EMRFD Message Archive 1260

Message Date From Subject
1260 2007-12-25 19:20:31 mdgolfbum Measuring Phase Noise?
Not much in EMRFD on phase noise measurements that I can find. Anyone
have some ideas for procedures or links to phase measurement techniques.

I've got one of Roger's power meters and a HP8640b.

jim ab3cv
1261 2007-12-25 19:37:36 Jim Miller Re: Measuring Phase Noise?

just remembered the cd and found demphano. reading it now...


jim ab3cv

----- Original Message -----
1262 2007-12-25 21:04:23 brainerd@wildblue... Re: Measuring Phase Noise?
On 25 Dec 2007 at 22:37, Jim Miller wrote:

> sri
> just remembered the cd and found demphano. reading it now...
> 73
> jim ab3cv
Now all we need is a PC program to take the output of the demphano into the sound card
and graph the phase noise vs freq offset.

Dave - WB6DHW
1263 2007-12-26 05:47:05 Jim Miller Re: Measuring Phase Noise?
the demphano article presumes access to an $8000 spectrum analyzer (waaay
beyond my means) but suggests that it would be possible to measure phase
noise with an analog spectrum analyzer. perhaps a ZOI model would work.

i'm finally starting to build my ZOI today after collecting parts for the
last year.

i'd still be interested in finding more phase noise measurement articles.


jim ab3cv
1264 2007-12-26 17:26:12 Gary Johnson Re: Measuring Phase Noise?
Here are a few methods for phase noise measurement.

1. Jim, since you have an HP 8640, there are a couple methods based on that. One is to
simply mix its output with the DUT, and use an audio spectrum analyzer to look at the IF
output, which will be the sum of ALL phase and amplitude noise contributions. This exact
method appears in the 2006 Handbooks, p. 10.11, under "Transmitter Phase Noise
Measurement in the ARRL Lab." This is equivalent to using a direct-conversion receiver,
and relying on the quality of its VFO's phase noise to determine the measurement floor.
One could also apply a narrow audio bandpass filter to the IF output and measure the AC
RMS value, instead of using an SA. But sometimes you get fooled by spurious signals,
hum, etc.

2. Another method appears on the very next page of that Handbook, and it uses a PLL
method very similar to the EMRFD Demphano article, with the HP 8640 serving as the VFO
in the PLL loop. The 8640 makes the system much easier to assemble. This is a VERY good
method. A simple and clear version of this also appears at: http://www.wenzel.com/documents/measuringphasenoise.htm

3. The "filter" method appears in EMRFD on p. 4.13, (and Fig. 4.19. In this case you DO
need an RF spectrum analyzer, and it does require pretty good resolution. The ZOI
analyzer could do the job if you obtain a very narrow bandpass filter for it, much tighter
than the 30 kHz one in the published articles. Otherwise, you can't measure close-in
phase noise. I'm one of those guys who spent $$$ on an ebay HP 8568B, and I use this
method with very good results.

3A. There is another version of the "filter" method I just read about. Paul Kiciak, N2PK,
measured phase noise of some crystal oscillators that are the reference for a DDS
generator that is part of his vector network analyzer project. He used his station receiver
(and an audio spectrum analyzer) to measure the residual noise. He is, in effect, using his
receiver as a spectrum analyzer, with "manual" frequency sweep. By using a VERY narrow
xtal filter (10 MHz with 180 Hz BW), he could measure phase noise very close in. So, you
don't necessarily need and RF SA... A detailed post
1265 2007-12-26 20:05:42 Jim Miller Re: Measuring Phase Noise?

thanks for the great tips!


jim ab3cv
1284 2007-12-29 12:28:48 Wes Hayward Re: Measuring Phase Noise?
Hi Jim and group,

The measurement of phase noise is always an interesting, sometime
difficult, and extremely important chore.

