Best solution for ESD protection of USB port.

[Asgards]

Beginner here. Im trying to find a good ESD protection IC for a simple USB to rs232 adapter

I am looking at  PRTR5V0U2X and IP4234CZ6, both from NXP.

To me it seems like most of the specifications are very similar. Both has low capacitance, both are ideal for USB, protects up to 8kv and can protect the data and Vbus lines. The main difference I see is that the IP4234CZ6 has 6 legs and the data lines goes through the IC whereas the PRTR5V0U2X only has 4 legs and the data lines are simply connected to the IC.

What is the reason for this? Which IC provides best ESD protection?

[German_EE]

If it's just static that you want protection from a 1M ohm resistor from each line to chassis may be enough. Actual isolation between the socket and the rest of the circuit needs a more technical solution.

[Asgards]

Thanks for the reply, however it does not really answer my question.

[Jeroen3]

The IP4234CZ6 includes resistors.
You'll be adding the 27 ohm resistors anyway, wouldn't pay to much attention to it. Get the cheapest 

[Alexei.Polkhanov]

Almost always when I look at USB device I find TVS diodes such as ones from company called Littlefuse (http://www.littelfuse.com/products/tvs-diode-arrays/general-purpose-esd-protection.aspx).

[Jeroen3]

Diodes are mode by almost everybody nowadays.
Even TI has some: http://www.ti.com/lit/ds/symlink/tpd2e009.pdf

[T3sl4co1l]

The second is best.  Two reasons: the trace length that necessarily tees off to the first device (if not in the layout itself, then the lead, bond wire and supply pins) add stray inductance and resistance.  Which increases overshoot in those first few nanoseconds.

The second is better for two reasons: one, the clamping is two stage, so the leftovers from the first (usually on the order of 30V, a few ns wide, and a low impedance) are attenuated very effectively by the second stage, although still limited by supply/ground inductance.  The second reason: the resistor divides with the capacitance of both to provide a very small amount of filtering at high frequencies, which also helps with EMC.

If you can verify that your RS232 converter is USB 1.1 / 2.0-compatible Full Speed only (not USB 2.0 High Speed), the signals are plain ordinary 3.3V CMOS level, and can be filtered accordingly (some series resistance and relatively large C).  I would go for a cutoff frequency of 20-30MHz, which will help filter noise and residual ESD from the lines.

Tim

[free_electron]

just cramming on diode's is not enough.
care needs to be taken to split system and chassis groudn , do the bridging correctly and so on ...

[Asgards]

The second is best.  Two reasons: the trace length that necessarily tees off to the first device (if not in the layout itself, then the lead, bond wire and supply pins) add stray inductance and resistance.  Which increases overshoot in those first few nanoseconds.

The second is better for two reasons: one, the clamping is two stage, so the leftovers from the first (usually on the order of 30V, a few ns wide, and a low impedance) are attenuated very effectively by the second stage, although still limited by supply/ground inductance.  The second reason: the resistor divides with the capacitance of both to provide a very small amount of filtering at high frequencies, which also helps with EMC.

If you can verify that your RS232 converter is USB 1.1 / 2.0-compatible Full Speed only (not USB 2.0 High Speed), the signals are plain ordinary 3.3V CMOS level, and can be filtered accordingly (some series resistance and relatively large C).  I would go for a cutoff frequency of 20-30MHz, which will help filter noise and residual ESD from the lines.

Tim

Thanks a lot for all the replies. Tim, it sounds like you have a lot of experience regarding this. If you get a chance, please take a look at the layout I have attached here. Do you recommend any changes or improvements regarding the ESD protection for the USB port? or for the entire circuitry? Should I remove the R2 and R3 resistors on the USB data lines since the IP4234CZ6 already has resistors built-in? Any suggestions for improvements or changes are much appreciated. 
The converter should work with USB 1.1, 2.0 and 3.0 ports. Max baud rate is 1Mbps.

[Alexei.Polkhanov]

Perhaps for UBS 3 you need to use low capacitance diodes - these are just a bit more expensive.

Here I have attached example from FT232 datasheet which shows 2 TVS diodes for protection.
(http://www.ftdichip.com/Support/Documents/DataSheets/ICs/DS_FT232H.pdf)

[Asgards]

Perhaps for UBS 3 you need to use low capacitance diodes - these are just a bit more expensive.

Here I have attached example from FT232 datasheet which shows 2 TVS diodes for protection.
(http://www.ftdichip.com/Support/Documents/DataSheets/ICs/DS_FT232H.pdf)

Thanks for the suggestion Alexei, I do see that the PGB1010603 which FTDI use in their circutry has only 0.06 pF where as the IP4234CZ6 I was planning to use has 2.5 pF. I guess the lower the better (?).

[T3sl4co1l]

Existential question... why are you making this at all?  Surely you can buy one of these off the shelf.

FT231XS is a "Full Speed" device, so its signalling tops out at 3.3V CMOS and does not require high speed differential considerations (like single digit pF caps).

You can save the extra caps by using somewhat meatier devices (like zeners directly, or schottky).  I would just as well nix the 47pF caps and special ESD device and use a pair of BAT54S (of course, this costs stray inductance due to package design and layout).  But either is OK.

The IP4234 doesn't specify the "R" between clamp diodes; there's an "Rs" spec, but it's not defined what this is...  A better choice might be one that does explicitly incorporate RC filtering, and specifies series resistance and capacitance.  Think there's an On Semi USB filter/ESD clamp device out there but I don't remember the part number offhand.  In that case, you won't have to worry about the series resistors either.

C1 should be much larger if using clamp diodes, otherwise, the local zener/TVS in the clamp device will handle transients.

The ferrite bead is unnecessary; if you want to reduce power line noise, or hot-plugging transients, use separate filtering, or a series resistor, or an active current limiting circuit (kind of unnecessary as the other end of the USB port is required to have such a limiting device as-is).  This might be interesting on the serial end, as RS-232 related devices I don't think are supposed to run over 10 or 20mA, or something like that.  But if you're not following a particular standard, word for word (i.e., the traditional DE-9 async serial connector, TIA-574), you can of course do whatever suits your purpose.  As shown, a short on the serial end (pin 5 to 9) will brown out the USB port, causing it to disconnect (and throw a power consumption fault, if so equipped).

Only thing that jumps out to me: GROUND THOSE SHIELDS!  There IS no "oh, I'll just connect this through a jumper / resistor / capacitor / etc. and it'll be alright", no, the days of that are long gone.  Anyone who says you can connect a shield loosely is at best ignorant (repeating a very old suggestion, but never having tested it himself).  Even if you limit signal bandwidth on your end, you can't escape the low-frequency energy delivered by ESD or EFT events without compromising the signals themselves, and you can't address the transients that propagate back up the cable to the host end.

Tim