[rescue] Re: JTAG (was: Happy happy happy! Joy joy joy!)

[Skeezics Boondoggle]

On Tue, 3 Jun 2003 Joshua D. Boyd wrote:

> Are you aware that there is a JTAG IEEE standard, and I'm begining to
> suspect that Sun's JTAG is just that, except with undocumented
> registers (which are left to the implementor anyway).

Yup.  I've seen references where Sun/Cray called theirs "JTAG+", and they 
mention a small addition... Ah!  Found it:

"The IEEE Joint Test Action Group (JTAG) standard describes a 4-signal 
interface for serial scan of digital logic devices:  ICs, modules, boards, 
and systems.  The four signals are TCK (clock), TMS (test mode select, a 
serial control line), TDO (test data out) and TDI (test data in).  Sun has 
added a non-standard signal TAS (test address strobe)."

So the Sun JSCC Sbus card supports that extra signal, and they use a
software package called ScanTool to do debuggin, factory test and
diagnostics on the SC2k; the SSP software drives it instead on the CS6400. 
The cable uses 18 pins out of that DB-25 connector; the docs say they use 
RS-422 type differential signalling and "the 20' length is fixed and 
cannot be changed due to JTAG+ timing restrictions."

Neat stuff. :-)

In another followup:

> Of course.  But we can at least probe the chain length, and what
> registers of sizes are in each chip on the chain.  From there one can
> record and reverse engineer other information, but it's certainly
> extremely labor intensive.

Back in the early '90s I was doing software manufacturing for a company
called TSSI (they've been through a thousand renamings, reorgs, and been
bought and spun-off so many times now I have no idea what they're called
anymore!) who had software that took CAD output and simulation data and
produced test programs automatically, for doing ICT on a wide range of
board and chip testers.  They were into boundary scan and supported the
JTAG efforts to come up with standard interfaces for all that stuff, so I
was peripherally exposed to it, but hey, I was a sysadmin and tape spinner
and shipping grunt, not a chip designer... :-)  But everyone who was
anyone used TDS and WaveMaker, including Sun, and most of the other big
systems vendors, chip makers, even some "three letter agencies" of the
gubmint...their customer list really was a "who's who" of the industry.

So the idea was that a lot of that labor was made a lot less intensive,
since you didn't have to spend months writing test programs by hand - TDS 
could grind things out in a couple of hours or days.  It was pretty cool.

Ironically, when I worked there the offices were at 8205 SW Creekside
Place in Beaverton, OR - the Cray BSD office was at 8300 SW Creekside, and
I didn't know that until years later - d'oh!!  Sun's Beaverton offices are
still in that same business park, I think, but I haven't been out that way
in years...

-- Chris