A superb video. This highlights for me how much of the PCB design-rule "received wisdom" passed between engineers is based on a series of engineering old-wives-tales and rules of thumb, rather than techniques based on hard-facts and physics-based design principles.
What a treasure. This fills in so many gaps in the knowledge that I have learned from others over the years. I wish I had run across this sooner. Thank you.
Thanks Altium, instead of just providing an awesome software you're putting lot of efforts to bring this kind of quality content to engineers. This is an awesome talk, lot of myth Busters revealed...thank you very much again Altium.
Wow, this has been really enlightening. I only started working with differential pairs this past year and to think I was so wrong on some of the things I thought about them. I'm really wondering if there's much benefit to having differential pairs for same-board transmission vs just a well routed single-ended signal. Considering he said about 85-90% of the coupling is between ground underneath, that still leaves 10% of the coupling for the trace to trace influences. Maybe that's enough to make a difference in some designs. Considering DDR is single-ended, it would seem that single-ended transmission for same-board transmission is just fine! I had used differential pairs for deserializing a data stream from off-board so that's why it was needed in my design, but I think I was starting to travel down the path of just thinking they were better for signal integrity inherently. Great talk.
I am playing with a differential dac and i can confirm that the differential to single ended signal has much less "noise", that was common mode before. it was on spagetti wired prototype and also on optimized pcb to observe.
My god, I try to explain this to PCB layout folks and I am like a bumbling fool compared to how eloquently and simply you put it. Everything you say I've been echoing for so long and I just get thought of as a controlling asshole, that it's 'what I want'. No, it's what the E-fields want to behave well...
Haha same here. At my work the bosses think any one who enjoys sitting in front of a PC can be a layout guy. Better a guy off the floor who has interest in PC than a real engineer that understands the physics. As long as the boards come out in time the lower pay to the layout guy the better. We have huge EMI problems, always spending weeks fixing products to pass, and we have horrible layouts, but nobody believes the board is a source of EMI! Todays high frequency challenges require a true understanding of electromagnetics. Layout is now a post graduate study, or the very least the person needs true interest in the physics involved and constantly willing to stay educated.
Digital sygnal has not only main clock frequency in its spectrum. If the IC able to have nanoseconds of rising and falling time even 1kHz signal will generate many harmonics in the spectrum.
@@creedo8301 He made a video that upset electrical engineers at first, about a 1 light-mile switch. So veritasium made a more in depth video about how power flows through the fields, in the dielectric, not the copper. Clips from this video were shown there.
The main benefit of DF pair is to get ride of "ground noise", You transfer your signal between two line that are suffering the same GND noise so at the receiver only the signal is extracted and Common noise is rejected. If you route the Pair's lines far apart "at each edge of the board" , how do you guarantee they are suffering the same "GND" noise if they are not and each line see a different noise level the noise rejection at the receiver will not be effective
He only routed them far like that to proove that the distance between them and the lenghts aren't as important as that guy through and the circuit would still work fine. He's not suggesting you actually do that as ofc, like you mention, there are other issues with it, but that was not the point of that particular demonstration as they were not discussing noise rejection and just discussing that the circuit it will still function.
Haha he's fun to listen to. I'd have to disagree with him on "all app notes are useless", e.g. app notes written by Jim Williams are worth their weight in gold.
it's probably one of the most underrated subjects for engineers. Indeed while this video is important, if he made a video on how to speak to others it'd be mandatory viewing for me at least.
Has anyone done any youtube tutorials that show how to implement these things? I’m not an expert, so I’d appreciate it if someone could walk me through it.
19:00 JUST in the plastic? Not in the surrounding air? Ok, he said "and in the air". Isn't that the field forms a region around the wire, and the plastic just happens to intersect that region? Which seems different than "in the plastic".
Ew, then if you need to switch layers with a differential pair on a 4 layer board, you are screwed; you need continuous GND on your layer 2 and 3 at least underneath the differential pair.. Say bye bye to solid power fill on that layer
some points were wise, but some were a brute oversimplification bordering on inaccuracy. I have a few years of analog, RF, mixed signal chip design in mutli-GHz range. take that example of routing the diff lines on opposite sides of the chip. well... what if you were to rout a bunch of vias running an aggressor signal orthogonal to the differential lines plane? Ampere and Maxwell would come back from the dead to smack you in the face in a form of bad differential interference 🙄 and then some peculiar definition of jitter 🤔
If those vias were close to the diff pair, then it would be fine because the noise would induce in both lines equally? Is this different from the example in 36:32 which implies asymetric noise injection anyway?
