It would be really cool to have a video talking about laying out high power switch-mode power supplies. It would be interesting to know the EMC effects of switching amps in something like a boost converter. Also great video, I have learned so much from watching these things.
Keith Armstrong has written many great articles and they helped me a lot clarifying many EMC myth. Thx, Robert! So great you can have him on your channel, can't wait to see next talk ! Next time perhalps talks more about shield, chassis ground connection and ESD protections. Sometimes I can't compromise between ESD requirement and EMC requiremnt. Oh, as Keith mentioned in the video, low cost spectrum analyzer is really affordable now days! This year I bought Siglent's SVA1032X , somewhere around 3000 usd, spectrum anayzer with VNA upto 3GHz, I couldn't imagine this price five years ago. There is simply so much you can with it!
Thank you Robert for this video. Keith is one of the EMC gurus out there, and his website is full of material freely available after subscribing, which is free of charge. You can even enjoy a quiz about EMC, best design rules for EMC and troubleshooting. I work as an EMC test engineer since 2008, let us spread good electronics design practice around the great world of Electronics designers! By the way, if you want a spectrum analyzer go for either Rigol or Siglent, tracking generator is mandatory, and experiment with it. Don’t forget to buy near field probes… you can build your own or buy a complete set for around 300 $… cheers!
I had the honour of having two times a one-week seminar with Keith, I couldn't agree more, the man is a genius. Thumbs up for both of them for making this video
I just wanted to clarify on what a cavity is for people to better understand. A cavity in electromagnetism is a structure where the E fields reflect multiple times on the walls. In EM (electromagnetism), a perfect conductor is one that will reflect ALL the E field, so that no E field can penetrate inside the conductor. Now an actual metal plane in real life is not a perfect conductor, in low frequencies the metal will allow the E field to penetrate the whole metal structure meaning that the E field will not reflect, but at higher frequencies, due to skin effect, the E field will be reflected back in a metal plane (if the penetration length of the skin effect is much smaller than the thickness of the metal plane), now if you have multiple metal planes which meet this condition, then you have effectively formed a cavity because the E fields will reflect back and forth between those metal planes at higher frequencies. Now the whole talk is about the resonances inside those cavities because in physics cavities have a very interesting feature: they are resonant to only certain frequencies. And these resonant frequencies are given by the physical characteristics of the cavity (the main one is usually the length of the cavity - i.e. the spacing between conductors). This is why by changing the grounding bonds in the heatsink example the resonance frequency changes, you have now a different cavity geometry. The software he is showing effectively solves Maxwell equations for the geometry he introduced and is able to tell you which frequencies are resonant. This way you can know if your geometry is optimized at shifting the resonant frequency as far off as you can.
That first tool is amazing, being able to see what’s happening on all layers and then it selecting a capacitor with a resonance at exactly the same frequency. Yep, I can see why they could charge 100K for that
39:25 “you couldn’t shield the printer cables […] because a shield terminated at one end becomes an antenna”. I’ve read that it’s OK just to earth one side of a shield? (because of ground loops). He’s obviously the expert here, I’m just trying to understand the context of this.
I'm not an electrical engineer but it seems to me that because I've done things with Tesla coils and other RF stuff that it easier for me to understand the cavities and resonance of things. Very interesting video though, thanks for sharing this online.
Didn't we learn from some of your invited experts that we should never connect the shield of a cable or any other chassis piece to 0V GND? The shield is the continuation of the chassis - is what they said. Now, listening to Keith all chassis structure, shields etc would be connected to 0V GND. I know this is impossible for unbalanced signals but for balanced signals such as Ethernet which are non-earth referenced or any balanced/symmetrical audio lines, the shield has nothing to do with the signal and therefore should not be connected to ground. So that Ethernet housing should not be connected the ground. Can you please clarify?
