V3 Sensor Time Domain Frequency Range Question

Just a quick question regarding the ODR and related frequency ranges on the V3 Vibration Sensor lineup.

This post has some of the detail, but I want to double check something:

If I choose 3200sps, and take the Time Domain Sample, and run it through my FFT Calculator, the output data, and then graph that.

My X axis is starting at 50Hz and ending at 800Hz right? all data lower than 50Hz is being removed entirely and should not be represented at all on the graph?

I’ve been reviewing other solutions to see how they do it, as far as the visualization. I see some sensors that show graphs from 10Hz to 2500Hz, and I don’t see how that is possible, seeing the list of ranges shown.

I believe its not just not possible, but mathmatically impossible to show valid data from 10Hz to 2500Hz unless you are stitching together several FFTs?

For a more targeted question, what determines the Upper and Lower Bounds of that Frequency sample, the Nyquist Theorem, right?


the selected range means the harmonics in this range will be unattenuated and trusted, but as you know , the nature of all filters is that they have a roll-off rate that attenuates the signals outside the bandpass range, but it “does not kill the signal” completely. So for example, if your range was 50 to 800 Hz, and there is a signal at 40 Hz, it can appear on your graph (with much attenuated amplitude though). We 're currently working on a new release that should minimize this effect further.

Thank you Bhaskar, so if I plot my chart from 50 to 800, I think I am doing it wrong then.
I am assuming the very first datapoint in the array is referring to that 50Hz signal.

I should plot from 0 to 800?

Yes, you can plot from 0-800. The data wont be true data for 0-50Hz ( it will be attenuated)

Thank you Bhaskar, is that graphing 0-800 statement, true whether I enable Filtering or not?
In the above, and other posts you’ve explained filtering and it seems like Filtering should likely be turned on.

However I was asked by our Vibration Specialists to turn filtering off. That seems to contradict what I have learned elsewhere.

if you turn off the filtering then you will get true data for the entire frequency range and there will no attenuation.
Filtering can be done on the receiver/software side as well. we do it on hardware side because we use hardware filters.

If turning Filtering off gives True data for the entire range, it would sound like you would always want that.

In what situations would you want to turn Filtering On?

(If you don’t want to go into detail and instead just send a link or the title of the subject that I can research further, that is fine too).

As far as why you would want to turn Filtering On, it sounds like you want to turn it on when you want to very precisely look at a specific range of Frequencies so

Sorry for the basic questions. I did try to review original emails from way back when we discussed this, but I think I lacked some knowledge at that time.

without the filters you would see the DC noises ( acceleration due to gravity/mounting orientation).

if filters are disabled then set the ODR to 25600. This will ensure you get all the frequencies.

With Filters Disabled, and ODR set to 25600, I would get that entire frequency range, including the DC Noises you mentioned, just confirming for my own sake, and thank you again for explaining.

I think thats more a question for our Reliability Engineers on if they want to see that, but I am glad we have that options. I don’t quite understand some of their reasoning for certain sample rates and their desire for filtering off.

So the reason to set to a specific range is to be able to see precisely that frequency range in greater detail right?

They also asked me about resolution. I did a number of FFTs and recorded my results.
3200ODR provides 3200 Samples (1 per each axis, so total of 9,600 samples).
However, anything 6400 or above, (12800, 25600) caps out at 3400 samples (1 per each axis, so total of 10,200 samples).

When looking at ODR 6400 and above, is the number of samples set at 3400, is there anyway to icnrease that, or is that the nature of the ‘Resolution’ changing depending on which ODR you choose?

i have met a handful vibration analysis and all of them have their own ways to look at data.
most folks are still used to mechanical vibration sensor where there was no way to set ODR, filters, samples duration, g rage. it used to dump all it can in one go. now we are using mems sensors and we can fine tune the sensor in a way, so that it exactly does what we ask it to do.

There is a an internal buffer and max it can store is 3400 samples per axis. while performing FFT you will need 2’s compliment samples. so it could be 256,512,1024,2048,4096 etc
More number of samples doesn’t mean more data.