"pawan jadia" <p...@yahoo.com> wrote in message
news:6...@posting.google.com...
>
> ---> fred, yes, the system we are designing is for detecting the
> velocity of the slowly moving object(reflector), that is the echo will
> definitely contain the changed frequency(doppler shift). This is the
> thing we want to measure. So i think you got my point. ya definitely
> there will be range and velocity ambiguity (similar to heisenberg
> principle) but even measurement with some errors(very minimum) is
> acceptable to us. The precision(its precision, not resolution) we need
> is 90Hz. Since there will be frequency change so we can't use
> correlation process, or still we can use??? or suggest some other
> appropriate method

Pawan,

You didn't say if you got rid of the chirp waveform.  In view of the
complexities it involves I would suggest a single-frequency tone.  To
discuss further requires this be made clear in your responses.

I always wanted to be able to remember the difference so I Googled on it and
came up with:

a. Precision. This is a measure of repeatability, i.e. the degree of
agreement between individual measurements of a set of measurements, all of
the same quantity.
b. Accuracy. This is a measure of reliability, and is the difference between
the True Value of a measured quantity and the Most Probable Value which has
been derived from a series of measures. The True Value is, of course, never
known.
c. Resolution. This is the smallest interval measurable by an instrument.

From this it appears that in order to obtain precision to some level it
appears you need resolution to support it.  Otherwise, the measurement can
jump from resolution cell to resolution cell and the precision will
deteriorate accordingly.

You did not say that you want to measure range very accurately at all.
However, the short pulse suggests that you really do - if so, with what
resolution?  The tone pulse will have a time-bandwidth product around 1.0
unless you window the pulse shape.  The time provides Doppler resolution and
the bandwidth provides range resolution.  So, you must trade between the
two.

I have already noted that 90Hz resolution (perhaps inadequately small for
90Hz precision) requires the pulse length be around 11msec.  We need not
continue to repeat this do we?

I feel that something is missing in this discussion.  One thing is range
resolution necessary.  The other is getting rid of the chirp.  The other is
an explanation why the pulse is so short to begin with. .......etc.

Fred

Fred Marshall wrote:

   ...

> I feel that something is missing in this discussion.  One thing is range
> resolution necessary.  The other is getting rid of the chirp.  The other is
> an explanation why the pulse is so short to begin with. .......etc.
> 
> Fred

Fred,

My guess -- I love it when I'm wrong about things like this -- is that
the specifications and proposed implementation arise from false
understanding. A chirp, a short pulse, and band-limited white noise all
have the same spectrum: uniform over a frequency interval. A chirp can
be generated by passing a short pulse through a black box with a
frequency-dependent delay*. The short pulse can be reconstituted by
passing the chirp through a complementary delay. When so reconstituted,
conservation of energy increases the effective peak power. This allows
long signals at moderate power to behave like short ones with the same
total energy. which provides good range information with long signals.
(No contradiction here: the time at any one frequency is short.) Chirps
in which the frequency varies linearly with time work well with
stationary targets, but doppler shifts muddy them. Chirps in which the
period varies linearly with time are not decorrelated by doppler shifts.

Pawan's chirp might well yield good range information. I'm not familiar
with the way doppler information is used, or if it can be.

Jerry
______________________________________
* 'Spheric whistlers are an example.
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

"Jerry Avins" <j...@ieee.org> wrote in message
news:40d1b4c8$0$3010$6...@news.rcn.com...
> Fred Marshall wrote:
>
>    ...
>
> > I feel that something is missing in this discussion.  One thing is range
> > resolution necessary.  The other is getting rid of the chirp.  The other
is
> > an explanation why the pulse is so short to begin with. .......etc.
> >
> > Fred
>
> Fred,
>
> My guess -- I love it when I'm wrong about things like this -- is that
> the specifications and proposed implementation arise from false
> understanding. A chirp, a short pulse, and band-limited white noise all
> have the same spectrum: uniform over a frequency interval.

***I understand but that's a relatively gross statement.  I don't think they
have the same spectral character.... it depends on the analysis interval
doesn't it?  As you know, they each have a different ambiguity function:
- a chirp has a linear, diagonal ridge
- a pulse has a sinc in frequency and
- pseudo noise has a "thumbtack" with a noise floor that can be too high...
Said another way, the noise will have a noisy spectrum and the other two are
also definable and different.

Very long waveform yields good spectral resolution - just like in DSP..
Very short waveform yields good range / temporal resolution - just like high
sample rate in DSP
One has to trade so that the combination of spectral and range resolution
are balanced for the application.
The idea of using time-bandwidth products > 1.0 only works in some
circumstances.  Otherwise TW=1.0 is more like the only practical value
that's useful.

..................

> Pawan's chirp might well yield good range information. I'm not familiar
> with the way doppler information is used, or if it can be.

***Tracking..... prediction .... Kalman filter inputs..... where is the
object likely to be now as compared to where it was last observed?  Helps
when observations are noisy, subject to dropouts .....

***It is useful to note that radar applications often rely on repeated range
measurements to get radial range rate.  Sonar applications can't get
repeated range measurements so rapidly (speed of sound vs. speed of light)
and can usefully apply Doppler measurements to get radial velocity / range
rate.  Sonar applications are helped in this regard because the Doppler
shift is relatively high - whereas in radar, the Doppler shift for most
manmade objects is small thus harder to measure (the speed of sound vs.
speed of light once more).

***It is also sometimes useful to note that target responses to an active
radar or sonar are referred to differently as are the measures for their
reflectivity:
RADAR talks about "returns" and "radar cross section" RCS
SONAR talks about "echoes" and "target size" T or TS
RCS and TS are the same thing in principle subject only to which reference
one uses for 0dB.

Fred