USB 2.0 4-Port Aluminum Hubs … and Chinese manual labour (WHL #20)
For today’s blog post I have something that I found at work yesterday. It’s a plain old USB 2.0 hub in an aluminium case, which is used inside of our measuring equipment. Just a jellybean part, it’s sold as Logilink here but I bet other brands just slap on a different logo on the front and they’re good to go. We chose this because it is relatively small, has all USB jacks flat in line (the smaller ones have 2×2 configuration or tend to stack them), and can make use of external power supplies, rather than sucking too much from the uplink port.
I was told these hubs fail after 2-3 years, so they get replaced in advance whenever a device is back for service. This time, the replacement didn’t really fit, because even though order number, EAN, packaging and everything was the same, the hub has gotten a different external power jack. For replacing the thing, we now have to put in a different (yet more common) power cable.
As we had an issue with USB regardless of which type of hub was used, I had a closer look and I also decided to take apart the old hub for fun. These things are highly integrated and cost-optimized…but something caught my attention.
Here’s the top view: (sorry, these images are all taken with the phone I had at hand, no Canon goodness this time!)
Awesome. The five USB plugs and the power receptacle aside, this thing uses one 28 pin SSOP main chip (“FE 1.1s USB 2.0 HUB LD3A040B12146C”), one crystal to match USB frequencies via internal PLL circuitry, one green status LED, 3 electrolytic 100µF/10V caps for general voltage stabilization, 3 ceramic caps for more decoupling whizzbang, two resistors (one for the LED…) and a diode for crude reverse polarity protection. That’s it. That’s the whole USB hub. Dude!
But: Have a look at the pins 17 to 19 at the lower left side. They have been reworked, someone put an iron on there and shorted three pins. While there is actually a product brief available on that chip (from a company called Terminus Technology Inc. in Taipei, Taiwan), it doesn’t unfortunately tell pinouts of the chip (or its 48 pin QFP or LQFN brothers).
However, given this is a 4 port device and has 28 pins, 5x 2 data pins are already gone (pin 4-11 and 15-16, that’s easy to spot). From the remaining 18, we need two for the crystal, four for individual port activity LEDs and one for the hub status (power) LED. According to another data sheet, these LED ports are even configurable so that you can exchange them for a resistor between the data lines, mimicking a USB charger per port. So I guess they are not 1 pin each, but two…which means we’re down at 18-2-8-1 = 7 pins reserved for power and everything else.
One more pin is some current monitor sensing thing, but I’m not sure how that is implemented on the board. I mean, I’d expect some shunt resistor or the like if it’s an external thing and not fully integrated…but the LED dropper resistor aside, the only resistor implemented is the “272” 2.7kΩ fly shit, which cannot work that way. Still, that’s one more pin used from the chip – bringing it down to a maximum of 6 available. BTW: Have a look at this nice application note from Microchip about ganged power port control for USB: AN2230
And with that 6 pins, the device powers itself and also all of the 4 ports, which would mean 2000 mA port consumption + 115 mA juice for itself. Even with 6 of those tiny legs, I’d check first if that is acceptable in terms of voltage drop and thermal load. And this is probably why they botched three legs together…
…but why the hell did they do that for both boards? They are clearly marked as revision D and E – why would you keep the botch over multiple revisions if you even rearrange the LED resistor and change some of the power PCB traces? Make your friggen traces bigger and join earlier if you need to, don’t let the Wan Hung Lo factory worker put a blob of solder on EVERY single unit! Once again, how cheap must Chinese manual labour be if you can do such awful PCB layouts for devices that are produced 100’000 times? Geez…
As for the bottom side of the board: Have a look yourself. (rev D is top, E is bottom just like shown above)
Well, at least now they’re soldering both mechanical tabs from the USB plugs. In rev D, they only had solder for one of them, and you can even see some solder on the other pad where the Wan Hung Lo worker accidentally touched the metal when moving from one block to the other. If solder for individual solder joints is more expensive than labour…
So that’s it for today, I rambled much longer than I expected to, but that’s nothing unusual…however the USB power gating tech doc was a nice insight, and figuring out the pinout was fun, too. Stay curious for the next crap of the week