Split Ground/Power planes for sensitive analog, vs glitchy digital

This isn't a noise issue; it's a matter of keeping a section of ground equipotential at low frequencies.

But I thought the issue was having separate analog and digital grounds that meet at one point under some delicate part. I can't agree on that one.

It's a thermocouple input. Nothing in that section is high speed.

Oh.

"They are joined?" What does that mean? Is a single ground "joined" to itself?

Sounds like new-speak gendering, where one person is "they."

Hmmm... so if they are not "equipotential" at low frequencies, that's not a noise issue??? Sure sounds like a noise issue to me! If you preserve the "equipotential" aspect of the analog section of the ground plane by making cuts in the plane, how is that different from what I propose??? What exactly, are the cuts in your ground plane separating? How do you choose where to put them?

It's not that the part is "delicate". You will find nothing to show I said that. The parts would be devices like ADCs, which have both high speed digital (often arranged on one side of the chip package to allow this sort of ground plane design) as well as a sensitive analog section.

Of course. That's the circuitry that is completely within the analog section of the ground plane, separated from the rest by the cuts.

But that's not the question I asked. Again... "if this corner of the board had a high speed analog signal coming in with a 500 MSPS ADC, and some noisy circuitry to the left, would the ground plane be continuous or cut into sections as you have done here? Where would the split be?"

Please try to answer this question as simply and directly as possible, rather than dodging the question.

Your work is different. Mostly you design products that you sell. Many people here can't show their designs. I recall one guy who was designing toys and could not even tell us what he was designing.

This should be of no surprise to you.

I suppose it should not surprise me that you resort to being snarky about this. You know exactly what I mean. All along, I have talked about two *areas* of the ground plane, joined at an optimal point, so that it is a single ground plane electrically.

Now, that it is very clear you acknowledge using the same technique, you are trying to hide by not discussing the facts, and playing your usual games.

This is about the point where you will become offended, saying I am being snarky, and refuse to discuss the actual topic further, or to answer my questions in a meaningful way. That's ok. I'm not expecting anything different.

In fact, this is an implementation of the classic star-ground approach, where the connection between ground areas is a bit of copper foil an inch or two wide.

Joe Gwinn

Yeah, to some eval board and that's often it. With academians I meant people tasked with teaching, research and publications. As found at universities or ivory tower labs of large corporations. They often have no or very little exposure to real "electrical life".

It is almost generally not possible in real life. It somehow seems you don't want to believe me. Then just go ahead and split the grounds, it's not my board.

Another (noisy) IC, transistor stage or whatever somewhere off center

It allows more current to flow, making it a more efficient (unwanted) antenna that receives stuff.

Resistance has almost nothing to do with this. It's the impedance that counts and, most of all, where stuff couples in.

RF from, for example, large motors starting up, nearby lightning, welding equipment, a big AC unit being hard-switched on, et cetera.

Other paths is via externally connected sensors, power sources, monitors, user interfaces and so on. I had explained that.

If you had read the explanation of the gamma match I brought up you'd understand. But maybe you don't want any advice, then so be it. I've seen plenty of folks who didn't believe it. Their designs failed and mostly during EMC testing.

They are nearly always connected somewhere because they have to. And that's when the problems arise.

He is not afraid, he uses full ground planes just like I do. Because that makes sense.

We don't. If you don't have an audio application I'd never even consider a split. Because in audio when you do it that causes a lot of other grief such as susceptibility.

I do but that's because the grounds are _not_ connected. Even the power is supplied in an isolated fashion, no connection.

I do not agree. Cutting voids into ground planes generally reduces the nosie immunity of circuits and, therefore, I do not do it.

Nonsense. But go ahead, split the grounds, it's your design.

If you read the thread carefully, you will see the class I took was taught by Lee Ritchey, about as far from an academician as you can get. I don't know why you are talking about them.

Your use of "It" is misplaced and unclear. But thank you for the permission to continue to design my own boards. That's very generous of you.

Ok

Again, ok. So, you are claiming that Larkin's use of cuts in the ground plane will cause it to pickup voltages from lightning and motor starts?

Connection of external grounds is a whole 'nother problem. It is easy to create ground loops that will mess up any design.

My designs have always passed EMI testing, first time!

Which is why it's better to talk not about "split planes", but planes with cuts. I'm sorry I used the term "split planes". This obviously confuses you greatly.

If you believe that, you didn't read his posts. He posted a link to an example where he added cuts to his ground plane to create a low noise region, one to isolate the ground currents so the low noise region was not disturbed. Maybe this link will work. If not, you will need to dig for his post.

?rlkey%3Denbp8fd2l451jkxmv9jqz8blr%26raw%3D1&source=gmail&ust=1690942233249000&usg=AOvVaw1ECzm002Td0u3S4Y_5zY4L Does this design not have two ground areas?

And yet, it works. You have acknowledged that you cut ground planes and Larkin has provided an image of one of his designs with a very significant cut in the ground plane, creating two ground areas. Why are you now denying this?

Ok, so with no ground connection, this is not relevant to this discussion.

And yet, you said you did. I'm not sure what you mean by "voids" though. What is this "void" you are talking about? Maybe that's the problem. We are talking about two different things.

You are very gracious. Thank you.

Snark? Me?

No, I don't know what you mean. Please post a pic of an actual board.

Picture, please.

I posted one.

You have to admit, that sometimes, Larkin is very funny.

Please explain how it is different from the plane splits I talk about.

Show us.

If you can't understand what I've written, why have you been arguing so long that I'm wrong? My design is largely like yours, but the two areas are separated by a 0.05" gap, connected by a 0.4 inches wide join.

Everyone can see easily that you don't really want to discuss this. You are trying to say that your isolated ground area is not the same as mine. But there is no significant difference. The isolated area is virtually all analog with a max frequency (of interest) of 20 kHz. The only digital signals are control lines to configure the circuit through analog switches. The digital lines are set, and not changed until the analog circuit needs to be reconfigured.

Breaking out the analog area with a connected ground, but surrounded by a moat, just as in your design, prevents any stray ground currents from flowing through the analog area, introducing voltage drop in the analog region. This is a daughter board with the off board, ground connection in the same area as the analog/digital join. It works very well.

Instead of playing games, why don't you explain why your isolated ground area is ok, when you seem to claim the concept is fundamentally flawed? I know that others have said the concept is fundamentally flawed. You seem to have said that too, before you said that you also use the idea.

All hat, no horse. Show us a board.

John Larkin can't talk about why he designed his boards to turn out that way that they did - probably because design is an intellectual process and he doesn't understand what he is doing - so he wants people to show him images.

Larkin is frozen in place. He doesn't understand words. He only knows pictures.

Whatever. He has very clearly been using the same technique that I've described. So when he says it can't work, he is talking about his own designs. It's not surprising that he can't explain his designs. He has said all along that he seldom actually designs circuits, rather than experiment with them until he can make them work "good enough". That's likely why he doesn't understand he is using exactly the same technique I'm talking about.