Bogatin’s so clear in his thinking and presentation. I think the key is how clearly he separates concepts and breaks problems down to the most basic principles. He’s a pure pleasure to listen to 👍😁
I took a Signal and Power integrity course at my University that followed Bogatin's -Simplified book. I learned so much and everything was so accessable and walked through concepts that were possibly missed from earlier coursework. Fantastic speaker, fantastic teacher, fantastic writer. Thank you for helping us access this fountain of knowledge
Another maybe obvious point of view is that if we are driving some receiver, those have some limits on common mode voltage (voltage of shorted diff line to IC ground) they can handle. In that case we still need DC ground to be at least routed somehow, even if we have ideal diff pair, otherwise we have two floating inputs with some differential voltage between them. Eg. for CAN bus between to battery powered devices wee need ground (or luck). If we are not driving input transistors in some IC but simple coil as in ethernet transformer than we can "delete" the ground as utp cable.
24:04 there are two factors pointed out: 1) DC resistance 2) tolerance but how about actual resistance of the signal against EMI? Is it not affected by pair lines being farther apart?
Excellent presentation on differential pairs! One question about return currents: In your example with a twisted-pair cable carrying differential signals, the fields cancel out within the cable. But what happens when the differential pair transitions to a PCB with a return plane nearby? Eric Bogatin says that most of the return current will be coupled to the plane. If the PCB return planes of the two connected devices aren't directly tied together, wouldn't the return current struggle to complete its path? In that case, wouldn't it be better to connect the planes to ensure controlled coupling and minimize EMI issues? Thanks.
I still missed one thing. In the case where transformer drives diff line (e.g. UTP Ethernet cable) - where will the single ended signal come from? Isn't the function of the transformer to eleminate the SE signals?
Imagine there is a length difference between the two traces from the transformer to the Ethernet jack so the two sides of the pair are not perfectly out of phase. The net (common mode) signal at any given point in the pair is now an AC wave at your signal frequency. Your board, traces, connector, etc. have capacitance to ground for the return path. At least, that is my understanding 😅
I see these 'differential pair' schematics in my audio realm explorations all the time. But, some have it set with a 'Main' (POS) signal, adding an Inverter to it for the Inverse (NEG) signal.. Doesn't that introduce a temporal offset between the two transmission lines?
The peak current distribution inthe reference plane (i.e. J(0) just underneath the trace) should be same for all h's (i.e. J0/PI) acording to the formula given in slide 4, unless J0 is also a function of d. What am i missing here?
What does Eric say at 3:06? All differential pairs are two single-ended transmission lines? Edit: ok yes I was right as you can see later on the slides but it's really hard to tell as he seems completely swallow the word "ended".
Odd mode impedance is by definition half the differential impedance. Even mode impedance is the common mode impedance. Removing the ground makes the even mode impedance infinite.
So the secondary of a transformer that is not center taped to the ground is still a two single ended pair? If you feel this doesn't sound right and smells bad, you acre correct, because it's a terrible generalization.
Another question. The practical case - we have a PC (a computer), and an Internet router/switch. They are connected by a 50 meters unshielded Ethernet cable (U/UTP). Will the differential currents cancel out between every two wires in every four pairs and common currents will find their way in the air as a radio wave between a PC and a router/switch? May we then treat that cable as a 50 meters radiating antenna? And what if we have a shielded ethernet cable (F/UTP) and both ends of that shield are connected to the ground in both PC and a router, so we have the metal, not air, return current path. Will the common currents of four pairs go back in that shield? And what if we will connect that shield just to one device (a PC i.e.)? Will that shield act as 50m antenna? Will it radiate to an air by entire 50 meters shield? Or it will radiate just by that few centimeters between the unconnected shield and a ground in the other unconnected device - in this case the router/switch?
I think this is because, the differential signal will have a clean return path (the difference ie., D+ - (D-)), but the common single ended (avg.) would actually not have a return path and would oscillate as antenna (without reference) and thus causing EMI issue.. Anyone can feel free to comment and carry this discussion.. Thanks!!
