©
Alexander
Slocum,
last
update
2016.03.27
Deterministic
Design
of
Machines
using
Error
Budget
Methods
In
the
process
of
designing
a
new
machine,
the
designer
typically
starts
with
ideas,
sketches,
quick
calculations
(at
least
first
order
guestimates)
of
feasibility,
and
then
gets
ready
to
start
investigating
ideas
in
more
detail.
However,
before
investing
a
lot
of
time
in
solid
models
and
finite
element
analysis,
it
can
be
helpful
to
more
deeply
consider
in
a
deterministic
way
the
performance
of
different
concepts
either
wholly
or
partially
(e.g.,
a
particular
axis).
The
technique
of
first
order
error
apportionment
among
axes
and
components
followed
by
more
detailed
error
budgeting
can
be
helpful
to
investigate
design
ideas
and
options.
First
step:
Error
apportionment
The
designer
needs
to
have
an
idea
of
the
number
of
axes
needed
to
accomplish
the
task,
and
the
overall
accuracy
(or
repeatability
)
of
the
machine.
Assuming
no
particular
axis
should
“work
harder”
than
another
to
deliver
performance,
the
total
allowable
error
can
be
apportioned
amongst
axes
and
components
within
axes.
This
can
be
done
with
the
spreadsheet
Axis_error_apportionment_estimator.xls
We
now
have
an
idea
of
the
dimensional
performance
requirements
(e.g.,
accuracy
or
repeatability)
for
the
primary
elements
in
the
system.
When
very
high
performance
is
required
where
it
seems
like
we
will
have
to
work
very
hard
to
meet
the
requirements,
The
next
step
is
to
assign
coordinate
systems
and
then
to
estimate
the
error
motions
in
the
elements
and
stiffnesses
of
connections
(e.g.,
bolted
joints,
bearing
elements)
between
coordinate
systems
and
the
compliance
of
structural
elements.
Second
step:
Concept
Development
with
the
aid
of
an
Error
Budget
1
©
Alexander
Slocum,
last
update
2016.03.27
An
error
budget
can
be
used
to
first
assess
just
geometric
errors,
and
the
designer
can
assume
if
geometric
errors
can
be
managed
according
to
limits
allocated
in
the
error
apportionment
phase,
then
the
load-‐induced
errors
could
also
be
dealt
with
by
appropriate
component
sizing.
The
designer
can
then
proceed
using
simple
calculations
to
size
elements
to
solid
model
the
concept,
and
then
use
FEA
to
predict
deflections
and
make
sue
they
are
within
the
limits
set
by
the
error
apportionment
phase.
This
requires
an
iterative
process
and
some
hope
that
when
the
entire
machine
is
assembled
all
will
be
well.
A
more
deterministic
approach
is
to
do
a
second
round
or
error
budgeting
where
applied
loads
and
the
stiffness
of
joints
and
bearings
and
compliance
of
structural
elements
are
added
to
the
model.
This
discrete
element
model
effectively
considers
the
net
effect
of
geometric
errors
and
load
induced
deflections
to
predict
overall
machine
performance.
Being
spreadsheet
based,
the
designer
can
rapidly
spot
trends
and
adjust
member
and
bearing
sizes
(stiffness)
accordingly
to
converge
on
a
workable
design.
THEN
the
design
can
be
solid
modeled
and
checked
with
FEA
against
the
and
the
model
considers
To
illustrate
the
latter
approach,
this
document
will
guide
users
through
using
error
apportionment
and
error
budget
spreadsheets
by
applying
them
to
a
bridge
style
machine
intended
for
use
as
a
wood
router
in
a
home
shop.
Where
to
start?
A
designer
often
starts
at
the
tool
and
works
backwards
to
the
base.
The
rational
is
that
the
designer
should
have
an
idea
of
the
stiffest
tool
that
will
be
used
and
the
power
and
speed
of
the
spindle.
Figure
1
shows
some
stick
figure
concepts
for
the
to-‐be-‐designed
router.
Keeping
early
stage
designs
in
the
hand-‐drawn
phase
enables
the
designer
to
spend
more
time
with
young
family
members
creating
concepts
in
Crayon,
as
shown
in
Figure
2,
as
well
as
preventing
business
overlords
from
assuming
the
design
is
finished
and
so
they
start
taking
orders…
It
also
helps
the
designer
to
relax
and
more
freely
think
of
design
options…
Drawing
animals
and
fun
shapes
within
a
design
helps
keep
young
helpers
happily
engaged
and
promotes
free-‐spirited
creative
thought
in
the
designer.
It
is
like
a
dither
signal
that
is
sometimes
used
to
help
a
servo
overcome
friction.
2
©
Alexander
Slocum,
last
update
2016.03.27
Figure
1:
Early
Stick
Figure
concepts
for
routers
as
motivated
by
consideration
of
Functional
Requirements,
Design
Parameters,
Analysis,
References,
Risks,
&
Countermeasures
(FRDPARRC1)
Figure
2:
Crayons,
big
sheets
of
paper,
blocks…,
and
a
young
helper
do
wonders
for
creativity
and
happiness
of
all
involved.
1
Big
Awesome
(J)
Design
acronym
(created
to
make
fun
of
TMS-‐TLAs).
See
for
example
Graham,
M,
Slocum,
A,
Moreno
Sanchez
R.,
“Teaching
high
school
students
and
college
freshman
product
development
by
Deterministic
Design
with
PREP”,
ASME
Journal
of
Mechanical
Design
(Special
Issue
on
Design
Engineering
Education),
July
2007,
Vol.
129,
pp
677-‐681.
