FEDEVEL
Platform forum

nachodizz990 , 07-15-2016, 09:22 AM

Hi all, i´m reviewing some Texas Instruments board layout designs such as -> TI_AM572x_PM_RevA3.brd <- with the cadence aleegro free viewer and i sensed that they use a lot global shared shielding where all connectors (excepting audio) are touching the shield border instead of shielding each individual connector.




What are the pros and cons of this technique ?

Thank you all in advance

robertferanec , 07-15-2016, 12:14 PM

I used to do this ring, but I do not do it anymore. The advantage is, that you can connect the ring to your chassis / enclosure or you can connect it to the board GND. For example, depends on EMC or ESD test results you can fit resistors, capacitor or nothing between the ring and your board GND.

For the boards which we normally design (CPU boards), the best test results were always when the ring was connected directly to the board GND through 0R resistors. So I stopped using it as it requires quite a lot of space. The picture which you attached has unusually small gap between the ring and the board signals, in your design you may want to keep the gap as big as possible.

nachodizz990 , 07-15-2016, 12:27 PM

Ok i agree, suppose the best approach for EMI certification its to make the ring and place dummy compomponent footprints for testing.
In automotive ECU (Engine control) there is a huge guard ring (Bosch EDC series for example) but there are no AC mains...
Somewhere i read that it depends on the number of the io cables attached to the board for example single usb or a lot of cables including industrial profibus etc.. with noisy enviorment.
I read that It depends on how it´s connected on the other side...
But i can´t find solid guidelines
There are a lot off contradictory terms on the shielding field between manufacturers so i will not place the guard ring this time because it´s a whim design

robertferanec , 07-19-2016, 05:28 PM

I used to read EMC recommendation articles, but there are so many and they tell you nothing - or to be precise - they will tell you the things which may be completely opposite of what you just read in the other article. And, do not be surprised, when next year you will read, that everything what you have been doing until now is completely wrong

Therefore, for me are much more important my past projects and I do the things the way I know that it has never failed in past (even if some articles say it's wrong). The biggest problem around EMC is, that there is no one universal or general solution. What works in one case, may not wok in other (or it can even make it worse). That is the reason why I do not teach about EMC, because it depends not only on the board but the whole system where the board is connected.

nachodizz990 , 07-19-2016, 05:40 PM

I hate the "Black Magic Topics" and contradictory themes too...
I agree ...
Perhaps with a 12 GHZ DSO all the black magic will dissapear ...

alazareff , 08-09-2016, 10:20 AM

Hi,

Very interesting topic, i will try to add some informations based on my EMC experience :
In EMC, there is 2 schools for the ground : global and local.
From my experience, and EMC expert i met in automotive, the global is preferred today.
This means, one or more GND plane for shielding, but also to have the lowest impedance for the return current. And you should add GND stiching via in order to create a grid with a minimum distance of 1 to 3 cm depending of the EMC regulation. This guard ring will also help for ESD

Of course, there is exception like isolated power supplies or Ethernet. But if the Ethernet connection length is short (below 10 meters) you can directly connect it to the ground plane of you board.
If it is a switch, then you can isolated the ground, and keep a path with a capacitor for RF return current.


Now here is what i do when i design boards :

• I list all the frequencies and their harmonics (very helpfull during the EMC qualification)
• -add ferrite (or at least 0R in case of) for every IC power supplies, between the decoupling capacitor and the power supply => i have seen too much IC with weak power input filters. And it is difficult to find and solve during EMC qualification.

• -high speed tracks in internal layers => Basically, follow the 1/10 wavelength rules for external layers. In automotive it can go down to 1/30
• -Reduce the high current loop of switching power supplies
• -add Pi filter on the power supply input + ferrite => this is for the conducted and radiated emission
• -for some specific signals which goes to connector, i also add 0R in series with a 100pF capacitor
• -Guard ring and GND stiching via in order to have one homogeneous 0V reference.

Now, if you have a metal enclosure, i only add filter and ferrite on the connector signals and the related components.

