[time-nuts] Short-Term Stability

SAIDJACK at aol.com SAIDJACK at aol.com
Thu Mar 16 15:42:23 EST 2006


The DDS system you describe in your paper is very similar to our FireFox  
Signal Generator unit (_www.jackson-labs.com_ (http://www.jackson-labs.com) ).
 
We included a GPSDO than can achieve parts to the E-011 accuracy in it  basic 
form and about 10x better (being quantified now) with a slightly more  
expensive high-stability double-oven OCXO.
 
The output ranges from 0.00001Hz to 1640MHz with a 1GHz, 10 bit DDS  driving 
the 0-400MHz output directly, and the frequencies above 400MHz are  generated 
by frequency multiplying the DDS output using a low-noise VCO and  PLL.
 
Some results on the DDS output: we achieve DAC generated spurious  below 
about -52dBc with the 10 bit DAC clocked at 1GHz. Drawbacks of this DDS  are some 
digital noise crosstalk, and over 2W of power consumption at 1Gs/s! (We  are 
desparately waiting for a 2GHz, 12bit DDS :)
 
The big advantage of using this DDS output, as mentioned in the paper,  is 
the frequency hopping capability up to over 400MHz output frequency  
(essentially instantaneous, <<1ms) since its a digital system. The hopping  time thus 
depends realy only on the output reconstruction low-pass filter.
 
Another advantage of using a high speed DDS is that it essentially  doesen't 
increase the phase noise floor like a wide-band VCO would do. So using  a 
1Gs/s or 2Gs/s DDS you can choose the best single frequency VCO you can get to  
drive the DDS, run it with a very small PLL loop bandwidth (say <10KHz) and  
thus get extremely low phase noise. The DDS will thank you with even lower phase  
noise at its output (20log(fin/fout) relationship).
 
This compares to having to use a wide-band VCO for multiplying up the  
frequency, which inherently has higher phase noise due to its wide band  capability. 
The VCO doesen't generate wide-band spurs of course like the DDS  does, if 
designed well.
 
Another huge advantage of using the DDS is that using a DDS and some tricks  
we can get about 49 bits of frequency resolution over the bandwidth. This, in  
combination with the internal GPS disciplined OCXO gives a nice capability  
to really generate essentially any frequency with extremely high  accuracy. 
 
So the trade-off when using a DDS versus PLL frequency multiplying is less  
phase noise, but higher spurs.
 
BTW: we are working with one of the members of this discussion  group to try 
to get the best accuracy possible from the GPSDO  (thanks much for the help!!).
 
Let me know if you would like to get more detailed info on how  we 
implemented the GPSDO and DDS, and the caveats we ran into etc.
 
SJ


More information about the time-nuts mailing list