[sdiy] PCB layout rules

[Tim Daugard]

From: "Antti Huovilainen" <ajhuovil at cc.hut.fi>

Reversing the order:

> Has anyone experimented with sense/force ground?
> Idea being that sense ground has minimal current through it and is only
> used for references and no loads.

I use what you are calling a "sense" ground in most of my modules!? As one of the other
posts said; opamps don't actually use the ground. In +/- supply systems the grouhd is the
reference at the center of the range. With a single sided supply like mine, the half way
point has to be created. Here is many ways to do this. My methods are (in order of
expense):

1) A two resistor voltage divider and a capacitor - giving a 6db reduction of any noise on
the power supply line after the regulator. This is used for non-critical applications such
as using the opamp as a comparator (outputs are either high or low and exactly how high or
how low doesn't matter.

2) A three resistor voltage divider with two caps (two low pass filters in line with a
known load to provide the voltage. This reduces the power supply noise by 12 db.

2a) A three resitor voltage divider with two caps and the center resistor being variable.
This is used to maximize the output swing of the opamp without clipping at either rail.

3) A diode stabilized voltage divider for greater accuracy and less power supply induced
noise. (Think zener.

4) A dedicated local voltage regulator for the module.

Note that on the resistor and diode voltage dividers, I try to make the voltage dividers
consum at least 10 times as much current as the opamp circuits draw from the voltage
divider. this keeps the opamp from overly influncing the bias point.

Note also: I call this a bias reference instead of a sense reference - but it doesn't
matter what you call it. I've been influnced by too many tube circuits.

> Doesn't this mean that the ground might get contaminated by power supply
> noise? From what I've understood, a cap is only needed between the rails
> for stability.

Yes. Even with voltage dividers all points get contaminated. The easiest thing to do is:

1) Consider the ground as the absolute reference. What ever the ground potential is,
that's the ground.

2) Consider current flows. The old RS-232 standard and one previous to that was current
referenced. A current flowing in one direction was a mark. The current flowing in the
other direction is a space. In simple terms each circuit would source a current and that
current would be sunk when the loop returned it to the original circuit. Noise voltages on
the line didn't matter as long as the didn't blow up the circuit. It is the current in the
loop that is the desired signal.

Even with the ground carrying all the various currents, it is still relatively noise free
if the circuits deal with current loop inputs and not voltage inputs. Of course this plays
hell with trying to process fine control currents for VCO's! 1 mA per octave anyone?

My wife is dragging me off to do yard work. I can discuss more later if any one cares.

Tim Daugard
AG4GZ 30.4078N 86.6227W Alt: 12 feet above MSL
http://home.sprintmail.com/~daugard/synth.htm