[Lowfer] Ultra accurate soundcard calibration

[Garry Hess]

When I started monitoring LF a few years ago lowfer TAG was transmitting 
Wolf10. To receive and decode that very accurate tuning is necessary. 
Fortunately, at that time Loran-C was available, as was John's excellent 
article "Using the WOLF GUI to Measure Sound Card Sampling Rates" at 
http://www.w1tag.com/WOLFSamp.htm. There are two sources of frequency 
error, receiver tuning error and soundcard sampling rate error. The 
former can be avoided with AM demodulation and fortunately the Icom 
706MkIIG radio I was using at the time had very little tuning error at LF.

More recently, K1JT, Joe Taylor added frequency calibration capability 
to his WSPR software ("Accurate Frequency Measurements
with your WSPR Setup" at 
http://www.physics.princeton.edu/pulsar/K1JT/FMT_User.pdf).

And now Jay has described a calibration procedure using Spectrum 
Laboratory program. The learning curve for that program is significant 
but well worth the effort. It's hard to imagine anything that program 
can't handle!

The focus of the above methods seems to be avoiding direct GPS reception 
and instead relying on sources whose frequency has atomic accuracy. Yet, 
thanks to eBay, purchasing a GPS disciplined 10 MHz source is quite 
reasonable. Cheaper yet, you can build a GPS-locked source easily with a 
"Jupiter" receiver 
(http://www.gpskit.nl/documents/rockwell/jupiter-gps-board.pdf) and GPS 
antenna/amplifier. This particular receiver provides a locked 100 kHz 
output which can be applied to a phase-locked loop to control the 
frequency of a main oscillator. Pyrojoe of eBay fame has a simple PLL 
circuit for accomplishing this with a simple exclusive-OR for the phase 
comparator. I prefer using the more complicated 74HC7046 IC since it 
allows for unlock/lock indication.

The point of the above is that having a GPS-locked 10 MHz source in the 
ham shack is inexpensive and very desirable. Applying that source to a 
divide-by-100 counter yields 100 kHz and a second divide-by-100 yields 1 
kHz. I use 10 MHz in the AM mode to establish the true clock frequency 
of my SDR-IQ, then 1 kHz directly to the soundcard to calibrate its 
sampling rate. But I apply 100 kHz to the SDR-IQ tuned to 99.6 kHz USB 
to establish my final LF sampling rate to 3 decimal places of precision 
with the Wolf frequency measurement mode. Calibration only takes a few 
minutes so it's no big deal to repeat it often if you're concerned about 
drift. Having the GPS-referenced signals available allowed an average 
accuracy of 60 ppb in the April 12, 2011 Frequency Measuring Test, 
improved to 17 ppb in the April 19, 2012 test. Of course a lot of 
processing of Spectrum Laboratory files with Excel was necessary to 
reach that accuracy.

-- 
73, Garry, K3SIW, EN52ta, Elgin, IL