I'm new to DSP and working on a simple hand-held balance instrument for analyzing mass imbalance on rotating machinery (think out-of-balance fans), utilizing accelerometers and zero-degree phase opto sensors.
There seems to be two common approaches:
1. Apply hardware filtering to (noisy) raw accel waveform, which generates a near perfect sine wave, and perform analysis on this sine wave, to resolve the required velocity and pahse angle report.
2. Apply (noisy) raw accel signal direct to the AtoD inputs on a dsPIC and (apparently) perform digital filtering and FFT to resolve the required velocity and phase angle report.
Can anyone comment on the relative complexities of either appraoch, and refer me to info on the specific precedures, or perhaps on-line tools?
Rotating imbalance analyzer
The digital approach will be cheaper and more satisfactory. The FFT is a nice approach, because it not only is it easy to construct a filter, but you can take the imbalance measurement directly from the fundamental frequency. If you want to know the angle of the imbalance, you'll need two accelerometers, but the complex FFT will give you that rather directly.
If you know what you're doing, you're not learning anything.
I've used several commercial propeller balance analyzers that require only one accelerometer and an opto sensor to generate the required velocity and phase angle report.
The frequency is known and fixed in these applications (airplane propeller balancers) so I don't see the value of using an FFT, since it will indicate the frequency that is already known.
In addition, frequencies on either side of the fundamental freq are not of interest because mass imbalance refers to the fundamental frequency.
Many people confuse "balance analyzers" with "vibration analyzers". The latter will work for balancing applications but is likely a much more complex instrument, since it considers wide band noise from bearings, loosness, etc. The former only considers the fundamental frequency. Everything else is filtered out.
Can anyone comment on or direct me to a source of info on the specific algorithms employed for generating the required velocity and phase angle from the raw acceleration and opto sensor data?
The frequency is known and fixed in these applications (airplane propeller balancers) so I don't see the value of using an FFT, since it will indicate the frequency that is already known.
In addition, frequencies on either side of the fundamental freq are not of interest because mass imbalance refers to the fundamental frequency.
Many people confuse "balance analyzers" with "vibration analyzers". The latter will work for balancing applications but is likely a much more complex instrument, since it considers wide band noise from bearings, loosness, etc. The former only considers the fundamental frequency. Everything else is filtered out.
Can anyone comment on or direct me to a source of info on the specific algorithms employed for generating the required velocity and phase angle from the raw acceleration and opto sensor data?
Of course, the FFT will help if the frequency is known. It gives you the magnitude of a known frequency, while filtering out everything else, if you chose. The complex FFT (which is in the mE library) will also give you the phase angle.
The FFT is used very commonly in power line analyzers of known frequency. It's not primarily used to determine frequency, in fact it's kind of clumsy for that.
The FFT is used very commonly in power line analyzers of known frequency. It's not primarily used to determine frequency, in fact it's kind of clumsy for that.
If you know what you're doing, you're not learning anything.
The phase angle between two inputs. If you capture the data streams for two inputs simultaneously, and then transform, you'll get complex frequency pairs for the two inputs. Even if one is a digital input, you'll get imaginary frequencies, and you can calculate the magnitude from the real/imaginary pair, and the phase angle from the relationship between the two imaginary components.
A DSP text will explain it all. It's rather involved.
A DSP text will explain it all. It's rather involved.
If you know what you're doing, you're not learning anything.
-
- Posts: 46
- Joined: 05 Aug 2007 07:15
- Location: Shizuoka JAPAN
Hi
Even if you take digital approach , you need analog low-pass filter to avoid aliasing.
So if speed of rotor is fixed, I think analog filter is most simple way.
If rotor speed is not fixed, analog filter( Anti-aliasing ) and digital filter( variable freq) is best way, ( Be careful to filter-delay)
If there are some rotor rotating different speed in same machine,
and you intend to analyze which rotor has imbalance, I recommend FFT.
Below are good site to learn FFT. and DSP.
. .http://www.dspguide.com/ch8.htm
Even if you take digital approach , you need analog low-pass filter to avoid aliasing.
So if speed of rotor is fixed, I think analog filter is most simple way.
If rotor speed is not fixed, analog filter( Anti-aliasing ) and digital filter( variable freq) is best way, ( Be careful to filter-delay)
If there are some rotor rotating different speed in same machine,
and you intend to analyze which rotor has imbalance, I recommend FFT.
Below are good site to learn FFT. and DSP.
. .http://www.dspguide.com/ch8.htm