In article , "John Moriarity"
writes:

The "need" for low-noise RF sources was prompted by the
electronics industry going hot and heavy on cellular telephony
which uses partly phase demodulation and clock recovery
circuits in digital electronics. Because of those particular
markets, "low noise" has become a Big Buzzword.

Whether you have one pole or twelve or whatever, you will
NOT need a specific "low noise oscillator!" The very ordinary
sweep oscillators of ten, twenty, or thirty years ago are quite
fine.

Len,

I'm sorry, but I disagree. Concern with phase noise
in *all* kinds of communications systems really
got hot in the 1970s. Adjacent channel rejection
is limited by phase noise performance. It is presently
the limiting factor in HF receiver performance.

The limiting factor was, and continues to be, the
receiver front-end noise (usually in the first RF stage
or 1st Mixer, often described by NF or Noise Factor),
the ultimate selectivity (usually done by a quartz
crystal filter). Some things exacerbate the front-end
noise such as cascaded low-gain stages, the second
stage contributing some slight amount of extra noise.

Ham transmitter phase noise can easily be heard
during CW DX contests as a keyed increase in noise
floor.

Whether that is done in contests or not, the "phase noise"
heard is MUCH more likely to be generated by the Xmtr
RF section. That noise is more than likely to be just AM
noise from that transmitter or the AM-type noise from the
local receiver front-end.

I'm at a loss to see where an AM-type demodulator is going
to see a small amount of phase noise or even frequency
noise...a diode or "product detector" (multiplier of signal input
times local IF oscillator or "beat" oscillator) is going to pick up
the amplitude-modulated noise of the distant Tx or the local
Rx front-end. That demod isn't going to be able to "hear" much
in the way of phase or frequency noise.

Measurement of a crystal filter with steep sides
could be compromised by PM to AM conversion
on the slopes.

Perhaps...if the PM or FM was quite large. The actual PM or
FM from even moderately-stable L-C oscillators is still quite
small even with the large shape-factor values of very good
quartz crystal bandpass filters.

I did a bunch of measurements of quartz-substrate SAW
filters back in 1974 at RCA EASD which had Very steep
skirts, steeper than 9-pole quartz crystal lattice filters. The
only reason for concern was the _amplitude_ noise out of the
RF generator, solved by getting an HP 608F. [60 to 70 MHz
center frequency of the SAWs...608F has a nice amplitude
stabilizer] That was better than trying for some averaging of
multiple sweeps via a minicomputer data logger-averager
which would have done the same thing with a controlled
stepping RF synthesizer.

Anyone who cares can find out what the PM or FM "noise" is
by simply modifying an FM receiver to tune to an oscillator's
output and calibrating the FM receiver in terms of deviation done
by a signal generator and frequency counter combination. Since
PM and FM are - in theory and practice - nearly identical in terms
of sideband content versus deviation. That same setup can also
determine what the modified FM receiver's own LO PM-FM is as
well as the signal generator's stuff. In actual practice, that PM-
FM is so low that it doesn't matter one whit insofar as a basically-
AM demodulated HF signal is as far as "phase noise."

Oscillator phase noise IS important with various levels of QAM
where BOTH AM and PM are going on. The BER there is a very
good indicator of whether or not phase or amplitude "noise" is
a culprit and the Eye diagram can show which one is worst.
There's lots more already been done on various QAM systems
but the subject thread wasn't about digital modulations.

The subject thread was about voltage-controlled crystal oscillators
used for filter measurement. With good control to suppress
garbage on the (VCO or VCXO in more common terms) control
line and some ordinary good (and old) practices in making stable
oscillators, plus a wideband limiter stage to damp out the AM as
the oscillator sweeps, ain't going to be much of a problem except
calibrating the setup.

There are all sorts of frequency-variable RF sources that could be
made up for this kind of measurement, nicely described in
literature all over the industry, but "doing the numbers" on the
theoretical "phase noise" of a particular source is what counts.
The way I see it, after doing a bunch of that kind of thing (besides
in the 1974 time frame mentioned) is that the now-touted "phase
noise" figures just don't matter for this application.

Phase noise can be a factor in obtaining very precise stats on a
local frequency standard. "Allen Variance" and all that stuff (which
is described in detail on the NTIS website in one section) isn't going
to be needed for rather simple PPM tolerance measurements. This
is not PPB tolerances.

Lots of late-WW2 era FM-PM systems were used by the military,
notably the TRC-1, -3, -4 VHF radio relay sets using PM through
a reactance modulator from an ~500 KHz (?) crystal and then
doing something like 96 times (? forget the details, long chain of
doublers and triplers) frequency multiplication to reach 70 to 90
MHz output. After all that multiplying, the output modulation was
so close to FM that ordinary FM demodulators had no trouble
receiving it. Bandwidth went a bit beyond 12 KHz. Nobody talked
about any "phase noise" then even though the CF-1 and CF-2
carrier bays were also requiring phase and frequency stability of
the radio part.

Oscillator "phase noise" specs got in there in more modern times
as terms like Bit Error Rate (BER) and Eye Diagrams entered the
scene along with the explosion of various cellular telephony and
digital modulations. By that time, roughly the beginning of the
1970s, the ultimate in tuneable receiver sensitivity had already
been achieved for HF bands and those receivers were on the
market. Sensitivity hasn't really changed since that time for
most signals involving AM-type modulations.

For a setup to test filters with very precise frequency control as
well as stability of that frequency (in all three AM, FM, PM), a
DDS is probably the most versatile RF source. But, some DDSs
have unique problems of their own, involving the generation of
spurious sidebands. That's an entirely different matter than "low
phase noise" oscillators.


retired (from regular hours) electronic engineer person.