ADXL345 Sampling Rate

So I found this on SparkFun and I'd like to use it to monitor vibration on a machined part. Max frequency I expect to see is 1500 Hz and the ADXL can go to 3200 Hz internally but I'm concerned about the processor. I'd be using an Uno to poll the registers and log the data to an SD card, which has some unknown number of instructions/cycle and the Uno has only 16 MHz.

Is logging at 1500 Hz realistic or am I crazy?

Sorry, I realized that I forgot to include the links to the product I was talking about:

The accelerometer's page is here: https://www.sparkfun.com/products/9836 The datasheet is here: https://www.sparkfun.com/datasheets/Sensors/Accelerometer/ADXL345.pdf

Also, I only really need to log one channel (the "Z").

Basically I'm just curious if 1500Hz is closer to "will require some work with the code" or "dude, you're crazy."

Thanks!

Max frequency I expect to see is 1500 Hz and the ADXL can go to 3200 Hz internally

You're awfully close to the Nyquist limit there then.

Right, but only 3000Hz is required to give me basic coverage for anti-aliasing, right?

They have a single-axis sensor that has a much higher internal sampling but it appears to have a 400 Hz filter: https://www.sparkfun.com/products/9332

Well the ADXL345 datasheet recommends using SPI at 2MHz (5MHz max) or faster for the 3200Hz output rate , so if you use a 4MHz SPI clock I'd think you'd be fine, especially reading only the two Z registers and not all six.

Further, the ADXL345 has a FIFO buffer so you can check the status register (to see how many reads you need to do) and then do a burst of repeated reads of the Z registers (with 5uS between) to drain the FIFO. The FIFO will store up to 30-ish samples as you process the data and send it to SD etc.

The SD card should be able to run at 8MHz so you can flip the SPI speed back and forth between 4MHz for the ADXL and 8MHz for the SD card.

I don't know what kind of overhead there is in the SD card FAT driver, there might be a bit of delay there, but this seems to be within the realm of possibility.

Hmm, sounds like I should get this home and try it out...

I found a Freescale sensor that would be much more work to implement but has a slightly higher rate: http://cache.freescale.com/files/sensors/doc/data_sheet/FXLC95000CL.pdf?fpsp=1&Parent_nodeId=1366833099850701025628&Parent_pageType=product

See any major differences that would make this one worth it?

1500 Hz is rather low for vibration analysis, but you have to start somewhere. and yes you only need one direction. and after that a FFT to find the peaks, so that will be a hard thing for an atmega.

pgm575:
So I found this on SparkFun and I’d like to use it to monitor vibration on a machined part. Max frequency I expect to see is 1500 Hz and the ADXL can go to 3200 Hz internally but I’m concerned about the processor. I’d be using an Uno to poll the registers and log the data to an SD card, which has some unknown number of instructions/cycle and the Uno has only 16 MHz.

Is logging at 1500 Hz realistic or am I crazy?

Sorry for bring this up…

As i am also planning to use the ADXL345 for vibration analysis.

Can someone please kindly explain to me what is this frequency you guys are talking about? What and how to set the most ideal frequency to use? I am currently really confused now about natural frequency/ sampling frequency / spi clock frequency and frequency output rate ??

I basically uses the adxl345 to measure the acceleration. I am required to convert it to velocity as per the ISO 10816 standards, however according to the formula, i need find the frequency at this particular instance, which i am have been trying to figure out how to get, before going into FFT.

My lecture was also saying how adxl345 isnt a legit accelerometer for frequency analysis?