[Open-graphics] A really good immediate use for OGD1: PCI bus analyzer

[Timothy Miller]

On 12/31/06, Jack Carroll <mykrowatt at comcast.net> wrote:

>
>         Tapping a PCIe signal should be doable, but it would take some
> serious engineering at the electrical level.  What I'd try is to tap the
> transmission line with a 3-arm T attenuator, with most of the power passing
> straight through, and a greatly attenuated signal coming out the side arm.
> Then terminate the side arm, and amplify back to standard logic levels for
> compatibility with a logic analyzer.  The tap adapter would have to be
> implemented as a microwave PC board, with connectors to plug between the
> device and the cable.

I envision having 1x and 16x "extender" sockets.  The extender plugs
into the mobo, and the device being debugged (let's call it the
"subject" or "agent" or "object" or "patient," perhaps) is plugged
into the extender.  We can then use what you're describing as a way to
get at the signals.  PCIe, with its differential signalling and
self-timing should handle a reasonable amount of trace lengthening.
Reflections are what's going to kill us.  You can get reflections from
tees, as well as from impedance mismatched connectors, and a number of
other things that I don't know as much as you do about.


-- 
Timothy Miller
http://www.cse.ohio-state.edu/~millerti