[time-nuts] Collecting raw timing data

Bruce Griffiths bruce.griffiths at xtra.co.nz
Wed Nov 28 10:29:45 UTC 2007


Jeff Mock wrote:
> Bruce Griffiths wrote:
>   
>> Jeff Mock wrote:
>>     
>>> Thanks so much, your description of collecting the raw timing data is 
>>> really clear.  I like the technique of doing timing interval 
>>> measurements from a slower known reference on one channel to a DUT on a 
>>> second channel.  This definitely gets me started.
>>>
>>> jeff
>>>
>>>   
>>>       
>> An algorithm for unabiguously unwrapping the rollovers can be found in:
>> http://horology.jpl.nasa.gov/papers/picket_uffc.pdf
>>
>>     
>
> Thanks, that's a nice paper, 3-pages, lots of short declarative 
> sentences that make sense.  As I understand the problem, Fig. 1 in this 
> paper is the key for collecting good data for measuring Allan variance.
>
> jeff
>   
It all depends on the resolution you need.

When comparing a DUT with PPS pulses produced by a GPS receiver this
method can have more than adequate resolutioin.
If you need/want better resolution (and lower system noise level) than
around 2E-11/tau then a more complex method is usually necessary.

A resolution of around 1E-14/tau or better is achievable using the dual
mixer time difference method when comparing 10MHz sources.
However to achieve this performance attention to a lot of fine detail is
required:

The mixer and distribution amplifier temperatures have to be stabilised.
(Mixer and distribution amplifier phase tempcos can be as high as
several picosec/C)
Distribution amplifier reverse isolation and channel to channel
crosstalk needs be 120 dB or better.
The distribution amplifier needs to have low input and output VSWR.
The distribution amplifier output should have relatively low distortion.
Mixer LO and RF port VSWR has to be low.
A very low noise zero crossing detector is required.
The RF connectors used should be TNC, SMA, N or other stable connector
definitely not BNC.
The temperature of the interconnecting RF cables needs to be stable.
The mixer IF port should be capacitively terminated to reflect the sum
frequency back into the mixer. This reduces the mixer output noise and
enhances the slew rate of the beat frequency signal.
Both mixer ports should be saturated for maximum SNR.
etc.,

To justify the complexity of such a system the sources to be measured
need to have good short term stability (hydrogen maser, Cryogenic
sapphire whispering gallery resonator, mercury Ion standard,
Oscilloquartz 8607 or similar high performance OCXO, etc.,).

Bruce





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