[time-nuts] Sound Card Spectrum Analyzer

Bruce Griffiths bruce.griffiths at xtra.co.nz
Fri Feb 19 21:20:11 UTC 2010

Gerhard Hoffmann wrote:
> Bruce Griffiths wrote:
>> Not necessarily, it depends on the phase detector circuit details.
>> What phase detector output termination network did you use?
>> Aren't BF862's somewhat noisier than 2SK369's at low frequencies?
> SRA-3H because I had them,  49R9 in series with 3n9 for sum termination.
> LC low pass with fairly large capacitance, high load impedance.
> as recommended in Crawford, Frequency Synthesize Design Handbook page 11.
> he cites
> Fisher, M.C. "Frequency Domain Measurement Systems",
> 10th Annual Precise Time an Time Interval Applications and Planning 
> Meeting
> as the source.
> I could not see much of a difference between NXP BF862 and Toshiba 
> 2SK369.
> BUT  ---  the BF862 has MUCH smaller input capacitance, so one could 
> at least
> parallel 3 times as many of them for the same input load.
> Now that I have some access to an Agilent signal source analyzer on a 
> good-will-base
> my own phase noise measurement capability has dropped in importance, 
> so I can
> concentrate in oscillators. I could re-use them as a clock for a 
> digital PN solution with
> 2* Inphi S&H, 2* state of the art ADCs and FPGA, which is more on my 
> home turf.
> I think that the Wavecrests use the Inphi S&H, too.
> Gerhard
That termination should result in a quasi trapezoidal beat frequency 
waveform for low beat frequencies.
The noise would be somewhat lower (as NIST found) if the 49R9 were 
replaced by a short.
Most of the improvement would be due to reflecting the sum frequency 
back into the mixer.

The SRA-3H is likely to have a somewhat lower output impedance than an 
RPD-1 or similar.
In which case your preamp with the paralled BJTs may be close to optimum.
If an RPD-1 were substituted a preamp using a single SSM2220 should have 
somewhat lower noise than if your preamp were used.

If the output impedance (as a function of input levels) for several 
mixers were to be measured for various IF port terminations this would 
be useful.
Alternatively one could always ask Minicircuits about the characteristic 
for their mixers and phase detectors.

The Wavecrest models that correspond to the extremely detailed patents 
that I have seen use relatively conventional TAC interpolators.


More information about the time-nuts mailing list