Access Interface modeling

Discussion and feedback on the SJTAG Initiative Group weekly meetings.
Post Reply
User avatar
Michele
SJTAG Established Member
Posts: 31
Joined: Wed Nov 11, 2009 8:39 am
Location: Grenoble, France

Access Interface modeling

Post by Michele » Fri Oct 09, 2015 4:20 pm

I am waiting for my shuttle pickup after a really interesting ITC, and some of the discussions I had (especially after the 1687.1 meeting) helped me flesh out the last really important open point in our System Modeling: Access Interface. Sorry for the bad image quality.

I believe that our difficulties come from tha fact that TAPs and such have in fact TWO functions:
- provide DR chain selection
- provide external connection.

Only the latter point is an Access Interface per se, the first one is no different from a scanmux/linker. So my proposal is to model these two function separately, and most importantly :

"the Access Interface is responsible for providing a set of commands (protocol) allowing operations over one or more register chains ".

This way, an 1149.1 AI would look like this:
AI.pdf
(21.77 KiB) Downloaded 65 times
In the node you have the protocol, that is to say the set of basic commands. In this example, I just put the ShiftIR and ShiftDR SVF commands. Then, for each derivation you can define a specific methods that combines protocol commands to access the chain. Once more, for 1149.1 it is simply calling SIR/SDR with the vector value extracted from the circuit model transversal.
A complete 1149.1 TAP would look like this:
TAP.pdf
(24.09 KiB) Downloaded 59 times
The linker simply references the IR register, which appear explicitely in the Circuit Model, which is much clearer and robust that in my previous modelling where it was a private value used during the "preconditioning". The nice thing is that now that is part of the CM, the configuration cycle we discussed is immediately able to handle it, without having to change anything!
And you can easily do composition, for instance here are two daisy chained TAPs, which I had some problems in figuring out before:
Daisy TAP.pdf
(28.55 KiB) Downloaded 50 times
And now that the Access Interface node is defined, new AIs can be added simply by filling in their protocol, straight from the behavioural documentation without any need of internal knowledge. So, here is the Brocade TAP mux:
Brocade.pdf
(28.67 KiB) Downloaded 65 times
The protocols are simply the I2C operations needed to access the control register. Once more, as the register appears explicitly in the CM, SJTAG can directly use it for configuration without any specific code.

And once the I2C AI is defined, what would 1687.1 be? Simply some registers accessed through an I2C AI, like this:
1681_1_1.pdf
(25.9 KiB) Downloaded 59 times
Last edited by Michele on Mon Oct 12, 2015 7:08 am, edited 2 times in total.

User avatar
Michele
SJTAG Established Member
Posts: 31
Joined: Wed Nov 11, 2009 8:39 am
Location: Grenoble, France

Re: Access Interface modeling

Post by Michele » Fri Oct 09, 2015 4:22 pm

And in the second, more complex version of 1687.1 where you have I2C driving a TAP? In these terms, you would just need a conversion table where the new AI gives an equivalence for the protocol defined in the base AI. Once more, this is something found directly in the documentation:
1681_1_2.pdf
(26.12 KiB) Downloaded 56 times
Hey, the standard has not started yet and we already support it :lol:

Apart from jokes, what do you think about this idea?

Regards,

Michele

User avatar
Ian McIntosh
SJTAG Chair
Posts: 415
Joined: Mon Nov 05, 2007 11:49 pm
Location: Leonardo, UK
Contact:

Re: Access Interface modeling

Post by Ian McIntosh » Sun Oct 11, 2015 7:24 pm

Michele, not to detract from the purpose of your posts, but the AI.pdf and Daisy_TAP.pdf files seem to have been repeated in your first post - I don't think that's what you intended.

Actually, I think there could be a Newsletter item in this.
Ian McIntosh
Testability Lead
Leonardo MW Ltd.

User avatar
Michele
SJTAG Established Member
Posts: 31
Joined: Wed Nov 11, 2009 8:39 am
Location: Grenoble, France

Re: Access Interface modeling

Post by Michele » Mon Oct 12, 2015 7:10 am

Thanks Ian,

I edited out the typos. I am still not too familiar with the forum interface, especially for inline images. I supposed I should upload them as JPEG to have them directly displayed in the text, I will try it next time!

Regards,

Michele

User avatar
Michele
SJTAG Established Member
Posts: 31
Joined: Wed Nov 11, 2009 8:39 am
Location: Grenoble, France

Re: Access Interface modeling

Post by Michele » Fri Oct 16, 2015 12:33 pm

In 1687 terms, I would say that an "Acces Link" would be composed by what I call "Access Interface" + "DR Mux" ....which should probably get a better name.
I remember Brad using a specific term, but I cannot quite recall it!

Post Reply