Much appreciated for the Korean subtitle. Though I can speak and read English, the Korean subtitle helps me understand the contents much more quickly and accurately. I'm enjoying watching lectures by Zach Peterson on this channel. Thanks to his intonation and pronunciation, the lectures have good delivery. And also for the subject of the video, the ESD issue is a problem I actually have gone through these days, making a USB product. Thank you for the great video!
In products I have worked on, the chassis ground is connected to the board ground via a parallel R/C, but not acting as a filter. The C is to shunt transients between the grounds, and the resistor is to bleed off any potential difference. This can be required in cases where the externally connected signals could potentially be shorted to the chassis (outside the product) and leakage current must be limited.
Great practical tips. Thanks Zach. We used TVS components to mitigate lightning strikes for 2 major aircraft makers. Amazing to think how well they work against lightning bolts!
Great video I'm still confuse when using cap for mating the chassis and the system ground to have same potential and not generate EMI. I saw the video of Rick Hartley, he said the chassis closure around the PCB is just the metal form the Faraday cage to contain the field not going to outside or enter your cage to have EMI problem and even if it not connects to the ground, it still works. The only thing I can think is it use for diverting the high transient energy to the large metal chassis by the cap to absorb it and ideal this chassis need to connect to Earth. Because this kind of caps have Kv rating
When you have your shield trace around the board (chassis ground), you say to have a low impedance connection to your 0V system ground. If you want a low impedance connection, why even have two separate nets? The lowest impedance connection would be a 360-degree connection to the 0V system ground all the way around the board. This has always confused me about the guard trace. I don't see the benefit. Thanks for the video!
The issue with that approach is that the current that you are shunting is now going through your system ground. Depending on how much energy the transitient has, and how sensitive your components are this might not be okay. By placing the guard rail you are providing a *separate* low impedance path to system ground.
@@theondono I follow your logic but if the Chassis GND is connected to the circuit GND at the connector then a transient at the connector would only flow through a short path to the TVS diode and back to the connector and there would be no need for a separate guard rail.
@@chromatec-video That's a good way to do it if you have only one connector (power + data), but that section of the video discussed at least two connectors, one for power and one with the data lines you are protecting. Connecting the shield at two points to the GND plane (on both connectors) will cause ground loops, so you want to avoid that.
@@theondono Are you saying that the goal would be for the connection between chassis and system grounds to be sort of a fuse, so that if you get a huge transient, it burns out instead of allowing your system ground to get too far out of compliance?
@@jimjjewett Nope. if that were to happen your system would probably be out of compliance anyway since you've lost your connection to the chassis! Imagine that you don't use a separate guard trace, you just use your system ground plane. If you were to have a very big transient spike, that's going to create a very big (but extremely short) current pulse through your ground plane. The connection between your GND plane and the chassis will have a parasitic inductance, it will be small, but it will be there. You can also have other points on your ground plane that might present a higher than ideal parasitic inductance. This creates ground bounces on your board (voltage spikes of the ground net), and those bounces can potentially exceed the ratings of your internal electronics and break them. If you can keep those current spikes outside of your system, you'll have way less issues.
Hi! As I understand, applications with high precision and low noise are sensitive to anything present in the signal path, such as high resistances, leakage currents and differential mismatch (which degrades common-mode rejection). Should TVS diodes be avoided in this cases? If so, what would be an alternative?
Hey Zach, thank you for the video! I realized that you have continuous guard ring around the PCB, which is not suggested by some guides as it creates a loop that can interact with magnetic field changes. What is your take on this?
It depends on what you do with the guard ring. If you include it then it is not recommended to leave it floating, it should be connected to some ground if it is present. Also using small clearances between your system ground and chassis ring will minimize the exposed loop area, so that becomes a non-issue. In my example, it is used as a point to connect to chassis ground; this is a typical usage for the guard ring. Another usage is to connect the board to a metalized (shielded) enclosure, and as long as the metalized chassis does not carry return current then you will not have any problems with shocking a user. Typically you would not need to create a ring if you did not have a chassis/earth connection in the device, instead you would just connect everything to system ground. You can include the ring in cases without a chassis ground and use it as connected to system ground, but it should use a star ground approach. The use of vias is used to increase the shielding effectiveness for radiation and it ensures minimal loop area in the cross-section of the ring.
Hmm, if the circuit GND and Chassis GND are to be at the same potential then is there really any benefit in running a separate Chassis GND strip around the PCB on all layers? It would be a lot easier to just connect Chassis GND and electrical GND at the connector?
The point is to divert the current into the chassis ground and thus the chassis rather than into the system ground. Normally in systems that use chassis ground there could be an earth connection somewhere, then that current will eventually go back to earth. Just about anything that connects to an AC mains line will do this for that reason. The point of doing it at one point near the input connector is because the earth connection tends to also be at that connector. Sometimes you don't do this connection at all because you have a floating ground for your system ground, such as if your input power comes into an isolated power supply.
