Fw: Re: Fw: Re: [Test-Equipment] Spectrum Analyzer BW measurement
Dave Emery
[email protected]
Sat, 31 Jan 2004 16:31:38 -0500
On Sat, Jan 31, 2004 at 08:53:45AM -0600, [email protected] wrote:
> That, I understand. Thanks Dave.
>
> Just for curiosity what do you read with your SA when you look at a
> signal at the -55dB points using the "few %" RBW vs 10x that RBW. I
> realize actual filter passbands are going to be different, today's
> spectrum analyzers probably compensate, etc. but I'm interested if you
> see something around 5x-10x BW at -55dB.
>
> 73 Kees K5BCQ
I will check a couple of mine in a few minutes and see...
This is related to the shape factor of the filters, which varies
between different spectrum analyzers. It is a function of
implementation technology and decisions by the designers balancing cost,
sweep speeds possible, and overall accuracy on various signal types.
But spectrum analyzer filters are classically Gaussian shaped
because it can be shown that filters with that shape ring the least when
swept past a signal. This is important, as the ringing represents
momentary errors in determining the level of the signal in the bandpass.
Spectrum analyzers with truly sharp filters have to be swept
slowly in order to avoid the distortions due to ringing - thus making
the display update rate lower than it would be otherwise.
But you have indeed recognized one of the pitfalls of using
spectrum analyzers - if you want to measure a weak signal accurately
near a strong one you do have to pay attention to the shape of the
bandpass many db down of the filters involved.
Another issue you have not commented on is the different
behavior of narrow filters versus wide with noise like signals versus
signals with narrow band energy (relative to the filter bandwidth). On
a purely noise like signal (many modern digital signals) a narrow filter
will capture proportionately less of the signal energy which is
distributed more or less uniformly across a wide bandwidth relative to
the filter than a wider filter will. This makes the amplitude of a
noise like signal much less as measured in a narrow bandwidth filter
compared to a wide band one that includes more or even most of the signal
energy inside its bandpass.
But if the signal contains mostly strong narrow band spectral
lines (a carrier, sidebands from modulating tones etc) which are
narrower than or close in width to the resolution bandwidth in use those
will show up with close to the same amplitude indication in a narrow
bandwidth as they would in a significantly wider resolution bandwidth.
What this means in practice is that wide band digital noise like
digital signals when seen with resolution bandwidths less than the
bandwidth of the signal are much shorter peaks than unmodulated carriers
or narrow band signals that actually contain less total energy. This can
be quite deceptive, if one assumes the amplitude of the peak is
proportional to total signal power. And especially so if the sweep
width and other parameters are set such that it is not obvious the
wide band signals are actually wider than the resolution bandwidth in use
(or the narrow band signals narrower).
>
>
> --------- Forwarded message ----------
> From: Dave Emery <[email protected]>
> To: [email protected]
> Date: Sat, 31 Jan 2004 01:21:40 -0500
> Subject: Re: Fw: Re: [Test-Equipment] Spectrum Analyzer BW measurement
> Message-ID: <[email protected]>
> References: <[email protected]>
>
> On Fri, Jan 30, 2004 at 07:15:44PM -0600, [email protected] wrote:
> > Thanks, Dave. That answers some of the questions as to why 3Khz.
> >
> > > Your RC remote control spec seems have used the later form -
> > >specifying how many db down from the carrier level the noise power
> > >reading in a 3 khz RBW centered 20 khz from the carrier is supposed to
> > >be. This obviously is trivial to measure correctly if you have a
> > >spectrum analyzer with a 3 khz bandwidth, but a little more complex if
> > >you have one (as you do with only 500 hz and 5 khz bandwidth).
> >
> > Right, and what is the "little more complex" proceedure ? Other than my
>
> > "rough estimate" indicated in a previous note.
>
> Depends a little bit on what you are expecting the energy
> to look like. Basicly you need to synthesize a 3 khz bandwidth
> from your 500 hz data by summing a list of 500 hz samples multiplied
> by a weighting factor (window). The crudest is a rectangular window
> of 6 samples. A fancier window is a gaussian shape matching a typical
> 3 khz bw filter with more than six samples and unequal weights.
>
> You have to convert dbs to noise power in each sample, multiply
> that by the weighting factor and sum the samples , then convert that
> back to dbs. An antilog and log operation.
>
> To get a curve, you slide the window over one sample and do
> it again for the resulting set of samples.
>
>
> > There are actually two specifications. One is the FCC spec for R/C
> > transmitters
> > which specs BWs at -25dB, -45dB, -55dB, and peak detect (based on the
> > +/-10Khz BW spec at -55dB a RBW/BW ratio of a few percent is implied).
> > Of the 5 RBWs I have, 500Hz RBW fits best and I have no problems with
> > those measurements.
>
> Yup, should work fine.
> >
> > The other is an AMA spec for -55dB for (apparently +/-) 20Khz using
> > 3Khz RBW. That's the one I have problems with ....correlation to my
> > measurements using 5Khz RBW or correlation to the FCC spec.
> >
>
> As I say, this can for most signals be synthesized from
> the sum of weighted narrower band measurements.
>
> It actually gets a bit trickier if you want real precision
> and the signal has signal components that correllate with each other,
> but to a first approximation my algorithm will work.
>
>
> > 73 Kees K5BCQ
> > _______________________________________________
> > Test-Equipment mailing list
> > [email protected]
> > http://mailman.qth.net/mailman/listinfo/test-equipment
>
> --
> Dave Emery N1PRE, [email protected] DIE Consulting, Weston, Mass
> 02493
--
Dave Emery N1PRE, [email protected] DIE Consulting, Weston, Mass 02493