This is an extension of transient analysis rather than a separate analysis mode. When activated, real time noise sources are added to all noisy devices with a magnitude and frequency distribution calculated using the same equations used for AC noise analysis. This allows noise analysis to be performed on sampled data systems and oscillators for which AC noise analysis is not appropriate.
Real time noise is not available with all versions of the product. Contact sales for details.
In this topic:
This specifies the sampling interval of the noise generators. You should set this to a maximum of about 1/3 of the highest noise frequency of interest. Note that the interval also forces a maximum time step so short intervals can result in long simulation times.
The time at which the noise generators are switched on. Defaults to 0.
The time at which the noise generators are switched off. This defaults to the stop time of the transient run.
If you think you may wish to restart the transient run after it has completed and you wish the noise generators to continue to be enabled after the restart then you must specify this time beyond the initial stop time before starting the analysis. You should avoid, however, using inappropriately large values for this stop time as this may noticeably slow the simulation and in extreme cases could cause an out of memory condition.
This affects how the noise sources are handled between noise steps. The choice is between Mode 0 and Mode 1. Mode 0 is nearly always the best mode but this can underestimate the noise is some cases. The difference between these modes is explained as follows:
Real time noise, introduces current sources across all noisy junctions. The magnitude of these sources is determined at each noise step according to the operating point of the device and a randomly generated value whose magnitude is determined from the noise equations.
The RTN mode affects how this current source is set between noise steps. This is not a problem if the operating point of the device is unchanged; the source simply ramps linearly to the next noise step. The problem occurs when the operating point changes, especially if it changes profoundly as would be the case if a transistor switches rapidly from an on-state to an off-state. In this scenario, the magnitude of the noise current would be high in the on-state but fall away to near zero in the off-state. At the same time the switch moves from a strongly conducting state to a non-conducting state.
In Mode 1, the source ramps linearly between noise steps and the operating point of the device is not considered until a new noise step is reached. This method can in some cases grossly over-estimate the noise. In Mode 0, the noise source is recalculated at each time step and adjusted according to the operating point of the device. This method tends to underestimate the noise but not by the same excessive amount that Mode 1 overestimates.
In general, we can't think of a good reason to use Mode 1 except as a confidence check. Both methods should give similar results if the noise step is small enough so a useful check is to run a circuit for a small time using each mode but with a very small noise step. The results for each should be similar.
Real time noise waveforms are created using a pseudo-random number generator which requires to be seeded by a starting value. The seed controls the random number sequence; two runs with the same starting seed will give identical results.
Two option settings provide some control on how the seed is chosen. In addition the MCFIXEDSEED option is designed to be used with real-time noise. These option settings are not available through the GUI but may be set using a .OPTIONS command in the schematic F11 window or directly in the netlist.
| RTNSEED | Fixes the seed used by real-time noise at the start of each step. Note that when this option is set, all runs in a multi-step run will use the same seed |
| RTNRESEED | Boolean: if specified a new random seed is generated for each real-time noise run. This is overridden by the RTNSEED option. If neither RTNSEED or RTNRESEED are specified, the real-time noise seed is fixed at the start of the simulation to a consistent value meaning that successive real-time noise runs will be yield the same noise waveforms. Note that each step in a multi-step run will use a different seed |
| MCFIXEDSEED | This option does not affect the real-time noise seed but is designed to be used with real-time noise. If greater than zero, all Monte Carlo steps in a multi-step Monte Carlo run will use the same seed and thus select the same component values |
Real Time Noise analysis in the Simulator Reference Manual. This includes the results of some comparisons between AC noise and real time noise.
| ◄ Noise Analysis | Transfer Function ▶ |