Hi zach ! Great topic as ever. want to know that is it a good practice to have a gnd plane for chassis and make sure that there are only connections with the chassis of the shielded blocks
You could have an entire plane for the chassis ground if you want, maybe if you expect really high current then that would make sense. I think one of the reasons you would not do this in practice is that it takes up an extra layer in the PCB. Also it would overlap with a system ground plane, so then current from signals or power could enter that plane capacitively and could then flow through the chassis. The chassis ground and the chassis itself should not carry currents because those currents could interact with the user of the device, who might then get an annoying shock.
A GDT provides much higher level of protection and would be used when very large transients might be experienced during operation. Small GDTs are also used in residential surge protectors. These are sometimes used in parallel with TVS diodes. So you would put GDT closer to the ESD entry point, and then put the TVS diode after that. Together they provide multiple levels of protection with different response rates.
I still can't understand why bidirectional TVS DIODE is better than unidirectional TVS DIODE. When there is ESD noise on GND, the bidirectional will also start to discharge the noise energy to the positive pole of the power supply, like unidirectional TVS DIODE. At this time, GND may even have a reverse voltage.
The bidirectional TVS diode is better for two reasons. First, it can handle differential lines that might switch between positive and negative polarity, this is one of the main reasons to use it on data lines. 2nd, when there is a strong ESD event that brings the ground potential above the signal line potential, the bidirectional diode will not get forward biased, it is additional protection that allows the injected pulse to dissipate into your system's protective ground.
Bidirectional TVS diodes generally don't make sense to me. If you can't rely on your GND plane to have low impedance, you've got bigger problems than ESD protection. A bidirectional TVS creates a symmetrical clamp about GND, which is not what you want for negative transients on a CMOS logic circuit, for example, which has a Vil(min) of -0.3V. If you use a bidirectional diode here you are clamping at -3.3V. If you just use a unidirectional diode here you will clamp at Vf which is much closer to -0.3V.
The bidirectional diode is for two reasons: 1) in the instance where the GND plane potential suddenly rises above the signal line potential (negative ESD polarity), and 2) (the more common case) where the signal has alternating negative and positive polarity. AC-coupled differential pairs would be one example. If you look at the specs for bidirectional diodes the clamp voltage is not dependent on direction, they are symmetric devices and the clamp voltage can be seen as a magnitude value.
you're the best amazing helpful guy who actually teaches professional high level design lessons... thank you man
Much appreciated for the Korean subtitle. Though I can speak and read English, the Korean subtitle helps me understand the contents much more quickly and accurately.
I'm enjoying watching lectures by Zach Peterson on this channel. Thanks to his intonation and pronunciation, the lectures have good delivery. And also for the subject of the video, the ESD issue is a problem I actually have gone through these days, making a USB product. Thank you for the great video!
You're welcome 😊
In products I have worked on, the chassis ground is connected to the board ground via a parallel R/C, but not acting as a filter. The C is to shunt transients between the grounds, and the resistor is to bleed off any potential difference. This can be required in cases where the externally connected signals could potentially be shorted to the chassis (outside the product) and leakage current must be limited.
Great practical tips. Thanks Zach. We used TVS components to mitigate lightning strikes for 2 major aircraft makers. Amazing to think how well they work against lightning bolts!
Very cool!
Great video
I'm still confuse when using cap for mating the chassis and the system ground to have same potential and not generate EMI. I saw the video of Rick Hartley, he said the chassis closure around the PCB is just the metal form the Faraday cage to contain the field not going to outside or enter your cage to have EMI problem and even if it not connects to the ground, it still works.
The only thing I can think is it use for diverting the high transient energy to the large metal chassis by the cap to absorb it and ideal this chassis need to connect to Earth. Because this kind of caps have Kv rating
When you have your shield trace around the board (chassis ground), you say to have a low impedance connection to your 0V system ground. If you want a low impedance connection, why even have two separate nets? The lowest impedance connection would be a 360-degree connection to the 0V system ground all the way around the board. This has always confused me about the guard trace. I don't see the benefit. Thanks for the video!
The issue with that approach is that the current that you are shunting is now going through your system ground.
Depending on how much energy the transitient has, and how sensitive your components are this might not be okay. By placing the guard rail you are providing a *separate* low impedance path to system ground.
@@theondono I follow your logic but if the Chassis GND is connected to the circuit GND at the connector then a transient at the connector would only flow through a short path to the TVS diode and back to the connector and there would be no need for a separate guard rail.
@@chromatec-video That's a good way to do it if you have only one connector (power + data), but that section of the video discussed at least two connectors, one for power and one with the data lines you are protecting.
Connecting the shield at two points to the GND plane (on both connectors) will cause ground loops, so you want to avoid that.
@@theondono Are you saying that the goal would be for the connection between chassis and system grounds to be sort of a fuse, so that if you get a huge transient, it burns out instead of allowing your system ground to get too far out of compliance?
@@jimjjewett Nope. if that were to happen your system would probably be out of compliance anyway since you've lost your connection to the chassis!
