UCRL-12409

I~CNNIC^L
INFOIIMAiiON
I~iSION

University of California

Ernest O. Lawrence
Radiation Laboratory

THERMAL EFFECTS IN DIMENSIONAL METROLOGY

Livermore,California

UCRL- 12409

UNIVERSITY OF CALIFORNIA
Lawrence Radiation
Laboratory
Livermore~
AEC Contract

THERMAL EFFECTS

,.

Californla

No. W-7405-eng-48

IN

DIMENSIONAL

Win.
Jarr~es
Eldon

Brewer
B. Bryan
R. McClure

J.:.W.
February

Pdarson
22, 1965

METROLOGY

DISC/LAI
MRR
Thi~ documentwas prepmeda~ an account of work sponsored by m~agency of the United States
Government.Neither the United States GnvernLuont
nor the University uf California aor m~yof their
employees,makesany warranty,expressor implied,pr assumeaany legal liability or responsibility for
the accuracy,completeness,or usefulness of any information,apparatus, product,or process disclosed,
or represents that its use wouk[not infringe privately ownedrights. Referenceherein to any specific
commercialproduct, process, or service by trade name~trademark~mamtfitcturet9or ofl~erwise, does
not necessarily ~mstitut~or implyits endorsement,recommendation,
or favoring by the United States
Government
or the University 0f California. The views and opinions of authors expressed herein do
not necessarily stale or t~flect those of the UnitedStates Government
or the Universityof Calihmda,
andshall not be used for advertising or product endorsementpurposes.
This report has been rep~ducexi
directly fromtile best availablecopy.
Available to DOEand DOEcontrac|ors from the
Office of Scientific and TechnicalInhmnation
P,O. Box62, OakRidge, TN3783~
Prices aw~ilableflXHtx(615) 576-8401~
PTS626-840I
Availableto the public fromthe
National Tedmical
InformationService
U.S. Departmentof Commerce
5285Port RoyalRd.,
Springfield, VA22161

Workperformedtinder Lheatmpicesof tile U,S. Depilrimei~tof llriergy by LawrenceLiveriuoreNational

Laboratory under Contract W-Td05-ENG-41t.

1TI.IE[~MAL

EFFECTS.IN

DIMENSIONAL METROLOGY

Win. Brewer~ ~" James B, Br~an, 3 Eldon R. M.cCture, 4 5and J. W. Pear,on

ABSTRACT
A Lawrence Radiation
dimensional

metrology shows that,

thermal effect
correctiv~

is the largest

action

an awareness

solutions.
error

their

limits

if

quantitative,

accuracies.

to thermal

vibration,

frequencies

to avoid resonant

the thermal

environment

effect

tolerance
large

work,

and machine

This paper is .an effort

environment

is..as

and theoretica].ly,

important as the absolute

the sensitivity

A relatively

and automatically

some

method of thermal

of machine

can be minimized by selecting

conditions.

to

problem and to suggest

s.emiexperimental

and that

in

enough to make

systems

It is shown, experimentally

variation,

structures

of close

source of ~rror,

.of temperature variation

of the temperature

of thermal

modern measurement

potential

is developed.

that the fr.equegc.y

in the field

of the thermal

A simple,

evaluation

investigation

single

necessary

tools are to attain

create

Laboratory

effect

.si.mple

error

environmental
device to monitor

compensation

is

proposed.
1Work performed

under the auspices

2Graduate Student of Mechanical

East Lansing, Michigan.
3Chief Metrologist,
Livermore, California.
4Assistant Professor
Corvallis,
Oregon.

of the U. S. Atomic Energy Commission.

Engineering,

Lawrence Radiation
of Mecltanical

Michigan State

Laboratory,

Engineering,

5Mechanical Engineer, Lawrence Radiation

Livermore,California.

University,

University
Oregon State

Laboratory,

of California,
University,

University

of California,

-2I.
During a routine
Laboratory

(LRL), it

with time.

was negligible.

that

the electronic

but a careful

temperature

machine at the Lawrence Radiation

the measurements

that

The measurement

and gage output

a high degree

at first,

was the cause,

of a sensitive
ture

of a measuring

was observed

It was thought,

measurements
drift

test

INTRODUCTION

system

of correlation

was then

the electronic

monitored

by means

near the gage.

over long time periods.

between temperature

significantly

gage used to make the

check showed that

pickup mounted in the air

were recorded

varied

variation

Tempera-

The results

showed

and measurement

variation.
Similar

tests

were conducted

systems with similar

significance

of this

working tolerance
tolerance

results.
effect

Figure

is clear

of the gage.

of this

began an investigation
investigation

progressed,
In the field
si..ngle

2.

when the observed drift

is 100 microinches

of the gage is only 10.0 microinches.

As a result

distur.bing

of thermal
it

types

of measurement

1 is an example of such results.

The drift

accounts for 10070 of the tolerance

1.

on many different

The

is compared with the

and the working

In the case shown, the drift

of the gage.
development,
effect

in Dimensional

became increasingly

of dlose tolerance

the Metrology Section

clear

Metrology.

of LRL

As the

that:

work, thermal effect

is the greatest

source of error.

The ugu....a.1
" correction,"

efforts

to correct for thermal error by applying expansion

or by air

conditioning

the working area do not always

solve the problem and are based on an incomplete

the problem.

understanding

of

-33.

The specifiedaccuraciesof modern precisiontools and gages are

attainableonly if the thermalenvironmentmatchesthe requirements of each measurement system.

~t has been helpful

the effects

of average temperatures

temperaturevariation
this

to think of the temperature

arbitrary

about tltis

division

and means of reducing

used to discuss

other than 68 degrees,

avei-age.

of theproblem.
thermal

thermal

effects

errors.

II.

frequency
EFFECTS

response

o.f,

(2) the effects

The paper ttrgariization

There is also a discussion

Appendix A is a glossary

problems.

reflects
of ways
of terms

Appendix B is a detailed

of a " drift check and Append.ix C is an outline

thermal

problem in terms of (1)

procedure

of a method to determine the

a measuring system.

OF AVERAGE TEMPERATURES

OTHER THAN 68F

An inch is the distancebetweentwo fixedpointsin space.It is defined

as 41,929.398742
wavelengthsof the orange-redradiationof krypton-86when .
propagatedin vacuum. An inch does not vary with temperature.This fact
is obscuredbecausethe lengthsof the more common representations
of the
inch such as gage blocks,
The lengths
ture.

of most of the materials

In April,

Meeting in Paris
automatically
agreement

lead screws,

do vary with temperature.

we deal with also change with tempera-

1931, the International

agreed that

and scales

Committee of Weights and Measures

when we describe

the length

mean its length when it is at a temperature

was preceded

by intensive

international

[ 1, 2, 3, 4, 5, 11 ]6. This agreement., meansthat .....

t_h..e...m....e_.~..s.~.r..e.ment

of an object

we

of .6_~_de_g.rees.

This

debate and negotiation

it is not necessary to specify

temperature on every drawing (no more necessary than it

is to define the inch on every drawing).

6Numbersin bracketsdesignateBibliography
at end of paper.
0

-4If dimensionsare only correctat 68 degrees,how have we been getting

by a. ll these years by measuringat warmer temperatures?The answer is
that i~ o~t~ wo~,l~ i~ s~.e~t and out- ~eate is steer ~he two expand ~ogethex. a~td ~he
resultant

errors

tend to cancel.

such as aluminum, the errors

to this

error

however, the work is another

are ~tifferent

and they don,t

expansion."

We can get :i.nto

as " differential

if our work is steel,

If,

cancel.

As a result

development of practical

laser

of the discovery

fringecoun.ting

be using more of these " honest" inches

thisproblem.

We refer

the same trouble

but we are measuring with the " honest"

from an interferometer.

material

inches

of the laser

interferometers,

that

come

and the

.we expect to

and will have to be very careful

of

Knowledgeablemachinistphave always made differentialexpansion

corrections.

The thing

corrections

that

are not exact.

is sometimes

overlooked,

however,

Our knowledge about average

coefficients

pansion is meager and we can never know the exact coefficient

This inexactness

we call "u.ncertainty

This inexactness
68 degrees,
Its

is at least

for different

consulted

in expansion of a material
expansion

experimental
materiat

errors

is zero when the average temperature

to the thermal

materials.

distance

from 68 degrees.

We have reason to think that

factor

also includes
directions.

and the handbook or ~ nominal"

and the material

is

stated

it

One
that

the co-

iron may vary as much as 4Y0 between thin and

in different

and because

of ex-

expansion."

in the course of our investigatiop,

This uncertainty

these

of each part,

and on up to 25~]0 for other materials.

of expansion of cast

thick sections.

thermal

according

5~]o for gage steel

metallurgist,
efficient

or uncertainty

and increases

magnitude varies

of differential

is that

of dissimilarities

of our workpiece.

the possibility
Diffe.Fences
expansion

of differences

between the actual

occur because

between the experimental

of

Coml~lete
epai~sion

the

of the errors

are notably

coefficient

be expected
that

studies

the

The data

of expansion

for

so,

of the

by Goldsmith

of common materials.

properties

as ~chard

in the e~timates

presented

some of the more exotic

knowledge

necessarily

absent.

introduced

materials~
of steel

K. Kfrby

This

of the

of thermal

et al.

disagreement

but intuition

might

would indicate

would be m~e e~act.

National

[ 9]

Bureau

Not

of Standards

port 7:
s
" The accuracy
expansion

is about

of the steel
many heats
~3 percent,
about

for

can be reduced

thermal
part

part

expansion,

to part

indiv~dua~
deviation

among several
and (c).

of steel

that

will

same chemical

treatments

(a)

content

of the

is about

cut from different

annealed,

of d~fferential

expansion

the uniformity.of

nominal

percent)."
cannot

~d by limiting

of

of expansion

controls,
the

be made. The

coefficients

coefficient

and metallurgical

locations

is about ~2 percent

of about:~5

more accurate

among

same steel

cause a difference

chemical

treatment

of the coefficient

has been fully

and gage expansions,

of uncertainty

of thermal

and mechanical

among samples

by improving
better

heat

the

heat

uncertainty

from 68 degrees.

Control

the

The .precision

by establishing

through
part

if

of a coefficient

of nominally

or cold roll~ng

Corrections
error

of steel
(b)

value

~5 percent

~s indicated.

~10 percent,

in a large
(hot

of a tabulated

from

by determining

room temperature

..
of ~ifferential

maintaining

a ~68 degree

averag~

temperature."

edge of all

coefficients,

the con~usion

expansion

is the primary

Even ifwe

and pdssibility

reason

had an exact

of mistakes

for

knowl-

in making

7A personal
comm~ication
from Richard
K. Kirby,
In charge,
Thermal
Expansion
Laboratory,
Len~h Section,
Metrology
Division,
U. S. National
Bureau of Standards,
Washington,
D. C.