Virtually all of the communications equipment we enjoy today is
limited in performance with the synthesizers as the critical
Although not "projects" to be built, there are some things that are
useful in Chapter 7 with regard to phase noise measurement. The
discussion starting on page 7.40 deals with the measurement of
synthesizer spurious responses. The measurement of reciprocal
mixing resulting from phase noise in a receiver LO is identical and
this is discussed. I would urge any of you who contemplate a phase
noise measurement system start with some simple experiments of this
sort. (Owing to the design of the synthesiser in our
usual "appliances," these measurements are often all to easy to

An HF measurement receiver is presented on page 7.33. This uses a
250 Hz wide crystal filter with a Gaussian to 6 dB shape. This is a
wonderful measurement filter for phase noise or SSB transmitter
IMD. I would of course dump the NE604 these days in favor of an

Figure 7.64 shows a very simple block converter that will let you use
a good signal generator (8640, crystal oscillator, etc) to convert
signals down to an audio baseband that can then be examined with PC
software. You can also use the audio output of some high end
receivers for this purpose. KK7B has reported in QST some
measurements of this sort with his Racal. Fig 7-63 shows the
baseband output of Bob Larkin's DSP-10 transceiver which shows
oscillator noise spectra.

If you take a look at virtually all of the schemes that are used to
measure oscillator noise, you will conclude that they all seek to do
one thing: they try to eliminate the strong carrier from an
oscillator so that a measurement system can be used to look at the
noise while not being overloaded by the carrier. If you look at the
numbers for oscillator noise, you quickly conclude that receivers or
spectrum analyzers generally lack the needed dynamic range. Gary
Johnson described some phase lock systems in his discussion (EMRFD
Yahoo #1264 and a later one.) Chuck (K7QO) mentioned a collection
of older articles from RF Design that would do some of these
things. Alan, G3NYK, talked about a noise discriminator in his
posting, #1266. In this regard, EMRFD Fig 4.37 shows how a diode
ring mixer is used with a length of coaxial cable to build a

The discriminator is a system that gets rid of the oscillator
output. Essentially, it uses the signal being measured as a LO to
heterodyne a sample of the signal down to base band. The delay
line (phase shifter; call it what you like) serves to let the LO
signal to the mixer be at a different point in time than the RF that
is down converted.

Years ago K7TAU and I built a phase noise measurements system using a
discriminator. This work happened at TriQuint Semiconductor. The
signal from oscillators (F=0.5 to 2 GHz) that we studied was applied
to a power amplifier with about 50 mW output. The signal was then
split (3 dB hybrid) with the louder portion going to the LO port of a
diode ring mixer. The other portion went into a long delay line.
This consisted of about 100 to 200 feet of low loss coax cable
followed by a "trombone" line stretcher that would let us adjust the
line length to obtain a total length that was an odd multiple of 1/4
wavelengths. This then resulted in a DC mixer output of zero, which
served to eliminate the carrier. The RF signal was attenuated
before it was applied to the RF port of the the ring mixer. The IF
output port included a blocking capacitor that prevented DC from
reaching the output. We then used a low noise amplifier to get the
signal level up. The amplifier had a gain of 35 dB, NF of around 2
dB, a bandwidth of 80 MHz, and very good S11 and S22. This
amplifier is one we built from a paper in the trade literature and
used a Motorola RF input transistor followed by a Com-Linear Op-amp
that had a bandwidth pushing 1 GHz. This amplifier was a wonderful
general purpose gain block.

By the way, I originally wanted to include the schematic for that
amplifier in EMRFD, but National Semiconductor who had purchased Comm
Linear dumped my request to their legal department. The response I
got back was extreme enough that all that made it to EMRFD was
a reference. I didn't want to republish the paper, but just the
schematic. (Arr, &%^#**@!) For this application, a 2N3904
feedback amplifier with a gain of 20 dB and a NF of 6 dB will do most
of the good work.

We calibrated this system by generating a low level FM signal at a
few kHz. Our oscillators all had on chip tuning diodes. The FM was
observed with both a Tek 492 analyzer as a sideband that was around
60 dB below the carrier, and with the baseband analyzer, a Tektronix
7L5, attached to the discriminator output. Once we knew what the
baseband output meant with regard to the single sideband signal, we
could interpret the baseband data that we observed.

