grounds in active crossover

[georgiano]

I'm designing PCB for Rod Eliott's active crossover

here is the link to the project
http://sound.westhost.com/project81.htm

as there is no high currents involved, is star grounding necessary?
may I skip it and do groundings in more free manner (of course with no ground loops)

 

[justblair]

I'm designing PCB for Rod Eliott's active crossover

here is the link to the project
http://sound.westhost.com/project81.htm

as there is no high currents involved, is star grounding necessary?
may I skip it and do groundings in more free manner (of course with no ground loops)

Watching closely, I have this design in mind for my next amp, gathering the parts as we speak.

What package are you using for the design work?

Care to share the files?

And how do you plan implementing, are you putting the whole crossover on one pcb. I was thinking about on multiple repeated crossover sections to produce a modular system.

Haven't started the design of this bit yet, so I cant really answer your question with suggestions....

 

[georgiano]

I'll be using Eagle 5.0, shure can share files.

any help with grounds, anyone?
also, is capacitor type selection critical?

 

[sek]

Hi,

will you have the PCB manufactured (sent out), or do yo DIY?

I'm asking because doing the PCB yourself has a whole lot of implications on your PCB design, including grounding. Through-hole capacitors cannot be soldered from the bottom side, layers (board sides) can only be changed at component pins (or dedicated pin vias), etc. And yes, capacitor choice can have a significant influence on frequency accuracy and sound quality.

But generally, ground planes are as good as star grounds, as long as care is stil being taken on the current paths, impedance levels and IC decoupling measures. In essence: do your layout as if a star grounding scheme was being used. Once this has been accomplished, a ground plane is easier to design/manufacture and can be still as good.



Cheers,
Sebastian.

 

[gainphile]

I've done many active crossover, all on veroboard. Yes, star ground is very important, otherwise you will get hum.

 

[sek]

I've done many active crossover, all on veroboard. Yes, star ground is very important, otherwise you will get hum.

I would argue this, although I agree that proper grounding is mandatory. But as veroboard won't allow ground planes, we'll never find out.

This would of course all be different in case we were talking high frequency applications...

 

[georgiano]

which grounding scheme is better?
do you see any obvious design flaws? (ground-wise)

here are pdf files for schematic and PCB
www.georgiano.net/images/crossover.pdf
www.georgiano.net/images/crossover_1.pdf
www.georgiano.net/images/crossover_2.pdf

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

I higly approciate any help

 

[AndrewT]

I count ten items connected to the ground trace around the PCB edge.
Some are connected to the IC pins and others to passive devices.
I suspect you are combining clean and dirty returns into a series of linked impedances (traces with resistance and inductance). = big problems.

 

[sek]

That and this is not a ground plane!

Also, you could shorten some traces (and thus board size) considerably.

 

[georgiano]

is it that bad? both of them?

do you mean by ground plane double sided board with second layer as a ground?

as I have concluded from what I've read on the net low current grounds should "flow" into high current ones, just like rivers. and if there are no level differences grounding should be no problem.

wrong?

 

[gfiandy]

Hi, You should seperate the grounds for the power supply return and the signal return with seperate paths back to where the ground enters the board. The signal and ground should, if possible, follow each other through the circuit and this should only be joined to the star ground at the input to the board.

However I think you have another problem I can see no local decoupling on any of the opamps. This risks instability in the opamps. Each opamp should have a 100nF cap from its power rails to ground, these grounds should return to the sart point seperately from the signal ground. For the opamps on the far side of your PCB from the PSU it would be sensible to add the option of another pair of bulk decoupling capacitors.

So bottom line neither of your options is optimal and you are missing some parts from the design. You may get away with and the circuit might work but it won't be optimal.

Regards,
Andrew D.

 

[jzagaja]

Can I use virtual GND in active crossover? It should be decoupled with some capacitance, how high?

 

[sek]

do you mean by ground plane double sided board with second layer as a ground?

No, not neccessarily, although it is an option.

What I meant is more of a true star grounding arrangement with a ground plane then spread atop. The key concept here is: impedance vs. current.
Intertwining high impedance return paths with low impedance return paths would let the currents through these paths induce voltage drops across these paths.

So, as Andrew pointed out, power supply return paths shouldn't be shared with signal return paths. Ideally, different power or signal return paths should be split completely.

If all this is accomplished, in a sense the resulting network supports the idea of your rivers flowing from smaller streams into larger streams. But the trick is to still prevent signal return streams from interchanging with power return streams. Imagine a hierarchical structure, with the signal return path stream set strictly disjoint from the supply return path stream set. And with the respective stream sets as small, short, low in count and in overlap as possible.

Note that a complete ground plane on the other board side also fulfills these requirements, just with excess copper outside the current path/stream area. Even with the 100% fill approach, splitting signal ground plane and supply ground plane will help. This can be seen on mixed signal (digital/analog) boards with the ground planes on different board areas, and also on multilayer board designs for high frequency applications.