[time-nuts] How do time-nuts measure phase noise?

Bruce Griffiths bruce.griffiths at xtra.co.nz
Tue Aug 18 14:21:15 UTC 2009


Gerhard Hoffmann wrote:
> On Mon, 17 Aug 2009 23:18:58 +1200, you wrote:
>
>
>   
>> 1) Use 3 sources, S1, S2 and X.
>>
>> 2) Phase lock S1 in quadrature to X using a diode double balanced mixer
>> using a low bandwidth PLL.
>>
>> 3) Phase lock S2 in quadrature to X using a diode double balanced mixer
>> using a low bandwidth PLL.
>>
>> Then for frequencies outside the PLL bandwidth the crosspower spectrum
>> of the 2 mixer outputs should be predominantly due to the phase noise X.
>>
>> In practice its best if the phase noise of the 2 standards S1, S2 isn't
>> significantly worse than that of X.
>>
>> It is also desirable that both S1 and X and S2 and X be locked at their
>> respective mixers quadrature point (not necessarily the same as that for
>> zero dc output)
>>     
>
> Has anybody out there already written that cross correlation /
> user interface stuff? I can / will build a neat multi channel digitizer,
> perhaps with USB2, but trying to link it to Linux/Windows would probably 
> make it a never ending story. 
>
>   
There are a few commercial sound card based spectrum analysis software
applications that actually measure and display crosspower spectra.
Unfortunately most sound card based spectrum analysis software doesn't
include crosspower spectrum measurement capability.
Although you can always save the samples to a file and use something
like Matlab or SciLab to produce cross power spectra.
> BTW when building the DG8SAQ vector network analyzer 
> <http://groups.yahoo.com/group/VNWA>  I had to learn that the sound system 
> of my deLuxe Core Duo motherboard sports ridiculous 40 dB of channel separation. :-(
>   
As long as the crosstalk is well defined, measurable, and stable then in
principle it can be easily corrected with a little digital signal
processing.
> I would also like to try Inphi's sample&holds in front of a decent RF ADC 
> for a direct RF three cornered hat implementation (maybe a ANSI/VITA 57 
> mezzanine board for a contemporary Xilinx development board), 
> but with all the software it's too much for one person.
>
> I have done a new incarnation of my discrete NIST-flavoured distribution
> amplifier. The input return loss still needs to be cleaned up, then
> it's ready for publication.
>
>   
Did you consider using an output stage like that attached?
This should allow only a single cascade (npn EF+ pnpCB + pnp CB) to be
used instead of paralleling 2 such cascades.
> Gerhard, DK4XP
>
>   
Bruce
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OutputStage.gif
Type: image/gif
Size: 7842 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20090819/38a01863/attachment.gif>


More information about the Time-nuts_lists.febo.com mailing list