3
©
Alexander
Slocum,
last
update
2016.03.27
Given
that
a
full
sheet
of
plywood
is
ideally
to
be
cut,
a
bridge
style
machine
is
selected
as
the
approach
to
first
consider.
Even
in
the
stick
figure
phase,
it
is
a
good
idea
to
start
to
think
about
coordinate
systems
that
will
represent
the
major
elements
of
the
machine.
The
best
way
to
assign
coordinate
systems
is
to
think
of
the
centers
of
stiffness
of
the
system,
particularly
where
the
bearings
attach,
which
are
simply
represented
in
the
stick
figure.
Start
simple
and
think
in
terms
of
cantilever
and
simply
supported
beams,
so
you
might
have
more
coordinate
systems
than
moving
degrees
of
freedom.
The
bridge
concept
path
that
has
been
decided
upon
can
be
envisioned
as
being
modeled
with
7
Coordinate
Systems:
1. Tool
in
Tool
Holder
2. Spindle
3. Z
axis
4. Y
axis
5. Vertical
Riser
6. X
axis
7. Cutting
point
on
workpiece
As
discussed
in
more
detail
below,
every
Coordinate
System
N
has
an
origin
to
where
it
will
attach
to
the
next
Coordinate
System
N+1,
and
a
Point
Of
Interest
(POI)
to
which
the
previous
Coordinate
System
(N-‐1)
is
attached.
Figure
3
shows
a
first
pass
isometric
view,
still
hand-‐drawn
to
prevent
too
early
lock-‐in
to
design
details
as
can
be
the
case
when
solid
models
are
too
early
created
in
the
design
phase.
4
©
Alexander
Slocum,
last
update
2016.03.27
Figure
3:
Whiteboard
sketched
bridge
machine
concepts
for
a
router
(created
at
white
board
based
on
inspiration
from
early
sketches
created
as
in
Figure
2).2
As
the
bridge
structure
in
Figure
3
was
sketched,
the
designer
started
thinking
about
how
it
might
be
built
and
thus
also
modeled:
where
should
coordinate
systems
be
placed?
Indeed,
many
corrections
and
adjustments
were
rapidly
made
along
the
way
using
an
eraser
and
markers.
A
big
DO
NOT
ERASE
note
ensured
that
the
design
could
remain
for
several
days
so
it
could
mature
with
the
help
of
peer
review
by
colleagues
who
wandered
into
the
office
were
very
valuable.
Sticky
notes
were
sometimes
added
and
used
to
capture
ideas
that
2
Can
you
find
all
the
smiley
faces?
Ideally
there
would
also
be
small
wild
animals
hidden
within.
Staying
a
little
silly
helps
keep
the
designer
free
thinking
so
as
to
not
get
too
enamored,
and
hence
resistant
to
change,
with
their
own
design
too
early.
5
©
Alexander
Slocum,
last
update
2016.03.27
were
not
quite
ready
to
replace
existing
lines.
The
process
was
catalyzed
and
lubricated
by
lunches
and
coffee
breaks.
Throughout,
the
process
enabled
thinking
about
coordinate
systems
that
represent
motion
axes
and
connections
between
major
structural
elements.
The
entire
process
shows
that
it
is
good
to
use
a
whiteboard
for
early
development.
This
“fuzziness”
is
an
important
part
of
the
creative
phase.
Analysis
(as
in
equations
and
also
critical
thinking
about
obtaining
components,
manufacturing…)
can
be
thought
of
as
a
lens
that
helps
bring
a
design
into
focus.
A
note
about
assigning
coordinate
systems,
which
will
be
discussed
in
more
detail
below.
Every
Coordinate
System
(CS)
has
a
Point
Of
Interest
(POI)
and
a
Point
Of
Attachment
(POA),
which
by
convention
here
is
at
the
origin
of
the
CS.
Accordingly,
one
CS’s
origin
(POA)
is
located
at
another’s
POI.
Meanwhile,
errors
in
machines
are
dominated
by
systematic
errors
(e.g.,
axes
are
not
bolted
together
exactly
square,
or
repeatable
straightness
error
in
an
axis),
random
errors
(e.g.,
bearing
noise),
and
deflections.
The
latter
are
due
to
limited
stiffness
at
joints
and
bearings
and
compliance
in
structures.
It
is
thus
logical
to
follow
the
structural
loop
of
the
machine,
which
traces
the
path
of
forces
through
the
machine
from
toolpoint
back
through
the
machine
and
up
into
the
workpiece
where
the
tool
is
acting,
and
place
a
CS’s
POA
at
a
connection
between
major
elements.
Hence
each
CS’s
POI
will
be
collocated
with
the
next
CS’s
origin
which
is
at
a
connection
between
major
elements…
The
stiffness
of
the
connection
between
coordinate
systems
can
thus
be
represented
by
springs
between
one
CS’s
POA
and
its
adjoining
CS’s
POI.
Within
a
coordinate
system,
the
compliance
of
the
structure
that
connects
the
POA
to
the
POI
can
then
be
represented
by
a
full
(6x6)
compliance
matrix
as
discussed
below.
Figuratively
the
sequence
is
shown
in
Figure
4.
Figure
4:
Modeling
a
machine
as
a
sequence
of
compliant
beams
and
discrete
spring
connections.
6
©
Alexander
Slocum,
last
update
2016.03.27
Once
the
overall
concept
has
mostly
congealed,
different
views,
close
ups
and
cross
sections
of
regions
can
be
sketched
to
help
the
designer
better
understand
as
shown
for
example
in
Figures
5
and
6.
