Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. It only takes a minute to sign up.
Sign up to join this community
I am learning about the USB interface. My current understanding is the following:
The Host Computer contains an I/O interface, called root hub. This root hub is connected with either other hubs or I/O devices.
Now my question is. What is exactly the difference between hubs and I/O devices? There is a diagram in the textbook I am reading that shows the two as clearly being difference from each other:
A USB hub is formally a USB device too. As such, it is a subject of USB enumeration process as any other USB device on the bus. Therefore a hub gets its own dedicated USB address, and responds to various host inquiries on its own, like reporting status of its downstream ports, and responding to interrupt pipe when something is connected or disconnected.
However, hubs are special devices in the sense that their main function is to relay/channel all USB packets to downstream ports (and back), so other devices can be connected and enumerated. If there are connected and enabled USB devices in HS (High-Speed, 480 Mbps) mode on its downstream ports, anything that comes to its upstream port will be re-transmitted (repeated) downstream, and anything that comes from downstream devices will be repeated upstream. This is a fairly simple two-way repeater functionality.
However, if the connected downstream device is Low Speed or Full speed, USB hubs have additional responsibilities when relaying the USB packets - they translate the fast upstream transactions into slow LS/FS transactions. For this purpose hubs have dedicated buffers with enough size to accommodate longest FS/LS packets. For this purpose the USB host uses special class of communication, so-called "split transactions". The rules of split transactions are fairly complicated, and USB specifications have nothing better than list many examples that took more than 100 pages to explain. The host feeds the hub TT (transaction Translator) just enough to keep slow devices running, and this saves the upstream HS bandwidth for other devices. In this sense USB hubs are fairly sophisticated communication processors.
In short, normal USB devices perform specific I/O functions, like storing data in flash memory, or adapting a WIFi modem, or bridging/converting USB I/O data into other interfaces like UART or I2C, or relaying keys pressed on a keyboard. The hubs have an universal function to repeat/broadcast/ any data regardless of their origin or purpose.
A clearer image of the architecture might help you: http://www.usblyzer.com/usb-topology.htm
All endpoint devices are of course I/O devices if they are using the USB serial protocol.
The USB controller is separate from the USB Root hub (though built together in the controller chip).
Read the documentation for the NEC720201 controller chip to understand the current architecture. The USB controller is split in two ...on one side is the PCIE to USB controller interface (called the xHCI) ...and on the other the USB Root Hub and Phy.
The HCI contains all the parallel to serial logic and the memory management/DMA logic. The USB Root Hubs handle only serial data for the USB data/command protocol.
If you are not learning about USB 3.0, then you might be better to start with the upd720101 which was the classical reference chip for USB 2.0 implementations.
I agree with the other answer that the nomenclature is inaccurate. The USB specification provides lots of vocabulary to clearly define every device and every bit in the USB protocol. This tree graph doesn't reflect that.
However, I think the best answer to your question is simply that a USB Hub allows you to connect to more USB devices besides itself. Not all devices do and therefore aren't hubs. Hubs can connect to other hubs, and therefore provide even more ports, up to 5 levels, if I recall correctly (someone check me on this).
This answer may also be interesting to you.
That reference is making invalid assumptions - what they are labeling an "I/O device" is merely a USB device - it might or might not have any I/O capability.
For example, in a USB flash drive, nothing goes any further. Communication with it via USB could be argued to be I/O from the perspective of the computer, but so could communication with an intermediate hub, and those aren't being categorized as I/O devices. Therefore the proposed model is meaningless, and invalid.