Imagine that you don't use a separate guard trace, you just use your system ground plane. If you were to have a very big transient spike, that's going to create a very big (but extremely short) current pulse through your ground plane.
The connection between your GND plane and the chassis will have a parasitic inductance, it will be small, but it will be there. You can also have other points on your ground plane that might present a higher than ideal parasitic inductance. This creates ground bounces on your board (voltage spikes of the ground net), and those bounces can potentially exceed the ratings of your internal electronics and break them.
If you can keep those current spikes outside of your system, you'll have way less issues.
Hi! As I understand, applications with high precision and low noise are sensitive to anything present in the signal path, such as high resistances, leakage currents and differential mismatch (which degrades common-mode rejection). Should TVS diodes be avoided in this cases? If so, what would be an alternative?
Hey Zach, thank you for the video! I realized that you have continuous guard ring around the PCB, which is not suggested by some guides as it creates a loop that can interact with magnetic field changes. What is your take on this?
It depends on what you do with the guard ring. If you include it then it is not recommended to leave it floating, it should be connected to some ground if it is present. Also using small clearances between your system ground and chassis ring will minimize the exposed loop area, so that becomes a non-issue. In my example, it is used as a point to connect to chassis ground; this is a typical usage for the guard ring. Another usage is to connect the board to a metalized (shielded) enclosure, and as long as the metalized chassis does not carry return current then you will not have any problems with shocking a user. Typically you would not need to create a ring if you did not have a chassis/earth connection in the device, instead you would just connect everything to system ground. You can include the ring in cases without a chassis ground and use it as connected to system ground, but it should use a star ground approach. The use of vias is used to increase the shielding effectiveness for radiation and it ensures minimal loop area in the cross-section of the ring.
Hmm, if the circuit GND and Chassis GND are to be at the same potential then is there really any benefit in running a separate Chassis GND strip around the PCB on all layers? It would be a lot easier to just connect Chassis GND and electrical GND at the connector?
The point is to divert the current into the chassis ground and thus the chassis rather than into the system ground. Normally in systems that use chassis ground there could be an earth connection somewhere, then that current will eventually go back to earth. Just about anything that connects to an AC mains line will do this for that reason. The point of doing it at one point near the input connector is because the earth connection tends to also be at that connector. Sometimes you don't do this connection at all because you have a floating ground for your system ground, such as if your input power comes into an isolated power supply.
Thanks for a good explanation of the bi-directional TVS with respect to GND.
Hi zach ! Great topic as ever.
want to know that
is it a good practice to have a gnd plane for chassis and make sure that there are only connections with the chassis of the shielded blocks
You could have an entire plane for the chassis ground if you want, maybe if you expect really high current then that would make sense. I think one of the reasons you would not do this in practice is that it takes up an extra layer in the PCB. Also it would overlap with a system ground plane, so then current from signals or power could enter that plane capacitively and could then flow through the chassis. The chassis ground and the chassis itself should not carry currents because those currents could interact with the user of the device, who might then get an annoying shock.
dear friends, does tvs or gdt provide better protection?
A GDT provides much higher level of protection and would be used when very large transients might be experienced during operation. Small GDTs are also used in residential surge protectors. These are sometimes used in parallel with TVS diodes. So you would put GDT closer to the ESD entry point, and then put the TVS diode after that. Together they provide multiple levels of protection with different response rates.
I still can't understand why bidirectional TVS DIODE is better than unidirectional TVS DIODE.
When there is ESD noise on GND, the bidirectional will also start to discharge the noise energy to the positive pole of the power supply, like unidirectional TVS DIODE.
At this time, GND may even have a reverse voltage.
The bidirectional TVS diode is better for two reasons. First, it can handle differential lines that might switch between positive and negative polarity, this is one of the main reasons to use it on data lines. 2nd, when there is a strong ESD event that brings the ground potential above the signal line potential, the bidirectional diode will not get forward biased, it is additional protection that allows the injected pulse to dissipate into your system's protective ground.
Bidirectional TVS diodes generally don't make sense to me. If you can't rely on your GND plane to have low impedance, you've got bigger problems than ESD protection. A bidirectional TVS creates a symmetrical clamp about GND, which is not what you want for negative transients on a CMOS logic circuit, for example, which has a Vil(min) of -0.3V. If you use a bidirectional diode here you are clamping at -3.3V. If you just use a unidirectional diode here you will clamp at Vf which is much closer to -0.3V.
The bidirectional diode is for two reasons: 1) in the instance where the GND plane potential suddenly rises above the signal line potential (negative ESD polarity), and 2) (the more common case) where the signal has alternating negative and positive polarity. AC-coupled differential pairs would be one example. If you look at the specs for bidirectional diodes the clamp voltage is not dependent on direction, they are symmetric devices and the clamp voltage can be seen as a magnitude value.
@@Zachariah-Peterson thanks for the explanation. Much appreciated!
Martin Robert Martin Robert Hernandez Kevin