PSS Question in IC 5.0

Vincent,

This information is (and always has been) irrelevant for PSS. QPSS needs to
know whether the signal was large or moderate, but PSS does not use this
information. You used to have a choice of large, moderate, or small, with the
default set to moderate. Since it made no difference, and PSS treats all signals
as large, the form was changed to only allow large.

Similarly since QPSS only allows large and moderate, small is no longer
available as a choice for QPSS.

Depends on what you're doing.

What frequencies do you have going in - what is the PSS fundamental. What
tolerance settings have you set (reltol, iabstol, vabstol) and pss options
(errpreset, relref, method, maxacfreq, maxstep, no. harmonics, highorder),
How big is the circuit (the circuit inventory at the start of the ouput will
tell you)?

Andrew.

 

Hi there:



I have two questions regarding to PSS simulation in IC 5.0:

1) When I open PSS analysis form, there are always "large" in "fundamental
tones-signal" column, looks like I can't change this property. I thought
there should be "moderate" in IC 4.4.x. so I just wonder if Cadence made
any modification here.

2) the memory was eaten up during PSS analysis (my RAM is 512M). that makes
analysis speed extramely slow. Is it a bug?

Thanks a lot.

 

Hi Andrew:



Thank you very much for your reply.

According to your point, new PSS don't not have choice other than "large".
However, the problem is new PSS always treats my input signal source (could
be set as "small" one in previous version) as "large" signal. It could make
some errors I guess. In this case, how can I make PSS recognize it as a
"small" signal just for PAC? or I add this kind of sources after PSS
analysis?

As of the memory issure, my circuit info is:

Circuit inventory:
nodes 102
equations 138
ahdl simulator 1
bsim3v3 100
capacitor 30
inductor 10
isource 6
opamp 1
quantity 6
resistor 31
vsource 12
Important parameter values:
start = 0 s
outputstart = 0 s
stop = 5.15 us
step = 5.15 ns
maxstep = 103 ns
ic = all
skipdc = no
reltol = 1e-03
abstol(I) = 1 pA
abstol(V) = 1 uV
temp = 27 C
tnom = 27 C
tempeffects = all
errpreset = moderate
method = trapgear2
lteratio = 3.5
relref = sigglobal
cmin = 0 F
gmin = 1 pS

PSS sideband: 0



what do you say? is there any thing I can set to minimize RAM usage?


Thanks a lot





Andrew Beckett wrote:

 

Hi Andrew:



I guess I solved half of the memory problem. By typing in a swap file name
in "swapfile" column, spectre could save tons of memory. however, there are
still lots of memory being occupied by disk cache and finally slowing down
the performance. So, I'm try to figure out how to disable cache function in
Linux now. Do you have comments about this?



Regards

 

Hi Vincent,

Well you didn't mention the frequencies involved (which is probably
useful for me to know, since that will be a major influence on the
time.

PSS treats all large signal, time varying inputs as large (because they are).
If you want to have a small signal for PAC, then use a vsource, isource, or
port, and set the "PAC Magnitude". No need to make that input a sine
source - DC would do.

So in a mixer you could have just the LO as large, with the RF input as
DC with PAC magnitude set, Then the PSS would only have to analyse
a single period of the LO, which is pretty efficient. Now, it depends on
what you're doing - you might need large signal inputs too (if you're doing
1dB compression point, say) - but if there's no need fo a signal to be large,
make it small...

I'd definitely advise attending a training class if you can - it definitely
helps to get the best out of a complex tool like spectreRF.

You mentioned in a subsequent using "swapfile". swapfile doesn't prevent memory
being used, it just uses it more efficiently by mapping that memory into a file
in a way that is under spectre's control, rather than using OS swapping. If you
can avoid using that memory, then that's a better thing.

To be fair 512Mbytes isn't that much memory, though...

Andrew.