Ground Isolation

isolate the grounds and you can connect them at just one point. Keeps ground loops from happening. you can even use a ferrite bead to connect the two grounds, keeps the trash out.

Whenever one sees "digital and analog grounds" you can expect trouble.

John

Isn't there a school of thought where you just try to make the ground as low impedance as possible and don't deal with separate grounds. On a chip, the grounds might have to be within a diode of each other, depending on technology. From a marketing standpoint, the chips have an analog and digital ground because if you combine them on chip, there is no fallback position to separate them.

The chips have digital and analog grounds so that current in the digital ground pin doesn't make the analogue ground bounce. Both should be tied through the shortest possible path to the same solid ground.

(What does marketing have to do with that??)

Jeroen Belleman

snipped-for-privacy@j19g2000prh.googlegroups.com...

If only it was so simple. Isolating grounds at high frequency with a ferrite bead is more likely to cause problems than solve them, and blindly applying such "rules of thumb" without understanding the cause of the problem is a troubleshooting technique commonly known as "Easter egging" - not known for it's high rate of success.

It is necessary to understand the parasitic circuits which are causing the problem in order to deal with them properly, and no one can tell you how to do that in a short message. Read "Grounding and Shielding" by Ralph Morrison for a good introduction to the analysis and mitigation of the effects of parasitic circuits.

Glen

Every board that I do has one solid ground plane, which I bolt to the metal enclosure at every opportunity. I handle any ground loop issues locally.

Separate grounds implies that the grounds are at different potentials, a recipe for trouble.

Mixed-signal chips often have separate analog and digital ground pins to reduce ground bounce from wirebond inductance. I nail both to the board's ground plane.

John

I would first try and figure out how the 'ground noise' is getting into your system. Is it 100kHz 'crud' from the H bridge? Is it magnetic coupling from the coil? (Can you substitue a resistive load?) Once the source is identified (Or at least the biggest source.) Then the solution may be obvious. Does your circuit sit inside a metal box?

George H.

I'll second that opinion. Meaning I do the same on my designs. Unless there is a safety-related reason to split, such as full patient isolation or far apart industrial control modules.

I thought the common term was "winging it" :-)

We have a bunch of VME modules with isolated channels, usually 8 to 16 per board. In that case, each channel has its own little ground pour rectangle. We generally have a small cap, hundreds of pF maybe, from the channel common to real board ground.

John

Tek has a few great scopes where they did that as well, and the signal (and probably power) is transferred via very flat cores that cradle PCB traces. That was a great piece of engineering, unlike their TDS2xx series.

I always wonder how they dunnit. Dragged the old crews out of nursing homes for just one more gig?

All you have to do is watch your power return paths.

Cheers! Rich

And, of course, that's what optoisolators and 20 mA current loops are for. :-)

Cheers! Rich

:-)

Someone posted a bunch of pictures of the innards of a more modern Tek scope, the DPO 4034, here:

Maybe I accidentally blew by it but I didn't see a close-up of the couplers in there. I think they were very flat cores going through slits in the board.

You won't get 40MHz BW signals from noise floor to >100V through any of those, got to pull some other tricks out of the hat :-)

... and when you see a cherry-red glow in those, pull the plug :-)

I didn't see any couplers either, but he doesn't show anything underneath the shields around the probe inputs where presumably all the "interesting" analog bits would be.

Plus that scope doesn't have isolated inputs anyway, so there might not be much in the way of wideband magnetics.

I was thinking of it more of an example of the type of workmanship you get with Tek today.

--Joel

Go 7 times higher or 700kHz. This will get you the first three odd harmonics.

The reason to separate planes is to avoid common impedance noise on a cable somewhere. If you tie the grounds together then run the currents through the same cable you defeat the purpose. Make sure your two circuits are powered separately and that your processor has a regulator on the board near the processor. Tie the grounds together at the power supply such that there are separate current loops.