[time-nuts] FMT

[Didier Juges]

Hi Mike,

Interesting, I was just doing pretty much that, except that I did not 
think of using the tracking generator of the 3586A as a reference, I 
used an 8657B synthesizer phase locked to the Thunderbolt GPSDO to 
inject a reference signal 20Hz above the test signal (I used WWV at 5 
MHz for test), so that I could display both signals at the same time in 
the FFT using Spectrum Lab. Setting the two very close eliminates the 
risk of selecting the wrong sideband and reduces the error caused by the 
sound card sample rate being off.

I found out using the slowest sampling rate (5 kHz or so) and the 
largest FFT size seems to give the best resolution (10 mHz?) and good 
filtering (several seconds of time constant). I also set the software to 
display one bin per pixel.

There is a screen plot of Spectrum Lab with the 2 signals on my web site 
http://www.ko4bb.com/Test_Equipment/SpectrumLab.png

Exporting the data to a file and using a spreadsheet to do the post 
processing is a good idea, thanks.

To display the generator frequency using the peak function, I simply 
disconnect the antenna, which causes the generator to be the largest 
signal. I intend to do that a few times during the test to make sure 
everything is stable.

I have not found out how to export data to a file so that it includes 
the peak, but I'll be looking (I started using Spectrum Lab a couple of 
weeks ago and there are a lot of settings I am not familiar with yet)... 
If I don't find soon, I may send you a message :-) I am interested also 
in finding how you can output two peaks (signal under test and 
generator) to the log file.

It's interesting to look at the peak on WWV and see it wobble at the 
rate of the fading (or close). During a few minutes of observations, the 
frequency was about 0.5 Hz low and it fluctuated by about 0.4 Hz peak to 
peak with a period of a few seconds.

I only have an 80m antenna at the moment (quarter wave slopper hanging 
from the 50 feet tower), so that will be the band where I will have my 
best shot. I may try 40m as east coast signals tend to be good on the 
80m slopper, in spite of the severe mismatch.

Didier KO4BB


Mike Suhar wrote:
> John and I use a similar technique.  I am running Windows so I use DL4YHF's
> Spectrum Lab FFT software.  Spectrum Lab will output each line of the
> waterfall display to a tab delimited text file.  Using the function
> peak_f(f1,f2) I output the frequency of the peak within F1 and F2 to the
> file.  I also output the peak of the signal generator's carrier in a band
> above or below the band of the desired carrier. The difference between the
> two is used to calculate the final carrier frequency.  You just have to
> remember to be in the same sideband USB or LSB and on the same side of the
> signal generator carrier for each measurement in order not to confuse adding
> or subtracting the delta.  At the end of the test the output file contains
> several lines of data. I import the file into a spreadsheet.  The
> spreadsheet calculates the delta and averages all the readings.  I may have
> several hundred lines of output.  
>
> As for the audio sample rate and FFT size I have used several the past
> couple of years.  I am sure they are optimal but they work.  I am currently
> running the audio sample rate at 11025 with the FFT size of 32768 with a
> decimate factor of 2.  Bin size is around 168 millihertz.  This helps cut
> down on any CW that may be near the desired carrier as I won't see those
> signals as they are too fast to fill the bin. 
>
> For this year's FMT I will use my HP 3586C with its internal tracking
> generator looped back into its own antenna port (via a step attenuator).
> This will serve as the signal generator used in the past.  A  Z3801 feeds
> the external timebase of the 3586C.  
>
> It has been interesting to measure the carriers of local AM broadcast and
> shortwave stations.  Our local stations are a few hertz off but WLW (700
> KHz) is dead on. I wonder what they are using for frequency control?
>
> Mike
>