AnsweredAssumed Answered

vrf More on the comparator question

Question asked by rsb on Nov 18, 2004
Digging into the nitty gritty:
> >
> > Or just throw the "build arbitrary waveform" object in front of the
> > comparator. It won't change bevhavior at all but it does make it
> > more obvious what is happening- and the behavior is documented.
> >
> > with the extra object you'll have resampled input to match
> > the X values you want- so the comparator will have explicit
> > point-to-point comparison (rather than implicit).
> >
> For me, this comes down to a trade-off.  On the one hand, it's more
> preferable to rely on the build-arb-waveform than on the comparator, because
> the BAW's algorithm is documented.  On the other hand, to use the BAW in the
> way you suggest means resampling ie modifying my measurement data; I might
> risk false passes doing that - such scenarios exist although they are
> unlikely.  No, I have to leave the measurement data unchanged, which means
> if I want a point-to-point comparison I must resample the spec to match the
> measurement data x values.  I can't use the BAW for that because the
> measurement data x values will not be linearly distributed, maybe not even
> monotonic. 

I think that you need to define more clearly what you actually need-
and a good place to start is the basics.

You have time-discrete data, and as soon as discrete data exists
you need to consider how to deal with it. Simple math no longer
functions ( add, subtract, >, < , etc ) UNLESS all the data
is discretized exactly the same way.

In your case you have measurements ad more-or-less random times
so the discretization isn't regular- which will indeed complicate

Presumably you have a spec of some sort, probably with its own
set of discrete control points.

The first question then that needs answering is what does your SPEC
say about time points that aren't specified? That it your spec
might specify what happens at times 1,2,3- but what does it say
about other times- say 2.3 or 1.8?

Once you know what your spec says then you'll know what sort of
resampling (interpolation) is needed. At that point I'm sure that
folks here will be able to help determine the best way to accomplish
that resampling. But without that data all we can do is
stumble about with various ideas that may or may not be of help.



Stan Bischof  Agilent Technologies  707-577-3994

You are currently subscribed to vrf as:
To subscribe send a blank email to "".
To unsubscribe send a blank email to "".
To send messages to this mailing list,  email "". 
If you need help with the mailing list send a message to "".