3

I am designing a keyboard which will connect to the computer via usb-miniB to charge a LiPo battery. The datasheet for the charging IC I am using specifies overvoltage protection on the USB power line, and I have been looking into various TVS diodes. I am finding differing information about where to connect the TVS diodes to:

This FTDI doc provides an example to shield ground, and also advocates and RC filter connecting shield ground and signal ground.

This TI part, which is advertised as for USB, provides an example to signal ground.

I am wondering which is the better practice, and whether that practice changes depending on whether it is a TVS intended for OVP rather than ESD protecton.

pbandjazz
  • 337
  • 1
  • 4
  • 11

3 Answers3

2

The choice of where to ground the ground pin of a ESD chip depends on the type of device you are making.

If the device is powered from an individual power source that goes to wall AC outlet, it should have the shield separated from signal plane, and the shield should be connected to signal plane only at one point, preferably in the area of DC power jack (if there is no special third-pin on AC cord). In this case the ESD protective device should be grounded to SIGNAL GROUND PLANE, to prevent excessive ESD spike from getting across pins of the IC that you want to protect. If grounded to the shield (with usually high inductance), the TVS device will not be able to shunt the voltage across actual signal pins, and the efficiency of ESD protection will be lower.

If your device is battery-powered or USB bus-powered, there is no alternative path provided by shield for the ESD spike. In portable hand-held electronics it is advisable to make the shield tightly connected to signal ground, just as the same copper plane. In this case there is no difference where to connect the TVS protector.

However, the path for ESD discharge can be complicated, and will depend on overall system layout and internal grounding and shield placement. It is still a CAD challenge to model all the 3-D complexity of electric field with all device conductors and conductive covers, so the most practical way is to build several configurations of the device, and determine the best variant experimentally.

Ale..chenski
  • 38,845
  • 3
  • 38
  • 103
  • Thank you for the thorough answer. I have been reading a lot about the shield-signal ground connection - whether it should be an RC filter, direct connect - but where you say "there is no alternative path provided by shield for the ESD spike", I am a little confused. Doesn't the ESD spike come from the actual USB cabling? And since (usually) the shielding is inside the cable as well, wouldn't that be a valid alternative path? – pbandjazz May 30 '17 at 14:52
  • @pbandjazz, I should say "explicitly engineered" path instead of "valid alternative". In case of protecting power line of a charging IC in a portable device, I would connect a TVS between points that are closest to IC pins you need to protect. TVS works by shunting the ESD surge, so you should shunt the pins as best as you can. How the shield is connected to your signal ground, some considerations are presented HERE https://electronics.stackexchange.com/a/269313/117785 and HERE, https://electronics.stackexchange.com/a/291380/117785. – Ale..chenski May 30 '17 at 17:28
1

The recommended circuit has a few TVS diodes, but doesn't recommend what to get. TVS diode selection needs to consider the capacitance of the TVS diode because that could reduce the speed of the USB line (from 2.0 to 1.2) if the capacitance is too high. It's easier to look at IC's because they are designed to a specific ESD standard (like IEC 61000-4-2).

The TPD4S012 has a "±10-kV IEC 61000-4-2 Contact Discharge" ESD rating and a low capacitance of under 1pF. Although it would appear from the diagram that the TPD4S012 has OVP protection, nothing in the datasheet would indicate that. It is only rated for a surge current of 3A for 20us additionally the \$V_{BR}\$ starts at 6V to 9V depending on the part for the data lines. The \$V_{BR}\$ for Vbus starts at 20V. Sot its a really poor OVP protector.

They don't provide any package information in the datasheet about thermal ratings, but they do say it can't get hotter than 85C, but if we consider that most of the heat is going to leak into the PCB the juction to board thermal resistance of 343C/W would suggest the part would exceed its absolute maximum rating after dissipating 250mW which wouldn't be hard to do in an OVP situation.

While the part is great for ESD protection, I would use an OVP diode on the Vbus ahead of this part or a Zener.

If ESD protection is not required (by regulatory standards and testing) (they zap your desing in regulatory with an ESD gun) then go with the simplest\cheapest design like TVS diodes.

enter image description here

Voltage Spike
  • 75,799
  • 36
  • 80
  • 208
-1

The FTDI-version shown in the previous post will not work.

The TVS diodes at VBUS, D+, and D- shall protect the connected ICs from being damaged. Those ICs have their voltage supplies referenced to GND, but not to the shield. Hence, the TVS-diodes need to go to GND, but not to the shield.

The shield is intended to avoid the cable becoming an antenna. For that purpose the shield shall be "grounded" for RF-signals by a capacitor with a resistor in parallel to discharge that capacitor from ESD-events - typical values are 4.7-10nF and 1MOhm.

Additionally, one may put a TVS-diode in parallel to that capacitor to prevent the capacitor from being damaged by ESD-events.

Moreover, if the TVS-diodes especially from the data lines are connected to the shield, that capacitive coupling may cause EMC-problems and EMC-tests fail and the signal integrity is significantly disturbed. Do not forget that the shield of the attached USB-cables is typically accessible and can be touched.

W5VO
  • 18,303
  • 7
  • 63
  • 94
Österreichischer Flochlandla
  • 11