We were building some GaAs integrated circuits that included an on-
chip oscillator. Even the resonators were on chip. Nothing ever came
of the experiments, but they were great fun. I'm glad I got a
chance to build one of these measurement systems. Unfortunately, I
don't have any more data on this work, for I don't have the notebooks.

Somewhere in that same time frame I purchased a 1000 ft role of RG-11
coax, brand new. It was labeled as "shorted" and I gambled. Sure
enough, the short was at one of the ends, and was easily fixed. The
goal was to build a discriminator that would work at VHF and perhaps
even HF. I never got around to doing it, but the coax is still in
the garage. Perhaps someday. The role of coax weighs 80 pounds.
I got the stuff at the Tek company store for $30. It was used by
the TV division.

Crystal filters can also be used to enhance the dynamic range of a
normal spectrum analyzer. K7HFD did this in connection with his
oscillator. This circuit is in EMRFD and was originally described
in SSD, page 126. Gumm used a 3 kHz wide 10 MHz crystal BANDPASS
filter for this measurement. I'll not repeat the details here; you
can all look it up if interested. Essentially though the oscillator
output was run through the filter, but the carrier was tuned about 10
kHz away from the filter frequency. The peak in noise in the filter
passband was a measure of the single sideband noise density of the
oscillator spaced 10 kHz from the signal.

I recently built a couple of crystal notch filters. They are used
by dropping an oscillator carrier right into the hole in the filter
response, which was from 85 to 95 dB deep in the filters I built.
(So far.) The filters had about 1 or 2 dB of insertion loss. A
low noise amplifier (the same one described above) followed the notch
filter and then went into a spectrum analyzer. While it was not
possible to get close to the carrier with my homebrew analyzer, which
has only a 30 kHz filter, it was suitable to measure the noise peak
from a particularly bad DDS synthesizer that I had recently put on
line. The DDS noise peaked at -105 dBc/Hz at a spacing of 42 kHz
from the carrier. I was also able to see noise from my old and
rather abused HP8640B that was -140 dBc/Hz.

I got identical spectral density results when I observed the noise
with an old HP spectrum analyzer that had a 1 kHz resolution
bandwidth. The notch filter is a wonderful tool, but it is
restricted to measurements with the oscillator operating at the notch
frequency. It also produces noise outputs that are asymmetrical,
which are probably a combinati
1285 2007-12-29 17:36:23 Russell Shaw Re: Measuring Phase Noise?
Wes Hayward wrote:
> Hi Jim and group,
> The measurement of phase noise is always an interesting, sometime
> difficult, and extremely important chore.
> Virtually all of the communications equipment we enjoy today is
> limited in performance with the synthesizers as the critical
> element.

You could mix the signal to be tested down to baseband using a low
noise synthesizer such as a Singer FM-10. It covers DC to 500MHz
using nothing but the mixing of crystal oscillators and has no PLLs.
1286 2007-12-29 17:40:29 Russell Shaw Re: Measuring Phase Noise?
Russell Shaw wrote:
> Wes Hayward wrote:
>> Hi Jim and group,
>> The measurement of phase noise is always an interesting, sometime
>> difficult, and extremely important chore.
>> Virtually all of the communications equipment we enjoy today is
>> limited in performance with the synthesizers as the critical
>> element.
> ...
> You could mix the signal to be tested down to baseband using a low
> noise synthesizer such as a Singer FM-10. It covers DC to 500MHz
> using nothing but the mixing of crystal oscillators and has no PLLs.

Correction: multiplications, additions, and subtractions of the frequency
of a single crystal oscillator.
1287 2007-12-29 18:11:19 Jim Miller Re: Measuring Phase Noise?
I knew my single 1/4 wave of coax idea was too good to be true. ;-)

So I gather that many wavelengths of coax are required on the way to an odd
quarter wave being useful. The result being that the two sampling points are
highly uncorrelated?

