Different results in Hspice and Spectre

[wjlzhx]

which will i trust hspice or spectre?
I simulate a circuit--osc

using spectre i get the rusult ,frequcy is 22k
but using hspiec i can not get the rusult ,it can not osc .

when i simulate with hspice, there are no error.

so which i can trust??

 

[Ruritania]

Re: hspice vs spectre

Hi,

I think you should make sure that all the settings are the same in both the Spectre simulator and Hspice simulator... and, maybe you need to set some initial conditions to make the circuit oscillate in Hspice.

Anyway, Hspice is always the golden standard in the IC industry ...

Hope this helps.

Ruri

 

[xirix]

Re: hspice vs spectre

HSpice - good for analog simulations
Spectre - good enough; used for big digital circuits (very fast vs. HSpice)

 

[bigpop]

Re: hspice vs spectre

You should add some initial condition in hspice to let it oscillate.
If your circuit is not large, hspice and spectre will get same results.

 

[sunny153]

Re: hspice vs spectre

certainly Hspice is the king of analog simulators. As mention by other folks u need to check out the settings..

 

[vbhupendra]

Re: hspice vs spectre

hspice is very accurate. and as others have mentioned you have to give initial conditions. or you can just give a pwl to VDD from 0 to VDD in a small time.

 

[wjlzhx]

hspice vs spectre

thank you for your helps

I set the initial condition like this
.ic v(net108)=0.5v v(net137)=0.5v v(net110)=0.5v
net108,net110,net137 they connect cap to gnd!

but i got nothing.

and i also add the pwl ,but got nothing .

now sometime i got the wrong rusult ,it said "no convegence",is this the modle's problem or the circuit's problem???

 

[Fom]

Re: hspice vs spectre

Hspice or Spectre? It looks like religions.
Hspice believer or Spectre believer.
Why some people consider Hspice is more accurate than other simulators? Any proof except declaration?
May be it was during before BSIM3 era. When Hspice used level 28 MOS model. But today if simulator use BSIM3 models it has the same accuracy as other simulator that use the same model.
If simulator use the same differential equations accuracy is defined by options like reltol, abstol, etc.
Another thing is an algorithm used to find solution. But it affects only on simulation time not accuracy.
BTW, I was very surprised when I didn't find voltage nonlinearity for resistor model in Hspice.

 

[yibinhsieh]

Re: hspice vs spectre

I think this two simulator are good for us. The convergence in spectre is better than the one in hspice. But the waveform view in hspice is more friendly than the one in spectre.

Yibin,

 

[Fom]

Re: hspice vs spectre

yibinhsieh it is just your opinion. I am absolutely disagree with you. If to consider Analog Artist it has the most frendly interface.
Just an example.
For Hspice you have to care about what kind of sophisticated waveform you will plot after simulation and put it in simulation netlist as MEASURE statement. If you forget to do that or you need to plot something else you cannot plot it.
For Spectre (Analog Artist) you don't care about that because you can plot every expression AFTER simulation.

 

[Analog_starter]

Re: hspice vs spectre

xirix Posted: 12 May 2005 9:05 Post subject: Re: hspice vs spectre

--------------------------------------------------------------------------------

HSpice - good for analog simulations
Spectre - good enough; used for big digital circuits (very fast vs. HSpice)


But I found spectre is not faster enough than hspice.

Best Regards
Analog_starter

 

[dendrite]

Re: hspice vs spectre

If you use the large mixed signal circuit. you'll feel the speed.
The third choice is ultrasim. Maybe it's the best.

xirix Posted: 12 May 2005 9:05 Post subject: Re: hspice vs spectre

--------------------------------------------------------------------------------

HSpice - good for analog simulations
Spectre - good enough; used for big digital circuits (very fast vs. HSpice)


But I found spectre is not faster enough than hspice.

Best Regards
Analog_starter

 

[layes2]

Re: hspice vs spectre

you should add
option
.ic
to hspice
hspice is better

 

[simov]

hspice vs spectre

in your case, low frequency oscillator is autonomy circuit, so the simulation result depend on the arithmetic of simulator about autonomy circuit. From my experience, simulate low frequency autonomy oscillator not only in time domain but also in frequency domain, and base on some theoretics calculation and real chip test result

 

[selvaraja]

Re: hspice vs spectre

Hi,

It is trade off if want accuracy go for Hspice .... Spectre for fast & complex circuits

 

[hrkhari]

Re: hspice vs spectre

I think Spectre and HSPICE carries the same acurracy rate due to the common model used. Spectre are much user friendly and has good waveform display. The only thing distinguish both of the simulator is the cost of purchasing it.

Rgds

 

[piao]

hspice vs spectre

I think it is up to the accuracy options and all other options you used in your design.

But Hspice works as a standard in the IC design field.

 

[uncle_urfi]

hspice vs spectre

The supply voltage goes from 0 to Vdd in a short time, as someone mentioned is the solution to the convergence problem.

This works for me when i simulate LC oscillators in spectreRF, but the frequency is too low, and i don't know a direct solution for this in Hspice, if it is indeed a frequency-related problem.

One can also put an initial condition on the currents and check the biasing circuits at start-up. May be there is not enough current to establish DC conditions that will be used for the rest of the simulation.