star ground vs. ground plane

[Sebastiaan]

Dear All,

Most bridge/parallel PCB designs on this forum use ground planes for the LM chips.

Myself had problems with using ground planes for a LM3886 bridge/parallel design. Of course that can say more about my design skills then about ground planes but when I designed the next version with a consequent star ground THD+N improved highly, and noise get lower. Also the sound quality improved.

Also the National application sheets promote against ground planes.

My personal feeling is ground planes should only be used for digital designs or low-power opamp designs. Not for power amplifiers with high current flows in the ground scheme.

Maybe I am wrong or missed the boat somewhere, that is why I ask to share some thoughts. Pro’s and con’s for a ground plane. I also wonder why the designers on this forum decided to use ground planes for their LM3886 design.

With best regards,
Bas

[okapi]

I have recently been spending quite a bit of time thinking about the same question.

see: LT Dual Tracking 3A Regulated Power Supply 1.25V to 20V

If a star ground showed improvement, my understanding is that high and low power grounds were likely being mixed (close to a sensitive circuit/component). One possible solution (instead of a pure star ground) is to use two or more different planes for high and low power parts of the circuit and then bring the planes together at a star.

One possible problem that pure star grounds can cause is high trace impedance because star grounds inevitably require many individual traces and space constraints may force the traces to be thin. Thin traces can cause the trace impedance to be significant.

[pacificblue]

Trace impedance is solely determined by its length, not by its width. Trace width is good for low trace resistance, but increases capacitance.

The ideal ground has a potential of 0 V. A real PCB ground does not have that potential due to resistance of leads and traces from the power suppply. The higher the current the more voltage drops across those resistances and the further away from zero is the ground potential on the PCB. If you have a common ground for power and input, the input ground potential will change with the output current, changing the reference for the amplifier all the time. Therefore you should try to move the common grounding point as near as possible to the point of the lowest (and most stable) potential. Usually that is either slightly behind the main smoothing capacitors in the power supply or the point, where PE enters your amplifier. It can also be the RCA connector, depending on the layout of the other components.

A ground plane is good for low power applications, because the low currents there don't lead to a significant change in the PCB's ground potential.

My impression is that the designers, who use planes for chipamps do it for aesthetic rather than technical reasons.

[AndrewT]

Originally posted by pacificblue

My impression is that the designers, who use planes for chipamps do it for aesthetic rather than technical reasons.

My impression is that the non-experts, who use planes for any circuit, do not know what they are doing and further do not know why the ground plane does not work as expected.

[okapi]

Hi Andrew,

i think you are saying that if you know what you are doing, a ground plane can work just fine.

can you elaborate on where you think it is best to use a ground plane or a star ground and possibly comment on any trade-offs between the two approaches?

thanks

[AndrewT]

From the many reports, papers and comments I have read, I know I do not have the competence to design in a plane whether it's a ground plane or a power plane.

[hollow_man]

In my experience, neither design practice is panacea. We have had a mixed signal telecoms design at work, which used split analog and digital ground planes, suffer bad signal integrity problems. This can easily happen in practice, because often systems need to connect to the outside world at more than one point, e.g. power connector, various input/output connectors etc. And this is where split ground planes can actually do a lot more harm than good, by letting 'dirty' currents flow through sensitive circuit areas. In such instances, a solid ground plane is much preferred, although it is extremely difficult to eliminate circulating currents completely.
Star earthing is good as long as the ground trace impedances can be kept negligibly small for the job, and again no circulating currents are introduced by wrong start-earthing point locations.
In either case, one has to think very hard where unwanted currents could flow from connections to the outside environment and how to stop these from being added to the low-level signals.
Tough one!

[Sebastiaan]

Dear All,

Thanks for the input so far.

I think the main problem with ground planes for a BPA300 design is the fact that as well the high current output stage as the analogue inputs (since in all designs so far analogue signal is unbalanced internally and the chip inputs refer to ground) refer all the to the same ground.

Measurements turned out that PSU current flows get modulated at the line inputs from the chips which all refer to the same ground plane. And that explains for me what the THD+N ratio drop significant.

In one post above someone offer the option to use different groundplanes for power and line. In multilayer circuits this is a way to go. One ground plane for power, and one ground plane for line level and e separate ground return for the output. And those all joined together at the best common point closest to zero volts.

Other solutions would be to make the internal line signals balanced truth differential and floating above ground. Some virtual ground circuits for line applications would serve very well too.

The Jeff Rowland LM3886 design use ground planes with very good result.

With best regards,
Bas

[Mr Evil]

I think the main reason not to use a ground plane is that it is really hard to do right with only 2 layers to work with.

One advantage of a ground plane is the high-Q distributed capacitance. This makes for pretty good filtering of extremely high frequencies, but that's not really important for audio. Of course you don't actually want this capacitance to appear everywhere. You can fix that by removing the ground plane from underneath the affected tracks, but then you compromise that other advantage of a ground plane - low inductance.

I thought I had a bunch of links to useful articles about this sort of thing, but I can only find one, and it's not specific to the low-frequency analogue that we are interested in.

[okapi]

it has been suggested in another thread here on diyaudio to use a combination of star and a plane:

http://www.diyaudio.com/forums/showt...542#post887542

what do you think of the above idea?

Also, i recently designed two pcb's, one with a star ground and one with a ground plane, for a simple power supply built around 3 pin adjustable regulators. Which would you pick and why?





** for the purpose of this discussion please ignore the slice in the ground plane. i placed that there to allow for experimentation.