[time-nuts] low noise voltage regulation, again

Bruce Griffiths bruce.griffiths at xtra.co.nz
Thu Feb 18 05:29:40 UTC 2010


Scott Burris wrote:
> I'm finally thinking about building some real power supplies for my 
> Thunderbolts
> and LPRO-101, and that got me thinking about the topic of low noise power
> supplies again.
>
> I think Dr. Bruce and I exchanged some messages about this a year or 
> two ago,
> when I was looking at "RF regulators" and the conclusion seemed to be 
> that
> the ancient ua723 was still the low noise champ.  Is that still the 
> case, or
> are there better modern regulators?
>
> I was thinking of doing strictly linear regulators, but then I read this
> note at Wenzel:
>
> http://www.wenzel.com/documents/finesse.html
>
> I think of this as the "short out the noise before anyone sees it" 
> approach.
> Do you think the circuit of figure 4 can clean up the output of a 
> switching regulator
> down to 20nv?  I assume this was written as a post cleanup of a linear 
> regulator.
>
Yes its intended as a cleanup regulator and it doesnt short out the 
noise its a balanced bridge type cancellation circuit.
They can be cascaded and their performance can be improved.
Its also possible to use an opamp in a feedback circuit to reduce noise 
but one then has to have cotrol over the output capacitor 
characteristics etc.
> Alternatively, what do you think about these Maxim circuit ideas which 
> claim
> 7nv and 6nv:
>
> http://www.maxim-ic.com/app-notes/index.mvp/id/3656
This only produces 3.3V output.
You need to be careful that the selected regulator is stable (ie doesnt 
oscillate) since the filtering is included in the regulator feedback loop.
Such filtering may degrade the regulator line rejection.
The 7nV is presumbly at 1kHz?
Noise at lower frequencies will be greater.
> http://www.edn.com/contents/images/608157.pdf
>
The line rejection wont be particularly high as the opamp supply is fed 
from the unregulated input.

> Is it better to apply cleanup circuits like this to average regulators or
> go back to something derived from the 723 or its kin?
>
Starting from a quieter regulator is probably better as less filtering 
is then required.
There are also the ultra low noise super regulators favoured by the 
audio nuts.
These typically use an opamp, an LM329, a few resistors, a LED, some 
capacitors, plus some discrete transistors and diodes

> Just casting about for ideas about what I should be experimenting with.
>
> Scott
>
Bruce




More information about the time-nuts mailing list