flawless PCB design: Bad ground ideas many engineers believe - Part 6

flawless PCB design: Bad ground ideas many engineers believe - Part 6

Other parts in this series. These really help to get the full understanding of this video.

Get my free Electronics Product Development Checklist

Check out my courses

In this series, I'm going to show you some very simple rules to achieve the highest performance from your radio frequency PCB design for your electronics project. I'll do this by showing you how to improve the grounding and reduce the crosstalk. These rules apply to both analog , digital and RF electronics. I'll also share some rules of thumb for PCB parasitics that you can use in electronics simulations. I'll prove my points with actual measurements on dedicated test PCBs. In this sixth part I'm showing a number of bad concepts that are widely promoted as great ideas which can really ruin your day.

00:00 Introduction
00:19 Star grounding
04:30 Multiple ground planes
07:29 Why a single ground plane prevents interference between blocks
08:34 The via wall
11:38 Bad module pinnings
14:09 How to prevent mistakes
14:50 My attempt to be funny :-)

[xxmrrickxx]

I think a lot of your advice can also be extended to system design, not just PCB-level. I've found that in the world of power electronics it is still common thinking that 60Hz is the only frequency needed to be suppressed and filtered. Now that everyone uses solid state switching devices with digital controllers, everything has become an RF device. I've often had to help engineers troubleshoot comm issues by bringing out a spectrum analyzer and showing that the output of their inverter design with power cables routed in a circle made a 1 MHz transmitter that is killing their ethernet link :-)

[PeterA]

Thank you for debunking the DGND/AGND myth that some companies still propagate.

[yelectric]

Fascinating. Thank you for this advice. That ground return through the capacitor is beyond clever... I am slowly understanding how much I dont understand. Thank you.

[rfrisbee]

One point that was a bit misleading is that return currents are highest directly under the trace, but spread out further the lower the frequency is. At lower audio frequencies the return current will spread out quite a bit. For low level circuitry this probably isn't much of a problem, but a power amplifier with 10's of amps sloshing around is likely to not work as desired if a ground is being used as a global reference for the signal.

[DreadDeimos]

Having a Master in Metrology, I wish these kind of things were teached at my uni. Not only they're very useful for design (Metrology doesn't have a lot of that), but the concepts you use for explanation would greatly simplify the understanding of all the electronic courses I took.

[johnrobertson]

I realized that a (what turned out to be a) star ground configuration in pinball games (from the 1970s and early 80s) made by Gottlieb - back in the mid 1980s - was leading to coils and transistors burning out due to biasing shifting as the ground connections deteriorated.

I published this in the trade journals of the day, but Gottlieb never followed up and their games continued to fail - leading to an undeserved (an easily fixed problem!) reputation for failure in the field. All it took was ensuring that the ground connections at the power supply and logic board was shared with better connections. This defeated the Star Ground, and as I am not an engineer, just a self-trained tech I could only describe the problem and the solution as I found it.

I can say that games that had my ground modifications stopped failing and became as reliable as any other solid state machine. I wish I knew you then and could ask you how to do a better job or at least explain better than I did what was happening.

I won't post a link to the ground cure I developed - it can be found with internet searches I'm sure - I've said enough for now.

[HC]

Finally an electronics-engineer who's worth watching! 😃
Great video. I usually fall asleep when I watch YT videos that are more than 10 minutes, but not this time!
I guess Hans will make a very bad politician... 😉

[maxpetrelli]

Thanks Hans, your videos are very clear and instructive. You are helping a lot people with your clear explanations. Thanks again.

[rfrisbee]

One issue with the four layer SIG1-GND1-SIG2-GND2 stack up when transitioning between SIG1 and SIG2 is that on a 1.6mm GND1 is a long way away from SIG2 relative to GND2, so it's still a good idea to put stitching vias between GND1 and GND2 close to where signals change layers. Also, if trying to maintain a constant impedance, the trace widths necessary on the two signals layers will need to be different.

I think the whole star grounding for audio thing is popular because consumer equipment generally uses ground as a universal reference for all signals in the system, which is a convenient fiction that cannot be realised in practice. A better solution is to route all signals as differential pairs over a solid ground plane, using ground as a local reference. It is often quite easy to convert single-ended circuitry based on opamps (or with a differential input stage) to differential IO by adding an additional resistor to the non-inverting input, with the other end routed along with the signal to the next stage. This compensates for any difference in the potential of ground between the two sections of circuitry.

[jrodrigo]

This channel has a huge potential. Keep it real.

[NicleT]

That was very informative. Great content on your channel. Immediately subscribed!

[talkinghat]

Great sharing on common grounding problems. I feel young all over again. Thanks sir.

[TomStanley-piqi]

Like your videos, informitive and to the point. I'd like to see more on grounding and trace layout for analoge audio with power supplies and different signal levels.

[tangiblewaves]

Thanks so much for this! Absolutely brilliant explanation and super helpful! ❤❤

[MrHaggyy]

Great video, and awesome interaction in the comments. I'm really looking forward to new videos.

[hardywoodlandcypress]

I always figured that the star ground thing was mainly to avoid problems at DC levels between different boards etc., especially with high current like power distribution. But I'd always be happy to bolt grounds to the nearest piece of sheet metal (chassis) with a piece of braid. But up to now I've been fooled by the separate AGND and DGND thing. Instinctively, I always thought it was a bit suspect, because one quickly runs into problems deciding exactly where to put the tie-point, with any non-trivial layout. Now it makes a lot of sense what you say, which is to just have a single ground plane and try to keep the different types of signals in their own area. Great videos; I'm leaning a lot, thank you!

[AMalas]

8:34 there is a slight sneaky layout error here! Most PCBs do not have equally spaced dielectrics, so in a common four layer board today, your stackup looks not like:
[1 2 3 4] but more like....
[1 2 3 4]

So the signals on layer 3 are only very, very weakly coupled to the ground on layer 2, but strongly coupled to the ground on layer 4. Using layer 4 for power or other signals will give you trouble. Ofcourse, dont forget to have a nearby return via for the ground connecting both ground return currents!

[yunqingli]

Thanks Hans! Your video series really helped me as a newbie in this field a lot, as I'm a software developer trying to make some PCBs of my own! :)

[xxportalxx.]

One problem that's always scratching at the back of my mind is that there's two grounding rules that I find are often at odds with one another: one being that you want to keep the return path as close to the signal path as possible, preferably with no other signals sharing that path, and the other being that you want to prevent the formation of ground loops. In a pcb with ground planes and short distances this seems easy enough to rationalize, but in larger assemblies with multiple pcbs, modules, or even sub assemblies it seems like you get one or the other in most cases.

[ctid]

Thanks Hans.. My new favourite channel !