One
method
of
design
synthesis
is
to
start
at
the
tool
tip,
because
that’s
where
the
action
is,
and
keeping
in
mind
the
forces
and
accuracy
desired,
create
elements
along
the
structural
loop
to
enable
the
desired
motions
to
occur,
and
conceptually
size
them
in
accordance
with
their
distance
from
the
tool
point.
This
is
the
phase
where
one
moves
from
a
stick
figure
to
at
least
quasi
reality
J.
Following
this
logic,
consider
the
machine
in
Figure
3
and
the
cross
section
in
Figure
5.
CS_1’s
origin
(POA)
is
located
at
the
center
of
the
collet
that
holds
the
tool.
The
POI
is
the
tool
tip
and
the
compliance
of
the
tool
(1/stiffness)
can
be
easily
determined.
The
stiffness
of
the
tool
held
in
the
collet
and
the
collet
in
the
tapered
receiver
of
the
spindle
shaft
would
be
modeled
as
a
set
of
linear
and
rotary
springs
where
resultant
forces
and
moments
on
the
CS’s
origin
cause
linear
and
angular
displacements
respectively.
This
might
require
experimentation
or
a
literature
search
as
surely
this
has
been
tested
before.
Figure
5:
Whiteboard
sketched
cross
section
of
the
tool,
collet,
and
spindle
on
the
end
of
the
Z-‐axis
structure.
Note
Y
axis
is
positive
into
the
page
so
a
positive
X
direction
force
would
cause
a
negative
rotation
about
the
Y
axis.
CS_2
needs
to
be
able
to
model
the
stiffness
of
the
spindle
bearings
as
well
as
the
spindle
shaft.
If
it
were
located
at
one
bearing
or
the
other,
the
angular
error
created
by
the
resultant
moment
from
the
force
at
the
tool
point
may
not
be
completely
captured.
Yet
the
spindle
shaft
may
also
be
a
significant
source
of
compliance.
The
joint
between
the
spindle
housing
and
end
of
the
Z-‐Axis
structure
is
likely
to
be
far
more
rigid
than
the
spindle
bearings,
and
hence
we
will
focus
our
efforts
on
the
spindle
bearings.
Placing
CS_2
at
the
center
of
the
spindle
shaft,
and
modeling
its
attachment
stiffness
to
the
POI
of
CS_3
(the
end
of
the
Z-‐Axis)
as
the
linear
and
angular
(pitch
and
yaw)
stiffness
of
the
spindle
bearings
and
housing
will
accurately
reflect
the
effect
spindle
bearing
deflections
have
on
the
motion
of
the
tool
tip.
7
©
Alexander
Slocum,
last
update
2016.03.27
But
what
of
the
deflections
of
the
spindle
shaft,
the
compliant
structure
connecting
CS_2’s
POA
to
its
POI
which
is
the
origin
of
CS_1?
The
spindle
shaft
conservatively
looks
like
a
simply
supported
beam
unless
it
is
a
slow
speed
spindle
with
a
significant
set
of
back
to
back
mounted
rolling
element
bearings
at
its
front.
In
either
case,
linear
deflection
of
the
simply
supported
beam
is
not
significant
because
CS_1
is
considered
located
at
a
support
point
(the
front
spindle
bearing)
but
angular
deflections
are
important:
The
slope
deflection
of
the
spindle
shaft
at
the
POI
directly
affects
angular
deflection
and
hence
Abbe
error
of
the
tool
tip.
Since
the
relative
position
of
CS_2
and
CS_1
does
not
change
we
can
locate
CS_2
at
the
center
of
the
spindle
shaft,
and
model
the
compliant
beam
that
connects
CS_2’s
POA
to
its
POI
as
a
simply
supported
beam
whose
length
is
that
of
the
entire
spindle
shaft.
Moving
on
to
the
rest
of
the
system,
consider
Figure
3
and
Figure
6
which
shows
a
side
view
of
the
machine,
where
it
should
be
clear
that
CS_3’s
origin
is
at
the
Center
of
Stiffness
(COS)
of
the
Z
bearings
(e.g.,
two
linear
guide
(profile
rail)
cartridges
(generically
called
bearing
blocks)
on
each
of
two
profile
rails).
The
stiffness
of
these
bearings
with
the
local
structure
will
dominate
the
attachment
stiffness
of
the
CS_3’s
POA
to
CS_4’s
POI.
The
compliance
between
CS_3’s
POA
and
POI
will
be
that
of
the
Z-‐Axis
structure,
sometimes
called
the
ram,
which
can
be
modeled
as
a
cantilever
beam,
whose
length
changes
in
accordance
with
the
position
of
the
spindle,
which
is
controlled
by
the
Z-‐Axis
actuator.
That
Z-‐Axis
actuator
will
thus
be
the
Z-‐Axis
compliance
value
in
the
{6x6}
compliance
matrix,
where
the
other
entries
are
due
to
those
of
a
cantilever
beam.
Note
in
this
design
the
Z-‐Axis
bearing
blocks
are
attached
to
a
chunky
structure
to
which
the
Y-‐Axis
bearing
blocks
are
also
attached.
In
this
case
CS_3’s
POA
(origin)
is
attached
to
CS_4’s
POI
and
there
really
is
no
compliant
structure
between
the
two.
Hence
the
compliance
matrix
for
CS_4’s
“structure”
would
just
be
the
null
matrix
of
zeros3.
3
In
a
spreadsheet
this
is
easier
to
do,
because
old
values
can
be
deleted
and
the
default
value
is
“0”
in
subsequent
calculations
that
reference
the
cells.
