Connect with us

USB Cable shield continuity

Discussion in 'Electronic Design' started by Paul E. Schoen, Jun 11, 2010.

Scroll to continue with content
  1. I have been having seemingly random noise problems where my PIC-based USB
    device will go into a USBSuspendControl state, and it requires the cable to
    be removed and replaced to reestablish communication. The problem usually
    occurs in the field where there is switching of high current and high
    voltage, but I was able to duplicate it to some extent by running the USB
    cable along an AC power line to a current source which I switched on and
    off.

    Recently I suspected that the USB cable itself might be at fault, because I
    had bought a batch of 100 pieces for $0.69 each (but now about $1.50) from
    www.CableWholesale.com, and several of these were sent to a customer who
    reported problems, while a previous customer with an older cable did not
    seem to experience this very much, and another customer replaced his cable
    with a longer one, and he said his unit was working OK.

    So, I dissected one of the new cables by removing the PVC jacket in the
    middle, and I found a substantial tinned copper braid shield, and an
    aluminized Mylar wrap under that. When I removed the shield I could see that
    the four USB conductors were twisted together, which is generally good for
    noise induced by strong magnetic fields. So far, so good.

    But when I measured continuity from the connector shells to the exposed
    shield, I got intermittent readings of about 3 to 30 ohms and sometimes an
    open circuit. Then I measured the continuity from shell to shell on a couple
    other USB cables I had been using, and I found that one showed an open
    circuit and the others showed intermittent. This was the case for two new
    cables from different sources. Yet I measured the shells of a USB cable for
    my Nikon digital camera (with a mini-USB on the camera end), and I got a
    solid connection of less than 1 ohm.

    I still need to do more testing and I may also purchase a high grade cable
    with gold plated connectors and better shielding. They are about $20. I will
    also have my customer check the continuity and try a better cable. Perhaps a
    USB 3.0 cable will work better.

    I removed the PVC molded cover for the male type "A" connector, and there is
    a metal shell that extends back and tapers to a smaller "neck" where the
    cable is clinched or crimped. By bending the ears on the crimp I was able to
    separate part of the metal housing to reveal where the shield has been
    folded back and exposed so that the inner surface of the housing presses
    against it. But it seems that the jacket of the cable is a continuous
    molding that fills the shell of the connector, and the crimp mechanism can
    only apply light force to the exposed part of the shield. So the actual
    connection may degrade with time as a non-conductive film may form on the
    metal surfaces, and mechanical flexing may further degrade the connection.

    Here are pictures of the cable after dissection and exposure of the crimped
    shield connection:

    http://cygnus.smart.net/~pstech/photos/USB_Cable_23.JPG
    http://cygnus.smart.net/~pstech/photos/USB_Cable_Conn-A_26a.JPG
    http://cygnus.smart.net/~pstech/photos/USB_Cable_Conn-A_27.JPG
    http://cygnus.smart.net/~pstech/photos/USB_Cable_Conn-A_28.JPG
    http://cygnus.smart.net/~pstech/photos/USB_Cable_Conn-A_29.JPG
    http://cygnus.smart.net/~pstech/photos/USB_Cable_Conn-A_30.JPG

    I think this is a design defect and I should be able to get a refund or
    credit for the unused cables. It may not be the reason for the problem but I
    should be able to determine that if my customers replace the cables with
    high performance versions and their problems are greatly reduced.

    Anyone else have experience with this? One member of the Microchip forums
    reported that he found the following with some new cables he had on hand:

    Poundland 1.8 m A-A(F) yellow 8 ohms
    Signalex 1.5 m A-A(F) white OPEN CIRCUIT!
    Signalex 1.5 m A-B white OPEN CIRCUIT!
    CPC 1.8 m A-B translucent yellow 18.5 ohms
    IXIOS 3 m A-A(F) translucent/silver grey gold plated connectors
    <0.5 ohms

    Paul
     
  2. baron

    baron Guest

    Paul E. Schoen Inscribed thus:
    Hi Paul,

    The figures above don't cause any surprise at all !

