about monte carlo analysis - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>EE - 电子工程

about monte carlo analysis

about monte carlo analysis# EE - 电子工程

z*b2007-09-28 07:09

1 楼

跑一个运放的input offset volgate，monte carlo跑100次，可是得到的offset仅仅为
几十个uV。
问了一下好像是因为每次跑的时候所有的管子的mismatch都是朝同一方向变化，也就是
说每次所有的
管子都增加或者减小相同的百分比。这样根本跑不出来真正的offset。好像是库里面需
要修改一下参数，
具体不知道怎么修改，望高人指点一下。
以下是库里面的统计模型：
section statistical
//parameters sw_stat_global=0
//parameters sig_limit = 3
parameters mc_vth2=0
parameters mc_tox2=0
parameters mc_xl2=0
parameters mc_xw2=0
parameters mc_xj2=0
parameters mc_vthn=0
parameters mc_vthp=0
parameters mc_vth=mc_vth2
parameters mc_toxn=0
parameters mc_toxp=0
parameters m

s*g2007-09-28 07:09

2 楼

You need use mismatch instead of process variation in your statistical block.
Process variation will apply the variation to all devices.
Mismatch apply different variation for each instance.

【在 z*b 的大作中提到】

: 跑一个运放的input offset volgate，monte carlo跑100次，可是得到的offset仅仅为
: 几十个uV。
: 问了一下好像是因为每次跑的时候所有的管子的mismatch都是朝同一方向变化，也就是
: 说每次所有的
: 管子都增加或者减小相同的百分比。这样根本跑不出来真正的offset。好像是库里面需
: 要修改一下参数，
: 具体不知道怎么修改，望高人指点一下。
: 以下是库里面的统计模型：
: section statistical
: //parameters sw_stat_global=0

z*b2007-09-28 07:09

3 楼

but my model does not include mismatch parameters which is provided by chart
ered. should I write the mismatch parameters by myself? but how do to write
those mismatch parameters?

block.

【在 s******g 的大作中提到】

: You need use mismatch instead of process variation in your statistical block.
: Process variation will apply the variation to all devices.
: Mismatch apply different variation for each instance.

s*g2007-09-28 07:09

4 楼

Mismatch variation is not like process variation.
It is normally a function of mismatch parameter and size of device.
You need ask your foundary to provide, otherwise it won't be accurate.

chart
write

【在 z*b 的大作中提到】

: but my model does not include mismatch parameters which is provided by chart
: ered. should I write the mismatch parameters by myself? but how do to write
: those mismatch parameters?
:
: block.

z*b2007-09-28 07:09

5 楼

哎~~~~foundry的技术支持找了10来天了，发了好几封信了，这个推那个那个推给这个。
对于chartered的这些人真的。。。。。。。。

【在 s******g 的大作中提到】

: Mismatch variation is not like process variation.
: It is normally a function of mismatch parameter and size of device.
: You need ask your foundary to provide, otherwise it won't be accurate.
:
: chart
: write

e*l2007-09-28 07:09

6 楼

Is monte carlo analysis simulation necessary? Do any industry friends answer
this question? thanks. I just finish my chip(A/D converter) measurement.The
performance is good. No monte carlo simulation for my design. By the way, I
am a student.

j*j2007-09-28 07:09

7 楼

if no monte carlo analysis, how can you estimate how many chips can work if
you fab thousands of chips ? testing by hand one by one??

answer
The
I

【在 e*****l 的大作中提到】

: Is monte carlo analysis simulation necessary? Do any industry friends answer
: this question? thanks. I just finish my chip(A/D converter) measurement.The
: performance is good. No monte carlo simulation for my design. By the way, I
: am a student.