If you’re still running an age-old dinosaur system like me (well, make that around six years old), then you’re probably feeling the pinch from not having quite enough USB 3.0 ports. Maybe you have none at all. I’m lucky that the motherboard I have has two first-generation NEC chipset based ports, but that’s slowly becoming a bottleneck now that I have more USB 3.0 devices than USB 2.0, and when I’m trying to make the most of USB 3.0 SSDs. As a result, installing a second USB 3.0 host controller where the bandwidth is not shared between controllers will make life a lot more bearable.
Such add-in boards are normally quite simple products, and can be had quite cheaply from eBay. They’re hardly sophisticated when practically all of them are based around one-chip style solutions, often with manufacturers providing reference designs which can be copied to bring a product to the market quite rapidly.
A quick peruse on eBay shows a multitude of boards available – two ports, four ports, five ports, from a variety of chipsets, such as VLI, Asmedia, Renesas/NEC, (or less common Etron Technology and Fresco Logic). None of the boards so much as bother with a brand name or any packaging, but that shouldn’t really matter.
Ultimately, I decided to opt for a Renesas/NEC based four-port controller, as I have known it for having some of the best compatibility when it comes to peripherals from the USB 2.0 days and it seems that this trait has definitely carried over to USB 3.0 as well. Having more ports makes much better use of the I/O slot itself and reduces the chance I would need a hub in the future. I was quite happy that the whole thing cost me just AU$10.35 shipped.
As expected, the product does not have any real branding or retail packaging to speak of. It came wrapped in an anti-static bag, as a full-profile card (as expected from 4-port cards) with the driver CD wrapped in with the card (which seems quite wrong as the cellophane the CD is wrapped in might generate static). The external writing and label suggest this product has a code of TB01200. A search for that seems to show Shenzhen YK-S Technology‘s store has this product, also sold via Newegg.
The card itself is branded STAR-UPCIE201-V1.0, but has nothing to do with Startech as far as I can tell. The board seems to have nice solid capacitors for each of the ports, and also has the specified Renesas uPD720201 chipset. It seems to have two switching converters on the board, likely one to provide the 5v necessary for the ports and another to provide a lower voltage possibly for the ICs.
Rather concerningly, many of the footprints are left unoccupied and others almost deliberately mis-populated. For one, F1 through F5 appear to be positions for polyfuse protection to protect against USB port overcurrent events by essentially acting as a self-resetting fuse. Instead of populating polyfuses, they “shorted” it over using 0 ohm resistors. Another line of defense, this time against EMI noise, is the ferrite beads which should be mounted on the power lines near the ports – FB12 through FB15 which appear to be substituted for just regular o ohm resistors again. Contrast with FB11 where a real ferrite bead was installed, possibly because it was necessary.
U3 through U6 DIP packages may have been current sense amplifiers for each of the ports, all of which are unoccupied. Without overcurrent protection or current sensing, it seems each of the current sensing lines may be fed dummy signals from a voltage divider arrangement possibly (e.g. R14, R18).
Another interesting change is that D7 and D12 near the switching converters which appear to be reverse polarity protection diodes have also been swapped out with 0 ohm resistors.
I couldn’t find any information about the IC labelled U14 – AAW3ZC. I presumed it was a switching converter, but the inductor looks too small to work with the expected 4 ports x 0.9A = 3.6A current. D5 and D6 are too far away and too small for that current (SS14 are rated at 1A), and it seems D5 is dedicated to the bottom switching converter. How would such a converter work without a freewheeling diode? I hence suspect that this is not a switching converter but merely some inductor filtering of the incoming power supply 5V power from the molex to be supplied directly to the ports.
More skimping can be seen near the PCI-E port where L1, L2, L4 are not populated and are shorted through on the PCB itself. These appear to be places for interference suppressing inductors for the PCI-E differential pairs, but they were never fitted.
The rear of the card has no components at all, just a generic QC Passed label.
You get a full height backplate, and nicely cut-out holes for the four connectors you were promised.
The card accepts a molex plug for power, with the bevelled edges facing the board.
It also comes with a generic driver CD wrapped in cellophane. The CD has the recordable logo on the top, but is a pressed disc. How absurd.
Installing the Unit
Despite the fact the components installed seem to suggest this card is a bit of a skimpy nightmare, provided no faulty USB devices are plugged in, it should still work. As a result, I decided to go ahead and install it and try it out.
The first problem was the wretched CD, which was so poorly pressed that the installer could not be read by two different drives. The CD also contained drivers for other USB 3.0 based products.
As a result, I resorted to the norm of downloading Renesas drivers from station-drivers. Specifically, I installed Version 184.108.40.206 driver first to gain access to the card, and then upgraded the firmware to Version 220.127.116.11.
Originally, the card had firmware version 18.104.22.168 installed. Upgrading was successful.
It was also determined that the card’s coexistence with an older NEC uPD720200 was not a problem, which was good.
The performance of the newer chipset was slightly better than the older one under Windows 7 with the Seatay external enclosure and a “dirtied” SSD.
While peripheral add-on boards are generally very simple devices and low cost is quite normal, it seems that this low cost board cuts corners by using component substitution methods to eliminate what they deem non-essential functionality to reduce costs.
As a result, it seems that proper individual port overcurrent protection does not exist on this card – in the best scenario, the 5v is provided by a switching converter and overloading one port causes the converter to shut down due to overcurrent resulting in 4x more current to flow and all ports to lose power.
In the worst case, it could just be connected through to the 5V molex connection, meaning a possible transient 48A of current flow (provided the power supply can supply it, and provided the supply is ATX compliant with 240W short circuit power delivery). I really don’t want to find out – it could even get dangerous.
The loss of the ferrite beads on the port power filtering is probably also not optimal for RF suppression reasons, as now USB cables can radiate digital hash noise from the power rails inside the case. The loss of reverse polarity diodes won’t normally cause any issues, except for when someone forcibly plugs in the molex connector backwards and possibly fries all their connected USB devices and or motherboard.
Of course, most users will just plug it in, find it works properly with proper peripherals and claim the card is just fine. Which it is. Until something goes wrong, and causes damage which it otherwise shouldn’t. Just another case of being dangerously cheap.