FAQ - Accelerometer

320 - The accelerometer seems to be too noisy

We test every accelerometer before it leaves the factory, and it must pass all of our specs, including noise, before we will ship it. Remember, we specify an RMS noise figure. This is an average of the noise - peak to peak values will be much higher. If the sensor still seems noisy, check that you are supplying the correct voltage and are looking at the correct signals. Also check for environmental factors that could be adding noise to your system; for example, a nearby motor that is making your workbench vibrate.

363 - Whenever I move the DMU forward with a positive acceleration,the graph shows at first an overshoot of negative values and then follows an overshoot of positive values and oscillates around a value. What I do not understand is this first downshoot. Why is this happening, while I move the unit towards the positive direction of x-axis/ y-axis?

The DMU uses a non-standard definition of acceleration sign. With the unit sitting upright on your desk, the output will read (0, 0, +1) g. If you think about this, when you move the unit down (in the positive direction of the HDX z-axis), the unit will be in freefall, and go to zero g. So that is actually a negative acceleration, although according to the DMU z-axis, it should be positive.

You should just multiply all of the acceleration outputs by -1 to make them agree with standard definition of acceleration.

Second, when you move things by hand, you will not get smooth acceleration curves. Your hand/eye works by "position" feedback, not acceleration, so the acceleration curves will look odd. Also, even if you are sliding on a fairly smooth surface, you will see "jerks" in the acceleration. Try to integrate the acceleration readings to get velocity or position to see if that looks more familiar.

The accelerometer is just measuring what acceleration is present, so even though it looks funny, that is what you did to the device.

364 - I have mounted the IMU on a terrain mobile robot, and I get very noisy measurements for acceleration.Even when the robot stands still, the measurements show values of around -3 up to 3 cm/sec2,jerks and generally not a good profile. From your experience, is there some kind of filtering that should be applied in order to improve the acceleration profile ?

Are you using the analog outputs, or the digital outputs? If you are using the analog outputs, you may be getting some sort of electrical noise into your circuit that makes the data look noisier than it really is. If you are using the digital data, then that is probably what the robot is really doing in terms of acceleration. We filter the acceleration internally to 10Hz. We do this because we are using the accelerometers primarily as a vertical reference. Depending on what use you have for the accelerometers, you may want to add some additional filtering (to see long term motions) or may want to keep the information to see all of the motion of the robot.

369 - We are using the DMU to detect the angle position of a model helicopter. The DMU is placed on the helicopter. We use the analog signals pin 12 and pin 13. After starting the motor (without a take off) the sensor signals are drifting. So we can NOT depend on the sensor signals. We figure it's because of the noise (the vibrations). Is this normal or shoudn't it be corrected by the DMUÕs accelerometers? Is it possible that the vibrations will be integrated by the DSP? How can we filter it? The 'normal' functions without the vibrations are ok. Have we done something wrong?

The accelerometers make no distinction between the vibration from the motor or the real acceleration. We have 100Hz low pass filter on the outputs and hence any high frequency noise should be filtered out.

I would suggest you to measure the frequency of the noise and then using low pass filter, you should be able to filter it out. If it is actually caused due to the vibrations, try to isolate them.

378 - Can you explain the output of the ADXL202EB-232 evaluation board. The documentation states that four bytes are received from the board, 2 x and 2 y. What do each of the bytes represent? Should the two bytes be concatenated to get a signed acceleration value in g, or do the two bytes represent something else?

Take a look at http://products.analog.com/products/info.asp?product=ADXL202,
and go to the evaluation tools tab. You should be able to download a data sheet that explains this in detail. In short, the data sent is the duty cycle of the PWM signal output by the ADXL202.

379 - The system I am attaching the accelerometer to has a differential input. According to the data sheets I have, the accelerometer has only one output. Would you recommend connecting the negative differential input of our system to the ground of the accelerometer?

