Brad, > From: Brad Hards [mailto:[EMAIL PROTECTED]] > > "Dunlap, Randy" wrote: > > I hadn't seen the 5-limit anywhere lately either, > > and I looked for it a few weeks ago unsuccessfully. > > However, I don't usually spend much time looking at > > "Chapter 7: Electrical", but that's where it is since > > it is (worst case) timing-related. > > > > On pdf page 150 or spec page 134, at the top: > > > > "Figure 7-31 depicts the configuration of six signal hops > > (cables) that results in allowable worst-case signal > > delay. The maximum propagation delay from the upstream end > > of a hub's cable to any downstream port connector is 70ns." > > > > Note that this doesn't STRICTLY say a limit of 5 hubs ("six > > signal hops"), but that's what the graphic there shows. > > It's also what all of the USB Compliance/Interoperability > > Gold-Tree test configurations show IIRC. > Thanks very much Randy. I thought I must have imagined it. > > The Gold Tree tests is actually one of the reasons why I was asking. > When I write tests, I like to understand what failure of the > test means. > The Linux compliancy tests try for operation behind five hubs. So are you looking at what USB-IF calls compliance tests or interoperability tests? Or both? When will you be ready for a test spec review? What we need most for USB-IF PlugFest in August is interoperability tests...that should have priority IMO. > My understanding from this is that providing the cable delay > + hub delay > is <=70ns, there could be an arbitrary number of internal hub chips. > This means that it may not be possible to have a heiristic in the USB > code that can prevent excessive depth, since we can't tell the > difference between two hubs with a five metre cable between > them, and a > hub with two TI2040 chips wired together to make seven > downstream ports. > > Has anyone tried excessive depth? Maybe I have to get some > more 5 metre cables... Only 5 external hubs, some bus-powered, some self-powered. > > The USB 2.0 spec will be public soon. It clarifies > > this a bit better. > Can you say whether the maximum depth problem is easily solvable with > 2.0? No, because I haven't seen the final wording of the spec, only the "release candidate." I discussed it with the writers/editors/architects a few weeks ago and I'm hoping that they will clarify it some. > > Brad ~Randy --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