-6corrections

is a second reason

for maintaining

progressed,

it became necessary

to establish

68 degrees.

As our study

a more exact definition

of

.refer~o AppendixA. O~ossaryof Terms,definitionsNo. 15 through

whichare pe~tinent
to the discussion
in thissection.
Three examplesare given below to illustratethe Consequencesof
average

temperatures

other

than

68 degrees.

Possible

be 13~o, 37~0, and ~.0~/0 of the working tolerance.

the effect

of temperature

They do not include

standards

variation

the other

and comparison

only 10~/~ of the working tolerance

Example

of measurement such as accuracy

The traditional

for all

(500/~in.)

the Nominal Coefficient

The K for the steel
of Nominal Coefficient

part with a tolerance

of ten to one allows

of plus or minus a half-

is measured in a C-frame comparator

case,

by comparing

75 degrees.

A handbook lists

of Expansion (K) for the gage block as 6.5/~in./in./deg.

part

is assumed to have the same value.

The Uncertainty

of Expansion (UNCE) for the gage block is estimated

at plus or minus 5~. and for the part at 10~o (its

Uacertainty

rule

of

measurement error.

to a 10 inch gage block in a room which averages

For this

do not include

No, 1

A 10 inch long steel

thousandth

These errors

are shown.to

which is covered in the next section.

errors

technique.

errors

the Nominal Differential

of Nominal Differential

It is the sum of the two Uncertainty

Expansion

exact comPosition is unknown).

(NDE) is zero.

Expansion (UNDE) is,

The

however, significant.

of Nominal Expansion (UNE) values.

-7NDE

~. No correction

UNE gage

ffi 0

necessary

"- 10 in. 6.5/~in./in./deg

block

UNE part

ffi 10 in. 6.5/~in./in./deg

7 deg 10~o ffi 44 ~in.

UNDE = 86 ~in,

66
~ 100 -" 13~/0 0.f

Example

is measured
readings

long plastic

on a surface

Temperature
(K) for

6.5 pin./in./deg..

with a tolerance

of plus

composition

using

gage.

the

handbook

gage steel

The K value

for

nor heat

making the

differential

expansion.

He computes

Cadillac
value

gage

reading.

is computed

is listed

of past

to the contrary

to the dial

Nominal

75 degrees

indicator

A correction

measurement

experience

of

reading
for

gage

of Expansion
know the exact

with

the plastic

is thoroughly

the NDE correctly

plastics

familiar

and applies

be made.

and

is estimated

it

with
in the

which is used to transfer

UNDE cannot

(7

by the manufactuger

we do not

the UNCEfor

inch

to the

Coefficient

Coefficient

at 10~0 since

Because

The inspector

direction

compare it

averages

the

of Nominal

is estimated

treatment.

of any information

lists

the plasti6

The Uncertainty

the gage steel

stand.to

or minus 0.002

assUmed to be used in the Cadillac

at 25~]o.

proper

an indicator

The room temperature

Offset)..A

as 40/~in./in./deg.
(UNCE) for

part

plate

of a Cadillac

Expansion

a lack

,
~

No...2

A 10 inch

degree

tolerance

working

Its

the
possible

below:
Corrections

ffi

6.5/~in./in./deg

)< 7 deg 1070 ffi

46/~in.

UNEpart

= 10 in. ~< 40 ~in./in./deg

7 deg )< 25~]o ffi

700.. pin.

-- 10 in.

are made

NDE
UNEgag
e

UNDE = 746
746

~< 100 = 37~of

working

tolerance

-8Example

No. 3

A 10 inch long aluminum part with a tolerance

inch is measured on a surface
the readings
(2 degree

of a Cadillac

Temperature

plate
gage.

Offset.)

using an indicator

gage is assumed to be 6.5 Bin./in./deg.

example,

decides

with an NDEcorrection

in this
that

to

70 degrees.

the NCE for the

The UNCEfor the gage is estimated

the magnitude of NDE, and arbitrarily

error

averages

The NCE for the aluminum part

and for the aluminum part at 2070. The inspector

he does " not bother"

stand to compare it

The room temperature

As in the previous

assumed to be 13.5 Bin./in./deg.

of plus or minus 0,001

is

at 1070

case does no_./t

70 degrees

appreciate

is " close

to his readings.

enough,"

The possible

is computed as follows:
NDE

-- 10in. (13.5 - 6.5) 2 deg

140/~in.

UNDE:
UNEgage --

10in, XS.5X2 deg i07o = 13 Bin.

UNEpart --

i0 in. X 13.5 X2 degX2070 = 52/~in.

UNDE =

65 ~in,

NDE plus UNDE =

205 ~in.

205
1000 X 100 = 207~ of working tolerance
In the above examples we assumed that
gage and part
adequate
thermal

result

temperature

temperature

of the room. If

time has been allowed for the gage and part

equilibrium

assumption
Instead,

were the same as the average

the average

this

is a reasonable

assumption.

to " soak out" and reach

Unfortunately,

does not apply to the instantaneous.temperatu~.e

the environment

is that

differences

is continually
in temperature

varying

of the

of these

this
components.

around some mean value.

in the various

The

components are dyh~mically

-9
i.

inducedin the system.The next section,discussesthe errors causedby

.v~nriation
in thermalenvironment.
III..E.FFECTS
The Two-Element

measuring

elements

apparatus

arranged

The comparator

efficient

can be viewed as consisting

to form a "C." Figure

of a C-frame comparator.measuring
tubing.

IN THERMALE ..NVIR.0NMENT

System.

All length
of individual

OF VAI~,,IATIO~

the diameter

frame and the part

gage head will

might select.
thin section
comparator

read zero after

is slow enough to allow both parts

changes,

there

will

Somewhere in between these

of temperature

change that

somewhat similar

results

to resonance

To confirm our intuition

simplified

were made for

this

to that

of the

deviation.

The

If the rate

If

is so

there will again be a small change

there

will

be a frequency

in a maximum change in reading.

on the nature

we

to keep up with the temperature

extremes

in.vibration

that

Of change of temperature.

be a small change in gage head reading.

that even the thin tubing can~ respond,

the

however, the relatively

show a temporary

the rate

further

the same as the part,

of the tubing will react sooner than the thick section

will depend on the rate

of hollow

If the co-

~oak-out at any uniform temperature

frame and the gage head will

in reading.

section

form two elements.

If we induce a change in temperature,

amount of the deviation

fast

2 shows a schematic

of a short

of expansion of the comparator is exactly

of a number

This is

work,
of ~hese effects

shown in Fig. 3. Sample heat transfer

model and programmed on an analog

the above model was

calculations

computer.

The"

-10cylind~,r

with the displacement

Both cylinders
i.]

pickup can be considered

are made of steel

diameter

and Cylinder

computer-predicted

and are 4 inches

and miaus one degree

sinusoidal

cylinder

head reading

The effect

and reaches

We call

this

predicted,

Fir.teen

effects
example,

elements

workpieces

of temperature

have much longer

elements.
unusual.

in Fig. 4 shows the predicted

difference
(definition

frequency

at a point

in the lengths

of

24) of the system.

is plotted

in Fig. 5.

in between which we call

resonance.

14) of the system.

In

of course,

variation

overall

of the environment

was 68 exactly.

can have quite

responses

Magnitudes

gage

is small for very high or low frequencies

pin. may appear to be a negligible

these

of the

lags the thin

even if the Time Average Temperature

machines and machine tools,

and real

having a frequency

occurs at 1/2 cycle per hour and has a value of 15 pin.

would occur

mechanical

of a plus

than one-third

the " Frequency Response" (definition

case resonance

This error

the drift

as a result

temperature

~he " Thermal Drift"

a maximum amplitude

Figure 5 is called

out,

The do{ted line

of varying the " Thermal Vibration"

As our intuition

and .all

shows less
and its

A is 2 inches

Figure 4 shows the

temperature

which is the same as the instantaneous

the two cylinders.

this

cylinder

change of the thin cylinder

by 3 or 4 minutes.

Cylinder

of the two cylinders

change in air

of one cycle per hour. The thick

temperature

long.

B is 1/2 inch in diameter.

changes in length

the comparator.

magnitude,

of 150 ~in.

Rather than mass ~lone,

measuring

dont have uniform coefficients

different

much worse.

coefficients.

If the part

would be much more severe.

lengths

but real

and more severe

per degree

Real systems

at resonant

the more significant

This makes the

were Lucitc,

differences

for
generally

in mass between

frequency

factor

through-

are not

is the ratio

cubic inches of volume to the square inches of surface exposed to the air..

of
This

-11ratio is proportionalto the " Time Constant" (definition13) of the element. The
%in%e constant is discussed in the followingheat transfercalculations,which

necessary, because the important conclusions have been presented above.

~e have used a ga~e frame %o illus~ra%e the effect of temperature variation,
but i~ should be emphasized that %he same %hin~ happens ~Q m~chine ~ooI fr~es.
Deflection due ~o temperature variations is common %o all machine s~ruc%~es
~e%her ~hey be measuring machines or machine tools.

Calculations

for

Fr..equency

To simplify
a.

The bodies

resistance

to heat

added simply

raises

Response

the calculations,
always

the following

have uniform

transmission
the

between

temperature

The temperature

of Two-Element

of the

assumptions

temperatures,
the parts

at all
air

System

points

surrounding

i.e.,

Model
have been made:
there

is

no

of the body and any heat

uniformly
the

bodies

and instantaneously.
ts uniform

at

e
T
c.

All heat

transmission

to and from the

body is

governed

by Newtons

law of cooling:
q = hA(T - e)
where A is the surface area of the body in ftp, h is a film coefficientdefining
9" OF, and q is the
the abilityof heat to pass from air to the body, in Btu/hr-ft
rate of heat in-flux,Btu/hr.
d. The heat store.d in thebody is proportionalto the thermal
capacitanceof the body or that
qs = Cp ;~V dT/dt
8~eeping in mind that radiative and conductive envi.ronmentscan exist, we
shall limit the follo .wing discussionto the effect of a convectiveenvironment
on the measurement process and the resulting error.

-12where qs is the rate of heat storage,

Cp is the specific

heat in Btu/lb F,

V is the volume of the body in ft 3 3p is the density,

pounds-mass per ft

and dT/dt is the rate

with time.

inilux

of change of body. temperature

must equal the heat stored in any interval

Since the heat

of time

q "- hA(T - Te) -- qs -- CpVpdT/dt,

solving for Te yields the differential
equation describing

the system:

(1)

eT+r dT/dt -- T
where
Thermal capacitance
Thermal resistance
Because of assumption
ever,

for metallic

approximation

(a), equation

objects

(I) is only approxim:ately

the thermal conductivity

correct.

