I have a HP 141 series device which is reasonable from the point of view of
strong signal handling.
It is surprisingly easy to produce intermodulation effects at levels lower
than you would expect.
A "forest" of signals at -20 dBm will produce IM effects by the additive
effect of all their amplitudes.(you could easily get this from a 40M dipole
at night for example)
In fact, my friend Rod Green has done an article for QEX on a "figure of
merit" device to test receiver strong signal handling which consists of a
"comb generator" with harmonics every 20 KHz, and a bandpass filter covering
the band of interest - say 7.0 to 7.2 MHz. At a comb level of -20dBm most
receivers will be overwhelmed.

Richard

If I were designing a spectrum analyzer for the electronic instrument
market, I would shoot for at least meeting Hewlett-Packard Agilent
or Rhode&Schwarz specifications...R&D budget willing. That's a bit
steep for the hobbyist area.

The problem is that real incoming signals and the analyser's spurious
responses all look very much the same on the screen. When you can't
trust what the analyser says, it becomes very hard to understand what's
really going on.

Sigh. A spectrum analyzer, almost ALL of them, is one of the easier
instruments to characterize from the outside, using other instruments.
Frequency span, logarithmic linearity, passband of the final IF are all
relatively easy to determine from the outside.


Len Anderson
retired (from regular hours) electronic engineer person

I have a HP 141 series device which is reasonable from the point of view of
strong signal handling.
It is surprisingly easy to produce intermodulation effects at levels lower
than you would expect.
A "forest" of signals at -20 dBm will produce IM effects by the additive
effect of all their amplitudes.(you could easily get this from a 40M dipole
at night for example)
In fact, my friend Rod Green has done an article for QEX on a "figure of
merit" device to test receiver strong signal handling which consists of a
"comb generator" with harmonics every 20 KHz, and a bandpass filter covering
the band of interest - say 7.0 to 7.2 MHz. At a comb level of -20dBm most
receivers will be overwhelmed.

Richard

If I were designing a spectrum analyzer for the electronic instrument
market, I would shoot for at least meeting Hewlett-Packard Agilent
or Rhode&Schwarz specifications...R&D budget willing. That's a bit
steep for the hobbyist area.

The problem is that real incoming signals and the analyser's spurious
responses all look very much the same on the screen. When you can't
trust what the analyser says, it becomes very hard to understand what's
really going on.

Sigh. A spectrum analyzer, almost ALL of them, is one of the easier
instruments to characterize from the outside, using other instruments.
Frequency span, logarithmic linearity, passband of the final IF are all
relatively easy to determine from the outside.


Len Anderson
retired (from regular hours) electronic engineer person

In article , "Ian White, G3SEK"
writes:

Hans Summers wrote:

The W7ZOI homebrew spectrum analyzer article is not to be missed:

http://www.qsl.net/n9zia/wireless/pdf/9808035.pdf (part 1)
http://www.qsl.net/n9zia/wireless/pdf/9809037.pdf (part 2)

You'll find this design vastly superior to the "poor man's spectrum
analyzer" projects out there that are based on CATV tuners.

The W7ZOI project is very interesting. I'm interested in what ways you think
it's superior to the poor man's spectrum analysers based on CATV tuners? The
latter surely have greater frequency coverage. But in what ways are they
inferior

A spectrum analyser is simply a frequency-swept receiver with a
dB-scaled output to a screen. To give meaningful results, that receiver
must have a very high dynamic range with very low spurious responses.

"Meaningful results" are subjective to the hobbyist.

NO analyzer means NO results.

If I were designing a spectrum analyzer for the electronic instrument
market, I would shoot for at least meeting Hewlett-Packard Agilent
or Rhode&Schwarz specifications...R&D budget willing. That's a bit
steep for the hobbyist area.

The problem is that real incoming signals and the analyser's spurious
responses all look very much the same on the screen. When you can't
trust what the analyser says, it becomes very hard to understand what's
really going on.