The output from the accelerometer is single ended. You will need to ground the negative differential input of your data acquisition system. You should plan in some flexibility into where you ground it, as this can make some difference to the noise. You do not want to set up ground loops. Usually, you would have just the signal wire going from the sensor to your data acquisition device, and ground the negative input to the power supply ground.

380 - In the spec sheet, there's a resolution column for gyro and accel. The bias stability @ 25C is +/- 1 deg/sec, so if I convert it to deg/hr, it's 3600 deg/hr which is very huge a data, but if I mulitply it with the resolution figure, I get 180 deg/hr which is more like it. Do I actually need to multiply all my readings with this resolution figure for gyro and accel ? If not, what is the resolution figure use for ?

Bias stability is a funny number. In general, it is trying to describe the amount of variation you would see in the bias of the rate sensor signal over time or temperature. Generally, one will see different amounts of variation depending on how long one waits between measurements. The number quoted above is really meant to show that if you take a measurement of the bias one time, and come back any time after that (1 minute, one week, one year), your new bias measurement will be within 1 deg/s of the original measurement.

Resolution is the smallest amount of change you can measure with the sensor. It is related to the noise level. In fact, we typically quote the resolution as the standard deviation of roughly 1 minute of data collected from the sensor.

As you can see from the above, bias stability and resolution are independant measures of the sensor performance.

381 - I'm investigating the use of a tilt meter on a railway trolley to measure the superelevation ( Cant ). The trolley is towed at speeds of upto 30Kmh on tracks with curve radii of min 100m. Is your tilt sensor affected by centifugal force, accelerations and decelerations?

All of our tilt sensors are based around accelerometers. We output tilt assuming that the only acceleration we are measuring is from gravity. (This is true for all tilt sensors.) In your case, the centrifugal acceleration will affect the measurement. At 30 Km/hr and 100 m radius turn, this would correspond to an error of about 4 deg.

We make a different kind of sensor for dynamic applications. Look at our website http://www.xbow.com/Products/productsdetails.aspx?sid=29
This is a solid state vertical gyro. It uses a combination of angular rate sensors and accelerometers to distinguish between cases when the sensor is actually rotating or tilting and a pure acceleration. You can download a PDF data sheet and description from the website.

384 - We bought a CXL02TG3-S accelerometer. We tried now to make a position measurement system as you described in your Tech-Forum at the point inertial Navigation. So we integrate the analog output signal twice with the trapezoidal-rule. We recognize, that the Zero G voltage is not constant. So we filtered the DC-Offset with a special DC-Estimator in Labview. But the DC-Offset gives us a too big error. How would you propose to compensate the DC- Offset before integration. Is there an Application note available?

The measurement of position inertially is a very difficult task with a reasonable accuracy. The reason being that the bias/offset variation would result in large error build-ups over short period of time. You need to have a long term reference (such as GPS) to correct for these errors.

The offset/bias varies both with temperature and the supply voltage. The variation over the temperature can be compensated using the temperature sensor that is built in the CXL02TG3 sensor. You will need to run the unit through different temperature and measure the bias and then use the best fit curve to compensate these variations in bias. Unfortunately, we don't have any application notes available on how to perform this.

385 - We have been experiencing problem on HF Series of accelerometers. On several occasions the zero-G voltages of these accelerometers are shifted from their original zero-G voltages given above and stay at the shifted values for very long intervals. At times, they return to their original values and then again shifted. Could you please inform us as to the reasons for this shift and how can we resolve this problem?

The Offset voltage of the HF series accelerometers vary with the variation in supply voltage and ambient temperature. The 0G offset doesn't mean anything in HF series accelerometers, because these don't have a DC response. You should be able to obtain the absolute aceleration by subtracting the running average from the accelerometer signal.

403 - I am studying the G forces at impact on the metal and was wondering one thing. Will the orientation of the accelerometer affect the results recorded.? How do you compute the span of the accelerometers? Would each be +5V for the positive value and 0V for the negative extreme?

