RadioBanter - View Single Post

February 7th 04, 07:01 AM

In article , (Tom
Bruhns) writes:

(Avery Fineman) wrote in message
...
In article ,
(gudmundur) writes:

My current I.F. bandwidth is 8mhz at the 6db points. I am looking at
pulses
of .8microseconds length, or about 1.25mhz. If all else remains the same,
and I change the swamping resistors, and tweak the slugs for a 1.5mhz I.F.
bandwidth at the 6db points, what increase in signal to noise ratio should
I see?

Signal to noise ratio changes as the _square_root_ of bandwidth
change. Wouldn't be much of an effect going from 1.25 to 1.5 MHz.

Um, he was starting with an 8MHz BW...

Ooops, mea culpa! My fault. Please see reply to Paul in other post.

With 0.8 uSec pulses and a 1.25 to 1.5 MHz bandwidth (I presume
Mega Hertz, not milli Hertz), the output envelope will be very
rounded, almost Guassian or "cosine-quared" in shape. Rounding
happens because of the limitation of passing the harmonics of the
pulsed RF; all you have left is the carrier frequency.

Yes, the pulses will certainly be rounded when they come out of the
filter (though they may have started that way anyway). But depending
on the filter type, they may also incur lots of ringing, and if the
pulses follow one after another at the right spacing, the phase of the
energy in the pulse relative to the phase of the energy left in the
filter will matter a whole lot in what you see coming out. The
trailing edge of a rectangular pulse fed through a Chebychev filter
isn't very Gaussian looking!

True. "Gaussian" shape is so often used incorrectly when so many
equate that to the statistical distribution curve shape also referred to
as "Gaussian." Chebs and Cauers (elliptical) all exhibit ringing in
an L-C component application...but it isn't quite the same kind of
shape resulting with equal group delays of SAW or digital filters.

In terms of RF Envelope shape, the envelope has little hangovers at
the trailing edges (I call them "burbles" in my mind, heh heh). Now,
some of that comes from energy stored-and-released-at-a-later-time
(commonly called "ringing") but I think (from analysis and simulation)
that it is the result of pulse sideband energy content summation that
includes the relative sideband phases.

When working with SAW filters at the 3rd generation of RCA's SECANT
(previously mentioned in reply to Paul), there was almost NO ringing
possible (or observed) and the "matched-filter" effect was absolutely as
predicted and observed in the earlier generations using L-C filters.

As for the original question, the answer depends on the spectral
distribution of the noise...if it happens to be strongly peaked at the
carrier frequency of the pulses, the narrowing won't make much
difference; if it happens to be peaked at some other frequency, it may
help a lot. If it's uniformly distributed, AND you keep the filter
shape the same and narrow the bandwidth by 1.5:8, then you will have
1.5/8 as much noise _power_. You'll also have somewhat less signal
power, depending on the shape of the pulses. And of course, the
filter won't get rid of noise that's introduced after the
filter--fairly obvious but sometimes overlooked.

In looking at the basic system, one has to assume that the random
noise IS uniform in energy distribution. If the real world has a different
set, such as peaking at certain parts of the spectrum, that can be
calculated later and overall S:N modified...while keeping all the other
factors the same. Any other method of looking at too many variables
at once results in long hours and a marked increase in aspirin intake.
:-)

I've got a copy of Claude Elwood's original "Shannon's Law" 1948 paper
in the BSTJ and still need to keep the Tylenol bottle handy when I
study that again. It makes sense, but getting to the "sense" part isn't
intuitive. Neither is time-domain response of filters. If it weren't for
the
computer simulation programs, I'd still have to rely on old aphorisms
of a very general nature. :-(

I once spent a fruitless night trying to figure out the spectral content
of an RF pulse that was on for only one RF cycle. The next day I
brown-bagged it and hooked up some test equipment during lunch
hour to get the results. Rather remarkable spectral content shape
and the RF waveshape was exactly one RF cycle as observed on a
wideband scope. Never did finish the analysis. Sometimes ya hafta
get down and dirty on the bench to prove a point and get results.

Len Anderson
retired (from regular hours) electronic engineer person