NNTP-Posting-Date: Mon, 28 Oct 2002 05:31:52 -0600
From: "Big John"
Subject: Re: vibration measurement circuit (vibration gurus please help)
Date: Mon, 28 Oct 2002 06:26:20 -0500
X-Newsreader: Microsoft Outlook Express 4.72.3110.5
X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly
I have some experience with vibration testing from a job I had a couple of
years ago. Your method sounds OK but there are some details you need to
know about. First, you have to worry about the mounted resonance frequency
of the accelerometer you are using. To get accurate readings you need to be
far below the resonance frequency at 5kHz. I'm not familiar with the analog
devices accelerometer but the manufacturer will provide this data. Also the
mass of the accelerometer must be low enough not to load the structure.
Smaller accelerometers are less likely to load the structure and have a
higher mounted resonance frequency, but they are also less sensitive so it
is important to know what accelerometer you need for your application.
There are two general ways of specifying filter types for vibration
- constant bandwidth filters
- constant percentage bandwidth filters
Constant bandwidth filters divide the 10 Hz to 5kHz bandwidth into a fixed
number of "lines" or individual filter bands. FFT works like this. You
might have 512 or 1024 lines. Each "line" being about 5 or 10 Hz wide.
Constant percentage bandwidth filters have a bandwidth that increases as
the frequency increases. Common types are 1/3 Octave or Full Octave
filters. They are usually specified by their center band frequency. IIRC,
for Full Octave the center band frequencies are 16, 31.5, 63, 125, 250, 500,
1000, 2000, 4000, 8000 Hz. For 1/3 Octave the center frequencies are 10,
12.5, 16, 20, 25, 31.5, 40, 50, 63, 80, 100, 125, 160, 200, 250, 315, 400,
500, 630, 800, 1000, 1250, 1600, 2000, 2500, 3150, 4000, 5000, 6300, 8000
Hz. Full Octave bands are usually used for airborne sound, 1/3 Octave are
often used for vibration measurements and pro-audio work.
So 7 bands is fairly small a number, even at 1/1 Octave band frequencies.
You can make your own custom filters but it won't be consistent with
industry standards. I think analog filtering is the way to go here because
FFT requires a fair amount of processing power and memory, but if you choose
to go this route you'll need to add a Hanning window and an anti-aliasing
filter to make it work. You can make your own active filters using Op-Amps,
but it's easier to buy them. IIRC there are chips available to do this.
Once the filters are available you can use your microcontroller to read the
voltage at the outputs. It's been a while, but I believe the standard way
is to measure the RMS voltage output at each band to get the vibration
Lastly, reading the amplitude on a linear scale can make it difficult to
see anything. The standard way is to look at a Log scale of AdB
(acceleration decibels) referenced to 1 micro G. The formula is AdB = 20
Log(G's measured / 1 micro G). Of course, you normally work with the
measured voltages and a calibration factor for actual measurements. You'll
probably need at least 60 dB of measurement range to get usable data, but
this depends on your application.
Hope this helps,
mook johnson wrote in message
>I'd like to get a graph of the vibration level (Gs) over a frequency range
>of 10 to 5kHz
>The vibration to be measured would be continuous or slowly changing over
>Seven bands over that interval would be sufficient resolution of the
>Is it valid to do this using an Analog Devices accelerometer to feed 7
>then measure the peak amplitude of each of the 7 BP outputs?
>I was thinking of using peak detectors to hold the peak experienced in each
>band over a 300mS period then use an A2D to sample the peak in each band.
>I only have an 8051 or MicroChip PIC running at 7.3MHz to do the
>A2D manipulation and serial communication to the host at 9600 bps. 3
>different axes will be sampled over the 1 second period and the host will
>request an update at 1 /second.
>I think FFT is out due to limited processing power and limited local RAM
>storing the raw waveform.
>Is this a valid approach? Will this be useful data for verifying
>qualification vibration levels for a device to be used in the same
>The desired display on the computer screen would be 3 graphs with G(peak)
>the Y axis and frequency on the X axis.