Yes, the orientation of the accelerometer when it makes contact would affect the readings. If you want to measure the forces on impact, you should make sure that the sensitive axis of the accelerometer is perpendicular to the surface of impact.

The span of accelerometers is usually between 0.5V (negative extreme) to 4.5V (positive extreme) centered around 2.5V. For exact values at these maximum values, you need to refer to the calibration data sheet for individual sensors. You can estimate this by:
Output = Offset +/- Sensitivity * Span

404 - We have some problem with the analog output of the IMU400. The outputs of the accelerometers are available at pin numbers 5 to 7. The voltage at these pins varies between 0 to 5 Volts. Now problem is that when we connect IMU in z-down or x-down position then voltage at pin 7 or 5 increases from 0 g value,i.e, if voltage at 0 g is 2.536 Volts then at 1g this value is 2.710 and vice versa. But when we connect IMU in y-down position then voltage decreases from its 0g vlaue. This behavior is completly opposite to one that is observed in z-down and x-down position. is it true? if it is true then i want to know the calibration equation for x,y and z-axis. I have the calibration parameter sheet. Other problem is that accelerometers output at pin number 5 to 7 varies between 0 to 5 volts. But for gyros the output voltage at pin 8 to 10 varies between -5 to 5 volts. is it possible that this voltage at pin 8 to 10 also varies between 0 to 5 volts?

We recommend that you use the compensated outputs for rate (pins 8, 9 & 10) and acceleration (pins 12, 13 and 14). You will find that these outputs all match the sign convention described in our user manual. This is also true if you are using the digital outputs.

The IMU400 is set up so that when the accelerometer sensitive axis is oriented towards ground, the output is more positive. However, this means that when you accelerate the unit along the sensitive axis, the output will actually be more negative. This is a convention used by Crossbow and may not match normal SAE or IEEE definitions.

Pins 5,6 and 7 provide the outputs directly from the accelerometers and are 0 to 5V levels. These outputs do have one axis that is "backwards" relative to the other two. This is due to the orientation of the sensor within the IMU. The compensated outputs correct for this.

All of the compensated outputs are generated by a D/A converter and have a full range of +/- 4.096V. The is no option for a 0 to 5V volt output.

405 - Currently this version of the LP accelerometer comes with a rather long cable. We are trying to place this acclerometer into a fairly compact package and this cable is just taking up space. Is it possible to order a similar accelerometer with a much shorter cable? Or could you recommend a possible means to rig something up, because the wires are so small.

The standard length of cable on our accelerometers is 8 ft. I don't believe we have anything with shorter cable. If you order one with short cable, it would make it a custom product and will involve a significant NRE associated with it. It will be worthwhile doing if there is a volume requirement. Otherwise, I would recommend you to cut off the connector and use pigtails on the ends.

406 - I have an ADXL202EB-232A. I'm trying to do a 2 small projects using the Crossbow (X-Analyze) Software. What I am trying to do is calculate the velocity and distance if I'm on an elevator going up and down. Does Crossbow/X-Analyze do this? If so, how does it work?

The X-analyze won't let you calculate the velocity and displacement from ADXL202EB-232A. In order to do this, you will have to write your own software application.

We have a sample LabVIEW code that explains how to communicate with the device and interpret the data packet. This is located under the Software folder in the CD.

408 - We have two LF series Accelerometers, which is now connected to our data aquisition system. However, when I looked at the software provided "ACCLE-VIEW" in your website, it shows the CXL can be connected to COM1 directly? How does it work? What kind of interface card should I order in order to use this software? Can I use two LF Series accelerometers at the same time and log the signal from both channels (buy two interface card)?

You can not connect our accelerometers directly to the Serial port. In order to use the LF series accelerometer with AccelView, you need the CXLDK interface card. Please refer to http://www.xbow.com/Products/Product_pdf_files/Accel_pdf/CXLDK_Datasheet.pdf for more details.
CXLDK can take up to 4 input channels. If you have single axis LF accelerometers, you can use 2 at a time, if tri-axial modules, you would need 2 CXLDK interface cards.

