High Speed 4 Ports LED USB 3.0 HUB With On/Off Switch Power Adapter USB Hub For Desktop Laptop and the CAT S61 (WHL #58)
This post is filed under “old, forgotten crap that has issues with recently added hardware”.
Well, this might be the new record holder for ancient reviews – but only due to recent events. I bought this USB 3.0 hub for 3.59€ with free shipping back in July 2017 (pricing hasn’t changed all that much). It was a purchase following up the 3+1 Gigabit/LAN adapter described in WHL #34. I still run this thing to support four USB3 keystone modules that are roughly 2.5m away from the mainboard. That’s not only difficult for the USB signalling at the highest speed setting, but also a strain when running high power devices such as 2.5″ hard drives or a mobile phone that needs a bit of juice. Like the CAT S61, a recent upgrade (and downgrade!) at work due to ongoing problems with the CAT S60 (FLIR and battery issues). Adding an USB hub right before the keystone fanout buffers the signals, allows me to run only one cable instead of four, and also allows for an easier way to supply 5V where it is needed. The latter part didn’t happen until a week ago, when I received a PoE 5V splitter, which might be featured in another review.
Then there’s this slight issue, independent of an external power supply: Charging current with the CAT S61 isn’t exactly great. To be more precise, it is non-existent, as the internal power draw when trying to charge is HIGHER as it is allowed to slurp from USB, which is 100mA (good thing it does honour this limit). The phone actually slowly discharges when charging this way:
Reason enough to rip it open and have a look, right? Well, I had fun with it before deciding to make a blog post out of it, so this is the fully “fixed” unit that is currently in use. I promise new units will not only carry four annoyingly blue LEDs that indicate the switch status – they will also be populated with four switches
Not sure why the “4” needs to angled in the logo print. The ports aren’t at an angle, the only thing not at an 90° angle is the LED indicator light guide?
It comes apart pretty easily be removing the bottom piece. After that, the single-part light diffuser and the PCB will fall out. I might have lost the screws a long time ago, but I don’t think there are any. It’s just held together with clips and a simple metal pry tool will make it come apart.
This is the PCB front with one switch removed, plus the custom cable to supply the thing via MicroUSB from the PoE splitter to the tiny barrel jack (~3.5/1.3mm). The layout is mediocre – the chip would allow four power and four status indicator lights plus individual power switching, but none is used. Instead, the power to both the LEDs and the USB socket is cut via switch and the chip has no way to hard-reset devices. The former likely would have been possible with the smaller version of the controller chip, see below. The latter is also supported by one of the smaller versions. I would argue I found a blind via during tracing several pins, which indicates 4+ layers and must be more expensive than a full via. Maybe this is a manufacturing defect on top of all Wan Hung Lo cheapness?
Now the more interesting back side. The chip used in this one is a Genesys Logic GL3520-22, which comes in three different flavours: A physically bigger QFN88 version that supports everything, and two QFN64 versions, one with programmable ganging option for the ports (“OS8”), and one with fixed ganged USB power switching that affects all ports at once (“OS3”). Both of the smaller versions do not have enough pins for the Christmas lights LED show, so they should have been sufficient for what this USB hub is designed to do. I can only assume there’s something inside that requires a significant portion of wafer space that allows for a commercially viable third chip option for an item that is sold for less than 4€. Crazy.
Anyway, this hub uses the big QFN88 chip for no good reason. Quick overview of pin usage:
QFN88 |OS3 | OS8 USB3 RX/TX: 20 | 20 | 20 (4 Ports plus Upstream) USB2: 10 | 10 | 10 PWRGREEN: 4 | 0 | 0 PWRAMBER: 4 | 0 | 0 PWRENABLE: 4 | 1 | 4 OVERCURR: 4 | 1 | 4 PGANG: 1 | 1 | 1 (why?) PSELF: 1 | 1 | 1 (self-powering or external) CLK: 2 | 2 | 2 RESET: 1 | 1 | 1 SPI: 4 | 4 | 4 1V2 (a): 9 | 5 | 5 1V2 (d): 4 | 3 | 2 3V3 (a): 3 | 3 | 3 3V3 (d): 2 | 2 | 1 GND: 6 | 0 | 0 (WTF?) VBUS: 1 | 1 | 1 5V DCDC: 1 | 1 | 1 (internal 3V3 generation; 1V2 LDO however is an outside part) 5V: 1 | 1 | 1 RTERM: 1 | 1 | 1 (single 680Ω termination resistor?) TEST: 1 | 0 | 1 (enables some test mode) NC: 4 | 6 | 1
Why on earth would you skimp on features or power pins and then have more NC pins on the QFN64 than on the QFN88? Also how can you not use GND pins?! Even the design guide for this very chip doesn’t tell. I guess one needs more power pins when driving eight additional LEDs…but zero GND pins seem a bit…few, don’t you think? They do not mention a body contact beneath the chip, but the drawing shows additional dimensions for a thing centered inside the chip – maybe that is how they do it?