Sigh. A spectrum analyzer, almost ALL of them, is one of the easier
instruments to characterize from the outside, using other instruments.
Frequency span, logarithmic linearity, passband of the final IF are all
relatively easy to determine from the outside.

So what if a spectrum analyzer isn't "perfect?" It is much, much
better than having NO spectrum analyzer.

The majority of spectrum analyzer input signal levels are most likely
to be UNDER -10 dbm. That input level is not - generally - going to
cause all kinds of "imperfections" in the viewed spectrum.

As with lots of beginner-level test equipment, it sometimes needs an
expert to understand it!

I disagree. There are any number of application notes free for the
downloading on the Internet, from Agilent they are copies of older
(two decades at least) FREE paper application notes. Agilent also
has free application notes on the basic building blocks within an
analyzer and much information on the characteristics of those blocks.

CATV tuners and low-level NE602 mixers are simply not the building
blocks for a high dynamic range receiver. The W7ZOI design uses much
more appropriate building blocks so its readouts are much more
trustworthy.

Yes, its possible to "conquer" the dreaded too-high-signal IM mountain
peaks with high-level mixers and higher-milliWatt first LOs. See Mini-
Circuits' catalog as one place for modules.

I haven't seen the "W7ZOI design" so I won't critique it at all. Having
used spectrum analyzers for about 4 decades, the high-input-level IM
bogeyman seldom goes "boo!" for most spectrum observation. That
includes transmitter output monitoring.

and can you quantify it?

Very easily, in the same ways as you test a receiver for strong-signal
handling.... but most graphically by looking at the same spectra with
two analysers side-by-side. The one showing fewest signals is the one
you can rely on most.

Have you priced the used spectrum analyzers lately? Do you expect
others to have ready access to "another" spectrum analyzer?

In checking my own little special-purpose IF strip Sweeper (a quasi-
spectrum-analyzer), I am fortunate to have a pair of H-P rotary step
attenuators (salvaged, checked for calibration by another) to determine
if the Log response curve of the Analog Devices chip is correct...rather
very basic stuff. Unfortunately, the accuracy of the external attenuator
is about the same as the AD accuracy. I'd love to have a Weinschel
Precision Bench Attenuator to use as a comparison, but don't, can't
afford one.

I do have a very good pair of H-P signal generators (very much previously
owned) which have been put in order by a good friend of mine. With the
help of an outboard lowpass filter (easy to make) which is also
characterised by response testing using a linear detector, I can guarantee
an RF signal with all harmonics down 60 dbc. The RF power output of
those generators is also separately characteriseable/calibrateable within
2 db of absolute level at the high-power output (0 dbm) all the way down
to -120 dbm, probably lower. The lower levels are determined by a
waveguide-below-cutoff internal attenuator which hardly ever jumps out of
calibration unless the mechanicals get goofy. The Weinschel attenuation
standard uses the same basic below-cutoff principle.

With all of the above (two generators, overlapping ranges) I can absolutely
guarantee a true "two-tone" testing setup for an analyzer to determine
what the 1 db or the 3 db IM values are of any receiver front end. So far,
I've never ever seen any cause for concern with any input level up to -10
dbm. Have I ever done any such? Of course, even checking out a used
(and rare) Tektronix spectrum analyzer plug-in on a borrow.

I have YET to encounter any receiver input that goes higher than -10 dbm
equivalent input with the one exception of being within two blocks of
local AM broadcast station KMPC pushing 50 KW into their antennas.
I'm sure there are all kinds of "exceptions" to that and I'm sure there will
be commentary coming back on that. :-)

I'm not going to get in about the relationship of sweep rate and final SA
IF resolution...which are VERY important in SA work, particularly in
relatively narrowband observations. Close-proximity signals can "hide"
if wide resolution IFs are in place...that "hiding" visible at input RF
levels
well below ANY intermodulation distortion level. The resolution aspect of
practical SA design is a separate matter but should not be forgotten..

When push comes to shove in all this, I'd say that SOME KIND of
spectrum analyzer is a LOT better than NONE. EOF.

Len Anderson
retired (from regular hours) electronic engineer person