| Acceleration
Direction |
In
the past we only marked the "X" and "Y" axes our triaxial
accelerometers to
keep costs down. We have recently added laser marking into our
processes, so our newer triaxial accelerometers will have a "Z" axis
arrow on the side. If the "Z" arrow is not present on the
side of
the unit, it would be pointing up when the accelerometer is sitting on
a table with its marked lid surface facing up. If
you place the X or Y arrow pointing up with the unit sitting on a
table, that axis will also read +1g.
When the accelerometer is sitting on a table, the table provides an
equilibrium force against the bottom of the unit which keeps the unit
from free falling due to gravity. This
is the
polarity convention that we have chosen because it is also consistent
with free body acceleration such as in a car. If
you glue the unit to the headlight face of a car with one of the axis
arrows pointing towards the front of the car and step on the gas
(making the car accelerate), the unit will record positive acceleration
for that axis. This is consistent
with the
headlight pushing on the face of the unit opposite to the arrow head.
|
| BIAS |
The
accelerometer output with no acceleration present.
For
our differential output analog accelerometers it is a signed quantity
that is expressed in terms of either g or output volts and is ideally
equal to zero g or zero volts.
|
| Bias
Calibration Error |
Expressed
as
a percentage of span.
For
example, a 2% of g-span error for a 5 g full scale device would equate
to +/-2% of 10g = +/-0.2g. This
+/-2% error in
terms of output voltage for a 5g analog accelerometer would be +/-.02 x
8V or +/-160 mV. This same 2% error
in terms of
output frequency for a 5g digital accelerometer operating with a 250
kHz clock would be +/-.02 x 250 kHz = +/-5 kHz.
|
| Bias
Temperature Shift |
Expressed
as
ppm of span per deg C.
For
example, the percent of span bias shift that would occur for a 25g full
scale device with a +/-200 ppm of span per deg C rating and a 55 deg C
rise from room temperature would be: +/-200
/
1,000,000 x (80C - 25C) x 100% of span = +/-1.1% of span.
The
g shift would be +/-1.1% of 50g = 0.55 g. This
error in terms of output voltage for a 25 g analog accelerometer would
be +/-1.1% of span = +/-1.1% of 8 V = 88 mV. This
same error in terms of output frequency for a 25 g digital
accelerometer operating with a 250 kHz clock would be +/-1.1% of span =
+/-1.1% of 250 kHz = +/-2.75 kHz.
|
| Cross
Axis
Sensitivity |
A
percentage
of the level of applied acceleration in a direction that is 90 degrees
to the accelerometer’s sensitive axis. If,
for
example, a 5g acceleration is applied at 90 degrees to the sensitive
axis, the output would measure no greater than 3% of 5g or 0.15g.
The majority of cross axis error stems
from
the sense element structure not being perfectly parallel to the bottom
of the LCC package. This happens
due to the lack of
perfect parallelism from three sources; the LCC case's cavity floor vs
its bottom surface, the sense element die's bottom surface vs the LCC
case's cavity floor separated by an epoxy layer, and the sense element
die's top & bottom surface finish.
We only measure cross axis sensitivity during qualification testing but
it is controlled by the specification drawings for the piece parts that
go into making the accelerometer. The
ceramic
substrate drawing specifies a 1 microinch Ra surface finish.
The LCC package drawing specifies .002"
flatness for the die attach surface and .003" flatness for the bottom
surface. The die attach epoxy
adhesive used to
secure the sense element die to the LCC package cavity contains 2%
glass spacer beads. Each die is
pressed towards the
cavity surface so that the die seats on the beads which provides a
reliable .0029" thick bond line.
|
| g |
A
unit of
acceleration equal to the standard value of the Earth's gravity.
1g = 9.8085 m/s2 or 32.180 ft/s2
|
| General
Accuracy |
Determined
by
you for the operational range you intend to use the device over. Each
error source's contribution must be taken into account. The errors can
then be summed up but this may yield a pessimistic view of the overall
error because these individual errors are non-correlated (one does not
affect the other). Alternatively, you can calculate the total error as
the square root of the sum of the squares of all the error sources.
|
| Non-Linearity |
A
percentage
of span and is the deviation of
the accelerometer
output from its least-squares-fit scale
factor
(sensitivity) line slope. Measurements
are limited
to -90% to +90% of full scale or from -65g to +65g, whichever range is
smaller.
|
| PSSR
|
Power
Supply
Rejection Ratio: A measure of the variation in the
accelerometer's bias shifts as its power supply voltage is varied.
|
Scale Factor
(or Sensitivity) |
The
ratio of
the change in output to a unit change in the input acceleration.
Since the output of most accelerometers
is
slightly non-linear, the scale factor value is defined as the slope of
the least-squares-fit line to the acceleration input vs output curve.
Our measurements are limited to -90% to
+90%
of full scale or from -65g to +65g, whichever range is smaller.
For our analog output accelerometers,
the
units for scale factor are in millivolts per g.
For
digital output accelerometers, the units for scale factor are in counts
per g. We measure the output at 33
total points, 16
between 0 and +90% of full scale, 16 between 0 and -90% of full scale
and at 0g. The Scale factor is the
slope of the
least squares fit line for these 33 points.
|
| Scale
Factor Error |
The
percentage difference between the slope of a least squares fit line to
the measured (actual) scale factor curve and the slope of the ideal
scale factor line. A +/-2% error for a 50g full scale analog
accelerometer would be +/-2% of 8V/100g = +/-2% of 800 mV/g = +/-16
mV/g. A +/-2% error for a 5g digital accelerometer operating with a 250
kHz clock would be +/-2% of (250 kHz)/10 g = +/-2% of 25 kHz/g = +/-500
Hz/g.
|
| Scale
Factor Temperature Shift |
Expressed
as
ppm per deg C. For
example, the percent shift
in scale factor that would occur for a device with a +300 ppm per deg C
rating and a 60 deg C rise from room temperature would be:
+300
/ 1,000,000 x (85C - 25C) x 100% = +1.8%. For
an
analog 10g device, the scale factor would rise from its nominal (8
V)/(20 g) = 400 mV/g at +25C to 400 mV/g +1.8% = 407.2 mV/g.
For a digital 10g device operating with
a 250
kHz clock, the scale factor would rise from its nominal (250 KHz)/(20
g) = 12.5 kHz/g at +25C to 12.5 kHz/g + 1.8% = 12.725 kHz/g.
|
| Span |
The
magnitude
of the overall measurement range of the accelerometer from its minus
full scale to plus full scale. For
example, our 5g
full scale accelerometers have a 10g input span (from -5 to +5g).
The span can also be expressed in terms
of the
output where the Sensitivity (or Scale Factor) of the device is used to
convert between input span and output span. A
5g
analog accelerometer has an output voltage span of 800 mV/g x 10g or
8000 mV. A 5g digital accelerometer
has an output
span of 25 kHz/g x 10g or 250 kHz (assuming a 250 kHz clock is being
applied).
|
| Thermal
Transient Error |
Occurs
when a
drastic change in operating temperature results in a larger change in
output than would be predicted by the device's bias and scale factor
temperature coefficients. Piezoelectric
type
accelerometers are subject to this error because piezoelectric crystals
develop a voltage across them if the temperature change rate is within
its bandwidth. Silicon Design's
capacitive
accelerometers are insensitive to this type of error and see at most a
.1% of span change during drastic thermal changes.
|
