Effect of cable specifications/length with USB 3.0?


#1

Hi,

I have started accumulating a number of USB 3.0 products (hubs, docks, enclosures, drives, etc.), including, most recently, one of Plugable’s lay-flat docks.

My “routine” when I get a new USB 3.0 device recently has been to run a series of tests, involving some large file transfers, and testing with HDTach (I know that there are newer tools than HDTach, but I just like it, and having used it for a number of years, I can correlate/compare results easier for myself).

Besides the read speeds (burst and average) and access time, HDTach produces a graph of the read speed over time, and one thing that I’ve noticed is that sometimes I see drastically different results in that graph when I use different USB 3.0 cables (either from the target device direct to the PC, or from the target device to a hub, or between a hub and the PC).

Given that you all appear to have quite a lot of experience with USB 3.0 (it’s fairly new for me, as I only got my 1st PC/laptop that has USB 3.0 ports late last year :)), I’m wondering if you’ve noted the impact (or non-impact) of cabling and cable length, with USB 3.0?

For example, as a manufacturer/vendor of USB 3.0 devices, do you consider cabling (length, gauge, etc.) when you prepare your products for the market?

I guess what I’m really wondering is if a customer, like myself, uses an after-market USB 3.0 cable: (1) Is that potentially going to affect the performance of the product and (2) does Plugable support that type of thing?

Thanks,
Jim


#2

Hi Jim,

Thanks for asking!

Yes, we see the same thing – there is significant variation in cable quality for USB 3.0, and cable length magnifies those issues. USB 2.0 doesn’t have these concerns, because it just doesn’t push the cable and silicon as hard.

The main cause is the high frequencies on the wire for USB 3.0’s 5Gbps operation. Small variations can result in reduced throughput or devices which drop from the bus (depending on host controller and device resilience to signal variation and errors).

From a testing perspective, it’s possible to see this with a particular cable (e.g. see eye diagrams http://www.evaluationengineering.com/… for background), but production variations are still common.

As a result, we’re very careful about cable quality and length. We ship 1 meter or shorter cables with all our Plugable brand USB 3.0 products for this reason. This isn’t to say longer cables won’t work, but they will have a higher problem rate, and given all the confusion a bad cable causes, we want to avoid that at all costs.

We don’t currently sell any aftermarket cables (if we could find a production partner who could achieve a measurably higher unit quality level, we would). But if I were buying one for myself, I’d stick with one meter or less to minimize chance of a bad cable keeping things from working. And if I wanted to go father and be paranoid about performance, I would construct a test setup where the differences in cable quality could be measured via throughput (basically as you have), knowing I’d want to “bin” some of the cables, and only keep the best.

Definitely more to think about than with USB 2.0. Hope that background helps. Thanks again for asking!
Bernie


#3

Hi Bernie,

Thanks for the response to a probably strange question :)…

FYI, one of the reasons that prompted the query was that since I started using USB 3.0 stuff (again, fairly recently), I’ve been trying to find some *shorter* cables than than the ones that came with the devices I’ve gotten thus far. The shortest cable with a product was with a Toshiba Canvio drive, and that was about 18". One of the devices I got, a USB 3.0 hub (sorry :)!!) came with a cable that was probably a meter long.

I have a small setup in our kitchen, and trying to figure out what to do with all of those “long” USB 3.0 cables is a real mess :)!

So anyway, I started finding some shorter (less than 12") cables, and, while I’ve found some, thus far, to my surprise, the performance with these shorter cables has been worse than the original cables. By “worse”, what I mean is that I get more dropouts, disconnects, and “file is gone” problems.

I’ve been surprised about this, which is what prompted the post. Some of the cables I’ve tried are pretty heavy gauge, on the order of the cable that came with the lay-flat dock that I got recently, at least from what I can tell externally, and both the cables and the connectors are all marked with the USB 3.0 Superspeed markings.

I’ve even gone so far as to get some clip-on ferrites, and experimented with adding those to the short cables, but those haven’t helped thus far.

Anyway, I appreciate your response!

Thanks,
Jim


#4

Hi Jim,

(Just getting back from watching the game. Wow - thought after the start of the 2nd, it would be boring … no way)

I’m not sure! Because, all things being equal, I wouldn’t expect what you’re seeing.

It could be that the shorter cables are getting binned from lower quality batches by someone. Or, because there’s extra EMI around USB 3.0 connectors, the shorter lengths could be generating more interference (see this paper for more on USB 3.0 EMI:
http://www.intel.com/content/www/us/e… )

But, if it’s just a few datapoints, it might not be something you want to read too much into – could be bad luck. How many very short cables (like on the Toshiba Canvio) do you have?

Hope you can figure it out. And please let us know if we can help in any way - we’ll do anything we can to help, as it relates to our Plugable products.

Thanks!
Bernie


#5

Hi,

I was aware of the EMI thing, but as I understand it, that was more the case of EMI from the USB 3.0 interfering with 2.4 GHz equipment.

I keep wondering if the problem may be something like signal reflection, which might be less of an issue with a longer cable, but more of an issue with a shorter (e.g., 8") cable. There are connectors on both ends of the cable, which I think would cause a change in impedance at each end, which I think is a potential cause.

It’s a kind of strange problem, because, for example, I have one cable, that’s about 18" long, but much smaller gauge that the 7-8" cables that I have, but the 18" cable works a lot better than the shorter ones. I would have thought that the longer, smaller gauge (=higher impedance) cable would have had more problems.

Maybe, with the shorter cables, the signals are too “hot”, i.e., since short cables may have less impedance, the amplitude of the signals may be “too high”.

Anyway, sorry for rambling…

Jim


#6

Thanks! Let us know if you find any conclusive pattern, or if we can help!

Bernie


#7