shorting of analog and digital grounds

Or slot it with a global keepout over the slot to keep current where you want it and signal traces away.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

If you slot the ground plane (I no longer recommend it), please don't run any high speed signal lines over the slot, especially digital signals. Doing this will force the return currents to flow around the slot, turning it into an excellent radiator (read "antenna").

Tom

That's why the keepout.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

In article , John Larkin wrote: [... splitting ground plane ...]

If you split a ground plane to avoid thinking about things, you are doing it for the wrong reason and are likely to have trouble. The reason to split the plane is because you have to think about more things by the nature of the design.

If you have 20A pulses of current and 10uV low frequency signals on the same PCB you need to think about a lot of things.

--
--
kensmith@rahul.net   forging knowledge

Of course it's not really a problem if the return current is low.

Graham

In article , John Larkin wrote: [...]

[.. faints and recovers ..] you split planes

BTW: Digital sections almost always drive you to having more layers than the analong needs. If you fill the unused space by making more ground planes, you can improve matters.

You can bring all the switcher stuff together on a local plane above the system ground and then take it to system ground at one place. Ideally, the one point of connection is the same as where all the input and output filters come together.

--
--
kensmith@rahul.net   forging knowledge

No, at least that's not what we call a split plane; a split plane has different nets, by our definition (ie, by PADS.) I meant to just cut a little C-shaped insulating path into the plane. It's still fully connected along one edge. You can align the orientation of the C with respect to known or suspected ground plane current directions. All we're doing is steering circulating currents around the tender parts. You can similarly do link-shorted gaps to break thermal paths, to reduce thermal gradients in sensitive circuits like thermocouple amps. It's still all the same node.

I can't recall ever having more than one ground layer in my stuff, or pouring ground regions on other layers.

Again, I don't do that. It would tend to make the switcher ground have a lot of ac noise relative to real ground, which could radiate ringy things. I just always use a single ground plane.

Oh, another heresy: I bolt the pcb ground plane to chassis ground in as many places as possible.

John

I do that, and worse, routinely. Splitting planes generally makes the emi situation more hazardous, and emi can wreck a 10 (or 0.1) microvolt signal just as badly as a ground loop, but it's a lot less analyzable. We go for max emi integrity first (single ground plane, hard-bolted to the chassis often) and then work around any low-frequency loop problems locally.

I know people who use separate analog and digital ground planes with a single-point connection, then pepper the board with big plane-to-plane bypass caps everywhere to kill hf plane potential differences!

John

That's not usually a real problem. For normal (say, 1 ns or slower) risetimes, a reasonably-sized slot is invisible. And more importantly, a thin slot in the ground plane is usually spliced over by the presence of nearby (by design!) power planes. With thin dielectrics, the ground+power plane structure is really a single ac-equipotential structure. HoJo and his apologists conveniently ignore plane-plane capacitance when they analyze "return currents" and bypassing, and proceed to make some mighty silly recommendations. The gap is actually shorted by adjacent planes.

Just TDR a trace that runs over a ground-plane slit on a multilayer board. Even at 20 ps risetime, not much happens.

John

That occurs in water re-use situations. I've not heard of Legionnaire's from an ordinary "swamp cooler".

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
     It\'s what you learn, after you know it all, that counts.

Hello Jim,

And mind the legionaire's disease that spreads when the water didn't change state for a long time ...

Regards, Joerg

Usually there's so much mineral content that the sump won't support any form of life ;-)

When I had one (~15 years ago) I had a bleed so that there was always a fair amount of fresh water in circulation.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
     It\'s what you learn, after you know it all, that counts.

Hello Jim,

Maybe. When I walked into our local HW store one day they just turned on the swamp cooler. One guy mumbled "Oh rats, they didn't drain it last fall" and one heck of a stench filled the air.

Regards, Joerg

Hello Jim,

Not out here. They keep the tap water at high chlorine levels to avoid forms of life in there. Last time I measured it was almost 2ppm. Problem is, when it sits in that pan on a hot roof the chlorine is gone within a week.

Regards, Joerg

Hello John,

I think what Spehro meant by global keepout is truly a 100% keepout. No parts, traces or planes crossing that slot.

Still, I am not a believer in slots. I usually take them back out of existing designs. Slots can make nasty antennas and can lead to egg in the face when an unexpected redwood forest shows up on the analyzer at the EMC lab (that is usually when my phone rings off the hook).

Regards, Joerg

Just planes- more like what John was talking about.

Too bad most of the time the layout can't be dictated by electrical considerations. There's always some high voltage or current stuff, a heatsink or a terminal block or a big chunk of routing that forces things to be sub-optimal from an electrical pov. Well, maybe it isn't too bad. If it was easy anyone could do it.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

In article , John Larkin wrote: [...]

Its been a few years since I did my own layout. I was using Orcad PCB. I'd do this:

******************************** ******************************** **** ******** **** ****************** ******** >You can bring all the switcher stuff together on a local plane above the

No, it makes the ground of the supply less noisy and less ringy. The copper under the switcher is an added plane. Its impedance is as low or lower than the one that extends over the whole PCB.

At UHF frequencies, the capacitance between the two planes effectively shorts them together. At the frequency the switcher runs at, it ensures that all the AC currents remain under the switcher and don't go crawling all over everything else.

Do you used to work for Power10? No, on second thought don't tell me.

They tied things to the chassis in many places. As a result they introduced about 1.5Vp-p spikes into all the sheet metal on a 19 inch rack. If you hooked scope ground to one back corner of their chassis and the tip the the far front corner you'd see all this crap. We had to create isolated rack mounting hardware and stuff to use them.

--
--
kensmith@rahul.net   forging knowledge

Hello John,

That's because you only design high-end stuff. There are situations where it all has to be on single layer phenolic because it's a nickel cheaper than double sided FR4.

Same here. Then I make sure that the plating and surface coatings are compatible so that a low impedance path stays low impedance over the years.

Regards, Joerg