Designer’s Guide Community

‹ Previous topic | Next topic ›

Pages: 1

How to run Envelope Analysis of Cadence Spectre without "Internal error (Read 16162 times)

pancho_hideboo

Senior Fellow

Offline

Posts: 1424
Real Homeless

How to run Envelope Analysis of Cadence Spectre without "Internal error
Dec 28^th, 2009, 3:49am

I have question about how to run Envelope Analysis of Cadence Spectre without failure.

Simulation scenario is very simple.

PA is driven by MSK RF signal, e.g. like following MSK RF signal.
http://www.designers-guide.org/Forum/YaBB.pl?num=1195226387/9#9
However PA is power on and off during simulation by control logic signal "master_en".

"master_en" rises at 0.1usec with Trise=0.1usec and falls at 280.1usec with Tfall=0.1usec.

But Shooting-Envelope Analysis of Cadence Spectre fails at 280.714usec like following.
At 280.714usec, PA is enoughly power off state.
Why does Shooting-Envelope Analysis of Cadence Spectre fail at this point ?

If I use HB-Envelope Analysis of Cadence Spectre, it fails more early,e.g. at around 22usec.
Results HB-Envelope Analysis between 0usec and this 22usec are completely wrong.
On the other hand, results of Shooting-Envelope Analysis between 0usec and 280.683usec seem to be reasonable.

Of course, there is no NPORT which Cadence Spectre have many troubles in treating.
http://www.designers-guide.org/Forum/YaBB.pl?num=1258119869
http://www.designers-guide.org/Forum/YaBB.pl?num=1237187544

How to run Envelope Analysis of Cadence Spectre without failure ?

Any expert of Cadence Spectre, teach me.

[Error log of Cadence Spectre] Quote:

EnvStepIndex = 9592, envTime = 280.62 us, completed = (93.5 %)
EnvStepIndex = 9593, envTime = 280.651 us, completed = (93.6 %)
EnvStepIndex = 9594, envTime = 280.667 us, completed = (93.6 %)
EnvStepIndex = 9595, envTime = 280.683 us, completed = (93.6 %)

Internal error found in spectre at time = 280.714 us during envelope following analysis `envlp'. �Please run `getSpectreFiles' or send the netlist, the spectre log file, the behavioral model files, and any other information that can help identify the problem to support@cadence.com.
� �FATAL (CMI-2010): Assertion failed in file `tranAnalysis.c' at line 412.
� �FATAL (SPECTRE-21): Assertion failed.

[Extraction from my Netlist of Cadence Spectre] Quote:

parameters en=1 vdd_core=1.55 frf=2450M

ahdl_include "$CDSHOME/tools/dfII/samples/artis/rfLib/IQ_modulator/veriloga/veriloga.va"
subckt V_ZigBee_Sig_Gen minus plus
� �SRC_Iref (bb_I 0) vsource type=pwl scale=0.15 file="~/AHO/DATA/I_all_burst.pwl" �
� �SRC_Qref (bb_Q 0) vsource type=pwl scale=0.15 file="~/AHO/DATA/Q_all_burst.pwl"
� �I0 (bb_I bb_Q RFsig 0) IQ_modulator I_gain=0 Q_gain=0 I_cpdb=100 \
�Q_cpdb=100 rin=50 rout=50 nf=0 quad_error=0 flo=frf
� �R0 (RFsig 0) resistor r=50 isnoisy=no
� �E0 (plus minus RFsig 0) vcvs gain=1.0
ends V_ZigBee_Sig_Gen
// End of subcircuit definition.

SRC_in_p (0 in_p) V_ZigBee_Sig_Gen
SRC_in_n (in_n 0 in_p 0) vcvs gain=-1.0

I0 ( in_n in_p master_en out vdd 0 ) heppoko_pa

V0 (vdd 0) vsource dc=vdd_core type=dc
V1 (master_en 0) vsource dc=en*vdd_core type=pwl delay=100n \
� wave=[ 0 0 100n (en*vdd_core) (280u) (en*vdd_core) (100n+280u) 0 ]

PORT0 (out 0) port r=50 type=dc mag=0

simulatorOptions options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 temp=25.0 \
� �tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 rforce=1 maxnotes=5 maxwarns=5 \
� �digits=5 cols=80 pivrel=1e-3 sensfile="../psf/sens.output" checklimitdest=psf

envlp envlp harms=3 stop=0.3m fund=frf errpreset=liberal \
� �envmaxstep=(1.0/2M)/16 fixstepsize=no annotate=status outputtype=both restart=no

If I use Envelope Analysis based on Time-Balance(TB) Steady State Solver in Agilent GoldenGate,
I can finish envelope analysis even with including noise effects without fail.
http://www.designers-guide.org/Forum/YaBB.pl?num=1232855139

Quote:

Beginning Large-Signal Carrier Analysis for Noise...
--------------------------------------------------------------------------------

