Yes. I regret now that I rounded off the WA6ZTY frequency to the nearest
0.1 Hz.  I might have been exactly in agreement if only I had rounded to
the nearest 10 millihertz.

I used a Ten-Ted RX340 receiver, which has the advantage that it accepts
an external 10 MHz frequency reference at its rear panel and phase-locks
all its oscillators to the 10 MHz reference. That made the measurement
comparatively easy. I would set the receiver for CW reception with its
BFO offset 800 Hz from the center of the passband, and feed the output
to SpectrumLab, running on my laptop with an external USB sound card. I
calibrated the sound card against WWV audio before the test.

I captured the Spectrum lab measurements to a text file, which I
imported into Microsoft Excel to compute the mean and standard deviation
of the peak frequency within a passband centered on 800 Hz. This worked
well for the W1AW measurements: once having tuned the receiver, I left
its dial alone for the duration of the test. As I recall, the one-sigma
jitter on the received W1AW frequency was 1 or 2 Hz -- so I certainly
did NOT consider my measurement more accurate than about 0.1 Hz.

Unfortunately, when it came to the WA6ZTY part of the test, I
kepttweaking the receiver frequency by a Hz or so (it tunes in one-hertz
steps) and that made the spreadsheet file captured from SpectrumLab less
useful.  I didn't fell confident that I had made a good estimate of the
standard deviation of the measured frequency. because of my mistake of
tweaking the receiver frequency. So I rounded that measurement, too, to
the nearest 100 millihertz.

If I had only NOT disturbed the receiver tuning during the WA6ZTY
frequency measurement, I could have made a better estimate of the
one-sigma frequency jitter. Then I would have had a better estimate of
the accuracy of the measurement, and might have rounded it to the
nearest 10 millihertz rather than to the nearest 100 millihertz.

--
James Maynard
Salem, Oregon, USA

Connie Marshall wrote:

I also sent the Spectrum Lab data to a spreadsheet.  Once I got the carrier
at the frequency I wanted I did not touch the receiver.  I made two mistakes
this time: 1. I did not record a wav file, 2. my FFT settings were such that
it made tuning difficult.  The settings were fine once I found the carrier
but getting there was another matter. Because of this I was unable to tune
in the 40-meter signal.  I found it just as the test ended but it was too
late to get any data.

I used an HP 3586C Frequency Selective Voltmeter with the tracking generator
looped back to the RF input via a step attenuator.  The audio output went to
the PC.  An ICOM IC-745 was tuned to the 160M signal to hear the test
announcements throughout the test period.  The antenna on the HP was a SWL
slopper antenna that favors the lower frequencies.  The ICOM was on a
40-meter dipole.

My numbers:

160M:  1 854 317.77  [ARRL: 1 854 317.5 Hz]
80M:  3 587 117.82  [ARRL: 3 587 117.5 Hz]

I did not attempt to receive the west coast signal.

Mike

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of James Maynard
Sent: Friday, December 29, 2006 2:14 PM
To: k5cm@azalea.net
Cc: tim-nuts mailing list
Subject: Re: [time-nuts] FMT

Yes. I regret now that I rounded off the WA6ZTY frequency to the nearest
0.1 Hz.  I might have been exactly in agreement if only I had rounded to
the nearest 10 millihertz.

I used a Ten-Ted RX340 receiver, which has the advantage that it accepts
an external 10 MHz frequency reference at its rear panel and phase-locks
all its oscillators to the 10 MHz reference. That made the measurement
comparatively easy. I would set the receiver for CW reception with its
BFO offset 800 Hz from the center of the passband, and feed the output
to SpectrumLab, running on my laptop with an external USB sound card. I
calibrated the sound card against WWV audio before the test.

I captured the Spectrum lab measurements to a text file, which I
imported into Microsoft Excel to compute the mean and standard deviation
of the peak frequency within a passband centered on 800 Hz. This worked
well for the W1AW measurements: once having tuned the receiver, I left
its dial alone for the duration of the test. As I recall, the one-sigma
jitter on the received W1AW frequency was 1 or 2 Hz -- so I certainly
did NOT consider my measurement more accurate than about 0.1 Hz.

Unfortunately, when it came to the WA6ZTY part of the test, I
kepttweaking the receiver frequency by a Hz or so (it tunes in one-hertz
steps) and that made the spreadsheet file captured from SpectrumLab less
useful.  I didn't fell confident that I had made a good estimate of the
standard deviation of the measured frequency. because of my mistake of
tweaking the receiver frequency. So I rounded that measurement, too, to
the nearest 100 millihertz.

If I had only NOT disturbed the receiver tuning during the WA6ZTY
frequency measurement, I could have made a better estimate of the
one-sigma frequency jitter. Then I would have had a better estimate of
the accuracy of the measurement, and might have rounded it to the
nearest 10 millihertz rather than to the nearest 100 millihertz.

--
James Maynard
Salem, Oregon, USA

Connie Marshall wrote:

I

was

to

got

TX

old

was

time-nuts mailing list
time-nuts@febo.com
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts

Mike Suhar said the following on 12/29/2006 02:32 PM:

I too use the 3586C but there is one thing to be aware of: while the
main frequency synthesis system can be locked to an external reference,
the BFO signal comes from an independent oscillator that is not
disciplined.

The last IF is at 15.625 kHz and the BFO frequency (generated on board
A22) is (for the "C" version) is either 17.475 kHz (USB) or 13.775 kHz
(LSB).  The BFO is derived from a crystal -- 1.7475 MHz for USB and
1.3775 MHz for LSB; the crystal frequency is divided by 100.
Interestingly, it appears that the BFO signal is injected into the
product detector as a square wave; unless I missed it, there's no filter
following the divider.

After a bit of warm-up, the BFO is very stable (dividing by 100
certainly helps reduce any drift) but the nominal 1850 Hz tone generated
by a signal on the tuned frequency is likely to be off by a fraction of
a Hertz.

One of my plans is to build a BFO synthesizer to replace the crystals
and thereby eliminate any frequency offset.  Having done that I should
be able to directly determine frequency by measuring the output tone.

John

I didn't enter the FMT due to the lack of an HF antenna, but my original
plan was to downconvert the signal with a GPS-locked LO, then run it through
a logamp and a very narrow CW filter and straight into my 5370B without any
demodulation at all.  As long as there were no interferers within the CW
filter's bandwidth -- which it sounds like there may have been -- that
should have worked about as well as any other technique.

Does the 3586C have an AGC-controlled IF output?  A BFO synthesizer would be
a good idea, but it might not be necessary if you can filter and measure the
IF directly.

-- john, KE5FX

John Miles said the following on 12/29/2006 03:24 PM:

It has an IF output as well as a built in counter with 0.1 Hz
resolution.  However, there is no AGC as such.  You can put it into a
100dB dynamic range mode, which reduces sensitivity, or in an
autoranging mode where the gain is stepped in 10dB increments.  Or you
can turn the ranging off and manually set the gain.  The autoranging
results in transients that are annoying in a waterfall, so I usually set
the range manually and live with a bit of distortion if it overloads
from time to time.

John