Someone please help me understand what the return current does at the interface where the cable enters/exits the PCB? If we have RS422/485 differential signals on Shielded Twisted Pair, connecting to a PCB with no direct connection of grounds between the two systems, what happens to return current at that interface? It can follow the ground plane in PCB up to the connector, and the differential twisted pair return current is on the other wire in the H/L pair. But at that interface where the return current in the PCB has no where to go, is that similar to what Eric was saying it will give a convoluted return path but they'll be equal/overlap and cancel out?
If you are not sending a ground along with the twisted pair (which you're probably not doing) then the only way to do it right is to use a transformer as a balun. The centertap of the receiving side should be connected to the ground. This way, you have 2 single-ended signals again which are both referenced to ground. If you open up an ethernet switch, you'll see a large black block right behind the tcpip connectors that contains these transformers. Often they also include extra CM transformers to make sure CM interference doesn't get onto the cable.
Regarding the slide number 5, what if we put the N signal instead of the ground plane right below the P signal? Will the return currents cancel out almost completely as they are between P and N line below both of them right on the center?
While mr. Feranec is an amazing layouter and a great tutor in this area, it is embarrassing to see how little knowledge of electronics he possesses. Not mentioning his very poor English. I find having him next a giant (IMO) like prof. Bogatin is just unfortunate.
Bogatin’s so clear in his thinking and presentation. I think the key is how clearly he separates concepts and breaks problems down to the most basic principles. He’s a pure pleasure to listen to 👍😁
I took a Signal and Power integrity course at my University that followed Bogatin's -Simplified book. I learned so much and everything was so accessable and walked through concepts that were possibly missed from earlier coursework. Fantastic speaker, fantastic teacher, fantastic writer. Thank you for helping us access this fountain of knowledge
Your videos with Eric are always extremely useful.
I agree, I learn a lot
Amazing thank you both for educating us for free
I could listen to the professor speak all day he’s just wonderful
Videos like this are the reason I'm a Patreon supporter. Many thanks for doing this.
Thank you very much
Awesome content! Can you cover the case of less than 1MHz for us mixed signal/audio folks?
“A wire sitting in space”. Lol. Also known as an antenna?
Thank you Eric and Robert!
You guys are awesome. I can learn abour the most important topics in a circuit that didnt even mentioned in schools ans university's
Another maybe obvious point of view is that if we are driving some receiver, those have some limits on common mode voltage (voltage of shorted diff line to IC ground) they can handle. In that case we still need DC ground to be at least routed somehow, even if we have ideal diff pair, otherwise we have two floating inputs with some differential voltage between them.
Eg. for CAN bus between to battery powered devices wee need ground (or luck).
If we are not driving input transistors in some IC but simple coil as in ethernet transformer than we can "delete" the ground as utp cable.
Great presentation
24:04 there are two factors pointed out:
1) DC resistance
2) tolerance
but how about actual resistance of the signal against EMI? Is it not affected by pair lines being farther apart?
Excellent presentation on differential pairs!
One question about return currents: In your example with a twisted-pair cable carrying differential signals, the fields cancel out within the cable. But what happens when the differential pair transitions to a PCB with a return plane nearby? Eric Bogatin says that most of the return current will be coupled to the plane.
If the PCB return planes of the two connected devices aren't directly tied together, wouldn't the return current struggle to complete its path? In that case, wouldn't it be better to connect the planes to ensure controlled coupling and minimize EMI issues?
Thanks.
I still missed one thing. In the case where transformer drives diff line (e.g. UTP Ethernet cable) - where will the single ended signal come from? Isn't the function of the transformer to eleminate the SE signals?
Imagine there is a length difference between the two traces from the transformer to the Ethernet jack so the two sides of the pair are not perfectly out of phase. The net (common mode) signal at any given point in the pair is now an AC wave at your signal frequency. Your board, traces, connector, etc. have capacitance to ground for the return path. At least, that is my understanding 😅
@@jessicav2031 for my simple mind length difference is a good enough explanation. It makes sense to me. Thanks.
I see these 'differential pair' schematics in my audio realm explorations all the time.
But, some have it set with a 'Main' (POS) signal, adding an Inverter to it for the Inverse (NEG) signal..
