We're back and more shielded than ever

Happy 2022 y'all! We've had a busy start to the year, with a LOT of time spent researching (see: destroying) some samples to get us in a good place on new projects... and oh boy, does Kyle have a story for you!

If you haven't seen our recent LTT videos featuring our new cable tester, they are definitely recommended (though not required) viewing to boost your understanding of some of the concepts outlined in this update.

https://youtu.be/u6lx1ntNoxE
https://youtu.be/XFbJD6RE4EY
https://youtu.be/Vn2vdQZhs0w

Kyle Tharratt
Engineering Manager whose shields are well-terminated

It’s no secret that lttstore is planning to sell cables at some point. Our goal is to provide good quality cables that you can trust to “just work” and have easily understandable capabilities, freeing you from having to navigate ridiculous naming conventions and ever-changing specs – I’m looking at you, USB 3.2 gen 2x2 and HDMI 2.1.

And there’s no better place to start than the do-it-all cable, the USB type-C 20Gbps with 100W Power Delivery (PD).  We ordered a bunch from the most recommended factories (or at least the most recommended factories who would deal with us without any sort of existing large order quantity), ran them though our cable tester, and… EVERY SINGLE CABLE FAILED SIGNAL INTEGRITY??

Some failed so badly that they couldn’t even do 5Gbps! What a fantastic start…

At this point, there was only one thing to do… I pulled out a knife and a Dremel, and I went to town cracking open the housing on each cable.  Here’s a look under the hood on one of the failed units:

It’s a PCB with a SMT USB type C connecter on one end, an IC (called an emarker, mainly for PD negotiation), some MLCC caps, and the cable's conductors are soldered to the other end of the PCB.  If you’ve got some experience with high-speed signals, you may already see exactly what’s happening here…

For the uninitiated let’s take a quick look at the USB Type C cable spec - which is 373 pages long. Don’t worry, we only need one… page 75:

Note that SDP stands for Shielded Differential Pair and there are 4 pairs in total that form the high-speed lanes. All of them failed signal integrity testing. All failing cables used a shielded twisted pair that had no drain wire, and NONE of the 5 cables’ shields were terminated to ground… NONE. Shields aren’t wireless people… this is like putting on an un-grounded ESD band and expecting the static to be magically absorbed.

So let’s find a perfectly working cable that passes all signal testing and destroy it - for science, obviously - to see what it looks like inside:

Breaking away the outer housing, we get a stamped sheet metal shield around the PCB - nice, none of the other cables had that! Removing this we get some over-molded plastic (gotta have that plastic protection baby) and then finally we get the PCB WITH FREAKIN TERMINATED SHIELDs (red arrows)!!! To further drive the point home, I cut open the cable and found one of the twisted pair high-speed lanes.  Guess what… it had a drain wire with the twisted pair and all the drain wires are terminated correctly as well.

Its almost like following the spec actually makes a cable work… who would’ve thought. 

Now armed with this knowledge, would I be able to make the bad cables work by soldering the shields correctly with a soldering iron, some wire, and a can-do attitude? Short answer…

Nope.

I soldered the outer cable shield to the connector housing, and lo and behold I got some previously failed parts of the test to pass - namely resistance tests.  Signal integrity on the high-speed lines still sucked, probably because each of those lines have their own shield and, of course, it was also left un-terminated.

Trying to solder the high-speed lines didn’t really work as the flimsy shield just shriveled and melted. (I also accidently severed 2 of the lanes with my knife and didn’t feel like spending more time redoing this on a different cable :P )

I am guessing here, but it probably takes at least 3-5 extra manufacturing steps as well as extra material cost for twisted pairs with drain wire to make the cable that worked. And I suppose some manufacturers just choose not to do it to save a buck.

Rest assured, we will find working cables for the store. But in the meantime we might have another project for the LTT team to take on with our cable tester.

Let us know on Twitter @Linustech - or on the LTT forum - if you wanna see some further investigation on this topic!

As always, we're hard at work on projects like this. If you want to show your support to these efforts, we just added a classic Canadian-designed Plaid Flannel to the store, and we've got some restocks in popular styles like the WAN Hoodie! Keep an eye out for more updates coming soon.

Nick
COO
lttstore.com / Creator Warehouse
@nicklmg