Else
if
the
analysis
was
based
on
using
a
stiffness
matrix
(instead
of
the
compliance
matrix)
the
user
would
have
to
delete
values
and
enter
1E12
or
some
such
super
huge
value
8
©
Alexander
Slocum,
last
update
2016.03.27
Figure
6:
Whiteboard
sketched
side
view
of
the
machine.
The
bridge
of
the
machine
represents
a
closed
structural
loop,
whereas
the
matrix
methods
that
are
to
be
used
to
analyze
errors
(Homogeneous
Transfer
Matrices,
HTMs)
is
a
serial
based
approach.
The
solution
to
this
quandary
is
to
visualize
how
the
machine
deforms
and
consider
the
philosophy
of
POA
springs
and
the
compliance
matrix
that
represents
the
structure
between
the
POA
and
the
POI.
With
this
in
mind,
place
the
origin
of
CS_5
at
the
top
of
one
of
the
risers
that
connect
the
X-‐
Axis
bearings
to
the
bridge
beam.
The
bridge
beam
is
considered
a
simply
supported
beam,
when
the
Y,
Z
axis
assembly
is
all
the
way
to
the
right
and
thus
far
away
from
the
origin
of
CS_5
one
might
think
this
represents
a
HUGE
moment
that
will
cause
an
inordinately
(and
inaccurate)
Abbe
error.
However,
note
in
Figure
3
how
the
axes
directions
have
been
donated
as
CS_5’
and
CS_5’’
(e.g.,
Z’5
and
Z’’5).
This
indicates
that
the
attachment
points
are
far
apart.
Given
that
they
are
each
very
stiff,
the
moment
stiffness
of
two
linear
springs
representing
the
attachment
points,
which
are
spaced
far
apart,
is
also
HUGE
(see
the
linear
springs
spaced
apart
yielding
rotational
spring
analysis
method
on
the
left
side
of
Figure
4)
9
©
Alexander
Slocum,
last
update
2016.03.27
and
thus
the
resulting
inaccuracy
of
an
Abbe
error
caused
by
angular
deflection
of
the
coordinate
system
is
very
small,
and
in
the
worst
case
just
adds
to
the
conservative
nature
of
the
model.
The
simply
supported
beam
model
of
the
bridge
is
used
to
determine
the
compliance
matrix
for
the
structure
between
CS_5’s
POA
and
its
POI
(which
is
also
CS_4’s
POA).
As
the
Y-‐Axis
servo
moves
the
Y,
Z
axis
assembly
from
left
to
right,
the
slope
will
change
from
negative
to
zero
(when
its
in
the
middle)
and
back
to
positive.
Hence
the
Abbe
error
at
the
tool
tip
will
be
accurately
accounted
for.
The
same
philosophy
is
used
for
CS_6
which
models
the
X–Axis.
Note
that
the
compliance
matrix
for
the
structure
between
CS_6’s
POA
and
POI
is
a
little
more
interesting
because
of
the
two
risers.
This
just
means
that
the
bending
stiffnesses
are
the
combined
stiffnesses
of
the
two
riser
elements,
but
the
torsional
stiffness
about
the
Z
axis
is
based
on
the
bending
stiffness
of
the
risers
spaced
apart
(analogous
to
the
method
shown
on
the
left
side
of
Figure
4).
Note
that
the
effective
center
of
CS_6
is
thus
at
the
center
of
the
table.
However,
in
terms
of
the
spreadsheet
model
(and
coordinates
entered)
it
will
be
kept
at
the
center
of
stiffness
of
the
left
side
X
bearings.
The
final
coordinate
system
is
then
placed
on
the
workpiece
on
the
table
at
the
point
where
the
tool
will
contact
the
work.
To
model
the
effect
of
springs
and
compliances
associated
with
CS_7
takes
some
thought
experiments.
This
is
essentially
the
same
condition
as
with
the
bridge,
where
CS_4
(analogous
here
to
CS_7)
moved
along
the
bridge
and
CS_5’
and
CS_5’’
(analogous
to
CS_6’
and
CS6’’
here)
were
located
at
the
ends
of
the
bridge.
Reciprocity!
The
POA
for
CS_7
is
thus
modeled
as
the
stiffness
of
the
workpiece
from
where
the
tool
contacts
it
to
where
it
is
attached
to
the
bed
(in
most
cases
this
should
ideally
be
infinite
with
respect
to
the
stiffness
(1/compliance)
of
the
tool).
The
compliance
of
CS_7
from
its
POA
to
its
POI
is
the
compliance
of
the
machine
frame
and
bed
and
is
modeled
as
a
simply
supported
structure.
However,
some
finesse
is
required
to
apply
the
resultant
forces
and
moments
from
CS_6
to
the
compliance
matrix.
The
resultant
forces
and
moments
on
the
origin
of
CS_7
are
applied
to
the
compliance
matrix
to
obtain
the
apparent
deflections
of
the
POI
of
CS_7.
The
structural
loop
is
thus
complete
and
modeled.
Figure
7
summarizes
coordinate
systems
1
through
7
and
their
origins
(points
of
attachment)
and
points
of
interest.
Such
a
table
can
be
helpful
for
the
designer
when
building
the
detailed
error
budget.
Indeed,
recording
sketches
made
and
freeform
writing
to
record
the
thinking
associated
with
placement
of
coordinate
systems
and
modeling
of
stiffnesses
and
compliances
can
be
invaluable
to
the
designer
and
associates.
The
act
of
writing,
knowing
others
will
read,
helps
to
hone
thoughts
and
resolve
quandaries.
It
also
makes
it
easier
for
peer
review
to
help
identify
and
correct
issues.
