Question about LIS3LV02DL

[sarling]

Posted on July 07, 2009 at 23:40

Question about LIS3LV02DL

[sarling]

Posted on May 17, 2011 at 13:16

Hi,

this is off-topic but I don't where to post, so I give it a try here. I have an evaluation kit for the accelerometer LIS3LV02DL connected to an STM32. I'm pretty new to accelerometers and I have noticed that there's slight difference in gain (and perhaps offset) on each axis, which produces erroneous results. x-axis reads 16240 when perpendicular to gravity, y-axis 16480, z-axis 16033. Can anyone point me in the right direction on how to compensate and/or calibrate for these errors? Is it something wrong with the device or is it common with some deviation? I understand that you could multiply each axis value with a constant and add a constant to make all axis the same, but I want to know how You do this. Also a bit confusing when the datasheet says it's factory trimmed and no further calibration is usually needed. Kind regards, Sarling

[jj]

Posted on May 17, 2011 at 13:16

Have not used your specific device - have success with other accels.

Questions: a) Readings when ''perpendicular to gravity.'' Have you a calibrated (usually optic-type) x-y-z table? If not - how do you determine that you are perpendicular. How physically stable is your set-up? And how does the output vary over time (5, 10, 30 secs?) b) Several of ST's mems accels. use their ''thelma'' process - which provides basic physical ''likeness'' in the X & Y planes. The Z plane is impemented somewhat differently - thus I would expect there to me greater difference between the X-Z, Y-Z, than X-Y outputs. c) Understand you have an eval board - we are always more comfortable/confident when we can compare multiple devices - highlights misunderstandings or device ''irregularities'' quickly. In our design we purchased a ''deluxe'' accelerometer - and ran our lower cost devices in parallel. We measured and logged both dynamic and static acceleration - carefully noting where/how the deluxe device differed from our lower cost ones. In our case - we were able to individually calibrate and compensate our low-cost devices to achieve almost equal performance to that of the much more costly deluxe. Hope this helps - good luck...

[jj]

Posted on May 17, 2011 at 13:16

A source for optic/mechanical exercise/experimentation is www.edmundoptics.com.

We needed to devise a means to generate ''repeatable accelerations'' - so that we could compare the outputs of deluxe and inexpensive devices. We employed gravity and solenoid-driven methods to produce accelerations. By ''tuning'' the mass of our transport we could tweak accelerations. Capturing the data is best/easiest with a fast, multi-channel storage scope. We implemented our set-up w/in a small temperature chamber - so that we could observe temperature sensitivity. Again - a 2nd eval kit or several sensors on your own pcb will speed and eliminate ''single unit'' irregularities... Good luck...

[sarling]

Posted on May 17, 2011 at 13:16

Hello,

I have just used a water level to get it near perpendicular, the maximum reading does not change much when the PCB board's normal vector deviates slightly from gravity. I read values from the accelerometer 20 times per second. I would say the output vary about +/- 3 units for each axis, this is the raw data, unfiltered. I'm not really sure about how to calibrate these devices, what kind of rig is needed and so forth. There are some factory trimmed values for offset and gain in six registers that are reloaded at boot. I don't know what an optic xyz table is, sounds expensive, do you have a link to one? Kind regards, Sarling

[obtronix]

Posted on May 17, 2011 at 13:16

Quote:

On 05-07-2009 at 23:39, Anonymous wrote: Hi, this is off-topic but I don't where to post, so I give it a try here. I have an evaluation kit for the accelerometer LIS3LV02DL connected to an STM32. I'm pretty new to accelerometers and I have noticed that there's slight difference in gain (and perhaps offset) on each axis, which produces erroneous results. x-axis reads 16240 when perpendicular to gravity, y-axis 16480, z-axis 16033. Can anyone point me in the right direction on how to compensate and/or calibrate for these errors? Is it something wrong with the device or is it common with some deviation? I understand that you could multiply each axis value with a constant and add a constant to make all axis the same, but I want to know how You do this. Also a bit confusing when the datasheet says it's factory trimmed and no further calibration is usually needed. Kind regards, Sarling All mems accelerometers are pretty crude devices, for your device the spec says a 1 g input will yield an output between 14720 and 18016 counts at room temperature at Vdd= 2.5 volts and then change with temperature at a .025%/degrees C rate (typical). Changes in Vdd will also change the output. (this is all after the factory trimming) The amount of change will vary from part to part, so calibrating is a very expensive task.

[sarling]

Posted on May 17, 2011 at 13:16

Thank you both for your detailed answers!

/Sarling

[jj]

Posted on May 17, 2011 at 13:16

''so calibrating is a very expensive task...''

Depends - in our ''limited'' case we sought to replace an existing device which cost 17-20x more than the simpler accel. We had 10 channels of fast, accurate analog data collection - devoted 2 to the expensive devices and ran 8 ''lo-cost'' devices thru excitation simultaneously. Set-up by engineers but then run by tech - we were able to plot & tie data to each sensor. A one-time calibration program enabled us to approach the accuracy/performance of the higher cost device - and we were able to sell both our micro and accel. at above target profit - and save client money too. The plunge in cost of small, powerful micros opens many such opportunities. Would make little sense - as you suggest - in ''mini-volume'' but we won adequate commitment and were able to reuse many elements of this technique/set-up for other projects...