I've so far been using an ADXL 335 3-axis accelerometer and have spent a good amount of time researching data sheets and am hoping I'm just misinterpretting the specifications.
I'm working on a project that requires me to sample a wave at 11Khz. That is 360 points per cycle and the period of the cycle i'm measuring is 0.0324 seconds.
So assuming I need 360 points per 0.0324 seconds, that means each point is 9.90091 x 10_-9 seconds or 11099.87 Hz.
I've looked at several of Analog Devices' accelerometers but all 10Khz+ accelerometers have a poor mV/g
The datasheet for the 335 frequency response says that the X and Y axis have a bandwidth of 1600Hz. Meaning it can sample 1600 points in 1 second.
I think I have a misunderstanding of how to interpret these datasheets.
Can someone explain and/or recommend an accelerometer for my situation?
According to my own math, I'd need 11Khz sampling rate for a 30Hz signal to measure 360 points per cycle and I'm measuring on in single-digit g scale. So no more than 10Gs.
Why do you think you need to measure 360 points per cycle?
TheWinkleSnail:
According to my own math, I'd need 11Khz sampling rate for a 30Hz signal to measure 360 points per cycle and I'm measuring on in single-digit g scale. So no more than 10Gs.
For a signal with a bandwidth of 30Hz, you must sample it at no less than 60Hz. In other words, you don't need an 11KHz sample frequency to measure a 30Hz signal, only at least 60Hz. Look up the Nyquist Sampling Criterion.
Sorry i should have added some things.
We are measuring a 30 Hz sin wave for noise essentially. Noise within that wave to detect faults or uncharacteristic behaviour. I'm familiar with the Nyquist Sampling Criterion but is this just a task that requires much more than the standard accelerometer can provide?
Frequency I'm operating at: 30Hz
Period:
1/30hz = 0.0333 seconds
360 points per period:
0.0333 sec / 360 points = 9.25E-5
Frequency of that desired sampling :
1/9.25E-5 = 10800Hz.
I'm analyzing the wave very closely, not just confirming a waveform of 30Hz
What is the bandwidth and expected amplitude of the noise?
That's the problem. We're trying to detect all kinds of noise ranging from 0 to that 10Khz range. Amplitudes of the noise frequency would be much smaller in comparison. In only double digit milli-g's .
You MUST sample at a frequency at least twice the highest expected frequency of interest.
Furthermore you must make sure that there are no frequencies in the input that are higher than one half the sample frequency, or you will have "aliasing artifacts". So with a 10 kHz sample rate, your noise input must be somehow limited to 5 kHz.
If you are serious about this project, read up on the "sampling theorem" and input aliasing, and obey the strict rules, or the results will be useless.