10
©
Alexander
Slocum,
last
update
2016.03.27
Figure
7:
Summary
of
coordinate
systems
1
through
7
and
their
origins
(points
of
attachment)
and
points
of
interest.
“Stiffness”
(Compliance)
of
elements
between
a
CS’s
POA
and
POI
In
order
to
assign
reasonable
values
to
the
distances
between
coordinate
systems
origins
for
the
concept
shown
in
Figure
3
so
the
error
spreadsheet
can
be
utilized,
a
sense
of
proportion
of
the
size
of
the
structural
elements
is
needed.
First
start
with
an
estimate
of
the
loads
on
the
tool,
and
given
the
allowable
error
of
structural
elements
(Figure
1),
the
stiffness
required
of
each
axis
can
be
estimated.
Given
that
some
of
the
parameters
are
essentially
a
“guess”
based
on
what
may
seem
reasonable,
the
resulting
stiffness
guestimates
should
be
treated
as
fuzzy
estimates
subject
to
change
and
barter
as
one
idea
or
element
is
traded
off
for
another.
The
tool
tip
is
thus
the
point
of
interest.
A
spreadsheet
is
a
good
way
for
a
designer
to
quantify
initial
guestimates,
as
shown
for
the
router
for
example
Figure
8.
11
©
Alexander
Slocum,
last
update
2016.03.27
Figure
8:
Initial
assessment
of
forces
on
the
cutting
tool.
Next
from
the
first
guess
error
apportionment
shown
in
Figure
1,
we
see
for
the
three-‐axis
router
we
are
designing
here,
each
axis’
structure
is
allocated
10
microns
of
error
(deflection).
This
means
that
each
axis
must
have
a
structural
stiffness
of
about
85.5N/0.010mm
=
8550
N/mm.
Assuming
for
now
the
joints
are
much
stiffer
than
any
one
beam,
something
we
will
have
to
pay
careful
attention
to
during
the
design
phase.
Given
the
bridge
type
structure
of
Figure
3,
for
1.3
m
travel
(so
as
to
cut
a
48”
wide
sheet
of
plywood)
the
bridge
would
be
at
least
1.4m
between
the
uprights.
Remember,
the
stiffness
(compliance)
goes
like
L^3
so
be
frugal
with
adding
space
for
what
seems
like
ease
of
design
or
packaging.
In
order
for
a
1.4
m
long
simply
supported
beam
to
yield
the
a
stiffness
of
8550
N/mm
in
the
middle
of
the
beam,
the
equation
for
deflection
of
a
simply
supported
beam
loaded
by
a
force
in
the
middle
is
FL^3/48EI,
a
square
aluminum
cross
section
beam
must
have
a
moment
of
inertia
of
about
7E6
mm^4.
This
seems
large,
but
can
be
obtained
with
a
standard
stock
extruded
aluminum
square
tube
6in
x
6in,
x
1/8”
wall,
which
would
be
a
very
efficient
beam
(high
stiffness
to
weight).
To
help
build
design
intuition,
its
interesting
to
explore
possibilities.
Assume
we
were
wanting
to
machine
metal
with
higher
precision
than
a
simple
router
would
provide
and
we
wanted
a
stiffness
in
our
simply
supported
beam
of
about
64
kN/mm
(8X).
This
requires
a
square
beam
of
about
200mm
x
200mm
x
12mm
wall
(stock
US
size
is
about
8in
x
8in
x
½”
wall),
or
equivalently
a
250mm
x
250mm
x
6mm
wall
thickness.
Considering
the
beam
is
1400mm
long,
the
ratio
of
length
to
height
of
the
latter
is
about
5.6.
A
300mm
x
300
x
5mm
thick
beam
would
have
the
same
cross
sectional
area
and
be
40%
stiffer!
Think
of
Saint
Venant
for
beams
that
have
the
same
stiffness,
the
length/cross
section
dimension
proportions
are
7,
5.6,
and
4.7,
and
the
relative
areas
(weights)
are
1.00,
0.65,
and
0.39
respectively.
How
to
choose?
The
bigger
cross
section,
smaller
wall
thickness
beam
will
be
much
lighter
and
hence
have
higher
natural
frequency,
but
can
it
be
obtained?
Indeed
there
are
limited
sizes
of
stock
square
aluminum
tubing,
with
8in
x
8in
x
1/2in
wall
the
largest.
One
has
to
go
to
steel
to
get
larger
tube.
A
1400
mm
long
8in
x
8in
x
¼”
wall
square
steel
tube
would
12
©
Alexander
Slocum,
last
update
2016.03.27
have
a
stiffness
of
6392
N/mm.
What
is
the
“opposite”
of
a
rectilinear
tube?
A
Round
tube!
There
are
far
more
options
in
round
tubing,
but
what
are
the
implications
for
bearing
mounting
and
mounting
the
tube
to
the
rest
f
the
machine?
A
10in
diameter
round
tube
with
½”
wall
thickness
has
the
same
cross
sectional
area
as
the
8in
x
½”
wall
square
tube,
but
is
20%
stiffer
in
bending
and
60%
stiffer
in
torsion.
What
would
you
choose?
Back
to
the
router,
consider
Figure
4
and
Figure
8
below.
The
compliance
matrix
relates
the
stiffness
of
a
Coordinate
System’s
structure
from
the
POI
to
the
center
of
stiffness
(Coordinate
System
N’s
origin).
This
matrix
can
be
created
by
the
user
in
the
coordinate
systems’
worksheets.
A
6x6
compliance
matrix
is
needed
so
the
angular
deflections
of
a
structure
in
a
coordinate
system
N
are
added
to
the
next
coordinate
system’s
(N-‐1)
origin
deflections.
