The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> AMS Simulators >> AMSDesigner's speed
https://designers-guide.org/forum/YaBB.pl?num=1150384120

Message started by Cri Azzolini on Jun 15^th, 2006, 8:08am

Title: AMSDesigner's speed
Post by Cri Azzolini on Jun 15^th, 2006, 8:08am

Hi all,

I recently perfomed mixed-mode simulation of a design containing both analog (schematic) and digital (verilog) section with AMS-Designer. I noted that the same transient simulation performed on a schematic containing only the analog section of the design takes more time with Spectre: the difference is huge, about 50% or maybe more!
The simulation is run on the same workstation and the simulation queue was empty in both cases.

I cannot explain how this may happen: I mean, as far as I know, AMSDesigner uses Spectre for the simulation of the analog part of the circuit!
Any ideas?

Thanks,
Cri

Title: Re: AMSDesigner's speed
Post by Geoffrey_Coram on Jun 16^th, 2006, 5:05am

Does the digital block control the analog, and if so, how did you replace it in the analog-only simulation?

Title: Re: AMSDesigner's speed
Post by Cri Azzolini on Jun 16^th, 2006, 5:24am

Hi Geoffrey,

the analog block drives a digital post-processing digital machine. Actually there is some feedback signal from the digital block to the analog core. Does it make difference?
In the analog-only simulation I simply delete the digital block and no post-processing is done: in this way the output analog data are raw and not corrected by the digital machine.

Thanks, bye!
Cri

Title: Re: AMSDesigner's speed
Post by Geoffrey_Coram on Jun 19^th, 2006, 5:52am

I haven't used AMSDesigner, so I don't know if it runs Spectre in a low-accuracy mode or something ... What does the feedback do? In the analog-only simulation, if this signal is fixed at a "bad" bias point (I'm assuming it's not completely missing), then you could be balancing the analog circuitry at a sharp nonlinearity that could be causing convergence problems in the analog solver. But that's pure conjecture.

Title: Re: AMSDesigner's speed
Post by Cri Azzolini on Jun 20^th, 2006, 12:00am

Yes, I understand what you suppose about the "bad" bias point: it could be a possible reason. I am going to verify!
Thanks for your kind reply and regards,

Cri