ESD protection on USB 2.0 ports. Any advice?

[knotlogic]

I'm building a small custom USB 2.0 hub for a project, and one thing I'd really like is to include ESD protection on the user facing ports.  Needless to say there are a *lot* of parts out there catering to this application...

The more interesting parts are the "all in one" protection devices.  Usually a 6 pin IC that offers ESD protection on both data lines, and optionally protection for the Vbus line.  Given I'll be hand assembling this, I figure this will be easier than dealing with micro-sized chips for each line.  Especially considering my hands aren't as steady as they used to be.  So far I've found the following parts:

Onsemi - CM1231-02SO
Nexperia - IP4234CZ6
Bourns - CDSOT563-0502
Bourns - CDSOT236-0502

I'm sure there are more like these, I just haven't found them yet.

What should I be looking out for when picking one of these protection arrays?  I know I want to minimise capacitive loading on the data lines, but all these parts seem more or less the same in that regard.  And on the plus side I'm only looking at USB 2.0.  The Bourns parts lack a series resistor, and there are some other circuit differences between them, but that aside they seem to be about the same to my untrained eye...

Are there any other parts I might want to look at/avoid?

Thanks for any advice you can give!

[Berni]

Yep these combined ESD diodes for high speed lines are the right thing to use even when doing this in volume. Not only are they designed for minimum loading of the data lines, but they are also matched between the lines and usually give you a pinout for nice "flow trough" routing to keep the lines going fairly parallel as a dif pair.

Most of the difference between ESD protection diodes is how much abuse they can take before they fail into a dead short (so still protected your device but the port no longer works). In general USB lines should not be exposed to particularly harsh abuse so probably any diode is good enugh.

The series resistor is a good idea because adding resistance before ESD diodes can drastically increase there effectiveness. This is because the resistor limits the amount of fault current that can flow into the diode, protecting the diode while the resistor is the one that dissipates most of the thermal power during a fault. The 22Ohm resistors you often see before the USB chip pins also help here since they limit the amount of current that can flow into the chips built in weak ESD diodes. Those resistors also tend to save the chip from blowing up in the event of 5V getting into the data lines (very common thing since data and 5V is next to each other in the USB connector)

[knotlogic]

Thanks Berni!  Good to know I'm heading down the right path.

[Leo Bodnar]

I have never seen inline resistors on HS USB 2.0 design.  Are you sure?

If you stick 22R resistors into existing HS USB design it will immediately violate USB impedance spec (7.1.6.2)

Leo

Yep these combined ESD diodes for high speed lines are the right thing to use even when doing this in volume. Not only are they designed for minimum loading of the data lines, but they are also matched between the lines and usually give you a pinout for nice "flow trough" routing to keep the lines going fairly parallel as a dif pair.

Most of the difference between ESD protection diodes is how much abuse they can take before they fail into a dead short (so still protected your device but the port no longer works). In general USB lines should not be exposed to particularly harsh abuse so probably any diode is good enugh.

The series resistor is a good idea because adding resistance before ESD diodes can drastically increase there effectiveness. This is because the resistor limits the amount of fault current that can flow into the diode, protecting the diode while the resistor is the one that dissipates most of the thermal power during a fault. The 22Ohm resistors you often see before the USB chip pins also help here since they limit the amount of current that can flow into the chips built in weak ESD diodes. Those resistors also tend to save the chip from blowing up in the event of 5V getting into the data lines (very common thing since data and 5V is next to each other in the USB connector)

[wraper]

Series resistor like 22R is only good when USB device requires external resistors for impedance matching. Most of the modern ICs do it internally. ONsemi and Nexperia ESD protection devices which were mentioned have negligibly low series resistance.

[poorchava]

I recall using such TVSs arrays from ST. USBLC6-2 was the name IIRC. Just does what it says. Quick and easy.

The CDSOT stuff is also available in RS485 specific variant.

[knotlogic]

I have never seen inline resistors on HS USB 2.0 design.  Are you sure?

If you stick 22R resistors into existing HS USB design it will immediately violate USB impedance spec (7.1.6.2)

Leo

Some parts like the FTDI FT231X show 27R resistors on the data lines in their datasheet examples.  It's probably device dependent, their datasheets for the FT245R and FT232R don't have them.

