Why install the payload at the pad?

"Lee Jay" wrote in message
ps.com...
Herb Schaltegger wrote:
Here's a free clue: "vibration" does not have to be "periodic" - it
can and usually is exceedingly random, all your abstractions and
analyses to the contrary, the point you AGAIN refuse to acknowledge.

And you refuse to acknowledge that random noise has periodic components
- a fact we use daily to complete modal tests using pseudo-random
forcing functions from electrodynamic or hydraulic shakers as inputs to
our flexible structures.


IIRC, white noise has a flat power vs. frequency Fourier Transform, i.e. its
power for any frequency is the same. Stuctures definately have many
different frequencies as they vibrate at different modes (e.g bending vs.
torsion).

Danny Dot
www.mobbinggonemad.org



Lee Jay

"Herb Schaltegger" wrote in
message .com...
On Tue, 21 Nov 2006 14:11:41 -0600, Lee Jay wrote
(in article om):


And you refuse to acknowledge that random noise has periodic components

Wrong. What I HAVE said, time and again for you who seem so hard of
comprehension, is that vibration does not have to be periodic the way
you think, and that your analytical tools and models are only
approximations of reality, as anyone with any sense should be able to
understand.


Let me say I think you are BOTH right. In theory any vibration can be
modelled with a periodic Fourier Transform -- i.e. one that models zero Hz
to infinate Hz. This works OK for math majors that can work with infinate
Hz. I am an engineer and can't model infinate Hz in my computers, so my
ability to make a perfect model of the periodic nature of vibration is
limited. Perhaps one of you is talking mathimatics and one is talking
engineering.

Just a thought on my part.

Danny Dot
www.mobbinggonemad.org



Your continued denials of reality are more than tiresome. Enjoy your
stay in the cooler with the rest of the net.kooks.


--
Herb Schaltegger
"You can run on for a long time . . . sooner or later, God'll cut you
down." - Johnny Cash
http://www.angryherb.net

On 20 Nov 2006 11:38:41 -0800, "Lee Jay" wrote:

Herb Schaltegger wrote:
Than as an "experimentalist" you should realize that what goes into the
system is NOT the same thing that your FFT approximates coming out.

No but an FT (fast or not) of the input is a very good approximation of
the input. What's coming out has to go through the transfer function
(linear, nonlinear or otherwise) that is the system first. An FT of
the output is a good approximation of the output and, in many ways,
more useful than the actual output time series.

Arrgh! I'm three months late responding to this, but I can't let it
go by.

This is absolutely wrong, wrong, wrongitty, wrong if your system has
any nonlinearity at all. The FT, being linear, will discard the
nonlinearity in the output and you'll never even know it. It's not a
good approximation and it's certainly not more useful than the actual
time series output.

Give up the outdated frequency-domain techniques and try using a
time-domain technique. Try something like PEST; it works for
non-linear systems quite nicely.

Mary "x-dot = Ax +Bu; let's hear it for the state space equations"
--
Mary Shafer Retired aerospace research engineer
We didn't just do weird stuff at Dryden, we wrote reports about it.
or
Visit my new blog at http://thedigitalknitter.blogspot.com/

No but an FT (fast or not) of the input is a very good approximation of
the input. What's coming out has to go through the transfer function
(linear, nonlinear or otherwise) that is the system first. An FT of
the output is a good approximation of the output and, in many ways,
more useful than the actual output time series.
This is absolutely wrong, wrong, wrongitty, wrong if your system has
any nonlinearity at all. The FT, being linear, will discard the
nonlinearity in the output and you'll never even know it. It's not a
good approximation and it's certainly not more useful than the actual
time series output.

Say what? He explicitly said that the FT applied to the output of the system,
when any potential non-linearity had already done its work!

For any periodic function - and any real (as in real-world) function can be
made periodic by virtue of the fact that you only have a finite time for
measuring, and the warp-around artifacts can suitably be taken care of - the
Fourier transform is a homomorphism on this space (which happens to be one of
the square-integrable Hilbert spaces of functions). As such, it is
information-conserving.

Jan

On Sat, 24 Feb 2007 23:45:48 -0600, Reunite Gondwanaland (Mary Shafer) wrote
(in article ):

On 20 Nov 2006 11:38:41 -0800, "Lee Jay" wrote:

Herb Schaltegger wrote:
Than as an "experimentalist" you should realize that what goes into the
system is NOT the same thing that your FFT approximates coming out.

No but an FT (fast or not) of the input is a very good approximation of
the input. What's coming out has to go through the transfer function
(linear, nonlinear or otherwise) that is the system first. An FT of
the output is a good approximation of the output and, in many ways,
more useful than the actual output time series.

Arrgh! I'm three months late responding to this, but I can't let it
go by.

This is absolutely wrong, wrong, wrongitty, wrong if your system has
any nonlinearity at all. The FT, being linear, will discard the
nonlinearity in the output and you'll never even know it. It's not a
good approximation and it's certainly not more useful than the actual
time series output.

Thank you for your support. I could've used it a few months back. :-)

--
You can run on for a long time,
Sooner or later, God'll cut you down.
~Johnny Cash