Hspice Vs. Spectre which is better ?

[analogweb@yahoo.com]

Hi All,
I am trying to reach to a conclusion on the subject line before I
finalise one of them for my new project.
I have been a rigorous Hspice user so far and used Spectre now and
then. Both the simulators come with their advantages and
disadvantages, and this precisely make the choice difficult.
** Biggest advantage of Spectre:
a. Tightly integrated within Cadence DFII framework
** Biggest dis-advantage of spectre:
a. Seemingly simple tasks like sweeping multiple parameters across
corners requires 500 mouse clicks , not to mention a license for
doing process corner analysis !!!
b. The GUI based approach "hides" the different options,
specifically related to time stepping .. integration method and stuff

** Biggest advantage of hspice:
a. Single license for almost everything except the hspiceRF stuff
(phase noise and things).
b. Mutiple sweeps across corners a cakewalk
** Biggest Dis-advantage
a. Prone to simple netlisting errors.

To me doing a simple thing like:

..dc DATA=<SOME TWO VARIABLE DATA> sweep monte=30
..alter slow
...
..alter fast

is a big headache in Spectre and I have to buy two additional licenses
(one for monte carlo and one more for corner analysis)

Lets debate

-- Alex

 

[Svenn Are Bjerkem]

On Sep 4, 8:51 am, "analog...@yahoo.com" <analog...@yahoo.com> wrote:

Lets debate

Since we have the licenses needed to do corner simulations, I haven't
really investigated into their need when running spectre on the
command line.HSpice is a command line tool, spectre is IMHO not.
spectremdl is an attempt to achieve the same functionality, but the
work needed is a bit more. Corners can be done with aliases in mdl and
the measurements are a bit clearer in mdl as they resembles the
function you would use in ocean. Writing measurements in HSpice is
quite tedious, I think.

I am experimenting with ocean from within DFII. This give a very
powerful environment. With an enhancement of the ocean-menu posted on
this group I can have many simulations in sequence with results saved
to different directories which can be easily loaded with the results
load function in ADE for use in AWD.

--
Svenn

 

[Joerg]

analogweb@yahoo.com wrote:

Hi All,
I am trying to reach to a conclusion on the subject line before I
finalise one of them for my new project.
I have been a rigorous Hspice user so far and used Spectre now and
then. Both the simulators come with their advantages and
disadvantages, and this precisely make the choice difficult.
** Biggest advantage of Spectre:
a. Tightly integrated within Cadence DFII framework
** Biggest dis-advantage of spectre:
a. Seemingly simple tasks like sweeping multiple parameters across
corners requires 500 mouse clicks , not to mention a license for
doing process corner analysis !!!
b. The GUI based approach "hides" the different options,
specifically related to time stepping .. integration method and stuff

** Biggest advantage of hspice:
a. Single license for almost everything except the hspiceRF stuff
(phase noise and things).
b. Mutiple sweeps across corners a cakewalk
** Biggest Dis-advantage
a. Prone to simple netlisting errors.

To me doing a simple thing like:

.dc DATA=<SOME TWO VARIABLE DATA> sweep monte=30
.alter slow
..
.alter fast

is a big headache in Spectre and I have to buy two additional licenses
(one for monte carlo and one more for corner analysis)

Lets debate

-- Alex

--
Regards, Joerg

http://www.analogconsultants.com

 

[okguy]

In Europe, many companies such as ST prefer Eldo.
Did any Uni performe a test benchmark ?
Several test cases would be nice :
- RF LNA
- PLL
- ADC or DAC
- simple AOP

 

[Andrew Beckett]

On Mon, 03 Sep 2007 23:51:04 -0700, "analogweb@yahoo.com" <analogweb@yahoo.com>
wrote:

Hi All,
I am trying to reach to a conclusion on the subject line before I
finalise one of them for my new project.
I have been a rigorous Hspice user so far and used Spectre now and
then. Both the simulators come with their advantages and
disadvantages, and this precisely make the choice difficult.
** Biggest advantage of Spectre:
a. Tightly integrated within Cadence DFII framework
** Biggest dis-advantage of spectre:
a. Seemingly simple tasks like sweeping multiple parameters across
corners requires 500 mouse clicks , not to mention a license for
doing process corner analysis !!!
b. The GUI based approach "hides" the different options,
specifically related to time stepping .. integration method and stuff

** Biggest advantage of hspice:
a. Single license for almost everything except the hspiceRF stuff
(phase noise and things).
b. Mutiple sweeps across corners a cakewalk
** Biggest Dis-advantage
a. Prone to simple netlisting errors.

To me doing a simple thing like:

.dc DATA=<SOME TWO VARIABLE DATA> sweep monte=30
.alter slow
..
.alter fast

is a big headache in Spectre and I have to buy two additional licenses
(one for monte carlo and one more for corner analysis)

Lets debate

-- Alex

I'm not going to give to much of an opinion here because I obvious have an
interest in one of them (working for Cadence), but I have used both simulators
in my career extensively.

Be vary careful in your comparison. You are not comparing like with like in much
of the above. With most of the above, you're comparing spectre in ADE with
hspice from the command line. You can run spectre from the command line very
successfully, and you can run hspice in ADE. Much of what you're saying about
licensing is related to the fact that you're using the ADE interface, not what
spectre does.

Spectre itself only has two license options - spectre, and spectreRF (from
MMSIM62, these are called spectre-L and spectre-XL).

You can run alters and monte-carlo with spectre from the command line, and it
has a rich measurement language which is very consistent. It can also run with
SPICE-syntax netlists and control files.

Regards,

Andrew.
--
Andrew Beckett
Senior Solution Architect
Cadence Design Systems, UK.