## Simulating noise in spectre with measured noise parameters

### Simulating noise in spectre with measured noise parameters

Is there an easy way to simulate noise in spectre using measured noise
parameters?

--
S?ren Laursen http://www.tele.auc.dk/~slau/

### Simulating noise in spectre with measured noise parameters

Yes. Various sources in spectre (e.g. vsource, isource, port) accept a
parameter
noisefile which is a file containing spot noise data - see spectre -h
vsource
for more details, or look in openbook).

Of course, you need to be careful doing this. If the circuit you have
measured
noise data for has cyclostationary noise properties, and the circuit
you're analysing
in spectre is also periodic (with a related period to the measured
circuit), you
won't get the right answer, since you'll have lost important frequency
correlation
information when you captured the measured noise. The only real solution
in this case is to simulate the whole circuit.
If however, the measured circuit is not periodic, or the simulated circuit
is
not co-periodic with the measured circuit, then all should be OK. I'm
also primarily talking about the RF analyses here.

Regards,

Andrew.

> Is there an easy way to simulate noise in spectre using measured noise
> parameters?

> --
> S?ren Laursen http://www.tele.auc.dk/~slau/

### Simulating noise in spectre with measured noise parameters

> Yes. Various sources in spectre (e.g. vsource, isource, port) accept
> a parameter noisefile which is a file containing spot noise data -
> see spectre -h vsource for more details, or look in openbook).

> Of course, you need to be careful doing this. If the circuit you
> have measured noise data for has cyclostationary noise properties,
> and the circuit you're analysing in spectre is also periodic (with a
> related period to the measured circuit), you won't get the right
> answer, since you'll have lost important frequency correlation
> information when you captured the measured noise. The only real
> solution in this case is to simulate the whole circuit.  If however,
> the measured circuit is not periodic, or the simulated circuit is
> not co-periodic with the measured circuit, then all should be
> OK. I'm also primarily talking about the RF analyses here.

> Regards,

> Andrew.

> > Is there an easy way to simulate noise in spectre using measured
> >noise parameters?

> > -- S?ren Laursen http://www.tele.auc.dk/~slau/

What I have as input is really measured s-parameters and corresponding
values of Fmin, Gammaopt, and Rn for a transistor at various
frequencies.  If needed, these could be converted into a noise voltage
source, a noise current source, and a third noise current source.  The
first and second noise source are correlated with a complex factor.

These three parameters are needed to fully describe a noisy two-port.
As a understand the documentation, the spot noise data file only
includes one parameter.  Therefore, I can only simulate for a specific

Perhaps it is a limitation of Spectre?  HPADS and other RF simulators
accepts noise parameters directly in a 2-port s-parameter file.  Those
are not time-domain simulators, however.

--
S?ren Laursen http://www.tele.auc.dk/~slau/

### Simulating noise in spectre with measured noise parameters

Hi Soeren,

Currently this is a limitation of spectre's nport component - in current
releases
it is noiseless. In IC446 there will be some improvement (it can generate
thermal
noise itself, I believe), but not quite all the pieces are in place to
parameters from the s-parameter file and use them - this is the subject of

several enhancement requests, and it should be coming in a future release
(I
believe it has been partially implemented, but is not completed yet).

That way it should be possible to describe the correlation between noise
as
well, and therefore not have loss of noise information in periodic
systems.

Regards,

Andrew.

> What I have as input is really measured s-parameters and corresponding
> values of Fmin, Gammaopt, and Rn for a transistor at various
> frequencies.  If needed, these could be converted into a noise voltage
> source, a noise current source, and a third noise current source.  The
> first and second noise source are correlated with a complex factor.

> These three parameters are needed to fully describe a noisy two-port.
> As a understand the documentation, the spot noise data file only
> includes one parameter.  Therefore, I can only simulate for a specific

> Perhaps it is a limitation of Spectre?  HPADS and other RF simulators
> accepts noise parameters directly in a 2-port s-parameter file.  Those
> are not time-domain simulators, however.

> --
> S?ren Laursen http://www.tele.auc.dk/~slau/

Hi,

I am able to see flicker noise upconversion in my phase noise
simulations. However, the noise amplitude starts off from a value way
higher than 0dB (about 60-70 dB). The output is supposed to be in
dBc/Hz. So, should it not be less than 0dB at all times? I have tried
improving the tolerances but it is to no avail. Can someone suggest
something?

thanks
vivek