a
Mounting Considerations for ADXL Series Accelerometers
by Mike Shuster, Bob Briano and Charles Kitchin
ONE TECHNOLOGY WAY
•
P.O. BOX 9106
•
NORWOOD, MASSACHUSETTS 02062-9106
•
617/329-4700
As with all accelerometers, optimum performance
depends on proper mounting of the device. It must
be mounted so that the sensor is properly coupled to
the object for which acceleration is to be measured.
Also the effects of mechanical resonances must be
minimized.
All mechanical structures, no matter how ”solid,“ have
mechanical resonances that result in an output signal at
the resonant frequency that is larger than expected. Me-
chanical resonances cannot be eliminated completely.
The key to successfully reducing any serious problems
is to keep the resonant frequencies above the frequency
band of interest or to attenuate them to acceptable lev-
els. Structures that seem benign at rest often mechani-
cally resonate when vibrated at high frequency.
The ADXL Series of accelerometers are intended for PC
board mounting, and careful mechanical design of the
PC board is important. Resonances can be difficult to
determine before prototyping, but some precautions
can be taken.
1. Replace or firmly attach any individual components
that are flexible such as large through-hole compo-
nents. Surface mount resistors and capacitors are
preferred because they will not vibrate on their leads.
2. Wire-wound resistors and electrolytic capacitors can
change values when they are vibrated. Again, ce-
ramic surface mount components are best.
3. Mechanically couple a potentially problematic struc-
ture to one that has a high resonant frequency. An
example of this is to firmly attach PC boards to thick
bases.
4. Couple together several structures such that the com-
bination has a dampened response: PC boards, card
cages, etc.
5. Cables inside moving structures should be tied down
firmly. Cables between a moving structure and a sta-
tionary one should be flexible and have a service
loop.
AN-379
APPLICATION NOTE
6. Cavities, such as housings, can resonate. Filling cavi-
ties with suitable potting materials or epoxy is a good
solution for this, and also prevents components from
vibrating.
7. Keep all dimensions as small as possible because, in
general, smaller dimensions mean that the reso-
nances are at higher frequencies.
8. Use silicon grease or other materials on mating sur-
faces to improve coupling between components that
are screwed together.
Many applications do not have strict requirements for
being resonance free, but in some cases resonances can
be critical to a successful system. The best way to test a
design for resonances is to shake it, using a calibrated
shaker, over the frequency band of interest using a refer-
ence accelerometer attached at critical points.
Figure 1 illustrates the ideal response of the ADXL50
which is achievable with proper fixturing or potting ma-
terial. The sensor response is flat from dc to the
dominant pole set by the demodulator capacitor, typi-
cally at 1.3 kHz. Figure 2 shows the response of the
ADXL05.
NORMALIZED SENSITIVITY – dB
FREQUENCY – Hz
6
–21
1 10k10 100 1k
3
0
–3
–6
–9
–12
–15
–18
–24
Figure 1. Ideal Response of the ADXL50
–2–
PRINTED IN U.S.A. E1895a–5–5/95
FREQUENCY – Hz
9
6
–21
3
0
–3
–6
–9
–12
–15
–18
1 10k10
NORMALIZED SENSITIVITY – dB
100 1k
Figure 2. ADXL05 Ideal Response
(This can be varied by changing the demodulator capaci-
tor. See application note AN-377. The sensing beam has a
natural resonant frequency at 24 kHz.
Using the ADXL50 as an example, possible solutions for
mechanical resonances will be illustrated. (The response
of the ADXL05 will be very similar.)
All of the following measurements were made using an
aluminum cube designed to hold a PC board which
mounts directly to a shaker.
If the accelerometer is mounted to a PC board by soldering
the leads only, then it will resonate in a complex manner
in the range from 7 kHz to 9 kHz. This is caused by the
accelerometer‘s package moving on its leads with respect
to the PC board. Figure 3 illustrates a typical response
from the ADXL50 when soldered to a PC board. The ampli-
tude and frequency will vary depending on how it is held
when soldered.
FREQUENCY – Hz
190
19
0.19
100 1k 10k
1.9
SENSITIVITY – mV/g
Figure 3. Response of the ADXL50 when Soldered to a
PC Board
The resonance at 7 kHz will not affect the response of the
ADXL50 over its typical rated bandwidth of 1.3 kHz. How-
ever, if signals at 7 kHz are potential problems, then low-
pass filtering or better mounting methods may be taken.
By applying a small drop of epoxy under the seating plane
of the accelerometer‘s package, firmly attaching it to the
PC board, the resonant peak will shift to a higher fre-
quency (approximately 12 kHz) as shown in Figure 4. This
example illustrates the effectiveness that simple changes
can have on this or any other resonating component.
The ultimate solution to the problem of vibrating compo-
nents is to pot the entire board and all of its components
with a suitable material. Figure 5 demonstrates that when
encapsulated with paraffin wax (used because it is easily
removable) the resonant peak is pushed out in frequency
to approximately 20 kHz. Other more permanent potting
materials can further reduce or eliminate the resonances.
The ADXL50 when properly mounted will not likely be the
limiting component. Many other structures can resonate
in a complex subsystem in the range of a few kHz if you are
not careful.
FREQUENCY – Hz
190
19
0.19
100 1k 10k
1.9
SENSITIVITY – mV/g
Figure 4. Response of the ADXL50 when Soldered and
Epoxied to a PC Board
FREQUENCY – Hz
190
19
0.19
100 1k 10k
1.9
SENSITIVITY – mV/g
Figure 5. Response of the ADXL50 Potted with Paraffin
Wax