Referencing a model, not a model instance

Well, a better way is to use the new noiseoff_type/noiseon_type option that was introduced in MMSIM151 ISR10 (and also SPECTRE161). In IC617 ISR8 and later, this shows up on the Simulation->Options->Analog form:

So this way you can turn off flicker noise, although not absolutely every device supports controlling the flicker noise, but most modern devices do.

If you don't have this on the form, use the Additional Arguments field and set it to noiseoff_type=[flicker]

Regards,

Andrew

I'd be surprised if the PDK is that fussy about precise simulator versions, but worth checking nevertheless.

I'm not sure why your alter statements weren't working - I'd have to see the models to understand precisely what you need to do and what's not working; perhaps you could contact customer support (although I suspect by far the best option would be to use a newer simulator version; after all, if  you're disabling the flicker noise, one shouldn't get too hung up on any difference in results you get by using that version as that's likely to be a smaller effect than turning off some of the noise sources!)

Regards,

Andrew.

Hi Andrew,

Thanks - I am trying to go the new MMSIM route (and hope everything will work with the PDK).

I asked our CAD administrator to install the latest MMSIM version. But it contains Spectre 15.1.0 and its version is "MMSIM 15.10.585". Supposedly it is the latest available.
I am a bit confused about "MMSIM151 ISR10" - is there an alternative version scheme for this that matches the scheme above?

Thank you!

MMSIM subversion numbers show a build number (which increments in fairly big chunks), but if you run "spectre -W" it will also add the subversion on the end.

Looking at the downloads.cadence.com site I see that ISR10 is still listed:

Looks as if your CAD administrator hadn't installed the latest (which is sub-version 15.1.0.679.isr14, and shows on downloads as MMSIM15.10.679 - the manufacturing number on downloads is a slightly different format than the version number in the simulator itself, but it's aligned), but precisely ISR10.

Regards,

Andrew.

The versions are fine. If you use spectre -W it shows the subversion number, whereas spectre -V it shows the major version umber and build date. This is a hotfix of MMSIM15.1 which is why it came out during 2016. It wouldn't be 16.1 - not sure why you think that.

If you get a warning, then I think it could be that you aren't really running that version, or you've misentered the setting. Note I think you also have to use noiseoff_inst at the same time for this to have an effect. 

What version does it show at the top of the spectre log file? What do you have in the options/analysis statements in the input.scs? What exactly does the warning show?

You might also want to take a look at these two solutions:

Regards,

Andrew

Hi Andrew,

I again played around for an hour. I managed to get rid of the warning but the option still need to be ignored.

I think the spectre version is correct, the noiseoff_type option is known in "spectre -h options". I read your reference and I am sure I am using it correctly.

When I use noiseoff_type=flicker (or [flicker] or any other variation coming to my mind) all the noise is turned off (incl thermal). Without it, I get flicker. So it seems the option is still ignored.

What I am wondering: Why does spectre log list my option as noiseoff_type=1? What does 1 mean?

Cadence (R) Virtuoso (R) Spectre (R) Circuit Simulator
Version 15.1.0.585.isr10 64bit -- 18 Aug 2016
Copyright (C) 1989-2016 Cadence Design Systems, Inc. All rights reserved worldwide. Cadence, Virtuoso and Spectre are registered trademarks of Cadence Design Systems, Inc. All others are the property of their respective holders.

Includes RSA BSAFE(R) Cryptographic or Security Protocol Software from RSA Security, Inc.
[...]
Notice from spectre during initial setup.
    The noise for all instances in this netlist has been turned on.
    The noise of the subckt instance `N1' is turned off.


Global user options:
             reltol = 0.001
            vabstol = 1e-06
            iabstol = 1e-12
               temp = 27
               gmin = 1e-12
             rforce = 1
           maxnotes = 5
           maxwarns = 5
             digits = 5
               cols = 80
             pivrel = 0.001
           sensfile = ../psf/sens.output
     checklimitdest = psf
      noiseoff_type = 1
      noiseoff_inst = N1
               save = allpub
     subcktprobelvl = 2
               tnom = 25
             scalem = 1
              scale = 1

Scoped user options:

Circuit inventory:
              nodes 8
             iprobe 16    
             assert 12    
               ccvs 1     
              utsoi 2     
            vsource 5     

Analysis and control statement inventory:
                 dc 1     
               info 7     
              noise 1     

Output statements:
             .probe 0     
           .measure 0     
               save 0     

Time for parsing: CPU = 79.988 ms, elapsed = 94.121 ms.
Time accumulated: CPU = 439.932 ms, elapsed = 599.417 ms.
Peak resident memory used = 59.1 Mbytes.

~~~~~~~~~~~~~~~~~~~~~~
Pre-Simulation Summary
~~~~~~~~~~~~~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~~~~

Warning from spectre.
    WARNING (SPECTRE-16707): Only tran supports psfxl format, result of other analyses will be in psfbin format.
[...]

This is the relevant section of the netlist:

simulatorOptions options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 temp=27 \
    tnom=25 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 noiseoff_type=[flicker] noiseoff_inst=[ N1]
dcOp dc write="spectre.dc" maxiters=150 maxsteps=10000 annotate=status
dcOpInfo info what=oppoint where=rawfile
noise ( vo 0 ) noise start=1 stop=1T iprobe=V1 annotate=status
modelParameter info what=models where=rawfile
element info what=inst where=rawfile
outputParameter info what=output where=rawfile
designParamVals info what=parameters where=rawfile
primitives info what=primitives where=rawfile
subckts info what=subckts where=rawfile
saveOptions options save=allpub subcktprobelvl=2

The circuit contains exactly one PMOS (not connected to output node) and one NMOS ("N1"). Both are utsoi, supposed to be supported for noiseoff_type according to spectre manual.

I am starting to think about a bug ... ?

OK, a couple of things here. First, the reporting of noiseoff_type in the log file. This is not the friendliest reporting in the log file, but the way that all options in spectre which have names as option values work is that they're all actually enumerated. So if I look at "spectre -h options" for this option:

33 noiseoff_type=all
The noiseoff_type defines the noise sources which are
disabled by the noiseon/off_inst option. The default
value is all. Possible values are thermal, flicker,
shot, ign, and all.

This means that thermal=0, flicker=1, shot=2, ign=3 and all=4. It will report these numbers - just a single number if one is selected; if more than one is selected it will show (say) [0 1] to indicate that thermal and flicker has been selected. 

There's a (strong) argument to say we should report this better using the words corresponding to the choices. You could (for example) specify these options as numbers, but the words are much more friendly.

As for the flicker noise not working, I just tested this (with MMSIM15.1 ISR10) using bsim4 models, and I could disable flicker noise for my subckt (so it doesn't just have to be a leaf instance) and also thermal (I tried thermal and flicker). I did find that the flicker noise for one of the unsupported devices still appeared within that subckt, but for the transistors it was gone.

Now, according to the second link I posted above, utsoi is supposed to be one of the supported devices (that's the model you're using from your circuit inventory), so I'd expect it to work. Unfortunately I don't have any utsoi models handy that I can try (and I'm rather short of time this week).

So, I'd suggest that you:

1. Try with a newer versions still (maybe the initial version didn't support utsoi; the feature was officially rolled out in SPECTRE161 I believe, so it may be worth checking a SPECTRE161 version instead of MMSIM151).
2. If that doesn't work, contact customer support to see if we can identify what is going wrong; then an AE can spend time on investigating it.

Thanks,

Andrew.