Hi Robert, When you were asking the question about when you need to be concerned about preventing emissions, I think the answer is when the final product has to comply with regulations. The most common regulations for consumer products is FCC Part 15, class A and Class B. A is for industrial applications and B is for the home. These are performed at a system level, not usually at a board level. I assume that there are European equivalents of the FCC regulations. Are you familiar with them, and how do they compare to the FCC requirements?
@1:04:20: In the case where signals are single ended and share the same return reference potential [essentially two boards with each sharing the same return path potential], do we need to filter both the return path for the current also? When we do that, we are intentionally adding an impedance between two ground planes [one on each board]. Wouldn't that lead to cavity resonance if they are stacked? Or act like a dipole? What is the best practice when the return reference is common for two boards [for eg: control board and signal conditioning board]?
1:51:00 what if we throw components on pcb and provide metal shield box soldered at all 4 cornet at multiple point and put that box in enclosed casted aluminum enclosure. Will it work
6:20 Did the video ever reach a clear explanation of what the vertical, color and time axes are actually portraying? I guess Keith does say that basically the vertical axis duplicates the color axis. But what variable are those axes portraying? And what variable is the time axis cycling through? Frequency? (If so, from what to what?) Or is it cycling through some test digital sequence? (Again, if so what?). And why do these visualizations show flapping parts of the board as distinctly torn out of the overall plane? Are these actually isolated parts of the ground or power plane?
The simulation is looking for the frequency with the highest resonant peak, and them showing a full cycle at that frequency. The difference between vertical displacement and color is that color is determined with a desired target.
I should specifically ask that question. I guessed, it is showing electric field or voltage for specific frequency at specific place. So, for example based on the impedance chart you find the frequency of of the peaks (resonance) and then you have a look at the voltages at that peak frequency in each place of the power plane ... you will find where the voltage is the highest and that is the place where you need to put decoupling capacitor with resonance at that frequency. But I may be wrong - that is just my theory. I think, Keith mentioned these are the places with highest voltages.
@@theondono "showing full cycle": showing what variable? For a full cycle of what? In this example the frequency is 748 MHz, so are you saying you think that the flapping up and down corresponds to cycles of 748 MHz? That seems kind of pointless -- why not just plot amplitude, either as a heat map, or as a height map. But answering Robert's question, Keith says that the flapping represents sweeping a range of frequencies -- though I don't believe that answer. And as I mentioned, Keith says that the color axis and the vertical axis are showing the same variable (whatever it is) in two different forms. I do buy that, because it's commonplace to use color gradients on height maps.
@@RobertFeranec I think your description makes some sense. Perhaps voltage deviation from nominal (ie: zero, for ground). And it would be nice to have an idea of scale. But then what phenomenon corresponds to the cycling of the animation? Is is sweeping across a range of frequencies? Or is the animation showing data over some timeframe?
@@Graham_Wideman I’m speaking based on what I understood and some guesswork, but what makes sense to me is that the simulation is making a full frequency sweep from the injection point, and then showing the simulated voltage distribution for that input signal over a full cycle of the input. This means that the output can’t be completely represented just by amplitude, since the output can contain frequencies different from the input.
On the section on connecting mounting holes, could someone explain which ground are we bonding? Is it the zero volt reference planes on each pcb? Or the signal grounds if they are the same on multiple pcbs? Do we use the zero volt reference plane also as a safety ground or that is only put on one of the pcbs? Do we connect the safety ground to the zero volt references somewhere?
What is the nought-ball plane (called notebook plane by the captions)? Is that just his word for the ground plane, (nought-volt?) because he wants to separate ground-as-return from ground-as-safety? I'm assuming that fire-holes and firewall are really via holes, rather than a safety barrier of some sort, but if that is wrong, please say so.
