Subtle Problems in Multilevel USB Hubs

Home / Blogs / Subtle Problems in Multilevel USB Hubs

A project recently got me through tough times.  We have connected almost all instruments using USB and ended up using multilevel USB hubs.  The hardware connection is something like below.  We thought that everything will be fine because

  • All of these USB ports are detected by Windows without any issues
  • The payload of these instruments’ data are very low and
  • We’re able to test all these instruments in modular level.

However we faced some subtle issues which is not lucid at the beginning.

Fig1. Initial hardware connection with all instruments connected through USB hubs.
Fig1. Initial hardware connection with all instruments connected through USB hubs.

The project in itself is about a multi-site product calibration.  The golden values from master instrument looked perfect, however the calibrated product looked pretty off from the golden device values.  On top of that, the parameter being calibrated is inconsistent over time, i.e., the timing of the calibration played important role here.

Being unknown about the root cause of this issue, we started to narrow down the issues in several ways.  One of the narrow down method also included the verification of these USB instruments one by one.  The method is to run each sites by literally disconnecting USB Instruments wire for other untested site.  Then running 2 sites in parallel, then 4 sites, then 6 sites.  All by literally wiring only those instruments required for the sites being tested.  Until 6 sites the calibration results were perfect.  Nevertheless, going beyond 6 sites (i.e., 7 sites or more) started to disturb all other sites’ result.  On the other hand, the calibration results for all sites (7 and more sites) were perfect when we connected few instruments to PC directly (not through USB Hub) as shown below.

Fig2. Directly connecting few USB instruments to PC solved the issue
Fig2. Directly connecting few USB instruments to PC solved the issue

Conclusion?

We couldn’t arrive at the source of this issue yet, but the strong assumption is that when we connect more instruments via a USB hub, then there could be subtle overheads even when the payload of these connections are very low that will disturb the results.

Ajay is a professional developer and architect of LabVIEW applications with extreme interest in getting the hardware connected to LabVIEW and automate the stuffs.  He is also a good mentor for the various interns in his career.  He is ready to help the people in techie roles.

2 Comments

  • Tom V G

    We had the same issue with only a setup of a pc with 2 USB-hubs (10 ports).
    First we did think the problem was the use of non-intelligent devices, but even the connection of a spectrum analyzer was unreliable, we did some testing but we didn’t find the real cause, also because at that time it was easier, possible and much faster to connect the critical devices trough ethernet

Leave a Reply

Your email address will not be published. Required fields are marked *

%d bloggers like this: