15
\$\begingroup\$

I was looking over the schematic for the GumStix Palo 43 and noticed they used a common mode choke coil on the data lines coming in from USB.

I understand how this design can help remove noise coming in on the USB lines, but I wonder if it is actually something I should start doing on my designs. The datasheet for the FT232R has no mention of adding common mode choke coils, and I have used this chip before with out one.

So, would you recommend I change my USB design or keep it the way it is?

\$\endgroup\$
  • 2
    \$\begingroup\$ Quite rarely do datasheets or even application notes actually give you all the information to use a part successfully in regards to EMI emissions and noise generation in the rest of the system. Classic examples that should hardly ever be followed are that a single 0.1uF cap is sufficient power supply decoupling for a power pin and that you should have separate analog and digital ground planes (doing this one especially almost always results in EMI disasters). Another tip as you found is to be liberal with ferrite beads and common mode chokes on any signal that leaves or enters the board. \$\endgroup\$ – Mark Apr 30 '10 at 22:09
  • \$\begingroup\$ Also, what speed is the USB port? \$\endgroup\$ – endolith Aug 27 '10 at 20:18
  • \$\begingroup\$ The FT2232H is Hi-Speed, so I would plan for the "worst case" where the computer is Hi-Speed also. \$\endgroup\$ – Kellenjb Aug 28 '10 at 16:03
  • \$\begingroup\$ @Mark - Very apropos for me right now that you cited separate analog and digital ground planes as almost always resulting in EMI disasters - that's exactly what I'm going through right now. Traditional "star" topology rules were more or less followed, theoretically bringing analog and digital grounds separately back to a central ground point. And I have an EMI disaster on my hands, sigh. OTOH, common grounds might introduce digital noise into the analog stuff which is what the separate grounds are supposed to avoid... what to do? \$\endgroup\$ – user4102 May 4 '11 at 0:50
  • \$\begingroup\$ @Mark, I would not agree on the separate ground planes causing issues if done properly. You should be able to segregate a large amount of noise so that your analog has higher SNR. You can however do this very very wrong, very easily. \$\endgroup\$ – Kortuk May 4 '11 at 0:54
21
\$\begingroup\$

The USB signal is not entirely differential, so it's not a great idea. (The end-of-packet (EOP) signal is both pins pulled low, which, I believe, is why there's always noise at 1 kHz and harmonics in USB systems, since it's sending common-mode signals every 1 ms.)

  1. A common mode (CM) choke should be used to terminate the high speed USB bus if they are need to pass EMI testing. Place the CM choke as close as possible to the connector pins. See Section 5.1 for details.

Note: Common mode chokes degrade signal quality, thus they should only be used if EMI is a known problem.

Common mode chokes distort full speed and high-speed signal quality. The eye diagram above shows full speed signal quality distortion of the end of packet, but still within the specification. As the common mode impedance increases, this distortion will increase, so you should test the effects of the common mode choke on full speed and high-speed signal quality.

High Speed USB Platform Design Guidelines

Note: additional filtering may be achieved by winding the 4 wires through the ferrite bead an additional turn. As with the use of ferrite beads in signal paths, care should be taken to insure that the signaling meets rise and fall times, especially the EOP signaling. EOP signaling is single ended and may be strongly affected by a single bead, which acts as a common mode only filter.

Intel EMI Design Guidelines for USB Components

\$\endgroup\$
  • 2
    \$\begingroup\$ +1 that's my experience, too. Common mode choke will degrade the signal. As pingswept mentions, you should include the footprint for the choke (and ESD protection too!); you can always just not populate those components during assembly if they are not needed. \$\endgroup\$ – ajs410 Apr 29 '10 at 21:31
  • 1
    \$\begingroup\$ @ajs410, I think adding the footprint and not populating increases the chance of a need as you will almost guarantee extra mismatch at the pad location. \$\endgroup\$ – Kortuk Jun 20 '10 at 21:26
  • \$\begingroup\$ Would you really get that much extra EMI from a pad that has no solder mask, but otherwise makes no stub on the trace? \$\endgroup\$ – ajs410 Sep 1 '10 at 21:39
  • 1
    \$\begingroup\$ yes. Actually, the main issue, and the reason you need a common-mode choke is due to the ability of a USB cable to radiate. if the signal enters your board, collides with a mismatch, the reflected signal will traverse the cable again, radiating. whenever you have a long wire is when you want to do your best to minimize reflections entering the coil. The reason you add a common-mode choke is due to the large amount of EM generated by common-mode currents compared to differential mode. But, even though a common mode choke should not affect differential mode, there are imperfections. \$\endgroup\$ – Kortuk Sep 19 '10 at 20:42
6
\$\begingroup\$

There's a good chance that the choke is included on the Gumstix board because they have to pass FCC Title 47 CFR Part 15 emissions testing for unintentional radiators to sell their device.

You might not care about that, but if it's for a commercial product I think it's cheaper to add the choke now and then remove it if you discover it's not needed.

\$\endgroup\$
3
\$\begingroup\$

If your product sits "floating" on the USB cable, the common-mode choke is probably not necessary.

However, if your device is electrically attached through any other paths that could form a loop back to where your USB cable originates, you will have a potential for inductively picking up or generating noise that can affect the performance of your product, or other product attached to it. At that point, you will need the choke.

\$\endgroup\$
0
\$\begingroup\$

Yes, I would recommend using a common mode choke. It helps with EMI but as everyone has said, you will have a degraded signal. If you don't need the choke, just add jumper wires.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.