Designer’s Guide Community

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

stb and noise analysis fail at 125C
Jan 03^rd, 2012, 4:18am

Hi Guys,

I ran dc, ac, stb, and noise analysis for a voltage buffer that is implemented in 40nm LL technology. The stb and noise fail in almost all corners at 125 degree Celsius, but dc and ac analysis have good results.

For example, the stb and noise analysis fail at 125 degree Celsius if all devices are in ss corner. But I get expected stb and noise results at 124 degree Celsius. In other corner combinations, I cannot get correct stb and noise analysis results even at 124 degree Celsius.

�

Back to top

RCpara_125C.PNG

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: stb and noise analysis fail at 125C Reply #1 - Jan 3^rd, 2012, 4:21am This is the correct results at 124 degree Celsius. I think this is probably a spectre model fault. But what can I do to verify it? Best Regards, Yawei
Back to top	RCpara_124C.PNG IP Logged

rfidea Senior Member Offline Posts: 159 Europe	Re: stb and noise analysis fail at 125C Reply #2 - Jan 3^rd, 2012, 12:32pm Looks definitly like a model/simulator error, or a combination. You should send the test case to either the simulator vendor or the PDK vendor. Usually there is very difficult for the user to know in which the bug is.
Back to top	IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: stb and noise analysis fail at 125C Reply #3 - Jan 3^rd, 2012, 3:45pm did you check the output file regarding some warnings like some kind of jakobian error, once i got this kind of error which some kind of convergence error, try tp initialize some node may be help full. Thanks, Raj.
Back to top	IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: stb and noise analysis fail at 125C
Reply #4 - Jan 3^rd, 2012, 7:59pm

I find that Operating points had not been recomputed when stb and noise failed. It seemed that the simulator found DC solution for AC analysis and skipped this step for the succeeding stb and noise analysis.
Quote:

******************************************
AC Analysis `ac': freq = (1 Hz -> 100 GHz)
******************************************
Finding DC approximate solution failed. Try again with try_fast_op set to no.
Trying `homotopy = gmin'.
Trying `homotopy = source'.
Trying `homotopy = dptran'.

...
...

**************************************************
Stability Analysis `stb': freq = (1 Hz -> 100 GHz)
**************************************************

Notice from spectre during STB analysis `stb'.
Operating points will not be recomputed: the circuit and conditions were not changed since previous analysis.

...
...

************************************************
Noise Analysis `noise': freq = (1 Hz -> 100 GHz)
************************************************

Notice from spectre during Noise analysis `noise'.
Operating points will not be recomputed: the circuit and conditions were not changed since previous analysis.

I select DC analysis and stb only in ADE-L. Now the simulator finds DC solutions for stb analysis and gives correct loop response.

Yawei

Back to top

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: stb and noise analysis fail at 125C Reply #5 - Jan 3^rd, 2012, 8:01pm Tools are spectre sub-version �10.1.1.056.isr5, and IC6.1.4.500.9. I choose APS and Auto for Mutlithreading options. Yawei
Back to top	IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: stb and noise analysis fail at 125C Reply #6 - Jan 9^th, 2012, 3:26am This problem has been resolved when I replace spectre simulator with sub-version 11.1.0.249.isr2. Yawei
Back to top	IP Logged