I found one book i recommend, it is "Electromagnetic Compatibility Engineering" by Henry Ott. It helps me a lot to understand EMC.
The second recommandation is to pratice
As Robert says, EMC depends on the whole system, and what works on a product may not work on another. So the best is to practive to master the EMC related to your product and its environment.

Alex

nachodizz990 , 08-09-2016, 10:28 AM

Nice post @alazareff the hard thing it´s when the schematic designer plus the pcb designer + the EMC expertise i´s the same person, its hard to be an expert in all !!

Yor idea of placing dummy parts on posible problematic points its the best way to proceed with.

I like so much digital hardware design and route and about emi problems, i think i will purchase the book you pointed and study a little bit abouth it

nachodizz990 , 08-09-2016, 10:35 AM

Let me review the schematic of TI and i will post how they connected both the gnd and earth

robertferanec , 08-10-2016, 05:22 AM

@alazareff great post! Thank you.

JohnsonMiller , 09-03-2016, 09:07 AM

Thank you guys for a good discussion and knowledge sharing, one question which I cannot answer, The EMC/EMI is circuit theory and mathematics like other electrical eng fields, why you guys insist that there is no clean solution for EMC/EMI? Is it difficulty for modeling? or?

robertferanec , 09-03-2016, 06:29 PM

If you have the right software, if you play with simulation for enough time, you understand what you are doing and you understand the results you may be able to model it. However, I do not think it is so simple. I would say, even the board itself is quite a complex system + when you start placing it into an enclosure and you connect all the cables it may get really complicated. And if you would like to fully understand what's going on, you may not have enough time to be a hardware design engineer and you may become EMC expert instead.

nachodizz990 , 09-05-2016, 10:16 PM

In my case, i´m hardware - firmware and HDL designer, it´s impossible to be a master on EMI, FIRMWARE, HARDWARE, DFM, DFT, etc

Small companies don´t have acces to expensive instrumentation or can´t pay a salary every month to a EMC specialist that only knows about EMC.

Experienced and good engineers (like Robert) dont need to be "EMC specialists" to succesfully pass EMC or to quickly fix errors

Accurately simulate a pcb is a waste of time (if not impossible) , the best way it´s to design the best as you can remembering the welldone things you learned by the way and make REAL tests, then if something goes bad, you can have the ability to quickly fix the error

As @robertferanec said, it´s impossible to simulate the behaviour a board will have, how electrons and holes will decide to flow through decoupling-bypassing and bulk caps. In a complex systems, for example DDR3, sometimes the SoC 1V5 decoupling module and the Memory IC decouplig module will interact and will share capacitors depending on what is the easiest path to ground on that moment for the current xD

In my oppinion a good engineer is a person who is able to dont reinvent the wheel, and quicly fix problems without wasting time on innecesary mathematics (note that i know and respect mathematics and physics) but i ever like to follow practical approaches

The best way to play with emi its to have a good PDN (power supply network) and also read some integrity books to reduce the possibilitties of making mistakes.

An experienced engineer told me that if the power it´s ok, the board wil be ok

Some usefull notes (understanding this app notes will help a lot to understan manufacturer´s reference designs)

https://www.google.es/url?sa=t&rct=j...,d.d24&cad=rja

THIS Three ARE SO GOOD look also the references :

https://www.google.es/url?sa=t&rct=j...,d.d24&cad=rja

https://www.google.es/url?sa=t&rct=j...,d.d24&cad=rja

https://www.google.es/url?sa=t&rct=j...,d.d24&cad=rja

Here´s a very good pHd about powering (it´s based on in-package decoupling ) but xplains so well the main concerns about power quality an how SoCs behave

The Author is Mikhail Popovich

https://www.google.es/url?sa=t&rct=j...dD3JXA&cad=rja

robertferanec , 09-06-2016, 07:27 AM

I agree with @nachodizz990. It is not necessary simulate everything and understand all the mathematics around EMC. Simulation may be very time consuming and even if you get some results it doesn't mean they reflect the reality (I have seen boards passing simulations and failing in the real world).