Electronics tutorial - Crosstalk

#84 In this video I look at a common issue in modern high density digital circuits - that is of course crosstalk. I look at why this phenomenon is a problem, how it occurs and then what can be done about it.
Ti documentation:
www.ti.com/lit/sg/sdyu001ab/s...
Special Thanks to all my supporters on Patreon! Especially @afiskon !
If you want to support the creation of more and better videos please consider checking out: / feszelectronics
If you liked this video be sure to check out my other videos and you can also subscribe to be up to date with all the new ones!

[@KamleshGadhvana]

finaly i've seen crosstalk in practical way, thank you

[@FesZElectronics]

I'm happy you enjoyed it!

[@lad4694]

I love your videos so much! Signing up for your Patreon as I can now somewhat afford it!

[@urmok6iv]

Your teaching methods are really effective. Thank you so much for these videos!

[@FesZElectronics]

Glad you like them! Thank you!

[@TheMechatronicEngineer]

Really interesting!

[@AndrewKiethBoggs]

More videos like these! Love the simple lab implementation. Your quality of videos has been consistently amazing.

[@Juanpavz]

Amazing explanation and very useful for students

[@user-lo2xn9uh8c]

So simple and so informative at the same time. Thank you.

[@ango5421]

I learn a lot from your videos combining theory and real life examples. Thank you!

[@FesZElectronics]

Glad to hear it! I'm happy you are enjoying the videos!

[@MIsam-fv9kb]

Very informative as usual

[@nbuzkila]

Excellent video!

[@fixfaxerify]

You didn't consider shortening traces at all? Thumbs up for explaining the capacitive situation with groundplane as a capacitive divider, hadn't ocurred to me.

[@gingsSon]

Great video as always 👍

[@nikosant03]

Amazing video!! Very good explanation!! I didn't even know the benefiits of a gurd line

[@FesZElectronics]

Glad it was helpfull! The guard line is not always practical, but in certain cases it will be necessary...

[@p_mouse8676]

Really good explanation! 👍🏻👍🏻 Crosstalk is actually also really a thing with analog signals. It can even be part of your SNR, although nowadays crosstalk numbers are mostly more than good enough. A very difficult situation is when having high current switching signals (can also be digital) combined with analog signals. Humm and other introduced noise is essentially also a form of crosstalk. This was really an issue with tube amplifiers, especially because the current through the heater was quite significant.

[@FesZElectronics]

Regarding noise in amplifiers, I noticed quite an interesting phenomenon in a transistor amplifier that was supplied from the mains trough a transformer - with an E+I type of core there was a very noticeable buzz but with a toroid transformer it all went away (no other filtration or change was made) - basically its not just switching supplies that can cause problems, you can get pretty bad results caused by the magnetic fields coming from normal 50(60) Hz transformers also.

[@p_mouse8676]

@@FesZElectronics yes. Rotating a transformer will sometimes also help

[@SurvivalSquirrel]

Youre doing a better job at teaching, than all the other experts out there! Just some Details are missing: What is the thickness of the dielectric?

[@hanangonen8732]

Nice presentation ,keep it up . Maybe you will do another video about op amp basics ???

[@supaderrick3502]

Great Video.

[@reeraslan]

Hello, it was really nice topic you are explanining. Love your the way you show the complicated stuff. So I have a question. We see the voltage spikes on the second trace say shown at 6:47. What I dont understand from this scope pic is that current going through is not bidirectional. Current only flow to one way right. But why we see that coupled signal is passing both direction as it was induced by the ac signal. Thanks.

[@FesZElectronics]

Hello Eren! The spike is not caused by the direction in which the current flows but rather the "direction" in which the voltage varies - when voltage is increasing, it also increases the voltage on the "victim" line, which afterwards recovers to zero; then when the voltage decreases, it also decreases the voltage on the other line turning it negative. On the other hand, regarding current direction, the driver is of push-pull type (the 2 transistor configuration) - this can both source (send from driver to load) as well as sink (from load to driver) current. The trace has its own capacity and inductance so it does store a bit of energy which needs to be sink'ed so current does flow both ways.

[@wilmdrdo1228]

Please do more of this simulation and real life applications

[@FesZElectronics]

That's the plan! Thanks for watching!

