Taking Apart a USB Hub
When using a computer to operate electronic measuring equipment or power supplies, the various devices need to be able to communicate with each other. While consumer devices tend to connect wirelessly (using Wi-Fi or Bluetooth), it is still common for measuring equipment and power supplies for industrial use to use wired connections, due to the requirement for reliability and stability. Typical wired connection standards used in industry include General Purpose Interface Bus (GPIB), RS232C, Universal Serial Bus (USB) and Local Area Network (LAN). While GPIB used to be the standard means of connecting measuring devices, it is increasingly being replaced by USB and LAN. The USB standard, in particular, is simpler to use than LAN (which can be challenging for those unfamiliar with the technology), and a very useful technology when performing desktop experiments and building simple measuring systems using computers. In this column, I will be discussing the USB hubs used in such applications.
An Unstable Option?
You have probably had experience with an external hard disk or other peripheral which worked fine when connected directly to the computer but when connected via a USB hub either malfunctioned or was only recognized sporadically. In society, some people just don’t get along, and they say that the same applies to electronic devices. However, there is nothing worse than not knowing if two devices will work with each other until you connect them─if possible, you want to understand the underlying cause of the issue. With this in mind, and in somewhat of a homage to the article Qualities Bad Engineers Lack written by my colleague Kenichi Otsuka, I decided to dismantle a USB hub.
Inside a USB Hub
Multi-port USB hubs that allow four or more devices to be connected are commercially available for consumer use, for example as ten-port hubs. I purchased both consumer and industrial versions of a ten port hub, and took them apart to compare their inner workings.
I found that the consumer ten port USB hub (Figure1) contained three USB controllers. That is, rather than using a single chip to drive all ten ports, the hub used three chips to do the job. It would seem that generic USB controller chips are generally able to control a maximum of four ports. In the hub, controller chips are cascaded (Figure3) with one chip tiered above the other two. The internal cascade architecture allows the hub to provide ten USB ports.
In contrast, hubs made to industrial specification drive all ports using a single chip. Let us consider why these hubs do not use a cascade architecture. One would imagine that the cost of USB control chips depends on the number of ports they provide, with more ports meaning higher cost. In these days of belt tightening, one might expect manufacturers to use cheaper chips, even for industrial spec devices. So why don’t they?
The Truth about Compatibility
The USB standard allows up to 127 devices to be connected to a single host controller via a bus. It also allows a maximum of five hubs tiers between the host and any given device. In my experience, some combinations of hosts (such as PCs), hubs and USB devices are not conducive to a stable connection, despite complying with these above requirements. The issue may result from the use of a consumer use multiport hub. In other words, while the USB standard allows a maximum of five tiers of hubs, internal cascade architecture may mean that the system effectively contains more than five tiers, thereby not conforming with standard.
However, if you do not understand how a consumer multi-port hub works, you are likely to simply blame any malfunction on “incompatibility” and replace the hub. Then, when a randomly selected replacement hub (that happens not to employ an internal cascade architecture) does work, you will say, “I don’t know why, but it’s working now”.
The Difference with Industrial Specifications
The difference between consumer and industrial hubs is their approach to reliability. Put simply, everything comes down to economy versus reliability, with low costs being a trade-off for reliability. In consumer devices, cost is all important, while industrial applications place priority on reliability. Because of this, while the consumer and industrial versions of a device perform the same function, they may be different on the inside. While we tend to think of devices for industrial applications as being expensive, one reason for the higher prices of these devices is the philosophy that higher costs can be accepted in return for increased reliability. (Unfortunately, the smaller production volumes involved means that devices manufactured to industrial spec do not enjoy the same economies of scale as their consumer counterparts.)
Let us return to the topic of USBs. If a stable connection is important, you may be better off plugging the USB device directly into the host without using a hub. If you must use a hub, I suggest carefully selecting one made to industrial spec that does not rely on an internal cascade architecture.
There is a saying in Japanese about someone who thinks he sees a ghost, only to look closer and realize that it is just silver grass. The point is that often the thing you are scared of turns out to be a mere trifle on closer inspection. This is what happens when we baselessly dismiss certain electronic devices as “incompatible”. Rather than being put off by the urban myth of incompatibility, engineers need to search for the causes of issues and identify exactly what is behind these “ghosts” we see.
