spectre pnoise / timedomain - unexpected results

[oliver]

Hi all,

I have run a basic pnoise simulation of an ideal sine wave in series
with a 603.6 kohm resistance generating a 100nV/sqrt(Hz) white noise.

When plotting the output noise, the spectral density given by the
pnoise/sources and the pnoise/modulated (USB/LSB) is 100nV/sqrt(Hz)
as expected.

However the spectral density given by the pnoise/timedomain is
458.4 nV/sqrt(Hz) in both time and frequency domains [but the density
is flat vs time and frequency as expected].

I have tried in cadence versions 5.0.33_USR3 and 5.10.41_USR3: they
give the same result.

Does anyone have an idea why the noise given by the PNOISE/TIMEDOMAIN
is higher than expected?

Thanks.
Oliver.

 

[Andrew Beckett]

On 3 Dec 2006 12:50:46 -0800, "oliver" <screwi@free.fr> wrote:

Hi all,

I have run a basic pnoise simulation of an ideal sine wave in series
with a 603.6 kohm resistance generating a 100nV/sqrt(Hz) white noise.

When plotting the output noise, the spectral density given by the
pnoise/sources and the pnoise/modulated (USB/LSB) is 100nV/sqrt(Hz)
as expected.

However the spectral density given by the pnoise/timedomain is
458.4 nV/sqrt(Hz) in both time and frequency domains [but the density
is flat vs time and frequency as expected].

I have tried in cadence versions 5.0.33_USR3 and 5.10.41_USR3: they
give the same result.

Does anyone have an idea why the noise given by the PNOISE/TIMEDOMAIN
is higher than expected?

Thanks.
Oliver.

Are you sweeping the frequency above half the pss fundamental in the timedomain
pnoise simulation? Because timedomain noise is sampling the output (at the same
rate as the pss - it strobes the noise at a timepoint in the period), you cannot
sweep above half the pss fundamential - otherwise you get aliasing from this
sampling process.

Regards,

Andrew.
--
Andrew Beckett
Principal European Technology Leader
Cadence Design Systems, UK.