[time-nuts] re low noise regulators

[SAIDJACK at aol.com]

In a message dated 12/14/2007 15:00:00 Pacific Standard Time,  
bruce.griffiths at xtra.co.nz writes:

>>  A simple RC filter of say 2 Ohms into 4700uF has a -3dB cut-off at  
around  
>> 17Hz (4700uF caps are getting quite small these  days). That would take 
care  of 
>> most of the 100Hz to 10KHz  noise.
> 
>It still helps to have as low a noise as possible  before using such
>brute force filtering to get that extra few dB of  noise reduction.
>> Using a typical current of 0.16A at 12V for a  Euro-can OCXO we would only 
 
>> have 0.32V voltage drop across  the resistor.
>>   
>However the resultant drop will  depend on the ambient temperature.
>In principle the regulator output  could be given a suitable temperature
>drift characteristic to  compensate.
>However doing this without degrading the noise at the  filter output may
>be challenging.
Hi Bruce,
 
the tempco effect of the series resistor may not be a factor at all for  
typical OCXO's.
 
If we take a standard 1% resistor such as the Panasonic ERO-S2PHF2R00  
(available on Digikey) with 100ppm/C tempco (that's a fairly bad tempco),  then we 
get:
 
   0.32V * 100ppm = 32 microvolts per Degree C change (actually  it's 
+/-50ppm, so this example is worse than what we would see in real  life).
 
If we now take the Supply-Voltage-versus-Frequency characteristic of  a 
typical standard single oven OCXO  of 1E-08, at +/-5% variation for  a 12V part we 
get:
 
   1E-08 for 1.2V change, then we get: 8.33E-09/V * 32E-06V =  2.66E-013 
change per degree C.
 
For this particular OCXO with a stability of 1E-09 per degree C, the  change 
in frequency due to the change in the 2 Ohm resistor value is about 3700x  
less than the temperature stability of the OCXO itself.
 
In other words the effect of the resistor is  inconsequential.
 
For a good double-oven OCXO this is even less of a factor, for  exampe the 
MTI double oven we use only has 8.3E-11 per volt change -  two orders of 
magnitude less sensitivity than the example above.
 
The capacitors' microphonic sensitivity may actuallyy be more of a factor  
here.
 
>It still helps to have as low a noise as possible before using  such
>brute force filtering to get that extra few dB of noise  reduction.

It's not just a few dB, it's -3dB at 17Hz already, and that drops  at ~-20dB 
per decade. At 170Hz it could theoretically give -23dB already,  or in other 
words if the noise floor was entirely due to the supply voltage  noise, then 
the difference would be between say -140dBc/Hz and about  -160dBc/Hz at 170Hz - 
this could be a very significant improvement.
 
You are absolutely right, it is very difficult to get low noise  <10Hz.
 
bye,
Said





**************************************See AOL's top rated recipes 
(http://food.aol.com/top-rated-recipes?NCID=aoltop00030000000004)