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.
It depends on domain application you are looking / working!
If Digital & Analog HSPICE is very good & accurate..
I have not used Spectre, but heard that is OK for RF stuff..?
But My experience...HSPICE now completely Integrated in AWR Analog Office...
Hence there no prblem even for RF stuff...
i think that hspice is better in terms of accuracy also.
But spectre is faster than hspice and easy to use because it can be used interactively with cadence.
The best analog simulator is the one which you forget about his name. The future in analog design is a complete encapsulation of the
(netlist, comand)->result
engine into in enviroment which concentrate on the main analog design tasks. The cadence ADE is a small step in this direction. It need much more progress to to run corner, sensitivity, mismatch, hiearchical embedding, modelling, background verification and all task of >100k device analog projects in more automated and productive enviroment. The past research effort in analog design automation concentrate on synthesizers for a limited class of common used IPs. I did not see any approach, beside one cadence tool, to ask the question from the tools side what is the spec of the project/subproject/circuit. If a specification driven analog design enviroment is used many tasks could be automated because the circuit targets are formulated to be automatic verified by the machine. Also a limited insight into circuit topology could be supported today. For example the automatic identification of matching groups and the the automatic analysis of the mismatch impact could be done. Further possible tools are automatic dimensioning with abstract targets like maximum DC current, VDSAT, branch current distribution optimisation, floating node identification for power down.
There is so much lacking that the question which simulator is better target only a very minor area of improvement for the analog designer.
Did any one heard of ADiT from EverCAD? It said it is a fast spice simulator and can compete to Hsim from Nassda. Now Nassda is dead, can it occupy the position? Tell me any information if any one knows about it?
I use spectre for all first design and modification. And then I use Hspice for final post simulation. I use the extracted netlist after the layout was finished.
I have already said that few times but in my opinion the difference between the tool itself is of very low importance. One can run faster than the other or is a bit more precise what is hard to decide when you are designing IC and do not have any parasitics there...
To me what is important are the models. If the models are extracted by foundry for SPECTRE I want to use Spectre just because I believe the models more. If they tweak models for HSpiCE or ELDO then those models are probably ok but I think their accuracy is little lower.
Also I would prefer to use the simulator most foundry users are using (usually spectre) just because I believe that if 100s of engineers run simulations that it is probably safe to say that there are no big issues.
But that's just me - I want my design to work and I don't care what tool I am going to use.
i really really agree with Teddy. when simulating, the important thing is not the simlation tools but the model you get. the result IC chip work well or not depend on the accuracy of the model.
I have a few question to ask since there are so many HSPICE expert here.
could you guys give me examples on what is "convergence" mentioned above in HSPICE, what is it purpose.
Regarding the accuracy, what are the difference between other version of SPICE, any benchmark? i think for not so complicated circuit design, the results is almost the same.
I use other tools to generate the .cir files to run in HSPICE.
hspice has many control function to control the speed and convergence. Spectre can run with verilogX so is more suitable with mix design and it also provide rf simulation function e.g. phase noise
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
