[sdiy] "backwards" connected positive regulator (was: Problem with723 PSU)

[jhaible at t-online.de]

>----- Ursprüngliche Nachricht -----
>Absender: harrybissell at prodigy.net
>Betreff: Re: [sdiy] "backwards" connected positive regulator (was: 
Problem  with723 PSU)
>Empfänger: jhaible at t-online.de
>Kopie-Empfänger: synthusiast at gmx.net , synth-diy at node12b53.a2000.nl
>Datum: 18. Apr 2001 02:09

> My experience is that the "two positive regulator" approach is 
inferior
> in most respects. 
[...]
> Disadvantage:
>  Two Transformer windings (as you mentioned)

Yes, but many transformers come with separate windings anyway.
Especially toroidal transformers. I guess it's easier to bring
out two windings than to center tap one winding.


> Ground is not truly "ground" 

Sure it is. With everything else floating, I can pick *any*
node and call it "GND". The real question is, would the
negative voltage, referring to the GND ( = reference pin of
positive regulator be of different quality in both 
cases. I am convinced it is not. 

> it is actually the positive supply of the "negative"
> regulator. This is likely to be more noisy than the real "zero"  > 
voltage point of the pos/neg regulator pair.

A voltage is always a difference in electric potential between two 
points. Every regulator, positive or negative, will stabilize the difference 
between two of its pins - nothing else. I don't see 
why a negative regulator should do this job better than a positive
one. There are different circuit topologies in both regulators,
but if there were a substatial advantage for the negative type,
the answer would of course be to use a pair of negative regulators
then !

At first glance, the concept of a common GND from center tap of 
transformer winding to capacitor to positive and negative regulators'
reference pins (in the symmetrical pos/neg case) looks like an 
advantage. But in truth it is deceptive, as this common trace
is far from being the desired "GND" star point along its whole path.
(I learned it the hard way, when I thought that 10mm difference
wouldn't be harmful in my Modular - you surely know the story.)
So, even in the "symmetrical" configuration, the only thing to
rely on is the voltage difference straight across the reference
and output pins of your regulator chips. The rest is "floating"
here as well, floating on the voltage drop along that supposed GND path 
to the transformer center tap. That voltage being determined
by the difference of power consumption in upper and lower path,
and shaped by the ripples of your rectifier charging the big
capacitors.


> Current dumped into ground will likely bounce the lower (negative)
> regulator.
> Transient response may suffer.  They really are not quite like 
batteries
> in series...

They are, as long as everything else is floating. This is not
perfectly so (capacitive coupling etc.), but if this really
has an effect on stability, I'd like to see the mechanism.


Add three more disadvantages for the dual positive approach:

* harder to understand
* needs one more bridge rectifier
* more drop voltage because of 2nd bridge rectifier

I think the well-defined GND star point is a good
argument *for* separate windings. (Regardless if you choose
78' x2, 78' + 79' or 79' x2.) In systems with mixed signals
(weak - strong, analog - digital) freedom in routing and 
connecting separate GNDs is a blessing.

JH.