Then
when
the
error
gains4
for
Coordinate
system
N-‐1
are
applied,
they
will
effectively
include
the
Abbe
errors
caused
by
a
structural
member’s
angular
deflection
causing
other
elements
attached
to
it
to
tilt.
If
a
coordinate
system
is
placed
at
the
center
of
stiffness
of
bearings
that
allow
motion,
the
compliance
in
the
direction
of
motion
is
considered
to
be
the
stiffness
of
the
servo
(e.g.,
ballscrew,
coupling,
bearings
etc.).
At
the
origin
of
the
coordinate
system
there
will
be
some
sort
of
attachment
to
the
next
Coordinate
system’s
POI.
At
each
Coordinate
system
origin
there
are
three
resultant
forces
and
moments
computed
from
the
forces
and
moments
initially
acting
on
CS_1’s
POI
plus
any
additional
forces
and
moments
added
along
the
way
to
a
Coordinate
System’s
POI
or
origin.
If
there
are
forces,
such
as
from
center
of
mass,
in-‐between,
they
can
always
be
resolved
into
forces
and
moments
applied
to
the
origin.
Hence
only
three
linear
and
three
angular
stiffness
values
are
needed
to
describe
the
attachment
of
a
coordinate
system
(just
a
vector).
Figure
9:
Image
of
the
attachment
point
stiffness
and
compliance
matrix
for
a
coordinate
system
Since
users
will
tend
to
forget
to
enter
values,
and
the
default
in
Excel
for
an
empty
cell
is
“0”,
the
compliance
matrix
is
used
in
the
error
budget
spreadsheet.
It
also
makes
more
4
Some
positive
angular
errors
result
in
negative
displacements,
thus
since
random
errors
have
no
sign,
for
all
the
coordinate
systems,
the
magnitude
of
the
amplification
of
angular
errors
must
be
individually
computed
for
each
potential
angular
error.
A
table
of
these
amplification
factors,
or
error
gains,
can
then
be
used
to
compute
the
Abbe
errors
attributed
to
random
angular
errors
about
each
of
the
axes
of
each
of
the
coordinate
systems.
The
error
gains
for
the
systematic
errors,
on
the
other
hand,
do
have
a
sign.
13
©
Alexander
Slocum,
last
update
2016.03.27
sense
given
that
we
have
forces
and
want
to
determine
deflections,
so
X
=
C*F,
where
C
=
1/k.
Note
the
matrix
will
always
be
symmetric
in
this
analysis.
The
elements
for
which
6x6
compliance
matrices
are
needed
are
shown
in
Figure
7.
It
is
desirable
to
have
a
worksheet
within
the
error
budget
spreadsheet
for
entering
the
dimensions
of
members
and
to
calculate
the
stiffness
matrices
since
the
formulas
will
be
similar.
Note
that
when
equating
the
length
of
a
beam
by
referring
to
a
cell
in
the
worksheet
Axes'
CS
and
Loads
that
sometimes
the
coordinates
may
be
negative,
but
lengths
of
beams
are
always
positive,
so
use
the
ABS()
function.
Also
note
when
entering
formulas
for
compliances
remember
to
consider
the
sign.
For
example,
a
moment
may
cause
a
deflection
in
a
negative
direction5.
It
is
helpful
to
have
the
formulas
for
deflection
and
slope
of
cantilever
and
simply
supported
beams
subject
to
forces
and
moments
as
shown
in
Figure
10.
It
is
very
difficult
to
have
a
beam
with
a
true
zero
slope
condition
at
both
ends,
and
it
is
more
conservative
to
assume
a
simply
supported
condition.
Note
it
is
important
to
be
careful
of
the
sign
of
the
applied
force
or
moment
and
the
resulting
deflections
and
slopes.
These
formulas
take
this
into
account,
such
that
if
a
horizontal
beam
were
also
attached
to
the
point
where
the
force
or
moment
was
applied,
the
resulting
Abbe
error
would
have
the
proper
sign.
5
NOTE
Excel
has
a
strange
“feature”
where
entering
a
formula
such
as
“-‐L/(E*I)”
returns
a
positive
value!
Entering
“-‐1*L/(E*I)”
returns
a
negative
value.
The
same
is
true
when
a
reference
is
raised
to
a
power:
-‐1*B130^2
yields
–B1302;
however,
-‐B130^2
yields
B1302.
14
©
Alexander
Slocum,
last
update
2016.03.27
Figure
10:
Deflection
and
slope
formulas
for
common
beams
used
to
model
machines.
Position
of
the
force
on
a
simply
supported
beam
is
from
left
to
right.
A
right
hand
coordinate
system
is
assumed.
It
is
very
important
to
create
sketches
of
beams
being
modeled
with
their
assigned
coordinate
systems,
and
compare
the
orientations
to
the
right
hand
systems
used
for
shown.
A
check
of
the
equations
shown
in
Figure
10
corresponds
to
the
shapes
shown
in
the
sketches,
which
can
be
verified
by
supporting
and
loading
an
object
and
observing
how
it
deforms.
As
shown
in
Figure
11,
these
values
represent
the
compliances,
and
it
is
good
to
see
that
they
are
symmetric:
the
slope
caused
by
a
force
has
the
same
sign
and
value
as
a
deflection
caused
by
the
moment.
Figure
11:
Normalized
results
of
deflection
and
slope
formulas
of
Figure
11
applied
to
common
beams
used
to
model
machines.
