EMRFD Message Archive 1266

Message Date From Subject
1266 2007-12-27 07:02:17 Alan Melia re measuring phase noisee
Hi I dont think this was mentioned but I recollect seeing a system many
years ago maybe in EDN or Electronic Design that involved splitting the
signal in two and then mixing it with itself in a DBM (I cant remember
whether a phase shift (90deg.) was added to one of the paths) the output is
LP (passive) filtered and has twice the noise of the source. Presumably you
could use an audio analyzer on this and it does not depend on having a known
good source. This might at least be a very easy way of comparing systems.

Alan G3NYK
1267 2007-12-27 08:33:19 Jim Miller Re: re measuring phase noisee
Hi Alan

Although I'm not yet able to find the article you mention at either EDN or
ED archives I think you're describing the autocorrelation function.

If I'm correct we'd need to amplify the oscillator signal going to the
reference (LO) port as well as shift it by 90°. Since the phase shifter
would disturb the amplitude it would probably make sense to just square up
the reference signal before driving the LO port of the mixer?

I think some fine tuning would be needed to get to exactly 90°. Perhaps a
small varactor circuit and look for 0V DC at the IF output.

I'm not sure how the mathematics gets affected by the squaring. IIRC all of
the information is in the zero crossings of the waveforms so perhaps it is
just a scale factor to adjust for the difference between square and sinewave
multiplication due to the LO port.

Anyone have any thoughts? It does sound too easy so I'm probably wrong...

73

jim ab3cv
1275 2007-12-27 15:36:12 Gary Johnson Re: re measuring phase noisee
Alan,

I've usually seen the cross-correlation method used for microwave and optical
frequencies; we use it with lasers to determine coherence length, which is a cousin to
phase noise. The 90-degree shift is accomplished with a length of cable or a trombone
tuner. Maybe there are some microwave guys lurking out there who can talk about that.
Here is a cross-correlation instrument that measures phase noise from 5 MHz to 1.5 GHz.
Includes a nice block diagram. With some study, I would think an amateur homebrew
version of this could be built.
http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-CC.pdf

I did discover a couple more publications that might be of interest. Here is an old one
from NIST that covers a number of different phase noise measurement techniques for all
frequencies (and it includes correlation, too):
http://tf.nist.gov/timefreq/general/pdf/789.pdf

And if you really want some deeper reading on the subject, here is another NIST
publication, "Characterizati
1276 2007-12-27 15:49:20 Jim Miller Re: re measuring phase noisee
Gary

Assuming that a specific oscillator frequency is of focus, e.g. 10Mhz,
shouldn't a suitable set of amps followed by 25ft -30ft (depending on Vf)
provide a viable strong LO for autocorrelation?

I realize it isn't general purpose w.r.t. serving a range of frequencies.

73

jim ab3cv
1277 2007-12-27 16:00:33 Jim Miller Re: re measuring phase noisee
i wasn't clear...

the length of coax would be to act as a 1/4 wave phase shifter at the 10Mhz
frequency of interest.

jim ab3cv

----- Original Message -----
1278 2007-12-27 16:23:58 Lasse Re: re measuring phase noisee
Alan,
this sounds to me like an ordinary quadrature (discriminator) phase
noise test set. You split your source and phase shift one branch 90
degrees and feed both to a mixer
Low-pass filter the output and you will have the phase noise from that
oscillator. This link from Agilent (HP) will describe the basics of
phase noise measurement:
http://cp.literature.agilent.com/litweb/pdf/5952-8708.pdf

Usually one use a signal analyzer to do an FFT of the base band to get
the phase noise. Maybe a sound card and low-noise AF amps would be an
alternative There are several spectrum analyzer programs available as
freeware.

The major draw back is that it is really optimized for one frequency
only... but can offer a very low noise floor and is almost fool proof.

HP used to sell a test set that would phase lock a low-noise source in
quadrature to the DUT and then mix down to base band.. But the cost of a
HP 8662/3 synthesizer with suitable low noise option plus the HP 11848A
Phase Noise Test Set would be a bit steep :-)
Service manual http://cp.literature.agilent.com/litweb/pdf/11848-90004.pdf
One can pick up a few ideas by looking at the block diagrams!

Lasse SM5GLC

Alan Melia skrev:
> Hi I dont think this was mentioned but I recollect seeing a system many
> years ago maybe in EDN or Electronic Design that involved splitting the
> signal in two and then mixing it with itself in a DBM (I cant remember
> whether a phase shift (90deg.) was added to one of the paths) the output is
> LP (passive) filtered and has twice the noise of the source. Presumably you
> could use an audio analyzer on this and it does not depend on having a known
> good source. This might at least be a very easy way of comparing systems.
>
> Alan G3NYK
>
>