I am sorry but there are two opposite recommendations on how to connect cable shields: 1,. "be the signal" => EPSI-08-60 The most common sources of EMC Failures 14:43 2.- This video => 1:47:40
I am beginning to think that the ever increasing importance of Electromagnetic Field Solvers for the design of ever faster digital PCBs is going to kill-off the PCB design consultancy market. EM Field Solvers are becoming vital, but they can cost anything from £25 to £100K, and this is completely outside the budget of the individual PCB designer. Only a big company with a large PCB team could justify such an expense. As the digital interface speeds go up, the EMC testing requirement rules will also be extended to ever higher frequencies, thus making the cost of the test gear needed to pre-compliance testing of your prototype at home, go up far beyond the ability of an individual consultant to buy. In the future, the individual PCB designer will be relegated to designing low-end boards with no newer high-speed interfaces (e.g. PCIe6, DDR5, USB4 Gen 3×2 etc), so their services will not be as highly valued as they once were. The big companies who can afford the tools, will pull-up the ladder behind them.
I think, that there will be new area of consulting sercvices - EM simulation experts - most companies won't have enough workload to find purchase of such software viable option. So there will rise market for simulation consulting, because specialized consultant should have workload required to make profit. Another issue is steep learning curve, and high entry level in terms of required knowlege to conduct and properly interpret results of simulation.
@@krzysztofkwiecinski148 Mmmn, the business model doesn't look great for becoming an independent EM simulation expert. If I sold my house and slept rough in order to raise the $100K to buy a license (for 1 year!), I would then have to charge consultancy of $275 per day and work 365 days a year just to break even. That doesn't even cover the time needed between buying the license and training myself to be an expert with using that software, during which time I couldn't charge anyone for being an expert with it. EM simulation software does not need to cost $100K, it simply isn't that expensive to develop. It is artificially priced to ensure that individuals can't possibly buy it, so the vendor only has to provide support to a hand-full of big companies, which saves the vendor money. The big established companies also like this becuase they get great support, and it is anti-competitive against smaller rival start-ups without being illegal.
@@MaxWattage there are plenty of people out there with this niche knowledge and interest who happens to have an extra 100k laying around in investments. Instead of selling your house, just take out a mortgage on it for 200k, start a consultancy and now you don’t have the one year time constraint but 30 year Don’t ever go into business without a good biz dev
This is where cracked/pirated software comes in. Learn it all in the cracked software and if you can afford it, buy a license at some point to pay the developers for their effort. I used to work at a company where they had an air gapped workstation (a workstation which is never connected to the internet) that was full of cracked/pirated software.
@@rjordans I was asked this question once in an interview and still can't figure out the technical reason. Could be just a rule of thumb from experimentation?
4.6k views but only 200 likes??? Has Gurgle's killing of visible downvote count killed people's desire to even bother with voting?????? Or have 200 engineers each rewatched this video 25 times???
![KSI - Dirty [Official Visualiser]](http://i.ytimg.com/vi/VoDh1h1dPUE/mqdefault.jpg)
It would be really cool to have a video talking about laying out high power switch-mode power supplies. It would be interesting to know the EMC effects of switching amps in something like a boost converter. Also great video, I have learned so much from watching these things.
Or motor variable frequency drive manufacturers.
Keith Armstrong has written many great articles and they helped me a lot clarifying many EMC myth.
Thx, Robert! So great you can have him on your channel, can't wait to see next talk !
Next time perhalps talks more about shield, chassis ground connection and ESD protections.
Sometimes I can't compromise between ESD requirement and EMC requiremnt.
Oh, as Keith mentioned in the video, low cost spectrum analyzer is really affordable now days!
This year I bought Siglent's SVA1032X , somewhere around 3000 usd, spectrum anayzer with VNA upto 3GHz,
I couldn't imagine this price five years ago. There is simply so much you can with it!
Thank you Robert for this video. Keith is one of the EMC gurus out there, and his website is full of material freely available after subscribing, which is free of charge. You can even enjoy a quiz about EMC, best design rules for EMC and troubleshooting. I work as an EMC test engineer since 2008, let us spread good electronics design practice around the great world of Electronics designers! By the way, if you want a spectrum analyzer go for either Rigol or Siglent, tracking generator is mandatory, and experiment with it. Don’t forget to buy near field probes… you can build your own or buy a complete set for around 300 $… cheers!
