High Accuracy CMM
Measurements at NIST
by
John Stoup
National Institute of Standards and Technology, USA
2007 CMM Users Meeting - Mexico
October 22, 2007
Today‟s Discussion
We will describe the equipment and processes used at NIST for making
world class CMM Measurements.
• Describe what is needed to make the best possible
measurements.
• Outline some techniques used to assess the CMM
environment‟s thermal performance.
• Discuss optimizing probe performance.
• Present machine performance using gauge data.
• Uncertainty calculations.
• Special measurement setup designs.
Repeatability vs. Time
1.050
1.000
0.950
High Quality Industrial CMM in Good Lab
1000 millimeter dimension
repeatability data
0.900
0.800
Cost of improvement
increases substantially as
you attempt to drop these
lines closer together
0.750
0.700
0.650
0.600
0.500
0.450
0.400
0.350
0.300
Thermal issues dominate
$
$
$$
$$$$
$$$$$$$$
0.550
Short term
repeatability
Standard Deviation (micrometers)
0.850
$
$
$$
$$$$
$$$$$$$$
0.250
0.200
0.150
0.100
0.050
0.000
5 minutes
5 days
5 years
25 millimeter dimension
repeatability data
Mostly machine related issues
Probe limited performance
Repeatability vs. Time
0.850
0.800
0.750
High quality CMM in very good
laboratory
Standard Deviation (micrometers)
0.700
0.650
0.600
0.550
0.500
0.450
NIST PMM in very good lab
19.9˚ - 20.1˚C operating range
0.400
0.350
0.300
0.250
0.200
NIST Moore M48 CMM
0.150
0.100
Probe limited
performance
0.050
0.000
5 minutes
5 days
5 years
What do we need to make very high
accuracy CMM Measurements?
• Extreme high quality lab space.
- gradient control most important.
• CMM capable of exceptional positioning repeatability.
- error mapping will take care of the rest.
• Probe with exceptional gauging repeatability.
• Data collection techniques.
- redundancy.
- test for stability during long data collection runs.
• Operators that strive for the highest accuracy result.
The NIST Advanced Measurement
Laboratory
• Large laboratory
spaces.
• Airflow at the rate of
300 air changes/hour
in CMM space.
• 20.00 ºC 0.01 ºC
temperature
stability.
• Improved power
quality and
mechanical
reliability.
NIST M48 CMM in
AML laboratory
• Reflected room lights.
• Thermally controlled
floor.
• Vibration isolation.
• Laser scales.
• All heating sources
outside of room.
• Granite table added.
• 4 mm/s top speed!
M48 Motion
Mechanisms
• Roller bearing
twin V-ways.
• Lead screw
driven.
• All operation in
oil baths.
Average Temperature (°C )
AML Thermal Performance – short term
20.01
20.008
20.006
20.004
20.002
20
19.998
19.996
19.994
19.992
19.99
19.988
19.986
19.984
19.982
19.98
0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48
Hours
AML Thermal Performance – long term
20.015
Average Temperature ( °C)
20.010
20.005
20.000
19.995
19.990
19.985
19.980
19.975
19.970
19.965
19.960
792
768
744
720
696
672
648
624
600
576
552
528
504
480
456
432
408
384
360
336
312
288
264
240
216
192
168
144
120
96
72
48
24
0
Hours
Thermal Gradient Testing
• We need to find out if the moving parts of the CMM
maintain a constant temperature during operation.
• We need the temperature in the measuring volume of the
machine to be stable during operation.
Therefore, we must
• „Tune‟ the room to optimize these two requirements.
• Both axes – carriage and table motions.
Temperature sensor locations on M48
carriage
Right Side
Metal Sensor
#5 & #6
Air sensor #1
Left Side
Air sensor #2
Metal Sensor
#7 & #8
Air sensor #3
Air sensor #4
Metal Sensor
#9 & #10
Metal Sensor
#11 & #12
RAM
Camera
Thermistor difference data
Thermistor data - side to side differences
0.08
0.06
0.04
Temperature Scale (deg C)
0.02
0
A3-A4
A1-A2
M5-M6
M7-M8
M9-M10
M11-M12
-0.02
-0.04
-0.06
-0.08
-0.1
-0.12
-0.14
Time
Tuning the Room
Thermistor data - side to side differences after airflow adjustments
0.03
0.02
0.01
0
Temperature scale ( °C )
• Remove gradients
from around
machine by
removing some
ceiling tiles.
• Increased
turbulent airflow
with better air
mixing around
the machine.
• Differences
reduced by ~80%
-0.01
-0.02
-0.03
-0.04
-0.05
-0.06
A3-A4
A1-A2
M5-M6
M11-M12
M7-M8
M9-M10
-0.07
-0.08
-0.09
-0.1
Time
Error Mapping
Effort
0
Error (tenth microradians)
• Error mapping the M48
takes about 2 months.
