[time-nuts] Phase noise principle and measuring confusion

Wed Apr 8 09:23:25 UTC 2020

Hi Perrier

I try to explain phase noise in a simple way.
Assume you have a perfect oscillator which outputs a perfectly pure sine
wave. How would that oscillator's signal look like on a perfect spectrum
analyzer? it would be zero everywhere, except at the carrier frequency,
there would be an infinitely sharp "needle" peak.
However, a real oscillator does not behave like this. If you could look at
a real oscillator with a perfect spectrum analyzer, you would see what is
called "noise skirts", i.e. the signal has a peak at the carrier and decays
more or less steeply on both sides of the oscillator frequency, i.e. the
"needle" peak is broadened.
Since we usually don't have access to perfect spectrum analyzers, the
needle peak is always also broadened a bit by the spectrum analyzer's
response itself, so most of the time, we are not able to directly see the
noise skirts of a real oscillator with a real spectrum analyzer because the
analyzer is far too noisy.
So if we can't see the signal why do we want to measure it and why care
about it? answer: if the noisy oscillator is used as local oscillator for a
mixer in a receiver, the noise on both sides of the carrier also goes into
the mixer, and mixes down other frequencies we actually don't want to
receive. So the phase noise degrades the receiver performance. This is only
one example where the phase noise is important, there are many more.

A more exact description of what phase noise is can be given as follows.
The output voltage of any oscillator is:
Vo(t) = A*cos(2*pi*f*t + phi(t))
where
A = amplitude
f = frequency (obviously)
t = time
phi(t) = noise term
So for a noiseless oscillator, phi(t) is a constant. However, for a real
oscillator, phi(t) is fluctuating a bit. There are two contributions (maybe
more, but I know only of the following two):

- short term fluctuations: if phi(t) bumps around at very short intervals,
like 1sec and lower, we call this "short term" variations and it is counted
as phase noise as the short term variations make the oscillator signal look
wider on a spectrum analyzer.
- long term fluctuations: this is when phi(t) varies over long periods of
time, e.g. 10sec or even hours or days, and is called "stability" and
"drift". This is the parameter I am currently trying to measure.

If you look at above equation, the phase term is
phase(t) = 2*pi*f*t + phi(t)
so if you would plot phase(t), a perfect oscillator would yield a perfect
straight line which rises for increasing time. A real oscillator does, in
general, the same, but the slope of the line varies a bit, sometimes it is
a bit steeper and sometimes a bit less steep. Again, the time it takes for
these fluctuations to occur depends whether we call it phase noise or
stability.

I am not sure whether I understood your last question correctly, but I
assume you want to use the 10811 as reference for a 3336 synthesizer and
then use the synthesizer as local oscillator for a mixer?
if that is what you want to do, then it works fine. There are, however, as
always, several things to be considered:
- the 3336A synthesizer will increase the phase noise of the 10811A (the
amount of increase is called residual phase noise). Therefore, the phase
noise of the local oscillator generated in this way is for sure worse than
the original 10811.
- if you want to analyze the stability of an oscillator, the above is
perhaps (hopefully others can correct me on this if I am wrong!!!!) not so
much of a problem. If you analyze stability, you need to have a reference
which is more stable than your DUT, so if your DUT is worse than an 10811A,
then this setup can be used. You could not test an 10811A in this way!
- if you want to analyze phase noise, then the residual phase noise of the
3336A may be a problem as it degrades the reference oscillator's (your
10811A) phase noise. The residual phase noise of the synthesizer is the
absolute lower limit, so no matter how good your reference to the 3336A
will be, the phase noise at its output will be at least as high as the
3336A's residual phase noise. If your DUT is, say, at least 10dB or more,
worse than this residual phase noise, you are good to go and you can use
this setup to measure phase noise.
- if your DUT has a phase noise which is lower or in the same order of
magnitude as your 3336A's phase noise, then the 3336A cannot be used, no
matter how good your reference oscillator for it is, and you need to search
for a lower phase noise alternative.

There are the HP 8662A and 8663A signal generators which were exactly
designed for these types of measurement. They have very low residual phase
noise and can be electronically tuned, so you can build a PLL with it.

Does that somewhat answer your questions?

Best
Tobias
HB9FSX

On Wed, Apr 8, 2020 at 1:43 AM Perry Sandeen via time-nuts <
time-nuts at lists.febo.com> wrote:

> Learned Gentleman,
> I've read several articles on phase noise but I'm lost.
> I need a *Ding-Dong* school explanation of what it is, why it's important
> and how one goes about measuring it.
> <snip> Bob wrote:
> Phase measurement of my GPSDO
>
> Thequick way to do this is with a single mixer. Take something like anold
> 10811 and use the coarse tune to set it high in frequency by 5 to10 Hz.
>
>
>
>
> Thenfeed it into an RPD-1 mixer and pull out the 5 to 10 Hz audio
> tone.That tone is the *difference* between the 10811 and your device
> undertest.
>
>
>
>
> Ifthe DUT moves 1 Hz, the audio tone changes by 1 Hz.
>
>
>
>
> Ifyou measured the 10 MHz on the DUT, that 1 Hz would be a very smallshift
> ( 0.1 ppm ). At 10 Hz it’s a 10% change. You have ‘amplified’ the change in
> frequency by the ratio of 10 MHz to 10 Hz ( so amillion X increase ).
>
>
>
> *IF*you could tack that on to the ADEV plot of your 5335 ( no, it’s
> notthat simple) your 7x10^-10 at 1 second would become more 7x10^-16 at1
> second.
>
>
> OK, this seems to me that this is measuring frequency difference.
>
> wrote:
>
>
>
>
> Thereason its not quite that simple is that the input circuit on
> thecounter really does not handle a 10 Hz audio tone as well as ithandles a
> 10 MHz RF signal. Instead of getting 9 digits a second, youprobably will
> get three *good* digits a second and another 6 digitsof noise.
>
> OK, then would using a 3336A synthesizer work by using the *good* 10811 or
> other 10 MHz as an external reference and provide, say, a 10 MHz + 100 Hz
> or 10 MHz + 1KHz work with the 5335? (I have two). Or am I missing
> something (or a lot)?
> Regards,
> Perrier
>
>
>
>
>
>
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to
> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
>