The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> AMS Simulators >> trise/tfall in transition filter problem
https://designers-guide.org/forum/YaBB.pl?num=1176890273

Message started by morozmoroz on Apr 18^th, 2007, 2:57am

Title: trise/tfall in transition filter problem
Post by morozmoroz on Apr 18^th, 2007, 2:57am

Hi, everyone.
I have now a little problem, concerning mixed signal simulation.
I can not set rise and fall time for transition function for modules written in VerilogAMS less than 2ns. I have warning in SimVision console:
"../INV/verilogams/verilog.vams"
18: instance (INV+verilogams+0x10000001
behavioral) 3 of INV+verilogams+0x10000001:
Value specified for rise or fall time (1 ns) for transition filter is
too small. Value set to 2 ns.
This message will not be repeated for this filter.
How can I overcome this problem?

Title: Re: trise/tfall in transition filter problem
Post by Geoffrey_Coram on Apr 18^th, 2007, 3:52am

Do you have a default timestep/timescale for the digital blocks? Probably the analog section can't be too much smaller than that.

Title: Re: trise/tfall in transition filter problem
Post by morozmoroz on Apr 18^th, 2007, 4:08am

Geoffrey_Coram wrote on Apr 18^th, 2007, 3:52am:

Do you have a default timestep/timescale for the digital blocks? Probably the analog section can't be too much smaller than that.

No, I haven't.
Additional: I use UltraSim. This schematic isn't big, just 4 simple units. They are written in VerilogAMS and Verilog.
Could it be that my ncsim is very old and this couses problems? Our IT now are updating UltraSim to the IUS5.8
s004 version, may be it will help.
Fill free to suggest.
Thanks in advance!

Title: Re: trise/tfall in transition filter problem
Post by morozmoroz on Apr 18^th, 2007, 5:01am

Huh, I have solved the problem.
The root is in transient simulation time. I have used Cadence SourceLink and found:
The minimum rise or fall time for a transition filter in a transient
simulation is given by:
Tstop - Tstart / 500e6
I have used 1 sec and got 2ns minimum rise time.

Conclusion - using the SorceLink is very useful:)

Title: Re: trise/tfall in transition filter problem
Post by Andrew Beckett on Apr 19^th, 2007, 3:32am

BTW, there is a parameter transres which can be used to relax this requirement (don't push it too far though):

Code:

27 transres=1e-9*stop s
Transition resolution. The transient analysis attempts to stop at corners of
input waveforms. (ex. corners of rising/falling edge of a pulse). If such
events occur within a time less than transres, the analysis will combine the
events into one and force only one time point. The rest of the steps will be
determined by error control. This may lead to loss of detail.

The above is from the spectre help on tran, but it can also be used with AMS.

You need to be careful if you tighten this that you don't run out of numerical headroom on the timestep resolution - with sufficient accuracy to follow the transition.

Regards,

Andrew.