[sdiy] dare I say - decoupling?

[Magnus Danielson]

From: "JH." <jhaible at debitel.net>
Subject: Re: [sdiy] dare I say - decoupling?
Date: Sun, 4 Sep 2005 00:06:59 +0200
Message-ID: <002401c5b0d4$fae69d20$0200a8c0 at jhsilent>

> > > Would it be ok in the feedback loop of an op-amp like in some of the old
> > > Oakley stuff?
> >
> > Doesn't help. The point of Jürgens arrangement is that you locally consume
> as
> > much current at any time, regardless of the LED (or LAMP) being on or off.
> > By doing that, there is a much lower shift in current which needs
> decoupling.
> > The benefit is that the same DC current is pulled and there is almost none
> > shift which needs to be decoupled.
> >
> > Just tossing it into the feedback loop of an op-amp will not help to
> acheive
> > the constant DC-pulling.
> 
> Magnus is right.
> 
> But this configuration can give us an example for something else:
> 
> Asuming the opamp with the LED in its feedback loop (opamp 2) is used in
> inverting
> configuration (+inp tied to GND, input resistor connected to -inp
> acting as virtual GND), driven by another opamp (opamp 1).
> 
> Then all the AC current from LED driving is running thru the positive and
> negative supply,
> leaving the GND completely undisturbed.
> 
> And then you want your bypass capacitors from opamp 1's positive supply pin
> to
> opamp 2' negative supply pin, and possibly a second cap the other way round.

This is where a double-opamp chip will fit in nicely to make the decoupling
a little more simpler, since you then only decouple across the op-amp. The
ground rail is only used as a reference and is thus only weakly loaded.

It is however a cost that comes with this, each LED consumes a double op-amp.
If you want to acheive something similar at a cheaper price, I think that
maybe a transistor pair and a resistor would do the trick. The transistor pair
is hooked up as a diffrential pair with the LED on one leg and the dummy
resistor in the other. The led and dummy resistor both rests at the upper
rail and the emitters would see the usual current-limiting resistor. The dummy
resistor is chosen to cause the same current load for the same voltage drop as
the LED. Thus, the transistor pair only balance the current between the LED and
the dummy load. The benefit of this method is that you have a diffrential pair
input that may come in handy. The transistors can be any dirt-cheap transistors
capable of handling the current and voltage.

> (Not talking about the caps you need for stability - just talking about the
> caps to keep AC load currents local.)

Indeed.

Just to make people understand the subtle point of keeping DC current changes
local, a large known instrument maker has a nice instrument doing fairly high
resolution TDR when equiped with the TDR Sampling heads, which can do normal
sampling and is also equiped with TDR diodes. To let the user know that the TDR
diodes are active, that the input is now hot, they made the LED next to the SMA
blink RED. Since this is a precission instrument, averaging was a bit of a
mess. They fixed that mess by having the LED go solid RED when having hot TDR.
A simple software fix for the LEDs caused the averaging to get much better
result. What happend was that the shift in current caused sampling point shift
in quite measurable time and that caused a time-diffussion that was not
expected. Which reminds me that I should have a firmware upgrade performed.

Cheers,
Magnus