The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Behavioral Models >> Extracting Jitter for divider
https://designers-guide.org/forum/YaBB.pl?num=1058434360

Message started by Jason on Jul 17^th, 2003, 2:32am

Title: Extracting Jitter for divider
Post by Jason on Jul 17^th, 2003, 2:32am

I'm trying to extract Jitter for a divider to model PLL phase noise as per Ken Kunderts paper and have some questions--

1. What’s strobed PNOISE? I ran PSS, then PNOISE on a divider and plotted out the output phase noise in V/sqrt(Hz) relative to output signal. Is this ok?
2. To get Jee the psd is to be integrated between 0 and fo/2 as per the paper. If my divider takes an input of 6 GHz and divides it to 1.5 GHz, should I integrate from 1.5GHz to an offset of about 10 MHz when the phase noise is almost nil?

Thanks in anticipation of replies

Title: Re: Extracting Jitter for divider
Post by Ken Kundert on Jul 17^th, 2003, 10:02am

Jason,
Strobed noise is also referred to as time-domain noise. You use it by setting the "Noise Type" to "timedomain" on the PNoise "Choose Analysis" form. Then set the "number of points" to one and click on "Add Specific Points" and type in the time where you want to know the noise. That would be the time where the output signal crosses the threshold in the right direction.

To compute the integrated noise in your example, you would compute and integrate the noise from 0 to 750MHz. Why would the phase noise of your divider drop to nil after 10MHz? Are you driving it with an oscillator? If so, you should eliminate that and drive it with a source. That way you won't confuse the phase noise of the oscillator with the phase noise of the divider.

-Ken

Title: Re: Extracting Jitter for divider
Post by Jason on Jul 22^nd, 2003, 3:15am

Thanks Ken.
I wasn't using specific points for the PNOISE. And I ended up with a SSB phase noise plot relative to fo which falls off rapidly after 10 MHz offset.
I'll another simulation at specific points and proceed.
Thanks

Title: Re: Extracting Jitter for divider
Post by Jason on Sep 5^th, 2003, 3:37am

Hi Ken
I’m back on trying out your paper (I’m doing it in my spare time).
I set following pnoise parameters—
Sweetype—absolute 1 to 750MHz
Max side band 50
Timedomain noise # of points=1
Add specific points 1.333n(it’s a divide by 4 cct, input=3GHz, output=750MHz)

After simulation I setup PSS results as follows—
Output noise(instead of integrated noise)
Sweep—spectrum
And I plotted the magnitude.

The plot doesn’t seem to have the null which your paper shows. We can assume the discrete strobed noise to be a product of a continuos noise(with time) multiplied by a narrow pulse train, which is convolution in frequency domain and hence the noise profile repeats itself. Is that right?
I can’t see the repeating pattern in the noise plot.

Jason

Title: Re: Extracting Jitter for divider
Post by Ken Kundert on Sep 5^th, 2003, 9:31am

The lack of the nulls is not a problem. I expect that the spectrum will appear basically flat.

If you are really using the time-domain or strobed noise, the spectrum will be 1/T-periodic with a slight amount of roll off. The roll off will have a sin(x)/x shape with a null near 1/dt where dt is the timestep of the simulator. If the underlying spectrum is flat, you won't see the periodic nature.

-Ken

Title: Re: Extracting Jitter for divider
Post by Jason on Sep 8^th, 2003, 6:00am

Thanks Ken. The spectrum starts out at a positive value and falls off towards zero asymptotically. I’m not clear what the strobed noise feature computes. Your paper says “the noise is periodic in f with period 1/T because n is a discrete-time sequence with period T”. You also say, in the paper, “noise produced by circuit is periodically sampled to create a discrete time random sequence. SpectreRF then computes the PSD of the sequence”. But the timedomain feature appears to work out noise contributed by various frequencies to one time point. Is that what results from computing PSD of the sequence?

Title: Re: Extracting Jitter for divider
Post by Ken Kundert on Sep 8^th, 2003, 8:46am

The noise goes to zero asymptotically? Are you sure you are using the time-domain noise feature? If so, could you tell me what T is and what are the start and stop frequencies for your time-domain noise analysis.

Concerning the two different interpretations of the time-domain noise feature, the one I give in my paper is more correct. The problem with the "its the noise at a single time point" interpretation is that it consists only of the values at a single point in time. As such, there is no "signal", i. e., a value that varies with time, and so there is no PSD.

-Ken

Title: Re: Extracting Jitter for divider
Post by Jason on Sep 8^th, 2003, 9:05am

Input frequency to the divider is 3GHz. Output is 750MHz. So T is 1.33ns. I chose "output noise" and "spectrum" options in the tdnoise results form.
The rise/fall time for this bipolar divider is 180 ps. I was using a maxsideband of 5, which is too low I think. I'm running a simulation now with maxsideband of 120. The PSS has 20 harmonics and I've put in a maxacfreq of 100GHz.
I'll see if result of this simulation is different.
The "output noise"(as on tdnoise form) doesn't drop to zero with frequency but gets quite close (starts off at about 2 uV/sqrt(Hz) and falls off rapidly to less than 0.1 uV/sqrt(Hz).

Title: Re: Extracting Jitter for divider
Post by Jason on Sep 8^th, 2003, 9:14am

The start and stop frequencies are 1 Hz to 1GHz

Title: Re: Extracting Jitter for divider
Post by Ken Kundert on Sep 8^th, 2003, 5:50pm

I would expect to see a minimum at 750MHz. If the noise is dropping monotonically through 750MHz, then something is wrong.

-Ken