Teensy 3.1 in combination with Digital & Analog Accelerometer

Status
Not open for further replies.
C

Chopsticks

Active member
Hi there!

I've started a new hobby project where I want to test different kind of accelerometers in combination with my teensy 3.1.
Now i've seen lots of guides of how to select a suitable accelerometer.
But i'm getting stuck on the sensitivity part.

I see different kind of notations, such as:
4096 counts/g
1 mg/LSB
740 - 860 mV/g
1 - 3,9 mg/digit

what does this all means? And what is important to know when choosing a very sensitive accelerometer. For example, is 1024 counts/g better or worse compared with 4096 counts/g. Or is 10 mg/LSB better/worse then 1,h/LSB etc.

Any kind of help is welcome! Would appreciate it. :)

greetings!

Chopsticks
 
Chopsticks said:
Hi there!

I've started a new hobby project where I want to test different kind of accelerometers in combination with my teensy 3.1.
Now i've seen lots of guides of how to select a suitable accelerometer.
But i'm getting stuck on the sensitivity part.

I see different kind of notations, such as:
4096 counts/g
1 mg/LSB
740 - 860 mV/g
1 - 3,9 mg/digit

what does this all means? And what is important to know when choosing a very sensitive accelerometer. For example, is 1024 counts/g better or worse compared with 4096 counts/g. Or is 10 mg/LSB better/worse then 1,h/LSB etc.

Any kind of help is welcome! Would appreciate it. :)

greetings!

Chopsticks
Click to expand...

First, the mV/g value is the analog sensitivity, usually defined for a given current (mA that run though the accelerometer bridge). To this you have to add the ADC characteristics. If you use , say an 10 bit ADC with 1 V saturation voltage (sometimes also called Vref) then the LSB would indicate about 1 mV. If your accelerometer gives 750 mV/g then your digital saturation level (i.e.full scale at 1 V) would be 1.5 g and the total (digital) sensitivity would be 1.5 mg/LSB or 1.5 mg/digit.

In theory 4000 counts(digits)/g would be 4 times better than 1000 count/g, IF you are NOT noise limited. This depends on the quality of the accelerometer, the choice of the instrumentation amplifier and ADC. You always can average to obtain better noise statistics.

By the nature and original purpose of the widespread accelerometers, you cannot expect miracles. They are made to detect orientation of your screen, detect falling devices etc. Having that that, digital compasses, similar to the one suited to be used with teensy, are already a good compromise between performance and price.
 
What accuracy and operating range do you actually need? This is important when picking out one that is "suitable" for your application, in contrast to "very sensitive". "Very sensitive" can mean both: not sensitive enough or total overkill.

Example 1: You want to measure the orientation of a device that is not moving. You then need a range of +- 1g, and probably 8 usable bits of accuracy. A 10-bit digital output accelerometer will be enough for that. You can also always calibrate (you should do that anyway to eliminate orientation errors due to mounting). 10 samples per second is probably enough.

Example 2: You want to measure the acceleration of a model rocket. In that case, you need an accelerometer that is fast and has a wide range. The numbers depend on your rocket, and you should start with an estimation of the maximum expected acceleration by applying newton's law F = m a where F is the thrust generated by the rocket engine, m is the rocket's mass and a is the acceleration.

So what's your application?
 
christoph said:
What accuracy and operating range do you actually need? This is important when picking out one that is "suitable" for your application, in contrast to "very sensitive". "Very sensitive" can mean both: not sensitive enough or total overkill.

Example 1: You want to measure the orientation of a device that is not moving. You then need a range of +- 1g, and probably 8 usable bits of accuracy. A 10-bit digital output accelerometer will be enough for that. You can also always calibrate (you should do that anyway to eliminate orientation errors due to mounting). 10 samples per second is probably enough.

Example 2: You want to measure the acceleration of a model rocket. In that case, you need an accelerometer that is fast and has a wide range. The numbers depend on your rocket, and you should start with an estimation of the maximum expected acceleration by applying newton's law F = m a where F is the thrust generated by the rocket engine, m is the rocket's mass and a is the acceleration.

So what's your application?
Click to expand...

The application would be a simple office chair, i just want to detect the slightest movements, how people sit on their cheir etc. So the application needs to measure a change between ±15 degrees.
 
Oops, I didn't notice your answer up there, sorry. Well, a person in an office chair is constantly moving, but with low acceleration. I don't think I could push my chair around with my legs and apply more than 1g. This might be different for chairs that are pulled around in a room, but in that case you're probably not interested in the data anyway (or are you?).

The LIS3DH has three axes and can handle +/- 2g minimum. It should really be sufficient. Sample at 50 Hz or so and you've got enough data for detecting movement and static orientation.

You're asking for alternatives - what are you missing in your current choice? Is it not capable of doing something you are asking it to do?
 
Chopsticks said:
Hi there!

I've started a new hobby project where I want to test different kind of accelerometers in combination with my teensy 3.1.
Now i've seen lots of guides of how to select a suitable accelerometer.
But i'm getting stuck on the sensitivity part.

I see different kind of notations, such as:
4096 counts/g
1 mg/LSB
740 - 860 mV/g
1 - 3,9 mg/digit

what does this all means? And what is important to know when choosing a very sensitive accelerometer. For example, is 1024 counts/g better or worse compared with 4096 counts/g. Or is 10 mg/LSB better/worse then 1,h/LSB etc.

Any kind of help is welcome! Would appreciate it. :)

greetings!

Chopsticks
Click to expand...

Actually the mV/g value is the analog sensitivity, defined for the current which you are using. You must know about the ADC characteristics. If you use , say an 10 bit ADC with 1 V saturation voltage (sometimes also called Vref) then the LSB would indicate about 1 mV. If your accelerometer gives 750 mV/g then your digital saturation level (i.e.full scale at 1 V) would be 1.5 g and the total (digital) sensitivity would be 1.5 mg/LSB or 1.5 mg/digit. The use of 4000 counts(digits)/g would be 4 times better than 1000 count/g, if noise is not the problem in your case. This depends on the quality of the accelerometer you are using, the choice of the instrumentation amplifier and ADC. You always can average to obtain better noise statistics.


pcb prototypes
 
Last edited:
Status
Not open for further replies.
Back
Top