[Gb-ccb] Spartan-3 USB interface chip

[Martin Shepherd]

Apparently the USB interface chip on the Spartan-3 development board is
a Silicon Laboratories CP2101. This isn't what I was expecting at all. The
data-sheet at:

 http://www.silabs.com/products/pdf/CP2101_Short.pdf

says that it is an RS232 <-> USB-2.0 bridge. Now if this chip truly
were a native USB 2.0 device, running at USB 2.0 speeds, then this
would potentially be ok. However the data-sheet says that the CP2101
is a "USB Specification 2.0 Compliant; Full-Speed (12 Mbps)"
chip. This is a fancy way of saying that it is actually a USB 1.0
chip, going at USB 1.0 speeds. Thus their claim that it is a USB 2.0
device is based on the fact that any USB 1.0 device can run in a USB
2.0 system, albeit at a much slower rate.

Again, if this chip truly used the available capacity of a USB 1.0
link, then it would be just as suitable as the chip that I was going
to use. However the data-sheet says that the CP2101 has a maximum baud
rate of 921.6kbps, which is a factor of 8 slower. In fact this would
make it impossible to transfer all of the CCB data over one USB link,
since it has about half the throeretical minimum speed that we need.
There would be sufficient speed for just the 4 channels-worth of data
from a single development board, but since we would have 4 of these
USB connections time-sharing through a single USB hub, it isn't clear
whether there would actually be enough capacity. Furthermore, I don't
know whether the USB virtual serial-port drivers in the computer would
be able to keep up with this.

In summary, although the FIFO serial-interface that the CP2101
provides would be much easier to program than the general USB
interface chip that I was fearing, I am dubious as to whether it's
speed will be sufficient for our needs. Clearly this is something that
we need to discuss at the telecon.

Martin