How-

is high enough to make the

reasonable.

Equation (i) is well known in the literature

systems [10] in which all

elements

on the analysis

i~ represents

are called

of linear

" time-constant

elements"and 7 is called" time constant"of the element.

Given that

re varies

sinusoidally

around some mean Te00 i.e.,

"

Te = (Tema x - Te0) sin ~t

where ~ is the frequency,of

osciliation

in radians

(2)
per unit time, the solution

to

equation (1) gives:

T = (Temax - Te0) sin (~t + ~,)(1+
where the phase lag .angle
~ = tan

-1

isgiven

by

~2 v2)l/2

(3)

-13\.Vhcn

this

solution

relationships

Diift

is applied

between

to the two-element

the temperatures

Measurement

of the

(definition

26).

indicate

from the

drift

in a measurement

To make a drift

comparator

to the

record

the

relative

motion

between

of the

measurement

process.

This

development

of high

recorders.

sensitivity,

The electronics

means of proving
"electronics

the

drifts"

expected.

Details

Predicting

the

observations
know all

of less

Effects

of their

constants.

complexity.

installed

how large

decisions
of the

be for

are

to be made.

given

any system

conditions

made possible

by the

transducers

and

a simple
indicates

3 environments

are given

vibration

inAppendix

that

can be
B.

us to make quantitative

in sys.t.ems

check provides

will

Such information

before

is necessarY

the

from this

regardless

us with

our investigation

determine

any environment.

environment

systems

which we

a practical

can provide

be in real

Therefore,

for

uswith

in a given

approaches

ways to predict,
in

normal

Our exper{ence

above allows

systems

means to ~xperimentally
and to find

may be, ~nd

Variation

of these

the errors

the

case

25) provides

per day for

The drift

Neither

under

a " drift

necessary

as the

has been

devices.

of thermal

on real

merely

(definition

procedures

in a given environment.

ment systems
will

and the

is called

displacement

3 ~in.

approac,.h

is

or part

procedure

of Temperature

evaluation

of determining

than

it

the elements

of these

the effects

the time

master,

check

of drift-check

about

means of error

drift

system

check

drift-free

stability

The mathematical

study

of the tw.o elements

the

Check

check"

design

system we obtain

dynamic
information,

if

they

a means
are

rational

included
response

a
of measure-

what the "drift

-14A su:dy o~ the literature[ I01 on the analysis of linear systems shows that
it is possible to conduct "step input" tests, the results of which provide a

f~asib~lity ofapplyin~ this procedure %o a real measurement process, a series

of exporimonts was conducted in an L,~L inspection shop. [n these experiments
th~ apparatus consisted of a 15 .~nch Sheffield rotary con~gur ~a~e measurin~ a
hollow steel hemisphericalpart, as shown in F[~. S.
The Sheffield ~age chosen was particularly suited ~o these experiments
because i% was loca~ed in a room ~hat had a particul~rlygood air-conditioning
system. Room-air temperatures in the vicinity of the gage responded to a one
de~ree chan~e in the set point of ~he air-conditioningcontrollerwithin several
minutes.
Linearit~ of the system ~as established in three experiments which consisted of suddenlyraisin~ the st point of the controllerI degree in the first,
lowerin~ it i de~ree in the second, and raisin~ it 2 de~rees.inthethird. Air
temperature a~ a point just above the par% was recorded usin~ a thermister
magnetically held in contact with a I0 inch piece of 0.010,inch shim stock. Resulting drift was recorded by the equipment on the gage.
The results from this series of experiments were compared. All three
drift curves showed the effects of a high de~ree of linearity. They differed only
in magnitude and this disagreement was less than I0~.
Subsequently, an arbitrary temperature fluctuation was imposed on the
room by driving the controller set point with a motor-driven cam mechanism.
Temperature and drift were recorded as before.
The recorded drift and the correspondin~ recorded temperature chan~es
for the one-degree step input change experiment were used as shown in Appendix
C to compute a theoretical drift from the forced-drift temperature data. Figure

-157 shows the step i~put temperature

recorded

and computed drift.

change and drift

Considering

profiles

the fact

that

the experiment

tinued over a period of about 6 weeks, we think the results

applicability

fully

encouraged .us to use the computation

the frequencyiesponse

effect

of the s~stem ~ith the results

of additional

The next question

one environment

this

simple step input tests

time-.constant

shown i~ Fig. 9. Com-

to a similar

means that

pattern

"Can data obtained

system in another

a gage manufacturer,

in his lal~pratory,

as well as the

on a system in

environment?"

by conducting

can provide information

allow the customer to decide whether or not his environment

the gage.

If

these

that

will

is suitable

for

..

To answer the question,

w.e conducted a normal dri.~t

Fig. 1} on a second 15 inch Sheffield

room with a different

environmental

temperature

from this

variation

from the first

experiment

and actual

drift

large variety

Application
room large

control

check (shown in

contour gage located

system.

in a different

We used the recorded

~ystem and the frequency-response

system to.

are shown in Fig.

is impressive,

rotary

compute a predicted

drift.

10. The correspondence

information

The results
between

of

the computed

and though the method used must be tried

of. cases before we.know how general it is,

the affirmative

ability

the

elements,

to be answered is:

be applied

the answer is yes,

this

justify

method to calculate

paring these data with those in Fig. 5, we see the typical

obtained

con-

of this type of system testing.

These results

distorting

and Fig. 8 the

for a

we feel confident

that

answer has been obtained.

of this method require.

enough to house the completed

to hold a given temperature

s a high quality,
machine..

to a tolerance

temperature

controlled

The room must have the

that makes the step input

change

slgnificaat.

Tlm lime r(:quired

/or

tu~e

be a smal~ fraction

of the ~oak-out

should

the room o stabilize

predictions

by analyzing

to be difficult,
inch drift

Work on the soiuti.on

All.
elements,

variation

as well.

is now underway.

length

measurement

systems

can be discus~.~ed

a part,

amaster,

the

comparator,

The master

and a compa~ator

i.s

sometimes

are

of the device

not combined,

because

is the cornparator.
element

is

it

the micro..

of three

the

is

the

part

with

combined

with

screw

is

If the master

introduced

in

as i~s amplitude.

compare

In a ~nicrometer,

a ti~ne

check taken

in terms

usedto

obscured

as in a micrometer.

and the rest

comparator

of temperature

drift

is the depend:ence.~f

problem,

master.

at a methodio~, making

of an ordinary

of this

the

master

the results

bul;possible.~h.e
~ , difficulty

on the frequency

time of the machine.

I~ ~ w~.~ ~ui.ll.d al,~,tve

II w~.~ild l~.~ ~.1~ (~tlv~.~ll~.oai:

these

at the new tempera-

into

the

and the
the

measure-

merit

process

mastered,
lapse

or set,

the

between

We have already

in a two..,

element

tween

mas~er

there

is

comparator,

at the same time

is the difference

system.

the

because

seen

system

cau~es

maximum temperature

drift
variation

and compara~or

and between

this

variation

temperature

time cycle.

This

a master-comparator

master

error

it

.in

of calipers,

is used to J.ndicate

a two-clement
that

and comparator.

such as apair

the

r~ystem

different

a drift

error.

means that,
crror
error.

that

on a part.

A simiiar

This time

of part
error

and compara~or

will.occur

in athree...element
must also

system,

be coasidcred

The two drift

e-.e].ex~ent

be

and a three-.element

response

and comparator,
a ~hre,

cannot

curves,

can be used
system for

to get
betweenpar~

to approximatc
any mastering

-17.I~" lhe mastcringtime is zero or insignificantlysmall, then ~he comparator is slaved to the master and the temperalure variation error is the
drift between part and mas~er over a representative~ims periocl, This is
equivalent to the two-element system discussed previously. The representative time period is usually a working day, but may be shorter or longer
pending on environment control,and work habits. It should be long enough to
cover the entire temperature cycle of each measurement situation. The
of part and master cannot usually be compared directly but can be compared
indirectly by comparing the par~-comparatordrift curve wi~h the mastercomparator drift curve. The rr~aximum excursion of ~he two curves for the
same temperature

phase

will

maximum partr.to-master

provide

the

approximation

because

and am: plitude

the temperature

over the

representative

drift

error.

conditions

This

time
error

period

is

of the two drif.t

an

checks

will

never be identical.
Figure

11 shows part

made sinusoidal
an average

curves
error

measuring

are
will

The part
time,

and in phase..The

at

occur
is,

the

the

the part

i.s

measured

length

is

cannot

then

the drift

becomes

an additional

source

possible

temperature

greater

than

the

curves.

are equal

than

at

comparator,
master

comparator

of error.

determined
.i

a finite

length
It

are
from

can be

and master
The maximum

of maximum difference.
but

length

with respect

mastering

with

zero

at measuring

on a part

maximum error

they

length.

reading.

I~ can be shown.that

for

in

when the part

the point

be used to indicate

variation~rror
already

the

to the

of the

drif.t

in an error

measured

held

they

other

result

with

For simplicity

show absolute

any time

when the part

of course,

is mastered,

at

will

If the co.mparator
it

curves

same point

comparator

drift

of 68 at which point

temperature

seen that

and master

mastering
time.

at the same time

to part
,the
cycle

and master

maximum
time

from part-to-maste~

will

not be
drift

,:nlcss
eilhe~-tl:eI.(~I.~.[
lmvt..to..coml)arator
driftor the totalmaatet-tocoml)aralor
(Iri~(.(lu~ingthe maE~tcrin5cyc].etime is more than twice
maxin%unt
i)arl...io...mastc~r
driftet-t-oP.
]t~tlre

I~ ~.~hows

the ab~olutcdrifts Cot part, master, and comparator,

The phase lag shown is I.ypical

variation.

In the example

is the laaaximu~l
Mastering

the

drift

curve.

error

will

shown,

temperature

comparator

at

1he

with

true

be approximated

error

drift.

is true

for

lure

If

error

cyc]e

between that
error

because

variation
drift

error
rate

when the two drift

the

time

cycle

between

time

after

betwe.en
curves,

X1. This

and subtracting

can

the part-to-

drift,

the

trae

A, is

error

A -.

B is greater

for
the

X1. The same

those

con-.

maximum tcmpera-.

be done at any time during

rate

In this

the

drift

is A. Figure

of mastering
of master

pari:.,eomt)arator

of the

comparator

is

example the part..to-.master

B, so part..to-.master

l:hag

and

of part.-to-.comparator

would remain

when the drift

regardle~)s

variation

mastering.

and of the ~)art.