410 - It mentions in the applications notes that an accelerometer can be mounted with adhesives. I need to mount the accelerometer (aluminum body) to various points on a stainless steel frame. Is there a particular type of adhesive that I should use? Is double-sided tape acceptable? How about hot glue?

The type of the glue is very much application dependent. You can use the double-sided tape as long it does not cause the relative motion between the accelerometer base and the mounting surface. We do not really have a glue that we can recommend. There are also mounting wax available that you can consider.

413 - checking the DMU accelerometers it appears that X and Z axis are oriented 180 degrees out. I expect a + voltage gain when the axis arrows are at 90 degrees up in reference to the ground and - voltage gain arrows are at -90 degrees towards the ground. The Y axis responds correctly in this situation but the voltage in the X and Z axis are going doing exactly the opposite. Any thoughts on why the unit does not react in the same way for all the axis?

Are you referring to the analog voltage output from the DMU? What pins are you reading these outputs across? If from Pin 5, 6 and 7, these are taken directly from the accelerometers and the sign convention of these is dictated by the mounting constraints of the accelerometer.

However, if you either refer to the accelerometer outputs in digital signal or analog voltage from Pin 12, 13 and 14 in Scaled mode, these should be in agreement with the co-ordinate system convention on the DMU label.

423 - We are running the cable from the CXLDK interface board to the CXL accelerometer a distance of 75m. The cable is shielded and grounded to reduce interference on the output signal. I want to check that the power in will be okay over this distance (since it comes from the board).

As long as you have properly shielded the cable to eliminate the EMI you should be fine. In the past we have tested them to be fine for cable lengths up to 300ft. Make sure that the overall load impedance at the other end of cable is >10K ohm.

424 - We are designing signal conditioning for various sensors and we need to know the output impedance. Your specification sheet identifies Output Loading. It looks the same output impedance although I've never see it expressed along with capacitance in nF. Is it the same thing?

Output loading resistance (Rl) normally refers to the minimum load you should have on the sensor output to avoid creating voltage devider. The output loading capacitance (Cl) is the maximum capacitance you can have on the sensor output. If your load capacitance exceeds the recommended value, you may not get the full bandwidth of the device.

426 - I was connecting the accelerometer in a 'voltage mode' setup (i.e. individual connections for power and acceleration signals). From the CXL10HF3 spec sheet, it would appear that this board can be used to operate the accelerometer in "current mode". Can you verify this, and if so, define the connection (ie; for X-axis, use pin 3 and pin 2 (gnd))? Thanks.

The "current mode" version of CXL10HF3 has been discontinued some time ago. The unit you have should be just a "voltage mode" version. It takes 6-30VDC input (Pin 1 and Pin 2) and provides 0-5V output (Pin 3, 4, 5 and Pin 2).

431 - What is the accuracy of the specified low frequency roll-off point of the HF series accelerometers? The low frequency (-3dB) point for the HF accelerometers is specified as 0.3Hz. Do you have an accuracy figure for this? i.e. 0.3Hz +/-5% or so. In other words, from device to device, how much variability is there in the lower -3dB points?

Unfortunately, we do not have the accuracy specification for the low frequency. In general the frequency range for the sensor looks like:
(+/-5%) 1 to 4000 Hz
(+/-10%) 0.68 to 6000 Hz
(+/-3dB) 0.32 to 10000 Hz

437 - We are interested in your uniaxial accelerometers and also in your four channels CXLDK digital interface, but we are not sure if it is possible to connect four uniaxial accelerometers to the CXLDK digital interface. Do you have something like a junction box in order to do that?

It is possible to connect 4 uniaxial accelerometers to CXLDK, but you will have to manually wire the accelerometer outputs to the header pins of CXLDK. Unfortunately, we do not have any connector box that would let you do this, but it should not be hard to wire.

Memsic FAQ Accelerometer
Memsic FAQ Accelerometer