    Part of the problem with USB cables is that the shield is not needed for
    the cable to work but is vital to prevent interference effecting the
    signals on the cable and by the same token preventing the cable
    radiating and causing interference to other things.

    I built a test rig that checks continuity of all four wires and the
    screen. I used a 100ma test current from a 12 volt source. I found
    that most of the cables tested showed 10 ohms or more between the ends
    of the shells and I found a number with varying resistance on the
    individual conductors of a similar order.

    I also found that some receptacles showed varying degrees of contact
    resistance with increasing usage. The type "A" being the most
    affected, type "B" being more robust.
     
  3. The overall resistance of the four conductors should be much more consistent
    if they are formed with a proper metal-to-metal crimp. Solder or tack weld
    would be even better but would be very costly. Here is the USB spec:

    Universal Serial Bus Specification

    6.6.3 Electrical Characteristics
    ...
    The DC resistance from plug shell to plug shell (or end of integrated
    cable) must be less than 0.6 ohms.

    http://alsa.cybermirror.org/manuals/usb/usbcableecn_final.pdf

    The cables may work, but they do not meet specification. Your test at 100 mA
    is better than a simple ohmmeter check, but lower voltage might show poor
    connections better.

    I'm wondering if I could charge up a large capacitor and then discharge it
    through the cable shield so that a spot weld might occur at the junction of
    shield and shell. But it probably would not be very reliable long-term under
    normal use.

    Another "fix" might be to expose the shield near the connector and add a
    jumper to the shell, but that is ugly, time-intensive, and just plain wrong
    except to test if shield ground integrity improves the noise problem.

    I found some good deals on gold plated USB cables at www.monoprice.com and
    www.TriangleCables.com. Even if the same crimp was used, the gold plating
    might reduce the chance for oxides or other coatings to form at the
    junction. But the shield braid would also need plating.

    The manufacturer might be able to apply a drop of conductive paste at the
    junction point that would provide a decent connection and also exclude
    moisture and other contaminents. But there may be no reliable way to fix
    these cables as they are.

    Paul
     
  4. IanM

    IanM Guest

    If the braid is pigtailed and crimped into the housing separately from
    the overall strain relief crimp there would be no problem.
    Unfortunately that would require an extra step on the assembly line.

    Sturgeon's Law applies . . . .
     
  5. Baron

    Baron Guest

    Paul E. Schoen Inscribed thus:
    Grief ! I never saw one less than an ohm.
    I measured the voltage drop across the cable. (1.5mt nominal length)
    You might find that it just blows the bad joint open.
    I agree, they can't be fixed. All the cables that I built the tester
    for were of doubtful far east origin. The distributer had a financial
    agreement with what ever supplier, because the cables that were
    rejected didn't go back to the east.
     
  6. Baron

    Baron Guest

    Paul E. Schoen Inscribed thus:
    I forgot to mention, a temporary fix was to give the bit of plastic,
    where the cable enters the plug, a tap with a hammer. It doesn't last
    long though, a few flexes and its as bad again.
     
  7. Interesting, I have the same problem with a PIC project. Right now I'm
    moving to the latest Mchip USB stack.
    I did find a well placed Ferrite bead suppressed the problem, but it
    still occurs once in a while.
    Eventually I plan on programmatically Detaching the device and
    re-enumerating, after a set period of inactivity.

    This problem does not occur with other USB devices in the system.

    Cheers
     
  8. I found that I had to put a jumper from an unused pin (24 = RB3) on the
    PIC18F2550 to the D+ line of the USB connector to effect a reliable detach
    and reattach. When the device is stuck in the USBSuspendContol state, I use
    a countdown which eventually disables interrupts, changes the pin from an
    input to an output and drives it low to force a disconnect. Then I do a
    SoftDetach(), run a delay loop for about 5 seconds, and then a program
    reset.