~;han

time.

comparator

mastering

variation.error

could

cycl.e

number 1 is

drift

approximates

A B, ]-~

(B)

maximum temperature

the part-.to..comparator

is B and the rnaster.-comparator

be greater

displace

is {~rcater,

indicating

shortl.y

drift

mastering

time rmmber 2. In each case,

~.han

of the master

of a comparator
apparent

mastering

13 shows what happens

B can never

temperature

cycle

the maximum temperature

including

for zero

subsequent

B conservatively

B were greater

Figure

A -.

B is

error

the peak-.Co..val].ey

cycl.e

to temperature

maximum part.-to-ma.ster

mastering

whichever

mas~eri.ng

variation

mastering

drift

A .-

B so A ..

ditions,

the

In the example shown,

so that

responses

time shown wi].l

that

for

by measuring

greater

than

the

the

or master.-to...coml)arator,
master

the.

variatJo~l

I~ can be seen
now vary

dicating.

for 1:he varyJ.ll~l

drift
eycle

and part

13 shows that
remains

time.
drift

the

maximum

This

condition

rates

becomes

and master-.comparator,

-19arc compared for part-to-master

for temperature

drift,

If the two drift

phase are out of phase,

and part

the comparator

and the maximum temperature

of maximum excursion

between tl~e

curves when aligned

drift

variation

two curves

must be between

error

becomes ~hat

over the representative

time

period.
The general
can be stated

case for

maximum .temperature

small,

mastering

~[-Maximum excursion

TVE =

comparator

drift

is greater,

for the mastering

part-to-master

TVE

If the rbsult
variation

cycle time

of part-comparator

or the maximum master-comparator

drift

error

drift

cycle time period

previously

drift,

of A - B is greater
error

for the chosen mastering

drift.error

remains

Part-to-master
I
drift
.J

result

cycle

is the maximum

time.

error.

A will not be more than 2B, and will therefore

part-to-master

drift

error

prevent

is made to have a drift

of the part and the master regardless

the peak-to-valley

drift

If B is

the maximum tempera-

ture variation

A will not be more than 2B if the mastering

the

chosen.

than B then that

if the comparator

whichever

chosen and subtract

rMaster-comparator
drift
or
part-comparator
drift,
which-| _
I ever is greater, for mastering i
I! cycle time period chosen

then the part-to-master

tween that

approximation

Lfrmtativemastertime(part_to.masterperiod.cmparatr
driftdriftoVererror),
represenmastering cycle time, take either the maximumpart-

For a significant

greater,

error

as:

For zero or insignificantly

temperature

variation

of mastering

cycle time is kept.

o~ the comparator

from either

.cycle
short

not add. to the

rate

be-

time. Also,
enough to

part or mas~er

-21comparator combination while simula{ing the actual conditions of the measurerncnt

process.

checks

are

Both the

analyzed

occurring

within

comparator

to determine

a time

period

the

time period.
It should

checks

are then

Variation

Error

For zero

of 24 hours

substituted

maximum drift

time

period

drift
of each

cycle..Both

then

analyzed

occurs

the

to determine

wxthin"a

" representative"

is somewhat difficult

pattern

of temperature

is sufficient.

The results

in~o the following

expression

master-

variation.

of the drift
for.

to define.

or small

check

Temperature

masteri.ng

time.

mastering .cycle times.

Parttmaster
Master-cmparatrdriftr
part-con~parator
drift,

TVE =

Use whichever

cycle

~Maximum excu..rsion
of part-comparat0r
and mastercomparator
drift
curves when curves are aligned
for
in-phase,
temperature
conditions
.over representative.
time period.
::
(l~art-to-master
dri.ft)

For significant

~ ~
which- 1 -.~drift
chosen |
.]

ever is greater.,
for
mastering
cycle time
of above

The temperature
Expansion

TVE is

variation

error

(NDE).and

the

(UN. DE) to obtain

error

as above.J

greater.
is combined
Uncertainty

the Thermal

with

the

Nominal

of Nominal Differential

Error

Index

(TE:I).

The plan

consists

of:
1,

Computing

the

Nominal

2.

Computing

the

Uncertainty

3.

Determining
drift

In

(TVE):

TVE =

Differential

are

that

the full

part-comparator

mastering

curves

This " representative"

analyses

of the

drift

of the drift

a period

value
to the

be long enough to reveal

most cases,

and the

equal

and part-comparator

the maximum excursion

Expansion

master-comparator

the

check dat~.

Thermal

Differential

Expansion

of Nominal
Variation

:"

Error

(NDE).

Differential
(TVE).

Expansion(UNDE).
by ,evaluation

-224. Summing the absolute

5. If

NDE corrections

2he tlme-honored
limited

to 10~0 of the

values
are

rule
working

obtained

made, NDE is

of 10/1

suggests

tolerance.

in 1, 2, and 3.
not included
that

in the

the ~otal

We have found,

above

measu~ing

however,

that

sum.
e~o~ be

the

error due to temperatureis,lnmost c~ses, so large that in order to stay within

economic reality we must plan :~n giving up the full i0~o ~nd ~ore for temperazure alone.
The following example shows how the Evaluation P~an is used in practice.
This example is the same as the one used in Section II but now includes the
effects of temperaturevariation:
A i0 inch long steel,partwith a toleranceof plus or minus a halfthousandth (500 pin.) is measured in a C-frame co~parator by comparing
it to a i0 inch gage block in a room which averages 75 degrees. A
handbook lists the Nominal Coefficient of Expansion (K) ~or the gage block
as 6,5 pin./in./deg.The.K for the steel part is as~$umed to have the same
value. The Uncertainty of Nominal Coefficient of Expansion (UNCE) for
~he gage block is estimated at plus or minus 5~ and for the part at ~0~
(its exact composition is unknown). For this case, the Nominal
Differential Expansion {NDE) is zero. The Uncertainty of Nominal
Differential Expansion {UNDE) is, howe~er, significant. It is the s~
of the two Uncertainty of Nomina~ Expansion (UNE}.vaiues,
A 24-hour drift check between the comparator an~ master gage
block shows a 300 pin. peak-to-valley drift. The comparator is
normally remastered eveny 2 hours. Interpreting.,~hedrift checks for
maximum drift in 2 hours gives a maxim~ value qf 30.~in. Because
the part has fewer cubic inches of volume per square inch of surface
than the gage block (its time constant i~ smaller} the ti~e constant

-23mismatch

to the relatively

part-comparator

drift

into the Evaluation

heavy comparator

frame is Worsened.

is found to be 350 ~in. in 24:hour:s,

l~lan for the above conditions

yields

The

Substituting

the following:

NDE

= No correction

necessary

UNE gage block

= 10 in. 6.5 ~in./in./deg

>~ 7 deg 5~0 = 22 ~in.

UNE part

--10 in. X 6.5 ~in./in./deg

X:7 d~g X 10~o = 44 pin.

= 0

UNDE = 66/~in.
A
B
! Maximum comparator drift
from! Part-to-master
TVE = !part
or master over 24 hours
I " j drift = 50 ~in.
Use A - B or B, wl~ichever ~s greater
(30-50)

<

Therefore:
TVE

= 50

~16
500 X 100 = 23~ of the working tolerance
The above example shows a thermal
corrective

action

decided to accept
to normal.

is indicated.
a higher

.way of estimating

control

percentage

A" bad" environme:nt

whidh does not justify

each job.

If,

however, the tolerances

thermal

the temperature

or we

would return

become a " good" environment

The Evaluation

Plan is a

problem for each shop, each machine,

and

us whether or not we need to improye our temperature

of the large
problem.

it has become stylish,

increased

index the situation

the cost of any impro.vements.

and by how much. The plan provides

for investment

index of more than 1070 and

would suddenly

the temperature

It can tell

error

concrete

.amount of money that

It can also prevent

to have special

economic justification

may be necessary

overdesign

temperature

to control

in the situations

controlled

:~reas.

where

It substitutes

-24an o.~dc,ly

~hinking

priorities

are established

be made. Should
reduce

process

we try

a positive

In spite

response

is that

we are still

estimating.

significant

the plan

is,
V.

~
~verage

The error

than

FOR

the

at

all.

we ~r~,~o

these

questions,

may be made.
have been raised.

pr~ceddre

objection

when obviously

is to agree

that

the plan

by 2S~/~ or more and still

No plan

who arbitrarily

Than

for

at all

decide

that

means that
this

or that

be a
we must
en-

68 are

limited.

THERMAL

the error

resulting

INDEX

from average

They can be summarized

by making nominal

of coefficient

ERROR

68

controlling

more ac.curate

uniformity

that

to be an exact

DECREASING

Other

can be reduced

by establishing

answers

should

accept.able.

METHODS

other

no plan

efforts

some objections

may be in error

of experts

The possibilities
atures

It

over

or is not,

Temperature

p~etends

Our answer to this

advancement

vironment

the plan,

Natural

or shoul~i

not only

to any improvements

and not exact.

depend on the opinion

to S8 degrees,

The plan

of the advantages.of

set rules.

where our improvement

to move closer
variation?

One objection

is not perfect

emotion or arbitrarily

to indicate

our temperature

i~ gives

for

nominal

differential

coefficients

of expansion

frompart

in one sentence~

expansion

corrections,

of expansion,
to part

temper-

by improving

through

better

chemical and metallurgicalcontrols, by determining individual part expansions,

and by limiting the room temperature deviation from 68 degrees.

Temperature

Variation

Error

What are some of the th~ngs

system
thermal
drift

to withstandtemperature
response
between

the

of the
master

master

we can do to improve
variation?

Our first

and comparator

and comParator.

equal.

Shortening

the ability
reaction
This
the

is
will

mastering

of a gaging
to make the
result
cycle

in zero
has

-25the

same effect,

increased

mismatch

z,c.~spm~se
problem

but is

one part.

about

half-way

more detail
in several

Tho best
between

ways.

insulate

is

temperature

thing

effective,

thelr

ultraprecise

Another approach
room.
the

The frequency

advantage

frequency
are

so slow (in

lished
that
the

gage.
this

by increasing
a plus

same drift

Because

diagram
the

It

th~s

the resonant

the

enviro~ental

change ~ts wall

machine.
approach

temperature

or minus 4 degree

control

this

to control

the

The ~oore
in comparing

the rate
contour

frequencles

environment?

of the whole

of one part

frequency

frequency.

the

excursion

rotary

per c~cle),

itself.

env~ro~ent.

of increasing
of the

obtain-

Time constants

to the part

maybe cheaper
the

environmental

14 hours

as a plus

the

in

of wall thickness.

we cant

gages to an accuracy

case

or mlnus 1 degree

improve

uses

is the possibillty
response

can be applied

to be

is readily

design

more

~s discussed

of only 1 ptn./in.

around

Conn.,

used for

can be accomplished

Invar

the temperature

area

step

of mismatching

of the

practical.

expensive.

in a small

(th~s

response

~he

dr~ft

can be done to improve

reduce

Tool Co. of Bridgeport,

calibrating

that

but also

drift

coefflcient;

we can do is

is to simply

aompa~htor

and by proper

change its

are

an

thermal

ellmtnat~

most gages

of thermal

.solutions

things

is to make the

and the master

i~ quite

we may create

completely

and has a coefficient

we cant

excursion

goal

because

Adjustment

some of the

reaction

This

III).

none of these

The only

Special

the part:drlft

it;

because

~s to destgntho

by use of insulation

What are

room.

approach

compromise

cost

Unfortunately,

however,

"A worthwhile

The use of Invar

can be controlled

Our first

goal,

part,

not a practical

in Section

thickness.

a false

e~ement~ ~he ~ame, This

th~e

at a reasonable

We cant

is

to the

of a~l

than

able

This

and

in ten million.

of cycling
gage

and the
of real

mismatching

of the

(Fig.

