On Thu, 2002-04-04 at 09:11, Dan Streetman wrote: > >If you use other pipes then the transfers can take longer, depending on > >the type of transfer. Maybe 10 ms, maybe even 100 ms ... These transfers > >are guaranteed to succeed, but no obligations on "when" part... > > That is not true, for INT pipes the USB spec clearly states the > polling interval cannot be greater than what is requested. So the > timing must be at least the ep's bInterval.
Is there such requirement for bulk pipes? I guess not. You probably put comma in wrong place :-) - That is not true, for INT pipes the USB spec clearly states ... + That is not true for INT pipes, the USB spec clearly states ... USB pipes have priorities, and some even have guaranteed bandwidth. However on the lower end of the scale "the transfers can take longer". I don't think we can offer any better suggestions to the original poster until he reveals what is the _order of magnitude_ of his timing requirements. It can be anything from nanoseconds to seconds. I would not be surprised to hear anything in this range, because I already had such requirements (+-0.5 microsecond) on another project. Dmitri
signature.asc
Description: This is a digitally signed message part
