FX2 Introduction/Overview
 

EZ-USB FX2 offers a highly flexible and configurable Full Speed and High Speed USB peripheral function that's designed to achieve the maximum USB 2.0 High Speed bandwidth. FX2 accomplishes this through its auto-transfer and peripheral interface architecture. The GPIF (General Programmable InterFace) engine is one of the vehicles for the auto-transfer architecture, and is used to gluelessly move data between FX2, your external slave device, and the USB host. The following sections briefly touch upon the silicon architecture and implementation of FX2, and will lay the groundwork for understanding GPIF concepts discussed later on.

In order to achieve the maximum USB 2.0 High Speed bandwidth, it was proven that the CPU (in this case an enhanced 8051 core) should never be directly involved in moving the payload data from the external slave device to the USB host (and vice versa). The CPU would clearly be the largest bottleneck in a High Speed design. So instead an Auto-Transfer mode was invented, whereby the payload data is "auto-committed" from the USB host to the external slave device and likewise in the other direction. The GPIF engine uses this Auto-Transfer mode to move data to and from the external slave device. Figures 1 and 2 show a system level view of the data flow for both the IN and OUT directions.

 

Figure 1. System Level View of Data Flow in the OUT Direction

Figure 2. System Level View of Data Flow in the IN Direction

Endpoint FIFOs

In FX2 the USB endpoints share the same physical memory as the FIFO buffers. This is the very backbone of what allows the Auto-Transfer mechanism to work. Anytime the USB host sends a packet of data the packet is stuffed into an available Endpoint FIFO buffer (the multiple buffering schemes available allows FX2 to accept up to 4 packets before NAKing the host). The data packet is then ready to be distributed to the external device. Conversely, a data packet read from the external device can be automatically made available to the USB host. FX2 can be configured to allow the CPU to manually "commit" the data packets (called Manual Mode) or allow the data packets to be committed automatically (called Auto Mode) to either the USB or peripheral domain. Figures 3 and 4 show the maximum buffering scheme (4x) in relation to the overall data path the data packets follow.
 

Figure 3. Auto-Transfer Dataflow in the OUT direction

Figure 4. Auto-Transfer Dataflow in the IN direction

The CPU is the Traffic Cop

The CPU in FX2 has a minor role in the Auto-Transfer architecture as it does not participate in moving the payload data, but it does play a very important role nonetheless. The CPU configures and defines how the physical interface operates, sets up the endpoint configurations, triggers GPIF transfers, and can be allowed to manually commit the data packets to either the USB or peripheral domain. It can also monitor the status of the external world, thus giving it the capability of regulating GPIF transfers.