[time-nuts] Low Phase Noise buffer
warrensjmail-one at yahoo.com
Fri Feb 19 11:14:48 UTC 2010
I'll take a guess at the answer.
White noise is constant power over Frequency
If the Nose Bandwidth is reduced from 100MHZ to 10 Hz then power would be
reduced by 1e7 to 1
Voltage is square root of Power so 1e7 power ratio is = 3e3 to 1 (1/3162)
The 10 ns noise would be reduced to 3ps
Or put another way, if you want under 1ps ADEV nose,
Need to keep the Buffer's wide bandwidth phase noise down around 1 ns or 1%.
> Magnus said
>>White noise goes down into the lower range and do contribute to the
> Can you put some numbers to it to show how much effect there is.
> If the 10 MHz had an incredible total of 10 ns (10%) white noise at say
> 100 MHz bandwidth, what would be the Phase noise for the 1 sec ADEV
> optimum BW of 5 Hz?
>> The white noise which is the only dominant noise at higher frequencies
>> goes down into the lower range and do contribute to the ADEV measures
>> there, even if filtered. This filtering will affect lower-tau measures
>> more than mid or high tau measures. Telecom standards prescribe
>> filtering and tau0 for their reference measures for repeatable measures.
>> The filtering has been debated. The estimators will filter themselves
>> regardless when m has become a number of multiples, essentially when the
>> filter bandwidth becomes large compared to the inverse of the tau.
>> If the noise is benign white noise it should be kept at a low level, but
>> work on flicker noise is where most effort should be spent.
> WarrenS wrote:
>> T&F Nut question
>> Is the high frequency phase noise of a buffer important when the
>> Oscillator is used just as the Reference for taking ADEV data?
>> The current (2005) writings say that ADEV phase data should be bandpass
>> filtered which removes all the High freq edge jitter.
>> If that is the case, Does a buffer's added HIGH FREQUENCY phase noise and
>> jitter have any effect on the ADEV numbers when using modern, high
>> resolution, mixer type, Time interval counters? Of course the low
>> frequency type of buffer noise would still be important.
>> Previously there was some confusion and disagreement on the subject. What
>> are the latest opinions?
More information about the time-nuts