Doesn't that introduce a temporal offset between the two transmission lines?
Can you please explain what happens to voltage and current in long wires? If I apply 24V , how long can the wire be? and what happens for signals
The peak current distribution inthe reference plane (i.e. J(0) just underneath the trace) should be same for all h's (i.e. J0/PI) acording to the formula given in slide 4, unless J0 is also a function of d. What am i missing here?
What does Eric say at 3:06? All differential pairs are two single-ended transmission lines? Edit: ok yes I was right as you can see later on the slides but it's really hard to tell as he seems completely swallow the word "ended".
Excellent!
What is the difference between odd mode impedance and differential impedance? Also, removing the ground plane would not work for even mode?
Odd mode impedance is by definition half the differential impedance. Even mode impedance is the common mode impedance. Removing the ground makes the even mode impedance infinite.
A bifilar transmission line is a differential pair?
So the secondary of a transformer that is not center taped to the ground is still a two single ended pair? If you feel this doesn't sound right and smells bad, you acre correct, because it's a terrible generalization.
That configuration sounds single ended to me
Are erics slides for download somewhere? I'd like to read through the rest of his presentation.
How can I get the full course from Mr Bogatin? It is so interesting!
I'm not certain about that but he does has a couple books. They electronically link to videos for visual illustration of the point at hand.
Another question. The practical case - we have a PC (a computer), and an Internet router/switch. They are connected by a 50 meters unshielded Ethernet cable (U/UTP). Will the differential currents cancel out between every two wires in every four pairs and common currents will find their way in the air as a radio wave between a PC and a router/switch? May we then treat that cable as a 50 meters radiating antenna? And what if we have a shielded ethernet cable (F/UTP) and both ends of that shield are connected to the ground in both PC and a router, so we have the metal, not air, return current path. Will the common currents of four pairs go back in that shield? And what if we will connect that shield just to one device (a PC i.e.)? Will that shield act as 50m antenna? Will it radiate to an air by entire 50 meters shield? Or it will radiate just by that few centimeters between the unconnected shield and a ground in the other unconnected device - in this case the router/switch?
Very useful. Why does Shielded Twisted Pair cause EMI issues? Can someone explain? I think it has to do with common signal return path?
10:25 ? Who said shielded?
thanks man, i somehow heard shielded and was surprised lol@@dzidmail
I think this is because, the differential signal will have a clean return path (the difference ie., D+ - (D-)), but the common single ended (avg.) would actually not have a return path and would oscillate as antenna (without reference) and thus causing EMI issue.. Anyone can feel free to comment and carry this discussion.. Thanks!!
Someone please help me understand what the return current does at the interface where the cable enters/exits the PCB? If we have RS422/485 differential signals on Shielded Twisted Pair, connecting to a PCB with no direct connection of grounds between the two systems, what happens to return current at that interface? It can follow the ground plane in PCB up to the connector, and the differential twisted pair return current is on the other wire in the H/L pair. But at that interface where the return current in the PCB has no where to go, is that similar to what Eric was saying it will give a convoluted return path but they'll be equal/overlap and cancel out?
Differential signal will cancel out, but common signal will find its path as a radio wave between two systems. That's EMI.
If you are not sending a ground along with the twisted pair (which you're probably not doing) then the only way to do it right is to use a transformer as a balun. The centertap of the receiving side should be connected to the ground. This way, you have 2 single-ended signals again which are both referenced to ground. If you open up an ethernet switch, you'll see a large black block right behind the tcpip connectors that contains these transformers. Often they also include extra CM transformers to make sure CM interference doesn't get onto the cable.
Thumbs up for rocking the brass rat
coaxial screen grounded one end (source)!
Regarding the slide number 5, what if we put the N signal instead of the ground plane right below the P signal? Will the return currents cancel out almost completely as they are between P and N line below both of them right on the center?
While mr. Feranec is an amazing layouter and a great tutor in this area, it is embarrassing to see how little knowledge of electronics he possesses. Not mentioning his very poor English. I find having him next a giant (IMO) like prof. Bogatin is just unfortunate.
Why in the hel would you prevent Eric from talking??????