• Performed redundantly
over time to watch
warmup behavior.
• External laser used to
measure all rotational
errors directly.
• Full 21 component map
at 25mm intervals.
Y Axis Roll Map, Ryy
exisiting map
-5
no warmup
-10
1 hr warmup
-15
2 hr warmup
-20
3 hr warmup
-25
-30
-35
-40
-45
-50
0
50
100
150
200
250
300
350
400
450
500
Table position (mm)
550
600
650
700
750
800
850
CMM Probing – repeatability is key
• NIST uses a currently
unavailable probe
design.
• Stylus geometries
mapped for optimum
correction.
• Stem lengths kept as
short as possible.
• Probe trigger design is
important for dirt
detection.
Do what it takes to get probe
repeatability!
• Room airflow creates
vibrations in the probe.
• A cover is required for
highly repeatable results.
• X axis repeatability ~ 9 nm.
• Y axis repeatability ~ 13 nm.
• Z axis repeatability ~ 7 nm.
Average Puck Repeatability – 1m Step Gage
Average Repeatability Standard Deviation - Step Gage Measurements
0.000040
Move to AML July
2004
0.000036
Standard Deviation (mm)
0.000032
In Old Laboratory Environment
In AML
0.000028
0.000024
0.000020
0.000016
0.000012
0.000008
0.000004
0.000000
10/01 2/02
10/06
6/02 7/02
2/03
5/03
11/03
1/04
6/04
12/04 5/05
12/05
4/06
9/06
9/06
Step Gage Data – AML Comparison
Long Term Repeatability - Step Gage Data
all combined history vs. AML data
0.000120
AML Puck Side A
0.000110
AML Puck Side B
0.000100
History Puck Side A
History Puck Side B
0.000090
Prior History Average Linear Fit
0.000080
0.000070
0.000060
0.000050
0.000040
0.000030
0.000020
0.000010
Position (mm)
96
0
10
00
10
40
92
0
88
0
84
0
80
0
76
0
72
0
68
0
64
0
60
0
56
0
52
0
48
0
44
0
40
0
36
0
32
0
28
0
24
0
20
0
16
0
12
0
80
40
0.000000
0
Sigma (mm)
AML Average Linear Fit
~ 3 Day Length-Based Repeatability
Comparison – NIST M48 CMM
• Prior environment results:
• AML current results:
ulb = 0.035 + 0.022 L µm
ulb = 0.019 + 0.015 L µm
** A 45% improvement in performance with a better room!
This term is independent of error sources such as gage instability,
inaccuracy of the CMM error map, fixturing effects, thermal
gradient induced errors, and thermometer calibration.
But it does include CMM positioning, probing effects, error map
stability and thermal stability of the machine space.
Table Setups Designed for Long Operation
• Measurements of ring and
large plug gauges.
• Long gauge blocks, step
gauges and end standards.
• Grid plates and scales.
• 30 % of artifacts we
measure belong to NIST!
M48 CMM Uncertainty Components
Uncertainty Source
•
•
•
•
•
•
•
•
•
•
•
•
Standard Deviations
μm
ppm
Machine Positioning Uncertainty
Temperature difference in beam paths during calibration
Laser Frequency Difference
Measurement Reproducibility (probe effects are here)
Edlén Equation
Index of Refraction – Air Temperature
Index of Refraction - Air Pressure
Index of Refraction – Humidity
Artifact Temperature Measurement Accuracy (4mK)
Coefficient of Thermal Expansion (1ppm/˚C)( 0.05˚C)
Contact Deformation
Gage Surface Geometry
0.04
0.04
0.002
0.004
0.01
0.02
0.04
0.03
0.01
0.04
0.03
0.05
0.05
NIST M48 CMM Typical Uncertainty
Statements
• For 1D measurements:
Uc (k=2) = 0.11 + 0.2L µm
(L is in meters)
• For 2D measurements:
Uc (k=2) = 0.13 + 0.2L µm
(L is in meters)
Special Setups and Arrangements:
Silicon Spheres
• Required for even lower
uncertainties.
• Designed to reduce or
eliminate some
uncertainty components.
• Have achieved task
specific expanded
uncertainties (k = 2) of
about 0.03 micrometers.
Special Setups and Arrangements:
Double Corner Cube
• In one case, we created
better than a class 1000
cleanroom environment
around the machine.
• Designed for a 3D feature
measurement in a critical
component of a NASA
space interferometer to be
launched in the near
future.
Other Special Arrangements
Conclusions
• The NIST M48 CMM has state of the art performance.
• Everything is compromised or designed for the sake of
accuracy and repeatability.
• For the highest accuracy you must have all the required
elements as discussed earlier.
AND
• We are always making incremental improvements in its
performance.
• Measurand definition becoming important due to surface
imperfections of even the best of artifacts.
• We may become “probe-limited” soon.