Keystone USB 3.0 A Female To Female Jack Module Adapter Converter Power Socket (WHL #13)
My new desk has some extension cables running underneath to provide data and also power conveniently on one end. I decided to use a 19″ 1U “KeyStone” patch panel because I like the variety and modular nature of these adapters. Well, things like HDMI (or even this DisplayPort voodoo!) are hard to find in China, and 4mm banana plug keystones and audio jacks are basically the same price over here. But USB seems very common – so why not buy an USB 3.0 keystone from the Wan Hung Lo factory?
So I did. This arrived not too long ago:
The front dimensions are standardized to 15x17mm. Depth is not standardized, and this adapter clocks in at 37mm. Most (3.0) gender changers are a tad longer than that. Great, so this unit fits nicely into the 19″ panel. Once. I will explain in a minute, but I had to remove it again for testing, and the first thing to come off was one of the clips that usually hold the unit in place. So I was back into the “could have gotten a regular gender changer instead” world. Meh. Still, at 1.08€ unit price on eBay, not much was lost.
Why did I have to remove it? I copied a lot of stuff to a Sandisk Cruzer Extreme 128GB to test this unit. But I also really wanted to transfer these some tens of gigabytes from my desktop PC (with onboard USB 3.0) to my laptop, and using the USB memory stick with at least 150 MB/s transfer rate is much faster than sending it via Gigabit Ethernet to the router, which uses 802.11n 3×3 (450 Mbit/s) to beam it down to the laptop. But this transfer took ages. 35 MB/s the file dialogue said. 35 MB/s…that’s proper USB 2.0 speed. Hmmm…
Alright, bitch. Are you USB 3.0 compatible or not? First, testing the memory stick directly on the back I/O panel of the computer:
160 MB/s+. I maybe have to add that this device has some troubles with internal garbage collection and gets faster (back to factory speed) when manually writing zeroes to it. After a while of regular use, performance drops (and I’m thinking about RMA’ing because of that). So do not be puzzled why it gets faster when run at the end of 3 meters of Wan Hung Lo cable plus an old 3.0 gender changer. Each of the following AS SSD screens were taken after a full null-write = clear of the entire 128 GB:
160 MB/s, bingo. So, what about the keystone?
Mmmh…nope. USB 2.0 speeds. But maybe let’s repeat. Disconnecting from the keystone, attaching to old gender changer, wipe the thing, bench again:
Back to 160 MB/s. Wiping, then using the keystone again:
USB2. Am I doing anything wrong? Old unit:
Yep, 160 MB/s.
So – this USB 3.0 keystone does not work as a keystone module after a single unmounting cycle, and also does not provide USB 3.0 speeds. Guess what…I got my money back almost instantly. These guys have an idea what they’re selling.
But I was curious: All connections are there, my multimeter says below 1Ω for every pin, wiring of the differential pairs seems alright, what’s the problem?
Cracking it open produces a lot of chippings. The plastic is extremely brittle and cheap. Someone really brought down the BOM at the Wan Hung Lo factory for this item. Looks nice, bricks as soon as you touch it. After killing the ridges on both sides, the two end caps fell straight off.
Well, I now have a copy of the fingerprints of some poor girl at a Chinese factory. Construction of the adapter is extremely simple, one PCB (“SL-75”), two USB 3.0 connectors, tiny amounts of solder (the through-hole parts are not even fully covered). Done. Who needs to clean the solder paste residue on a high frequency PCB…
Having a closer look at the traces yields path length differences. Which lead to signal delay (phase) differences. Doing some measurements with GIMP, I reckon that the differential pair that is not routed per via has some 6mm and 20mm path length from soldering point to soldering point. Which is a delta of 14mm in this pair. I didn’t bother to take apart everything to get measurements of the other one. It’s probably better, but what does it save if the other pair already kills it…
Let’s have a look at official USB specs: http://www.usb.org/developers/docs/devclass_docs/CabConn_3_0_Compliance_Document_v1.02_2011-10-04.pdf
Case B-4-2: D+/D- pair Propagation Delay: Test EIA 364-103: 200 ps (10%-90%) rise time: 16 ns maximum, 10 ns maximum for Micro Series
This one basically is the reason for the inofficial 3m cable length maximum. They use 5 ns/m as reference delay, which is 200’000 km/s or 2/3 the speed of light. That’s a very typical figure for signals on copper cables and anyone with an oscilloscope and a few meters of coax can verify that. Set an arbitrary maximum delay of 16 ns between USB host and device, and there’s your 3m cable length. USB 2.0 had 25 ns, which translates to 5 meters of cable. USB 3.0 with micro connectors is at 2m due to the lower 10ns spec. I am using a 3m cable, which is on the edge of the official spec. If anything goes wrong, one should see lower speeds or total failure of USB 3 mode. Well, we just did…
Now, the more important one:
Case B-4-3: D+/D-pair Propagation Delay Skew: Test EI 364-103: 200 ps (10%-90%) rise time: D+/D- lines: 100 ps maximum
Propagation delay 100ps maximum on each differential pair. What’s 100ps at 2/3c? 20 millimeters. Well, this piece of shit already grabs 14mm of this because of some lazy and/or badly instructed guy with native 21:9 vision screwed up the PCB layout. 6 more millimeters in the chip->board layout->cable->attached device board->chip chain and we’re out of spec. And a few millimeters are easily obtained along 3 meters of cable, especially if it is made in the same factory.
But, as proven with the gender changer, it is not the fault of the cable, that is, it does not add tons of phase difference as well. In fact, another 50cm cable also does not work with the keystone but does with the old adapter. Maybe it is screwed up in the other direction and they cancel out
So apart from the phase difference, it also might be the violation of the impedance spec:
Case B-4-4: Differential impedance (SS) of Mated Connectors: Test EI 364-108: 50 ps (20%-80%) rise time: 75 Ω minimum, 105 Ω maximum
With this trace routing, I assume “impedance control” is just an unknown term to the designer. I tried incorporating that in my USB measuring adapter (WHL #4) for USB 1 or 2.0, but not knowing about it for USB 3.0 is probably just deadly for any signal on these pairs. I lack the testing equipment to verify that, but anyone with an Agilent 86100B with the modules listed in the USB spec sheet (low five-figure price tag?) is very welcome to contact me
To sum it all up: With that kind of PCB routing, it is a great standalone USB 2.0 gender changer. With massive design changes, it could be a keystone USB 3.0 adapter one day…