[time-nuts] Low noise power supplies - dont' use Electrolytics

[SAIDJACK at aol.com]

Were getting of off the original thread about Electrolytics versus other  
caps a bit..
 
>A switcher at 2.2MHz does not have single frequency spikes in its  power
>spectrum, but a rather wideband distribution.
 
Yes, that's Fourier 101 basics.
 
But what matters to Time nuts is if the switcher is running close to 1.0MHz 
 or 2.0MHz, as the 10th or 5th harmonics of these would fall right into the 
phase  noise spectrum of the 10.0MHz output and may show up on the usual 
phase noise  plots.
 
It is also easier to filter out a ~2MHz carrier than a ~1MHz carrier from  
AM-modulating our 10MHz oscillator through the power supply lines. Most of 
the  noise will likely be on the input of the buck switcher, not the output.
 
With a 1MHz switcher, you could have a nasty spur at 10.01MHz right in the  
spectrum of interest. Running at 2.2MHz moves that spur 1MHz away from the 
10MHz  carrier, and the possibility of the switcher "injection locking" with 
 the 10MHz oscillator is reduced. This is easier to accomplish with a 
2.2MHz  switcher than a 1MHz switcher.
 
>but a load-dependent feedback loop that will
>change duty cycle  or frequency (depending on device) to maintain a target
>output. Check for yourself.
 
The part I listed as an example is a fixed frequency device. Stay away  
from variable frequency devices for anything Time Nuts related - that is my  
opinion.
 
>Try to get >90% efficiency for instance, from 24V down to 3.5V 5A.  
Unlikely you can work at 2MHz.

A) why would you need 3.5V at 5A for anything time-nuts related?
 
B) you could parallel two or three 2MHz devices to get 5A output power, and 
 synchronize them to each other
 
C) If you are consuming 17.5 Watts at 3.5V in the first place, then  why 
would you care about 90% efficiency versus 85% efficiency in time-nuts  
related projects? The difference is less than one Watt, and getting 85% from 24V  
should be possible with the switchers that are out there.
 
We now have commercial Cesium Vapor atomic clocks running at 0.12W and  
less, so it should be possible to get power consumption down way below  17.5W 
for most time nuts related equipment I would think...
 
bye,
Said
 
 
In a message dated 11/28/2011 16:43:03 Pacific Standard Time,  
camaraq1 at quantacorp.com writes:

A  switcher at 2.2MHz does not have single frequency spikes in its  power
spectrum, but a rather wideband distribution. It is not a fixed  frequency,
fixed duty cycle oscillator, but a load-dependent feedback loop  that will
change duty cycle or frequency (depending on device) to maintain  a target
output. Check for yourself.

Try to get >90% efficiency  for instance, from 24V down to 3.5V 5A. Unlikely
you can work at  2MHz.