infrequent USB dropouts on SBC

Question

I have a full speed USB audio device connected to the USB port on a SBC running Ubuntu Linux. I'm acquiring data over this connection and I need 24/7 uninterrupted data acquisition. If the USB connection drops at all then I lose data since the device has no NVM and is powered via USB VBUS.

I have 20 of these systems deployed in a typical North American urban environment, in the rear yards of residences, away from any obvious EMI sources. Over the course of 2 weeks of data collection I've had 2 momentary USB dropouts (lasting < 1 second).

The SBC I'm using is this one and it exposes the USB host interface as four male header pins. I have a custom 6' USB cable which connects to the device using microUSB-B and at the host end has a 1" length of breakout to female headers. For this final inch there is no shielding and no twist in the data conductors. The drain of the USB cable is connected to a ground pin on the SBC. Neither the USB host nor USB device has a metal enclosure. The USB connections and all electronics are all well protected from water/dust ingress.

Prior to connecting the drain to ground I had a much higher rate of dropouts.

Questions:

1. Is there anything about the arrangement described that jumps out as a red flag as likely responsible for these dropouts?
2. Is the dropout rate I'm experiencing expected for a typical USB system? I've never noticed USB dropouts in my day-to-day life with USB (e.g, using a mouse, keyboard, webcam, etc) but I don't use USB for activities that would notice this drop rate.
3. Is it reasonable to assume, since the port is exposed as header pins instead of a female USB-A port, that the SBC manufacturer expects the user to add additional circuitry after the headers and before the physical USB port? What would that typical circuitry be?
4. Following on #3, would the SBC manufacturer likely not have implemented onboard EMI and/or ESD protection, and would users be wise to implement their own using something such as this device? I was informed by the USB device manufacturer that USB Implementers Forum does not recommend adding any passive network on the line and as such their USB device does not have any filtering. Yet filters like the Wurth unit I linked presumably exist for good reason. I've read this excellent answer yet I'm still unclear if this filter falls under the solutions identified as "flawed".

Answer 1

There are many possible explanations for the symptoms described, but EMI is among them, yes.

Going on that as an assumption:

A common use-case for USB is to connect via headers -- which as you note, are unshielded for an inch or two, which ruins the EMI performance of the connection. The assumption is that the board will be integrated into a metal chassis/enclosure, which provides ESD immunity and limits emissions.

The most straightforward solution for your application might be to either switch to metal enclosures and use metallic shielded connectors, or to line a plastic enclosure with metal foil or conductive paint to the same end.

The shield doesn't need to be grounded to the PCB specifically, but it needs to join all connector shields/grounds together; most PC connectors are shielded already so this is a given, but the power cable may need grounding whether through the board or through an inlet filter or etc. Likely, the PCB has screw holes which are [at least RF-]grounded, and can be used for EMI shielding purposes.

The shield doesn't need to be all-around, but preferably should have few seams or openings, and mostly box or cage the board and connectors.

Alternately, you might consider USB isolators, so that the cables aren't conducting noise into the PCB (and across the unshielded length where it gets into the signal). This may be a safer approach when considering circumstances like lightning induced surge (affects long or buried cables) or freak incidents like downed power lines. I suppose if the cables are USB, they can't be too long anyway, but the conclusion remains the same whether it's pure USB, or only for some distance up to an interface device and onwards to the end device(s).

The linked connector might be handy for designing a PCB, where integrated protection could save component count and board space; it's not useful inline as the problem is the line itself -- what's known as common-mode interference, which a purely inline device can't do much about (short of a full isolator; and even then, a low-capacitance type may be required).

Note that an otherwise unassuming environment need not be particularly quiet. Interference is carried on power lines, from switching contacts, static sparks, mobile transmitters*, etc. Indoor sources can be as simple as moving around people or equipment, a poorly wired flame igniter, light switches, etc.

*CB radio isn't nearly as popular these days as it once was, but there are still operators out there running at illegal power levels, which can inject voice modulation into lots of things as they drive by.

Answer 2

Prior to connecting the drain to ground I had a much higher rate of dropouts.

This tends to confirm it's EMI being picked up on the USB lead. Current generated by EMI in the device<->SBC connection creates a common-mode voltage at the USB interfaces that disrupts their operation.

Before you go for shielded enclosures, try improving the cable grounding further, to reduce this voltage.

Make sure the drain of the USB cable is

• connected by as short a wire as possible
• to a ground on the SBC that is as close to the USB interface chip as possible (granted this may not be a convenient ground to find or connect to)
• and the same at the device end

If you've done this and are still getting dropouts, then you need a shielded enclosure.

The most important part of the enclosure is the point where the USB cable comes through the shield. If you simply pass the cable through, then the shield is not only totally compromised, but might actually make things worse. The USB cable shield must be electrically connected to the shield at the point it comes through. One of the benefits of a shield is that it makes an almost zero connection distance for this grounding possible, using a grounded shielded connector. If you don't want to use a connector, then remove a piece of the outer insulation from the USB cable to make contact with its shield, and connect that to the enclosure with as short a connection as you can use, ideally clamping it to the enclosure metal. Once inside the shield, you don't need to sweat the SBC connection to the shield so much.