9) shows

resonant
gaging

is

best

systems
accomp-

Interpreting

Flg.

at 0.07

per hour glves

control.w&uld

cycle

9 we see

g~ve at one cycle

per

"

-26hour.

In some cases

it

is

possible

to increase

the

rate

of room cycling

by a

simple readjustment of the thermostat. The results can be quite dramatic.

High cycling rates are generally achieved by circulating large velumes
of air. High volume air circulation is not too expensive and offers several
advantages. A greater volume of air requirQs a smaller temperature dlffercnce
between the inlet and outlet to maintain the same room average. This is simply
a matter of removing the same number of Btus with more pounds of air at a
smaller temperature difference. Another advantage of air volume is that the
increased velocity tends to scrub the whole gaging system and remove the hea~
that may be coming from external point sources of heat such as motors, lights,
people, and radiation from the sun. Stated more exactly, the increased air
velocity increases the convectiveheat transfer coefficientand decreases the
thermal resistance between the gage and the room air, which is the thing that
is being controlled.Still another advantageof high air flow is increased
operator comfort. The decreased difference between inlet and outlet air
temperatures means fewer cold drafts which are the real source of discomfort.
The benefits of high air flow, high cycling rates, and close containment
of sensitive equipment have recently been demonstrated at LRL. A new rotary
contour gage has ~ust gone into service which is completely enclosed in a
plexiglass box. Air is admitted through a plenum chamber at the top and leaves
through a plenum chamber at the bottom. The circulation rate is one complete
change of air every 3 seconds. The cycling rate is 2B cycles per hour. The room
temperature variation is 0.7 degrees, but a 24-hour drift check shows less
than 3 ~in. of drift~
The problem of standardization of room air temperature measurement is
illustra~edby the differentvalues obtainedon. this system with three differentways
of measuring. High sensitivity mercury thermometers show less than 0.05 degree

-27vari:uion.
degree

The thermister
vnriatlon

An Automatic

recorder-controller

and a high
Error

frequency

Correctin~

response

for

the

enclosure

thermo~raph

shows a 0.4

shows 0.7

de~ree.

Device

In reviewing our experiences with computing ~hermal drift from k~owlod~

of system frequency response and measurement of temperature variation,
J. W. Routh, of LRL, suggested that we consider the possibility of automatic
error correction. Preliminary investigationof ~his idea has convinced us
that it should be possible to design a thermal model of the system that can
sense the room temperatureand provide an electricaloutput equal to the drift.
This output can be used to zero shift the coordinatesystem of the gage and
provide direct, on line, compensation for ~hermal error. As it is now
visualized,~his device would be completelyautomatic once set for the specified
part to be measured. The operationat settings required would be nominal coefficien~ of expansion,time constant, and size of the part. The response of
the gage wouid be built into the device. Someadjustment might be req~red
for different setups that might be enco~tered.If ~he time constan~ of the
master could be tailored to match the gage, the bulk of the thermai error could
be eliminated. Error due to uncertainty of nominai differential expansion woutd
still remain.
~le this manuscript was being prepared, a report of a feasibility study
by a graduate student at the University of California, Berkeley [12) concerning
~he practicabiiity of such a device became available. This study was made at
our request and has shown that:
(i) A simple analog, consisting of on.ly two time-constantelements in
parallel, provides an adequate model of the system.
(2) The main problem enco~tered in constructing the compensating
device was in finding practical, Iong-~ime-constantelements.

-28This device, if realizable,will have far-reachingeffects on the use and

design of machine tools and measuring machines.

VI.

CONCLUSIONS

In conclusiontl%e following gcncrallzodapproach to the problem of

thermal effects in dimensional metroloEy is suggested:
A.

Evaluate existin~ conditions to determine whether or not a problem

exists. This,is acclo~nplishedby substitutingexistin~conditions
into the Evaluation Plan. If the thermal error index is more than
i0~o of the part tolerance,it is likely that a problemdoes exist.

B.

Review the workinE tolerances to be sure they are economically

and functionallyrealistic.

C.

If necessary, take corrective action to reduce the thermal error

index as follows:
To reduce error resulting from average temperatures other than

I.

iV~ake correctionsfor nominal differential.expansion.

9.. Establish more accurate coefficientsof expansion so as to

increasethe accuracy,of the corrections.
3. lVIinimize average temperature deviations from 68F.
To reduce error resulting from temperature variation:
I.

Improve procedures for soaking out workpieces and masters

so they are in thermal equilibritunwith the environment.
Shorten the mastering cycle time if indicated by the Evaluation
Plan.

3.

Increaserate of air flow and improve its distribution.

4.

Increase the frequency of temperature variation.

-29Dccrease the amplitude

of temperature

Redo.~ign ma~ters and temperature

coefficients
parts
"

sO thoi~

of expansion are in better

time constants

and

balance with those of tl~e

to be measured.
APPENDIX 2k:

Part or Work]~iece:
physical

variation.

object

GLOSSARYOF TERMS

In every length

for which a linear

determination

..

process,

there

dimension is to be determined.

is some
This

object is called the part or workp.i.ece.

Master:

In the length

of the part

measuring

process,

the unknown or desired

is compared with a known length

may be the wavelength of light,

the length

called

the master.

of a gage block, line

dimension

This length
standard,

lead screw, etc.

Comparatoz::

Any device

master is called
Mastering:

used to perform the comparison

a compar~/~or.

The action

of the part

and

{!

o~ n,ulling

~ comparator

with a master is called

mastering,
Mastering.
process

Cycle Time: The time between successive

is called

the mastering

Measurement Process:
composed is called
Measurement

System.

ment is called

(and sinks:)
all

the measurement
The entire

of which.~a measurement is .

process.

apparatus

..

used in making a measure-

system.

Any!physical

object

of heat energy which influence

such sources

of the

time of the process.

All of the activities

the measurement
0

Thermal .Environment:

cy.cle

master,rigs

and sinks form the thermal

is expose.d

to various

its thermal state.

environment

sources

Taken in tote

of the object."

In

-30-

the laboratoryor shop, the thermal enviro.nmen~_..of

any object can c6nsist
of all other objects with which the object is in ~hermai communication,
i.e., by convection,conduction,and radiation. Sources and sinks commonly
found in the laboratoryare:
Convection sources and sinks:
~ system and
Air atmosphere,~includingthe air-conditi~nin~
distributionor flow of the air. The air constitutesthe medium
of convectionheat transfer.
2) Radiant sources:
a) Sun (if ~vindQws exist)
b) Walls,floor..,:,and ceiling
c) Illuminating.lights
d) Electric motors
e) People
3)

Conductivesources are usually the most olbvio.~s.

, and include
all objectsin directcontact.

In this sense, then, an obje~t in an air-conditionedroom is in thermal

communication with the air-conditioner by, usually, convection. It may
also be in communicationwith an electric motor by convection, conduction,
and radiation.
Although,in the general case, it is probable that all types of thermal
communication exist between the environment and a given object, perhaps
the most common environment is the one in which the only significant
communicationis by convecion.In this case, the effect of the environment on the object can be describedin terms of thermal state of the
volume of air surroundingtl~e object.

Convective
When all

environment,
environmental

and a single

temperature
is.called

of an object

(changes in length)

correl~.ted

Environment
e.

environment

8(e)

in nature

the environment,

a convective

environment.

the
The response

in such an environment

can

with the environment" ten~perature,

temperature,

The temperature

are convective

describes

environment

directly

8(b)

influences

by which the thermal

is measured is called

sta~e of a convective

the environment

temperatuz~e.

T__sm_m__perature
offset.
The difference

between the time average,

of th.e

temperature .and 68 F is called .t.e,mpera:ture

environment

offset.

TO = Te - 68
Variations
9(a)

of Thermal Environments
environment.

!:

When the envigonment is invariable

in time, it

Stationary

stationary.
9(b)

Periodic

environment.

:.

in which every variable

manner .is called a periodic

Aperiodic
(1)

is called

..

An environment

9(c)

[!

environmen,t..
!:

environment.

Transient

well-defined

a transient

change is not periodic

pattern,

of temperature

..

environment:

When the environment

changes in a cyclic

such as a constant:rate

in a convective

environment.

envi.~onment,

but has a
of increase
it is called
""

-32(2)

Random environment.
Whe.u the
called

"

environment

changes

a random environment.

presence

of human beings

Although

all

environments

in a random manner,
Influences

or weather

tend

it

is

due to the
to be random.

have some random characteris.-

tics, deliberateattemptsat envirorfinen~lcontrol, e.g.,

by refrigerantair conditioning,tend to introduce dominant
periodic characteristics. Also, in uncontrolled environments
transient characteristics may be found to dominate. For
example, t.he outside air temperature may dominate in a
room which is well ventilated.

i00 Standard Temperature~f__or, i.L.engthMeasurements:Unless otherwise

specified,the dimensionsof an object given in drawings or specifications
shall be for an object with a uniform temperature of. 68F (20 C).
length of an object at stand:ardtemperatureis called:thestandardlength of
the object. This procedure follows the April 1931, resolution of the
International Committee of Weights and Measures that the temperature of
20 C (68 F) sh6uld be universally adopted as the normal temperature of
adjustment for all industrial standards of length. Also, Recommendation
No. 1 of the InternationalOrganizationfor Standardization,issued in 1954,
promulgates the standard temperature of 20 C among the 40 participating
countrieS.
ii. _T...emperatures
of a Bqd_y_
ll(a) Temperature(a.t a poi.nt).
When discussin~ a body which does not have a single uniform
temperature,it is necessary to refer in.~ome manner to the
distribution of temperature throughout the body. Temperature

-33at a point in a body is assumed to be th@ temperature

very small volume of the body centeredat

t~at

of a

point.