    However, this causes a problem on a new installation where the driver must
    be installed. I found this from a customer. I already had the driver
    installed, so it enumerated before the timeout, but without the driver it
    went into a loop which prevented it from being installed. There may be a way
    to pre-install a driver, but that could be tricky.

    I switched to the new stack (2.6) recently, and found several changes needed
    to be made. There also seems to be a bug in the code if you are using a
    PIC2450 as I have on one of my older boards.

    I have two versions of PIC code, one for a CDC and another that uses the
    generic mchpusb.sys driver. There is supposedly a bug in the Microsoft
    supplied usbser.sys driver, but I have found that both work equally well (or
    not well), and the errors seem to be random. These are comm errors where the
    USB port seems to stop receiving data for a second or longer, or where
    characters are skipped. The SuspendControl lockup is a separate problem
    which seems noise related.

    Paul
     
  9. I'm using the 4550, as a HID device. A hard reset is not an option for
    me. But the USBSuspendControl state is an annoyance. Have you tried a
    pull down resistor on D+ ? I could detach my device and reattach it,
    since I don't need a driver with the HID model and my application will
    happily ignore it.

    I have a early USB stack, and the 2.7 seems to be really finicky. I just
    got it to enumerate, but that's all it does at the moment. Have yet to
    debug it, but I haven't seen any comm errors while using the HID.

    Yes it is noise related, I can reproduce it by discharging a hand held
    Tesla coil to earth ground near by ;)

    Cheers
     
  10. I tried some experiments with decoupling (2.2uf tant) the Vusb output
    pin. Some improvement which leads me to believe there is a noise related
    problem with the internal USB core. ( I don’t use the Vusb pin, and
    apparently the USB core runs on 3.3v ) Also I see that there is another
    lockup beside the SuspendControl, I haven't dug deeper since I was
    successful in detaching and Attaching to the USB bus. But this is a
    bandaid fix, and I'm sure most will find this not acceptable, but in my
    system its acceptable.

    I was able to detach by setting UCON=0 and UIEF=0 ,waiting 1sec and
    reattaching by setting USBEN=1 and UPUEN=1 and initializing the stack.
    I left my version 2.1 stack in the code, since 2.7 required a little
    more work and we don't have any other problems with it. I added a com
    timeout to handle the unknown lockup, since it is not related to the
    SuspendControl lockup.
    Now everything happily recovers, without a hard reset, when that 120vac
    relay switches in the system.


    Cheers
     
  11. JosephKK

    JosephKK Guest

    That sounds like a foil shield instead of braid.
     
  12. Baron

    Baron Guest

    JosephKK Inscribed thus:
    I belive most of them probably are. The only ones that I saw that were
    braided had clear jackets.
     
  13. The cables I had bought for $0.69 each from www.cableswholesale.com had an
    inner foil shield and a substantial outer shield of tinned copper braid.
    Most other cables at least appeared to have a braided shield. I don't think
    it would be possible to meet the USB spec of 0.6 ohms with just a foil
    shield. I am returning 80 pieces of the 100 I had bought that are still
    unopened and unused.

    I also contacted www.monoprice.com because they have very good prices, and
    some with gold plating, but my technical questions have remained unanswered
    even though several people have responded. I asked for a free sample (list
    price is only about $1), but nothing yet, although they all say they will
    issue a refund if not satisfied. But I'd have to take the time to place an
    order, pay maybe $5 S$H, then test it, and then pack it up and spend a
    couple dollars more to ship it back to get my $1 refund.

    I found www.trianglecables.com, and they say that their cables are custom
    made to their specs. They have some that have ferrite filters built into the
    cable, so I ordered an assortment of six cables with and without (6ft and
    10ft, black, white, clear) for a total of about $25 including shipping. I
    should get the shipment in a couple days and I'll report back with findings.

    It seems that the USB3.0 cables have a larger type B connector so I can't
    use them.

    Paul

    Paul
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-