Prev: Error probability of Recursive Least Square (RLS) algorithm
Next: biquad for resonant lowpass
From: FatScouser on 28 Apr 2010 10:49
Hi,

We have a low frequency (<100Hz) motion control system which we need to
spectrum analyze. Does anyone know of an easy-to-use spectrum analyzer tool
that is applicable for low frequencies? It also needs to support low
frequency bandpass filters, e.g. around 5 Hz.

We're currently using SigView, but this seems more tailored towards audio
processing and high-frequency analysis.

Any pointers or advice would be most welcome!

Many thanks,

John

From: Rune Allnor on 28 Apr 2010 11:27
On 28 apr, 16:49, "FatScouser" <john.hague(a)n_o_s_p_a_m.truebit.co.uk>
wrote:
> Hi,
>
> We have a low frequency (<100Hz) motion control system which we need to
> spectrum analyze. Does anyone know of an easy-to-use spectrum analyzer tool
> that is applicable for low frequencies? It also needs to support low
> frequency bandpass filters, e.g. around 5 Hz.
>
> We're currently using SigView, but this seems more tailored towards audio
> processing and high-frequency analysis.

Any SW that can do spectrum analysis ought to be able to
do the job, with a couple of provisos:

- That the software can read data with arbitrary sampling
parameters, and not only the standard audio formats
- That the spectrum displays can be adjusted to show
the relevant bands
- That a number of audi-specific filters and scaling features
can be switched off.

You could also try some of the matlab clones, like octave or
scilab, that allow you to write your own functions to load
data and display the spectra.

If all these factors are taken care of, all you otherwise need
to do is to ensure that you have a sufficiently long recording.
The time-bandwidth product limits what frequency resolution B
you can achieve with a signal of duration T seconds.

The rule-of-thumb formula is

B*T = 1.

If you want a frequency resolution B < 100 Hz you need
a recording that is at least T > 1/B = 0.01 seconds long.
The longer duration the better.

Rune
From: Vladimir Vassilevsky on 28 Apr 2010 11:35


FatScouser wrote:
> Hi,
>
> We have a low frequency (<100Hz) motion control system which we need to
> spectrum analyze. Does anyone know of an easy-to-use spectrum analyzer tool
> that is applicable for low frequencies?

http://www.iowegian.com/scopedsp


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: Tim Wescott on 28 Apr 2010 13:22
FatScouser wrote:
> Hi,
>
> We have a low frequency (<100Hz) motion control system which we need to
> spectrum analyze. Does anyone know of an easy-to-use spectrum analyzer tool
> that is applicable for low frequencies? It also needs to support low
> frequency bandpass filters, e.g. around 5 Hz.
>
> We're currently using SigView, but this seems more tailored towards audio
> processing and high-frequency analysis.
>
> Any pointers or advice would be most welcome!

Like Rune, I'm going to mention Scilab/Octave/Matlab even though their
"easy to use" factor is only high when you compare them to things like a
Cray with Fortran and a line printer, or a ream of paper, a book of sine
tables, and a pencil.

Are you extracting transfer functions, or just looking for resonances
and the like? For the first, you need a lot more smarts; for the second
you just need a vector of data and an FFT.

When I am writing software to implement motion control systems my
preferred method of extracting data for system identification is to
instrument the controller -- i.e. to have software in the controller
that injects a test signal into a summing junction, then picks the
resulting signals off of various points in the system.

The reason I like doing this is because it assures me (if I don't screw
up and change the delay significantly) that I'm measuring exactly the
same system that my controller sees. It also makes it easy to do my
system identification with the control loop active. This is good both
for verifying the robustness of the loop, and for iterative controller
tuning when the plant is nonlinear enough that its apparent linear
equivalent changes with loop tuning.

My book* has a chapter on doing this in situ, if you're interested in
going in depth. It does need revision to include more methods than just
swept-sine, but all in all it's still a good resource.

* http://www.wescottdesign.com/actfes/actfes.html

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
From: HardySpicer on 28 Apr 2010 21:42
On Apr 29, 2:49 am, "FatScouser"
<john.hague(a)n_o_s_p_a_m.truebit.co.uk> wrote:
> Hi,
>
> We have a low frequency (<100Hz) motion control system which we need to
> spectrum analyze. Does anyone know of an easy-to-use spectrum analyzer tool
> that is applicable for low frequencies? It also needs to support low
> frequency bandpass filters, e.g. around 5 Hz.
>
> We're currently using SigView, but this seems more tailored towards audio
> processing and high-frequency analysis.
>
> Any pointers or advice would be most welcome!
>
> Many thanks,
>
> John

LabView would be fine.

Hardy
 |  Next  |  Last
Pages: 1 2
Prev: Error probability of Recursive Least Square (RLS) algorithm
Next: biquad for resonant lowpass