Examples of BAD Ground in PCB Layout And How to Fix It

[@Peter_S_]

Everyone should watch Hans' videos. They are excellent presentations with real world data and practical information.

[@HansRosenberg74]

Thanks 😀

[@pyrokinetikrlz]

i have a few years of experience in RF design (mostly 2.4GHz). I learned all this lessons about grounding with blood and tears while doing my designs. Great explanation by Hans!

[@HansRosenberg74]

Thanks a lot :-)

[@Electromechaniac]

I don't know exactly why but people too often ignore basic things/knowledge and this is a root cause of many problems. Great video, great explanation!

[@HansRosenberg74]

The real problem is that this is one of those. I didn't know I didn't know about that.....people are not even aware that a problem exists, that makes it hard to solve. There are many more of those that I'm going to address in the course I'm making.

[@dmitriyv4557]

many people are trying to pretend more "smart" and "complex" than they are indeed, so overcomplicating things without any good reason to do this. )

[@YM-qf8iz]

Excellent video...one of the best, and you have many good ones.

[@AdityaMehendale]

The 27:03 section is especially brilliant :) Thanks!

[@icestormfr]

Note: Digital signals, it's the rise/fall times that relates to the highest relevant frequency, not directly the clock frequency (stated ~9...10×f_clk). That rule has more to do with the relation, that you need a good rise/fall time to still have a good clk/digital signal. Using i/o with 0.1 ns tr/tf for 1 MHz clock/digital signals still has frequencies in the >1GHz region

[@robegatt]

BW around 0.35 / RT

[@eropis]

I love these videos. Technically i should know all of these from university but its always a nice refresher.

[@Jindraxx20]

An other great video ! Thank you Robert !

[@newdimension-q8n]

please more videos with Mr Eric xD he is exceptional with all my respect to the other guests

[@HansRosenberg74]

Noooooo, that hurts :-D. But yeah, he's good :-)

[@newdimension-q8n]

@@HansRosenberg74 sorry sir xD I'm still watching your video

[@philsaunders65]

The best video I have ever seen on this subject. Thank you very much for providing this, Robert.

[@HansRosenberg74]

Thanks!!!!!

[@hanelyp1]

A good ground network provides low impedance between ground points, and return for signal and power lines. The latter works best if the ground provides a close path parallel to the trace. An uninterrupted ground plane does this all very well. If the ground must have gaps, parallel a high frequency trace tends to work better than across forcing the return path to depart from the signal line. Consider the difference between a transmission line, close parallel conductors, and an antenna, conductors without a return path nearby.

[@newdimension-q8n]

the most worth subscribe i've ever did

[@Boltvin]

Great! Thanks so much.

[@f33net]

Fashionable arguments about reverse currents alarm me personally. This is true for coaxial lines, where the potential difference between the receiver and the transmitter interferes. In digital circuits, it is more important for us to ensure a strong ground so that it does not jump when transmitting a signal. I don't see any current under the track, there is a load on the resistance of the group, which makes it jump. The solid layer has the least resistance - this is true, but it is the lining of the capacitor on the ground for each line. The experience with the ground cut under the track at high frequencies speaks about the manifestation of wave properties at very high frequencies during the transition of medium density. It seems to me more understandable to discuss the topic from the point of view of the reliability of the ground, relative to which a useful signal is being pushed. If the ground is soft like a trampoline, it will be difficult for you to push off from it and at the same time there may be a resonance that will throw you into the sky. :) That is why it is more reliable to stand on solid ground of a solid layer than through the track. Imagine: you are standing on a metal structure podium (track to ground) at a concert, it is a little shaky (have resistanse) , of course. You are trying to shoot a video, and next to you an enthusiastic fan begins to joyfully jump and shake the supporting structure - the image is shaking. If you get down earth, the situation will improve. If you like the performer and you're having fun, then you can become a nuisance to boring people yourself. And you will advise the organizers to fill the podium with concrete. However, until a train goes near you. Everything in the World is relative. :)

[@theIpatix]

52:48 Honestly I think I just realized why ICs actually have separate pins. Because if you wouldn't have separate pins, you wouldn't be able to "unmix" the currents. So it'd be impossible to avoid crosstalk.

[@87Spectr]

Oh no, another great video

[@theIpatix]

56:00 Honestly, I don't understand why the designer of this circuit has not placed the capacitors very close to the audio input, without separated grounds. Wouldn't that also have fixed the unintentional demodulation? Also, I know for a fact that multiple ground planes with star topology are very commonly used in professional audio equipment across many brands. Though, the example in the video is probably a poor one. I'm not 100% sure why they do it, but I at least have one theory, which is to avoid crosstalk of low frequency signals (e.g. mic amp with 60 dB gain, yet still good SNR). And from all I understand, audio frequencies will still prefer the paths of lowest resistance. However, you will of course see capacitors connecting the ground planes so that RF loop area stays small.

[@petrsuchomel9639]

Even better that JLCPCB allows for free via in pad on 6 layer PCB - that really helps with grounding individual pins

[@HansRosenberg74]

Wow, seriously, is that true? I didn't know that. I've ordered my first pcb's there a few weeks ago and had a very close look at all their specs. It looks really good. But via in pad is a very big advantage. Especially when you're working with BGAs, in that case it is a life-saver. I need to design a new fpga module for my course so I will definitely look into that. Great tip! Thanks!

[@f33net]

In these discussions about the grounding layer, there is no thought of an increase in parasitic capacity to the ground. But it spoils the fronts and violates the integrity of the signal. It is not always possible to avoid long digital lines. The digital signal is not harmonic and some propagation features arise in it. What would you say about it?

