The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Star connections, good or bad?
https://designers-guide.org/forum/YaBB.pl?num=1239402236

Message started by Godfrey on Apr 10^th, 2009, 3:23pm

Title: Star connections, good or bad?
Post by Godfrey on Apr 10^th, 2009, 3:23pm

It is common to star connect different circuit blocks to the power/ground in order to prevent noise transfer.

However, I have heard that it is actually better to strap analog supplies to digital supplies because the capacitance of the (digital) supply is much higher and this can actually reduce noise. This sounds wrong, but since the star connection is on the chip side of a bond wire, the vdd pin is not really the supply and the vss pin is not actually the ground.

Does anyone have a reference to the arguments AGAINST star connections?

Title: Re: Star connections, good or bad?
Post by rf-design on Apr 11^th, 2009, 2:06am

A star connection is the simplest way (some argue stupid way) of controlling the interference of ground loops.

There is is no argument against a random ground connection. But you have to know exactly for your critical signals what is the signal voltage loop and how it is shared which more noisy currents. A ground or a supply tree helps to made that work a simple calculation independend of current redistribution effects.

With a full chip extracted simulation (signals are Rc extracted while supply metal is extracted only by R) you can see the effects. But then a lengthly story of finding the sharing loop beginns and next often you have to relayout the hole supply nets.

So non-engineering is not substituted by more tooling.

One of the biggest supply loop sharer is the bond wire. So typical spending pins is expensive but the return is high.

For a quick starting change your thinking; imagine that each sensitive voltage is delivered as a loop, a voltage with a reference. Also differential signals have a ground reference because the common mode rejection is not infinite or depending on differential signal.

Start with the star and begin thinking in loops.