Keith has some of the best EMC books available. Very practical and easy to follow.
Keith Armstrong is one of my heroes. Thanks to you both for this video.
I had the honour of having two times a one-week seminar with Keith, I couldn't agree more, the man is a genius. Thumbs up for both of them for making this video
I worked for Keith in 1988 in an engineering design team. He was probably the best person I ever worked for.
I just wanted to clarify on what a cavity is for people to better understand. A cavity in electromagnetism is a structure where the E fields reflect multiple times on the walls. In EM (electromagnetism), a perfect conductor is one that will reflect ALL the E field, so that no E field can penetrate inside the conductor. Now an actual metal plane in real life is not a perfect conductor, in low frequencies the metal will allow the E field to penetrate the whole metal structure meaning that the E field will not reflect, but at higher frequencies, due to skin effect, the E field will be reflected back in a metal plane (if the penetration length of the skin effect is much smaller than the thickness of the metal plane), now if you have multiple metal planes which meet this condition, then you have effectively formed a cavity because the E fields will reflect back and forth between those metal planes at higher frequencies.
Now the whole talk is about the resonances inside those cavities because in physics cavities have a very interesting feature: they are resonant to only certain frequencies. And these resonant frequencies are given by the physical characteristics of the cavity (the main one is usually the length of the cavity - i.e. the spacing between conductors). This is why by changing the grounding bonds in the heatsink example the resonance frequency changes, you have now a different cavity geometry.
The software he is showing effectively solves Maxwell equations for the geometry he introduced and is able to tell you which frequencies are resonant. This way you can know if your geometry is optimized at shifting the resonant frequency as far off as you can.
Excellent call Robert! Very informative! We greatly appreciate these calls with the SME’s! 😊👍
At 1:46:32 you come with a question that help me a lot. I did it like the old days, I have been so confused - thank you Robert and Keith
That first tool is amazing, being able to see what’s happening on all layers and then it selecting a capacitor with a resonance at exactly the same frequency. Yep, I can see why they could charge 100K for that
His voice is very calming
39:25 “you couldn’t shield the printer cables […] because a shield terminated at one end becomes an antenna”. I’ve read that it’s OK just to earth one side of a shield? (because of ground loops). He’s obviously the expert here, I’m just trying to understand the context of this.
I'm not an electrical engineer but it seems to me that because I've done things with Tesla coils and other RF stuff that it easier for me to understand the cavities and resonance of things.
Very interesting video though, thanks for sharing this online.
PS, thanks Robert, excellent video. I've been working on this stuff for years and still learning!
Note to Ansoft: If you are going to charge 100K$ for your software, at least don’t use Comic Sans in your slides 😂
Didn't we learn from some of your invited experts that we should never connect the shield of a cable or any other chassis piece to 0V GND? The shield is the continuation of the chassis - is what they said. Now, listening to Keith all chassis structure, shields etc would be connected to 0V GND. I know this is impossible for unbalanced signals but for balanced signals such as Ethernet which are non-earth referenced or any balanced/symmetrical audio lines, the shield has nothing to do with the signal and therefore should not be connected to ground. So that Ethernet housing should not be connected the ground. Can you please clarify?
Should we have Chassis GND in the plastic enclosure? If yes how it will help in the EMC/EMI. Or should we have single GND plane in the board ?
When you are on a slide 5 of 319 and it is already good. This is a presentation I'd gladly sit through.
thank you! i watched this video all week!
Another top tier video! Thank you for sharing this with the world! 🙏
I would appreciate video on the topic of ferrite beads. It's still kind of mysterious component with unclear usage.
I look forward for a board interconnect and cable shielding video. Especially for more down-to-earth smaller projects.
Hi Robert, When you were asking the question about when you need to be concerned about preventing emissions, I think the answer is when the final product has to comply with regulations. The most common regulations for consumer products is FCC Part 15, class A and Class B. A is for industrial applications and B is for the home. These are performed at a system level, not usually at a board level. I assume that there are European equivalents of the FCC regulations. Are you familiar with them, and how do they compare to the FCC requirements?
