Hi,
Here's a summary of parts of the USB 2.0 Developer's Conference
that I went to last week.
There were several keynote speakers. I'll only tell you about
two of them.
1. Martin Reynolds (Dataquest VP and Fellow):
Reynolds is a dynamic speaker who kept his presentation moving
rapidly, with lots of good content also.
Everyone else there talked only about Win2000. Reynolds told the
USB device vendors (IHVs) that they should consider Linux driver
support for their devices as well as Windows driver support.
2. Bill Machrone, Ziff Davis publications:
Machrone asked his readers to send him messages that should
be delivered to the USB (2.0) Developer's Conference.
Most of his feedback was about particular problems with devices
and/or configurations (like "doesn't anybody ever test this stuff?")
since the people with problems are more likely to write to him.
3. There's a new (additional) USB compliance spec (at usb.org)
which addresses power management in more detail.
4. USB 2.0 Technical Overview:
. Same host/device model
. Same basic protocol (with a few new PIDs, like PING, NYET,
Start-Split, Complete-Split). Split transactions are
HC<->hub, not to devices.
. Same device framework (descriptors, messages)
. Same software interfaces (USBDI) (N/A)
. Same power distribution & consumption
. Same power management model
. Same topology management
. Same cables & connectors (mostly; low-speed devices with
unshielded captive/attached cables are not compliant)
. Primary changes are in the host controller, host controller
driver, and hub driver (at least in Windows)
. New 480 Mb/s signaling rate
. 2.0 Hubs must be more intelligent to handle low/full/high-speed
traffic/protocol
. USB 2.0 is an extension of and interoperable with USB 1.x
. The USB 2.0 spec replaces the USB 1.1 spec.
. USB 2.0 high-speed devices are required to support full-speed
signaling at least for enumeration, preferably for some
(reduced) backward compatibility; high-speed devices must not
support low-speed signaling.
. USB 2.0 hubs must support low/full/high-speed downstream ports.
. USB 2.0 defines several test modes and fixed control messages
to put devices into those modes.
. USB 2.0 splits 1-ms frames into 8 125-us micro-frames for greater
interrupt and isoc bandwidth. No change/effect on bulk or control
messages.
. USB 2.0 high-speed devices always have control max packet size of
64 bytes (no changes for low/full speeds) and bulk max packet size
of 512 bytes (no changes for low/full speeds).
. Since endpoint interval was always implemented in power-of-2
brackets, high-speed intervals are defined as
2^(bInterval - 1) microframes, not as bInterval microframes.
. 2.0 hubs have optional but standardized LED indicators (green,
amber, off, blinking).
. The USB 2.0 compliance program is being strengthened and a
"compliance" logo is being added.
5. Microsoft's USB 2.0 Roadmap & backport plans:
. Microsoft's USB 2.0 stack for Windows 2000 is being used
in demos (control, bulk AFAIK; maybe int.; no isoc yet).
. They are trying to support "early adopters" (IHVs) in Q4/
2000 with discrete PCI/USB addin adapters. Expecting
broad availability by mid-2001 (in chipsets).
. Tentative full USB 2.0 support is planned for Windows
"Whistler" (consumer/workstation/server) in 2001.
. Tentative backports for Windows ME and Windows 98 SE, but
not Windows 98 gold (first edition), which disappointed
several people there, but Microsoft explained that it's
due to other infrastructure (besides USB drivers) being needed,
such as streaming audio/video, which are outside of the USB
stack. [It sounded as though it would be similar to us
backporting USB to 2.0 or 1.x and having to also add V4L,
OSS, etc., to the backport.]
. Windows 2000 is now the development platform of choice (better
tool chain; easier to develop on and then port to other Windows
versions).
. Microsoft wants to test USB 2.0 host controllers, hubs, &
devices for at least 3 months (July-September), and must have
at least 2 (different) host controllers, 3 hubs, and 6 devices
and 3 months of solid testing in order to ship USB 2.0 support.