@@dzidmailif an alternating current passes through a wire loop, it induces some alternating current in that loop. think of the separated diff lines as a loop that closes through the loading diff stage input cascade. you would have induced a diff noise this way
Ohm's law isn't V = IR. That's a gross oversimplification, and the full formula is still just an approximation which makes many assumptions about materials and behavior. Real world is a bit different, especially with high frequencies and RF.
This talk should be mandatory for every EE. Thanks for sharing
A superb video.
This highlights for me how much of the PCB design-rule "received wisdom" passed between engineers is based on a series of engineering old-wives-tales and rules of thumb, rather than techniques based on hard-facts and physics-based design principles.
What a treasure. This fills in so many gaps in the knowledge that I have learned from others over the years. I wish I had run across this sooner. Thank you.
Rick Hartley is my love....i saw his every video, every discussion, every post....
Thanks Altium, instead of just providing an awesome software you're putting lot of efforts to bring this kind of quality content to engineers. This is an awesome talk, lot of myth Busters revealed...thank you very much again Altium.
Wow, this has been really enlightening. I only started working with differential pairs this past year and to think I was so wrong on some of the things I thought about them. I'm really wondering if there's much benefit to having differential pairs for same-board transmission vs just a well routed single-ended signal. Considering he said about 85-90% of the coupling is between ground underneath, that still leaves 10% of the coupling for the trace to trace influences. Maybe that's enough to make a difference in some designs. Considering DDR is single-ended, it would seem that single-ended transmission for same-board transmission is just fine! I had used differential pairs for deserializing a data stream from off-board so that's why it was needed in my design, but I think I was starting to travel down the path of just thinking they were better for signal integrity inherently. Great talk.
46:00 wow, just wow.
This was a fantastic talk, thanks for uploading.
5GHz can tolerate 5mm skew 🤯
@@JoeTaber you mean 1,5mm?
@@FluorescentApeIt's 3mm of skew. 1.5mm in each direction (+/-).
You just changed the way i look at differential pair. It was awesome 👍 Thanks Altium for inviting him...
Such a wonderful RUclips video that every EE should watch!
one of the best videos on this platform by far
As for peanut butter filler, creamy works much better than crunchy.
Good point!
Awesome presentation!
One of the first applications for differential pairs... Twisted pairs as used by the phone company, well over 100 years ago.
More than a few mic drop moments in this presentation. It's truly an eye-opening experience for anyone hearing these concepts for the first time.
I am playing with a differential dac and i can confirm that the differential to single ended signal has much less "noise", that was common mode before. it was on spagetti wired prototype and also on optimized pcb to observe.
Thank you, Rick and Altium. It helps me a lot.
My god, I try to explain this to PCB layout folks and I am like a bumbling fool compared to how eloquently and simply you put it. Everything you say I've been echoing for so long and I just get thought of as a controlling asshole, that it's 'what I want'. No, it's what the E-fields want to behave well...
Haha same here. At my work the bosses think any one who enjoys sitting in front of a PC can be a layout guy. Better a guy off the floor who has interest in PC than a real engineer that understands the physics. As long as the boards come out in time the lower pay to the layout guy the better. We have huge EMI problems, always spending weeks fixing products to pass, and we have horrible layouts, but nobody believes the board is a source of EMI! Todays high frequency challenges require a true understanding of electromagnetics. Layout is now a post graduate study, or the very least the person needs true interest in the physics involved and constantly willing to stay educated.
Why have only 42870 people watched this? Why only 1315 likes??? Why isn't it viral????!!!!!!!!
What a great explanation
this guy is freaking hilarious. "If there are any IC companies here, I'm sorry. Well I'm not actually sorry but I will pretend that I am." LOOOL
I have the original Vacuum tube op-amp Philbrick k2 Very informative talk.
I love this guy
Digital sygnal has not only main clock frequency in its spectrum. If the IC able to have nanoseconds of rising and falling time even 1kHz signal will generate many harmonics in the spectrum.
Veritasium dudes mark your attendance here.
That would include me.
Me too
did he recommend the video or smth ?
@@creedo8301 He made a video that upset electrical engineers at first, about a 1 light-mile switch. So veritasium made a more in depth video about how power flows through the fields, in the dielectric, not the copper. Clips from this video were shown there.
@@P4nDA_pls i've seen both of them but at that time i wasn't into PCB Design, damn
it's all coming together
Extremely useful
The main benefit of DF pair is to get ride of "ground noise", You transfer your signal between two line that are suffering the same GND noise
so at the receiver only the signal is extracted and Common noise is rejected.