The

continuum.
ll(b)

The temD~ratur@

9 f abody.

When the differences

between the temperatures

in a body are negligible,

temperature.

at all

point~

the body is said to be at a ~iform

This temperature

is then the temperat~e

of the

body.
11(c) Instantaneous

a~verage temperature

of a ~ody~

Whenthe body .~s not at a uniform temperature

but it is desirable
a single
total

to identify

temperature,

heat storq~

homogeneous this

ll(d)

of a body.

the time-mean t.emperature

12. Soak Out: One of the characteristics

experienced,
to another,

such as occuzs
there

" forgets"

about its

only on its

current

environment

until

is called

will

soak-out

the object.is

of the body. :

when a complete
when an object

change

is transported

and exhibits

The time elapsed

influenced

it has.a

in enviro~ent

of time before~he

enviro~ent

environment.

of a body is called

of a thermal system is that

be some period
previous

over the vol~e

This,iS called the

The time average of the average temperature

words,

the

of the body.

Time-mean temperature

" memory." In other

of the body by

~en the body is

is the ave.rag.,

point temperatures.

temperature

points,

which represents

in the body may be used.

of the body, of.all

average

the thermal state

the temperature

temperature

at all

is

from one room

object

completely

a response

dependent

from .a change in

only by the new environment

time. After " soak out"the

objec}~is

said to be in

-34equilib,-ium with the new environment. In cases where an

environment is time-variantthe response of the object is also a
variable in time, "vVhen the object exhibits a response dependent only
on the environmentit is said to be in dynamic equilibriumwith its
environment.
13, TAme Constant of a Body: Tile tlmo constant of a body Is a measure of the
response of the body %o envi, ronmental temperature ehan~es. It is defined
as the time requiredfor a body to achieve ~3.2~0 of its total chan~e after
a sudden step chan~e in the environment,
14..~req%lency Res~5onse: The frequency ~esp0nse of a,measurement systsm
is definedas the ratio of the amplitudeof the drift in microinches%o the
amplitude of a sinusoidal environment temperature oscillation in de~rees
Fahrenheitfor all frequenciesof temperatureoscillation.
15. Thermal Expansion: The difference between the len~h of a body at one
temperat,,ureand its length at anothertemperatureis .calledthe thermal
:

expansionof the body.

16. Coefficient

of Expansion,
The true coefficient
of.expansion,
~, at a temperat.u,
re., t_, of
of a body is the! rate of change of length of the. body with respect
to temperature,at the given temperaturedivided by the length
at the given temperature.
I dL
~ = L dt
The average true coefficientof expansion of a body over the
range of temperaturesfrom S8F to t is defined as the ratio
of the fractionalchange of length of the body to the change in
temperature, .

-35-

Fractionalchnnge of length is based on the length of the body

L- L68
a68, t - L68 (t - 68)
Hereinafter

the term ~ coefficient

,,

the
17,

average

Nominal Coefficient

value

over

of expansion.

.,,

the

shall

,~

range

refer

from 68 =F to another

of Expansion: The e~9}ir~.a.te....of....t..h

expansionof a bodysha}! ,bg. ca!l..ed .th.e..no.~i.nal_.

t:

temperature,

e coefficient

c.oefficient.

To distinguish
this value from the average true
)
(K88 t it shall be denoted by the symbol K,

only to

:.:

of

_. of expansion,

coefficient

Of expansion

18.

Uncertainty
percentage

of Nominal coefficient
difference

between

of Expansion:

the.actual

The maximum possible

coefficient

of expansion,

a, and

the nominal coefficient

of e~,pansi~n shall be denoted by the s~..y~.b.o~..~,
expressedas... ~ ..p~.r..c.e..nt.age of the t.r..U.e. 9o.effic~.n..t o.f..exp.an,~i0.n,

-K
Variations

in material

as well as .inherent
orientation

"
composition,

anisotro,

cause objects

different Shermal

and

formin,

pic properties

of suppo,s.edly

expansion.lcharacteristics,

g processes,
and effects

identical

and heat

treatment

of preferred

composition

to exhibit

Also, d~fferences

in ex-

perimental technique cause, disagreement among thermalexpansion measurements. As a result, it is difficul~,solely from published information,to
obtain an exact coefficientof expansi.onfor any given:pbjec.t.
This value like that of K itse,lf must be an estimate. Various methods can
be used to make this estimate,

For e,.~..a....mp.!e.:

18(a) The estimate may be based on the dispersion fDund among

published data. .

-3618(b) The estimate may be based on the dispersion found among

results of actual experiments conducted on a number of like
objects.
Of the two possibilitiesgiven above, (b) is the recommended
procedure.
Because the effects of inaccuracyof the estimate of the
uncertaintyare of second order, it is consideredsufficient
that good judgment be used.
19. Nominal Expansion: The estimate of the expansion of an object from 68F
to its time-mean ternperatu~-eat the time of the measurement shallbe
called the nominal expansion and it shall be determined from the following
relationship.
NE = L(t -.68)(I~)
20..Uncertainty of Nominal Expansion: The maximum difference between the
true thermal expansion and.the nomiflal expansion is ,called the uncertainty
of nominal expansion. It is determined from:

21.

Differenti"al

Expansion~

Differential

expansion

is def.ined

as the difference

between the expansion of the part from 68 F to its time-mean temperature

at the time of the measurement and the expansion of the master from 68 F
to its time-mean temperature at the time of the measurement.
22. Nominal Differential Expansion: The difference between the nominal
expansionof the part and of the master is called the inominaldifferential
expansion.
NDE -- (NE)part - (NE)master

-37,,3. l~jnce,,tainty
of NominalDifferential
Expan.s.ion.:The sum of the uncertaintics of nominal expansion of the part an~l ma~er is calle~
uncertaintyof nominal differentialexpansion.
UNDE = (UNE)I~,~
24, Thermal D~ift.

25.

Drift

is de~ned a~: the differential

or f}~e

master

and the

in fl~e

thermal

environment,

Electronics

Drift

when it

" pure"
head,

the drift

etc.

the ~ransducer

involves

the output

of the drift

making a transducer

o~ cartridge...

a small

against

the inside

similar

devices,

in a direct

of.a

cap ove.r
of the

~manner so that

~rift

over a perio~

indicate

cartridge-f~e

Finger

Drift

Check...

by blocking

Blocking

on its

a transducer

own frame,

gage head,

this

as

base,

is accomplished

so the plunger

registers
with

to see fhaffhe

b~locking

of temperature

on the blocking

fh~ influence

is

done
de-

..
conducted

in a measurement

system

normal

check.

the

usual

the

correlation

28) involves

on the sensing

t~e gage heads can be blocked

exercised

An experiment

Since

~s the s~ of the

of time at least

vice is negIigible.
26,

drift

being

performed

fo be performed..

the end of the cartridge

cap.,

Care must:be

to that

check is

the

and recorders

of the enviro~ent

effectively

In the case

by mounting

test

by time-variations

amplifiers
similar

drift

of the part..

to determine

The electronics

The ~lectronics

and observin8

long as the duration

associated

an~ ~ the effect

drift

caused

conducted

environment

check itself.

electronics

amplifier,

and its

fo a thermal

~9ve~;enf

in microinches

An experiment

transducer

is subjected

evaluatedby

comparafor

Check:

in a displacement

+ (UNE)~

~der

to determine
operating

method of monitorin~

the

t~e a~tual

conditions

is

envinonment

of one or more temperatune

drift

inherent

called
(see

recordings

a drift

definition
with

drift, a drill cheek will usually consist of simultaneousrecordingsof

drift and cnviromnental temperatures. The recommended procedure for
the conduct
27.

of a drift

.Temperat.urc

check is given in Appendix B.

Variation

measurement

error

Err~p~i

induced

.TVE: An emtimate

solely

b~ deviation

of the

of the

maximum possible

environment

from

average conditions is called the tem..__P..SAa_tDr_e_y_a.._ria_dg..n...e__rror.

determined

from the results

comparator,
For zero

and the
or small

of two drift

other

of the

mastering

part

cyele

For significant

mastering

cycle

Use whichever

of above

Total

Error:

Thermal

usually

29.

error

temperature

the total

thermal

Index:

It

does not establish

and penalties
which

the

error

is

defined

as the

maximum possible

estimate

conditions
the actual

to processes

not even possible.

Therefore,

below.

The evaluation

existence

may be " good"

and in many cases,

is outlined

more than

environment

doubt about

greater.

thermal

to do,

Thermal

thermal

for chosen
time

from temperatures
other than a uniform,
conof exactly 68 F. It is, of course,
~desirable
to determine
error induced in any measurement. :~ However, this is

procedure

nothing

Part-to-master
drift error as above l

resulting

not practical

Error

TVE is
Total

an alternative

does

time.

times.

i mastering
ever is greater,
c:ycle

stant

and

comparator.

~ Master-cor~para~or
drift
or
~ part-comparator
drift,
which-i

TVE =

measurement

and the

on~ of the master

Maximum .e.xcursion
of part-comparator
and mastercomparator
drift
curves when curve~ are aligned
for inphase temperature
conditions
over representative
time
period.
(Part-to-master
drift)

TVE =

28,

checks;

TVEis

technique
the

maximum possible

affecting

a particular

magnitude

of ~he errors

of any error.

and to establish

which are combinations

and some "bad."

proposed

in this
error

section

caused

measurement
It

serves

a system

:,
of ]technzques,

by
process.

to remove
of rewards
some.of

The The1"mal Error Index shall apply only so long as conditions do

not change,
Th~ proposed plan consists o~
Computing the nominal differential expansion, NDE.
In this computation(and in %he next), the temperatureoffse~
is assumedto be the averagedifference...betw:een
68=i? and
the air temperaturein %he vicinity of the process over ~he
mastering cycle of ~he process.
Computing the uncertainty of

NbE,UNDE.

Determining the thermal variation error,, TVE, by means

of a drift check.

(4).

Summing ~he absolute values obtained in i, 2, and 3 to obtain

an index relatedto the qualityof.the process, yields the
~emperature error index.
TEl = NDE + UNDE + TVE

(5)

If an effort is made to correct the meas~remen~by computing

the NDE, partl is to be deleted.

The plan penalizes a measJrement process on two counts:

(i)

Existenceof er~vironmenttemperatureo~fset~ resultingin

differential
exl~ansion.

(2) Existence of environment variations.

"

The plan rewards good technique by reducing the thermal error index for:
(I) Attemptinga. correctionfor differential.expansion.
(2) Keeping envirof~mental variations to a minimum.
Thermal error index can be used as an adminis~ratiyetool for certification
of measurementprocesses. ,as is discussed in the next section. It can also
be used as an absolute index of accep~abili%yof the ~roce.ss. For example,
a good rule of thumb for establishing the accep~abil~y ofa measurement

-40process
error

with respect

to ti]ermal

errors

is to limit

index to l O~/0 of the working tolerance

30. Monilorin~

To perpetuate

the

to monitor

the process

conditions

are recognizable.

