Jim Lux wrote:
At 05:58 AM 1/19/2006, John Ackermann N8UR wrote:
Jim Lux wrote:
Of course... the SDR1000 does have an advantage because it's a very
simple receiver with fewer stages to screw things up.
While the SDR-1000 can do a credible job on its own, I just want to
reinforce that anyone with a signal generator that's locked to a
decent reference, and a sound card, can measure to small fractions of
a Hertz using the "delta reference" method I describe at
http://www.febo.com/time-freq/fmt/technique.html. You don't need a
lab full of expensive gear to do it.
Indeed.. The two accuracy limiting things in John's scheme that I see are:
1) Group delay variations across the band in the presence of any drift.
If the unknown and the reference move across the passband, and there's
some group delay ripple, the relative phase will change between the
tones. If you're integrating for 10 seconds, a phase change of 360
degrees would be a 0.1 Hz error. THis might be an area where older
receivers using a long string of tuned LC networks for selectivity would
be at disadvantage (because each section flips the phase 180 over the
passband, and they all add up). You also clearly don't want that exotic
10 Hz CW crystal filter in the circuit.
Yup. I normally use a fairly wide bandwidth; in fact, for the last
test, I used the widest 3100Hz bandwidth in the 3586C selective level
meter; good thing I did, too, as I mistuned on a broadcast hetrodyne on
40M, and only after the fact did I find W1AW 1kHz away, but still in the
bandpass.
2) Uncertainty in the sampling clock of the audio card. I suppose that
one can calculate this by measuring the frequency of the "known"
reference at two frequencies. A lot of signal generators have a way to
put an AM modulation on the signal with precisely controlled frequency,
and that would give you two marks, say, 400 Hz apart, with which to
calibrate the FFT box.
We've been around the soundcard clock issue several times on the list.
An advantage of the delta reference method is that you don't care about
the absolute frequency, only the delta between two closely spaced
frequencies. So the clock error is applied to a smaller number (the
delta between the two tones) and thus has less absolute impact than if
you were measuring absolute frequency. My measurements of the Delta 44
card I use shows that its clock error is in the noise for this test.
Remember that frequency measurement in the HF range doesn't really
require a super-accurate source -- 1x10e-8 gets you down to 0.1Hz at
10MHz. That level of accuracy can easily be met with a $100 surplus
ovenized oscillator (like an HP 10811A or the older 10544), a GPS
receiver, and a $100 surplus counter like the HP 5334 (my personal
low-cost favorite)*. And that's assuming you're fairly
fumble-fingered. With a bit of care, you can set to at least one, and
maybe two, orders of magnitude better than that (though the aging rate
will limit how long that accuracy holds).
The GPS receiver needs a 1pps output of some sort. (I tried this with a
serial output only once, and while the timing of the messages was
reasonably stable, it wasn't great: about 1ms uncertainty)
Yes, I should have said that. The good news is that there are lots of
old Oncore receivers available for a pittance that will do the job.
73,
John
