
esi RV722 Decade Voltage Divider
This standards grade Kelvin-Varley voltage for use in bridge circuits, providing two arms of a
divider is highly accurate, stable, and linear bridge with a very well known ratio. Applications
instrument for use in many applications requiring include linearity determination, the measurement
accurately known voltage or current ratios. In of voltage and resistance, and the calibration of
particular, the RV722 is especially appropriate voltage, current, and resistance.

Equivalent circuit: A Kelvin-Varley voltage divider may be thought of to zero at the full scale and zero settings. This series resistance has
as being equivalent to a digital potentiometer. However, it has an ad- no effect in balanced bridge type applications, where these dividers
ditional, but variable, resistance in series with the wiper arm, which goes are often used.

1.1 HIGH
1.0

LOW LOW

SPECIFICATIONS
TERMINAL LINEARITY (Relative to Input Terminals) Same as
RATIO RANGE: 0 to 1.0 of input.
absolute linearity except for end voltage drops
not exceeding 0.05 ppm for 100 kΩ divider
RESOLUTION: 0.1 ppm with 7 decades.

ABSOLUTE LINEARITY:
COMPENSATED (Relative to Output Common Terminal) Same
[VOUT/VIN]-S WHERE S ±0.5 ppm at mid-scale, improving at zero
TERMINAL LINEARITY as terminal linearity except that voltage drop
IS THE DIAL SETTING. and end settings
at zero setting is compensated to ± 0.002 ppm
for 100 kΩ divider
SHORT-TERM LINEARITY 0.2 ppm/30 days under standard
STABILITY laboratory conditions and VIN <100 V. SWITCH CONTACT & Less than ± 0.004 ppm for 100 kΩ divider
WIRING RESISTANCE
LONG-TERM LINEARITY ±1.0 ppm of input/year at mid-scale. improving at VARIATIONS
STABILITY: zero and end settings
CALIBRATION DATA ISO-17025 Accredited Certified test report
TEMPERATURE COEFFICIENT supplied with the unit gives calibration data
OF LINEARITY: <±0.2 ppm/°C. accurate to ± 0.2 ppm linearity. (at the time of
final inspection). Calibration presented in
form suitable for interpolation calibration of
POWER COEFFICIENT OF ±1 ppm/watt improving at zero and end settings. correction at any dial setting.
LINEARITY:
TERMINALS: High quality low thermal emf gold plated
MAXIMUM INPUT POWER: 2.5 watts; 5 watts intermittent.
tellurium copper binding posts.
MAXIMUM INPUT VOLTAGE: 700 V rms for 100 kΩ
DIMENSIONS: 48.3 cm W x 13.3 cm H x 21.3 cm D
BREAKDOWN VOLTAGE: 1000 V peak to case (19.0" x 5.25" x 8.4").

INPUT RESISTANCE: 100 kΩ ±50 ppm. WEIGHT: 5.7 kg (12.5 lb).

MAXIMUM OUTPUT 66 kΩ, determined by shorting across the input and

RESISTANCE: measuring the resistance across the output terminals