[Noisebridge-discuss] cheap USB Analyzer, or signal integrity tools?

[Michael Prados]

Hi All,

This has been a topic of increasing interest to me, as USB becomes
less of an optional luxury in hardware hacking and more of a
necessity, and it's come up again lately in the context of Adafruit's
bounty for the Kinect:

http://www.adafruit.com/blog/2010/11/05/our-kinect-arrived-today-you-gonna-get-modified/

So many devices that I care to use or build have a USB interface, and
there are a lot of tools out there for embedded device developers to
add this functionality.  That is all well and good when it just works,
but frequently the reality is less than ideal.  Here's some scenarios
I've had to deal with:

* the interface for a USB device is not, or not fully, documented
* the interface for a USB device is nominally documented, but doesn't
perform exactly as the documentation suggests
* environmental noise, connector, or transmission line issues lead to
signal integrity problems (not to mention PCB design issues!)
* you make a working system, set it up somewhere where it is difficult
to maintain, and the USB subsystem stops working. How can you diagnose
it remotely, or outside of the lab?

Traditionally, for signal integrity issues- essentially physical layer
issues, you need a high-bandwidth oscilloscope.  I feel like it is not
unreasonable for a hacker to get a hold of a 100 MHz scope, which
might suffice for Full Speed USB, but for High Speed USB at 480
Mbits/sec, you need a scope at up around 2 GHz or above.  Even a used
scope in this range typically runs $5k and up.  And heaven forbid you
need to debug a problem outside of the lab- are you going to strap
your nice scope to the top of a car, or a helium balloon?

What I really fantasize about for the physical layer is a
self-diagnosing USB hub.  Imagine if your hub could provide even a
rough estimate of the eye size.  Has anyone encountered anything like
this?

So far as the data layer, this is where the USB Analyzer typically
comes in.  It seems to cost about the same as the multi-GHz scopes.
For this, it really seems like something running on Linux, perhaps
with special hardware, could do the trick.  Anyone encounter something
like this?  I've used Windows based USB monitor software before, but
I've found this kind of limiting, especially if you can't run this
software on the host device.

All too frequently, I end up resorting to RS-232 or RS-485 when I have
the choice.  This may still be right decision, even in 2010, but I
hate to be forced into it by the inaccessibility of good USB debug
tools.  Seems like a major barrier to hardware hacking, which is only
going to get worse if a next generation technology such as USB 3.0 or
Light Peak gains in popularity.

Any one have some good solutions to these problems?  I'll probably
post elsewhere too, but I figured on giving it a try here first.

-mike


-- 
[REMOVE THIS TEXT BEFORE SENDING AN EMAIL!]