. Microsoft says that they have 2 USB 2.0 host controllers
for development/test.
. Microsoft is "worried" about 2.0 hubs.
. Microsoft says that their schedule is very aggressive.
. They say that USB 2.0 storage is fine for secondary
add-on storage but have no plans to support a USB 2.0 storage
device as the primary boot device. Recovery from a USB 2.0
CD-ROM driver will be supported (as long as the BIOS supports
it).
6. EHCI (Enhanced Host Controller Interface) and PDK:
. Rev. 0.95 of the ECHI spec for discrete HCs will be the first
public version (Q3/2000). Gated by validation of 2 discrete
HCs. Rev. 1.00 to be available in 2001.
. Developed by Intel with contributions from NEC, Lucent,
Philips, Compaq, & Microsoft.
. There is a host controller compliance program to ensure that
ECHI host controllers are compliant to the spec.
. ECHI interface is largely like OHCI (PCI config space,
operational register space [memory-mapped], and a shared
memory work interface. It uses two schedule lists, one for
periodic (int./isoc) and one for async (control/bulk).
. No hardware/software sync required to add/delete work on a
queue.
. Architecture is optimized to provide efficient memory accesses.
. PDKs (Peripheral Development Kits) are planned for late June
availability. They contain a PCI/USB 2.0 addin card, a USB
2.0 software stack for Windows 2000, and USB 2.0 transaction
generator software.
. There is also a USB 2.0 Compliance Device that is planned for
July/2000.
7. Phoenix Technologies has a BIOS that will boot from USB
floppy or Zip drives, working on CD-ROM and LS-120, others
to follow as needed.
. To support topology changes, Phoenix wants to see consumer
pressure on floppy vendors to add serial numbers.
8. contacts, promised items:
. Met Jim Blackson of Y-E DATA there. Hi Jim. Would you like
to add anything to this DevCon summary?
. Met NetChip Technology VP Sales/Mktg Robert Wong.
. Discussed Linux USB 2.0 support with a GM from Philips.
His answer to other people was "waiting for Windows 2000
driver support," but for me his answer was "waiting for
silicon validation."
. NEC didn't want to discuss host controller availability
or Linux support.
. Lucent was interested in Linux support for USB 2.0 but
doesn't have devices available yet. They are sending me
a USB 1.x 4-port (or 4-bus) OHCI adapter (called QuadraBus).
It has 4 USB ports and one USB/OHCI chip on a PCI addin
card. Each port is 12 Mb/s, total 48 Mb/s.
. In-System Design is supposed to send a USB-ATAPI converter
device to me. Just hook it up to an IDE drive and +5V/Ground
and it should work. :) Matthew Dharm, can I send this to you?
. Lexar is interested in having their flash devices supported
by Linux. I'll plan to have the devices sent directly to
Matthew Dharm.
----------------------------------------------------------------
> Randy, how was the conference? Are USB 2.0 devices starting
> to show up?
It was good. Lots of technical presentations, but the tech.
detail could have been more in-depth IMO.
Only a scanner (Mustek with NetChip technolog USB device)
and storage (Zip driver IIRC), plust host controllers
(NEC was used everywhere AFAIK).
----------------------------------------------------------------
> Noticed that
> http://www-us2.semiconductors.philips.com/usb/products/host/isp1561/
> refers to a device that "Integrates two USB 1.1 OHCI and one USB 2.0
> EHCI compliant host controllers: works with all USB 1.1 and USB 2.0
> devices"
>
> Search engines don't seem to know about the enhanced host controller
> interface - can anyone who was at USB 2.0 tell us more?
I've left out a huge amount of detail here, but I don't
want to type the USB spec into email.
The EHCI info is based on presentation material. No spec
available yet.
~Randy
___________________________________________________
|Randy Dunlap Intel Corp., DAL Sr. SW Engr.|
|randy.dunlap.at.intel.com 503-696-2055|
|NOTE: Any views presented here are mine alone |
|and may not represent the views of my employer. |
|_________________________________________________|
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
