Ground Fill Trivia

Personally, I've historically done prototyping on the Vector 8013, which doesn't seem to be available any more - on one side it's got pad-per-hole, (it's got holes everywhere on a .1" grid) and on the other, a "ground plane" which is more like a grid or mesh, because it's got enough naked space around each hole so that the chip leg shoulders don't short to the ground plane.

But I've never needed to hack the ground plane itself - when I need to hack the prototype, I just redo it.

But as far as fill, doesn't your layout program have a "flood fill?" It seems like plunking down polygons could get tedious. I've never had the luxury of a proper layouter - in my heyday, I was laying out boards by hand. (I did fills partly with that red tape, and partly with pen and ink.) ;-)

Cheers! Rich

Sounds like the Vector 8007. I last used one in February for a client emergency--it had an AVR in it, so dead bug wasn't as attractive as usual. 'Twere it for my own use, I'd probably do normal perf board construction for the micro, and do the analogue part dead-bug on the ground plane side.

Cheers

Phil Hobbs

It calls that "polygon" -- you define a polygon, it fills it with either solid or hatched copper. So the only work is to define the edges of the polygon.

This particular tool (Eagle) has the unfortunate propensity to assume that any via that's punched inside the polygon will have something to connect to, so you have to change the polygon net, route, then change the net back to ground. But it's not that bad to do, and it makes a nice fill.

And when I need swaths of ground plane upon which to dead-bug that emergency bit of circuit -- there it is.

That's a great hint, Tim.

I am a proponent of solid fill, so I have a side question; is it better to use hatched or solid fill and please explain the justification.

Hatched uses less copper; there might be a difference in signal propagation if you're operating at frequencies where the dimensions of the hash mesh might be on the order of more than, say, 0.1 wavelength. It also makes it easier to tape up by hand and be sure that your tape sticks.

But I'd think that at those kind of frequencies, they'd have already gone to ceramic substrates, strip lines, and so on.

Hope This Helps! Rich

Tape? What century are you in?

As to .1 wavelength, that's about 1/2 inch at 2.5 gHz. Hellova hatch.

I use hatch fill because I'm paranoid (probably baselessly) about the solid copper warping the board, because I generally get prototype boards without solder mask and it's easier to control solder on a hatched plane, and because a hatched plane is easier to whack up!

Oh -- and it looks keen.

None of these are very good reasons, but I don't see strong reasons to go to solid fill unless you're doing high current work or really high frequency stuff.

I've been told about the warping before, but I wanted another opinion. Thanks for that confirmation.

Our layout guy is always worrying about differential copper but we've yet to have a problem with it. It's likely one of those problems that is ancient history. It may even go back as far as FR-2. ;-)

On my two layer boards, I usually have a ground fill on the bottom, and then have power fills on the top, especially when I have mulitiple power zones in the design. This gives you that low value distributed 'capacitor' that can help ease some noise problems. I might also have some ground fills on the top if that helps lower ground impedance for some SMT parts...

Charlie

I finally went to always using soldermask, even on prototypes. It is too easy to bridge SMT pins especially on dense parts if there is no mask there...

Charlie

Here is one I did recently

You did not specify the frequency range in which the circuit operates. Any microstrip construction require a solid ground plane.

Anyway, if the actual circuit is operating at low frequencies, you still need quite good ground planes due to the EMC requirements close to the connectors/PCB edges.

Are PCB manufacturers actually refunding for the etched out copper ?

Nah, he means without going to 4 layer. I do this all the time, you can get a pretty good (i.e. continuous) ground plane for most microcontroller/mixed signal circuits. Route it 2 layer, then minimise the length of traces remaining on the bottom layer. This allows you to flood fill the bottom layer, with a few clear islands remaining for tracks that need to dip down momentarily.

Here is a 2 layer design with a LPC2478 with external SDRAM.

...until things get so small that you can't get soldermask between pins. For the small money, it still makes no sense to leave it out for the 99% of the cases where it works. It also adds some protection for the traces.

If impedance matters, it makes no sense to stay with two layers. IMO, it rarely makes sense anyway. Of the eight or nine boards I've done so far this year, one was two layer. It had four SOPs, a TSOP, three LEDs, and maybe a dozen passives. ...and operated at finger speed. ;-)

My last board was six layers, to try to help EMI problems from the 180MHz and

Does the ground actually do anything? Wouldn't it be better to sake the ground pins directly to ground rather than adding that slab in the middle? Oh, I see the ground is exiting on the bottom left. I'd think you'd be better off with a pour exiting the bottom left. I'll stick with four (or more) layers. Solves a *lot* of problems.

I'm not sure I understand what you are saying.

if you add a via in a polygon it will not be connected to the polygon, because it is on it's own net unless you route to it _or_ give it the same name as the polygon

-Lasse

That would be why I assign the polygons to ground. But if I put the polygons in first, then I end up with a bunch of ground connections that go nowhere. This is deleterious to circuit operation, so if I need to re- route the board from scratch, I either rename the polygons or I delete them before autoroute, then replace them afterward.

Well, the 21st now, but when I was doing hand taping, it was the 20th. ;-)

And, as a free-lance layouter making starvation wages, I couldn't afford thousands of dollars for layout software. )-;

OK, fair enough - but these days guys talk about clock frequencies that used to be the in the realm of klystrons an' s*it! =:-O

Thanks! Rich