For
a
servo
controlled
axes,
which
controls
the
position
of
a
CS
origin
(POA)
wrt
another
coordinate
system’s
POI
CS
origin
(POA@Center
of
Stiffness),
the
axial
stiffness
(compliance)
of
the
system
at
the
Point
Of
Attachment
(POA)
will
be
dominated
by
the
actuator,
such
as
a
ballscrew
or
belt.
The
compliance
of
the
shaft
(or
belt)
can
be
computed
and
then
a
factor
(typically
3
for
the
attachment
(e.g.,
nut)
to
the
moving
member
and
the
support
bearings)
accounts
for
the
compliance
of
other
elements
used
to
attach
it.
Figure
12
shows
an
image
from
the
worksheet
CS_3
Z-‐Axis
.
15
©
Alexander
Slocum,
last
update
2016.03.27
Figure
12:
Results
of
deflection
and
slope
formulas
applied
to
common
beams
used
to
We
are
now
ready
to
apply
the
above
to
each
of
the
coordinate
systems
as
shown
in
Figure
7
using
the
worksheet
bearings,
servo,
structure.
CS_1
Tool
Holder
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
For
the
Tool
Holder,
see
Figure
5,
the
compliance
is
that
of
the
tool
which
is
modeled
as
a
fraction
of
that
of
a
cylinder
to
account
for
the
flutes.
Figure
13
shows
the
Tool
Holder
system
compliance
calculations
done
in
the
worksheet
CS_1
Toolholder.
Note
how
figures
are
pasted
to
help
the
designer
check
for
the
sign
of
the
compliances
in
accordance
with
coordinate
system
placement
on
the
beam
element.
This
should
be
done
for
each
coordinate
system.
Note
the
“Note:”
on
the
last
cell.
The
designer
should
leave
notes
on
how
they
modeled
the
system.
Notes
can
also
be
attached
to
individual
calculation
cells.
Note
that
the
worksheet
is
made
generic,
so
if
certain
cells
with
their
entires
are
not
needed,
those
cells
should
just
be
colored
blue.
No
need
to
delete
or
clear
entries
as
they
may
be
used
in
the
next
round
of
use
of
the
spreadsheet.
Figure
13:
Compliance
calculations
for
the
Tool
Holder
coordinate
system.
Attachment
compliance
(POA
to
next
CS
POI)
16
©
Alexander
Slocum,
last
update
2016.03.27
The
attachment
compliance
for
the
tool
holder,
a
tapered
collet
for
example,
will
be
an
order
of
magnitude
greater
than
the
stiffness
of
the
tool
itself,
and
even
of
the
spindle
shaft.
It
is
helpful
that
whenever
copying
a
set
of
cells,
pasting,
and
then
updating
the
formulas,
to
change
the
cell
background
color
as
a
reminder
to
make
sure
to
change
all
the
cells!
The
compliance
matrix
references
the
cells
to
the
left,
so
if
additional
coordinate
systems
are
to
be
added
for
a
new
machine
(or
CS),
the
entire
block
can
just
be
copied
and
pasted,
and
then
checked
for
the
orientation
of
the
structure
with
respect
to
the
coordinate
systems
and
the
compliance
matrix.
As
cells
are
changed,
change
their
background
color
and
thus
prevent
missing
one.
CS_2
Spindle
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
The
spindle
coordinate
system
is
placed
at
the
center
of
the
spindle
shaft
between
the
spindle
bearings,
as
discussed
in
detail
in
conjunction
with
Figure
5
above,
and
the
entries
for
determining
the
compliance
matrix
from
the
worksheet
CS_2
Spindle
are
shown
in
Figure
14.
Figure
14:
Compliance
calculations
for
the
Tool
Holder
coordinate
system.
Attachment
compliance
(POA
to
next
CS
POI)
Note
that
the
forces
applied
to
the
simply
supported
beam
are
considered
to
be
at
the
front
spindle
bearing
and
thus
they
cannot
create
an
angular
deflection
and
the
corresponding
compliance
terms
compute
to
zero,
as
do
the
off
axis
moment
effects
on
displacement.
Attachment
compliance
(POA
to
next
CS
POI)
The
stiffness
of
the
POA
of
CS_2
to
the
POI
on
the
Z-‐Axis
is
assumed
to
be
dominated
by
the
spindle
bearings.
The
attachment
points
are
assumed
to
be
at
the
equator
points
of
the
four
bearings
CS_3
Z-‐Axis
17
©
Alexander
Slocum,
last
update
2016.03.27
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
The
structure
between
the
Z
axis
POA
and
its
POI
(the
origin
of
the
Spindle)
is
dominated
by
the
tube
that
is
the
Z-‐Axis
structure
itself:
The
Z-‐axis
bearing
blocks
are
fixed
to
a
plate
structure,
which
also
bolts
to
the
Y
axis
bearing
blocks.
The
Z-‐Axis
profile
bearing
rails
are
bolted
to
the
structure,
which
is
modeled
as
an
aluminum
square
tube.
A
square
tube
is
considered
because
the
cutting
forces
can
be
equal
in
the
X
and
Y
directions
and
so
equal
stiffness
is
needed.
Note
the
length
of
the
structure
is
tied
to
Attachment
compliance
(POA
to
next
CS
POI)
The
stiffness
of
the
POA
of
CS_3
to
the
POI
on
the
Y-‐Axis
is
assumed
to
be
dominated
by
the
Z-‐Axis’
bearing
carriages
(blocks).
CS_4
Y-‐Axis
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
The
Y-‐Axis
POI
is
the
Z-‐Axis
origin
and
the
two
are
very
close
together
so
there
is
no
appreciable
structural
compliance
between
the
two:
Its
just
a
chunky
plate,
so
the
compliances
are
all
set
to
“0”.