[@Wholistic1]

Excellent video, thank you! I think it would have been nice to touch upon the subject of parasitic capacitance as well. The subject at matter is about 50 Ohm transmission lines, and you are comparing two different stack-ups with two different reference layers for the transmission line, at different depths, at minute 40:00. If nothing is done to adjust the trace width and clearance to coplanar GND plane (if any), then the line is not 50 Ohm in both cases. Then comes the question: how much impact does the same via on the transmission line still have, once you introduce the adequate parasitic capacitance (adjust the transmission line) to maintain the 50 Ohms? I'm afraid without having this in mind, you are more or less likely comparing apples with pears. In the end it does not only come down to the inductance... oversimplifying things this way is dangerous. There are so many effects and aspects to signal propagation.

[@Wholistic1]

Also, minute 42:52: unless backdrilling was performed, the via impedance is identical to before. It is the return loops' impedance that changes by exiting the via on a layer closer to the direct current's layer. But keep in mind this introduces one additional dangerous effect: you have now created a via stub of approximately 1.4mm length. Depending on your designated operating frequency, this can become comparable to it's wavelength and therefore dangerous.

[@Wholistic1]

Regarding the multiple GND planes. In my opinion, the idea behind it and what is actually of interest is not to have separate GND planes, but separate return paths. One personal preference and suggestion for an everyday layouter is to design your layout without placing a Plane Shape on top of the entire board (not until the very late stages of your design). Try to route the GND traces (return paths) the same way you route the signal traces (corresponding direct paths). This will help you get at least a roundabout understanding of where your return currents will flow throughout the board. In the end, you should of course add the GND plane shape and interconnect everything, reduce impedance and all other advantages that have been presented throughout this video. Don't forget about the golden rule: signals follow the paths of least impedance... do not mistake that if you have a signal on Layer 1, GND planes on layers 2 and 4, and if you route the return path on 4 it will actually flow on 4. Electromagnetic fields exist in the dielectric, so the return path will always be layer 2, the one sharing the dielectric with your signal's layer.

[@Wholistic1]

One good example is actually, on the same topic, when having Audio speaker outputs and Microphone inputs on the same board, even around the same area of the board. 99% of designers will create a separate GND plane for the Microphone functional block. This is like, let's say, an insurance taken in order to make sure you don't mix the two functional block's return paths. But it can be prone to huge, hard to depict errors. I recommend understanding return currents and how things flow around the board, stick to the same GND plane, and simply make sure you don't cross the return currents from the two different blocks.

[@fedimakni1200]

On boards more than 6L if L1 is sig, L2 is gnd and L3 is sig, can the return path of signals from L1 and L3 cause any interference or crosstalk or things like that?

[@HansRosenberg74]

I never tried this, but the impedance of a ground plane is soo low that it will probably be hard to detect. But I think this would be a great one to try at some point. I think there will be a frequency dependency, at very high frequencies, the skin depth is way less than the thickness of the copper, so there cannot be any crosstalk. I'm not sure at what frequency this would be, probably quite high.

[@fedimakni1200]

@HansRosenberg74 I asked because I am currently designing a motor control board (6L) and i have signals to control the H bridges and power on L1 but i also have analog signals to measure the voltage and current on L3, i was worried if the return path of power and signals on L1 (more noisy...) can interfere with the return path of the sensitive signals of L3 and cause errors in the measurements.

[@robegatt]

​​@@HansRosenberg74at 100 MHz skin depth is about 6 um... a 1 oz/ft2 is 35 um.. so plenty of distance in the copper... obviously no capacitive coupling because there is no dielectric and maybe some mutual inductance if traces are parallel?

[@HansRosenberg74]

@@fedimakni1200 ah, now it is clear. Well, you could make sure traces are not routed 'on top of each other' on each side of the ground plane. Also, motor voltage and currents probably don't need to be measured with an incredibly high accuracy, so you are probably ok using a good ground plane. If you want to do some calculations: Just calculate the resistance of a 1cm wide strip of ground plane (assuming the current stays in that width), figure out how long the piece of pcb is where those signals run on top of each other (you need that to calculate this resistance) and multiply that with the current running in your H-bridge. That will give you a voltage across the ground plane. Now see how big that is compared to the accuracy you want to measure voltages with. It is probably not going to be a problem, but this way you can make some educated estimations.

[@electgpl]

Good video Robert! With the addition of AI to voice dubbing, you will have more visits in Latam!

[@Aighthandle]

No.

[@tommihommi1]

KZbin is creating auto translated titles that you have to remove manually. They're based on the working title. Right now the video shows as "Hans Erdung v1" to me.

[@RobertFeranec]

I had not idea! Good catch. I am going to disable that. Thx

[@HansRosenberg74]

Hahaha, that is a very creative translation to German :-D

[@Mr1Spring]

Hans, there is a difference between Hz and GHz, Check the axis legends. Oh, I see it a bit better now, you put a B after the number. Why this confusement?

[@HansRosenberg74]

Sorry about that, this is material from my first video and I did not have the time or energy to figure this one out as I had 100 things to figure out. This python plot package uses B for Billion as standard. I've fixed it for all subsequent videos :-)

[@Mr1Spring]

@@HansRosenberg74 Hi Hans, Try QtiPlot, Veusz or SciDaVis - They are alternatives to Origin pro

[@zorabixun]

Mr Robert, most of our viewers are normal people, not quantum rocket scientists .. 😂😅 haha .. so we don't need going to molecular structures 😂 Previously your videos were interesting, now they are not practical, boring, so ... good bye ..