Noise FMAX setting in transient noise simulation

Dear Itai Finfter,

Itai Finfter said:

"you want your maximum noise frequency to be about 40 times the clocked frequency of the ADC".

What does the 40 times value based on? where does this magic number come from?

I cannot speak definitively for the motivation for the factor of 40 used to set noisefmax in the ADC transient noise simulation as I do not work for Cadence. However, in light of my experiences with transient noise with modulated signals and my reading of Cadence's documentation, please allow me to provide a few thoughts that might be insightful.

1. When simulating transient noise of a circuit that contains some type of modulated output, to accurately capture the resulting output noise, it is necessary to include the effect of the modulation on

Figure 1

Dear Itai Finfter,

Itai Finfter said:

"you want your maximum noise frequency to be about 40 times the clocked frequency of the ADC".

What does the 40 times value based on? where does this magic number come from?

I cannot speak definitively for the motivation for the factor of 40 used to set noisefmax in the ADC transient noise simulation as I do not work for Cadence. However, in light of my experiences with transient noise with modulated signals and my reading of Cadence's documentation, please allow me to provide a few thoughts that might be insightful.

1. When simulating transient noise of a circuit that contains some type of modulated output, to accurately capture the resulting output noise, it is necessary to include the effect of the modulation as it will alias noise sources about the modulation frequency. As a simple example, Figure 1 shows the impact of sampling a 50 MHz tone with thermal noise at 50 MHz. Note that the sampled signal contains noise at low frequencies of not only the thermal noise in the unsampled waveform, but also includes the aliased noise about the harmonics of the 50 MHz modulation signal.

2. The value of noisefmax in a transient noise simulation sets the upper bandwidth of the noise source(s) in the simulation. Hence, if the value of noisefmax is not set to a sufficiently large value, the aliased noise at low frequencies will be underestimated. In this case, "sufficiently large" implies a value that the highest sample clock frequency harmonic will alias to near DC.

3. Now, the assumption of  a noisefmax as a factor of 40 times the sampling clock rate is one that I believe Cadence estimates as the maximum number of harmonics of the 1.2 GHz sample clock signal. You can verify this if you run a conventional transient analysis and example the output spectrum of the 1.2 GHz sample clock to determine how many harmonics are significant. In a Cadence application note describing a transient noise analysis for a sampled noise source which compares the low frequency noise to that of a conventional pnoise analysis, Figure 2 shows that the value of noisefmax must be 5 GHz - or 50 times the clock rate of 100 MHz - to provide good correlation. When the value of noisefmax is set to 500 MHz or 100 MHz, the resulting power spectral density does not correlate well with the expected noise at low frequencies.

II hope this helps explain my thoughts as to why a factor of 40 was selected. I think that in lieu of just using afactor of 40 or 50, the actual clock spectrum might be provided to illustrate the motivation for the choice of noisefmax.

Shawn

Figure 1

Figure 2

Thanks, Shawn, for this clear explanation!

Dear Itai Finger,

I realize it has been some time since you asked your question concerning the motivation for Cadence recommending the noisefmax parameter as 40 times the clock frequency of the ADC. However, the question concerning the appropriate choice for the maximum noise frequency in a Transient Noise simulation is relatively common and is seen often in the Cadence Forums. As a result, I have attempted to study the issue a bit more and have come up with some guidelines for choosing the maximum and minimum noise frequencies based on the relevant signal's properties. I considered a Transient Noise simulation where one if primarily interested in voltage noise and a Transient Noise simulation where one's concern is in phase noise or both voltage noise and phase noise. I've placed the note at URL:

1drv.ms/.../s!AnM-GsAEZPoSry7qIOYgZd02_-C0

and have attached its summary and conclusion section below to save you or anyone with an interest some time if you are still interested in the topic.

Shawn

Section 6 of note at URL:

1drv.ms/.../s!AnM-GsAEZPoSry7qIOYgZd02_-C0

Dear Itai Finger,

I realize it has been some time since you asked your question concerning the motivation for Cadence recommending the noisefmax parameter as 40 times the clock frequency of the ADC. However, the question concerning the appropriate choice for the maximum noise frequency in a Transient Noise simulation is relatively common and is seen often in the Cadence Forums. As a result, I have attempted to study the issue a bit more and have come up with some guidelines for choosing the maximum and minimum noise frequencies based on the relevant signal's properties. I considered a Transient Noise simulation where one if primarily interested in voltage noise and a Transient Noise simulation where one's concern is in phase noise or both voltage noise and phase noise. I've placed the note at URL:

1drv.ms/.../s!AnM-GsAEZPoSry7qIOYgZd02_-C0

and have attached its summary and conclusion section below to save you or anyone with an interest some time if you are still interested in the topic.

Shawn

Section 6 of note at URL:

1drv.ms/.../s!AnM-GsAEZPoSry7qIOYgZd02_-C0