[@realchristopher4334]

I thought this man is a fan, supporter, supplier, distributor, or being sponsored by Texas Instruments. But I realise that I am rather similar to him! 🤣 The 'problem' is Texas Instruments sells their products at reasonable prices, and they have friendly technical support agents (I do not know those are engineers or not because most enquiries are not answered immediately, even very very very very basic ones. Very strange the support is called 'technical'. Most enquiries will end up in their E2E forum. ) . Above all, Texas Instruments people and notes do not keep telling people USE TI, USE TI, USE TI, USE TI, USE TI. Analog Devices is the direct opposite of Texas Instruments. Their engineers ought to stick to their profession and NOT marketing. And, Analog Devices prices are exorbitant.

[@thezodiace7399]

keep them coming good man, One question: why does high impedance source implies an E-field coupling and low impedance implies H-field coupling?

[@FesZElectronics]

Hello! Well an E-field is generated by voltage applied between 2 terminals and H-field by magnetic fields, which are created by current going trough conductors; If we have a given amount of power (say 1W); the highest voltage (that will generate E-fields) will be obtained by applying the power onto a high impedance load - this will be accomplished with a low current ; on the other hand to get magnetic fields, so high current we need a low impedance, that will cause low voltage. Let me know if this makes sence

[@thezodiace7399]

@@FesZElectronics Thank you good man.

[@sajidnaseeb5386]

Hey, I have been watching your videos for a long time. Your work is legit and far more precise than other youtube channels. I am very thankful for your videos. Talking about precision and accuracy, can you tell me how to accurately model the capacitance value that represents the coupling between a DUT (Device under test) and the ground plane when carrying Conducted Emissions Test for compliance. I really need a legit answer and I was hoping if could help me with that? Thank you once again.

[@FesZElectronics]

Hello Sajid! Well there is no easy answer to this. Depending on how the DUT is built, you need to see what trace is closest to the ground plane below - if its for example a metalic rear cover, or some other structure. This will be the bit that is capacitively coupled to the ground plane (a few pF). Then you need to see how this structure is connected to the rest of the circuit, and of course to your main noise generators. The common mode noise will be coupling trough the LISN into the ground plane, and back to the unit. So you need to model the noise generator the path of the noise - any input filters, the LISN and then the few pF back to the unit. On the other hand if the noise is differential, than the ground plane coupling shouldn't matter that much since the noise goes out one supply line and back the other. May I ask what sort of device you are working on?

[@sajidnaseeb5386]

@@FesZElectronics Basically, its the E.field parasitic coupling between the Buck converter switching node and ground plane as well as the between the metallic cover of an Iphone and Ground plane. In that experiment, they considered the capacitive coupling between the Buck converter switching node and ground plane = 20pF and between the metallic cover of an Iphone and Ground plane = 100pF. So, I am going to share a picture of the whole experiment, I took it from that webinar. It will help get the idea. drive.google.com/file/d/1O-Sw54TMwXtER8zCVgSJnK9cS_4SMYaQ/view?usp=sharing

[@FesZElectronics]

Well there is quite a long cable between the Buck and Phone (seems to be ~1m) so that adds a bit of inductance - that should be modeled. You can of course model the cables as transmission lines - to have distributed capacitance and inductance. I think the most important part of the simulation needs to be the noise generator though - the wave forms and the spectrum of the noise in the switching node need to be accurately reproduced to get the same results in real life and in the simulator. So to summarize, you need the noise source (the buck) the cables modeled as inductors or transmission lines, the LISN on one end and the phone just as a capacitor from circuit to ground plane on the other end, and the buck parasitic capacitance to the ground plane. This should offer a fairly precise model. Do you know what model of spectrum analyzer is used?

[@sajidnaseeb5386]

Yes, the spectrum analyser is ROHDE & SCHWARZ FPC1500. I got the idea, but the only thing I am confused about is how to calculate the values for those parasitics? How I determine the values for inductors and capacitors, used to model the cables and parasitic coupling, respectively?

[@FesZElectronics]

Well you can find quite a lot of online wire inductance calculators - simply plug in the wire length and thickness and you will get a good approximation of the inductance. For capacitance, its about surface are over distance - you can get a decent approximation by using plane parallel capacitor calculator for that. And if you have the measurement results - the real life ones, you can simply fine tune the values until the simulator shows a good match.

[@breedj1]

Excellent video!