[time-nuts] V standards

SAIDJACK at aol.com SAIDJACK at aol.com
Tue Dec 2 07:57:00 UTC 2008


Hi Warren,
 
actually, we need to differentiate between phase and frequency errors;  Allan 
variance is typically a measure of phase drift over time. It's not quite  the 
same as a spectral power plot on a spectrum analyzer, but it's  related.
 
Some things to keep in mind:
 
* A 12 bit DAC may have an error of 1/2, 1, or even more than a couple  of 
LSB noise/monotonicity error/etc. Every DAC has a minimum quantization error  of 
1/2 LSB.
 
* On a 5V 12 bit DAC, you are going to switch the OCXO with a massive  1.22mV 
jolt when the dac LSB changes (if the DAC doesn't have any other errors).  
Most Euro-can OCXO's have about 8Hz/V frequency sensitivity, or even more than  
this, so 1.22mV could actually result in 9.76E-010, or almost 1ppb  frequency 
change!
 
On a Micro Crystal DIP-14 OCXO the change would be closer to 15Hz/V, so  
almost 2ppb steps per LSB. Not good.
 
When controlling a Rubidium with only +/-0.2Hz overall control range this  is 
a totally different story of course.
 
* GPSDO's don't really have "tracking times", they rather just use low-pass  
filters with natural cutoff-frequencies around 100s to 1000s (1/100Hz or  1
/1000Hz). Meaning they measure the phase error, low-pass filter this phase  
error, and apply the result to the DAC (oversimplified of course).
 
* The overall error of the system is a combination of the DAC quantization  
error (1/2 LSB) PLUS the error of the OCXO PLUS the error of the DAC voltage  
reference PLUS the inherent DAC errors.
 
All noise sources add since they are typically stochastic, so to expect an  
Allan variance to be the same as just the DAC quantization noise won't work 
out,  since we need to take all of the other noise sources into account too.
 
Lastly, to measure Allan variance down to xE-011 per second, we can use  
Agilent 53132A counters, TSC5110A or TSC5120A measurement systems, CNT-90  
counters,  Wavecrest DTS series counters, or other 12-digit per-second  resolution 
time intervall measurement systems. Most of these would work by  measuring phase 
drift between two oscillators.
 
So there is no easy answer to your problem, but yes - depending on your  
requirements a 12 bit DAC may just work out ok if you have a very good OCXO  with 
less than +/-0.5Hz overall adjustment range (1E-07 at 10MHz).
 
Hope that helps,
bye,
Said
 
 
 
 
 
 
 
In a message dated 12/1/2008 21:48:01 Pacific Standard Time,  
warrensjmail-one at yahoo.com writes:


BTW  I've  head from others that a 2.44E-11 resolution Dac works OK for  
most applications.  More is better of course. Maybe there is some  
additional filtering going on or maybe it is they just can't measure short  
time period noise very well. That comes out to about 10 ps or so of 1 sec  
phase noise.Not many instruments are going to be able to measure  that!
What is it you use using to measure the 2.44E-11 freq noise 
at 1  second? 




More information about the time-nuts mailing list