[time-nuts] fast freq. synthesis schemes

BriMDavis at aol.com BriMDavis at aol.com
Wed Oct 21 01:50:33 UTC 2009

Jim Lux wrote:
> Look at how the PTS synthesizers work.. A series of decade modules, 
>lots of adding, mixing, filtering. Off the shelf they do sub microseconds.
 Following a DDS with mix divide stages works great, but can be 
very hardware-intensive.
 As an example, the Wavetek 5130A is esentially a PTS 160 with the 
lowest decades replaced by a decimal DDS; each sucessive decade 
division stage knocks down the DDS spurs by 20 dB.
( PTS and Rockland/Wavetek must have had a licensing agreement
at some point, as I have an older Rockland 5600 synthesizer that 
_is_ a PTS-160 but with Rockland labels on the outside. )
 The final PTS output stage does bandwidth expansion using an
offset drift cancel loop with of a comb of the master OCXO.
>  A DDS feeding a multiplier (not a PLL mult, but a step recovery 
> diode ( or similar) also will have very fast switch time, and might 
> even be phase continuous.  Your challenge with this approach is 
> going to be spurs.
 Yes, the close in DDS spurs can really kill you when multiplied. 
 Even when not multiplied, the normal close in phase noise floor of
the DDS can be masked by spurs and intermods near many tune words.
 I put together some plots of this a few years back, modeling just the
numerical spurs ( phase truncation and and amplitude quantization ) but
_not_ the harmonically related intermods, near some 'pile-up' frequencies.
 There are a few off-by-two errors in here that I need to fix, but
the plots still illustrate the problem well ( put Acrobat in single
page mode and mouse wheel scroll to 'animate' the plots ) :
  _http://fpgastuff.googlepages.com/dds_oddities.pdf_ (http:/
>  The spurs might not be as bad as you think... The newer crop 
> of DDS include some forms of error cancellation for close in 
> spur reduction, so the "in the loop bandwidth" part is cleaner.  
 I haven't kept much track of this in the past few years;
is there a DDS now in production having some sort of close 
in noise shaping?
 ADI has a patent using feedback of the phase error, but I don't 
think that'll help much with the amplitude quantization spurious:
 They did have a 2006 press release about licensing something 
else from Pentomics; this looks like more of an spur analysis 
tool and CORDIC-like algorithm than a noise shaping or error
cancellation scheme:
 I think the ADI "spur-killer" parts have only two(?) cancellation 
channels, intended for dominant spur cancellation :


More information about the time-nuts mailing list