Attachment
compliance
(POA
to
next
CS
POI)
The
stiffness
of
the
POA
of
CS_4
to
the
POI
on
the
Bridge
is
assumed
to
be
dominated
by
the
Y-‐Axis’
bearing
carriages
(blocks).
CS_5
Bridge
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
The
bridge
spans
the
risers,
and
to
the
bridge
are
attached
bearing
rails
on
which
the
Y
axis
bearing
carriages
move.
The
structural
compliance
of
the
bridge
structure
thus
must
be
calculated.
As
discussed
above,
for
a
closed
structural
loop
machine,
an
open
kinematic
model
can
be
used
by
considering
the
attachment
points
for
CS_5
to
CS_6
to
be
widely
spaced
apart
(at
the
top
of
each
riser)
and
then
calculate
the
angular
stiffness
of
the
attachment
accordingly.
This
gives
a
huge
rotational
stiffness
so
the
angular
deflection
will
be
very
small.
The
structural
compliance
is
that
of
a
simply
supported
beam,
so
when
the
Y
axis
is
positioned
in
the
middle
of
the
bridge,
the
angular
deflection
of
the
bridge
will
actually
be
zero.
Near
the
ends
of
the
bridge,
the
compliance
becomes
huge.
Attachment
compliance
(POA
to
next
CS
POI)
18
©
Alexander
Slocum,
last
update
2016.03.27
The
stiffness
of
the
POA
of
CS_5
to
the
POI
on
the
X-‐Axis
(the
top
of
the
risers)
is
a
bolted
connection
which
if
done
correctly
should
be
as
stiff
as
if
welded.
It
is
modeled
as
4
points
spaced
at
the
corners
of
the
joint.
CS_6
X-‐Axis
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
From
the
plates
to
which
the
X-‐Axis
bearing
carriages
are
attached
up
to
the
bridge
are
the
risers.
The
compliance
of
these
structures
must
be
taken
together,
so
for
example
in
the
X
direction,
the
compliance
of
the
two
risers
is
one-‐half
of
that
of
a
single
riser.
The
torsional
compliance
about
the
Z-‐Axis
is
the
effect
of
the
lateral
compliances
K
of
the
risers
acting
as
a
force
couple
over
the
length
L
of
the
bridge
and
is
thus
2/KL2.
The
roll
compliance
about
the
X-‐Axis
is
a
function
of
the
compression
stiffness
of
the
risers
ad
the
bridge
length
and
thus
is
quite
huge.
Just
before
we
come
to
the
end
of
the
structural
loop
back
at
the
toolpoint,
it
is
important
to
note
that
there
are
essentially
two
options
for
placement
of
the
bed’s
CS
origin.
The
first
is
to
simply
locate
it
in
the
center
of
the
bed
and
call
that
the
reference
frame
origin.
Then
the
entire
machine’s
performance
is
wrt
the
reference
frame
(center
of
bed).
The
second
option
is
to
assume
that
Bed
coordinate
axis
origin
is
also
at
the
toolpoint,
and
thus
the
position
of
the
X-‐Axis
coordinate
system
with
respect
to
the
bed
should
be
such
that
the
origin
of
the
Bed
coordinate
system
is
located
at
the
toolpoint.
This
can
be
achieved
with
the
product
of
-‐1
and
the
sum
of
the
coordinates
along
the
structural
loop
chain.
This
is
what
will
be
assumed
for
this
design.
Attachment
compliance
(POA
to
next
CS
POI)
The
stiffness
of
the
POA
of
CS_6
to
the
POI
on
the
Bed
is
assumed
to
be
dominated
by
the
X-‐
Axis’
bearing
carriages
(blocks).
CS_7
Bed
Structure
compliance
(from
a
coordinate
system’s
POA
to
its
POI)
The
structure
of
the
bed
is
assumed
to
be
a
sandwich
panel
such
as
plywood
bonded
to
either
side
of
Styrofoam
insulation
for
a
home
machine
or
1/8”
aluminum
sheets
bonded
to
either
side
of
honeycomb.
The
bed
really
thus
is
more
of
a
plate
structure,
but
for
simplicity
sake
here
it
is
assumed
to
be
a
beam
of
length
the
same
as
the
bridge,
and
width
ten
times
its
thickness.
It
ends
up
being
as
stiff
as
the
bridge.
The
X-‐Axis
servo
position
thus
ends
up
not
being
a
sensitive
parameter,
what
is
more
important
is
the
position
of
the
spindle
as
controlled
by
the
Y
axis.
Attachment
compliance
(POA
to
next
CS
POI)
19
©
Alexander
Slocum,
last
update
2016.03.27
The
stiffness
of
the
POA
of
CS_7
to
the
POI
od
the
“next”
reference
frame,
given
that
the
next
frame
is
to
be
at
the
toolpoint
would
actually
be
the
attachment
of
the
part
being
cut
to
the
bed.
If
properly
attached,
it
should
be
very
much
ore
stiff
than
the
stiffness
of
the
bed
itself
for
example.
Discussion
on
the
layout
of
the
spreadsheet
to
perform
the
required
analysis
This
section
is
under
construction…
it
is
not
needed
for
operation
by
a
user
of
the
spreadsheet,
but
it
is
good
to
discuss
the
analysis
process.
It
will
be
written
in
coming
weeks….
Error
gains
The
following
figure
illustrates
why
error
gains
are
needed,
and
for
systematic
errors,
some
of
the
gains
are
negative
20
©
Alexander
Slocum,
last
update
2016.03.27
21