The recoi/amended

thermal

procedu~e

thermal

of the part.

error

in such a way that

the acceptable

index

it

significant

is to establish

will

be necessary

changes

in operating

a parti~ular"temperature

recording station which has a demonstrable correlation with the magnitude

of the drift. In a " convectiveenvironment" this could simply be the
" environment temperature."

The temperature of the selected station should be r~corded continuously

during any measurement piocess to which the index~ls to be applied. If
the temperatureshows a significantchange of condlt~ons,the index is
null and void for that process, and a reevaluationshould be accomplished,
or the conditionscorrected to those for which the index applies.
In additionto continuousmonitoringof environmentalconditions,it is
recommended that efforts 5e made to establish that ~he process is
properly soaked out.- This may be done by checking the temperature of
all elements before and after the execution of the measurements.
APPENDIXB:

DRIFT-CHECK

PROCEDUI~E

The following is the recommended procedure for the conduci of a drift

check for a process in which the proposed monitoringm.e, thod is based on the
measurement of environment temperatures.
A.

Equipme,~t
The major

equipment

necessary

includes

very

sensitive

displacement

transducers and sensitive, drift-free temperature s~nsorswith associated

-41,a~p~1~lers and recorders. A linear variable differential transformer

with provision for r~corder output has proven quite successful. Also,
variou~
proven

resistance-bulb
successful

thermometers

as temperature

with

rec~din~

monitoring

provi~i~

devices.

The required sensitivity of the displacement transducers used may be

,

adjusted according to the r~ted accuracy of the meas~rer~ent system.

B.

~ment

Testing

The temperature
tested

for

measuring

accuracy

of sensitivities
.elements
Before
they

the

of calibration,

30 sec az:e

displacement

should

be calibrated

and observing

the output

volves

of the drift

Check" should

making a transduce~

for

drift

of the

An essential

feature

possible.

duplicate

to be performed.
effectively

"Therefore,

timing
actions

process

System for

apparatus

in the

indicate

used

environment.

by blocking

"Blocking"

are

An

the transducer

as long as the
a transducer

on its

ov~ frame,

conditions

during

inbase,

Test

of the drift

check is that

conditions
" normal" for

the

must be determined.
subject

of sensing

Preparation

must

The availability

of time at least

or cartridge.
C.

be thoroughly

Time constants

be performed

over a period

test

and drift.

and associated

and checked

drift

should

recommended.

transducers

" electron{cs

duration

respbnse

0.1 F is desirable.

of at least

of about

~ apparatus

and recording

before

the check is started,

The gctual

step-by-step

must be followed

in the drift

check.

the.... process

in the

inmastering

importaht
and all

and with
in terms

:lSreliminary

as

conditions

proce~lurefollowed

This is especially

of any human opera~ors

as closely

"..normal"

same sequence

the check

in the
the

same

of the
setup

steps.

-42With as little deviation from normal procedure as possible, the displacement 1.ransducers should be introduced between the part {or master,
depe~d~g on the type
that

it

measures

subjcc~

of dz.~t

relatively

measurement

The temperature
which is

placed

to measure

measurement

pickup

with

the

case,

In situations

pattern

some period

of time

a 7-day

work week is

observed

then

during

the first

Postcheck
After

duration

either

of the

a temperature

and error

may be

may have
elements

to be
of the

fs

Check
to continue

from " normal"

of activity
which

in the

area,

events

and each

recommended.
cycle

operating

is observed
most

If

should

as long as
conditions.

its

duration

are

repeated.

should

day is much like

a 5-day
be used

work week is
or checks

the drift

test,

days of the week.

apparatus

Evaluation

of the

Following

the drift

the displacement
should

Drift
check,

~ransducers

and the

Check (Drift-Check
the data

should

temperature

"

be recalibrated.
Report)_
be assessed

for

When
any
ob-

performed

Procedure

recording"

values.

~easdre

pickups

be allowed

during

a full-week

and last

to

of the active

For a Drif~

check should

where a set

served,

of all

a minimum of deviation

a 24-hour

of action

be over

other,

of the ~-f~ame ~uah

line

Some trial

temperature

temperature

Time Period

with

drift.

..

up the drift

possible,

~ the

must be placed

the

extreme

loop.

Representative
Once set

displacement

sensing

In the

and the rest

process.

correiatable

necessary.

checl~

~he following

-43(a)

Nonperiodic

Effects

disappear
should

with

For zero

elapsed

be described

compensated
(b)

- the effects

These

and the portion

by soak out should

Temperature
or small

time.

of the operator

Variation

masterii~g

Error

cycI~

and similar
of this

be included
(TVE)

to

effects

error

not

in the

TVE .

":

time.

tend

"

[Maximum ex.cursion
of part-comparator
and master-,
I comparator drift
curves when curves are aligned
for
TVE -- ]~ in-phase temperature conditions over representative
time period.
(Part-to-master
drift)

For significant

TVE

mastering

cycle

times.

i Master-co ,mparator drift or

~ part-comparator
drift,
which= I
~ ever is gre. ater, for chosen
mastering cycle time

Use whichever
A complete
Thermal
Items

of above

report

Drift-Check

of the drift,
Reports

in parenthesis

are

under a heading.
I.

Tv:E is

!]Part-to-master

drift

error

as above

greater.
check findings

should.inclu.de

the following:

0htline
s~gested

as a guide

to what might

be pertinent

.:

Description of System

a)

Identification
(Mfgs.,model, . {pertinentspecifications,
.and dimensions)

b)

Component Mo{ions
(Active

c)

element.s,

Operati6ns

linesof action)

1) Type of op6ration

-442)

Typical

workpiece

Sizes
Materials
Minimum tolerances
3) Method of n~asterlng
.. 4) Cycle tima~
(Operating, mastering)
Environment Description

a)

Room Features
(Size; solar exposure;exits; wall, floor, ceiling,and other
heat sources)

b)

System Features
I)

Location with respect to " room features"

2)

Internal heat sources

(Motors,l.a.mps,electronics)

c)

Air Circulation
(Inlet-outletlocations,sizes, numbers, drafts, air volume
circulated)

d)

Temperature monitoring and control

Test Apparatus Description

a)

Temperature monitoring
(Identification,
response, sensitivity,
lo~ation)

b)

Displacementm..onitoring
(Identification,
response, sensitivity,location)

Procedure
a)

Stepwise descriptionof testing

-4"3-

Results
(Displacement-temperature vs time graphs; ~naximum displacements
and temperature variations; cycle times; causes if known)
TVE
7.

R, ecommendatlons

APPENDIX

C:
I~ESPONSE

A METHOD

FOR DETERMINING

OF A MEASUREMENT

FREQUENCY

SYSTEM

Data obtained from step-change experiments performed on the 15 inch

Sheffield rotary contour gige gave: (i) an indicationthat the.system was
linear with respect to thermal variations,and (2) the set of data correlating
temperaturevariationwith drift.
The basic characteristicof a linear system is that the output (in this
case, drift) c.orrespondingto any input (temperaturevariation)is the
of outputscorrespondin~to the., componentsof the input,:,:.
This characteristicpermits the use of the data of Fig. 71 in computing
a predicted drift for a variation of ~emperature in the ~nvironment of
the Sheffieldgage as follows:.

.~

Suppose ~that the temperat.uTevariation isrecorded~asshown in Fig. C-i.

This record can be approximated by straight lines over 7-minute increments.
This procedure decomposes th.e temperature variation into a series of
componentssimilar to that of F~ig. 7.
The drift correspondingt0. any one of these components, can be determined
by scalingth.e data of Fig. 7.
Figure Q-2 shows the resulting set of drift components and their sums
which is an approximationto the drift caused by the temperature variation

-48This
easily

computation

and quickly

8 nnd I0 of the
tl~e

Lawrence
There

computations

procedure

done by a digital

paper

computer,

kind:

when done by hand but is

The results

shown in Figs,

an IBM 7094 digital

computer

at

Laboratory.

two practical

of this

cumbersome

were computedusing

Radiation

are

is

considerations

to be observed

in making

It must be possible to make an accurate approximationto the

temperature ~ata. For example, in the case described above,
a temper,~turevariation consistingof a sine-wave with a period
of less than ? minutes cannot be approximated.bythe data
available.
(2) Because of the " memory" of the system, the computed drift is
in error until a period of time equal to the soak-outtime has
elapsed. For example, in computin~ the data.for Fig. 8, the
drift computedfor the first 12 hours of ~emperatur.evariation
record was inaccdrate and was omitted.
ACKNOWLEDGMENTS
The authors

wish to express

special

gratitude

to Mr. I.

H. Fullmer

of NBS for his aid in assembling bibliographicalmaterial, and to Mr. T. R.

Young of NBS for suggestions concerning the uncertainties of thermal expansion properties.

-47BIBLIOGRAPHY
1.

Ch. Cochet,

"On th~ Choice of a Uniform Temperature

Calibration

of Measuring Instruments

de temperature

pour letalonnage

Gen. de lElectricite,

for ~he

(Sur le choix .~lun

des instruments

degre uniforme

d~ mesure),"

Vol. 4, Nov. 16, 1918, pp. 740-2.

Commission de Normalisa{~on de~ Ingenieurs

Boulogne-sur-Seine

Rev.

Report of

des A~Tts e{ Metiers de

recommending adoption

of 0 degree

centigrade

as

standard.
2.

A.Pcrard,

" Temperature

temperature

of Adjustment

dadjustage

des calibres

of Industrial

industriels),"

June25,1927,.pp.621-4.

Genie Civil,

necessity
3.

discusses

req.~isite

for standardization

L. Graux, " Temperature

temperature

dadjustage

des pieces

mecaniques),"

pp. 250-2,

Discusses

precision

~djustment

Vol...

on basis

" Temperature

(La temperature

de definition

of standard

m..~eter

degree c~ntigrade as temp..rature

necessity of unification

des calibres

,:

ef In,dustrial

Gages

industrie~s),",Rev.

Gen.

article

made by German C:omm. ission

of standardization;:,

0f t~.mperature in all countrie..s

which have arisen,

author

gives some reasons

stating

effect

Qerman decision

tr.oublesome

at 0 degree

of Standardization

of de.cision

et

91, Sept, 10, 1927,

Vol. 23, Jan~, 14, 1928, pp. 73-6. Object .el

emphasize importance

(La

des ver.ificateurs

..

G.E. Guillaume,

1,Electric%te,

industriels,

Genie Civil,

and

of adjus.tme,nt.