Steady-state solver �type � �= Time-Balance (TB)
Number of fundamental tones �= 1
fundamental tone 1 � � = 2.4500GHz, 7 harmonic(s)
Convergence error = 0.1 (conv_err)
Fixed time step �= yes (time_domain_hb_variable_time_step=0)
Nyquist rate oversample factor � � = 2 (time_domain_hb_over_sample=4)
Gear Integration order = 0 (time_domain_hb_integ_order )
Number of stabilization periods � �= 0 (time_domain_hb_relaxation_nb_period)
Newton Jacobian factor type �= Krylov
Max number of krylov iterations � �= 500 (krylov_max_iter)
Krylov tolerance heuristics �= robust (krylov_dynamic_tol = 0)
Continuation mode = On (subsequent HB always use available previous solution)
--------------------------------------------------------------------------------

Small-signal solver type � � = Frequency-domain
Preconditioner type � �= time-domain
Small-signal accuracy �= Krylov tolerance 1 (rtol=1.000000e-05)
Krylov max iteration number �= 500
Krylov restart iteration number � �= 100

Beginning Fast Envelope Transient...
--------------------------------------------------------------------------------

Fast envelope level � �= 1
Accuracy parameters � �= [Var = 1.0]
Integration method � � = Steady-State
Number of envelope sources � = 2
Number of output probes = 6
Number of output frequencies = 4
Number of output time steps �= 9599
Output start time = 0.000 sec.
Output stop time = 300.000usec.
Output time step = 31.250nsec.

« Last Edit: Dec 28^th, 2009, 8:08am by pancho_hideboo »

Kita━━━━━━(ﾟ∀ﾟ)━━━━━━ !!!!!
http://www7.plala.or.jp/ungeromeppa/flash/kita.html
http://www.youtube.com/watch?v=mjIxGh55bMM&feature=related

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: How to run Envelope Analysis of Cadence Spectre without "Internal error
Reply #1 - Jan 3^rd, 2010, 12:36pm

Hi Pancho Hideboo,

I couldn't find any reports of similar problems - but it was a little hard to trace with the information you gave (you didn't mention the version number, command line arguments, etc etc).

What version of spectre were you using? There were some problems with SST2 as the output format for envlp (fixed earlier this year), so maybe it's that - although the symptoms were slightly different.

Essentially the error is an uncaught non-specific internal error, so really needs the testcase to investigate further.

If you're not using the most recent MMSIM version, I suggest you try that first, then contact Cadence Customer Support - this looks like a good old-fashioned bug (all software has occasional bugs, of course), with the data to reproduce the problem, and then we can get it fixed.

Regards,

Andrew.

(you couldn't resist the temptation to have a dig about nport, despite the fact that this has nothing to do with that!)

IP Logged

pancho_hideboo

Senior Fellow

Offline

Posts: 1424
Real Homeless

Re: How to run Envelope Analysis of Cadence Spectre without "Internal error
Reply #2 - Jan 3^rd, 2010, 11:34pm

Andrew Beckett wrote on Jan 3^rd, 2010, 12:36pm:

I couldn't find any reports of similar problems

This shows that there is very few people who try to use envelope analysis, NPORT, etc. of Cadence Spectre in actual design.
And they didn't compare it's results with other simulators.
http://www.designers-guide.org/Forum/YaBB.pl?num=1261540239

Andrew Beckett wrote on Jan 3^rd, 2010, 12:36pm:

There were some problems with SST2 as the output format for envlp (fixed earlier this year)

I don't use "sst2" as output format, because I run envlp analysis from unix command line with specifying "psfbin".

Andrew Beckett wrote on Jan 3^rd, 2010, 12:36pm:

you couldn't resist the temptation to have a dig about nport, despite the fact that this has nothing to do with that!

pancho_hideboo wrote on Dec 28^th, 2009, 3:49am:

Of course, there is no NPORT which Cadence Spectre have many troubles in treating.
http://www.designers-guide.org/Forum/YaBB.pl?num=1258119869
http://www.designers-guide.org/Forum/YaBB.pl?num=1237187544

Originally, my test circuit have NPORT, but envelope analysis of Cadence Spectre results was "blow up" and failed, as I expected.
So I replaced NPORT with simple LCR equivalent circuits.

My results here are envelope analysis without NPORT.

I don't think NPORT, envelope analysis and QPSS of Cadence Spectre can all reach to practical use level.

Kita━━━━━━(ﾟ∀ﾟ)━━━━━━ !!!!!
http://www7.plala.or.jp/ungeromeppa/flash/kita.html
http://www.youtube.com/watch?v=mjIxGh55bMM&feature=related

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: How to run Envelope Analysis of Cadence Spectre without "Internal error
Reply #3 - Jan 4^th, 2010, 4:08am

I'm sorry, but this is ridiculous. There are many thousands of users of spectre who are using it in real design. I know you find that amazing, but it is absolutely true.

You clearly are completely biased, and seem to have a large chip on your shoulder about something, and are not interested in anyone who wants to help. As I keep saying, all tools have occasional problems, and if you are continuing to use old versions, or base your information on historical information, you are doing all other users a disservice.

I really value constructive criticism - there are always ways that tools can be improved (I'm always asking for enhancements and fixes), but constant negativity helps nobody.

And by the way, you still didn't say what version you were using. Will you contact Cadence Customer Support about this so we can fix the internal error (assuming it has not already been fixed)? (I still want to help, despite being rather irritated by your comments)

Regards,

Andrew.

IP Logged

Pages: 1

‹ Previous topic | Next topic ›