Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

New technology - what to do

Status
Not open for further replies.

CAMALEAO

Full Member level 4
Joined
Jul 29, 2016
Messages
201
Helped
2
Reputation
4
Reaction score
2
Trophy points
18
Activity points
1,868
Hi everyone,

I would Mld like to ask the following: when you are going from one technology node to another one, in this case, lower one, what kind of things one should pay attention to? What one should simulate to check differences, etc?

Does anyone has been in this situation? Can anyone give some highlights, and share their experience?

Thank you.
 

Hi everyone,

I would Mld like to ask the following: when you are going from one technology node to another one, in this case, lower one, what kind of things one should pay attention to? What one should simulate to check differences, etc?

Does anyone has been in this situation? Can anyone give some highlights, and share their experience?

Thank you.

This is rather vague... but generally speaking you look at how to build an inverter. There might be new layers you are not familiar with (fins, cuts, M0). The wells for the process might be different ie triple vs twin, etc.
 

I start by investigating mos characteristics, pmos-nmos Ion ratios, leakage, gate capacitance, temperature variations etc .So I can think about what will be the impact on my design and modifications that has to be done.
 

I think there should be EDA tools that help with process migration of your designs...
 

Thanks for all the comments.

It is not a question about migrating the designs. It is more about familiarising with the new process node. How normally you guys do it. What's your flow, tests, etc. This will be the first time I am going to do this hence the question.
 

Depending on your node1, node2 and relationships that
may or may not exist with the foundries, this could be
anything from poking PCM slivers with microprobes, to
blind trust in the PDK.

Personally I like the poking. But trying to probe 10nm
minimum geometry transistors meaningfully without just
breaking every one (core device ESD/EOS tolerance being
roughly nil) could be a challenge, and at the advanced
nodes DC attributes are largely "care little", sacrificed
for any speed*power advantage and good luck with
validating GHz-range device models without a really
fancy (spendy) setup from tips to rack to characterization
software.

Which is one of many reasons I sit back comfortably
near the trailing edge.
 

Thank you all for your inputs.

I start by investigating mos characteristics, pmos-nmos Ion ratios, leakage, gate capacitance, temperature variations etc .So I can think about what will be the impact on my design and modifications that has to be done.

In general that looks reasonable. But how would you do this? Via simulations or documentation? Does the documentation have this kind of information?
 

You may find some of this in the back pages of PDK design
manuals, but there would be much more in (say)
"process modeling documents" that would be part of
internal library / model development closure. Whether you
can extract them from the foundry depends on the nature
of your relationship. Simple curiosity probably doesn't
punch their ticket, but you never know until you ask.

Somewhere in design related docs should be a declaration
about where and where not to trust the models. Then to
the extent that the models are represented good for the
things you care about, design simulations can be a good
first comparison. But leakage is one thing that often gets
hand-waved away, esoteric analog / RF attributes of what
the foundry considers "digital" devices is another.

You probably want to take it on from a few different angles
with an eye to how consistent the "story" is. Curves you
get from a marketing brochure and what you get from a
N=100 Monte Carlo analysis and what you get from a
signed off modeling document and what you'd get from a
physical sample on the bench, stand a pretty good chance
of disagreeing (less so, the MC results and the modeling
doc, since those "come from the same shop").
 

Thank you all for your inputs.



In general that looks reasonable. But how would you do this? Via simulations or documentation? Does the documentation have this kind of information?

Both documentation and SPICE simulations.. In general the Design Kits come with all kind of documentation, not only DRM. Based on what you are looking for, one of these documents might be the one you are searching for.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top