If you route the Pair's lines far apart "at each edge of the board" , how do you guarantee they are suffering the same "GND" noise
if they are not and each line see a different noise level the noise rejection at the receiver will not be effective
He only routed them far like that to proove that the distance between them and the lenghts aren't as important as that guy through and the circuit would still work fine. He's not suggesting you actually do that as ofc, like you mention, there are other issues with it, but that was not the point of that particular demonstration as they were not discussing noise rejection and just discussing that the circuit it will still function.
18:00
The thumbnail looks like an aged Harry Potter. Fitting for the subject.
Close your eyes: Mike Ehrmantraut is speaking to you
Thank you!
You're welcome!
3:38 120 is a lot? @ $165 ea, i can understand why you didn't buy more.
Great talk!
Pure gold! Thanks.
Great content
Pure Gold knowledge.
Haha he's fun to listen to. I'd have to disagree with him on "all app notes are useless", e.g. app notes written by Jim Williams are worth their weight in gold.
Yeah, and layout for switching power supplies called out in the datasheets are often critical. You can deviate some, but you'll cause problems.
Not taking away anything from his engineering knowledge he sounds like he was developing himself more in soft skills (to a great effect).
it's probably one of the most underrated subjects for engineers. Indeed while this video is important, if he made a video on how to speak to others it'd be mandatory viewing for me at least.
Omg I'm fieldpilled.
3:14 Non-grounded connector at both end? Or receiver only?
This is the type of guy you want working on your design.
Has anyone done any youtube tutorials that show how to implement these things? I’m not an expert, so I’d appreciate it if someone could walk me through it.
great sharing!
I'm emailing this to my Professor's haha
great video
53:15 Is it correct to say that the misalignment causes current which must be "absorbed"? And the two-R plus cap serves to absorb that current?
It gives the generated, unwanted current a path to 0V, so it doesn't travel in the signal lines as far as i understand it.
At 35:00, how exactly skin-effect is causing this SI issue??? I can't seem to wrap my head around it.... Can anyone kindly elaborate???
Hell yeah! Its just sweet truth.
Wow just wow!!!
Спасибо Большое
23:21 But just if digital signals, right? Analog, eg balanced mic cable, cares about whole waveform.
Depends on your rise time, but rise time for an analog can be much lower.
I’m ready to challenge my boss!
Just kidding, I’m still dumb.
Next time someone asks me why I don't use autoroute, I will point him to this video
19:00 JUST in the plastic? Not in the surrounding air? Ok, he said "and in the air". Isn't that the field forms a region around the wire, and the plastic just happens to intersect that region? Which seems different than "in the plastic".
Yes, it's assumed you already understand that.
@@VeritasEtAequitas Unclear comment.
difficult concepts explained easily. that's great. It gives the impression to be intelligent :-) thanks a lot for that.
Is this Nina Hartley’s husband ? Wow what a power couple 😂
53:03
🔥🔥🔥🔥🔥🔥🔥
👍
36:08 How high?
36:30 did you stop watching 😂
I was asking about frequency, not distance.
Ew, then if you need to switch layers with a differential pair on a 4 layer board, you are screwed; you need continuous GND on your layer 2 and 3 at least underneath the differential pair.. Say bye bye to solid power fill on that layer
I believe he explains this situation here:
ruclips.net/user/liveySuUZEjARPY?feature=share
Example starts at 49:57
Bookmark: lecture on diff pairs
um now I need to go redo.... every board I ever made. jesus
some points were wise, but some were a brute oversimplification bordering on inaccuracy. I have a few years of analog, RF, mixed signal chip design in mutli-GHz range. take that example of routing the diff lines on opposite sides of the chip. well... what if you were to rout a bunch of vias running an aggressor signal orthogonal to the differential lines plane? Ampere and Maxwell would come back from the dead to smack you in the face in a form of bad differential interference 🙄 and then some peculiar definition of jitter 🤔
If those vias were close to the diff pair, then it would be fine because the noise would induce in both lines equally? Is this different from the example in 36:32 which implies asymetric noise injection anyway?
@@dzidmailif an alternating current passes through a wire loop, it induces some alternating current in that loop. think of the separated diff lines as a loop that closes through the loading diff stage input cascade. you would have induced a diff noise this way
A serious tractor was serious problems
20:00 "Forget voltage and current" seem extreme. Throw out Ohm's Law?
Yes
Yeah, it's all little antennas at high frequencies.
Ohm's law isn't V = IR. That's a gross oversimplification, and the full formula is still just an approximation which makes many assumptions about materials and behavior. Real world is a bit different, especially with high frequencies and RF.
'i own over 120 books'
hahahah
Great presentation!