@1:04:20: In the case where signals are single ended and share the same return reference potential [essentially two boards with each sharing the same return path potential], do we need to filter both the return path for the current also? When we do that, we are intentionally adding an impedance between two ground planes [one on each board]. Wouldn't that lead to cavity resonance if they are stacked? Or act like a dipole? What is the best practice when the return reference is common for two boards [for eg: control board and signal conditioning board]?
Awesome - thank you for bringing so many greats from the world of PCB to us! :)
Some very useful info in here, thanks for sharing!
Wow, lots of information in this video.
Thank you for this awesome video!
1:51:00 what if we throw components on pcb and provide metal shield box soldered at all 4 cornet at multiple point and put that box in enclosed casted aluminum enclosure. Will it work
6:20 Did the video ever reach a clear explanation of what the vertical, color and time axes are actually portraying? I guess Keith does say that basically the vertical axis duplicates the color axis. But what variable are those axes portraying? And what variable is the time axis cycling through? Frequency? (If so, from what to what?) Or is it cycling through some test digital sequence? (Again, if so what?). And why do these visualizations show flapping parts of the board as distinctly torn out of the overall plane? Are these actually isolated parts of the ground or power plane?
The simulation is looking for the frequency with the highest resonant peak, and them showing a full cycle at that frequency.
The difference between vertical displacement and color is that color is determined with a desired target.
I should specifically ask that question. I guessed, it is showing electric field or voltage for specific frequency at specific place. So, for example based on the impedance chart you find the frequency of of the peaks (resonance) and then you have a look at the voltages at that peak frequency in each place of the power plane ... you will find where the voltage is the highest and that is the place where you need to put decoupling capacitor with resonance at that frequency. But I may be wrong - that is just my theory. I think, Keith mentioned these are the places with highest voltages.
@@theondono "showing full cycle": showing what variable? For a full cycle of what? In this example the frequency is 748 MHz, so are you saying you think that the flapping up and down corresponds to cycles of 748 MHz? That seems kind of pointless -- why not just plot amplitude, either as a heat map, or as a height map. But answering Robert's question, Keith says that the flapping represents sweeping a range of frequencies -- though I don't believe that answer. And as I mentioned, Keith says that the color axis and the vertical axis are showing the same variable (whatever it is) in two different forms. I do buy that, because it's commonplace to use color gradients on height maps.
@@RobertFeranec I think your description makes some sense. Perhaps voltage deviation from nominal (ie: zero, for ground). And it would be nice to have an idea of scale. But then what phenomenon corresponds to the cycling of the animation? Is is sweeping across a range of frequencies? Or is the animation showing data over some timeframe?
@@Graham_Wideman I’m speaking based on what I understood and some guesswork, but what makes sense to me is that the simulation is making a full frequency sweep from the injection point, and then showing the simulated voltage distribution for that input signal over a full cycle of the input.
This means that the output can’t be completely represented just by amplitude, since the output can contain frequencies different from the input.
On the section on connecting mounting holes, could someone explain which ground are we bonding? Is it the zero volt reference planes on each pcb? Or the signal grounds if they are the same on multiple pcbs?
Do we use the zero volt reference plane also as a safety ground or that is only put on one of the pcbs?
Do we connect the safety ground to the zero volt references somewhere?
Brilliant Brilliant video!
It's the prophet of EMC 🙌
What is the nought-ball plane (called notebook plane by the captions)? Is that just his word for the ground plane, (nought-volt?) because he wants to separate ground-as-return from ground-as-safety? I'm assuming that fire-holes and firewall are really via holes, rather than a safety barrier of some sort, but if that is wrong, please say so.
I am sorry but there are two opposite recommendations on how to connect cable shields:
1,. "be the signal" => EPSI-08-60 The most common sources of EMC Failures 14:43
2.- This video => 1:47:40
Very interesing! Call this speaker again!
