From: Jim Thompson (jim_at_musenki.com)
Date: 2002-04-13 18:41:53 UTC
On Sat, 13 Apr 2002, Jouni Malinen wrote:
> similarily initialization routines for PCI/PLX in separate files. At
> some point, there might also be more changes that would be needed for
> USB devices, but I'd like to first get PCI cards working since they
> will probably require less changes.
Interesting you mention that. I was thinking about ways to reduce overall cost for embeded controlers (for hosting AP's on), and it seemed obvious to me that significant cost is involved in the I/O of the system. That is, you want miniPCI, you need to have the logic there, the connector, the matching etched conductors, etc. Why not build a platform arround say, a 4 to 6 channel USB interface, tied directly to a connector-less PCI interface? Ethernet could be interfaced via this method as well. I could imagine a rooftop installation lacking in large coax -- with radios placed mere inches of coax away from the antennas, yet the host controler is located somewhere more convienient. USB 2.0, obviously would be a great choice in terms of providing ample bandwidth to devices. But, even usb 1.1 (at 12mbits) would be more than enough considering the throughput 802.11/b realises (about 5 mbits or less). With 4 channels, I would imagine contention for the USB interface would be rather low.
Heh, we thought of this last year, and rejected it.
First, none of the USB 802.11 adaptors work well (perhaps 'yet'). I haven't seen any of the 11g/11a chipset vendors mention USB, but, as they say, its just silicon, so this issue could be fixed with time.
More importantly, the USB 1.1 and 2.0 specs say that max length is 5.0 meters (16.4 ft). I don't think you'll have a lot of luck pushing it farther, because the 2.0 spec calls for the same cables and connectors used in the 1.1 standard, but the voltage swing in the controllers has been cut to 400 millivolts, from 3.3 V in the version 1.1 USB chip sets.
And now you know why we use miniPCI @ Musenki. :-)