But in this case, as wraper noted, the resistor is a 1R internal to the TVS device.  I can't imagine the designers would have included that if it was going to mess up the data lines that badly.

[exmadscientist]

I really like the ST ECMF02-4CMX8 for USB 2.0 HS protection. The flow-through routing makes it really easy to integrate, and it's got a common-mode filter as well as the TVS array. Its downsides are that it's difficult to manually solder (being a DFN package will do that) and it's had some availability issues even in the Before Times.

But, man, does it work well.

[Berni]

I have never seen inline resistors on HS USB 2.0 design.  Are you sure?

If you stick 22R resistors into existing HS USB design it will immediately violate USB impedance spec (7.1.6.2)

Leo

As others have mentioned this is a bit chip specific.

Some USB chips have pin drive that does not match the required USB impedance and so they recommend adding some external resistance on the data lines. Once you sum together the resistance of the chips internal drive transistors and the external resistor you get the 90 Ohm impedance that USB requires.

Conveniently this external resistor provides a good deal of surge protection to the chip and can be used as a way of forcing most of the fault current trough the external ESD diode. The external ESD diode can typically handle more abuse before blowing up, and if it does blow up its easier to replace a simple diode rather than a MCU (especially when you don't have the firmware that was inside it).

The 1 Ohm of resistance built into some of those ESD diodes is not going to hurt anything either since most of your USB cables are not 90Ohm +/- 0.5Ohm anyway. They also run trough 0.1 inch pin headers on a motherboard that are most certainly also not 90Ohm.

Sure some things do sacrifice some signal integrity but the difference is so minuscule that it by far outweighs the much grater fault resilience

[knotlogic]

I really like the ST ECMF02-4CMX8 for USB 2.0 HS protection. The flow-through routing makes it really easy to integrate, and it's got a common-mode filter as well as the TVS array. Its downsides are that it's difficult to manually solder (being a DFN package will do that) and it's had some availability issues even in the Before Times.

But, man, does it work well.

Thanks, that looks like a really interesting part.  ESD rating seems a hair lower than the CM1231-02SO though?

[knotlogic]

So, one more question I have after browsing though an XMEGA app note (specifically Atmel AVR1017: XMEGA - USB Hardware Design Recommendations):

Should the TVS array anodes be attached to the circuit ground or the USB shield?  This is assuming a case where the two are separated by an RC combination like in the aforementioned app note.  Atmel chose the latter, but the datasheets for the TVS devices themselves appear to tie to circuit ground while ignoring the shield completely.

[Berni]

That looks odd to connect it to the shield.

The TVS diode is protecting the MCUs USB pins, and the voltage they see is measured in reference to its GND pin so it makes sense to also connect ESD diodes to that so that they start conducting in reference to GND.

Now that i think about it this could even be harmful. If you plug a static charged USB cable (Or a static charged user holding the cable) the first thing that touches is the shield. So in this case a static discharge into the shield could charge up that EMI capacitor to fairly high voltage at what point the ESD diodes conducts and pull the USB data lines to that very negative/positive voltage. Tho realistically the Vcc diode likely also conducts and diverts the current into the supply rail where there are usually quite a few uF of capacitance to eat up the energy.

[wraper]

So, one more question I have after browsing though an XMEGA app note (specifically Atmel AVR1017: XMEGA - USB Hardware Design Recommendations):

Should the TVS array anodes be attached to the circuit ground or the USB shield?  This is assuming a case where the two are separated by an RC combination like in the aforementioned app note.  Atmel chose the latter, but the datasheets for the TVS devices themselves appear to tie to circuit ground while ignoring the shield completely.

That appnote is an utter garbage. Do it like that if you want EMC/EMI problems (especially if you run in high speed mode) and susceptibility to ESD. Shield must be directly connected to the GND plane period.
https://www.unit3compliance.co.uk/hdmi-more-like-hdm-why-thoughts-on-cable-shield-grounding/

Worth to watch a whole video, but certainly watch from 30:50

[knotlogic]

Thanks all!

Worth to watch a whole video, but certainly watch from 30:50

That's definitely worth watching.  Thanks for the link!