The Chronicles of GND, Part 2

Single Supply = Multiple Problems?

When I get a printed circuit board layout back for review, there are a few things I look for first. I can usually get a feeling for what kind of board designer has been at work, too, specifically if s/he is (or s/he thinks s/he is) a ‘digital’ designer, an ‘RF’ designer, or a ‘power’ designer.

You might have formed the impression by now that I can be something of a picky customer. I know what I like in my PCB designs. In my younger years I designed a fair few – over 300, from what I can piece together from the archeological residues of a lifetime in electronic engineering. And while some of these designs had modest digital content, the vast majority were for systems intended for low- to medium-frequency analog signal conditioning and processing done to (I hoped) a world-class performance level.

When the first PC-based board design tools appeared, I threw myself into board design with gusto. At last, control over how I wanted things done! No more would I have to lean over the shoulder of the guy with the arcane layout skills, the tape and the scalpel, and yell “No, route it there! Route it there!”. (We are still friends, by the way, after meeting the first time some 47 years ago during my first job interview, when he was testing a home-build amplifier while on his lunch hour. Our old crew tries to meet up each Christmas for drinks.)

Naturally, I didn’t know nearly as much about how to do good layout as I thought I did, and I made plenty of mistakes. “The man who never made a mistake never made anything,” my dad used to preach, and it’s a wise observation. I’ve made plenty of mistakes in the past, in multiple contexts, including in the search for how to get the best performance out of a circuit when it makes that transition from your head to FR4.

So, let’s get back on track (spontaneous and unintended pun alert). What do I look for first in a design? Well, since this is another chapter of The Chronicles of GND, it won’t surprise you to learn that my first checks are GND-related. Specifically, related to a system being, as is nearly every design I encounter these days, a ‘single supply’ system. What does that mean?

Well, analog circuitry from the Golden Age of Operational Amplifiers, during which I learned much of my craft, was almost always powered from a ‘split supply’, typically +15V and -15V by the time I started in the industry. Instead of having two terminals on your power supply box (plus and minus), you had three (plus, ‘nought volts’ and minus). The reasons for this value are lost in the fog of time, but one thing is clear. Much of the development work on the kinds of analog circuits I’m referring to – amplifiers, integrators, and filters – came from early Analog Computers. In such systems, voltages (and sometimes currents) were the ‘analog’ of a variable or result that the computer was required to ‘compute’. And since that result could be either a positive or a negative number, it was sensible that the circuits could operate with positive or negative voltages, with respect to a reference that represented a result of zero.

The circuits in a typical ‘split supply’ system draw current from the plus terminal of the supply and return it to the minus terminal. Sometimes a little current may end up routing to ‘nought volts’, usually due to resistive voltage dividers and resistors from various circuit nodes to ‘nought volts’. But ‘nought volts’ was never thought of as a supply rail like plus or minus, even though it’s a terminal on the power supply box. It is a reference point, where your scope ground clip or the (mandatory, of course, see part 1) second lead of your two-leaded voltmeter would connect. It was also likely to be connected to a nice ground sheet as a shield under the insulating surface of your bench, and to the safety grounds of all the test equipment that kept the lab warm in winter.

Now of course ‘nought volts’ was the spiritual precursor to GND. In fact, we often called it ‘ground’, and things connected to it as ‘grounded’. Since it was often connected to “green/yellow”, the electrical safety ‘earth’, it was even possible that it found its way back to some metal buried deep in the actual ground, snuggling up to Mother Earth itself. So fundamental are the mythic origins of GND.

Fast forward to today’s ‘single supply’ systems. Here your circuits have a positive supply, but they don’t have a negative supply. Or (pause for theatrical effect…) do they? Well, yes they do. It just happens to be at (ostensibly) the same potential as GND. Effectively, the value of the negative power supply voltage is zero. So that means you can just connect it to GND and be done with it, right? No one in their right mind would go to the bother of building a power supply whose output voltage is zero.

Well, maverick that I am, I’m here to tell you that – in principle at least – that is exactly what you should do to get the best and most predictable results from your board layouts.

This is because that negative power supply has a job to do. It has to sink the current ‘coming out’ of the “negative” supply pin of each of your amplifiers, comparators, and whatnot. And this current is likely to be signal- and load-dependent (remember the Kirchhopamp from the Bypass Capacitor sequence). You do not want this flowing, on its way back to the power supply box, through all the same conductors that direct your lovely clean GND reference to the sensitive points of your amplifier circuits and converters.

For anything (in our analog world at least) that is a ‘power’ pin required to be at GND potential, you should (IMHO, or perhaps that should be IMH0…) connect it to a separate net, often called 0V (‘nought volts’!), to represent this negative power supply of value zero. 0V will have to be connected to GND at some point; this is best done at a ‘star point’ where the power actually arrives at the board or subsystem. This connection should not be done as fixed copper on the board that your layout software can recognize as conductive. I sometimes define a small copper component object that I can place as needed. The netlist should contain two completely distinct nets, GND and 0V, so you can eyeball the netlist to be completely sure that pins are connected to the correct nets. You do do a visual read of your design netlists, don’t you? Become familiar with one of the older simple, easily-readable netlist formats and export your netlist into that format to check it for misspelled and mis-attributed nets.

Some people call 0V “dirty ground”, though really it has nothing to do with GND at all except that it is very close in potential to it.

By this point you might be asking “What’s wrong with letting all these currents mix and just having a super low-impedance ground plane to keep all the voltage drops small?” Well, small isn’t zero, for one thing. In some systems, microvolts can matter. And there are some very specific reasons in op-amp based circuits why you should be suspicious about the currents coming out of those negative supply pins. For instance… no, I’m going to leave you hanging on that thought until part 3.

Agreements and disagreements are welcome through the usual channels. Has something about a single supply system ‘ground’ you down?