I'm in the middle of building my ZOI-SA so it will be a while before I get
around to doing phase noise measurements but this will be interesting for
futher study.

jim ab3cv
1288 2007-12-31 13:32:04 bkopski Re: Measuring Phase Noise?
Hi Jim and all,

I've been observing with interest the phase noise discussion and some
reference to "trombones" as used with some measurement schemes. This
brought to mind an incident well over a decade ago wherein I needed a
line stretcher "NOW". The only quick option I had was to try to
fabricate one – and it did the job within a few hours effort.
Thinking that perhaps such an HB test implement might be of interest
to the group, I reconstructed that long-ago effort and have posted
some photos showing how.

The HB stretcher is approximately 14" long with an adjustment range
of about 11". It is constructed of 4 pieces of telescopic "K & S"
brass tube from my local hobby shop. I selected tubing diameters to
approximate 50 Ohm air line.

The handbook gives Zo for such a line as:

Zo = 138 * log (b/a)

Wherein "b" is the inside diameter of the outer conductor and "a" is
the outside diameter of the inside conductor. Zo = 50 Ohms when (b/a)
= 2.3. All that's needed is to select tubing pairs that meet this
criteria as closely as possible.

The "K & S" brass tubes are stepped in OD from 1/16" up to about ¾"
in increments of 1/32". Wall thickness is specified as 0.014".
Taking all in to account, I found the best selections to be paired
sets of 7/16" and 3/16" for the "outer" conductor and 13/32" and
5/32" for the inner conductor. This pairing yielded Zo of 46.8 and
53 Ohms respectively – close enough.

The only other items needed are two BNC's, some buss wire (eg #22,
24, 26 A/R), some grommets, and a simple fixture to aid in assembly.

The wire is used to "build up" the BNC center pins and the BNC outer
diameters to "fill" the inner diameters of the various tubes at the
solder connections – this assures fairly close centering of all the
pieces. Just wind a few turns of the best suited gage to do this for
your connectors.

I used a simple wood fixture to align the inner conductor tubes to
the BNC center pins during soldering. I used a close fitting grommet
stuffed in the open end of the larger tubes to center up the inner
conductors during soldering of the former to the BNC bodies. The
outer tubes are full lengths of 12" and the inner tubes are trimmed
A/R to just protrude past the outers. This trim will vary with the
BNC's chosen; mine needed about 5/16" trim each. Incidentally, be
sure to do some brass polishing (eg. with Scotch Brite pad or
similar) and do remove flux buildup with solvent as you proceed thru

I tested my finished stretcher two ways. First, using a 10 mHz sq.
wave from a 50 Ohm source to drive the line, and terminating the line
output in 50 Ohms at my `scope, I observed just about 1 nSec delay
motion as I extended the line over its 11" range. This was at my max
sweep speed of 5 nSec / Div so this is a challenging observation for
sure! But the observed motion is reasonably consistent with the
calculated expectation so all seems well.

Next, I set up an RLB into the terminated line and observed the
return loss at 10, 20 and 30 mHz from "short line" to "long line".
The average of the six data points was 32.3 dB with the worst single
point being 28.8 dB.

K & S brass tube (and many other metal materials and shapes) are
available but not always in convenient quantities. You can obtain
single piece quantities at good prices from:


Two other thoughts: BE SURE the BNC's you choose have readily
solderable bodies before you begin – so you don't have to start over
as I did! Next, while I've not tried this, I believe you can get
greater adjustment range by arranging two such lines side by side
1301 2008-01-04 09:04:02 KD5SSJ Re: Measuring Phase Noise?
I found an interesting calculator applet for calculating jitter from phase noise.


Cash Olsen KD5SSJ

Kits to build Scotty's Spectrum Analyzer http://www.zianet.com/erg
Scotty's Spectrum Analyzer website http://www.cpu-net.com/host/wsprowls
Sam Wetterlin's website http://www.wetterlin.org/sam/
Yahoo Builders Group http://groups.yahoo.com/group/spectrumanalyzer/

[Non-text portions of this message have been removed]