Need help on mix different clock tree DFFs in one scan chain, has timing issue?

[yx.yang]

[solved] Need help on mix different clocks' DFFs in one scan chain, has timing issue?

Hi, All:

I meet a problem about mixing different clock tree DFFs to one scan chain. I don't know whether this will causing timing issue for the DFFs at cross clock tree point (for exampel:data come from clk_tree_1 DFF and send to clk_tree_2 DFF).
As we using 90ns process now, we need consider the OCV iuuse, this cause even worse timing issue to the cross clock tree point. Becaue the two clock tree are totally different. Then the delay variation of the two clock tree will be a little big number under OCV analysis mode.
I see DC DFT user guide has said: dc_dft can insert a "lockup latche" at that point, while the DOC doesn't explaint this clearly.
Can anyone help me solve this problem or give me some info.
1): can we mix different clock tree DFFs in one scan chain or not? Consdier the OCV issue under 90nm process.
2): where can I find some more infos talking about this.

Thanks ahead.

 

[lostinxlation]

OCV is irrelevant. You just need lockup latches wherever the scan chain crosses the clock domain and set false paths on the domain crossing paths.

 

[yx.yang]

OCV is irrelevant. You just need lockup latches wherever the scan chain crosses the clock domain and set false paths on the domain crossing paths.

Hi, lostinxlation:
Thanks for your reply. While some guy said: "No, Scan tool won’t insert latch for capture mode, only shift mode. Remember, the capture is through the normal data path, scan tool is not allowed and should not touch the normal logic at all !! ".
It seems that this guy has different understanding of the use of lockup latches.
I have go through the synopsys dft user guide, and I don't think they told this matter clearly. Would you please upload some DOCs or put some links where I can articles which talk about this matter more clearly.
If possiable, would you please send me some articles talking about this through e-mail? My e-mail address is: yangyuf1@126.com

Thanks a lot.

 

[lostinxlation]

The guy is correct. Lockup latches are inserted only on shifting paths and capture paths are usually untouched and are the same as the functional paths.
just search lockup latch on google, you can find a lot of articles.

I'm now confused which of shifting mode or capturing mode your concern is about.

 

[yx.yang]

The guy is correct. Lockup latches are inserted only on shifting paths and capture paths are usually untouched and are the same as the functional paths.
just search lockup latch on google, you can find a lot of articles.

I'm now confused which of shifting mode or capturing mode your concern is about.

Hi, lostinxlation:
I'm concern whether I can mix DFFs coming from different clcok tree to a single scan chain? Whether this will causing some timing problem or increase the clock tree gate count a lot.
So, I concern both shifting mode and capturing mode.
Thanks.

 

[lostinxlation]

You can mix the clocks on the same chain.
For shifting, having lockup latches can give you a solution to avoid metastability issue.
For capturing, you just can't toggle all the clocks at once. You need to toggle one clock at a time and all works fine.

 

[yx.yang]

You can mix the clocks on the same chain.
For shifting, having lockup latches can give you a solution to avoid metastability issue.
For capturing, you just can't toggle all the clocks at once. You need to toggle one clock at a time and all works fine.

Hi, lostinxlation:
Gets you point. Thanks for your wonderful reply and great help.
I just wonder whether the DFT tool can toggle the scan clock at different time or not? I will find it out.
Thanks a lot.
^_^