That aside, I did four modifications to the board:
1) Removed the power switch to fully isolate the USB power from any chip interference. Power isn’t routed through the chip, but there’s four external overcurrent shunts that carry power to their respective port. By removing the switch and wiring directly from the barrel jack pin to the power pin of the USB connector, nothing can stop a stable 5V supply. There’s also no visible low side switching MOSFETs, so the return path is the entire GND plane.
2) Removed the power indicator LEDs for stealthy living room operation. Those bastards basically double the idle power draw.
3) Bridged pins 1+2 of the unpopulated U5 chip on the bottom right. Pin 2 of U5 connects to PSELF on the main chip which decides if an external power supply is used or the thing is fed via USB. I’m not entirely sure if that makes a difference without this very chip present, but if that suppresses backfeeding the 5V of the external supply back into the computer, that’s a solder joint worth making. Pin 1 just happened to be 3V3, so I hope this is the external “high” power level (instead of 1V2). With that link active, the idle voltage rises from around 4.7V fed over the lengthy USB cable to 5.1V from the PoE splitter in close proximity. Voltage under load also drastically improves.
4) Bridged the USB2 data pins on port 4. There was nothing to be done in terms on configuration that would enable the CAT S61 to reliably charge above 500mA or even charge faster than that. I think 2.1A is possible on the phone and 1.5A via controller signalling. When plugging it in before supplying the hub itself, it would support 500mA, otherwise it just uses 100mA as shown above. This however isn’t convenient to use in my configuration. The S61 does its own load estimation thing when D+/D- is bridged and regulates down so that voltage never drops below ~4.7V. This can be nicely demonstrated by using the USB A-to-Micro bridge back from WHL #4. The GL3520 has a software option to make a dedicated charging port with increased wattage – also at the expense of the data lines, so the entire port is useless for data transfer. So I made this a hardware option instead of a software one. If anyone can provide the configuration utility shown in the design guide on page 13, please let me know.
Finally put everything together, installed it back under the table, and charging current is now around 1A:
That’s still below the capabilities of phone, hub and external power source, but it is charging at a suitable rate and it eliminates the need for a regular charger in that area, plus it is always on. I’ll mark that one port as “power only”, three regular 3.0 ports will do the job just fine.
Would I recommend this USB hub? Well, it’s dirt cheap, so if that is a requirement, sure, go ahead. Same if you need the hardware switches (there’s also a 7 port model available). In any other case – this chip is at least 4 years old now and has been surpassed by the GL3521 and GL3523 from Genesys and literally dozens of other manufacturers. If you don’t mind spending a couple of bucks for a more compatible product that can charge a phone properly – please do so. Also, USB 3.2 Gen2 and Gen2x2 (fuck all the fucks that are responsible for that name!) are a thing now, so if you need the additional bandwidth, those more expensive hubs might be your best option.
[…] chip available that day in Shenzhen. The GL3520 comes to mind as it was used in the USB hub on WHL #58, however, USB hub and host chips are different. On the other hand, from the routing on the PCB I […]
Low end microchip processor used in after market USB hub could be easily hacked by cyber criminals and spied on your computer or your passwords. Or it could be hacked, compromised by the company that sold cheap processors. Stay away from low end PC accessories.
I very much appreciate the security concerns but this is a dedicated USB controller chip from a manufacturer that does produce these things on a regular basis. These aren’t microcontrollers or FPGAs that could indeed be easily modified to do additional things, but would perform poorly as a drop-in replacement. Regular µCs do not have the throughput and/or I/O options to perform as a shady USB 3 controller, and FPGAs likely will have a substantial power draw doing so.
I think the total number of individuals on this planet worthy (as a target) of custom silicon that performs as USB controller with additional spy features is really, really small – and those people use tested, certified, probably hardened hard- and software only and certainly not a $5 USB hub they found extra-cheap on eBay.