I am beginning to think that the ever increasing importance of Electromagnetic Field Solvers for the design of ever faster digital PCBs is going to kill-off the PCB design consultancy market.
EM Field Solvers are becoming vital, but they can cost anything from £25 to £100K, and this is completely outside the budget of the individual PCB designer. Only a big company with a large PCB team could justify such an expense.
As the digital interface speeds go up, the EMC testing requirement rules will also be extended to ever higher frequencies, thus making the cost of the test gear needed to pre-compliance testing of your prototype at home, go up far beyond the ability of an individual consultant to buy.
In the future, the individual PCB designer will be relegated to designing low-end boards with no newer high-speed interfaces (e.g. PCIe6, DDR5, USB4 Gen 3×2 etc), so their services will not be as highly valued as they once were.
The big companies who can afford the tools, will pull-up the ladder behind them.
Let's see where this goes....
I think, that there will be new area of consulting sercvices - EM simulation experts - most companies won't have enough workload to find purchase of such software viable option. So there will rise market for simulation consulting, because specialized consultant should have workload required to make profit. Another issue is steep learning curve, and high entry level in terms of required knowlege to conduct and properly interpret results of simulation.
@@krzysztofkwiecinski148 Mmmn, the business model doesn't look great for becoming an independent EM simulation expert.
If I sold my house and slept rough in order to raise the $100K to buy a license (for 1 year!), I would then have to charge consultancy of $275 per day and work 365 days a year just to break even. That doesn't even cover the time needed between buying the license and training myself to be an expert with using that software, during which time I couldn't charge anyone for being an expert with it.
EM simulation software does not need to cost $100K, it simply isn't that expensive to develop. It is artificially priced to ensure that individuals can't possibly buy it, so the vendor only has to provide support to a hand-full of big companies, which saves the vendor money. The big established companies also like this becuase they get great support, and it is anti-competitive against smaller rival start-ups without being illegal.
@@MaxWattage there are plenty of people out there with this niche knowledge and interest who happens to have an extra 100k laying around in investments. Instead of selling your house, just take out a mortgage on it for 200k, start a consultancy and now you don’t have the one year time constraint but 30 year
Don’t ever go into business without a good biz dev
This is where cracked/pirated software comes in. Learn it all in the cracked software and if you can afford it, buy a license at some point to pay the developers for their effort. I used to work at a company where they had an air gapped workstation (a workstation which is never connected to the internet) that was full of cracked/pirated software.
A very nice video like always! Good job Robert and Keith! Should be also very interesting to talk about fringing effect about power plane, for example
Even i want to have Spectrum analyzer. currently i use DSO for EMC ! Try using SDR
1:45:00 conductive tape will do the trick!
Thank you for sharing.
Multi point bonds spaced at
The factor of 10 sounds like an engineering margin, read it as "significantly less than" lambda in this case and 10 is a reasonable margin usually
@@rjordans I was asked this question once in an interview and still can't figure out the technical reason. Could be just a rule of thumb from experimentation?
It's a rule of thumb indeed, more is better but aim to have at least 10.
"It screamed with noise at the SkyTV downlink frequency" - so does SkyTV. How could anyone tell?
Printed Circuit Board Boards
4.6k views but only 200 likes??? Has Gurgle's killing of visible downvote count killed people's desire to even bother with voting?????? Or have 200 engineers each rewatched this video 25 times???
Probably, it's worth taking time to watch and I might need to come back and check some things again
Hanzhen harmonic drive gear , robot arm gear , over 30 years experience
Whenever he say spectrum analyser .. i feel poor from the inside! i have been trying my luck onkeysight giveaway...... still poor!
Firrrst!!, now to watch the video hehehe
As always, great video! BUT it took me several days to get time to se it all! Please shorter Robert, please! 😀 And again, thanks alot!
This is the shortened version. Cut down from 3.5h of video :)
@@RobertFeranec OMG! :-)