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.

a question on the Mote Carlo analysis with hspice

Status
Not open for further replies.

John Xu

Member level 3
Joined
Jul 22, 2005
Messages
59
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
1,938
hspice monte carlo

Hi,
I am doing the Monte carlo analysis for my comparator circuiut. To get the more accurate statistical charercter, I run 1200 times for Monte Carlo analysis. But after running some time, I found the CPU become slower and slower. After checking, I found the hpice generate a file named "monte_comparator.sw0". It records the detailed values for every sweep and every node. So it is very very large, even up to 1.2G. I doubt it caused the slow run for my simulation. Anyone can tell me how to solve this issue? e.g., to remove this file after every run?

Thanks in advance
 

flamingo

Member level 4
Joined
Jan 10, 2005
Messages
73
Helped
6
Reputation
12
Reaction score
5
Trophy points
1,288
Activity points
717
John Xu said:
it is very very large, even up to 1.2G. I doubt it caused the slow run for my simulation. Anyone can tell me how to solve this issue? e.g., to remove this file after every run?
First I think 1.2G is not too large to beyond ordinary computer's capability. Try a better machine.:D
Second, *.sw? is the file that record your DC sweep results, if removed it away, hehe, you know the aftermath. A feasible method is to record some selected nodes output information other than all of nodes. Use probe option in your option command, and probe the nodes signal whatever your like to in your probe line. e.g.:
.option probe
.probe dc v(1) v(2,3) v(4).......
.probe ac v(1) v(2,3) v(4.......

PS: do you really need a run of 1200 times MC, it's so extravagance, maybe several hundred is enough. Good luck.
 

John Xu

Member level 3
Joined
Jul 22, 2005
Messages
59
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
1,938
[quote="flamingoSecond, *.sw? is the file that record your DC sweep results, if removed it away, hehe, you know the aftermath.
PS: do you really need a run of 1200 times MC, it's so extravagance, maybe several hundred is enough. Good luck.[/quote]

Hi flamingo,
Thanks for the reply! Let's elaberate my issue more clearly. Pls. see the below script.

***********
.DC Y 2.41 2.60 0.0001 2.60 2.41 -0.0001 SWEEP MONTE=500
.OPTION PROBE
.probeV(OUT_91_1)


.MEAS DCTHN_91 Find V(IN2) WHEN V(OUT_91_1)=2.5 fall=1
.MEAS DCTHP_91 Find V(IN2) WHEN V(OUT_91_1)=2.5 RISE=1
.MEAS VOS_91 PARAM='(THP_91+THN_91)*0.5-2.5'
.MEAS VWIN_91 PARAM='THP_91-THN_91'
***********
I use this cript to investigate the mismatch analysis for my comparator. What I want is the data postprocessed by measure command after every MC analysis. Then I hope to release the *sw file for every MC analysis. Remove them will not affect my postprocessing but will relieve more memory.

I used this function in other simulators, but not know how to use it in hspice.

Pls. give me some instructions.

Thanks
 

beargebinjie

Junior Member level 3
Joined
May 4, 2005
Messages
27
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,521
hello john xu, coulde you tell me how to extract the process viariations from the process lib before you do the monte_carlo analysis, thanks you very much!
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Top