8
\$\begingroup\$

Recently I had posted a seemingly innocent question on another forum. I was trying to keep a small and relatively simple circuit down to a 2 layer board and was asking about the shielding merits of a copper pour tied to analog ground, as opposed to dedicating unbroken separate plane used for ground. Though my question was only in reference to the shielding value of such a copper pour, a major controversy started on an entirely different point. In my layout I had created a virtual ground using the typical voltage divider feeding a unity gain configured OP amp, and then had run every single analog ground return to a point adjacent to the output of that virtual ground. I had also done a copper pour on the least busy layer and tied it to that same analog ground, again at that same single point. But there soon ensued a series of posts lambasting my use of such a convoluted "spider web" scheme, with many chiming in that I should just create a ground plane and just make all connections to the nearest available point of that plane ( through via when necessary). Well doing so certainly makes for a cleaner and simpler layout visually. But having gone the "single point" route for so many years, and having mostly successful (meaning quite and stable) designs, I'm hesitant to change. Especially as some of the arguments spoke of single point grounds as something archaic, and relegated to vacuum tube circuits. What?

So anyway, just poking around, my research on the subject seemed to show two camps on this subject. One argument is that when an unbroken ground plane is available, it is always superior. The other argument says that if its a high frequency circuit a ground plane is better, but for a low frequency circuit (including audio), the single point ground method is better, mostly to avoid ground loops.

Of course I still have a dilemma, as prototype costs are constraining me to a 2 layer board if possible this time, and this means that my pseudo ground plane will, at best, be a copper pour, broken by some short traces here and there. But before adding that complexity or possible exception to the rule, I'd like to put this specific question out for general discussion: For an audio design involving OP Amps, when is a Single point" ground scheme the bast way to go, and when is just the simpler "nearest path to a ground plane" the better (or at least an adequate) choice.

By way of example, these two layered screen shots show two reasonably similar versions of the same board, which in reality is about 4" x 2.5". In the first you can see lots of long traces on both sides of the board culminating at a single pad labeled AGND. The second is nearly the same circuit, but this time the blue copper poor area is part of the analog ground net, so all those long traces have gone away in favor of the nearest path to the ground plane/ copper pour. Aside from all possible critique about other layout issues one might see, this is just an example. I'd really like to confine this discussion to the original question.

single point ground version single point ground version

Nearest path to ground plane version Nearest path to ground plane version

\$\endgroup\$
  • \$\begingroup\$ This is a great question! (From my limited experience) I like the single point ground for DC- audio (maybe to 100kHz). (It's also somewhat gain dependent.) Once you get above that (f>1MHz.) I have found it better to have one big ground with multiple connections to the front panel/ case ground. I'm guessing there is some connection to the return currents of the signal, at some point you care more about the inductance of the ground line. \$\endgroup\$ – George Herold Oct 21 '14 at 15:59
  • \$\begingroup\$ You know, I've almost begun to think of inductance of those long ground paths as my friend. After all, they would seem to inhibit gain at frequencies where you wouldn't want gain in an audio circuit! And yes... on the gain question, this particular circuit is designed to be near unity all around, and only deal with line level (around 100mV) signals. So its arguable in a case like that it wouldn't matter much. But I think I'd better have an open mind if one way or another is truely better. Now aside from the routing hassle, have you experienced any negative effects of single point grounding? \$\endgroup\$ – Randy Oct 21 '14 at 16:09
  • \$\begingroup\$ Grin, well a Gain of 10^4, BW DC- ~2-3 MHz. The first single point ground pcb didn't oscillate...but almost. It had a wicked gain peak out a high frequency. At high frequency the inductance is not your friend. I think it leads to a phase shift. Have you looked at any of the signal integrity stuff, where they talk about where the return (gnd) currents flow? \$\endgroup\$ – George Herold Oct 21 '14 at 16:20
  • \$\begingroup\$ I have been hearing that at high frequencies, the ground plane is the better choice, likly because of that inductance issue. The trouble, of course, is that even a circuit intended for audio might oscillate at some unwanted high frequency. So I've gotten into the habit of adding those extra RC networks to stave off Mr Murphy. I guess regardless of which method is best, neither frees us from adding those circuit blockades against such problems. What's worse... inductance of long traces, or ground loops on ground planes! ;-( \$\endgroup\$ – Randy Oct 21 '14 at 17:14
4
\$\begingroup\$

There are pros and cons for either.

A signal full ground plane has the advantage that signal 0V is a high integrity 0V and can be relied upon but, not when there are currents flowing of any significance. Despite a ground plane being very low impedance, volt drops can still occur when significant currents flow. How much volt drop being a problem depends entirely on the smallest signal that you wish to amplify.

A star point system avoids those "significant currents" mentioned in the previous paragraph by making sure that tracks carry signals (0v return tracks) do not share with these significant currents. The down side is that you end up with magnetic loops that can have voltages induced from one another and the "significant current" flowing in a different track can be coupled.

\$\endgroup\$
  • \$\begingroup\$ Well I appreciate that. Your thinking reflects mine, and I have some history there. Now lets take a case where most of the ground references are pure references carrying minimal current. Like the + input of an OP amp referencing virtual ground through 1 meg ohm. And lets further assume its an all around low current circuit, where all gains are voltage (not power) gains. Would you say it even matters at that point? And what about a hybrid approach... where those pure references are tied to a plane, but all true signal returns used isolated traces back to the origin of the ground source? \$\endgroup\$ – Randy Oct 21 '14 at 21:52
  • \$\begingroup\$ I can also see that being confined to two layers, and using a copper pour in place of a solid plane, the need to put some traces on the "pour" side has to be minimized. In my second example, I can already see where I've shot myself in the foot, forcing ground currents between the lower right and the left side of the board to have to wrap their way around those long traces near the bottom middle. Since then I've added some vias to re-rout all of these to the top layer, to make my copper pour a step closer to a ground plane. \$\endgroup\$ – Randy Oct 22 '14 at 15:54
1
\$\begingroup\$

Sorry for reviving the old (sort of) thread, but OP (and other readers as well) might find this document of some interest:

http://www.maximintegrated.com/en/app-notes/index.mvp/id/5450

As I understand the above document, one should distinguish between high (1MHz and up) and low (1MHz and lower) frequency currents as these have different return current paths. Hence a solid/single ground plane can be used as long as the traces carrying signals of different frequencies are grouped together (i.e. high frequency signals with other high frequency signals on one side and low frequency signals together with other low frequency signals on the other side).

\$\endgroup\$
  • \$\begingroup\$ Interesting, and thanks. I'll say that in mixed signal designs I've done, such as the article describes I've always tried to use split ground planes for dig and aud areas. So far have had good results, but then again I seldom work with very high gain, and my analog work is in the audio realm. Also, I've opted for simple ground planes in several audio designs since my post, and the boards have been well behaved. As all of my ground points carry near zero current, I suppose it doesn't matter much. I suspect many of the star pattern args are most critical for hand wired perf board prototypes. \$\endgroup\$ – Randy Mar 31 '16 at 13:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.