Of Length Stan:.dard~

des etalons

adjustment

of 20 degree

for indus.t.rial;gages

of temperature

or at +20 degree centigrades:

4.

Vol. 90,

~,Plea
to German
Industri:,al
Standards.

Committee to readopt adjus~tment of 0 degree insteac~

centrigrade;

Gages

after

is to
fixing

stating

an, d difficulties

which.have,

20

moved hin~ ih

can have :on d~sired unificat..on.

-485.

A. Petard.

" Concerning

Gages (A propos

Standardization

de la temperature

Temperature

de definition

of Industrial

des calibres

I "~

industriels),
~

Discussion of reasons for not considering temperature of 20 degree

centigrade as temperature for standardization~ather than 0 degree
ccntigrade~takes into account coefficientof expansion..
6.

P. Grand, "Contributiona la metropole industrielle des longueurs,"

Rev. Gen. Mecanique, Vol. 36, Nos. 41,42, 45, May 1952, pp. i35-40:
Jm~e, pp. 167-70; Sept., pp. 293-6. Contribution to industrial measurements of length, with reference to parts of large dimension; influence
of ambient temperature in ~orkshops; permissible deviation between
temperature of gage and object being measured~ measurement of steel
par~s with steel and wooden gages, and light alloy p~rts with steel,
light alloy, and wooden gages: diagrams.

7.

.:

.~

P. Grand, " Influencede la.temperaturesur letalon:nage,des jauges,"

Rev. Gen. de Mecanique, Vol. 39, Nos. 74, 79, 83, :84, Feb. 1855,
pp. 39-4~; July, pp. 247-54; Nov., pp. 395-400~ Dec.., pp. 445-50~
Voi. 40, Nos. 86, 88, Feb., :.1956,.pp.81-7; Apr., pp. 15.9-64. Influence
of temperatureon calibrati.enof gages, environmentaltemperature in
measuring shop: temperature of materials subjected.to d~fferent
environments; calibration on measuring machine: comparator calibration:
9
errors of calibration due to temperature variations. (UCRL Translation-

9This reference%vas not receiveduntil after the completio

~ of this paper.
~r. Grand has covered much of the same material presented.here. His
approach to the problem is different, however, and the doplidationsdo not
justify~ithdrawalof this paper..

-40-

R. Noah and I.I. Huhn, " Investigaiions of the influence of temperature

variations

on

magnification,"
in G(~rman.

~h~

r~din~

of

Z. Instr~kde,

~e~h~ni~l

~ ~ompa~

wi~h

l~z

m~h~ni~l

Vol. 67, No. ii, Nov., 1959, pp. 285-8.

~omp~ra~o~s

with high ma~nifia~i~n

for ~ine~r

measurements can alter their readings due to the influenoe of small

changes of temperature. The paper deals with the influence of a slight
heat source, for example,

an operator or a lamp, on comparators

with

a ~~aduation of 0.0001 ram.

A. Goldsmith,

T. E. Waterman , and H. J. Hitschhotn,

ThermoPhysical

Ptoperties..of

i0.

D. K. Cheng, Analysis

ii.

I.. H. Fullmer,

Solid Materials~

of Linear Systems,

" Comments

on Various

National Bureau of Standards,

Handbook

Pergamon.

Addison-Wesley,

Temperature

N.Y., 19Sl.
Redding,

1961.

Combinations,"

1960, 4 pages. Give~historical

ground of. three standard te~pera~gres;

of

back-

discusses importance of knowin~

coefficieD% of thermal expansion ~cutately, and disq~sse.@ proposed

chan~e of standard temperature
12.

Evan

L. S. Lmn,

Measurin~

" Dynamic

Machine Errors,."

to.28 oC. (Not published.)

Corrections

of Thermally

Induced

(M. E. 209 report to Prof. H. Thal-Lars4n)

Department of Mechanical Epgineering, University o~ California,

Berkeley, January 1965.

~,

-50-

120

]--

"_.

MEASUREDDRIFT
I00

69.00

69.25

12

18
TIME (hours)

OUr ER CONTOUR
MEASUREMENT
(ROTARYCONTOUR
GAGE
No,I )

24

30

36

TEMP.
I

,.,,,
v

u= 69.50

f,,,/"\

~.- 69.7,5

70.00

-0

Fi.g, 1 .l.hermal.ly
No, ] with steel

12

induc,.ed

18
TIME (hours)

drift

of 15-inch Sheffield

30

36

~"otary co)~tour gage

MPLIFIER

RECORDER

DISPLACEMENT
TRANSDUCER

HEAVY C FRAME
THIN
WALL
TUBE

:~,ig.

2.

Schematic

of C-frame

comparator

and part.

-52-

~.

DISPI_ACEMENT
~

i,1<//////-/-/-/"~,,~
c~~ NDER_
A~,
" Iili~,

TRANS

DUCER

) il~zz~i!il

", l,i CYLI

NDER~
B

" C FRAME
-OF
I- !,
~

) I I
~-------~-.,

Ii
"---1: I

SURFACE

) ~.:

, =,~
l,

I"~ ~[n.DIA-----~

Iii
I

;ll .

,,

~" | 17-in~2
DIAM

MODEL OF "C"FRAM E
COMPARATOR AND PART
.l?iig.

3. Model. of (]--frame comparatot and pa).t.

~irl,

L I N DERIIKI AND CYLI N DERIIBII

/" CYLINDER
~,,,~~% ~D1FFERENCE
BETWEEN DISPLACEMENT
L~J

__./>----A.

TIME

CYLINDER
DISPLACEMENT

FIGURE

ENVIRONMENT
TEMPERATURE
+IOF

~ TAV

I HOUR

-IOF

I,:i.~. 4. InclJvitlLm]. and net (|zflt

.~hown
.in I,i.R, :~.

thermalresl.)Or~sp. (~" two-t;](-;m(mt lu()de].

ANALOG COMPUTER
RESULTS
TWOBODIES BOTH Z~in.LONG
AT 68 F
TIME CONSTANTS:.125 brAND.281 hr
INPUT AMPLITUDE =lF

20

15

I0
0

I i I ! !

INPUT FREQUENCY (CYCLES/HOUR)

Fig. 5.

Computed

frequency

response

of C-frame

comoarator

and part.

STEP

RESPONSE

20
z.

80

TIME

Fig. 7. Step input

response

on 15-inch

IN

HOURS

Sheffield

rotary

contour

gage No. 2.

-ACTUAL

DRIFT

/fCOMPUTED

81

1~.~-40~
0

DRIFT

zl

5 6
7
TIME (HOURS)

I0

II

12,

TEMPERATURE VARIATION
INDUCED BY MOTORDRIVEN CAM
ON SET POINT OF CONTROLLER

Fig. 8. Forced drift

check on 15-inch

Sheffield

rotary

gage No. 2.

"- iO0 --

z 80-0

!
|.--

4.0,

20-I ! I ! I!,
!
i ~ ~ I__J.Jj
0.1
1.0
FREQU
~- rt.C,,_
.
._" . .,,,-,
C ~t.,L.ES/hi"
Computed +..~"eque,qcy

response
of SheFfield

li

ga,Re

OUTER CONTOUR MEASUREMENT

~ ROTARYCONTOUR
GAGENo.l )
69.00
t

----~.$9.25

-4 69.50t,-...

--169.75

- 7o.oo

0,

i2

!$
TI i.i" C- ~.r hours)

24.

50

L._j 70. 25
5,,

Fig. i0. Conupuzedand ac~ual drifi of She:~ie;drotary contourgage ho. ! ~i~ s%eei part.

-60-

PART

LENGTH

AT.___~8~

lri~, 1.1. o I~art-to-masteP dPi..~t using in-phase, sinusoi.dal,

eu~ves,

-61-

PART-TO-MASTER

MAXIMUM ERROR

MAXIMUM COMPARATOR DRIFT

FROM PART OR MASTER
DURING MASTERING CYCLE
A- B = APPROXI MATE TVE FOR
MASTERING CYCLE 8HOW.N

MASTERING
TIME
I
I
~F""

PART

MA_S__TE
R ~"k

COM PARATOR/

DISPLACED
COMPARATOR

~gl ~
I X
1 =
I
,H X --MASTERING CYCLE

(~I+~I)-B~XI|FOR

TRUE ERROR FOR

MASTERINGCYCLE -#1
TRUE ERROR FOR
MASTERING CYCLE

S!NUSOIDALCURVES
SHOWN.

OCALCULATED APPROXIMATE ERROR (A-B}

(~,2+~,2)-B~X2~J
IS SOMEWHAT
LESS THAN TRUE ERROR(X)

.l?i.g.

]2. Drift error for two mastering cycle times.

-62B = PART-TO-MASTER MAXIMUM ERROR

A+A-- MAXIMUM COMPARATORDRIFT
FROM MASTER
XI ;TRUE ERRORIF PART IS
INDICATED AT MASTERING"rIME
X
2-"
A-B<B

TRUE ERRORIF PART IS INDICATED

AT END OF MASTERING CYCLE TIME
FOR EXAMPLE SHOWNREGARDLESS
OF MASTERING CYCLE TIME

MASTERING
TIME
COM PARATOR I

-~--~~~

1
~.--.[
MASTERING CYCLE
TIME

z~AT

68

DISPLACED
COM PARATOR

Fig. 13. Dlift error when comparator drift is betwt:en mast(,~ and pa.rl.

-63-

28
42
TIME (minutes)
Fig.

C-I.

Temperatule

variation

56

70

84

and approximation.

PREDICTED DRIFT

/ _ ~-,----.- __~ ...~-..._~....~._.,.../_~

_"-......
/ .. --; -. _.....- ~---..~--~ ~.._~.~._.
~._--.-- ......_....

!
14

I
28

Fig.

C-2.

I
I
42
56
TIME (minutes)

Graphical

addition

of drift

1
70

components.

I
84

LEGAL NOTICE
This report was prepared as an account of Government sponsored work.
Neither the United States,
nor the Commission, nor any person acting on
behalf of the Commission."
A. Makes any warranty or representation,
expressed or implied, with
respect to the accuracy, completeness, or usefulness of the information contained in this report, or that the use of any information, apparatus, method,
or process disclosed in this report may not infringe privately owned rights; or
B. Assumes any liabilities
with respect to the use of, or for dama.qes
resulting from the use of any information, apparatus, method or process disclosed in this report.
As used in the above, Ir person acting on behalf of the Commission i~
includes any employee or contractor
of the commission, or employee of such
contractor, to the extent that such employee or contractor of the Commission,
or employee of such contractor
prepares, disseminates,
or provides access
to, any information pursuant to his employment or contract with the Commission, or his employment with such contractor.